Understanding the

Fundamental Constituents
ofMatter

THE SUBNUCLEAR SERIES

Series Editor:

ANTONINO ZICHICHI
European Physical Society
Geneva, Switzerland

I.

1963

STRONG, ELECTROMAGNETIC, AND WEAK INTERACTIONS

2.

1964

SYMMETRIES IN ELEMENTARY PARTICLE PHYSICS

3.

1965

RECENT DEVELOPMENTS IN PARTICLE SYMMETRIES

4.

1966

STRONG AND WEAK INTERACTIONS

5.

1967

HADRONS AND THEIR INTERACTIONS

6.

1968

THEORY AND PHENOMENOLOGY IN PARTICLE PHYSICS

7.

1969

SUBNUCLEARPHENOMENA

8.

1970

ELEMENTARY PROCESSES AT HIGH ENERGY

9.

1971

PROPERTIES OF THE FUNDAMENTAL INTERACTIONS

10.

1972

HIGHLIGHTS IN PARTICLE PHYSICS

II.

1973

LAWS OF HADRONIC MATTER

12.

1974

LEPTON AND HADRON STRUCTURE

13.

1975

NEWPHENOMENAINSUBNUCLEARPHYS~S

14.

1976

UNDERSTANDING THE, FUNDAMENTAL CONSTITUENTS

OF MATTER

Volume 1 was published by W. A. Benjamin, Inc., New York; 28 and 11-12 by Academic Press,
New York and London; 9-10 by Editrice Compositori, Bologna; 13-14 by Plenum Press, New York
and London.

Understanding the
Fundamental Constituents
ofMotter
Edited by

Antonino Zichichi
European Physical Society
Geneva, Switzerland

PLENUM PRESS NEW YORK AND LONDON

Library of Congress Cataloging in Publication Data

International School of Subnuclear Physics, Erice, Italy, 1976.
Understanding the fundamental constituents of matter.
(The Subnuclear series; 14)
Includes index.
1. Particles (Nuclear physics}-Congresses. I. Zichichi, Antonino. II.
Sicily (Region) III. Rehovot, Israel. Weizmann Institute of Science. IV.
Title. V. Series.
QC793.I5551976
539.7'21
78-2898
ISBN-13: 978-1-4684-0933-8
e-ISBN-13: 978-1-4684-0931-4
DOl: 10.1007/978-1-4684-0931-4

Proceedings of the 1976 International School of Subnuclear Physics

(NATO-MPI-MRST Advanced Study Institute) held in Erice, Trapani,
Sicily, July 23-August 8, 1976 and sponsored by the Sicilian
Regional Government and the Weizmann Institute of Science

1978 Plenum Press, New York

Softcover reprint of the hardcover 1st edition 1978
A Division of Plenum Publishing Corporation
227 West 17th Street, New York, N.Y. 10011
All rights reserved
No part of this book may be reproduced, stored in a retrieval system, or transmitted,
in any form or by any means, electronic, mechanical, photocopying, microfilming,
recording, or otherwise, without written permission from the Publisher

PREFACE

During July and August of 1976 a group of 90 physicists from

56 laboratories in 21 countries met in Erice for the 14th Course
of the International School of Subnuclear Physics. The countries
represented were Argentina, Australia, Austria, Belgium, Denmark,
the Federal Republic of Germany, France, the German Democratic
Republic, Greece, Israel, Italy, Japan, Mexico, Nigeria, Norway,
Sweden, the United Kingdom, the United States of America, Vietnam,
and Yugoslavia. The School was sponsored by the Italian Ministry
of Public Education (MPI), the Italian Ministry of Scientific and
Technological Research (MRST), the North Atlantic Treaty Organization (NATO), the Regional Sicilian Government (ERS), and the
Weizmann Institute of Science.
The program of the School was mainly devoted to the elucidation and discussion of the progress achieved in the theoretical and
experimental understanding of the fundamental constituents of matter.
On the theoretical front we had a series of remarkable lecturers
(C. N. Yang, S. Weinberg, G. C. Wick) attempting a description of
finite size particles. Another group of lecturers covered such
topics as the understanding of the new particles (H. J. Lipkin),
whether or not jets really exist (E. Lillethun), and the unexpected
A-dependence of massive dileptons produced in high-energy proton-nucleus collisions (J. W. Cronin). Two other outstanding questions
were covered by E. Leader and G. Preparata respectively: whether
strong interactions are still within the Regge framework, and if it
is really possible to master strong interactions. A. J. S. Smith
convinced everybody that a large fraction of single inclusive lepton production in hadronic interactions can be accounted for by pair
production. The highlights of the School were the (~-e~O) events
presented by W. F. Fry. The program was completed by an excellent
series of review lectures on the more classical field of Subnuclear
Physics.
I hope the reader will enjoy this book as much as the students
enjoyed attending the lectures and the discussion sessions, which
are the most attractive features of the School. Thanks to the work
of the Scientific Secretaries the discussions have been reproduced
as faithfully as possible. At various stages of my work I have
v

PREFACE

enjoyed the collaboration of many friends whose contributions have

been extremely important for the School and are highly appreciated.
I thank them most warmly. A final acknowledgement to all those who,
in Erice, Bologna and Geneva helped me on so many occasions and to
whom I feel very much indebted.
A. Zichichi
January, 1977
Geneva

CONTENTS

THEORETICAL LECTURES
Critical Phenomena for Field Theorists
S. Weinberg
Monopoles and Fiber Bundles

53

C. N. Yang

Three Lectures on Solitons

G. C. Wick
Can We Ever Understand Hadronic Matter?
A Proposal
G. Preparata
Can Pedestrians Understand the New Particles?
H. J. Lipkin
Are Strong Interactions Still Within the
Regge Framework?
E. Leader

85

115

179

255

Hadronization of Quark Theories

H. Kleinert

289

Phenomenology of Neutral--Current Interactions

J. J. Sakurai

391

REVIEW LECTURES
Weak Currents and New Quarks
M. Gourdin
Review of Massive Dilepton Production in
Proton-Nucleus Collisions
J. W. Cronin
vii

445

485

viii

CONTENTS

Are Jets Really There?

E. Lillethun
Characteristics of ~-e~o Events Produced
by a Neutrino Beam
W. F. Fry

507

537

Hadron Physics at FERMILAB

T. Ferbel

555

A Review of the ISR Results

G. Valenti

611

The Highlights of the Tbilisi Conference

C. W. Fabjan

663

SEMINARS ON SPECIALIZED TOPICS

Hadron Nucleus Collisions in the Collective
Tube Model

G. Berlad

683

Production of Dimuons by Pions and

Protons at FERMILAB
A. J. S. Smith

701

Physics with the Single Arm Spectrometer

at FERMILAB
D. Cutts

741

Azimuthal Correlations in Particle Production

at Low

p~

G. Ranft

777

Monopoles
P. Vinciarelli

799

Quarks, Color and Octonions

F. Buccella

841

Field Theory Approach to the Statistical

Bootstrap

E. Etim

849

CLOSING LECTURE
Fifty Years of Symmetry Operators
E. P. Wigner

879

CONTENTS

ix

Closing Ceremony

893

List of Participants

895

Index

905

CRITICAL PHENOMENA FOR FIELD THEORISTS

Steven Weinberg
Lyman Laboratory of Physics, Harvard University
Cambridge, Massachusetts

02138

1. INTRODUCTION
~~ny of us who are not habitually concerned with problems in
statistical physics have gradually been becoming aware of dramatic
progress in that field. The mystery surrounding the phenomenon of
second-order phase transitions seems to have lifted, and theorists
now seem to be able to explain all sorts of scaling laws associated
with these transitions, and even (more or less) to calculate the
"cri tical exponents" of the scaling laws. 1 Furthermore, the methods
used to solve these problems appear to have a profound connection
wi th the methods of field theory - one overhears talk of "renormalization group equations", "infrared divergences", "ultraviolet cut-offs", and so on. It is natural to conclude that field
theorists have a lot to learn from their statistical brethren.

For this reason, I started a year or so ago to try to learn

the modern theory of critical phenomena. It has not been easy.
On one hand, there are a number of authors who use a language that
is unfamiliar to field theorists, involving concepts (like block
spins, lattice spacings, etc.) that refer specifically to crystal
lattices. It is not so hard to rewrite the formulas in a continuum language, but the physical insight is harder to translate.
On the other hand, there is a school of theorists who follow very
closely the formalism of quantum field theory. This makes the
theory even harder to understand, because many of the formal devices they use (like coupling-constant and field renormalization)
were motivated in field theory by the need to deal with ultraviolet divergences, a problem that has little to do with critical
phenomena. Above all, one wonders how renormalizable field theories,
involving only a finite number of interaction terms, can have any-

S. WEINBERG

thing to do with the effective Hamiltonians of classical statistical mechanics, which must surely involve terms of unlimited
complexity.
These lectures will present what I have been able to glean of
the theory of critical phenomena. After a brief review of the
field-theoretic formalism of statistical mechanics in Section 2,
the qualitative theory will be described in Sections 3 - 7, and
quantitative methods will be introduced in Sections 8 -11. I t hardly needs to be said that almost none of the theory I describe in
these sections is originally due to me. (For detailed references,
consult the reviews listed at the back of these notes. I) The only
material which may possibly have originated with me is the proof of
the invariance of the eigenvalues at a fixed point in Section 7;
the "one-loop equations" presented in Section 8; and the use of
renormalization-group methods to deal with the Bloch-Nordsieck
problem in Section 3. Even here, I would not be surprised to be
informed by a kind reader that some or all of this material already
exists in the published literature.
In the last section I will try to draw some lessons for field
theory from our study of critical phenomena. The formalism used in
studying critical phenomena guarantees that physical quantities are
cut-off independent for all theories, renormalizable or not. What
then determines which of the infinite variety of possible Lagrangians
in field theory is physically acceptable? Is renormalizability
necessary? How does one handle phenomena like gravitation, where
symmetries seem to rule out any renormalizable theory? Some tentative answers are offered, but the questions remain open.
2.

STATISTICAL MECHANICS AND FIELD THEORY

This section will present a very condensed review of the fieldtheoretic formalism of statistical mechanics. I want especially to
explain why it is that the statistical physicists who study critical
phenomena can live in a three-dimensional world, unlike field
theorists, who need to work in four space-time dimensions.
Most of you probably know all about this, but at least this discussion will serve to fix our notation.
The aim of quantum field theory is to calculate S-matrix elements. However, at a finite temperature there is no such thing as
an S-matrix; a finite temperature means that space is filled with
debris like black-body radiation, so any particle that participates
in a collision is scattered again and again before it gets out to
infinity. Instead of S-matrix elements, one wants to calculate a
partition function
Q = Tr exp(-H/8)

(2.1)

CRITICAL PHENOMENA FOR FIELD THEORISTS

(H is the Hamiltonian and 8 is the temperature in energy units, with

Boltzmann's constant equal to unity.) Of course, H may depend on
all sorts of external currents, in which case Q is a complicated
functional of these currents, not a mere number. The "temperature
Green's functions" are the derivatives of R.n Q with respect to
these currents.
To calculate Q, one separates H into a free-particle term Ho
and an interaction term V, and uses the thermodynamic version of
the Dyson formula:
e Ho /8 e-H/8

00

\L--'(_l)n

n=o

n.

f1 /8d

1: 1 d T

{()
()}
n TT VT I .. VTn

(2.2)

where V(T) is the "interaction-representation" operator

(2.3)

and TT is the operator which orders the operators in order of decreasing T-arguments. (To prove this formula, simply differentiate
with respect to 1/8.) It follows immediately that the partition
function is
00

\L -(_l)n
- In=o n.

f1 /8

. ~e -Ho/8 T {V(T 1) V(T ) }]

dT 1 dT Tr
n.
T
n

(2.4)

o
Both Ho and V can be expressed in terms of creation and annihilation operators as usual. (This is done even when there are
conservation laws which prevent actual particle creation or annihilation, because it is by far the most convenient way of incorporating the correct cluster-decomposition properties in the theory.)
Usually these operators appear !n Ho and V in the form of various
fields, linear combinations Ai(X) with simple translation properties.
In the interaction representation, V(T) is then a functional of the
interaction representation "fields"
(2.5)
-+

(The variable x may be continuous, as in field theory, or discrete,

as on a lattice.) We can apply Wick's theorem to Eq. (2.4), and
derive a set of Feynman rules which allow us to calculate Q orderby-order in perturbation theory, just as in quantum field theory.
(Details can be found in the
book of Fetter
and Wa1ecka. 2) The con-+
-+
traction of two fields AI(XIT I ) and A2 (X 2T2 ) yields a factor

~12(~I-~2,TI-T2)

= Tr[e-Ho/8

TT{AI(~ITI)A2(;2T2)}J

(2.6)

S. WEINBERG

with TT now defined with an extra minus sign for T2 > TI when Al and
A2 are fermion field operators.
The only important difference
of quantum field theory is that we
iT over imaginary values from 0 to
from ~ to +00. In consequence, we
Fourier integrals over momenta but

between this formalism and that

are now integrating the "time"
i/S, instead of over real values
can express the propagators as
Fourier sums over energies:

with w restricted to positive or negative integer multiples of TIS.

(On a lattice, the p-integrals are cut off at momenta of the order
of an inverse lattice spacing, while in a field theory they run over
all p.) Furthermore, the Green's functions have a remarkable periodicity property: for T in the range 0 < T < l/S, we have

lI 12 (;:,T) - Tr[e- Ho / S TT{AI(;:,i) A2 (0,i- T)}]

Tr[e-Ho/S eHo/S A I (;:) e-Ho/ S

eHO(~-T)

A2 (0) e-Ho(i- T)]

1 T) JTrLAI(;:) e-HoT A2 (0) e-Ho(e-

Tr[A 2 (0) e- Ho / S e HoT AI (;:) e-HoTJ

Tr[e-Ho / S

A2(0,~)

AI(;:.,T)]

{+ bosons

(2.8)

- fermions

It follows that the sum in Eq. (2.7) runs only over w-values with
w

TIS x even integer

TIS x odd integer

(bosons)
(fermions)

(2.9)

To calculate Q, we add up all diagrams with no external lines. The

momentum-space rules for these diagrams are the same as for the
vacuum amplitude in quantum field theory, except that every internal energy is replaced with a quantity iw satisfying the quantization conditions (2.9), and all energy integrals are replaced with
w sums:

(2.10)

CRITICAL PHENOMENA FOR FIELD THEORISTS

5
(2.11)

(2.12)

For instance, the Green's function for a scalar boson of mass m is

now
-i

-+

(21T) "

6(p,w)

-i
(21T) "

p +w
2

2 +m 2

(2.13)

(I should perhaps add here that Q is given by the sum of all

diagrams, while n Q is given by the sum of all connected diagrams.
The average value of any physical quantity A can be calculated by
introducing a term JA in V and noting that

(A)

TdA e -H/8}
Td e -H/8 }

d
8 dJ n Q

The nth derivative of n Q with respect to external currents is

given by the sum of all connected diagrams with n external lines of
appropriate types. For systems with a large volume Q, the quantity
n Q will contain a factor
and derivatives of (n Q)/Q with respect to various external currents will give the densities of the physical quantities coupled to
these curren ts . )
A second-order phase transition occurs when one of the renorma1ized boson masses of the theory vanishes; the value to which the
temperature must be lowered to make the boson mass vanish is known
as the critical temperature. (The renorma1ized mass, or inverse
correlation length, is defined in terms of the total inverse propagator at p=w=O, including all "radiative" corrections. It is
therefore a function of the temperature.) Our experience in field
theory makes it familiar that a vanishing boson mass signals a smooth
change in the properties of the physical states, including their symmetries. This is to be contrasted with a first-order phase transition, in which the properties of physical systems change abruptly.
For instance, water at sea level pressure undergoes a first-order
phase transition, with finite latent heat, at the boiling point, but
water at the critical pressure and magnets in zero external field
undergo second-order transitions at the critical temperatures.
It is in a second-order phase transition, when one of the boson
masses of the theory vanishes, that the quantization of virtual
energies has its most important effect.

s. WEINBERG

A quantum field theory at zero temperature would at most have

logarithmic infrared divergences, unless the Hamiltonian contained
super-renormalizable ~3 interactions. On the other hand, at finite
temperature we have only a three-dimensional momentum-space volume
element d 3 p instead of a four-dimensional element d 3 p dw available
to cut down the degree of infrared divergences, and the Feynman
integrals will therefore have power instead of logarithmic divergences. (As far as the three-dimensional integrals with w = 0 are
concerned, even ~4 interactions are superrenormalizable.) We therefore expect that at finite temperature there will be dramatic longrange correlations when any of the boson masses approaches zero.
The infrared divergences arise solely from the blowing up of
the w = 0 terms in the sums over boson energies as p-+ 0; in all
other terms the finite value of w acts as an infrared cut-off.
[See Eq. (2.13).] Therefore, in order to study the long-range effects which occur when one of the boson masses approaches zero, it
is very convenient to emphasize only the w = 0 terms, by burying all
other terms in the sums over w in "black boxes". That is, we construct an effective three-dimensional field theory, in which the
propagators consist purely of the w=O terms in (2.7), with all
terms having w'" 0 absorbed into the coupling constants of the effective Hamiltonian. Of course, this effective Hamiltonian will
consist not just of the simple interactions that may have been
originally present in V, but of an infinite number of interactions,
all with temperature-dependent coefficients. (For instance, in a
~4 field theory, the effective Hamiltonian will contain ~2n terms
produced by loops with n corners, in which all the internal boson
lines have w'" 0.) As long as we keep track of all these interactions, the three-dimensional effective field theory is supposed
to be fully equivalent to the original four-dimensional finitetemperature formalism.
In general, we would not expect to know very much about the
temperature-dependence of the parameters in the effective Hamiltonian. The one thing we do know is that these parameters are
given by sums over diagrams with w'" O. Therefore, the effective
Hamiltonian itself does not feel the influence of the infrared
divergences, and the parameters in this Hamiltonian are expected to
be analytic functions of temperature even near the critical temperatures. Fortunately, as we shall see, this is all the information
we need for most purposes.
In renormalizable theories with weak couplings and high temperature, it is possible to say a good deal more about the structure of the effective Hamiltonian. 3 (The material in the balance
of this section is offered to readers who want to see some concrete formulas for coefficients in the effective Hamiltonian. It
is not needed as a basis for the rest of the lectures.) If the
couplings are weak, loop graphs are generally small, and most of

CRITICAL PHENOMENA FOR FIELD THEORISTS

the coefficients in the effective Hamiltonian simply have approximately the value they would have had in the original zero-temperature
theory. However, some of the W-sums in these loop graphs may contribute powers of the temperature to the coefficients in the effective Hamiltonian, and if the temperature is sufficiently high,
these powers of temperature can compensate for powers of the coupling. For temperatures which are much larger than any masses or
momenta, the temperature dependence of a one-loop diagram of dimensionality D will be simply eD; here D is just the degree of divergence of the theory at zero temperature. Each loop also introduces
a coupling-constant factor, say f. Hence the leading contribution
to the effective Hamiltonian when e is large and f is small will
come from diagrams in which each loop is as divergent as possible.
The worst divergences in renormalizable theories are the quadratic
divergences in scalar boson propagators, with D= 2. Furthermore,
the only graphs in which ~ loop is quadratically divergent con~
sist of just a string of one-loop insertions in a scalar boson
propagator. Hence, we can obtain the effective Hamiltonian to
lowest order in f but to all orders in fe 2 by simply calculating
the one-loop corrections to the scalar self-energies.
For instance, consider a theory involving a number of scalar
fields ~" with an interaction
1.

The one-loop correction to the boson mass matrix in the effective

Hamiltonian is
L\M~,(S)
1.J

We are only interested here in the leading terms when e is much

greater than the zero-temperature mass M, so we can rewrite this as
~M~ ,(e)
1.J

-TIe
(2TI)"

It is straightforward to calculate that for large e

2

L\Mij (e)

-+

-1

24 e

f ijkk

There are also terms of first and zeroth order in e which are respectively linearly and quadratically divergent, but we neglect
these because we now only are keeping terms in ~M2 of order fS2
Thus, to all orders in fe 2 but 'lowest order in f, the mass matrix

S. WEINBt:RG

in the effective Hamiltonian takes the form

We expect the actual phase transitions to occur somewhere near the

temperatures where one of the eigenvalues of this matrix vanishes.
As expected, the mass matrix Mtj(8) is perfectly analytic near any
critical temperature.
Similar calculations have been carried out for arbitrary renormalizable gauge theories. 3 The general result is that

M~.(8)
~ M~.
1.J
1.J

- 2l482{f"kk+6(e e) .. + Tr[r.y r. y ]}
1.J

a. a. 1.J

1.

"

"

where eO. are the representations of the gauge generators on the

scalar fields (including gauge coupling constant factors) and the
r i are the matrices in the Yukawa couplinglJjril/JCPi of the scalars
and spinors.
It must be stressed that these formulas only provide an approximation to the "bare" mass matrix appearing in the effective
Hamiltonian. The "renormalized mass" (or inverse correlation
length) involves a sum over higher-order corrections, which become
more and more important near the critical temperature. In fact,
as we shall see, the renormalized mass has a dependence on temperature very different from that found here.
3.

THE FLOATING CUT-OFF

The lesson of the last section is that critical phenomena are

conveniently described in terms of an effective Euclidean threedimensional field theory with t~mperature-dependent interactions
of arbitrary complexity. We can now forget the four-dimensional
theory from which we started, and deal with the three-dimensional
effective theory on its own terms.
We are interested in what happens when one of the boson masses
M in the effective Lagrangian becomes small. Because the theory is
three-dimensional, the infrared divergences become so serious in
this limit that perturbation theory necessarily breaks down, no
matter how weak the coupling we started with may be. For instance,
if the theory contains an interaction ucp", then any graph with L
loops constructed solely from this interaction will contain a factor (8u)L.
[See Eqs. (2.11) and (2.12).] But i t follows then by
simple dimensional reasoning that when all momenta are of the order
of the effective boson mass M, the loop integrals must also contribute a factor MrL. (In fact, this is true for sufficiently
large L even when the external momenta are much larger than M.)

CRITICAL PHENOMENA FOR FIELD THEORISTS

Hence the effective coupling constant contributed by each additional

loop is 8u/M. No matter how small u may be, this becomes large when
M is sufficiently small, and perturbation theory becomes useless.
What can we do about this? The problem arises from very small
virtual momenta of order M, but the effective Hamiltonian we are
working with involves all momenta, or at least all momenta less
than an inverse lattice spacing. Is it possible that we are working with an inappropriate effective Hamiltonian?
Suppose we choose the effective Hamiltonian to emphasize the
degrees of freedom that are really important near a phase transition, by introducing a momentum-space cut-off A, and integrating
only over momenta t with Ipl < A. (Eventually A will be allowed to
go to zero.) We do not want the introduction of the cut-off to
change the physics, so the effective couplings must all be chosen to
depend on the cut-off, in such a way as to leave the partition function A-independent. This condition imposes a set of differential
equations on the couplings in the effective Hamiltonian, known as
the renormalization-group equations.
The idea of a floating cut-off, introduced in such a way as
not to change the physics, is actually very old. For an early example, let's return briefly to four dimensions, and consider the
classic problem of Bloch and Nordsieck,4 the radiative corrections
to scattering of an electron by an external Coulomb potential. In
each order of perturbation theory, we encounter infrared divergences
of the form n nmy, where my is a fictitious photon mass introduced
as an infrared cut-off. We all know that these infrared divergences
are cancelled by real photon emission, but let's forget that for a
moment, and ask what is the behaviour of the purely elastic matrix
element as my -+O? To answer this, we introduce an ultraviolet cutoff A on the photon momenta, and use an effective Hamiltonian, in
which the interactions are A-dependent black boxes which include all
effects of "hard" photons with momenta I q I > A. (To define the
cut-off, we can either simply take Iql as the non-invariant square
q2 + q 0 2 , or use a regulator of mass A.)
I t is assumed that the
photon mass my is much smaller than the smallest energy Echaracteristic of the scattering process. (Here E is me' or the initial
momentum It. I, or the final momentum IPfl, whichever is smaller.)
We take A iff the range
my A E
(3.1)
It is easy to see that the only graphs which can produce leading
singularities in my are those containing a single interaction of the
form
(3.2)

plus any number of emissions and absorptions of soft photons with

Iql < A from the incoming or outgoing electron line. (See Fig. 1.)

S. WEINBERG

10

Furthermore, since A E, the external electron momenta never get

very far from the initial and final mass-shell values Pf and p.
(with P~ = pi = -m~). It follows that the matrix element take~ the
form
m
(3.3)
M(Pf,Pi,me,my) = F(Pf,Pi,me,lf)MA(Pf,Pi,me )
where MA is the mass-shell matrix element uf r A ui of

r A' Note that

Figure 1. A typical graph which can produce leading singularities in

Here wavy lines are soft photons; straight
lines are electrons near their mass shell; and darkened
circles represent sums over graphs involving hard photons
and electrons far from their mass shell.

my.

HA does not depend on my and F depends on lily and A only in the ratio
my/A, because we took A E. For instance, the lowest-order radiative correction function may be calculated directly as
(3.4)
where

(3.5)
(3.6)

No~ it is actually pretty easy to calculate F to all orders in

a, and even to sum the series. Suppose however that we were all
combinatoric cripples, and were unable to carry out the 'calculation
directly. The "renormalization group" would do the job for us. We
know that the matrix element M must be independent of the floating
cut-off, so the derivative of Eq. (3.3) with respect to A gives

o=
or in other words

-my
A2

ClF

ClMA
M + FClA
Cl(my/A) A

Cl,Q,n F
ClMA
A-=
ClA
Cl ,Q,n(my/A) MA

(3.7)

CRITICAL PHENOMENA FOR FIELD THEORISTS

11

But MA does not depend on ~ for my E, so the derivative on the

right-hand side must be independent of my/A.
(3.8)

and therefore F is just a power of

ex:

myl A

(my/A)A

(3.9)

in agreement with the lowest-order result (3.4).

It can be shown in precisely the same way that the infrared
divergences in any QED matrix element always sum up to a power of
my. (However, the results are different if the electron mass is
zero; for instance, the lowest-order correction involves ~n2my instead of ~n my. The reason that our renorma1ization-group argument
breaks down for massless QED is that the matrix element MA in this
case is not independent of my, even though it only involves "hard"
photons with Iql ~ A my; this is because a massless photon can
produce an infrared divergence for any value of Iql, when it is
emitted parallel to the initial or final massless electron. Similar
remarks apply to non-Abelian gauge theories.)
If my \o!ere so small that ~n(E/my) was of order 137, the ong1na1 perturbation theory in powers of a ~n(E/my) would have been useless. However, by introducing a cut-off A with A my but
a ~n A/my 1, we can use perturbation theory to calculate the
function F as accurately as we like. True, we cannot then also
calculate the matrix element MA, because a~nE/A is of order unity,
but all the my dependence is in F, and this we can calculate. The
moral is clear: when it is infrared effects that invalidate perturbation theory, the introduction o.f a floating cut-off may not
restore perturbation theory, but it does allow us to say useful
things about the infrared effects themselves.
4.

RENORMALIZATION-GROUP EQUATIONS

The observations of the last two sections lead us to consider

the theory of one or more boson fields in three dimensions, with a
floating cut-off A, and with cut-off dependent and temperaturedependent interactions of arbitrary complexity, chosen subject to
the condition that the physics be cut-off independent. We will
take a single scalar field (p) for simplicity, but will work in a
Euclidean space with arbitrary dimensionality d, for reasons which
will become clear below. The effective Hamiltonian may be written

s. WEINBERG

12

\" (21T) d
HA/8 = L - - , H

n.

Jdd Pld d P

+
+
+
+
u (p ... p ;A)c/>(p ) ... c/>(p )
n n
1
n
1
n (4.1)

We do not explicitly show the temperature-dependence of the Un;

temperature is now just one of the many parameters on which the
real symmetric coupling functions un may depend.
Each of the

Un contains a momentum-conservation delta function

+
+
d+
+ _ +
+
u (p ... p ;A) = 0 (p + ... +p )u (p ... p ;A)
nl
n
1
nnl
n

(4.2)

The usual clustering properties allow each of the Urt to be expanded

as a power series in the momenta; the coefficients in these power
series are our coupling parameters.
In particular, the function u 2 may be written
(4.3)

We can think of the partition function and Green's functions as

being given by a sum of Feynman diagrams with propagators

1+1

(21T) -d G(p;A)8(A-p)

(4.4)

and wi th vertices - (21T) d u 3 , - (21T) d u,,' . . . .

[According to Eqs.
(2.11) and (2.12), the propagators and vertices in the finitetemperature graphs in d+l-dimensional space-time are associated
with factors
-d-l
(21T)-d
2i1T x -i(21T)
-(21T)

respectively.] The function 8 is taken here as the usual stepfunction, but our discussion could be easily adapted to deal with
a smoother cut-off function.
We are interested in the behaviour of the Green's functions in
the infrared limit, when all the momenta are scaled to zero together.
If the coupling parameters were all dimensionless and A-independent
this would be a trivial problem, because A would be the only dimensional quantity in the theory, and we demand that the Green's functions are A-independent. In this case, as all momenta are scaled
together to zero, the Green's functions would simply scale with
their naive dimensionality. Of course, life is not so simple, but
we try to use dimensional analysis for all it's worth.
To this end, let us define a new dimensionless momentum

t - piA

(4.5)

CRITICAL PHENOMENA FOR FIELD THEORISTS

13

for which the cut-off is

(4.6)

Also, we define a new field

X()

= a(A)(p)

(4.7)

with a(A) a constant to be chosen below. We can write the Hamiltonian in terms of new coupling functions

with
g

-+

(9,

-+

-n nd

-+

-+

9, ;A)
a(A)
A u (PI .. P ;A)
I n n
n

(4.9)

Again, we factor out a delta function

-+

od( + ... + )g ( ... 1 ;A)

-+

g (9, 9, ;A)
n

Inn

and find that

-

-+

-+

g (9, 9, ;A)
n 1
n

a(A)

-n nd-d- -+
-+
A
u (p ... p ;A)
n 1
n

(4.10)

(4.11)

In choosing the field scale factor a(A), we recall that our

aim is to study the infrared behaviour of the theory. For this
purpose, we must consider the low-momentum behaviour of the propagator. Before re-sca1ing, we would in general have
-+
-+
U 2 (p ,-p ;A)

-+

-+2-1

m (A) + p Z

(A) + ...

(4.12)

and the re-sca1ed inverse propagator therefore has the behaviour

(4.13)

It is a very great simplification to choose a(A) so that the coefficient of 2 here is a A-independent constant. (We will see in
the following sections the price that would have to be paid if we
made any other choice.) While we are at it, we may as well pick
a(A) so that this constant is unity, i.e.,
(4.14)
With this definition, g2 is dimensionless, and since HAle is dimensionless, the field X and all the coupling functions gn(!l ... !n;A)
are also dimensionless.
Using (4.14) in (4.11), the coupling functions are now related
by

s. WEINBERG

14

nd
-r-n-d n/2
- -+
-+
A
Z
(A)u (p . p ;A)

gn (t1 ... t n ;A)

(4.15)

If ui is the coefficient of a term of order N in the expansion of

in powers of p, and gi is the corresponding term in the expansion
of ~, then
= Z(A)n/2 And / 2 -d-n+N u.
(4.16)
gi
1

un

In particular, the mass has n

scaled coupling is

2, N

0, so the corresponding re-

8 2 (0)

(4.17)

The gi [including 8 2 (0)] comprise the dimensionless coupling parameters of the theory.
We demand that the Green's functions of the theory [for the
original fields (p)] should not depend on A. This imposes on the
couplings ui a set of differential equations, giving dUi/dA in
terms of u and A. These equations can then be rewritten in terms
of the dimensionless couplings gi' and must on dimensional grounds
take the form
(4.18)
(The calculation of the 8i will be taken up in Section 8.)

Note

:~:~s~ib~~:~~: ~~~e~~e!!;:c~~c~~O~:ed:::~:r~~u~:m;:r:~:;~:re~~~am-

only through the g(A); hence the condition that the Green's functions be A-independent must be expressed in terms of the gi themselves. That is, the temperature and similar parameters enter the
theory only as "initial conditions", determining the values of the
g(A) at some arbitrary point A = Ao . Since the 8 do not depend on
any dimensional parameters except A, and the 8' s are dimensionless,
they also cannot depend on A, except through the dimensionless
couplings gi(A). The 8 i are in general non-zero for all interactions, so even if we started with some simple (e.g. renormalizable)
theory, the introduction of a cut-off would force us to include in
the effective Hamiltonian "all conceivable couplings consistent with
the symmetries of the theory.
In addition, the renormalization constants Z(A) or a(A) satisfy
renormalization group equati-ons, which must be linear and homogeneous. [If Z(A) is a solution, so must ~ Z(A) be, because we could
have started with a field /12 instead of .] On dimensional
grounds, this equation must then take the form
A

d~

Z(A)

Z(A)y[g(A)]

(4.19)

Now let's apply this formalism to the Green's functions of the

CRITICAL PHENOMENA FOR FIELD THEORISTS

15

theory. If C~[PI ... PE;u(A) ,A] is a connected (not amputated or IPI)

Green's funct10n with E external lines carrying momenta Pl .. PE ,
then the corresponding Green's function in the re-scaled theory is

D
-E/2- -+
-+
A Z(A)
CE[p ... PE;u(A),A]
I

where

E + Ed - d
2

(4.20)
(4.21)

As always, a bar over r E or CE indicates that a delta function

Od(tl + ... ) or od(P I + ... ) has been factored out of these Green's
functions; this is the origin of the factor A-d in (4.20).
We do not include an explicit A-dependence in r , because this
Green's function is dimensionl~ss, and there are no ~imensional
parameters other than A on which it could depend. On the other hand,
-+
for any given theory, the Green's function CE must (at fixed p) have an explicit A-dependence which compensates for its dependence on A through
the coupling parameter u(A):

(4.22)

This is what we mean by the physics being A-independent.

The renormalization-group equation for the Green's functions
can now be derived by using (4.20) to express fE in terms of CE ,
then using (4.22) to change the cut-off, and then using (4.20)
again to express CE back in terms of rEo For an arbitrary momentum
scale parameter K, we have

-rE [K 7)(, I

..

Kif
-- Z(,)-E/2
,Dr~[AK!I ... ,u(A),A]
NE .g(')]
'
it
it
i t -E

[Z(KA) /Z(A)]

E/2 -D - -k
-k
K r E [)(,l . X,E; g(KA)]

(4.23)

It is convenient to suppress the A-dependence, writing

(4.24)
We can regard giK as the solution of the equation

s. WEINBERG

16

(4.25)
with initial condition
for K

(4.26)

Equation (4.23) can be rewritten (using 4.19) as

(4.27)
We see in particular that the behaviour of the Green's function for
low momenta is determined by the solution of Eq. (4.25) in the limit
K -+ O.
Each particular physical theory with a particular value of the
temperature is represented by a trajectory in coupling-constant
space, along which (4.25) is satisfied. Different points on a given
trajectory do not represent different theories, but only different
Hamiltonians, corresponding to different cut-offs.
5.

FIXED POINTS AND SCALING LAWS

In general, we would not expect the solutions of the renorma1ization-group equations to have any particularly simple behaviour
for fI.-+O. For instance, if m2 (fI.) does not vanish as fI.-+O, then
(4.17) suggests that g2(0) would blow up like fI._2. In order to keep
the physics fl.-independent, the coupling constants g.(fI.) would then
also have to blow up for fI. -+0. However, it might b that for some
special trajectories, m2 (fI.) vanishes for fI. -+ 0, in which case the
various dimensionless couplings g.(fI.) might all remain well-behaved
for fI.-+O. We are going to see th~t the trajectories for which this
happens are just those corresponding to the critical temperatures
of the theory.
The simplest kind of non-singular behaviour is for the g.(fI.)
to approach fixed values gt for fI. -+ O. According to Eq. (4.18);
this would require that, for all i,
(5.1)
If a trajectory leads to such a fixed point for K -+ 0, then in this
limit Eq. (4.27) gives
fE[Kl".KE;g(fI.)]

ex:

K-Dp;

(5.2)

where
(5.3)
In particular, the two-point function behaves like

CRITICAL PHENOMENA FOR FIELD THEORISTS

0::

17

K-2+y(g*)

(5.4)

This result is conventionally written in the form

(5.5)
and we see that the "critical exponent" n is

n
The renorma1ized mass
defined by

y(g *)

(5.6)

(or inverse-correlation length

~/
-

- lim f2
K+O

2-j

d f2
--2dK

1/~)

is
(5.7)

and (5.4) shows that this vanishes, just as we expect at a phase

transition.
Even if there is a point g* at which 8 vanishes, not all
trajectories hit this point. In order to see what is involved,
consider trajectories that pass close to g*
The renormalizationgroup equation (4.25) can then be linearized
(5.8)

==

M ..

[d8 (g)]
i

dg.

1.J

(5.9)

g=g*

The solution can be written (barring degeneracies) as

giK - gi =

\' c ~

l..

(~) A~
ei
K

(5.10)

where A~ is the ~-th eigenvalue of M, and e(~) is the corresponding

eigenvector (with a fixed but arbitrary normalization)

LMi' e ~~)
j

(5.11)

The eigenvectors are classified as infrared-attractive or infraredrepulsive, according as A~ > 0 or A~ < O. (The case A~ = 0 is a
nuisance, and will not be considered here.) Clearly, the condition
for a trajectory actually to hit the fixed point is that c~ = 0 for
all infrared-repulsive eigenvectors e(~). The number of parameters
which have to be adjusted to achieve this is just the number of
infrared-repulsive eigenvectors. For a phase transition of the

s. WEINBERG

18

usual type, in which there is just one parameter (the temperature)

that must be adjusted to achieve a vanishing renormalized mass,
there must be just one repulsive eigenvector. We will see in
Section 11 whether this is actually likely to be the case.
This approach yields useful information even when we are not
precisely at the critical temperature. In general, we would know
almost nothing about the temperature-dependence of the coupling
parameters. However, a coupling function ~(Pl"'Pn;A) may be
thought of as a black box in which are buried all effects of virtual
particles with momenta greater than A (or energies
0; see Sec. 2)
so it is immume to infrared effects caused by virtual particles
with very low momenta. That is, unlike the Green's functions, the
coupling parameters ought to be smooth functions of the temperature,
even near a critical temperature.

w"

If there is just one repulsive eigenvector e(R) , and if 8 c

is a critical temperature at which the trajectory hits the fixed
point, then the coefficient of this eigenvector in (S.lO) must
vanish at 8 = 8 c :
(S.12)
Therefore, by the above smoothness argument, we expect that for 8
near 8 c '
C
a: (8 - 8 )
(S .13)

As long as the gK are sufficiently close to g*, the couplings are

functions only of
g. _g~a:(8_8)K-1/V
(S .14)
1K

where V is another critical exponent, defined in terms of the single

repulsive eigenvalue AR by
V

= -lIAR> 0

(S .1S)

From Eq. (4.27), we find that the Green's function with E external
lines has infrared behaviour

K-~*

or equivalently

f [(8 - 8 )K-l/V]
E
c

(8 - 8 c )-

v~
-I'.;

FE [K(8 - 8 c )

(S.16)
-V

(S .17)

with D~ given by (S.3) and (S.6) as

D~

(~ +

1 -

) - d

and fE and FE unknown functions of a single variable.

(S .18)

Equation

CRITICAL PHENOMENA FOR FIELD THEORISTS

19

(5.2) can be regarded as a special case of the result (5.16), for

6 = 6c
It must be stressed that whether or not we are able actually
to calculate the critical exponents n and v, Eq. (5.16) or (5.17)
contains a remarkable quantity of information. Who would guess
that the Green's functions would take such a simple form, with scaling parameters n and v that do not depend on the number of external
lines?
One case of special interest is provided by the limit K +0,
in which all momenta scale to zero together. For 6
6 c there are
no infrared divergences, so we expect FE in Eq. (5.17) to be analytic at K = o. We can write the two-point function in this limit
as

with t;, a "correlation length". Comparison with Eq. (5.17) shows

that for 6 + 6 c ' th~ correlation length exhibits the scaling behaviour
t;, ex: (6 - 6 ) -v
(5.19)
c

This is the way V is usually defined. Note that the "renormalized

mass" l/t;, does vanish for 6 + 6 c ' showing again that this really
is a second-order phase transition.
As a special subcase, consider the partition function itself.
As we have already remarked, R-n Q is the connected Green's function
with no external lines. Since we divided by a a-function in defining fE' the function fo is just the free energy density

(5.20)
where Q is the volume of the system. Because there are no external
lines, fo cannot depend on a momentum scale K, so (5.17) gives

~
Q

ex:

(6-6 )-VDO = (6-6 )+Vd

c
c

(5.21)

The total energy density per unit volume is then

u = -

a
[~]
a(1/6)
Q

ex:

(6 - 6 c)

Vd-l

(5.22)

Finally, the specific heat per unit volume is

c =

where

CJ.

au
as
ex:

(6 - 6 c)

-CJ.

(5.23)

is yet another critical exponent

CJ. =

2-vd

(5.24)

20

S. WEINBERG

6.

EXTERNAL-FIELD PROBLEMS

We can also apply the same formalism to study the effect of

external fields. Suppose we add a "magnetic" perturbation to the
Hamil tonian
(6.1)

The effect is to change the free energy density to

00

W(h)

E=l

hE
CE(D,D, ... D) E!

(6.2)

But for zero external momentum, Eq. (5.17) gives

CE(D,D, ... D) ~ fE(D,D, .. D) ~ (8-8 c )

-\!~

\!d

(8-8)
c

(8-8)

-\!

d-H - .!l) E
2

2
(6.3)

Hence the free energy density takes the form

(6.4)

W(h)
with S some unknown function of a single variable.
tion" (<P) :: M is defined by the condition that

M=

d
- 8 dh W(h) ~ (8-8 c )

\!(~-1+!J.)
2

s'

f,Lh (8-8

c)

The "magnetiza-

-\!(~+l-.!l)J
2

(6.5)
In particular, it may be possible to have a spontaneous magnetization: MID for h = D. In this case, Eq. (6.5) gives
M ~ (8-8 )8

(6.6)

where 8 is another critical exponent

(6.7)

More generally, the value of the external field h required to produce any given magnetization is given by Eq. (6.5) as
h

= (8-8)
c

\! (~+ l-.!l)

function of M(8-8 )-8

c

(6.8)

or equivalently
h

MO x function of M(8-8 )-8

c

(6.9)

21

CRITICAL PHENOMENA FOR FIELD THEORISTS

where 8 is one more (the last!) critical index

8

v(d/2

1 - n/2)/S

(d/2 + 1 - n/2)
(d/2 _ 1 + n/2)

(6.10)

In field theory, it is very convenient to work with a potential,

defined like the Gibbs free energy as a Legendre transform
V(M)

W(h) - hM

(6.11)

From the above results, it is easy to see that

V(M)
7.

cr

M8+l

function of M(8-8

)-s

(6.12)

INVARIANCE OF THE EIGENVALUES: CUT-OFFS VS. RENORMALIZATION

The discussion in Sections 5 and 6 has shown that critical

phenomena are to a great extent governed by the number and the
values of the repulsive eigenvalues of the matrix (5.9). This
naturally raises the question whether the eigenvalues are invariant
to possible re-definitions of the coupling parameters or the cut-off
procedure. In particular, would we obtain the same eigenvalues if
we assumed the theory was renormalizable, and used a floating renormalization point (as in the original work of Gell-Mann and Low 5 )
instead of a floating cut-off?
Suppose we introduce a new set of dimensionless coupling para~
eters ga(A). They are defined in terms of a dimensional quantity A,
which may be a new kind of cut-off (perhaps smoother than a 8-function) or a floating reQormalization point. If the theory is nonrenormalizable, or if A is some sort of cut-off~ there are as many
ga as gi. If the theory is renormalizable and A is a floating renormalization point there are only a finite number of ga parameters,
and we are restricted to a finite-dimensional surface of trajectories.
The old dimensionless coupling parameters gi can only be functions of the new parameters ga and of the dimensionless ratio A/A:
g. (A)
1

g. [g(A) ,A/ AJ

(7.1)

Further, the value of the old cQupling parameters cannot depend on

how we choose the new quantity A, so
-

o = A -=
where

- is
S

dA

g(A)

the new beta function

'\ ago L ~ S

ga

+ _A

ago
1

A a(fi.; A)

(7.2)

S. WEINBERG

22

(3 a C)
g

(7.3)

The old beta function is then

or using (7.2)
ag. _
(3. (g) = I a- 1 (3 (iD
1
a ga a

(7.4)

That is, (3 transforms like a covariant vector in the space of coordinate parameters._ One immediate consequence is the invariance
of fixed points: if (3 vanishes at ga = g:, then (3 vanishes at the
corresponding point gi = g~. Now, how does the matrix (5.9) transform? From (7.4), we have immediately

a(3. ag

a 2 g.

j
Ij ~
--= I
- :: S
a ag}gb a
ag ag
b

- ~
+ ~ _a

ag. a(3

aga agb

(7 5)
.

This is moderately complicated (derivatives of vectors are not

generally tensors) but it simplifies at a fixed point, where the
first term on the right vanishes. At such a point, (7.5) reads

I
j

M.. Sjb
1J

S.1a Ma b

(7.6)

(7.7)

where
M.. - [a(3i (g) la g j
1J

*.

g=g

Mab - [aSa(g)/a gb]_ -*

g=g

(7.8)

s.1a

(7.9)

= [agi/agaJ
-g=g-*
.

It follows then that if e is an eigenvector of M with eigenvalue A,

I Ma b

(7.10)

23

CRITICAL PHENOMENA FOR FIELD THEORISTS

then there is a corresponding eigenvector of M

Is.1a ea

e.

with the same eigenvalue

I.
J

(7.11)

M. e.
1J

Ae

(7.12)

The eigenvalues are therefore independent of the cut-off procedure.

Also, if_we constrained the theory to be renorma1izab1e, and calculated S using a floating renorma1ization point instead of a floating cut-off, the eigenvalues we would obtain would be some subset
of the eigenvalues obtained with a cut-off. The use of renorma1izable field theories in statistical mechanics does not give all the
eigenvalues of M, but those it gives, it gives correctly. As it
happens, one of the eigenvalues of Min a renorma1izab1e q," theory
is the repulsive eigenvalue AR' so renorma1izab1e theories can be
used to calculate the critical exponent v, as done by Brezin et al. 1
8.

THE ONE-LOOP EQUATIONS

It is remarkable that the functions Si(g) can be calculated

exactly in terms of one-100o diagrams. The derivation is lengthy,
and will be outlined in an Appendix. Here, we just give the results.
The one-loop equations ~re w.ritten in terms of a modified rescaled coupling function ~ (Q, l' .. Q,n;A) . [These are not precisely
the same as the coupling functions discussed up to now - the relation is explained in the Appendix. This modification will not be
important in using th~ one-loop equations to study critical phenomena.] Let -(2TI)d Ln (Q,l ... ln;]J,A) be the sum of all one-loop, oneparticle irreducible gr~phs, with n external lines carrying outgoing rescaled momenta Q,l' .. In. These graphs are constructed with
vertices
dd-+
-t---t--t(8.1)
-(2TI) 0 (Q,1+"'+J(, )g (x, x, ;A)
n n 1
n
and propagators with an infrared cut-off
(8.2)

where
(8.3)

Propagators are ngt inc1~ded on the external lines of Ln, and a

delta-function O(Q,l+' "+~n) is factored out, as indicated by the
bar over L. The one-loop equations take the form

s. WEINBERG

24
d -

-+

-+

-+

7-

AdA gn (J/, 1 ... J/,n- 1,-J/,1-"'-)(,n- l;A)

For instance, let us suppose that our theory is invariant

under a transformation ~ -+ -~, so that all odd couplings are eliminated. For n = 2 and n = 4, the one-loop integrals are
_(2n)d
and
-(2n)d

L2 (Q: l ' -i].1

1"

A)::

L.. (i 1.. .1 .. ;].1,A)

-~JddJ/,8(lil-].1)L'1(i'A)-g
(i ' -i1
-1 'A)
,
..
, l'
l'

:: -

~ J dd J/, 8( lil-].1)\(t;A)g6 (i,-t,i1'"

+~ J dd J/, dd J/,'
x

(8.5)

.1.. ;A)

8( Itl-].1)8( Itl-].1')\(t;A)\(t, ;A)

[~(t-t'.i1 ,i2;A)g..(~,-tt3,i.. ;A) 8 (t1+t2+t-t' )

+ 2 permutations]

(8.6)

where 1 .. :: -t1-t2-t3'
(See Figure 2.) In consequence, the oneloop equations for g2 :: L'1- 1 and g .. take the form

A~ L'1- 1 (1 . A)
dA
l'

Y- 2+t
-

and

l'

atJ L'1- 1 (i 1 ;A)

~ (2n) -d fdd J/, 8 ( Iii-I) \(t ;A)g.. (1,-1, t1 ,-t 1 ;A)

d - 4 + 2Y +

I
~(21T)-d IiJ/,
~(21T) -d

r=l

(8.7)

J/,r'

! ~lg.. (t1, .. J . ;A)

aJ/,t

ddJ/,8 ( Iil-l)\(i ;A):g6 (l,-t

l ' t

.. ;A)

ddJ/,' {8 (Iii-I) 8 (1~I-l)+8 (11'1-1) 8 (I ii-I)}

x L'1(i ;A)\(1i;A)[g ..
+g.. (t,-~,i1

(t-~ .t 1.t2;A)g.. (~;-t~f%> 8d(1-1i+~+~)

J 3;A)g.. (~,-t.t2 .1.. )8 d (t-tl+t 1+i 3)

+g.. (t,-t'.t1.1 .. ;A)g (t,,-t,i 2 .i 3 )8 d (t-tl+t 1+t")-11

(8.8)

25

CRITICAL PHENOMENA FOR FIELD THEORISTS

Since the fields are normalized so that the coefficient of 2 in the

1

L2 =

L4 =

3
3

Figure 2. Graphs for the one-loop functions L2 and L 4

Here internal lines represent the cut-off propagators (8.2),
while the small circles represent the interactions gn.

power series expansion of /'>,-1(1;11.) is unity, Eq. (8.7) yields a

formula for y:

Jdd Q, 6( \\

-1)1'1(;II.)g4 ('-'1 '-1 ;11.)

(8.9)
Note that the integrals in all these equations are taken over a
closed d-l-dimensional surface, and are therefore automatically
finite.
These equations are all exact. The "mistake" we 'make in dropping graphs with more than one loop is cancelled by the "mistake"
we make in differentiating only the 8's, not the g's. However, the
fact that we have exact equations in closed form does not mean that
we can derive an exact solution. The equation for g2 involves g4;
the equation for 84 involves g6; and so on. Only by using some
sort of perturbation theory can we get useful results.

26

S. WEINBERG

9.

THE GAUSSIAN FIXED POINT

A fixed point is characterized by a set of coupling functions

~~ gt, g~, ... for which the right-hand side of Eq. (8.4) vanishes.
There is one obvious such fixed point, with

* -+

-+

g n (1"'n )

(for n > 2)

(9.1)

[Equation (8.9) gives y = 0.] This is known as the Gaussian fixed

point, because it corresponds to a free massless field theory, for
which field distributions are Gaussians. The Gaussian fixed point
turns out to be relatively unimportant for statistical mechanics,
but the reasons why this is so are enlightening. Also, the
Gaussian fixed point will play an importan t part for us in our discussion of renormalizability in Section 12.
To find the eigenvalues of the M-matrix at the Gaussian fixed
point, we note that there is a simple basis in which the matrix is
triangular. (That is, the matrix has only zero elements above the
main diagonal.) Observe that there is no term in Adgn/dA linear
in gm for m < n, while the term linear in gn is simply

(9.2)

Hence H is triangular in a basis in which the vectors correspond to

interactions i with definite numbers n i of fields and Ni of momentumfactors. The eigenvalues of a triangular matrix are just its diagonal elements, so (9.2) shows that the eigenvalues at the Gaussian
fixed pOint are the numbers

D.

n.d
1

(9.3)

l.Je recognize that Di is just the "dimensionality" of renormalization theory: an interaction is just super-renormalizable if D < 0,
just renormalizable if D=O, and not renormalizable if D>O.
For d > 2, the value of Di increases as either ni or Ni is increased. Here are the eigenvalues for the interactions with the
lowest values of ni and Ni:

27

CRITICAL PHENOMENA FOR FIELD THEORISTS

Interaction

</>2

-2

</>02 </>

0" </>, ( [J 2 </ 2

d-4

d-2

2d-6

We see that for d = 3, there are two repulsive eigenvectors with

D < 0, corresponding to the super-t=en"ormalizable interactions </>2 and
cp", and one marginal eigenvector with D = 0, corresponding to the
renormalizable interaction </>6. (We do not include </>02 </> as an
"interaction", because its coefficient is fixed by our field renormalization convention.) Even for 3<d<4, there are two repulsive eigenvectors of M. Hence in order for the Hamiltonian to approach the Gaussian fixed point as A -+ 0 it would be necessary to
adjust at least two free parameters in the Hamiltonian to eliminate
the components along both repulsive directions, not just the temperature. This is what is called a tricritical fixed point.
Tricritical fixed points do occur in nature, as for instance
in antiferromagnets and He 3 -He" mixtures. However, the second-order
phase transitions considered here (as in ferromagnets at zero field,
or water at the critical pressure) occur at a critical value of just
one parameter, the temperature, so they have nothing to do with a
tricritical fixed point, like the Gaussian fixed point. We must
look for some other kind of fixed point.
10.

THE 'HLSON-FISHER FIXED POINT

We have already noted that the renormalization-group equations

cannot generally be solved without some sort of perturbative expansion. However, perturbation theory would not generally be expected to work at all near any fixed point except the Gaussian
fixed point. The S-function for an interaction i involving ni

S. WEINBERG

28

factors of <p and Ni factors of momentum is given by Eq. (8.4) as

f\

D. g. + loop terms

D.

2-

where

n.d

d-n.+N.
1

(10.1)
(10.2)

and the "loop terms" arise from the y and L terms in (8.4). At a
fixed point g*, all 8i must vanish, so the first-order terms Di.g~
must cancel the higher-order loop terms. In general, this would not
be possible for small non-zero values of the g~.
However, suppose that one of the gi' say gI' has a very small
dimensionali ty DI , while DI1>DIII'... are all of order uni ty or
greater. If we tentatively suppose that gI is small, then we can
find functions g~I(gI)' g~II(gI)' ... as power series solutions of
the equations

These power series can then be inserted in the "loop terms" for 81 ,
generating a series in powers of gI:
(10.4)
For small DI , the equation 81 = 0 has a perturbative solution
DI

bD~

- -a + -a 3

(10.5)

Note incidentally that this solution for g* may be useful even if DT

is not so small, provided that the coefficient a is sufficiently
large; this is what actually seems to make the method work in
practice.
Of course, for any integral number d of spatial dimensions,
each dimensionality (10.2) is an integer or half-integer. Therefore,
in order to use the above expansions, we have work in a mythical
world, in which d is nearly but not quite equal to a whole number.
To see how this works in practice, we return again to our
standard example, of a single scalar field <p, with a symmetry under
the transformation <p + -<p to eliminate the interactions odd in <p.
Inspection of the table in Section 9 shows that when d is near 4,
the dimensionality of the <p4 interaction is small, while all other
interactions have dimensionalities of order unity or greater. (We
exclude <p0 2 <p from the list of interactions, because its coupling is
fixed by our field-renormalization convention to have the value
unity.) Since we are really interested here in d= 3, we will there-

CRITICAL PHENOMENA FOR FIELD THEORISTS

fore set

d=4-,

29

0<1

(10.6)

and hope that the expansion in powers of gives good results even
at = 1.
As we have seen, the first step in finding the fixed point
would be to express all other couplings in terms of the coupling gr
of the ~q interaction. rt is easy to see that, apart from the ~q
(and ~02~) interaction, the coupling parameter gt(gr) of any interaction with ni factors of the field ~ will have a power series expansion in gr which begins with a term of order
(10.7)

To check this, we suppose that it is true, and see if it gives

consistent results at a fixed point. Equation (10.7) shows that
g (p,p') is of order gr' except for the zeroth order term p2 0 (p+p')
wfiich is fixed by our field-normalization conventions to have coefficient unity. Hence the propagator takes the form

~(t) ~

;2 + O(gr)

(10.8)

R.

Also~ Eq. (10.7) shows that gq(tlt213tq) is of order g~, except for

the ~-independent term, which is of course equal to gr by definition.

A one-loop graph in Si that is built entirely out ?72~qcouplings
will have ni/2 vertices, and hence be of order grn1 ,in agreement
with the order of the term Digi in Si as given by Eq. (10.7). These
terms therefore can cancel at a fixed point satisfying (10.7).
(About the y term, see below.) On the other hand, if we construct
the loop from r ~q vertices and s other vertices with m1 ,m 2 , .. ,ms
external lines apiece, then we must have

+ ms - 2s + 2r
so the number of powers of gr in the graph will be
~(m

+ m2 + ... + mS ) + r =

n. + s
1

which is greater than~n . This incidentally shows that in lowest

order, the one-loop grapfis for Si can be calculated using ~q couplings only, a fact that will prove useful later.
Finally, we must give special attention to the ~02~ "interaction". This has D=O (for all d) and accordingly Eq. (8.9) gives
the fixed-point condition for this coupling as

S. WEINBERG

30

(The loop terms here are of order gi rather than gr' because the
one-loop graph constructed from a ~4 interaction is just a constant;
it therefore contributes to 8 for the ~2 interaction, but not to
the ~02~ or higher-derivative interactions.) The fact that y* is
of order gr2justifies the neglect of y in the estimates made above.
Note that if we had not renormalized our fields, we would not have
the term y available to cancel the one-loop terms in the 8-function
for the ~lJ2~ interaction, and in consequence there could not be a
fixed point of the type considered here. This is the whole motivation for field-renormalization in statistical mechanics - to make a
fixed point possible.
Now we are in a position to do some actual calculations. From
Eq. (8.8), we see that the 8-function for the ~4 interaction is exactly given by

-~(27f) -dJdd JI, 0 ( Iii-I) L'l(i ;1\)g6 (t,-t 0,0,0,0; 1\)

+-f(27f) -d dd

JI,

0 ( Ill-1M 2 (i ;1\)g~ (i ,-1 ,0,0; 1\)

(10.9)

[The two terms in the curly brackets in (8.8) add up to just o( 111-1).
The factor 3 arises because there are three equal terms in the
last integral in (8.8).] We have seen that when the fixed-point
equations are used to express all the other couplings in terms of
gr' it turns out that
y

O(g~)

g6

O(g~)

g4

gr + O(g~)

1:1- 1 = 12

+ O(gr)

Therefore in lowest order, we can simply drop y and g6' set 1:1 =
and replace g4 with the constant gr' so that (10.9) gives
8 r (gr)

-e:g r + z(27f)

-d

Sd gr + O(gr)

1/12 ,

(10.10

where Sd is the area of a unit sphere in d dimensions. Solving the

equation 8 r = 0 then gives the Wilson-Fisher fixed point

CRITICAL PHENOMENA FOR FIELD THEORISTS

31

(10.11)

CALCULATION OF THE EIGENVALUES

11.

We have found a fixed point in 4- dimensions with couplings g*

of order or less. Now we shall use perturbation theory to calculate the eigenvalues of the matrix a Si/Cl gj at this fixed point.
From Eq. (10.1), we see that

(a s/a gj]g=g*

Hij -

(11.1)

with ~Mij of order or less. In zeroth order, the jth component of

the ith eigenvector of M is just 0ij' so first-order perturbation
theory gives the eigenvectors of M as the "expectation value"
A. '" D.
1

+ ~M ..

11

asgi'J *
[~
g=g

(11. 2)

to order .
It is convenient to consider three special cases:
(a) Any interaction with Di - < 0 corresponds to a negative
eigenvalue of M, and hence to an infrared-repulsive eigenvector.
In our standard example of a single real scalar field with a
+ - symmetry, there is just one of these eigenvectors, corresponding to the 2 interaction, with D = -2.
(b) Any interaction with Di>O corresponds to a positive eigenvalue of M, and hence to an infrared-attractive eigenvector. There
are an infinite number of these.
(c) Any interaction with IDilof order corresponds to a borderline
eigenvalue of M, which might be positive or negative, depending on
the value of the interaction term ~M. In our standard example,
there is just one of these, the interaction 4, with DI = -.
(Recall that 02 is not counted as an interaction.) This is repulsive at the Gaussian fixed point, but we cannot tell whether it
is repulsive or attractive at the Hilson-Fisher fixed point without
taking interactions into account. From Eq. (10.10), we have

A =
I

[~]
ag
I

*=

or,using (10.11),
(11. 3)

s. WEINBERG

32

We see that this eigenvector is infrared-attractive, not repulsive,

so there is just one repulsive eigenvector, as required for an ordinary second-order phase transition. Now let us calculate the
critical index V to first order in . As shown in Section 5,
(11.4)
where All is taken as the single repulsive eigenvalue of M. This
eigenvalue is just the one mentioned in case (a) above, corresponding to the interaction ~2. The B-function for this interaction is
given exactly by the one-loop equation (8.7):

Bn

(-2+Y)gn -

~(21T)-dJddQ,O(ltl-l)Mt;A)g.. (t,-t,0,0;A)

(11.5)

The coupling fII is defined as the constant term in the inverse

propagator ~- , so that

So far, this is exact.

Now let us use the -expansion. At the fixed point,

is of
order E 2 ; -g .. equals gI* plus terms of order E 2 ; 0A equals 1 / 2 plus
terms of order ; so that to order E,

[ aBn]
agn

_ 2

~(21T)

-d S d gI

g=g*

or, using (10.11)

(11.6)

The critical index (11.4) is then

V

=2 +

12 + 0( )

(11.7)

Note the factor 1/12; this makes the correction to the lowest-order
value of V rather small even for = 1. If we neglect higher-order
terms, then (10.7) gives V = 0.5875 for e: = 1; the experimental value
is in the neighborhood of 0.6 to 0.7.
12.

RENORMALIZABILITY AND ALTERNATIVES

As I indicated at the beginning, the purpose of these lectures

is to present the theory of critical phenomena to field theorists
who want to apply this machinery in quantum field theory. We now
turn to one of these applications. From now on, our efforts are

CRITICAL PHENOMENA FOR FIELD THEORISTS

33

directed at phenomena in d = 4 dimensions, though we will again have

to consider formal variations in dimensionality. However, we will
continue to work in a Euclidean rather than a Minkowskian space,
leaving the analytic continuation to the physical region to be dealt
with separately.
A good deal of modern elementary particle theory is based on
the assumption that nature is described by a renormalizable quantum
field theory. However, the "floating cut-off" formalism described
in these lectures raises serious questions about the physical significance of the renormalizability requirement. In the formalism
described here, the effective Hamiltonian depends on the cut-off
in such a way that the physics (i.e., the set of Euclidean Green's
functions) is cut-off-independent for all theories, not only renormalizable theories. In this formalism, a renormalizable theory
merely corresponds to a subset of trajectories (characterized by a
few renormalized coupling constants) for which all but a few of the
couplings vanish (at least in perturbation theory) as A + 00
Why
should the real world correspond to such trajectories?
It might be argued that renormalizable field theories are distinguished because they have only a finite number of free parameters. This seems like a rather unphysical requirement. We can
eliminate all free parameters by demanding that at A = 1 MeV all
dimensionless coupling constants gi have the value 37. What is
needed is not uniqueness itself, but a rationale for uniqueness.
Nor is experiment much help here. Non-renormalizable quantum
field theories always inrolve a mass scale, such as Fn ~ 190 MeV
for chiral dynamics; GF-~ ~ 300 GeV for the Fermi theory of weak
interaction; and GNgY.JTON ~ 2 1019 GeV for general relativity. At
energies which are much smaller than the characteristic mass scale,
a non-renormalizable theory will look as if it were renormalizable.
(If symmetries do not allow any renormalizable interactions, it
will look like a free field theory.) The experimental success of
quantum electrodynamics only shows that any non-renormalizable interactions have a characteristic scale larger than a few GeV. The
success (so far) of renormalizable gauge theories of weak and electromagnetic interactions only indicates that any non-renormalizable
interactions have a characteristic scale greater than 300 GeV. We
need theoretical guidance to tell us whether physics will continue
to look renormalizable at really high energies, like 10~ GeV. And
if not, then what does pick out the trajectory corresponding to
the real world from the infinite number of possible theories?
I know of only one promising approach to this problem. Randomly chosen quantum field theories tend to develop unphysical singularities if extended to sufficiently high energies. (In the Euclidean region, any singularity is unphysical.) For instance, suppose
some coupling constant obeys the very simple renormalization-group

S. WEINBERG

34

equation
II. dg(lI.)
dll.

(a> 0)

The solution is, for arbitrary K,

g(KII.)

g(lI.)
l-ag(II.)!/,nK

For g(lI.) >0, this develops a singularity at!/,nK=I/ag(II.), and Eq.

(4.27) then suggests that some singularity occurs in Green's functions at finite Euclidean momenta. Nature must pick out trajectories
which avoid singularities of this type.
One way to satisfy this requirement is to demand that the
Hamiltonian lie on a trajectory which hits a fixed point for II. +00.
Equation (4.27) shows that in this case, the Green's functions
simply behave like powers of K when the momentum scale K goes to
infinity, and do not develop unphysical singularities.
[In particular, this is the case for our example above if we start with a
negative coupling, g(lI.) < 0.] Theories with this property, that the
trajectory hit a fixed point for II. + 00, will be called asymptotically
safe.
Of course, in order to be asymptotically safe, the trajectory
near the fixed point must have no components along eigenvectors of
the M-matrix,that are ultraviolet-repulsive. (Since II. now goes to
infinity instead of zero, ultraviolet-repulsive eigenvectors are
those with positive eigenvalues.) But in all cases that have been
studied, it turns out that there are only a finite number of ultraviolet-attractive eigenvectors (i.e., negative eigenvalues), and
all the rest are ultraviolet-repulsive. Therefore, the demand that
a theory be asymptotically safe imposes an infinite number of constraints on the coupling parameters, leaving only a finite number of
free parameters, i.e., the components of the trajectory at the fixed
point along the finite number of ultraviolet-attractive eigenvectors.
Thus, asymptotic safety can provide a rationale for picking physically
acceptable quantum field theories, which may either explain renormalizability, or else replace it.
To see how this works in practice, let us consider the theory
of a single real scalar field in five dimensions. Let us ask whether
or not it is possible for such a theory to achieve asymptotic safety
by hitting the Gaussian fixed point. As we saw in Section 9, the
eigenvalues at the Gaussian fixed point are simply equal to the
dimensionalities of the various interactions: for each interaction
with n ~-factors and N momentum factors, there is an eigenvalue
given by (9.3), which for d = S reads
D=

-S+tn+N

CRITICAL PHENOMENA FOR FIELD THEORISTS

35

If we do not impose the symmetry under </> + -</>, then there are 2 ultraviolet-attractive eigenvectors, corresponding to the super-renormalizable interactions </>2, with D = -2, and </>3, with D = -~. (As usual,
we do not include </>02</> as an interaction, and we do not include a </>
interaction because such an interaction can always be eliminated by
shifting </> by a constant.) Thus there is a two-parameter set of
trajectories which hit the Gaussian fixed point, and are therefore
asymptotically safe.
It is easy to see that this two-dimensional surface of trajectories simply corresponds to the super-renormalizable theories of a
scalar field in five dimensions. (Working in these super-renormalizable theories, we can calculate the matrix of derivatives of the
Gell-Mann-Lowbetafunction 5 at the Gaussian fixed point, and check
that the eigenvalues of this matrix are -2 and -~.) The two parameters needed to describe the trajectories which hit the Gaussian
fixed point are just the renormalized mass and </>3 coupling. Thus
in this case, asymptotic safety is achieved by requiring renormalizability. More generally, the effective Hamiltonian will approach
the Gaussian fixed point for A +00 if it corresponds to a superrenormalizable theory or an asymptotically free renormalizable theory.
However, it is not clear that nature really does choose trajectories which hit the Gaussian fixed point for A +00. The notorious
problem is gravitation: no one has been able to think of a satisfactory theory of gravitation which is renormalizable. Is it possible that nature achieves asymptotic safety by aiming the trajectories of the effective Hamiltonian at some fixed point other than
the Gaussian fixed point?
As an example of what is possible, let's return to our example
of a scalar field in five dimensions, but let's now impose the symmetry under the transformation </> + -</>. This plays a role here similar
to that of general covariance in the theory of gravitation - it
eliminates the only interaction </>3 that is renormalizable. There
still is one ultraviolet-attractive eigenvector, corresponding to
the "interaction" </>2, but a trajectory that reaches the Gaussian
fixed point along this direction simply corresponds to a free field
theory. An interacting asymptotically safe theory with this symmetry must hit some other fixed point.
What about the Wilson-Fisher fixed point? We are really interested (in this example) in five-dimensional space, but in order
to do calculations, let's work in 4+ dimensions. The eigenvalues
can then be calculated from the results of Sec. 11, by simply
changing E to -E. We see that there are two ultraviolet-attractive
eigenvectors at the Wilson-Fisher point, with eigenvalues

s. WEINBERG

36

\
A

II

2 - ~

+ 0(1':2)

and an infinite number of ultraviolet-repulsive eigenvectors, with

positive eigenvalues. If we assume that the eigenvalues do not
change sign for 0<1':~1, we can conclude that there are just two
ultraviolet-attractive eigenvectors in five dimensions. With this
assumption, there is a two-parameter set of asymptotically safe
field theories in five dimensions associated with the Wilson-Fisher
fixed point. These theories are not renormalizable in the usual
sense - the symmetry under ~ + -~ rules out the possibility of any
renormalizable theory in five dimensions. However, they are interacting theories with no unphysical singularities at high energy.
(In fact, this theory does have unphysical features, but of a
different kind. Changing I': to -I': in Eq. (10.11) shows that the ~~
coupling constant at the Wilson-Fisher fixed point is negative in
4+E dimensions. This means that the potential (6.11) goes to ~
for I <~)I + 00, so the energy is not bounded below. However, this
is an accident of this particular model. Any theory will have nontrivial ultraviolet-attractive fixed points in d+E dimensions if it
is renormalizable and asymptotically free in d dimensions. It is
well known that asymptotic freedom can only be achieved for a ~~
theory in four dimensions by giving the coupling constant an unphysical negative value. However, there are plenty of other
asymptotically free renormalizable theories, such as chiral dynamics
in two dimensions 6 and non-Abelian gauge theories in four dimensions,
which are not plagued by negative-energy problems.)
The asymptotically safe theories of a scalar field in five
dimensions are characterized by two free parameters: one dimensionless parameter picks out a particular trajectory in the twodimensional surface of trajectories which hit the Wilson-Fisher
fixed point, and one parameter with the dimensions of mass gives
the value of the cut-off A at which any given point along this
trajectory is reached. In order to perfect the analogy with general
relativity, we can eliminate the first parameter by demanding that
the theory have zero boson mass. This is done simply by assuming
that the theory lies on a trajectory which not only hits the WilsonFisher fixed point for A + 00, but also hits the Gaussian fixed point
for A + O. (To see that there is likely to be such a trajectory,
note that in d > 4 dimensions the symmetry ~ + -~ eliminates all
infrared-repulsive eigenvectors at the Gaussian fixed point except
for the one corresponding to the "interaction" ~2. Hence it is only
necessary to adjust one dimensionless parameter to hit the Gaussian
fixed point for A + 0, and we do have one free parameter at our disposal.) This theory is then described by a single free parameter,
with the dimensions of a mass, just like general relativity.

CRITICAL PHENOMENA FOR FIELD THEORISTS

37

Perhaps gravitation works this way. It may be that general

covariance rules out any renorma1izab1e theory of gravitation in
four dimensions, just as the symmetry under ~ ~ -~ rules out any
renorma1izable theory of a scalar field in 5 dimensions. Nevertheless, nature may achieve asymptotic safety anyway, by picking out
trajectories which hit some fixed point other than the Gaussian
fixed point for A ~ 00. For this to be possible, there must be some
non-trivial fixed point with at least one ultraviolet-attractive
eigenvector.
Unfortunately, it is not so easy to check that this is so.
General relativity becomes formally renormalizable in d = 2 dimensions, so one might try to study the fixed points in four dimensions
by working in d = 2 + dimensions and expanding in E:. However,
general relativity actually becomes a trivial theory in two dimensions (because R~v- ~ g~v R vanishes identically) and it is not clear
how to expand around d = 2. Work on this is in progress.
In any case, I think it is wrong to hope that we will learn how
to make sense out of arbitrary non-renorma1izable field theories.
It would be a disaster for theoretical physics if it were found that
the infinite variety of physical theories with all possible couplings
were all equally acceptable. We would then have no guide in understanding how nature picks out the particular theory that describes
our world. What we need instead is some principle, like asymptotic
safety, which picks out a very limited class of physically acceptable
theories - perhaps renorma1izable, perhaps not.
Acknowledgments
I am very grateful to Paul Martin and David Nelson for frequent
enlightening discussions on the theory of critical phenomena throughout the last year. I would not have been able to prepare these lectures without their help in learning this subject. I also wish to
thank Edouard Brezin, Sidney Coleman, Leo Kadanoff, Philippe Nozieres,
Kenneth Wilson, and Edward Witten for useful conversations, and
Antonio Zichichi for his kind hospitality at Erice.

s. WEINBERG

38

Appendix
DERIVATION OF THE ONE-LOOP EQUATIONS
\ole will now derive the renormalization-group equations satisfied by the coupling functions. For the present, it will be convenient to work with the original functions ~(PI Pn;A) rather
than the re-scaled functions gn(tl . tn;A). It will also be convenient to treat the Q4adratic coupling Uz as much as possible like
any other kind of interaction. In order to accomplish this, let us
write the original function uz,OLD appearing in the Hamiltonian as
u

(-+

-+ A)

z,OLD PI'PZ'

= U z,NEW (-+PI'PZ'
-+ A)

+ -I

-+

d -+

-+

GNEW(p)o (PI +p)

(A.l)

where GNEW is arbitrary but A-independent. (We can take GNEW of

the form (apZ+b)-I, but this is not necessary.) The propagator can
then be taken as
(2n)-d GNEW (p)6(A-lpl)
(A.2)
and U z NEW is now regarded as just another interaction. From now
on, we' drop the label "NEW": G and U z are to be understood as GNEW
and uz,NEW until further notice.
We consider the set of all connected Green's functions
C(PI .. Pn ), with outgoing momenta PI Pn. These functions are
calculated with vertices - (2n)d u Z' - (2n)d u q ' etc., and with propagators (2n,d G(p)6(A-lpl), except that the 6-function is omitted on
external lines. (We do not assume that the external momenta are
below the cut-off.) Our demand on the couplings is that these
Green's functions be independent of the cut-off:
(A.3)

for all n and all momenta. When we differentiate C with respect to

A, we encounter terms of three different kinds, shown in Figure 3:
(a) The derivative d/dA may act on one of the A-dependent coupling
functions in the graphs for C. Such terms may be written

id
-+ ... -+
(mn) (k
-+ ... -+
1-+
- (2n) d L\' dd k I .. ddkm
dA uml
(k
km
; A)-I r A
k I
P .. -+
pn )
Im
m
.
(mn) -+

-+ 1-+

-+

Here r A
(k . k P . p ) is a sum of graphs having m external
HI
ml
n
-+-+
lines carrying incoming momenta kl ... km and n external lines carrying outgoing momenta PI ... Pn. It is defined with cut-off propagators for the external k-lines but propagators without cut-offs on the
p-lines, except that if a k-line turns into a p-line with no interaction, there is just one propagator without cut-off. Since we

CRITICAL PHENOMENA FOR FIELD THEORISTS

d
dt\

39

X~I--_
(a )

( c)
( b)
Figure 3. An example of the variation of a connected Green's
function with cut-off. Here the shaded circle represents the
total connected Green's function with four external lines; the
circle with a vertical bar represents the sum of all graphs
such that every line leaving on the right is connected to at
least one line entering on the left; the darkened circles represent various interactions Un; the x in the first term on
the right represents the derivative of un with respect to A;
and the x in the following two terms represents She replacement of a cut-off 6-function by a a-function a(lql-A). For
the sake of clarity, the only graphs shown are those with six
lines entering the barred circle from the left.

want to include all connected diagrams in C, fA consists only of

the graphs in which each p-line is connected by some path to at
least one k-line, so that the interaction urn can connect the whole
graph together. Apart from this proviso, fl rnn ) may contain disconnected parts.
(b) The derivative d/dA may act directly on one of the cut-off functions 6(A-lql) associated with anyone of the internal lines of C.
The two ends of this internal line may terminate either in the same
vertex, or in two different vertices. If they terminate in the
same vertex, there must be some other set of m ~ 2 lines with outgoing momenta ~l ~m attached to this vertex. Also, since C is
connected, each p-line will be connected by some path to at least
one of the k-lines. The contribution of such graphs to dC/dA is
then

s. WEINBERG

40

(c) Finally, when d/dA acts on the cut-Dff function associated with
an internal line which ends in two different vertices, there will
be two sets of lines with outgoing momenta It~ ... It~ and It~' ... It~
respectively, attached to these two vertices. Again, since C is
connected, each p-line is connected by some path to at least one
It'- or It"-line. The contribution of such graphs to dcl dA is then

~(27f)d l: (r;s~!
Jiqik~ ... i k 'r
r.s.
r,s

ik'; ... ik"o(A-lql)

s

-+, ... -+"

-+ -+" ... -+"
G(-+)
q uri (-+
q,k
k ;it)u + ( -q,k
k ;A)
1
1
r
Sl
1
S

x rA
( r+s ,n) (k-+' ... -+k' k-+" ... -+"
k 1-+
p -+
p )

sIn

We can now put this all together, and write the result as
-+
d ~ J d-+
d-+
-+
-+
(mn) -+
-+ 1-+
-+
... p ) = -(27f) L d k ... d k BA (k ... k )r A
(k ... k p ... p )
m=2
1
m
1
m
1
mIn
n
where
(A.4)
-+
-+
d
-+
-+
B, (k ... k ) - d' u (k ... k ; A)
1'1m
it
m 1
m
d

-+

~(p

-'2

fd qo(A- 1-+1q )G(q)u

-+
If

'\

k-+k '&k"

-++ ... -+,

-+-+-+
(q,lC
k ;A)u + (-q,k'~
.. k";A)
r,+l
1
r
s lIS
(A.5)

The sum in the last term runs over all values of rand s with r + s =
m, and over all m!/r!s! ways of partitioning the momenta It l ... Itm
into subsets It~ ... It~ and It'; ... t~. We see that for C to be independent of A, it is sufficient that
-+

-+

B, (k ... k ) = 0
it

(A.6)

-+
-+
for all m and all k l ... k m. This is one form of renormalization-group
equation. (See Figure 4.) We note further that the kernel
r~mn)(kl ... ~Ipl ... n ) generally will have an inverse, because in the
absence of interactions it is simply proportional to

-+ -+
-+ -+
-+
-+
o o(k -p ) ... o(k -p )G(k ) ... G(k )
mn
11
mm
1
m
Therefore we expect (A.6) to be necessary as well as sufficient.
[Equation (A.6) is similar to a set of equations derived in
quite a different way by Wegner and Houghton. 7 They differ in that
G appears instead of u 21 , and that the sum in the last term includes

41

CRITICAL PHENOMENA FOR FIELD THEORISTS

Figure 4. The Wegner-Houghton equations for u 6 These are

derived here by demanding that the A-derivative in Figure 3
vanish.
terms with r or s equal to one. The difference arises because our
cut-off procedure is different for external lines.]
These equations have a well-known drawback: the last term in
Eq. (A.5) is highly discontinuous in the external momenta, contributing only when the total momentum of some subset with an odd
number of lines has an absolute value equal to A. Also, these
equations mix terms with different numbers of loops, so they are
not very convenient for use in iteration schemes. To avoid these
problems, we introduce a new set of coupling functions.
The reason for the discontinuities in the last term in (A.5)
is obvious: in calculating Green's functions we must include tree
graphs, whose internal lines contain cut-off functions 8(A-lql).
Even though all external lines of such graphs lie below the cutoff A, the total momentum of subsets of the external lines may be
above this cut-off, and the contribution of such graphs will contain discontinuities in the external momenta. The total Green's
functions certainly do not contain such discontinuities (they are
A-independent!) so in order to cancel them out the Un must contain
A-dependent discontinuities of some sort. These are automatically
generated by the last term in Eq. (A.5).
The diagnosis immediately suggests a cure. Let us introduce
a new set of coupling functions vn(~l ... tn;A), by the prescription
that -(2n)d un is to be constructed as a sum of all possible tree
graphs, with vertex functions -(2n)d vn , and with internal-line
factors
(2n)-d G(q)8(lql-A)
(A.?)
When we take the derivative of Un with respect to A, we then encounter terms of two different types (see Figure 5):
(a) The derivative d/dA may act on one of the cut-off functions
8(lql-A) associated with an internal line of the tree, changing it
to -o(lql-A). The two ends of this internal line must be connected
to separate trees built out of v-vertices.
(b) The derivative d/dA may act on one of the vertex factors, say
a v m. Since the whole graph is a tree, each of the m lines attached
to this vertex must itself be connected to a separate tree built out
of v-vertices and propagators (A.7).

S. WEINBERG

42

1K

~ 0_ +.~
+

x~+ X<+X~

+ ...
Figure S. A-derivative of a Un-function expressed in terms
of A-derivatives of vn-functions. Here dark circles represent Un-functions; light circles represent vn-functions;
the x in the first term on the right represents the replacement of a cut-off 8-function with a a-function; and
the x's in the other terms represent A-derivatives acting
on Un-functions. When inserted into the equations of
Figure 4, this yields the one-loop equations.

When we insert these results for dUrn/ciA in Eq. (A.S), we find

that term (a) cancels the last term in (A.S), so the sum of (b) and
the second term of (A.S) must vanish.
NOW,Urn+2 in Eq. (A.S) is itself a sum of tree graphs built out
of v-vertices. There must be a unique path from the line with
momentum q to the line with momentum -q, and each v-vertex along
this path is attached to a tree built out of v-vertices. The
v-vertices in term (b) are also connected to such trees, so the
trees may be divided out. leaving us with

o=

(A.8)
The last sum is over all values of ml

m2

,m, with

and over all m!/m l !m 2 ! ... m! ways of partitioning the ~.~omentq

into sets, with the j-th set consisting of momenta k~J) ... k~J).
I

It should be noted that the last sum in (A.8) includes terms

in which any of the mj are zero. These can be trivially summed:
their effect is simply to change' G(cl) into

CRITICAL PHENOMENA FOR FIELD THEORISTS

-+

-+

-+

-+

-+

-+

G'(q;A) = G(q) - G(q)V2 (q,-q;A)G(q) +

-+

-+

-+

-+-1

G(q) [1 + V 2 (q,-q;A)G(q)]

where

v 2 is v 2 with the o-function factored out:

-+-+

_-+-+

d-+-+

V2 (ql,q2;A) = V2 (q1,q2;A)o (q1 +q2)

This leads us to introduce a new v 2 function

-+

-+

V~(q1,q2;A)

-+
d-+ -+
= G- 1 (q1)O
(q1+q 2)

-+

-+

+ v 2 (q1,q2;A)

which has the same A-derivative as v 2

dv~

dA

dV 2
dA

but which yields the corrected propagator

-+

-+

1-+

d-+-+

V;(Q1,Q2;A) = G'- (Q1;A)o (Q1+ q 2)

That is, Eq. (A.8) still holds if we replace G with G', v 2 with v;,
and restrict the sums so that mj > 0 for all j. From now on we will
drop the primes on v 2 and G, ana interpret (A.8) to include only
terms with mj > O.
We now make the transition to the re-scaled variables.
new coupling functions:
-+

(~

Define

-+

Pn.A)
A'
-

(A.9)

and a new propagator

(A. 10)

so that
(A. 11)

Inserting (A.9) and (A.lO) in (A.8) gives

44

S. WEINBERG
x

m9, 2

()(,k

m1

17 I

-+ -+ -+( 1 )
-+( 1)
+2 (9.,-9,1,9: 1 .. 9,
;1I.)8()(,1 -1)L'I()(,1;1I.)
m
1

7 -t;(k)
-+(k)
,-)(',9,
9:
;A)
1
~

(A.12)

After factoring out a momentum-conservation delta function, this

becomes just the general one-loop equation (8.4).
REFERENCES
1.

For surveys of the modern theory of critical phenomena, including

references to the original literature, see the following reviews:
K. G. Wilson and J. Kogut, Physics Reports l2C, No.2 (1974);
M. E. Fisher, Rev. Mod. Phys. 46, 597 (1974~E. Brezin,
J. C. Le Guillou, and J. Zinn-Justin, in Phase Transitions and
Critical Phenomena~ ed. by C. Domb and M. S. Green (Academic
Press, New York, 1975), Vol. VI; F. J. Wegner, in Trends in
Elementary Particle Theory (Springer-Verlag, Berlin, 1975),
p.17l; K. Wilson, Rev. Mod. Phys. 47, 773 (1975); Shang-Keng Ma,
Modern Theory of Critical Phenomena-CWo A. Benjamin, Inc.,
Reading, Mass., 1976).

2.

A. L. Fetter and J. D. Walecka, Quantum Theory of Many-Particle

Systems (McGraw-Hill, Inc., New York, 1971), Chapter 7.

3.

S. Weinberg, Phys. Rev. D~, 3357 (1974); L. Dolan and R. Jackiw,

Phys. Rev. D~, 3320 (1974); D. A. Kirzhnits and A. D. Linde,
Zh. Eksp. Teor. Fiz. ~, 1263 (1974); C. W. Bernard, Phys. Rev.
D~, 3312 (1974).

4.

F. Bloch and A. Nordsieck, Phys. Rev. ~, 54 (1937). For a

version in modern notation, see S. Weinberg, Phys. Rev. 140,
B5l6 (1965).

5.

M. Gell-Mann and F. E. Low, Phys. Rev. 95, 1300 (1954).

6.

The fixed point in the nonlinear a-model in 2 + E: dimensions has

been under intensive study lately; see W. A. Bardeen, B. W. Lee,
and R. E. Shrock, Ferrnilab-Pub-76/33-THY, March 1976; E. Brezin,
J. Zinn-Justin, and J. C. Le Guillou, Saclay preprints, May 1976.
The motivation of these studies appears to be quite different
from that described here.

7.

F. J. Wegner and A. Houghton, Phys. Rev. A.., 401 (1973).

45

CRITICAL PHENOMENA FOR FIELD THEORISTS

DIS C U S S ION S

CHAIRMAN:

Prof. S. Weinberg

Scientific Secretary:

F. Posner

DISCUSSION 1
FERBEL:
Could you elaborate on the question of Higg's boson production
relative to W production in hadronic collisions?

WEINBERG:
Higg's production has been discussed in detail in a recent paper
by Ellis, Gaillard, and Nanopoulos. I believe that they do the calculations you ask about, but I do not know the answer. I will make
one point though: Higg's bosons' couplings are proportional to mass,
so Higg's bosons tend to be emitted from internal lines of heavy particles. A good place to look for Higg's bosons therefore is in neutrino reactions, where they are eIT.itted from the exchanged W line.
Ellis, Gaillard, and Nanopoulos, and LoSecco have calculated the
probability for Higg's production near threshold in neutrino-nucleon
reactions to be about 10- 5

FERBEL:
If charm violating processes occur, might one expect final state
correlations such as K+K+ or K+e+?

WEINBERG:
I have been assuming that the neutral currents do conserve charm.
If there is a milliweak ~C = 2 neutral Higg's exchange, then DO-Doooscillations will be much faster than DO decay. The result would be
that an incoherent mixture of D~ and/or D~ would be produced, each decaying equally into Ks or leptons of either charge.

s. WEINBERG

46
PARSONS:

How easy would it be to introduce more flavours into your model?

WEINBERG:

This model needs four quarks because if there were more, this
would be likely to introduce CP violation into the W-exchange process.
To include a bottom quark b, one requires

as the appropriate multiplet for weak interactions where

a~ + a~ + a~ = 1 .

The experimental data on S decay provides information on u + d.

The experimental data on A decay provides information on u + s. An
over-all normalization can be fixed by ~ decay. This sets tight
limits on al and a2, so that we must have la31 $ 10- 3 Such small
numbers in the theory are rather unsatisfactory.
It is, however, possible to introduce extra doublets into the
theory which do not mix with u, d, s, and c. These can be used to
cancel anomalies due to heavy leptons. The real problem would be if
neutrinos were found to cause transitions between u and b quarks,
for instance, thus indicating a non-zero mixing angle a3.
CRONIN:

In order to be sure of your ideas about the nature of CP violation, it is really insufficient to have the electric dipole moment
of the neutron come out ~ 2 X 10-2q e-cm. Thus one really needs to
see the CP-violating effects of the Higg's scalar. What is the nature
of the CP-violating effects of the Higg's scalar, if it could be isolated?
WEINBERG:

Let me first answer the question I thought you were going to

ask: What experimental evidence could be used to check that Higg's
exchange is responsible for CP non-conservation? Since Higg's bosons
couple essentially to mass, we would expect CP-violating amplitudes
of order 10- 3 in K~3 decay, but no corresponding effect in S decay.

As to the way that CP violations would show up in Higg's boson

decay, I have not worked it out. It would be a good exercise for a
student. The Higg's decays should violate CP strongly.

CRITICAL PHENOMENA FOR FIELD THEORISTS

47

LEADER:

You have offered us a beautiful and natural mechanism for producing a small number, namely 10- 3 , in amplitude. However, you rely
for this on a knowledge of the quark and the Higg's masses. Could
you explain what it is that gives you such confidence in your knowledge of these masses?
WEINBERG:

As regards the quark masses, I am just going along with all the
standard ideas on quarks and constituent models. As for the Higg's
mass, it is expected to be of order If x 300 GeV, where f is the 4
coupling constant. If f is of order a, as generally supposed, then
the Higg's mass is of the order of the intermediate vector boson mass.
Linde and I have recently shown that there is an effective lower
bound of order a Z In the simplest SU(2) x U(l) model, the Higg's
mass is greater than 3.72 GeV.
POSNER:

A rather elementary question: Why does a CP and P nonconserving milliweak interaction imply a detectable electric dipole
moment for the neutron?
WEINBERG:

If CP is broken, but not P, then there is no moment. If the

neutron is really two d and one u quarks, then the moment equals
4h (the moment of the d) - 3 (the moment of the u). The moments
arise from the virtual processes

H+ +

d + H

(or d)

+ c (or u)

MARCIANO:

Is the statement that the CP violation is due to the H+ propagator and not the W propagator, a gauge-dependent statement? Could
you clarify your statement that you feel that there will be CP violation even for the case l = z?
WEINBERG:

The sum of the W-propagator and the part of the H propagator

corresponding to a Goldstone boson is gauge invariant and conserves
CPo The remaining part of the H propagator is then also gauge invariant, and violates CPo Thus the amount of CP violation is gauge
invariant.

48

S. WEINBERG

Even if 1 = 2, there can be a CP violation in the interaction

of these Higg's bosons with other Higg's bosons that cannot couple to
quarks. I have not tried to calculate such effects.
FREEDMAN:

What classes of theories permit Higg's bosons which do not violate CP? Huw do these couplings differ from those discussed in this
morning's talk? What, if any, characteristics are different between
the two, mass, etc., and can one incorporate both in a single model?
WEINBERG:

Such theories are certainly possible. One can always impose

CP on L. Then Crs would have to be real, or have phases that could
be made real. However, for a finite range of parameters there will
still be a spontaneous breaking of CPo

DISCUSSION 2
YOON:
Does the fact that critical phenomena in boson systems has its
origin in the infrared behaviour of the system reflect the physical
picture of phase transitions arising from long-range correlations?
How does one understand critical phenomena in fermion systems where
there are no infrared divergences?
WEINBERG:

Second-order phase transitions can occur in systems composed

purely of fermions, because the fermions can have bosonic collective
excitations. That is a large subject, which I will not go into in
these lectures.
ALVAREZ:

In 4 quantum field theory, the two-point function is ultraviolet

quadratically divergent. In the finite temperature theory, the three
momentum integral is logarithmically divergent. When the energy sum
is performed, do the logarithms sum up to power behaviour?
WEINBERG:

Of course, they had better, because the divergences must be the

same at finite temperature as at zero temperature. The Lagrangian
does not know what the temperature is, so the counter-terms available
to absorb infinities are necessarily temperature-independent. However,
it is difficult to do the energy sums as you suggest after doing the

CRITICAL PHENOMENA FOR FIELD THEORISTS

49

momentum integrals. I have found it much easier to do the energy sum

first by converting it into a contour integral, and then do the momentum integral.
PAULI:

In the statistical mechanics formalism, if you take the fourth

derivative of the partition function, you get a Feynman-like graph
with four external legs. We know what this corresponds to in quantum
field theory vis-a-vis the S matrix and differential cross-sections.
We also know is field theory how to include symmetries. What physical and formal mathematical changes occur when we do this in a statistical mechanical form at finite temperatures?
WEINBERG:

Symmetries like rotation or isospin invariance govern the temperature Green's functions just as they govern the S matrix or the
Green's functions in quantum field theory. The only symmetry that
is really fouled up by a finite temperature is Lorentz, or Galilean,
invariance.
KLEINERT:

If you pass the critical point of a ferromagnet, however, do

the 4-point Green's functions not change symmetry?
WEINBERG:

No, the Green's functions as I have defined them have the full
symmetry of the underlying theory, i.e. of the Hamiltonian. The
symmetry-breaking affects the various expectation values calculated
using these Green's functions.
POSNER:

A useful and powerful formulation of quantum field theory is by

means of functional integrals. Are there any particular pit-falls in
dealing with statistical mechanics by functional techniques? In field
theory functional integrals, one rotates from t1inkowskian to Euclidean
space to avoid ambiguities. In statistical mechanics one has iw instead of po. How does this affect the functional integrals of statistical mechanics and their possible ambiguities?
WEINBERG:

Your classmate at Harvard, Claude Bernard, has worked out a very

nice formulation of statistical mechanics by using functional integrals.
He has derived Feynman rules at a finite temperature in a gauge theory.

s. WEINBERG

50

Because we are now in Euclidean space, the troubles with cut-offs due
to the metric are gone. Things work very well.
MARCIANO:

What plays the role of an infrared regulator in this formalism?

WEINBERG:

The partition function is not well-defined at the critical temperature. The temperature e regulates these divergences since for
e ec ' the partition function Q is finite and well-defined.

DISCUSSION 3

(Soientifio

Seoreta~:

O. AZvarez)

MARCIANO:

What does the renormalization group have to do with infrared

divergences?
WEINBERG:

It is unfortunate that Wilson, probably out of modesty, called

these equations renormalization group equations, because of the formal
appearance to the equation of Gell-l1ann and Low. The Wilson work concerns infrared behaviour while Gell-Hann and Low considered ultraviolet behaviour. The "floating renormalization point" introduced by
Gell-llann and Low has a formal similarity to the floating cut-off
used by Wilson, but the purpose is entirely different.
GARCIA:

Following your discussion of the Bloch-Nordsieck problem, can you

find a similar approach for coupled massless fields?
WEINBERG:

I believe there is a renormalization group argument, but I have

not been able to complete it.
GARCIA:

I do not clearly see the connection between your use of the renormalization group and the idea of "thinning" out of the degrees of
freedom as used by Wilson and Kadanoff.
WEINBERG:

This is an example of the trouble with translating solid state

language into the language of field theorists. Kadanoff's idea of
block spins, or changing the lattice spacing by integer multiples

CRITICAL PHENOMENA FOR FIELD THEORISTS

51

would be equivalent to changing the cut-off by integer multiples.

In quantum field theory we usually change the cut-off continuously,
but we could change it by discrete steps.
PAULI:

When you compared the work of Brezin et al. with the work of the
Wilson school, you stated that every eigenvalue of M is an eigenvalue
of M, but not all eigenvalues of U are eigenvalues of M. How do you
know that Brezin' s method will give the repulsive and the "important"
eigenvalues of MZ
WEINBERG:

The only way to check this is to calculate the eigenvalues, using

an approximation scheme such as the expansion. Brezin et al. work
in 4 - dimensions, and find one attractive and one repulsive eigenvalue. They cannot show, within a strictly renormalizable theory,
that all ignored eigenvalues are attractive, but this seems reasonable, and can be shown by introducing non-renormalizable perturbations.
PAULI:

You stated that if you have more than two repulsive eigenvalues
in three dimensions, the expansion is necessary. Why can you not
vary temperature and say magnetic field? Are there alternatives to
the expansion?
WEINBERG:

I must have been unclear. We believe that ordinary second-order

phase transitions are associated with fixed points that have only one
repulsive eigenvector, because these transitions can be brought about
by adjusting only one free parameter, the temperature. Even with a
magnetic field, the Gaussian fixed point cannot describe an ordinary
second-order phase transition, because introduction of the field destroys the ~ ~ -~ symmetry, and there are therefore three repulsive
eigenvectors at the Gaussian fixed point.
One alternative to the expansion is an expansion in liN,
where N is the number of fields. I will not discuss this in these
lectures.
FREEDMAN:

We learned that in three dimensions it was not possible to do

perturbation theory about the Gaussian fixed point because of two
repulsive eigenvectors. In order to circumvent this we go to 4 -
dimensions where only one repulsive eigenvector exists. Although
we can perform our calculations now, what can we learn about the

s. WEINBERG

52

physics in our original three dimensional problem?

WEINBERG:

The hope is that the eigenvalues do not change sign between

1 and = 1. The qualitative features of the physics will then
remain.

PHAM QUANG HUNG:

To which physical situations corresponds the limit A

O?

WEINBERG:

The Gaussian fixed point has a zero eigenvalue corresponding to

the interaction ~6. However, it is only possible to reach this fixed
point if two parameters are adjusted to eliminate components of the
trajectory along the two repulsive eigenvectors which correspond to
the ~2 and ~~ interactions.

MONOPOLES AND FIBER BUNDLES

Chen Ning Yang

State University of New York
Physics Department, Stony Brook, N.Y. 11794
Magnetic Monopole and Need to Introduce Sections
The magnetic monopole is the magnetic charge. While the idea
of magnetic monopoles must have been discussed in classical physics
early in the history of electricity and magnetism, modern discussions
date back to 1931 in the important paper of Dirac l in which he pointed out that magnetic monopoles in quantum mechanics exhibit some
extra and subtle features. In particular, with the existence of a
magnetic monopole of strength g, electric charges and magnetic
charges must necessarily be quantized, in quantum mechanics. We shall
give a new derivation of this result in a few minutes.
If one wants to describe the wave function of an electron in
the field of a magnetic monopole, it is necessary to find the
vector potential X around the monopole. Dirac chose a vector
potential which has a string of singularities. The necessity of
such a string of singularities is obvious if we prove the
following theorem 2 .
Theorem. Consider a magnetic monopole of strength g ~ 0 at
the origin and consider a sphere of radius R around the origin.
There does not exist a vector potential Xfor the monopole
magnetic field which is singularity free on the sphere. This
theorem can be easily proved in the following way. If there were
a singularity free
we consider the loop integral

around a parallel on the sphere as indicated in Figure 1.

53

By

C.N.YANG

54

FIGURE 1.

A sphere of radius R with a magnetic monopole

at its center.
The parallel divides the sphere
into two caps a and S.

Stoke's theorem this loop integral is equal to the total magnetic

flux through the cap alpha:

yA~dx~

Qa .

(1)

Similarily we can apply Stoke's theorem to cap S obtaining

jA~dx~

~S ..

(2)

Here ~a and ~S are the total upward magnetic flux through the caps
a and S, both of which are bordered by the parallel. Subtracting
these two equations we obtain

o=

~A

- ~B'

(3)

which is equal to the total flux out of the sphere, which in turn
is equal to 4ng I O. We have thus reached a contradiction.
Having proved this theorem, we observe that R is arbitrary.
Thus one concludes that there must be a string of singularities
or strings of singularities in the vector potential to describe the
monopole field.
Yet we know that the magnetic field around the monopole is singularity free.
This suggests that the string of
singularities is not a real physical difficulty.
Indeed the
situation is reminiscent of the problem that one faces when one wants
to find a parametrization of the surface of the globe. The
coordinate system that we usually use, the latitude and the longitude,
is not singularity free.
It has singularities at the north pole and
at the south pole. Yet the surface of the globe is evidently without
singularities. We deal with this situation usually in something
like the way illustrated in figure 2. We consider a rubber sheet
with nicely defined coordinates and stretch and wrap it down onto the
globe so that it covers more than the northern hemisphere.
Similarly,

MONOPOLES AND FIBER BUNDLES

rl rr7,

LIIJ7

y/ rrl

lO

rlTTJ
FIGURE 2.

Hethod of parametrizing the globe.

FIGURE 3.

Division of space outside of monopole g into

overlapping regions Ra and Rb

55

we consider another rubber sheet with nicely defined coordinates and

stretch and wrap it upwards so that it covers more than the southern
hemisphere. We now have a double system of coordinates to describe
the points on the globe. The description is analytic in the domain
covered by each sheet, if we had done no violence in the stretching
and wrapping. In the overlapping region covered by both sheets,
one has two coordinate systems which are transformable into each
other by an analytic non-vanishing Jacobian. This double
coordinate system is an entirely satisfactory way to parametrize
the globe.
Following this idea we shall now try to exorcise the string
of singularities in the monopole problem by dividing space into
two regions. We shall call the points outside of the origin,
above the lower cone in figure 3, region Ra. Similarly, we shall
call the points outside of the origin, under the upper cone, Rb.

c. N. YANG

56

The union of these two regions gives all points outside of the
origin. In Ra we shall choose a vector potential for which there is
only one non-vanishing component of A, the azimuthal component:
r

s~n 8 (1 - cos 8),

(4)

It is important to notice that this vector potential has no

singularities anywhere in Ra. Similarly in ~ we choose the vector
potential
(5)

which has no singularities in Rb. It is simple to prove that the

curl of either of these two potentials give correctly the magnetic
field of the monopole.
In the region of overlap, since both of the two sets of vector
potentials share the same curl, the difference between them must be
curlless and therefore must be a gradient. Indeed a simple
calculation shows
(6)

where <P is the azimuthal angle. The Schrodinger equation for an

electron in the monopole field is thus
2
1
(p-eA ) lj! + Vlj!
2m
a
a
a

Elj! ,
a

in R
a'

1
2
2m (p-eAb ) 1jJb + V1jJb

E1jJb'

in Rb ,

where lj! and 1J!h are respectively the wave functions in the two regions.
The fac~ that the two vector potentials in these two equations are
different by a gradient tells us, by the well known gauge principle,
that lj!a and 1J!b are related by a phase factor transformation

or

1jJa

1jJa

S1J!b'

(7)

exp (iea),

[exp (2iq<P)]1jJb' q

ego

(8)

Around the equator which is entirely in R ,1jJ is single valued.

Similarly, since the equator is also entifelyain Rb , 1jJb is single
valued around the equator. Therefore, S must return to its original
value when one goes around the equator. That implies Dirac's
quantization condition:
2q

integer.

(9)

57

MONOPOLES AND FIBER BUNDLES

Hilbert Space of Sections

Two ~'S, ~a and ~b' in Ra , and Rb respectively, that satisfy
the condition of transltion (8) in the overlap region, is called a
section by the mathematicians. We see that around a monopole the
electron wave function is a section and not an ordinary function.
We shall call these wave sections.
Different wave sections (belonging to different energies, for
example), clearly satisfy the same condition of transition (8) with
the same q. Thus we need to develop3 the concept of a Hilbert space
of sections. To do this we define the scalar product of two
sections ~,n (for the same q) by

(n,~) = fn*~d3r.
(The question of convergence at r
Notice that in the overlap
(na)*~a

(10)

0 and r

00

is ignored here).

(ll)

(nb)*~b

so that (10) is well defined.

It is clear that if

is a section, then

x~

is also a section,

since

Thus x is an operator in the Hilbert space of sections. Similarly

we prove that the components of (~-eA) are operators, but those
of p are not. Furthermore ~ and p-eA are both Hermitian.
Following Fierz 4 we shall now try to construct angular momentum
operators. Define
-+

-+

-+

-+

-+

r x (p - eA)

(12)

It is clear that L , Ly ' L z are Hermitian operators on the Hilbert

space of sections. x The following commutation rules can be easily
verified:

0,

[L ,x]
x

iz,

[Lx'Y]

[Lx'px - eAx ]

0,

[Lx'pz - eA z ]

[L ,z] = -iy,
x

[L ,p - eA ] = i(p - eA ),
x y
z
z
y
i(p

eA ).
y

(13)

It follows from these that

[L ,L ]
x y

iL , etc.
z

(14)

58

c. N. YANG

EQ. (13). together

Lz are the angular
the Hilbert space,
(The singularities
they occur outside

with its consequence (14). show that Lx. Lv'

momentum operators.
We emphasize that neither
nor these operators, possess any "singularities".
of Aa and Ab are not real singularities because
of Ra and Rb' respectively.)
Monopole harmonics Y

q ..... ,m

-+

-+

Since ~r , L] = 0, we can diagonalize r and study operator L

for fixed r . I.e. we shall study sections of the form
15 (r

- rOH:,

where ~ is a section dependent only on angular coordinates Sand

<p. t operates then on "angular sections".
Eq. (14) shows that [L2, Lz ] = O. Simultaneous diagonalization
produces the familiar multiplets with eigenvalues R.,(R., + 1) and m
L2y

n'

q, .... ,m

R.,(R.,+l)Y

q, .... ,m

LY a
=mY n
z q, .... ,m
q, . . ,m,

(15)

where R., = O,~, 1, .. and, for each value of R.,. m ranges from -R., to 3
+R., in integral steps of increment. The Yq.R.,
are the eigensections
which we shall call monopole harmonics. We AWall show later that
the allowed values of R., and mare
Iql,
m

-R."

Iql + 1, Iql
-R., + 1, .. ,R."

+ 2, ... ,

and that each of these R." m combinations occur exactly once.

shall choose each Y normalized so that

ro
7T

27T

sinSdS flY
0

R.,

q, ,m

I d<P

= 1.

(16)
We

(17)

(Notice. that in Rab , kYq,t,m)aI2 = kYq,R."m)bl 2.) Different Yq,R."m

for a flxed q) are orthogonal, a fact one easily proves in the
usual way fro~ (15). We shall choose the phases of Yq,R."m such
that the matrlx elements of Lz , L Lz between the Y's conform
to the convention adopted in ch. 2 of Edmonds' bookS. In
particular
(L

+ iL)Y
y

q, .... ,m

(R.,-m)~(R.,+m+l)~Y q .....n , m+l

(18)

These monopole harmonics will be explicitly exhibited. Each is

analytic. That is, (Y R., ) is analytic in Rand (Y R., )b is
analytic in~. The s~t o~ ~ll monopole harmo~ics for q ,m
a fixed q forms a complete set of sections. as we shall see.

59

MONOPOLES AND FIBER BUNDLES

Explicit expressions for Y

q,&,m
Stating from (12) one easily verifies
(19)
mY
mY
Eq.

q,&,m

L Y

(-ia

L Y

(-ia

z q,&,m

q,&,m

z q,&,m

<p

q)Yq , &,m' in Ra'

<p

+ q)Yq & m' in Rb

, ,

(20)

(20) shows that

Y
q,&,m

Y
q,,m

q, &,m
q,&,m

(e)ei(m+q)<p

in R
a'

(e)ei(m-q)<p

in Rb

(21)

The condition for a section shows that [8

(e)] =[8 n (e)]b
q,&,m
a
q,-<-,m
in the overlap. They are, in fact, the same function. Apply' (19)
to Y
An explicit evaluation of the operator [rx(p- eA)]2
.q,&,m
act1ng on Y
gives
q,&,m
[&(& + 1) - q2]8

q,,m

=[_

1 ~ sine ~ + ~
sine ae
ae
sin e

(m + q cos e)

2J 8 q,!C,m
n

(22)

Writing cose = x, this gives

2"
'1
.2
-(l-x )0 +2x8 + l-x 2 (m + qx) 8,
-1

1,

(23)

where prime means differentiation with respect to x. This

equation can be treated in the usual way, through analyzing the
indical equations at x = l. We shall, however, pursue a
different method which yields the normalization constant and phase
factor automatically.
Before proceeding we note that since Y is single valued in each
region, (21) shows that
m - q = integer.
Thus

&-

integer.

(24)

c.

60

N. YANG

Now (19) shows that

( + 1)

~ q

(25)

Eqs. (24) and (25) show that the allowed values of are among those
given in (16).
We shall now show that each value of in (16) is allowed, by
constructing, for each of them, the explicit function e 0
q, )(',m

q,,-

Jl~-q 11 +x Hq , -

q,!C

Iq I

= integer ~ 0,

(26)

where
N

>

q,

o.

(27)

To show this one substitutes (26) into (23) and verifies that the
latter is satisfied. The factor N is inserted so that
Yq,)(',
0
_0
is normalized in the senseqof (17).
!C
Repeated application of (18) onto the monopole harmonics
Yq ._ (given by (21) and (26)) leads to, (for ,m satisfying
(+6 the explicit expression for Y given 3 below. (As
stated above, this method leads to aatorn~tically normalized
Y n
starting from normalized Y 0 _0).
q,!C,m
q,)(', )(,
(Y

) = M
(1_X)a/2(1+X)S/2 p a'S(x)e i (m+q ),
q,,m a
q,,m
n
(28)

(Yq,,m)b= (Yq,,m)ae-2iq,
where
a = -q -m,

S = q - m,

n = + m,

!,;
(+m)!! 2

M
_ 2m!2+1 (-m)!
q,,m4n (-q)!(+q)!

x = cosS,

(29)
(30)

'

and pa,S(x) are the Jacobi polynomials,

n

pa,S(x)= ~(l_X)-a(l+x)-S __d__ [(l_x)a+n(l+x)S+n], (31)

n
2nn!
dx n
which are defined if
n,n + a, n + Sand n + a + S are all integers

o.

(32)

MONOPOLES AND FIBER BUNDLES

61

Completeness of Monopole Harmonics

For a given q(q may be negative) the set of Y .
with ~,m satisfying (16) form a complete set of orthonormal se~ti~s. I.e. every
continuous section (i.e. a section satisfying (9), with ~a and ~b
being continuous in Ra and
E

~,m

Proof:

~,m

can be expanded as a series

~)

q,~

can be expressed 3 in terms of

Y.

q,"',m
Now for fixed q = integer or half-integer, and q

+ m = integer,

there are four possible cases:

a 2!. 0, S

0, so that -m 2!. Iql

and

0, S

0, so that Iml s -q,

q~O

a "5: 0, S

9, s6 that Iml

0, S

0, so that m ~ Iql

q, q:..0 and
and

\l

=2 +

\l

and
\l

pial.
lsi (x)
\ l .

\l

Q.

m,

(33)

+ q,

(34)

q,

2 -

(35)
(36)

2 - m.

In case (33), the allowed values of ~, according to (16), are

g, = 1m I, 1m I + 1,... which are precisely
\l

0, 1, 2,

(37)

In case (34), the allowed values of g, according to (16) are g, = q,

+ 1, .. which are also precisely (37). Continuing this way we
conclude that given q
integer or half-integer, q + m = integer,
the allowed values of ~ according to (16) are always precisely those
given by (37).
-q

Now for fixed Ia I , I sI, the Jacobi polynomials P Iall sl , (v=0, 1,

\l

A.(m+ ) \l

2, .. ) form a complete set. The exponential functions e~~

q
m+q = all integers) also form a complete set. It can be proved from
these results that Yq, ..,..
Om forms a complete set of section for fixed q.
Examples and Analyticity of Y
For the case q
p-m,-m

RTm

= 0,

( _l)m

= S,

~:

!<

= ~ (g,+m):

(1

q,g"m

and (31) shows that

-x

2 ) m/2pm
2

(38)

62

C.N.YANG

where pR. is the associated Legendre function.

Substitution of (38)

into (28) shows that

yO.
= usual spherical harmonics Yo .
,"',m
N,m
We tabulate in table 1 a few of the monopole harmonics for
these examples illustrate the fact that Y 0
is analytic
q,N,m
everywhere. I.e., (Y q,,,,,m
0
)
(Y q,N.m
0
)b is
a is analytic in Rand
a

q =

~,1,

analytic

For

in~.

examp1e,(Y~)a

is clearly analytic in Ra'

which includes the point S=O, and

=I
is clearly analytic in

1 - cosS/ ~

(39)

which includes the point S=n.

Schrodinger Equation
It is simple to show by exp1ici.t evaluation, and with the aid
of (19) that
(p -

a
1
- 2ar
r

eA)2

(r 2

1
...2.)
ar + -[r
2
r

(p- eA)]

1
a
2
+ --.!.[L 2 - q 2 ].
- 2" ar (r L)
2
ar
r
r

(40)

The Hamiltonian thus commutes with L2 and L. Hence in solving for

we can choose spec~fic eigenvalues for L2
eigenfunctions of H
and L. I.e. we take
z
tfJ = R(r)Y

(41)

q,,,,,m.

obtaining
1
a (r2...2.) + R.(H1)-q
2 ar
ar
2
- -2
mr
mr

v _ E]R

o.

(42)

For the case that V = 0 this equation was solved by Tamm who found
that R is a Bessel function, if E~Q,
R = -

IiU

where

(kr) ,

(43)

63

MONOPOLES AND FIBER BUNDLES

k = v'zmE
If E

(44)

0, (42) has no meaningful solution.

Table 1
Examples of
9.

m
~

2"

2"

3
2

_~i4> 11-x(l+3x)

_~eo 11+x(l-3x)

3
2
3

(Y'4'iTY 9.! R, 2m) a

_e i 4>/1_x

R.
~

fuY in a region

e 0 /1+x
24>

1312e

1.

l1+x(l - x)

2"

-2"

13/2e-i /1-x(l+x)

.f3/4e 2i 4> (l-x)

-1

-h/2e i 4> h_x 2

h/4e o (l+X)

= cose.

To obtain Y
in
q,R"m

11,

apply (8).

Dirac E9.uation
Using the monopole harmonics discussed above, one can also discuss the motion of a Dirac electron in the field of a magnetic monopole. This was done in references 6 and 7 where bound states were
found.

64

C. N. YANG

Remarks
(A) It is important to realize that the above-described way
of using (A) and (A)b together to describe the magnetic field of
a monopole h~s an addltional advantage: It gives the magnetic field
H correctly everywhere. In older papers one oftentimes took a single
A with a string of singularities. Since by definition
11 (lIxA)

0,

the magnetic field described by IlxA must have continuous flux lines.
Thus its flux lines consist of the dotted lines of Figure 4, plus
the bundle of lines described by the solid line, so as to make the
net flux at the origin zero. Thus, IlxA does not correctly describe
the magnetic field of the monopole, a point already emphasized by
Wentzel. 8
(B) For ordinary spherical harmonics there are a number of
important theorems such as the spherical harmonics addition theorem,
the decomposition of products of spherical harmonics using ClebschGordon coefficients, etc. These theorems can be 9 generalized to
monopole harmonics.
(c) It is instructive to go back to the reasoning concerning
Figure 1 and try to repeat the steps for the combined A , A description of the magnetic field. Choose the parallel to be ~he gquator.
Then

jl(Afl\dX
Thus

41Tg

na - nS

nS

Jr (~)a

(~)8

] dl

which is, by (6), equal to the increment of a around the equator

i.e. 2g(21T) = 41Tg.
We have arrived at an identity. My reason for going through this
simple argument is that it embodies exactly the gist of the proof
of the famous Gauss-Bonnet-Allendoerfer-Weil-Chern theorem and the
later Chern-Weil theorem which play seminal roles in contemporary
mathematics.
As a matter of fact, gauge field, of which electromagnetism is the simplest example, is conceptually identical to some
mathematical concepts of fiber bundle theory. Table 2 gives 2
a
translation table for the terminologies used by physicists on the
one hand and mathematicians on the other. We notice that in particular Dirac's monopole quantization (9) is identical to the mathematical concept of classification of U(l) bundles according to the first
Chern class.

65

MONOPOLES AND FIBER BUNDLES

Table 20

Translation of Terminology

Gauge Field terminology

Bundle terminology

gauge (or global gauge)

principal coordinate bundle

gauge type

principal fiber bundle

gauge potential bk

connection on a principal
fiber bundle

S Eq.(8)

transition function

phase factor ~QP

field strength fk

parallel displacement

11

source

J l1

l1V

electromagnetism
isotopic spin gauge field

curvature
?

connection on a UI bundle
connection on a SU 2bundle

Dirac's monopole quantization

classification of UI bundle
according to firsE Chern
class

electromagnetism without monopole

connection on a trivial
UI bundle
connection on a nontrivial
UI bundle

electromagnetism with monopole

a I e ., electric source

66

C. N. YANG

"
\

FIGURE 4.

/
\

---..
\

Magnetic Flux lines due to A. Since V' (VxA) = 0,

flux lines are everywhere continuous. Hence there
is "return flux" along solid line.

TRIVIAL BUNDLE

FIGURE 5.

NONTRIVIAL BUNDLE
(MOEBIUS STRIP)

Examples of a trivial and a nontrivial fiber bundle.

67

MONOPOLES AND FIBER BUNDLES

The last two entries of the table identifies electromagnetism

with and without magnetic monopoles with connections on trivial and
nontrivial U(l) bundles. We can gain some understanding of these
facts by looking at (i) a paper loop and (ii) a Moebius strip
(Figure 5). If they are cut along the dotted lines, each would break
into two pieces. Looking at the resultant pieces we cannot differentiate between cases (i) and (ii). The two cases are different
only in the way the "resultant pieces" are put together. In case
(ii), a twist of one of the "resultant pieces" is necessary.
Thus case (i) corresponds to
Wa = SW b , where 5 = 1, (no twist);
and case (ii) corresponds to (q f 0),
Wa = SW b , where S = exp

(2iq~),

(twist).

A bundle where the transition function S is necessarily different

from 1 is called nontrivial, because a twist is needed. Hence
electromagnetism with a megnetic monopole is nontrivial [cf. (7)
and (8)].
FOOTNOTES
1

P.A.Mo Dirac, Proco Roy. Soco A133, 60 (1931).

2Tai Tsun Wu and Chen Ning Yang, Phys. Rev. D12, 3845 (1975).
3Tai Tsun Wu and Chen Ning Yang, Nuclear Phys. Bl07, 365 (1976).
4M Fierz, Helv. Phys. Acta 17, 27 (1944).
5
A.R. Edmonds, Angular Momentum in Quantum Mechanics (Princeton,1960).
6Yoichi Kazama, Chen Ning Yang and Alfred So Goldhaber, to appear in
Phys. Rev. D.
7Yoichi Kazama and Chen Ning Yang, to appear in Phys. Rev. D.
8

G. Wentzel, Prog. Theor. Phys. 5uppl. 37-38, 163 (1966).

9Tai Tsun Wu and Chen Ning Yang - to be published.

c. N. YANG

68

DIS C U S S ION S
CHAIRMAN:

Scientific Secretaries:

Prof. C.N. Yang

N. Parsons and B. Jancewicz

DISCUSSION 1
PHAM QUANG:

Could you comment on the differences between Dirac's quantization relation and Schwinger's quantization relation which has twice
the value of Dirac?
YANG:

We agree with Dirac's quantization relation for g and found no

reason for Schwinger's. Schwinger required two strings for reasons
of additional symmetry. In our approach, no strings are required at
all, provided we keep the vector potentials A~ and A~ in the allowed
regions Ra and Rb' respectively. One can deal with the singularities
in the forbidden regions arbitrarily. In our opinion, the quantization rule of Schwinger is groundless.
BERLAD:

Can one construct operators which will cause transitions between

states of different q?
YANG:

Maybe, but one does not know how to do it and be physically

meaningful.
KLEINERT:

From a purely group theoretical approach, such operators are

easily constructed. Your Yqlm for fixed q form a representation of

69

MONOPOLES AND FIBER BUNDLES

0(3.1).

The quantum number q seems to be analogous to the helicity

of a rotating top so that the angular momentum, of course, has to be
greater than the intrinsic helicity of such a top. The quantum number q seems to fill precisely the same role here. For different
values of q, the Yqlm seem to form representations of a larger group,
for example 0(4.1), in which you can construct these raising and
lowering operators.
YANG:

You may be right. The electron-monopole problem is certainly

more complex than that described by two coordinates e, for fixed r,
since the electromagnetic field has a momentum density distribution
and therefore has inertia. A complete clarification of this type of
problem is yet to be made.
GARCIA:

Could you clarify the motivation for introducing the extra term
in the expression for the angular momentum commutation relations in
the presence of a singular potential?
YANG:

It was first shown by Saha that if you have a monopole of

strength g and an electron charge e, then everywhere in space you have
crossed E and H fields. You will have a non-zero Poynting vector

The system has cylindrical symmetry about the line joining

tron and the monopole. Hence the Poynting vector, and the
momentum, points in an azimuthal direction with respect to
of symmetry, giving rise to an angular momentum -q~/r from
tron to the monopole.

the elecfield
the axis
the elec-

On the other hand, Fierz, in 1944, observed that the term -qr/r
is needed to obtain the correct commutation relations for the total
angular momentum. Both points of view should be equivalent in a
field theory of electrons and monopoles in interaction with the electromagnetic field. However, such a field theory is still to be worked
out.

C.N.YANG

70

There is, however, a problem in the paper of Fierz owing to

boundary conditions which were not definable due to string singularities. In our section approach, we have "opened up" the singularities instead of bundling them up into a string. In this way,
we are able to handle them rigorously.

JONES:
In the Dirac approach, the total flux leaving a sphere surrounding a monopole is zero because all the flux comes back in along the
string, even though you cannot see where the string is. In your
approach, is the total flux leaving the charge equal to 4ng?
YANG:

Yes. There is, in our scheme, no "return flux" at all. In

Dirac's scheme, the vector potential has zero curl. This inevitably
leads to a divergenceless magnetic field. In other words, there
must be a return flux, as you said. This return flux in the past
was supposed to be cancelled by the "Dirac veto", and led to much
confusion. In our approach, this confusion is removed from the
beginning by not considering a string of "return" flux at all.

LIPKIN:
To describe one monopole, you have divided space into two regions. Would you need more regions to describe systems with several
monopoles?
YANG:

sign.

One needs more regions when one has more monopoles of whatever

LIPKIN:
Is there a simple relation between the number of regions and
the wrapping number discussed in Wick's lectures?

WICK:
You need only two regions for one monopole, no matter what the
value of q = ge is. Therefore, I do not think there is a connection
with my wrapping number. However, there may be a somewhat different
way of formulating the question, to that it makes sense.

MONOPOLES AND FIBER BUNDLES

71

WIGNER:

If one writes down the ordinary Dirac equation for two oppositely
charged monopoles, the Hamiltonian is not self-adjoint -- it shows a
mathematical pathology as discussed by von Neumann. Does this difficulty manifest itself also in quantum field theory, and if so, how?
YANG:

This same difficulty exists. However, the difficulty does not

imply that quantum field theory equations for magnetic monopoles are
necessarily meaningless. It merely means that one cannot interpret
the "static approximation". I believe the situation is similar to
the case for ordinary QED in the thirties, or even now: the equations for QED are probably correct, but have to be understood correctly. With monopoles included, the renormalization theory, particularly in view of the large value of g, becmes a completely open
question.
WIGNER:

One thing that I am not clear about is whether a theory with a

coupling constant of the order of g ~ 137 can be useful because of
the difficulties of convergence.
YANG:

I agree, but there is one thing I would like to add. It is also

not known whether QED forms a convergent series in a; in fact, there
are many guesses that it does not converge in a.
WIGNER:

That is true, but QED is still useful; whether it would still

be useful with a equal to 137 rather than 1/137 is less clear.
YANG:

Yes, I agree completely;

it is 137 2 less clear!

WIGNER:

What are the experimental indications for the existence of the

monopole?
YANG:

Last summer, 1975, Price and collaborators published a paper reporting on experimental evidence for a magnetic monopole. The report
generated many discussions. It is generally regarded now as not
conclusive.

c. N. YANG

72
SOHNIUS:

Is the particular form of the Fierz term gauge invariant?

YANG:

Yes, any ~ describing a magnetic monopole would lead to the

same -qr/r in L.
Let me raise the following question: Can Dirac's monopole idea
be generalized to the case of SU(2) gauge fields?
The answer to this question depends on its precise interpretation. I believe the most important characteristics of Dirac's monopole is that it is the source of magnetic flux. Thus one should look
for a generalization of the following type. What SU(2) gauge field
has a point source of some "generalized flux"? It turns out that in
five-dimensional Euclidean flat space or on a four-dimensional
Euclidean sphere, there are such generalizations, but not in lower
space dimensions. Furthermore, in these two cases, you get only
two monopoles nnd not an infinite number as Dirac's quantization
rule allows.
MARCIANO:

Is your Yang-Mills SU(2) magnetic monopole a point source magnetic monopole with infinite energy? If so, then why can this not
exist in 3 + I dimensions?
YANG:

It is a point magnetic monopole with infinite energy. Foy

lower dimensions, such as in four-dimensional flat space, generalization do not exist which have a point source of a "generalized flux".
DISCUSSION 2
PARSONS:

You defined a quantity a~ by A~ - A~. Presumably, this quantity

is only defined in the intersection of Ra and Rb Is that correct?
YANG:

Yes. You can try to continue

singularities in these regions.

into Ra or Rb' but it will have

73

MONOPOLES AND FIBER BUNDLES

PARSONS:

Presumably, then, that is why you can allow the two conditions
to apply of having a zero four-dimensional curl of a and a non-zero
loop integral without inconsistency?
YANG:

Yes, that is correct. The fact that we have a zero curl means
that one can deform the four-dimensional loop in any way and not
alter the value of the integral provided it always remains in Rab'
Because of this, the loop is not contractable to zero and the integral is equal to a space-time independent numerical constant. This
means that we have a non-trivial fibre bundle associated with the
monopole.
JANCEWICZ:

I did not understand completely your picture:

How many dimensions does the cone have?

loop surrounding it?

How many dimensions has the

YANG:

The surface of the cone is a three-dimensional manifold and the

loop is a one-dimensional loop. If we consider the cone at fixed
time~ the surface is two dimensional.
As the time is varied, the
apex of the cone will move in a time-like direction so the surface
of the cone will trace out a three-dimensional surface in fourdimensional space-time. The loop is just a one-dimensional loop.
One can draw it at one time and then one is able to distort it in
four-dimensional space-time so that different parts of the loop are
at different times. Nevertheless, one cannot shrink it to a point.
MARCIANO:

Can your formalism be trivially extended to include dyons?

YANG:

Probably, but we have not done this. I would not say it is

trivial. Nothing is quite trivial since the joint idea is new to us.

74

C. N. YANG

MARCIANO:

Can you comment on the soliton type of monopole, such as the

t'Hooft and Polyakov type of soliton?
YANG:

If you have a pure non-Abelian gauge field, not interacting

with anything else; I would think that if you had a solution in this,
it would be most interesting, but such a solution has not been found
without singularities. Wu and I found three solutions about nine
years ago, but all of them are singular at the origin. The t'Hooft
and Polyakov solution consisted in coupling the gauge field with an
isovector scalar field. In that case, they found that you can have
an analytic solution.
Let me make a further comment about solutions with pure gauge
fields not interacting with anything else. I have said that you
cannot have an analytic solution without singularities. That is not
exactly correct. No-one has found analytic solutions if one confines
oneself to compact gauge groups. But, if one allows oneself the luxury
of non-compact gauge groups, one can write down analytic solutions with
no singularities. Wu and myself have found such a static solution
for a gauge field of the non-compact Lorentz group which has just
been published in Physical Review. For such a non-compact group, one
can find a simple solution with the following properties: i) it is
static; ii) it is everywhere analytic; and iii) the field goes to
zero at infinity. We believe that there is a theorem, which we have
not been able to prove, which says that unless the group is noncompact, one cannot have such a solution. This theorem was supposed
to have been proved in last year's lecture notes by Coleman. But
the proof breaks down, because it is based on a fallacious scaling
argument.
VINCIARELLI:

There exist examples of bona fide monopole-solitons within a

"pure" gauge theory based on compact gauge groups, such as SU(2).
In this case, the SU(2) gauge invariance must be broken down to U(l)
by the interaction of explicit mass terms for two of the three gauge
fields. The U(l) monopole which one obtains has a finite energy in
spite of the presence of a physical pole at the location of the
point-like magnetic charge.
YANG:

This solution seems to be in a different ball park, as it possesses a singularity at the origin. I feel that a solution without such a
singularity would be more interesting, because of the question of

MONOPOLES AND FIBER BUNDLES

75

convergence. A convergent solution may indeed point to the direction in which physics must move, but, unfortunately, this seems to
be impossible at present.
BUCCELLA:

Up to now, there is no convincing evidence for magnetic charges.

I wonder whether one can associate the magnetic charge with the unobserved colour degrees of freedom in the same way as electric charge
is associated with the observed flavour degrees of freedom?
YANG:

This is an interesting question.

I have not thought about it.

PAULI:

What are SU(2), or SU(3) or SU(2) x U(l), etc., fibre bundles?

YANG:

SU(2) bundles can be considered as generalizations of U(l)

bundles. It is a new topic for physicists and has not yet been very
much studied.
You divide space-time into regions and try to match the SU(2)
gauge fields in one region with those in another region in the overlap
of the two regions by a non-Abelian gauge transformation. If the
gauge transformation cannot be deformed to the identity transformation
then you have a non-trivial fibre bundle. The mathematical generalization from U(l) has been greatly studied by mathematicians. The only
known non-trivial generalizations in the physics literature are due
to Polyakov, reported in Physics Letters~ October 1975, and the cases
discussed by me earlier. In all these cases, you need a more complicated space than four-dimensional space-time in order to accommodate
the complicated matching between the regions.
GARCIA:

You made the remark that the condition A~~ = 0 is unnecessary

in the theory of monopoles and electrons. In the quantization of
ordinary electrodynamics, the subsidiary condition has an important
role in restricting the gauge freedom. What is the situation in the
quantized monopole theory?
YANG:

The situation is quite similar to that in the quantization of

electrodynamics. We think that one can quantize the theory if one
wishes to make a specific choice of gauge.

c. N. YANG

76

PHAM QUANG:

You talk about a theory of electrons, positrons, and monopoles.

What should be the spin of the monopole? Is it a fermion or boson?
YANG:

That theory is not yet complete. Tu, Wu, and I are working on
it. We believe the spin could be 0 or % or higher. For integral
spin, it would be a boson; for spin-%, a fermion, as usual.

ROSSI:
You spoke of the monopole-electron system as having some properties analogous to those of a spinning top. Besides that, the equations you have written to assure the integrability of the action in
the overlap region recall some general relativistic expressions which
appear in the study of non-irrotational manifolds, and, in that context, are referred to the gravitational potentials.
Are there any connections between these properties of the gauge
fields and the properties of the gravitational potential? In other
words, can we look at the gravitational field as a gauge field, and
vice-versa? Can we look at gauge properties as linked to geometrical
properties of space?
YANG:

It seems very likely that gravitation is, in some sense, a gauge

field, but how it is a gauge field is not yet clear. I discussed
this topic in Physical Review Letters 33, 445, 1974. The discussion
is, however, still incomplete.
--

DISCUSSION 3 (Scientific Secretaries:

A. Garcia and B. Yoon)

TELEGDI:
I think that it is not generally known that the work of Gibbs
was essentially paralleled in the first two or three papers of
Einstein's, where he reinvented the canonical and grand canonical
ensembles.
A second remark: if I remember correctly, Einstein also knew
about De Broglie's paper at the time that he received Bose's paper.
Once you had the light waves and you could do particle counting with
the photons, the jump to the real gas with the matter waves was maybe
made a bit easier by this idea of De Broglie's. I think he should
get some credit.

MONOPOLES AND FIBER BUNDLES

77

YANG:
About the first remark, I have only recently learned that
Einstein had duplicated a lot of what Gibbs wrote in 1901. As to
the second point, certainly De Broglie's contribution was extremely
important, but everything must have been in a fog at the time. It
required great insight and boldness, so characteristic of Einstein's
thinking, to marry two things together and reach the right conclusion.

WIGNER:
You mentioned the great accomplishment of Fermi's to have provided Fermi statistics. I was in Berlin at that time and I know that
Bose's article was an enormous surprise and Einstein was delighted
with it. The idea that particles behaved in the opposite was was, to
Einstein, quite obvious. It was interesting to see that this explained
what is called Duder' s paradox. But.; that there is an opposite behaviour of particles due to Pauli's exclusion principle was, in Berlin,
taken for granted. This shows that different groups of physicists take
different things for granted and are astonished by different things.

YANG:
I think this is a very interesting point and I would like to pursue it a bit further. A year and a half ago, when I was writing this
article and was particularly thinking about Fermi-Dirac statistics, I
had a conversation with Professor Wigner at Rockefeller University.
Then I remarked to Professor Wigner that, perhaps, Fermi-Dirac statistics is the greatest contribution of Fermi's in physics proper, with
which he immediately disagreed. I was completely taken by surprise,
specially when he asserted that the greatest contribution of Fermi's
was the theory of 8-decay. I was surprised because Fermi-Dirac statistics is, in many senses, essentially a final story; 8-decay theory,
although it is a great piece of work which produced a profound impact,
is not a final theory. In retrospect, as a final judgement of its position in the history of physics, I think we must put Fermi-Dirac statistics first. I would like now to ask Professor Wigner whether he agrees
with this assessment.

WIGNER:
You will forgive me, but I am afraid I do not. At that time we
all believed, perhaps foolishly, that there were electrons in the nucleus. I calculated myself the kinetic energy of the electrons and it
did turn out to be large but not terribly so. Then came the discovery
of nitrogen. The nitrogen nuclei obeyed Bose statistics, and this was
terrible; the nitrogen nucleus clearly consisted of 14 protons and
seven electrons and we did not know what to do. Fermi's paper had the
idea of the creation of particles, which was at that time a great surprise, even if it is now quite natural because we all create particles
with a+ and similar operators.

c.

78

N. YANG

YANG:
What you say illuminates the situation to me but there are other
aspects of it that are opaque to us at this time because we have seen
too much of what happened later. The creation and a-nihilation operators were familiar to you in the late twenties,and, furthermore, Dirac
derived the correspondance principle of atomic transitions using
quantum field theory. So creation and annihilation operators were
already used.

WIGNER:
We had operators for the creation of particles but the fact that
particles were really created and that neutrinos were created at the
same time -- well, we just did not think of this.

YANG:
Is it the case, therefore, that by the late twenties, it was recognized that photons could be created and annihilated, but a material
particle could not be created and annihilated, despite your paper with
Jordan?

WIGNER:
Yes, that is the way our thoughts went.

TELEGDI:
If you look in Pauli's Handbuch article, it seems that the a,a+
formalism was considered a luxury, a trick, because then one could, by
cheating, obtain most of the results on photon emission, it was very
close to classical physics. So, it appeared that these operators were
weird mathematical techniques one did not obviously need. When he
heard of Pauli's idea of the neutrino, Fermi recognized that there
was a wonderful application to try the usefulness of the formalism.

DISCUSSION 4
WILKIE:
I was always told that there is an ambiguity in the definition of
form factors. As you use one specific choice of form factors, Sachs,
as a measure of the matter distribution, can you give the physical
reasons for making this choice?

MONOPOLES AND FIBER BUNDLES

79

YANG:

In the limit that the mass + 00, but the magnetic moment remains
finite, e.g. for a hydrogen atom with proton mass = 00 and no proton
spins, the S~chs form factors, GE and GM, have definite geometrical
meanings. If the mass is < 00, these geometrical meanings become fuzzy
because of the recoil velocity, not recoil momentum, which is not
bothersome.

LEADER:
Although I do not believe that anyone really predicted the growth
of atot, it should be noted that on the basis of cosmic ray multiplicities, Heisenberg suggested in a published paper that atot would grow
like (log S)2 long before the work of Cheng and Wu.
YANG:

I did not mention Heisenber's work because, in my op~n~on, his

guess was based on the physics so far back in time that it really has
no relationship with our present picture of high-energy collisions.

LEADER:
You have given a nice heuristic derivation of the possibility of
having both the density p, and the current j, playa role in hadronhadron scattering. However, as early as 1969 Martinis et al., suggested that the natural generalization of p P should be j~j~, and
they examined the consequences in detail. Also, in the past few years,
the Marseilles group of Bourrely, Sofer and Wray have numerically investigated models based on an eikonal proportional to j~j~ and have
studied the resulting spin effects.
YANG:

I .am unfamilair with Martinis' work, I recently talked with

Bourrely and Sofer. Their spin effects are quite different from ours.

LEADER:
I think that the absence of a second dip in the recent ISR data
on elastic pp scattering at large It I will effectively eliminate models
based on taking Phadronic proportional to Pem' All such models predict
a sharp second dip inside the region covered by this experiment.
YANG:

I want to make two remarks. First, the ISR result in the

t ~ 7 (GeV/c)2 region is statistically not weighty. It certainly
has much fewer counts than the preliminary FNAL data.

80

C. N. YANG

Secondly, I would strongly disagree with your statement that if

the second dip is not found it would "eliminate" models based on
. t'"
Phadron ~ Pcharge "E I'ImIna
e IS a s t rong wor d . If a parame t er 1ess
fit works for a region of 10 10 and runs into trouble at smaller values,
I would hesitate to immediately eliminate it. That is based on simple
physical sense.
LEADER:

You rely on the growth of Gtot to produce spin effects. This would
imply that in the large region, 30 GeV/c ~ PT ~ 100 GeV/c, where Gtot
is flat, that no spin effects should be seen. Is this a prediction of
your picture? Moreover, if you insist on pa growth, then you certainly
cannot neglect the real part of the amplitude.
YANG:

I will comment on the second question first. Increase of GT

carries implications about the real part of the amplitude. When one
aims for an improvement of the theory, one would put it in.
About the first question, it is clear that at low energies there
are other effects that have spin dependence. Wu and I are working on
the problem of how to make a "two-component" approach to the spin
effects in the intermediate region. But, we do not know of a unique
approach. Without the "low-energy" effects, flat GT would, in our
picture, lead to an R parameter = 0, because there would then be no
right-left difference of blackness.
LEADER:

You justify your approach by considering scattering from a nucleus.

However, your method breaks down for the scattering of two nuclei with
high spin states. In your picture, you rely on the fact that sy, the
spin component perpendicular to the scattering plane, cannot change.
This will not be true for higher spins. Worse yet, for each value of
sy there will be many different spin states possible for the nuclei.
Thus your eikonalization procedure will fail. It therefore disturbs
me that you appeal to the nuclear picture to support your theory.
YANG:

It does not disturb me at all. For higher spin targets and beam,
the analysis is necessarily complicated. Brown and his collaborators
are working on this problem.
Let me emphasize thet we do not "justify" anything. The concept
of hadronic matter current is clear to us, although not precisely defined. The nuclear discussion is (a) for its own interest and (b) to
convince people of the usefulness of the hadronic matter current idea.

81

MONOPOLES AND FIBER BUNDLES

GOURDIN:
I suppose t~at in your approach, the matter density P, and the
current density J, are members of a Lorentz four-vector. However,
you identify these quantities which are frame dependent, with the
form factors, which are, by definition, Lorentz invariant scalars.
Therefore, such an identification has to be done in a particular frame
of reference. What happens in other frames? For instance, what are
the corrections in going from a proton at rest to a fast-moving proton?
YANG:

In my answer to Doctor Wilkie's question, I already touched upon

this point. I now repeat: if the recoil velocity is I 0, the physical
picture becomes fuzzy. Thus, for t/~ > 1, there are troubles with recoil velocity which we do not now know how to deal with.
BERLAD:

In 1968, working on my M.Sc., I tried to follow up on

Professor Yang's idea of the "coherent droplet model" by giving an
invariant formulation of the absorption phase shift. One starts from
two time-independent matter densities, no spin included, in the two
respective rest frames of the colliding particles. These densities are
then boosted to a common frame, and, hence, two currents are obtained.
The absorp.tive phase shift is then constructed by contracting the two
vectors j~1)j~2). As a result, one gets back Professor Yang's results, multiplied by a simple kinematical factor, i.e.
PI P2

V
PI P2
relative

where

i3

= .!c

This factor, of course, goes to 1 as the energy grows; and

Professor Yang's formula is recovered. This paper was rejected by
IZ Nuovo Cimento because it was not interesting enough. (This Zast

remark has been added by speaiaZ request of Professor Ziahiahi.)

WEILL:

The "geometrical" picture that you gave for hadron-hadron scattering is a very accurate description of proton-proton scattering.
How does it depend on the quark content of the projectile? For example,
if one scatters an L or ~ off a proton, does your model predict a new
value for the slope of the differential cross-section?

c. N. YANG

82

YANG:
We cannot make predictions of that type until we are given information about hadronic matter form factors of L or ~.

ZICHICHI:
The pion and kaon form factors have been measured in the timelike region. How does your calculation fit the time-like data?

YANG:
Chen, ~n a paper ~n The Physical Review~ 1975~ computed with the
geometrical picture from np and Kp scattering, the hadronic form
factors of n and K. No extrapolation to the time-like region has
been attempted.

TELEGDI:
In your talk, you have shown a graph for R as a function of t
for different scattering processes; why was K-p very different from
K+p?

YANG:
Phenomenologically, K+p total cross-section increases faster than
K-p at FNAL energies.

TING:
Can you comment on what happens if you use a photon beam?

YANG:
We have not considered it.

LIPKIN:
The parameter a used is really d/dp (log atot). Is this what you
actually use, rather than a power fit to the data over an extended region? If you are using both a(K-p) and a(K+p), you should use a(pp)
as well as a(pp). If the difference between particle and antiparticle
scattering is a low-energy phenomenon outside the scope of your analysis, you should disregard both a(K-p) and a(pp). If you think that
your model works also for the mechanism responsible for the particleparticle and particle-antiparticle difference, you should consider all
cross-sections. There is no qualitative difference between a(pp)-a(pp)
and a(K-p)-a(K+p). The reason why a(pp) is still decreasing at
200 GeV while a(K-p) is increasing is irrelevant; it is because a(K p)
increases faster than a(pp).

MONOPOLES AND FIBER BUNDLES

83

YANG:
Wu and I are working on some two-component picture.

BUCCELLA:
The hypothesis that the electric charge density is proportional
to the matter density seems to be disproved by the behaviour of
F~(x)/F~(x) near x = 1. The behaviour implies, in the framework of the
quark-parton model, that the Po and the no momentum distribution in
the proton are different.

YANG:
He stated in our paper that for the proportionality question we
are confused about the neutron form factors.

SAKURAI:
I understand that your hadronic density distributions are isoscalar, i.e., the same for the proton and the neutron; yet you identify your hadronic density with the proton charge density, which is
the sum of the isovector density and the isoscalar density. Perhaps
you would say that we should not use the neutron density because there
is a delicate cancellation in the neutron case. But, in the neutron
magnetic density, there is no delicat 7 cancellation! ~n) is large
and its sign is opposite to that of G~pJ. It appears totally arbitrary
to me that you identify the hadronic magnetic density with the proton
magnetic density. If you identify it with the neutron density, your
prediction for R changes sign. If you identify it with the isoscalar
magnetic density, you will presumably predict a very small value for
R.

YANG:
Let me first emphasize that we do not have the same confidence
in the proportionality hypothesis as we have in the main points of
our paper. We therefore do not predict any R value. We try to estimate it with the proportionality hypothesis. Your discussion is
logically correct; however, we venture to fix our attention on the
protons which, we believe, are more homogenized.

GOURDIN:
Do you use proton form factors which are a superposition of I
and I = I quantities for the deuteron target which has I = O? In
other words, have you solved the neutron problem?

C. N. YANG

84
YANG:

VIe are confused by the neutron problem.

KLEINERT:
Does anything go wrong if you prescribe using isoscalar form
factors?
YANG:

We have more information with the proton and good agreement

with the V-independent part of the theory with the proton data. Thus,
we stay with the proton. We are confused about the neutron.

THREE LECTURES ON SOLITONS *

G. C. Wick
Columbia University
Lecture I
It has*~een known for a long time, that certain non-linear twodimensional
wave equations possess "soliton" solutions, describing
solitary waves, that travel without changing shape or size. The speed
of the waves, however, may depend on their size. For illustration
purposes, I shall use the equation:
a 2ct>/at2 -

a2ct>/ax2

+ Sin ct>

=0

(S. G.)

which has a long history 1) Clearly it has stable static solutions

ct> = const (where const = 0, 21T, 41T, ) which will be regarded
as fully equivalent states of rest of the system. This is legitimate if
we view ct> as an angle, or as the phase of a complex z = e1ct>. Then
the position ct> =0 is identical with ct> = 21Tn. It is now very easy
G. equation of the form
to find solutions of the

s.

ct>(x,t)

= f(x-vt)

(1. 1)

* I apologize if, for the sake of clarity, one finds in these lectures
some repetition of matters already covered in S. Coleman's "classical
lumps and their quantum descendants," lectures held at this school in
1975. I have kept overlap with these lectures (quoted as Coleman In
the following) to a minimum, I hope, by: a) a different emphasis on
subjects b) using recent material and c) referring to Coleman for more
detail (especially in my second lecture) whenever this was possible without interrupting the flow of the main argument.
** i.e., in one spoce- and one time-dimensi on.

Research supported in part by U. S. Energy Research & Development Adm.

85

G.C. WICK

86

In fact, assume
f~)

=4

Iv I

<

2 -~
1, let y = (l- v )

and

Y~ +const J
Arc tan [e

Then (l.1) is a solution.

x, see fig. 1.

(1.2)

For a given time

t,

let us plot a<l>/ax vs.

a<l>
X

x-

xA slow soliton

A fast soliton
Fig. 1

We obtain a peak, that will move with constant velocity v without

changing shape, as t varies. The positive peak corresponds to the
+ sign in Eq. (1.2); the - sign gives a negative peak C'antisoliton")
In either case the peak indicates where the energy of the disturbance
is located. The width of the peak tends to a finite limit as v - 0
2)
(slow solitons); the peak is narrower and higher for faster moving solitons
Another venerable example is the Korteweg de Vries equation3 ) of
classical hydrodynamics. It describes waves in a canal; its solitons
move all in the same direction, with speeds depending on the size of
the soliton: bigger solitons move faster.
Since these wave equations are non-linear the superposition
principle does not apply in general. We can, of course, obtain by superposition the asymptotic form (as t - 00) of the solution corresponding to
the collision between two solitons. As soon as the two soliton peaks
overlap noticeably, however, the solution will differ from the sum of the
two one-soliton solutions. The surprising fact is that, nevertheless, a
closed analytic expression for the solution at all times can be given.
(This result extends further to n-soliton collisions~) We demonstrate this
in the case of the (S. G.) equation.
The calculation will also disclose a rather deep reason for the
result. It shows that, given a solution <1'(x, t) of the second order
equation (S. G.), we can get another solution ql(X, t) by solving a system

87

THREE LECTURES ON SOLITONS

of two first order equations:

aq>
ax - -at

aq,s _
aljJ _

aq>
x

aT--a
where

+ A sin ~ + ~ sin
L.

1\

tJI -

<I>

'\'
tJI+q>
1.
1\ sm
- 2 - + Asm

A is a constant.

a ax
atJI
We can calculaterr

(1. 3)
from t he first

equation, and:-. :tJI

from the second equation and the result must,
of course, be e~ual. t This is in fact the integrability condition of the
system (1.3), when tJI is the unknown. It turns out (as the reader can
easily verify) that the condition is satisfied precisely if q> is a solution
of (S.G.).'
In this case, tJI exists. But then notice that Eq. s (3) rema~
unchanged if we interchange tJI and q> and change A to - A ;
then the calculation we have just sketched also shows that tJI is a
solution of (S. G.) .
The gist of the argument is then the following:
a transformation 5)

Eq. s (1.3) "define"

of a solution q> of (S. G.) into a new solution tJI. Incidentally

BA belongs to a general type known in rrc thematics as "Backlund
transformations. "
Integrating (1.3), when <1> is given, is much easier than hunting
for solutions of the original 2 nd order equation. The process can be
repeated generating always new solutions. Setting q> = 0 in (1.3),
to begin with, the tJI one gets is the one-soliton solution, Eq. s (1. 1)
(1. 2). In the next step, using this tJI as the next q> , one generates
a solution which describes a collision between two solitons, . . and
so on.
There is a remarkable "commutability theorem" of Bianchi 6 ) which
states that the result of two successive Becklund transformations is
independent of the order BA BA <l> = BA BA <1>
In proving this
2
1
1 2
theorem, Bianchi obtains as a by-product a remarkable result. Let
tJl 1 = BA tl' , respectively tJl 2 = BA <I> , be the result of the first

88

G. C. WICK

Backlund transformation (for A = A1 or A2 ) . Bianchi proves by

direct verification (it is a little tricky) tliat the function tIJ deffned
implicitly by the equation

Tan 4 (tIJ-tp)
where

=k

Tan

4 (tlJ 1 -

t1J2)

is a suitable antisymmetric function of A1

(1.4)
and

A2 ,

satisfies the Backlund conditions (1.3) with tp = tp 1 ' A = A2 (or

alternatively = ct>2 ' and A = A1 ) on the right-hand side. This,
of course, proves the commutabil ity property, but we are more interested
in another consequence 7) of Eq. (1.4): it gives IJI in finite form; in
other words after the first step, no further integrations of the sy.;tem(1.3)
are required:
This method is to"berecommended, if one wishes to
gam some first hand knowledge of solitons, without delving too deeply
into the subject. There are more powerful methods, but their study
requires much greater effort.
The explicit form of the ~rsoliton-(or of the soliton-antisoliton-)
solution is given in many papers and it is easy to verify the following
remarkable property: after coalescing for a while, the two solitons
reemerge and separate with their shapes and sizes (and speeds) completely
unchanged, almost as if they had gone through one-another without
interacting. This is very remarkable since it is not required by any
of the obvious conservation theorems ). Energy, for instance, could
be transferred to other degrees of freedom (inelastic collision) but this
does not happen. It is this result, that has attracked the most attentiol)
and since several non-linear equations exhibit similar phenomena, it has
become customary in the Applied Mathematics literature to include
these peculiar properties in the definition of solitons.
The particle-physicist may be interested in solitons, insofar as
they constitute stable extended objects, that could conceivably serve
as interesting models for certain particles. In the traditiorYJ I scheme
of "second quantization" a wave theory generates particle-liKe states
only after quantization. The "elementary" particles, that is, appear as
"quanta" of the linearized form of the wave equation *; in subsequent
approximations, other particle-like states may appear as bound states
containing two or more elementary particies. Solitons, however,
already exist at the classical level, a circumstance that opens up new
possibilities for the description of particles in a non-linear field theory 10).

* In particular their mass (if it is not zero) is related to the parameters

of the wave equation via Planck's constant in the familiar way.

89

THREE LECTURES ON SOLITONS

It is amusing, that in a certain sense, after more than 60 years

speculation about the nature of elementary parti cles has
back
full circle to ideas that were first explored by Gustav Mie
Needless
to say, the setting is quite different, and the classical extended objects
are viewed only as a possible first approximation to states that, we
hope, will also exist in the quantized form of the theory.

c9rrr

I must now point out, however, that it is extremely unlikely that

any non-linear wave equation in ordinary 3-dimensional space will
possess solitons with the very special properties we have mentioned above.
We must therefore either enlarge the definition, or eventually drop
the name soliton altogether, as some authors propose. I shall go on
using soliton in these lectures.
I wish to describe now a particularly simple example 12) of a wave
equation in ordinary 3-dimensional space with "soliton" solutions in the
general sense mentioned above. In setting up 3)ch an equation, one
has to by-pass a well-known scaling argument 1 which shows that a
three dimensional soliton can only exist in a theory endowed with some
additional features, such as a conservation law. The reader may wish to
compare the present example to earlier models of "extended objects",
which contain fermions. In those models it is the conservation law of
the "number of fermions" that helps to stabilize the system. Here, for
simplicity, we stick to bosons.
The conservation law is obtained by
assuming that the Lagrangean contains, besides a real scalar field A,
also a complex field t[l.
The terms containing t[l are bilinear in t[l
and its complex conjugate t[l*. Specifically the Lagrangian density is:

~au

L = -

AaflA - aut[l* aflt[l- V(A,t[l,t[l*)

The first two terms have the familiar Klein-Gordon form.

term:
1
2 2
2 2
V(A,t[l,q,*) = 8" (A -1) + k A t[l*t[l
is chosen so as to have an absolute minimum at '" = 0,
we shall call the vacuum configuration 14). Any sol ution
equations) with a definite energy must have fields tending
v.Jcuum values as
--+ co (i. e. tjl --+ 0 and A--+ + 1 ,

1';1

(1.5)
The potential
(1. 6)
A2= 1. This
(of the field
to their
say).

The conservation law arises, of course, from invariance with respect

to the (global) gauge transformation t[l --+ e ia ljJ where a is an
arbitrary real constant. The conserved "charge" is

-.-

Q =

* 0ljJ

0ljJ *

f -{ t[l at - at

t[l} d x

(1. 7)

90

G. C. WICK

A static Solution (A and 1/1 time-independent) will have 0 = O.

Our solitons, however, will be solutions of minimum energy under the
constraint, that Q has a given non zero value. One can show, ~
familiar arguments (we shaTI not do it) that the usual "separation" of
the time dependence can be assumed for these solutions, and thC1t .in
fact A may be assumed time independent, while 1/1(-;, t};:: ..!.. e -,wtB(;).
fJ.
Furthermore we may assume B real. Now A(;) and B(X) obey a
set of two coupled partial differential equations, which are easily
obtained from the Lagrangian (1. 5). For the sake of simplicity we
shall consider only spherically symmetric solutions, i.e. we shall write
A = A(r}, B = B(r}, r =
The equations are then:

1;1.

drL

2
r

dA _ k2 B2 A _
dr

;B + 2 dB

rT

r2

+~-k

~ (i _1) A = 0
.

22J
A

B=O

(1.8a)

(1. 8b)

For such solutions, furthermore, we have:

(1. 9)
= 41TW 1o00 B2 r2 dr
We notice that these equations allow the transformation w -+ - W, 0
From now on we shall assume that w, and hence 0, is positive.

-+ -

o.

It is easy to see, that in a finite energy solution A2 -+ 1 and

B -+ 0 at infinity, so that also 0 is finite. Now a qualitative insight
into the nature of the solution is obtained if we assume that in some way
the function A(r} is known, and look upon (1 28b) as a radial Schrodinger
equati~ ~n which B is the wave function, w
is t~e energy parameter,
a~d k A (r) thefotential.
Since the potential -+ k at 00, the energy
w must be < k
Only in this case can B(r} tend to zero
(exponentially.') as r -+ 00. Clearly the condition: "0 is finite" is
the usual square-integrability condition for a bound state wave function.
No bound state will exist, however, unless there is qn inner region
where the val~e of the potential is lower than at 00. Thus we may
imagine k 2A (r) to be, roughly, aSS'l<'e'f'"ched in Fig.2.

Fig.2

2
k

ji ~w22- ~

r""';

THREE LECTURES ON SOLITONS

91

2
Then, of course, w must in addition, be one of the discrete eigenvalues corresponding to this potential. The wave function B(r) is also
sketched; it is large inside the potential and has a small tail outside.
All this is quite convincing, if we remember that the solution must have
minimum energy, when we look at the potential term in the Lagrangian,
Eq. (l.6). Outside the polrential well, A2 Z 1 and the first term is
negligible; but the second term is also small, because tJI, i. e. B is
small. On the other hand, inside, where A2 is smaller, tJI is allowed
to take larger values. Of course, once we have determined B(r) from
(1.8) for a !=liven A(r) we have hardly solved the real problem.
Let us consider, however, the limiting case when Q , Eq. (1.9)
is very large. This implies that B(r) inside the well is very large. In
this case the second term in the potential (1. 6) will be very large
inside the well unless A is very close to zero. Thus to keep the
energy down as much as possible it seems desirable to have a rather
abrupt transition from an inner region (the "potential well ") where A ~ 0
and B is larger, to an outer region when A Z 1 and B is negligible
One can show that a variational method based on trial functions A (r)
and B(r) consistent with these expectations is a good approximation
when Q is large. The radius R of the well, which is adjusted to
yield minimum fnergy for a given 0, has an optimal value which
varies like 04". This is then the "size" of the soliton; its energy is
.
I to 0 43
proportlona
In order to decide whether these solitons are really stable, we mL6t
examine whether their energy cannot be transferred to other degrees
of freedom. The space surrounding a soliton is, as we know, in the
normal vacuum configuration (tjJ = 0 and A = + 1, say). Could some of
the energy and charge of the soliton be transferred to this region in
the form of weak oscillations of the fields A and tjJ about the
vacuum val ues? These osci lIations are easy to study; they are of two
types. There are "pure A - waves" (with tjJ = 0 ; they carry no charge
and are of no interest in this connection) and there are tJI waves
which the authors call "free meson" waves. These waves carry charge 12)
and the relation between E and Q is E = k 0 for long wavelen~?
Clearly, if 0 is suficiently large, the energy of a soliton (E ~ 0 Lt)
is less than that of free meson waves carrying the same charge. The
emission of free mesons is energetically impossible. We have considered
long wavelengths only, but for short wavelengths the situation is even more
unfavorable. Please notice that the conservation of the charge 0 plays
an essential role in the stability consideration ~
There are many other interesting points aoo ut this example of threedimensional solitons, which are studied carefully in the original paper. But
time is pressing on, and we must turn to a different subject.

92

G. C. WICK

Lecture II
We shall now devote some attention to the so-called "topological"
solitons. Despite the esoteric name, the stability of these objects is
due to quite ordinary energetic factors, like potential barriers and energy
differences. The name refers only to the possibility of employing a
certain technique, to infer the existence of infinite barriers from qualitative considerations about the continuity of functions or "maps."
All the theories envisaged here contain a gauge-field and possess
"degeneracy of the vacuum." Assuming some familiarity with these
notions as background, I hope to present a reasonably self-contained
discussion of the general ideas involved and of their relation to unfamiliar (to physicists) but classical mathematical theories. In order to
get a good grasp of the subject, however, a physicist should look at
many different examples. I won't have time for this; even less shall I
attempt an exhaustive discussion and classification of all possibilities15).
The general procedure will be to show, that the set of accessible,
finite energy, field configurations of the system decomposes into disconnected components or sectors: there is no continuous path from one sector
to another, without going through configurations of infinite potential
energy.
The following treatment is entirely at the classical level; nevertheless, it may be pertinent to a quantised version of the theory for the
following reason. By field-configuration we mean a set of values of field
components as functions in space at a given time.
Now assumf6 that these components are commuting variables in the
quantised theory ) (the analogue of Lagrange's generalized coordinates
q ,q2' , q
for an ordinary dynamical system); in this case the study
o~ classical cJlnfigurations is also a stl,dy of the functional space, in which
-- -the Schrodinger functional is defined 17).
The following structure is general enough for our purposes. The
theory has a local Lqgrangian containing a gauge-field A (x) and a
1..1
real vector
field <I>(x) with values in some abstract n-dimensional *
space ~. With regard to proper Lorentz transformations the components
of <I> are scalars. The Lagrangian is invariant with respect to a continuous group G, which acts as an 0f~1ogonal group on the vector <I>
we identify G with this group;
Leaving out more general possibilities
simplifying further, we assume G is the group SO(n). (It could be a
subgroup).
The action of an element

of

is denoted :

* i. e. <I> (x) has n rea I components <I> 1(x), ... ; <I> n (x)

93

THREE LECTURES ON SOLITONS

II>t-+ gil>. We assume not only invariance with respect to G as a

global symmetry, but also with respect to the (much larger) group of
gauge transformations, of the form: II>(x)~(x) II> (x) As is wellknown, this requires an accompanying transformation of the gauge field
and a partlcular form of the terms in the Lagrangian containing
derivatives. We shall not discuss all this here.
We must concern
ourselves, however, with the remaining term in the Lagrangian, the
"potential" V (11)), whi ch because of our assumption (G = SO (n must
be a function only of the square of the vector 11>. For consistency, it
must be bounded from below; we may assume its minimum value to be
zero
(subtract a constant if necessary).

Vacuum Values: In an ordinary theory, V(~) has an absolute mmlmum

at II> - 0 and A (x) == 0, II> (x) ::
is a solution of the field equations,
corresponding to cJ: absolute minimum of the total energy. We may call
this solution the (classical) vacuum state. In the theories we are studying, V(II attai~s its minimum value on an infinite set M of values of
the field II> 19. A closer look at the field ~quations shows that, in
yields a solution20 ). II> (x), however, is not
this case also, A (x) =
zero but may be sS equal to any constant value II> belonging to the
set M. Owing to the assumption: G = SO(n) this set is an Sn-1
(an (n-1)-dimensional sphere) of equation:

222
11>1 +11>2 + . +II>n =const(,~O)

(2.1)

Notice that, if G were only a subgroup of SO(n), the set M could

be a submanifold of the sphere (2. 1). Barring "accidental" degeneracy,
the set would consist of all vectors of the form II> = g 11>0 (g any e G)
where 11>0 is some fixed vector. Just to remember this more general
possibility, we shall keep referring to the set of values of 11>, where
V(II attains its absolute minimum, as the manifold M (this also avoids
confusion with other spheres we have to mention).
Boundary Values: We now turn our attention to excited states, with
finrte excitatIon energy. Since the kinetic part ~f ~he energy density
is positive definite, and the "gradient terms" also 1 , the term
J V(II ci3x in the energy must be finite by its own virtue. Hence the
condition: "V(II must tend to zero at inf~ity" or:
/lIn a finite energy configuration II>(x) must tend to a vacuum
value (a value on M when
1 - CD /I At first sight one would
'Ci'SSiime that this limiting vacuum value is a constant, independent of
direction 22 ). The configuration would then differ from one of the zero-

r-x

* i. e. derivatives must appear as "covariant ll derivatives.

G.C. WICK

94

energy vacuum configuarations only in a finite region. V(C!, however,

will also vanish at infinity if we consider (using polar coordinates r,e,4
the more general assumption:
lim 4>(r,8,4 = f~,4
(2.2)
r-oo
as long as f(8,4
lies on M. We can describe f as a "mapping of
the boundary at 00 into M."
A finite energy configuration must also satisfy other conditions, for
instance the space-integral of the "gradient tenns" must also be finite.
Here the gauge field plays a crucial role, but we shall simply refer to
Coleman. The upshot of the discussion is, that everything is all right
as long as f is smooth, or at any rate continuous. This will do for
the following discussion.
Sectors
We must now examine the 9..uestion, whether any two field-configurations, 4> = u(~) and 4> = vex), say, can be connected by a continuous path (or deformation or homotopy). I shall use homotopy since
it is the technical term now in common use. Two configurations u and
v are said to belong to the same homotopy class (we shall write u "" v)
if they can be embedded into a common one-parameter oontinuous
family of configurations. At first sight, there is only one homotopy
class, since the linear interpolation 4>X (x) = (1- X) u(x) + Xv(x)
apparently supplies the desired family. The reason this cbes not work
is the boundary condition, Eq. (2.2). The boundary value of 4>X is
(1 - X) f + As , where f and g are the boundary values of
u and v respectively; this is not a value on M if X ~ 0 or 1,
see Fig. 3.
Fig.3

Owing to this, there can be more than one homotopy class.

To proceed further, it may be convenient to introduce the
follotng (now standard) notation. An arrow connecting two spaces:
X:.
Y denotes a continuous function f(x) defined for all xeX,
with values in the space Y. The formulation of a homotopy problem
always specifies that all functions concerned are from some fixed
"domain" X to a fixed "range" Y. For instanceour configurations
4>(;) are arrows like E-i.. ~, where E is ordinary space and i!?
is the vector space of values of the field 4>, whereas the boundary

95

THREE LECTURES ON SOLITONS

values f(9,cp) of Eq. (2.2) are properly described as functions: B--L.tm,

where B is the boundary at infinity. M, the manifold of vacuum
values, is of course, a subspace of ifl. *
We can easily surmise (and will verify later) that the set of
homotopy classes will depend on the domain and range of the functions
concerned. It also helps to remember that our homotopy problem is of
a special type: it concerns not just functions E~ i, but functions
subject to a restriction on their boundary values, Eq. (2.2). Obviously
any homotopy between configurations (u"" v) implies a homotopy f,.., g
between their boundary values. If we can classify maps ~ M intz3)
homotopy classes (which is, of course, in principle a different problem )
we will have learned something about our configurations. If there is
more than one class of maps ~ M, then there must be also several
classes (or sectors) in the space of finite energy configurations. Consider
then the sector containing the configurations with constant boundary
values, hence in particular those with cp = const. everywhere. That is
to say: the (degenerate) vacuum states of the theory are in this sector.
Now consider any other sector; a minimum energy configuration in it
must be sta~l~' since it cannot make a transition to the sector containing
the vacuum ). It is !-ere that we expect to find (and will find in
concrete examples, see also Coleman) solitons (or maybe in higher
sectors configurations containing several solitons), This is the qualitative
conclusion we were aiming at.
Field Theories in
to.

2-, 3- (or 4-) Dimensional Space.

Homotopy Groups

Let us have a brief look at th~ kind of problem we have been led
First notice that although B = 5 ---.sphysically the most interesting

* For the reader who is not repelled by "commutative diagrams": one

can summarize all these relations by the diagram in Fig.4.

Fig.4

)M

~-cp~~J

Here ordinary space E is turned into a compact space by including

"points at infinity" amongst its elements. (This is natural, since the
existence of the limit in Eq. (2.2) is then just a continuity requirement
on cp). The vertical arrows are just "inclusion maps." The diagram
then summarizes a) Eq. (2.2) b) f(9,cp) e M c) all the continuity
requirements we have stdted.

G. C. WICK

96

case, there is also some interest in two-dimensional configurations, as

in the examples studied by Nielsen and Olesen (see Coleman), with
B = 5' (a circle) or in configurations for a field theory in four-dimensional
Euclidian space, with B = 5'3 (see the instanton in our third lecture).
Thus, generalizing a bit, our problem is to study and classify maps
5m_
M, where m = 1,2,3, ... This is a very famous problem
or rather family of problems, as a glance at the history of mathematics
in our century will show.
In 1895 Poincare, in order to classify closed paths on a connected
manifold M (in other words, maps 5 1--+ M ~) chose to consider only
paths starting from, and ending at, some fixed point on the manifold.
The set of homotopy classes of such paths is denoted TT 1(M) and is
Poincare's famous fundamental group. A group structure is assigned to the
set by a) defining the inverse orapath f as the path obtained by running along f in the opposite direction, b) defining the product fg
of two paths f and g as the closed path obtained by running along
f and g in succession *, and c) showing that those concepts can be
transferred unambiguously to the homotopy classes of paths. Finally
one shows that d) the class of the "constant path ", i. e. the class of
all paths that can be shrunk to a point, plays the rule of unit element.
A simply connected manifold has, of course, a trivial fundamental group
(it consists of just one element).
Now the question for us is: suppose M is the vacuum manifold
of a theory, and we study two-dimensional configurations (B = 5'). Can
we disregard Poincare's fixed point constraint, and assert that TT] (M)
is identical with the set of classes we are seeking? 15)turns out that
the answer is yes, if the fundamental group is abelian
(as is always the
case when M is a group-manifold). In this case, therefore TTl (M)
is in one-to-one corrpondence with the "sectors" when B = 5' (twodimensional space). The simplest example of non-trivial group TT 1 is
given by the case: M itself is a circle. The group TT (5'), as one
shows quite easily, is the infinite cyclic group; equivalently, the
homotopy classes of maps 5' -+ 5' are characterized by an integer
p = 0, 1, 2, ....
Moving on to higher dimensions, a famous topological theorem is
Brower's result, in 1912, that homotopy classes of maps 52....... 52 are
characterized likewise by an integer (sometimes called the "wrapping
number"). This means for instance that in three-dimensional space a

* It is, of course, important here that the endpoint of

with the starting point of g.

concides

THREE LECTURES ON SOLITONS

97

2
gauge theory with G = 50(3) and M = 5
has an infinity of sectors.
AA example of soliton occurring in. such a theory is 't Hooft's monopole
solution, also described independently by Polyakov. Maps 52-; 52 can
be most simply thought of as maps of a sphere into itself. Trivial maps
(with "wrapping number" equal to zero) are homotopic to constant map 5
(mapping the whole sphere into a single point). If the image ot a
map leaves some part of S2 uncovered, it can be shrunk continuously
to a point; the map is then trivial. The identity map, which maps
every point into itself, has wrapping number = + 1 and is not trivial.
The reader will easily check that the boundary value of the Higgs field
in the monopole solution is such a map, and this is why the monopole
is a IItopological soliton. II

Around 1930 H. Hopf obtained some important results on spheres

of higher dimensionality. He showed that the maps Sm ... Sn (m < n)
are always homotopically trivial; and generalized Brower's result by
showing that for m = n the set of homotopy classes is always the same
(it is characterized by a "wrapping number II). Final!! he obtained some
surprising results for values m > n , for instance S--. S2 (see also
3rd Lecture). In order to summarize some results, we adopt the language
of higher homotopy groups, an invention of Witold Hurewicz.
In analogy with Poincare's definition, we consider maps
Sm ........ M (m = 2,3, ) and again we single out a special point on M,
and we only consider maps such that the image of the North Pole (say)
of Sm is the special point. "Trivial" maps are then homotopic to
the IIconstant" map, which maps the whole Sm into the special point.
IT (M) is then defined as the set of homotopy classes of the maps just
~fjned; a group structure can be assigned to 'IT (see Discussion
session) and in this case (m ~ 2) it turns out thar the group is always
abelian. Just as in Poincare's case, the introduction of a group
structure is useful: it puts a powerful tool in our hands. What we
are really interested in is the set of homotopy classes without the
fixed point6fonstraint. In many cases the fixed point makes no
difference2 and we shall proceed on that assumption.
The determination of the groups IT (Sn) (m > n) has been tackled
by many authors, with surprising results, Tome of which are summarized in
Table. It is often a very tough problem, and the general case is
sti II unsolved.

G. C. WICK

98

TABLE
Some Examples of Manifolds
With Given Homotopy Groups IT (M)
p
IT
p

SPHERES

UNITARy(l)
GROUPS

ORTHOGONAL (1)
GROUP5

SU(n)

ALL COMPACT LIE GROUPS(2)

S'

> 1
all (3)

5 (n

z -

> p)
U(n)

SO(2)

U(n) and SU(n)

n >2

SO(n) n = 3
or (4)
n

Z (5)
2

o (n)

(n

>5

and 50 (n)
> 3)

p-1

Remarks
(1)
The S in SU(n) or SO(n) means unimodular (i. e. with determinant

=1).

(2)
(3)
(4)
(5)

This includes S' which is i~entical with U(l) or 50(2) (as group
manifolds), and similarly S, identical with SU(2) .
All means p = 1, 2, ....
The exception, SO(4), has TT3 = Z
Z2 is the (one and only) group of order 2.

z.

Z, of course, is the infinite cyclic group, or the additive group

of integers, and 0 is the trivial group.

99

THREE LECTURES ON SOLITONS

Lectur e III

2
We shall now describ e an examp le 7) of soliton in a pure Yang(non-a belian)
Mills theory, that is a theory in which the only field is a
and is of
gauge- field B. The examp le is known as the "Instan ton"
other fields
interes t in conMection with more elabor ate models, in which
howev er,
n,
28
questio
the
of
part
This
).
are cOlJpled to the gauge- field
it. I
with
ned
concer
be
not
shall
we
is much more compli cated and
I
than
further
you
carry
will
ty
hope, neverth eless, that your curiosi
can go today.
is a
By pretend ing that the usual x4 = ict of relativ ity theory
ean
Euclid
nal
mensio
real variab le we turn Minkowski space into a four-di
2

fi
space WIth a POSItIve- de mIte metric x

r1
= .I 4=

2
x..
I

h
the
Th is is were

standa rd
Instant on lives. Aside from this, the field equatio ns have the
by
ng
Denoti
SU(2).
=
Yang-M ills form, with gauge- group G
T = ~ 0"
O"a (a = 1,2,3) , the customary Pauli matrice s, the genera tors a
a
ns
of the group satisfy commu tation relatio

(3.1)
[ Ta , Tb ] = if a b c Tc
index
The components Ba of the gauge- field four-v ector carry an
es:
a (= 1,2,3) , whicff disapp ears when we turn them into matric
likewi se,
B\l = - i Ba Ta (summation over a is, of course, unders tood).
are matric es: G "" -i GaT
the gauge- field compo nents G
I-N a
I-N
J.IV
by:
d
define
G

I-N

= a

B - a B + y[ B, B ]
VfJ

!-IV

fJV

(3.2)

9
is 2a couplin g consta nt. The Lagrangian matrix densi t/ ):
- ! r (G ) yields the field equatio ns:
I-N
I-N
(3.3)
+y[B ,G ]=0
aG
I-N
fJ
fJ I-N
but also
These equatio ns are satisfie d triviall y not only by B = 0
t of
el~men
/Jan
that
recall
us
Let
O.
=
by any B such that G
fJlJV
in the
the group of gauge transfo rmatio ns is a functio n g(x) with values
matrix
(in
group G. The corresp onding transfo rmatio n of the B - field

where

form) is:

1 -1
(x) B (x}g(x) + - g (x) a g(x) (3.4)
fJ
Y
fJ
fJ
on
rmati
transfo
gauge
the
zero,
is
B-field
so that, if the origina l
field:
genera tes a

B (x} - g

-1

100

G. C. WICK

\x)

.!

B (X) = 9 a 9 (X)
(3. 5)
Il
Y
Il
The G-field, however, remains zero30) as one can also verify
directly from (3.2)(3.5). The energy density of such a solution is, of
course, also zero. One assumes that such a state is physically indistinguishab
e from B == 0, and we shall call it a vacuum state.
Il
The Instanton is a non trivial solution of the field equations, which
tends asymptotically to a vacuum state as the point X tends to infinity.
An especially simple form of this solution is mentioned by the authors 27)
towards the end of their paper. In order to describe it, let us first define a particular gauge-function g(x). Recall that an element u of
SU(2) is of the form

(3.6)

where (u4 , uj is a unit four vector:

2
:2
2
2
u 1 + u2 + u 3 + u 4 = 1

(3.7)

To define g(x) we must define (3.6) as a function of x, i. e. define

four functions u (x) such that (3.7) is identically satisfied. We chose
the simplest Ansaz:

-1

u=x R ;R=(r
Il
Il
~=1
Having defined g(x), we now set:
1
B(x)=-

R2

2g

_1

(3.8)

(3.9)

(x)ag(x)

Il.
thY Rt + At.
till. I
.
Be fore we examine
e In eres Ing opo oglca properties

f h
I . 31 )
t IS so uhon ,
let us verify that it is indeed a solution. We first evaluate (3.9), with
g(x) = u, and u defined by (3.6) (3.9). An elementary calculation
gives:
2
2 _1
yB .(x) = i (R + A)
x CI
(3. 10)
Il
v VIl
where the symbols
CI

CI

v Il = -

vll
CI

are defined by the equations

fJV

CI

ab

e abc

CI

c ;

CI

4a =

CIa

(3. 11)

where a .. b, c are indices running from 1 to 3 32)

One easily derives
commutation laws for the CI symbols, and it is then a simple matter
to calculate:
fJV

101

THREE LECTURES ON SOLITONS

yG

flV

(x)

2 2
2 -2
= 21A
(R +A) (]

flV

(3. 12)

and to verify then the field equations (3.3). Let us notice in passing
!:be duali~y relations
(] 12= - (] 34 ,... etc., s~ that the dual tensor
G
= 2" e A G R obeys the sImple relation G = - G The
tJV
f.1V a t-' a to'
f.1V
f.1V
whole structure of the solution (3.1O) (3.12) is in fact remarkably simple.
Let us now understand why this solution is a "topological" soliton.
Let us first notice that in the limit R ..... CD the B-field of Eq. (3. 9) is
indeed of the form (3.5);
that is to say, it satisfies the general
condition that the physical state of the field must be indistinguishable
from the vacuum on any finite portion of the boundary B at infinity.
(Notice inciden.!2lly how rapidly the components of the G-field go to
zero: G ,..., R ). Notice also that here we have no scalar fields,
there is ~ degeneracy of the vacuum in the special sense of our earlier
discussion; nevertheless an analogous degeneracy arises from the possibility
of gauge transformations. In particular we can ask whether the Instanton
solution (3.9) can be gauge transformed to another form in which the
vacuum state at infinity is a normal vacuum, with B-field (and not ju;t
the G-field) equal to zero. By this we me.9p more precisely a B-field
that tends to zero at infinity faster than R
We can_ ~tate this requirement in another way, wh~makes it clearer why R is not lIfast
enough. II At large distances from the origin, neglecting terms of the
first (or higher) order in )..2/ R2 , the field defined by (3.9) (3. 10) is
indeed (as we noticed already) of the form (3.5) ; that is to say the
difference between this field and the normal vacuum (B J:1 = 0) is entirely
attributable to the gauge transformation g(x). With g(x) defined by Eq.s
(3.6) (3.8) there is, however, something quite peculiar about this
Jlvacuum state at infinity" in the sense that g(x) tends to a limit as
R .... CD, but this limit depends on the direction ~ The slow (l/R) decrease
of the B-fleld af infinlfyis indeed a dIrect manifestation of that dependence. We now r~~at our question: can the Instanton solution be
gauge transforme~ ) to another form, such tin t for large R the B-field
corresponds, in the Eq. (3. 5) to a gauge function which is constant at
infinity? On any restricted portion of the boundary B this is obviously
possible, but globally,trot is simultaneously over the whole boundary,
the answer is No.
The reason (you guessed it) is topological. Notice firstly that the
boundary value of the gauge function g(x} defined by (3.6) (3.8) is a
map B ..... G (= SU(2)). Since both manifolds are three-dimensional
spheres, we are reminded of the Hopf classification of maps S3 ..... S3
Treating the two manifolds as if thEo/ were the same sphere, we will

102

G. C. WICK

describe (3.8) as the identity map of an S3 onto itself, which as we

know is non-trivial, that is non-homotopic to a constant map. (It has
"wrapping number" equal to~, not zero).
Why is this relevant? Consider any map Bl.... G and suppose
f can be extended to a continuous function f(x) defined over the whole
space il"ltide B. It is easy to see that f(x) supplies a homotopy of the
map B4G to a constant map Gust consider the values of f(x) on
concentric spheres of smaller and smaller radius; as R -- 0, the limit
is the constant map with value f~. It is further easy to see that if
B is gauge transformed, as indicated in footnote 33 ), the gauge function
t~at generates the transformed field B' is f(x) g(x) and if the answer
to the stated question were Yes, it wcfuld mean that f(x)g(x) = const on
B. This is impossible since f(x) is "trivial" and g(x) is not.
Let me conclude this very brief description of the Instanton with
a couple of remarks, that may be of help to you, if you wish to do
some further reading on the subject.
The first remark concerns the expression of indices or IIwrapping
numbers ll in terms of integra~ The first famous example of such an
integral is due to Kronecker ). A four-dimensional integral expression
for the index of the Instanton is given in the quoted paper by Belavin and
0.; the integrand is then recognized as the divergence of a current3 5)
and transformed into an integral on the boundary at infinity. If you
engage in calculations of this kind, you probably ought to know that
quite considerable simpl ifications (and also a greater insight into what
you are doing) can be achieved if you adopt the notations of the
calculus of exterior forms. One way or another, the end product of
the calculation is that the three-dimensional integral reduces to the
invariant volume of the group space (see also the discussion session).
The second and final remark is that the boundary values of the
gauge function g(x), that we have studied, can be easily transformed
into a gauge function in ordinarY3 three dimensional space E3 by a
stereographic projection of B(= S ) into ordinary Euclidean space. If
the North-Pole (i. e. u4 = 1, U = 0) is used as the projection center,
that is if it goes to infinity in the projection, we then obtain a map
~--.. SU(2) (a gauge function in ordinary space) such that the group
efement tends to 1 at infinity. The static B-field generated by
this gauge function is then a vacuum-state, but a vacuum of a peculiar
kind. If we try to gauge transform this state to an ordinary vacuum
(B = 0) by means of gauge functions that have a constant limit at infinity,
wj discover that we cannot do that. The reason is the same as before.

THREE LECTURES ON SOLITONS

103

Whether this has a deeper significance, in connection with the "structure

of the vacuum" as some known AA. maintain is perhaps still a sub~gf
for discussion. I can only refer you to the papers cited in footnote
for further reading.
References
(1)

See e. g. Footnote 3 in Coleman. See also W. Blaschke "Oifferentialgeometrie" Vol. I. p. 207 (Chelsea Publ. Co. N. York, 1967)

(2)

This is a consequence of the Lorentz-contraction. Eq. (S. G.) is, of

course, relativisti c.

(3)

See for example, G. B. Whitham ,"Linear and Nonlinear Waves"

J. Wiley and Sons 1974, Chapters 13 and 17.

(4)

Another transformation is ( <1>, tjI, A, x) - (- tjI,<I>,A- -x).

(5)

Strictly speaking tjI is defined only up to an integration constant.

The notation (1.3') must be used with some caution.

(6)

See Bianchi's textbook cited by Coleman. The theorem assumes, of

course, a suitable choice of integration constants. (See earlier
remarks ~)

(7)

G. L. Lamb, Physi cs Letters 25A, 181.

(8)

See e.g. N. Christ, and T. o. Lee, Phys. Rev. C2H, 1606(1975);

R. F. [)ashen, B. Hasslacher and A. Neveu Phys. Rev. 011,
3424 (1975); J. L. Gervais and A. Jevicki, (1976) CCNY=FlEP76/2 preprint. L. O. Faddev, L. A. Takhtajan, T. M. P.21 , 1046(1975)

~)

See e. g. B. Yoon, Phys. Rev. 013, 3440 (1976).

(10)

T. Skyrme, Proc. Roy. Soc. A247, 260(1958); A262,L 237 (1961)

(11)

G. Mie Ann. Phys. 37, 511 (1912) 39, 1(1912) 40, 1 (1913);
See also W. Paull"lheory of Relativity" (pergamon Press 1958)
Part V S64.

(12)

R. Friedberg, T. O. Lee and A. Sirlin, Phys. Rev. 013, 2739 (1976).

The fields A and tjI in the original paper differ from those used
here by proportionality constants. We omit an overall factor g-2
in the Lagrangian. This factor is irrelevant at the c1assicm level.

104

G. C. WICK

(13)

Coleman (Ioc. cit.) refers to this argument as "Derrick's theorem"

compo G. H. Derri ck Journ. Math. Phys. 5, 1252 (1964) See
also J. Goldstone and R. Jackiw, Phys. ~ev. D11, 1486(1975)

(14)

There are, of course, two possible vacuum values for the field
A (= 1). This "degeneracy," however, will not play any
significant role in the following.

(15)

As indicated earlier, the omisSions are deliberate. The reader will

find much of the needed complementary information in Coleman,
Section 3. It should be possible to read the following, however,
independently of this reference.

(16)

A discussion of solitons involving Fermions (anticommuting fields)

is beyond the scope of these lectures

(17)

Our viewpoint is somewhat different from Coleman's. He claims

to be discussing a set of initial conditions (this includes
velocities).

(18)

In general G can act in the space of <I> through a representation

D(G). The representation need not be faithful. We shall also
occasionally assume that <I> is a complex vector, and G is
SU(n).

(19)

In the S. G. theory we have encountered an example of infinite but

denumerable set of minima of V(<I, i. e. <I> = 0, 21T, 41T,
etc.

(20)

Other, physically equivalent, solutions with A

by gauge-transformations. See Third LecturM.

(21)

We gloss over questions connected with

gauge field, see Coleman.

(22)

In ordinary field theories we have, of course, no other choice.

That interesting solutions arise if we allow a more general behavior, Eq. (2.2) was pointed out by H. Nielsen and P. Olesen
Nucl. Phys. B6 ~ 45 (1973).

(23)

That the two problems are nevertheless equivalent, when E is

ordinary space and B its boundary, is intuitively obvious,
but we shall not prove it.

0 are obtained

AO components of the

105

THREE LECTURES ON SOLITONS

(24)

This conclusion presumably extends to the quantised theory, since

there is an infinite potential barrier to be crossed.

(25)

There are quite simple cases of manifolds with non-abelian

fundamental group, e. g. the surface of a doughnut with two holes.

(26)

A sufficient condition for this has been stated by S. Eilenberg,

Fund. Math. 32, 167 (1939). It is satisfied if M is a sphere.
We cannot gOlnto this, however. See also: N. Steenrod "The
Topology of Fibre Bundles" S 16.5 .

(27)

A. A. Belavin, A. M. Polyakov, A. S. Schwartz and Yu. S. Tyupkin,

Phys. L. 59B, 85 (1975).

(28)

G. It Hooft Phys. Rev. L. 37, 8(1976) and Nucl. Phys. to be

published. R. Jackiwand C. Rebbi, Phys. Rev. D14,517(1976),
and to be published. C. G. Callan, R. F. Dashen, D. J. Gross,
to be published.

(29)

To obtain the customary expression for the Lagrangian

the trace. Remember Tr (TaT b) = ~ 6ab .

(30)

Unlike (3.4) the transformation law for the G-field:

_1
G (x) - g (x) G (x) g(x)
flV
flV

take twice

does not contain an inhomogeneous terms

(31 )

Please notice, although g(x) is ambiguously defined at the orlgm

and its derivatives are singular, the solution (3.9) (3. 10) is
regular everywhere.

(32)

See also It HooftIs ; symbol. One can write (j

= ~ (j
See especially Eq. (A.4) and (A. 13) in It HooftIs ra'per. aflV a

(33)

In pedantic detail: is there a gauge function f(x), well defined

and non singular over the whole Euclidean four -space such that
BI = f- 1 Bf +
1 a f has the requi red properties?
fJ
Y
fJ

.!. r

(34)

Interesting references may be found in J. Arafune, P. G. O. Freund

and C. J. Goebel Journ. Math Phys. 16, 433 (1975)

(35)

See also W. Bardeen Nucl. Phys. B75, 246 (1974)

G. C. WICK

106

DIS C U S S ION S
CHAI~N:

Prof. G.C. Wick

Scientific Secretary:

P. Rossi

DISCUSSION 1
ROSSI:
Can we study models contalnlng fermions, such as the SLAC bag,
with the same techniques as in the charged-boson case, in the case
of large charge quantum number or in the case of higher states?

WICK:
The anticommutativity of fermion fields, of course, requires,
in principle, different techniques. There is, however, some ingenious work by Neveu and others in which, by eliminating the fermion
variables, one is able to handle the problem by the usual technique.

GARCIA:
Does the Backlund transformation have a one-parameter group
structure? If this is so, could one learn anything about the solutions of the Sine-Gordon equation by studying the representations
of such a group?

WICK:
By combining an infinitesimal Backlund transformation with an
infinitesimal translation, one can construct the generator of a oneparameter group which, however, does not coincide with the set of
"finite" Backlund transformations. One does not get, unfortunately,
anything very exciting, since it is contained as a special case in
the general results of Faddeev.

107

THREE LECTURES ON SOLITONS

MARCIANO:
Are there uniqueness proofs which allow one to determine when
all solutions to coupled non-linear equations have been found?
For example, has it been done for a simple case like the Sine-Gordon
equation?

WICK:
I do not know of general proofs.
model at a time.

One would have to examine one

In the case of the Sine-Gordon equation, of course, the problem

has been thoroughly explored by Faddeev and others, and I think one
knows the properties of the system very well.
I assume you are asking about the classical, that is nonquantized aspect of the problem.

PAULI:
Could you illustrate roughly how soliton solutions could give
an oscillating bag and how possibly bag fissioning could be described
through non-linear wave equations?

WICK:
Let me try to ans,.,er especially the second part of your question.
There are well-known methods to study small oscillations around an
equilibrium configuration. On the other hand, I do not know of any
serious attempt to calculate bag fissioning. This would probably
require an enormous amount of numerical work, especially if you consider that the interesting problems would be in three-dimensional
space.

PARSONS:
You mentioned two categories of solitons: the mathematical
physicists' "collisionless" solitons and the topological type. Does
the Sine-Gordon "kink" type of soliton belong to both categories?

WICK:
To be more precise, I distinguished three different types. The
Sine-Gordon kink belongs simultaneously to two of these, that is it
is topological and it is collisionless.
A slight modification of the potential will give rise to a very
similar "kink", which is no longer collisionless, however. In other
words, it no longer satisfies the mathematicians' definition of soliton, as the numerical investigations of Ablowitz have shown.

G. C. WICK

108

DISCUSSION 2
POSNER:

I would like to know more about Hopf's work on the mappings of

Sm into Sn. Last year at Erice, Coleman used a theorem of Cartan's
to show that certain models could not possess monopole solutions.
What is the connection between Hopf's work and Cartan's theorem?
Are there any other physical consequences of Hopf's work?
WICK:

Hopf showed that all maps Sm + Sn, where m < n, are trivial,
that is homotopic to a constant map. When n = m, the equivalence
classes are in one-to-one correspondence with the integers. The
case m > n is much more difficult and is not completely solved even
now. Hopf gave some famous examples, such as a map S3 + S2, which
is, in essence, what physicists do when they associate a direction in
space to a normalized spinor or to an element of SU(2); as is well
known, the group space of SU(2) is homeomorphic to S3. The theorem
of Cartan that Coleman used is the statement that the second homotopy
group of a Lie group is always trivial. There is no simple connection between this and the results of Hopf -- of course, they belong
to the same chapter of mathematics.
The work on magnetic monopoles and on instantons are examples
of physical problems in which topology may be relevant.
LEADER:

Restricting ourselves just to one-dimensional equations, say,

of the type of the wave equation with some function f(~) replacing
sin ~, do the mathematicians know the conditions which f(~) must
satisfy in order that there exist "collisionless" soliton solutions?
WICK:

I am not aware of a general theorem that will tell you a priori

whether a wave equation possesses "collisionless" solitons.
ROSSI:

What are the relations between the homotopy groups of a space

and its connection properties?
WICK:

There are, of course, very deep connections. The simplest one

is the obvious statement of the equivalence between saying that a
space is simply connected and saying that the group TIl is trivial.

THREE LECTURES ON SOLITONS

DISCUSSION 3 (Scientific Secretary:

109

P. Sodano)

PARSONS:
In the paper by Polyakov et al. there was mention of the solitons
providing long-range effects and having a relevance to the infrared
problem. Do you have any comments on this? Could you generalize the
Polyakov solitons to SU(3)?

WICK:
To your first question, my answer is no. To your second question,
I choose the case of SU(2) as an example of the topological part I
wanted to illustrate. Undoubtedly, there are other examples. In
particular, the stability of the instnaton depends on the exi~tence
of non-trivial elements of the third homotopy group TI3, and this is
true for Lie groups other than SU(2). However, I have not looked at
these cases in detail.

ROSSI:
What about quantum solitons?

WICK:
The quantization of the classical solitons is, in a certain
sense, analogous to the well-known W.K.B. method. It was discussed
last year by Coleman. For a different method, I must refer you to
the paper by Christ and Lee.

ROSSI:
Topological solitons are often associated to spontaneous symmetry
breaking. Is it possible to study quantum solitons in Lagrangians
which do not possess classical ones, but show the phenomenon of spontaneous symmetry breaking by defining an effective Lagrangian in the
one loop approximation and obtaining Lagrangian equations of motion?
In other terms, can the variational equation for effective action
be used, not only to define the broken vacuum, but also for obtaining
non-constant solutions which can be dynamic topological soliton solutions?

WICK:
In principle, I see no objection to trying to do what you say.
As far as I know, it has not been done.

110

G. C. WICK

POSNER:

Last year at Erice, Coleman, in his lumps lectures, discussed

the patching together of two widely separated lumps. The process
involved the gauging of the invididual solutions at large distances
to a constant almost everywhere. The possibility of ambiguity increased greatly as the number of space dimensions went from three
to two. Is there also possible ambiguity in the procedure you discussed of generating two soliton solutions in one time and one space
dimension? You have discussed magnetic monopole solutions. Would
you elaborate on an antimonopole solution? Is there any possible
ambiguity in considering monopole-antimonopole solutions?

WICK:
In the procedure which I discussed of generating two soliton
solutions, there is no ambiguity. The possibility of ambiguity arises
perhaps in the case of topological solitons.
A magnetic antimonopole solution can be obtained from a monopole
by changing the sign of the Higg's field. A monopole-antimonopole
pair has a Higg's field at infinity which is homotopically equivalent
to the normal vacuum.
WILKIE:

vlliat is the equivalent for extended objects in relativistic

quantum field theory? In the non-extended case, it is impossible
to construct a local number operator; that is,that if one looks
closely at an electron, then one sees it, by the very act of looking
closely, as surrounded by virtual electron-positron pairs.

WICK:
It is possible, if one wishes, to describe a soliton. For
example, the kink solution of the Sine-Gordon equation, as a coherent
state, involving the excitation of one of the normal vacuum states
of the theory, such as the state with ()vac = 0 everywhere. The
number of elementary excitations, mesons, in the coherent state is
totally undefined.
JARVIS:

You gave the construction of the first homotopy group TIl, the
Poincare group, and the group operation. Could you also give the
definition of the group operation for TI2, TI3, and so on?

THREE LECTURES ON SOLITONS

111

WICK:

There are various ways of defining the higher homotopy groups

TI2, TI3, etc. of a space X. First one chooses in the space a distinguished point Xo. One then considers maps of spheres 82, 8 3 , etc.
into the space, subject to the restriction that a certain point on
the sphere, say the north pole, is always mapped into Xo. This
restriction is maintained throughout, for instance, in defining
homotopy between maps. Two maps are e~uivalent if they are homotopic;
that is to say the class of all maps 8 + X is subdivided into equivalence classes. These classes are the elements of the group TI2.
Analogously, maps 8 3 + X yield the group TI3, etc. For a connected
space, it does not matter much which point Xo is chosen as distinguished point.
In order to define the group operation, one may proceed as
follows: consider, for instance, TI2. Instead of a sphere 82, consider a square 1 2 , e.g. the set of all points (tl,t2) with 0 ~ tl ~ 1;
o ~ t2 ~ 1. One can obviously map 12 into 82 in such a way that the
boundary of 12 is mapped into the north pole and the interior of 12
is mapped homeomorphically, that is one-to-one and continuously
both ways, onto the sphere minus the north pole. We now replace the
maps 82 + X of the earlier definitions by maps 12 + X, such that the
boundary of 12 is mapped onto Xo. The composition is now defined as
follows: suppose f and g are two such maps, i.e. f(tl,t2) and g(tl,t2)
are functions of two variables tl, t2 satisfying conditions you can
easily state. We squeeze the map f into the lower half of 12 by defining

Likewise, we squeeze g into the upper half of 12 by

The two equations now define F over the whole square, and you can easily
see that the two definitions match on the dividing line tl = %. The
map F : 12 + X satisfied all the conditions stated earlier. It is not
hard to show that the homotopy equivalence class to which F belongs
only depends on the equivalence classes to which f and g belong. This
defined the group operation. It is also not hard to see that the operation is commutative, i.e. TI2 is always Abelian -- TIl is not always
Abelian! TI3 is defined analogously by considering a cube 1 3 , and so on.
GARCIA:

If I remember correctly, in 't Hooft's paper it is stated that any

Yang-Hills theory with a gauge group having a compact covering group
has a monopole solution. I am not sure if you have already explained
this in a different language. Could you clarify the statement?

112

G. C. WICK

KLEINERT:
This depends on the choice of the Higg's field. If you take an
isospin ~ field, all vector mesons become massive and no monopole can
exist.

JANCEWICZ:
You said during the lecture that g(X) is not homotopic to a
constant. One should rather show that B~(X) = g-ld~g is not homotopic
to a constant.

WICK:
There is no interesting homotopy question concerning B~(X).
The possible vacuum values of B~ are arbitrary. As a result, B~(X)
is always homotopic to a constant.

McPHERSON:
Expand upon the use of exterior forms.

WICK:
The general rules of the calculus of exterior forms are really
quite simple and more useful than most physicists realize. Allow me
to demonstrate this on an example.
In the discussion of the instanton by Belavin and others, as
mentioned in my lecture, there occurs the statement that a certain
integral, extended to the boundary of the instanton, must be an integer. It is further stated that this is so because the expression
under the integral sign
~

24n 2 Tr(g

-1

dg A g

-1

-1

dg A g

is the invariant volume element in group space.

second statement first.

(1)

dg)
Let us

exam~ne

the

I recall the notation: g is a matrix belonging to a three parameter Lie group G, specifically, SU(2); and g + dg is another element
of G differing from g by an infinitesimal displacement. The produce
under the trace sign in (1) is therefore a matrix product.
(1' )

THREE LECTURES ON SOLITONS

113

but also, as I have indicated by the wedge symbols A .. A, an

exterior product of linear forms. Since G = SU(2) is a three parameter group, I can write
-1

dg

r=l

a (u) du
r
r

(2)

I shall use the familiar parametrization

(3)

where
u~ + uf + u~ + u~ = 1

(4)

so that duo is linearly dependent on dUl, dU2, and dU3, and can be
eliminated as I have done in Eq. (2). Each ar(u) is, of course, a
2 x 2 matrix with elements (arhb that is gl dg has four matrix
elements (W)ik (i,k = 1,2) which are linear forms in dUl, dU2, and
dU3. The basic rule is that these differentials must be treated
as the generator units of a Grassman algebra, i.e. dUl A dU2 =
= -du2 A dUl, etc. There is only one unit of the third order, namely
in Eq. (1) must be a multiple of this. After
du
du
du and
some calculations, starting from (3) one gets:
Q

1
= 2n2

(uo dUl A dU2 A dU3 - Ul duo A dU2 A dU3 +

+ U2 duo

A dUl A dU3 - U3 duo A dUl A dU2)

(5)

This expression is nice and symmetric, but in order to demonstrate

what I first said, I eliminate duo and obtain
Q

= 2n2

-1

Uo

dUl A dU2 A dU3

(5')

The essential remark is now that this form is left- and rightinvariant. I recall that a left-translation is a mapping defined by
g + hg, where h is a fixed element of the group G + G. This entails
dg + hdg, and g-1 dg + (g-1 h- 1)h dg = g-l dg. Hence, each of the
linear forms is left-invariant, and this applies consequently also
to Q.
Under a right-translation g-1 d is not invariant. From g + gh
we get: g-1 dg + (h- 1 g-l)dg h = h- (g-1 dg)h. When we form the
triple product of Eq. 's (1) and (I') and take the trace~ the h drops
out again and Q is found to be left-invariant. As is well known,
these properties define the invariant volume element up to a proportionality constant. The constant in (1) is fixed by the requirement that the total volume be equal to one. This completes my comments on the second statement of Belavin and others .

G. C. WICK

114

Turning to the first statement, let us now assume that the

unitary matrix g, see Eq. (3), is a function g(X) of a point
X = (XO,X1,X2,X3) in a Euclidean four-space. Specifically, assume
X lies on a large sphere S3:
X~ +

xi

+ X~ + X~

= R2

(6)

which is the boundary of the region containing the instanton. The

assumption is, then, that the parameters Ur in (3) are given as
functions Ur = fr(X) (r = 0, , 3) satisfying (4) identically,
when X lies on S3.
This functional dependence then transforms the form
three-form in the X variables:

n into

a
(7)

The statement is then, that fn' extended to the sphere S3 of Eq (6)

is equal to a positive or negative integer. Without going into much
detail, I will sketch a possible proof. It is easy to see that the
form n of Eq. (5) or (5'), can be expressed, in many ways, as the
exterior differential of a two-form: n = dw. For example, we can
choose:
w

2~2

f(uo) (U1 dU2 A dU3 + U2 dU3 A dU1 + U3 dU1 A dU2)

(8)

provided f satisfies the differential equation:

(1 - u~)f' - 3uo f + 1

=a .

A solution is
f(uo) =

~ (1 -

uo)

-% arccos Uo - uo(l - uo) -1

(9)

This is, of course singular at one point in group space, namely

u = (-1,0,0,0). If this value does not occur amongst values
ui = fi(x) x 88 3 , then the two-form w' obtained from w,w' can be
approximated by transforming to x-variables the expression

It is then easy to see that the contribution of each singularity to

faw' is a positive or negative integer, which prove~ the assertion.
One final remark: the equations ui = fi(x) may be said to define
a mapping S3 + G, and since G = SU(2) is again an S3, as one sees from
Eq. (3) and (4), what we have studied is one famous Hopf mapping
sn + Sm, more precisely, the case n = m = 3. We have, of course, made
assumptions about the map f which are quite unnecessary; mere continuity is sufficient, but the proof then requires quite different tools.

CAN WE EVER UNDERSTAND HADRONIC MATTER?

A PROPOSAL

Giuliano PREPARATA
CERN - Geneva, Switzerland

The aim of these lectures is to discuss and clarify some important

questions concerning a recently proposed quark theory of hadronic matter 1)-3).
While the introduction and the discussion of quark spin effects,
internal quantum numbers and three-quark systems is necessary to turn very
theoretical proposal into a realistic theory of hadrons, it is nevertheless
crucial at this stage to ask ourselves whether it is possible to make this
general framework agree with some necessary general requirements.

And to

accomplish this we do not need to specify the otherwise important properties

like spin, and internal quantum numbers.
The two main questions which I shall try to answer in these lectures are the proper description of off-shell effects and an adequate definition of current operators on the space of physical states' wave functions.
In previous work

the importance of these problems was

adequately stressed, but the answer to them was left somewhat open, needing
a detailed analysis which had not yet been carried out.
I am quite aware that some of the conclusions I will
reach may have to be reassessed in the future, in the face of a deeper and
more accurate analysis.

However, I have not refrained from discussing

these pOints now, just to stimulate discussion on very important issues

which are not peculiar to this approach, but rather generally invest theories

115

116

G. PREPARATA

Quark confinement, i.e., the impossibility of observing quarks

as real physical states, is introduced through the requirement :

( 1 .2)
where

R4(p)

is a certain bounded space time domain (bag)

on the momentum

*),

depending

of our state, whose structure will be specified shortly.

To determine the quark "motion" inside the bag

some differential equations for the wf (1.1).

R4(p)

we need

The simplest choice appears

to be

_ 0 ...

i' l \>j )<.,"X.z.) =-

- 02 "-\'
for

x ~ R4 (p).

tn"Z

i> I,p; ", " ... )

Cr', 'X, ~ ') == tt?

..v tl"~

"1<,

'><2.)

However, as discussed at length elsewhere 1),3) , when one

converts (1.3) into equations for

*(p,x)

; i.e.,

or, equivalently

(1 5a)

one encounters the difficulty that (1.5b) carmot be matched with the solution
(1.2) outside the bag, due to its first order character.

The simplest way

out of this difficulty is to modify (1.5b) to

(1.5c)
with the further constraint that only that solution of (1.5c) should be kept
which corresponds to the minimum eigenvalue

woo

(1.5c) will only be compatible with those bags

Our equations (1.5a) and

R4(p)

which in the rest

UNDERSTANDING HADRONIC MATTER

frame,

p=(M,O),

117

factorize into a space bllg

(Rs)

and a time bag

(R t ),

i.e.,

(1.6 )
where the quark degree of freedom is unobservable (i.e., "confined").

We

can convince ourselves that the analyticity, the off-shell behaviour and the
currents' structure of a theory where the fundamental fields are confined,
must drastically differ from conventional quantum field theory, upon which
much of our intuition is based.

It is precisely the unavailability of

well-established reference points which makes an investigation of such

questions so difficult.

In order to voyage as safely as possible in such

quicksands, it is necessary that we check our answers by means of a number

of calculations of processes of fundamental importance.

For this reason extensive calculations will be outlined on such

processes as high energy scattering, deep inelastic phenomena and production
of hadrons at large

PT'

The very satisfactory picture one thus obtains

can be taken as an indication that we may be moving along a sensible track.

The plan of these lectures is as follows.

In Section 1 the basics

of our approach shall be briefly reviewed ; Section 2 contains a more detailed

discussion of the calculational rules, especially of the extension of our
rules in the case of complex loop momenta.

Section 3 takes up the question

of how to normalize the physical states, and therefore of how to introduce

currents in a physically meaningful way.

High energy hadron scattering and

the problem of Reggeization will be discussed in Sections 4 and 5.

Deep

inelastic phenomena are treated in Section 6, while hadron production at

high

PT

is the object of Section 7.

The final part will be devoted to

the conclusions.
1. - THE PHYSICAL HADRON STATES
A. - The meson

(qq)

wave function and its equations of motion

The simplest description of a meson state is given by the

--------------------------------------------------------------------------

*) The reader may notice some superficial similarities between this approach
and the so-called MIT bag 4).

G. PREPARATA

118

wave function (wf)

; f'l<.,+lt z

where

"f ("r;'Il"x,.,) = e

2.

is the total four-momentum of the

qq

four-co-ordinates of the

qq

pair.

(11)

~(\,;lt.)
system and

x 1 ,2

are the

Note that (1.1) is a trivial consequence

of translational invariance.
In the rest frame (1.5a) and (1.5c) become respectively

(1. 7)
and

( 1 .8)

The solutions of these equations which vanish at the boundary of the region
R4 (M,O)

given in (2.6) are easily found to be

with the eigenvalue conditions

( 1 .10 )
and

2~"t

( 1 11 )

~s (,10\)

Nn~(M)

is a normalization factor to be determined later,

spherical Bessel function of order

~(Q,~)

~'~n~

its

is the normalized spherical harmonic.

will give us a spectrum labelled by

~,

nth

j~(z)

is the

positive zero, and

Note that (1.10) and (1.11)

the angular momentum, and

n,

radial quantum number, only after we have specified the dependence of

and

Rt

on the mass

M of the meson state.

the
R

s
The important massive quark

model (MQM) criterion of universal quark mass behaviour of quark Green's

UNDERSTANDING HADRONIC MATTER

functions

5)

yields, as discussed elsewhere

119

1)-3)

(1.12)

which when inserted into (1.10) and (1.11) provides a set of exchange degenerate Regge trajectories which are almost linear and parallel.
tude of
Cl' ' ";;

R2
1 GeV- 2

The magni-

in (1.12) is fixed by the slope of our Regge trajectories: for

one obtains R2 ~ IT GeV- 2 We end this subsection by recording

the general structure of the high mass mesons wf in momentum space (see
Fig. 1)

Figure

The momentum space wave function

(1.13)
where

Ok

is the direction of the three vector

(1.14)
A-1(p)
function

being the Lorentz transformation bringing our state to rest.

0R2(Z)

is a "fat

The

function", characterized by the following

two properties :

(1.15)

(1.16)

120

G. PREPARATA

in the limit
0R2(z),

R2 ""(JJ

0R 2 (z)

tends to a Dirac

function.

A very simple

particularly useful in the calculations, is given by

(1.17)
Finally, in

(1.13)

the factor

Nn~

has been chosen in such a way that

(1.18)
one should keep in mind that this particular choice is only suggested by
convenience and has no physical motivation.

A physically meaningful norm-

alization can only be given by introducing the electric current ; this shall
be discussed in Section 3.

B. - Decoupling the high

states.

The fire sausage

We have just derived a meson spectrum consisting of approximately

linear and parallel Regge trajectories.

On one hand this looks very promising,

but on the other it may lead us into serious trouble if quarks do not
decouple the meson states whose angular momentum
0

is higher than a value

given by

..to =
where

R~(M)

( 1 .19 )

can increase at most logarithmically with the mass of the state.

The reason for this is quite simple; it stems from the fact that

if the strong interactions are to have a range which increases at most logarit~~ically

with energy, the maximum impact parameter of a

must also be so limited.

Calling

R~(M)

qq

such maximum value,

collisiom

(1.19)

follows

immediately.
Now, the quark coupling to the meson states is uniquely determined
by the wf, and we can eaSily check that if the normalization is given by (1.13)
2 2
at a given mass M all angular momentum states (up to ax = (R M )/n) are
coupled with comparable strength.

In our previous work

1),3)

an inadequate

definition of the electromagnetic current lead us to a normalized wf simply

proportional to

(1.13),

and to the consequent necessity to decree away the

UNDERSTANDING HADRONIC MATTER

troublesome states.

121

We shall see later that from a physically reasonable

electromagnetic current we can

this decoupling and we shall also be

able to give a very physical characterization of

Given that at a large mass
which

Lo'

RL(M)

appearing in (1.19).

M only those states are coupled for

we can get a nice intuitive picture of the space time struc-

ture of the typical meson state of mass

M.

We can in fact construct a set

of coherent states characterized by the direction

0= (Qo,(!Jo) ,

by taking

the following linear combination of the wave functions (2.13)

( 1 .20)

Notice that from the a.symptotic relation

( 1 .21 )
in (1.20)

is not an independent index.

Going back to configuration space

these states correspond to cylindrical space domain (see Fig. 2), which we
shall call fire sausage (FS), of height R :::::R 2M and width RL(M), containL

ing stationary quark waves moving with very small divergences

/(R 2ML! along the direction 00.

Figure 2

The space structure of a fire sausage

[ : : : (R L (M) )/

G. PREPARATA

122

On introducing an interaction among three bags along the lines

we shall discuss in the next section, we can study the decay properties of
such a peculiar hadronic structure.

This is a short summary of the main

results of such a study 2) :

i) the strongly preferred decay configuration of the FS of mass

is a

linear decay chain (Fig. 3) in which the initial FS cascades down

emi tting at each step a low mass state

(a) (n)

and a lower mass FS

oriented around the initial direction;

Tt

Tt

Tt

Tt

-I.I.L
p

Figure 3

r(M)

ii) calling

The linear decay chain of an

excited qq system

the width of FS of mass

M,

we get asymptotically

(1.22 )
where

depends on the coupling among bags

iii) the one-particle yield exhibits strong interaction scaling, i.e.,

2E
where

x= 2p

JiM

and

(1. 23)
and

Pol

are the particle momenta longi-

tudinal and transverse with respect to the FS direction

00

iV) the distribution in transverse momentum in (1.23) is strongly cut-off

with an average value given roughly by

(1.24 )

UNDERSTANDING HADRONIC MATTER

v) when

x--+O

f(x,p.L)

123

in (1.23) tends to a non-zero limit, as a conse-

quence the decay multiplicity of the FS increases logarithmically with

its mass

(1. 25)
where

is given by

These results suggest that what is experimentally observed, from hadronic

jets in

e+e -

annihilation to usual peripheral high energy hadron inter-

actions, originate from the decay of a small number of FS's which get produced in a variety of ways, depending, as we shall see later on, on the
particular process one considers.

C. - The

qq

Green's function.

The Regge states

From the bag states considered so far we can construct the

qq

Green's function following the usual procedure (see Fig. 4)

P/2+k

-p/2+k

P/2+k'

=
-p/2+k'
Fi~re

rn

rm

The diagrams making up the

Green's function

In order to have crossing invariance

?m=<
qq

G must be the sum of two

pieces,

(1 .26 )
representing the contributions of bag states in the
Note that the

and in the

channel.

channel, carrying non-zero triality, must be free from sin-

124

G. PREPARATA

gularities.

Let us study the

Gs

contribution first

we write

(1. 27)

where

11'

is a physically normalized wf (see Section 3), and

square of the centre-of-mass energy.

behaviour of

Gs '

s = p2

is the

We are interested in the high energy

Making use of (1.13), we can write

(1. 28)

where

<=.-, c.s ,I:;-,) =

ne

~
Zne

m;,2=((P/2)k)2, m1~2=((p/2)k,)2, Zn.t

normalize the wf and cosGs~(1+(2t/s.

t."2.eti) ?.(.lCa~G-s)

(1.29)

"-YY\~l"+ i ~r\,

is the factor needed to correctly

Note that in (1.29) we have added

to the dispersive denominator an imaginary part which, according to the dis-

cussion of the previous section ~ee (1.22 LJ is asymptotically constant.

shall now calculate the imaginary part
(1/Z n .t) = (2/Z o )G(.t o -.t)
easily get 1 )-3)

*) where

.to

1m Gs(s,t)

We

with the approximation

is given by (1.19).

From (1.29) we

exhibiting the "primeval pomeron" behaviour of the massive quark model 5)

As for the

Gt

contribution, we have for large

(1.31)

*) This approximation is adequate only for small t.

be more careful as will be shown in Section 7.

At large

one must

125

UNDERSTANDING HADRONIC MATTER

where

Gt,-"J-L..J
r s t'" _ ~ lX't'I) 11 c.-c.D~..')
and

cosQt = (1 + (2s/t-4m 2 )).

(1.32)

nt ~e (:b-m~r)

By a Sommerfeld-Watson transform we convert

(1.32) into
00

Ert lS.t) -= 1:, ltol~~) t:z.cnU:). 11

n=-,

Zlal,.\

SiTl1CClntt.)

..

'P

(1 .... 2s

~~) \

t-4.~

which is asymptotically Regge behaved :

G-t;lSlt ) _

1\d,lt-) t'2d,lt-) tol) _1_

S .... ~

Thus the Regge behaviour of

ao(t)

~W'l1tD(,~)

Gt(s,t)

Z,'-.oI,)

r (1+2",,)

rll+d,Y

~U:)

t:-~)

(1.34)

is controlled by a set of trajectories

corresponding precisely to the bag states.

These trajectories are

exchange degenerate, a crucial property in order to ensure the absence of

physical states (i.e., a discontinuity for
unobserved non-zero triality states.
s > 0,

s < 0)

corresponding to

This means that at sufficiently high

when we look at the discontinuities in the

channel we do not

find only bag states but also a continuum associated with the Regge pole.
These states must be looked upon as genuine physical states of a character
we shall call them "Regge states".

different from simple bag states

will fill up the final states with

n's

They

in a way similar to the bag states.

The only important difference between bag and Regge states at high energy is
that the former builds up diffractive scattering while the latter generates
normal Regge behaviour.
2. - HADRON SCATTERING AMPLITUDES
A. - Implementing unitarity in a perturbative way
In the preceding section we have explored the consequences of
some simple general ideas on the structure and the spectrum of meson systems.
We shall now take up the problem of how can a precise framework for hadronic
interactions be introduced.

As discussed elsewhere 1)-3) this can be done

along the lines of the massive quark model 5), and it can be shown that it
~ounts

of a

to a diagrammatical expansion of the

A~3

perturbation theory.

S matrix which parallels that

Actually further development of these ideas

has shown the incompleteness of such a scheme and the necessity for a piece
of the type

~~4

(see Section 4).

Enlarging accordingly our scheme, the

126

G. PREPARATA

parallelism between the

matrix expansion and the fictitious

ACP

+ fl.CP

perturbation theory is spelled out by the following transcription (see

Fig. 5)

t)
ii)
iii)

cP

propagator

cp3
4
cP

;:t

qq

vertex

;:t

irreducible six-point kernel

vertex

;:t

irreducible eight-point kernel

Green's function

V6
Vs

Figure 5

--~

Dictionary leading from Acp3 + fl.cp4

to our perturbative expansion

To any given order in

and

fl.,

which should then have the

role of relevant expansion parameters, we can draw the diagrams of

(Acp3 + f1cp 4 )

and then proceed to substi tu te the

Green's functions, the

the

cp4

cp3

with the kernel

cp

propagators with

with the irreducible six-point kernel

qq
V6 ,

and

VS.

Thus the whole construction of unitary hadron scattering amplitudes boils down to determining the structure of

G, V6 ,

governing their juxtaposition.

The structure of

the preceding section, the forms of

V6

and

Vs

Vs

will be discussed in the

next paragraphs.

Figure

and the rules

has been discussed in

The coupling of three-bag states

127

UNDERSTANDING HADRONIC MATTER

B. - The irreducible kernel

V6 ,

the coupling of three bags

and its extension to complex momenta

Let us look at the coupling of three-bag states (Fig. 6)

V6
bag, say

P1'

two-bag system

determines the probability amplitude for the quarks of one

to disappear from it and reappear as the constituents of a
(P2,P3)'

According to a geometrical picture of the inter-

action this amplitude should be zero whenever the space-time regions spanned
by the three bags do not overlap.

A way of achieving this is to write

(2.1 )

where

is a parameter, with dimension of a

of a coupling constant.

(mass)2,

playing the r8le

It can be interpreted as the amplitude for a quark

to tunnel from one bag to another.

According to (2.1) the three-bag coupling

is given in momentum space by the diagrams in Fig. 7, i.e., by the expression

[Pi == (Pi,nitimiLJ :

P,

Figure 7

P,

The diagrams describing the

three-bag coupling

Substituting the approximate expression (1.13) correctly normalized one

m2 = 0)

gets (we set for convenience

(2.3)

SR..

[f3~-~)1 \~(Q.. y~:Z(Ql~~) 1:5lQ.~-"i)

Cpr 1>3)'

128

G. PREPARATA

This quite formidable expression can be evaluated, to lowest order in

by making use of the

1/R2,

like properties of the "fat

We get quite easily:

z~ is the cosine of the angle between the vector k and the decay momentum
kp
2 2 2)
2
2
2
p=1/2m A(m 1 ,m 2 ,m 3 , where A(a,b,c)=a +b +c -2ab-2ac-2bc is the wellknown triangular function, which is positive provided we are in the physical
decay region
but for

A <

~1 2 (m 2+m 3

&~2

Ll

Notice that (2.4) is well behaved for

20,

develops exponentially exploding tails and we should try

and define such a coupling in a way which does not show this disease.

This

will be done shortly, but before doing it we would like to analyze the structure of (2.4).
The presence of the term
peaked for

zJq; = 1,

4
6 R2

makes the integrand strongly

and we get immediately

(2.5)

( 2.6 )

Thus (2.5) and (2.6) tell us that the matrix element strongly favours the
decay of the highly excited state
another high mass state.

P1

into a light meson (sayan)

and

This is the reason why the decay of a FS proceeds

through a linear chain, a fact which has been stressed in the previous
section.
Let us now turn to the case when

< O.

Its interest lies in

the fact that, in order to compute virtual effects coming from bag exchanges
we shall need to know how the three-bag vertex behaves in the unphysical
region.

To see what happens in Appendix 1 we report a calculation of a

129

UNDERSTANDING HADRONIC MATTER

simplified form of the vertex by means of Gaussian integration over the

loop momentum.

The disease for

A<

manifests itself by the presence in

the final expression of a logarithmically divergent piece.

We cure this by

a subtraction procedure, and then determine the subtraction by the requirement of continuity at

A = 0,

the boundary of the physical region.

As a

result ~ee (A 1.8) and (A 1.9 L], we get the following prescription
a complex momentum flows through a quark line the related "fat
is to be defined as

A < 0,

Thus for

(a,b

/)

whenever
function"

real)

the vertex is given by (2.4), where the expression

is to be interpreted according to (2.7).

c. -

The kernel

V8

and the four-bag coupling

The ideas of the preceding paragraph can be immediately carried

over to the description of the kernel
(Fig. 8).

V8

and the four-bag coupling

Thus we write

X,

P,

Y,
X2

P2

Y2

Fi~re

X3

Va

Y3
X4

Y4

P3
P4

The four-bag coupling

\I~l~) ... ,j",)-; b"'(X'-~)~\~')(4) b\':I"-"2) ~\)(&-~,)

;p2

~uta~iOTls.

is a mass parameter which should be equal to

tion of

~2

(2.8 )

~,

if our interpreta-

as a quark tunnelling amplitude is correct.

From (2.8) the

four-bag coupling in momentum space is given by the diagrams in Fig. 9.

P'TI P

130

P2

permutations

G. PREPARATA

P4

Figure

D. - Graphical rules.
The hadrons'

The four-bag coupling diagrams

Off-shell effects
S

matrix is therefore given by a collection of

diagrams which can be reduced to a network of quark lines joining quark

Green's functions and hadron wave functions.

The rules to evaluate any

such diagram are the following *) :

i)

each incoming bag of momentum

k.l.

ii)

Pi and relative quark momentum

is introduced through the properly normalized wf V(Pi,k i )

each outgoing bag state of momentum

momentum

iii)

kf

is given by

Pf

and relative quark

'*(Pf,kf )

each "exchanged" bag state (see Fig. 10) of momentum

relative momentum

k.e

is given by

(i.e)t(PA,k.e)'

P.e

where

and
.e

is

the angular momentum of the state ;

Figure 10
iV)

v)
vi)
vii)

An exchanged bag state

for each quark line we must multiply by

four-momentum at each two-quark hadron vertex must be conserved

for each loop we have
qq

an integration

d 4 .e/(2n)4

Green's functions are totally connected and their disconti-

nuities correspond to physical hadron states.

These rules, however, are not unambiguous.

In order to lift

the ambiguity we must specify the procedure by which off-shell effects

are to be calculated.
in

"TT-TT

To be concrete, let us take the simple Born diagram

scattering" (Fig. 11).

According to our rules we are led to a

*) For the reasons previously mentioned these rules turn out to be slightly
different from the ones given in Refs. 1)-3).

UNDERSTANDING HADRONIC MAnER

Figure 11

131

The lowest order diagram in nn

scattering and its quark counterpart.

two-loop integral of four wf's and one

qq

Green's function.

The question

now is how to calculate the Green's function to be inserted in the two-loop

calculation.
results *)
1)

There are two procedures which give in principle different

Calculate first the Green's function as a function of

and

fixed quark masses, i.e., put the imaginary part (1.30) into an
dispersion relation at fixed

t.

at
s

Insert then such a Green's function

into the two-loop integral.

2)

Calculate the

discontinuity of the whole diagram first and then

insert it in a fixed

dispersion relation.

While, obviously, the two procedures give the same imaginary

part (on-shell), the real parts in principle differ drastically.

It is

clear that in general only procedure 2 is in agreement with usual dispersion relations, and therefore should be followed throughout.

In more

complicated situations we shall have to perform a thorough analysis of

the correct analyticity properties of the amplitudes under consideration
before we can decide about the correct calculational procedure.

3. - CURRENTS AND NORMALIZATIONS

A. - A simple way to introduce currents

Let us now suppose that our quarks, still Lorentz scalars,
form an isospin doublet.

The states one can build out of a

will then carry isospin

and 0, and their wave functions are given by

qq

(a=0, ... ,3; i,j=1,2)

*)

In the MQM framework 5) procedure 2 was implicitly assumed.

*)

This was done in Refs. 1)-3).

pair

G. PREPARATA

132

(3.1 )
where

~ntm(p,k)

is the wave function (1.9), and for simplicity the iso-

triplet and the isosinglet have been assumed degenerate.

most natural way

*)

The simplest and

to introduce an isospin current is to write for the

current quark vertex operator (see Fig. 12) :

2:n
Figure 12

q/2+k'

The vector meson dominated currentquark vertex operator

Any current matrix element will be calculated by first folding (3.2) with
the relevant wave functions and then getting the real part by dispersing
in

q2.

This definition coincides in an obvious manner with generalized

vector meson dominance 6), whose phenomenological success is well known.

Using (1.5c) we can check that the current is conserved, i.e.,

In Appendix 2 a calculation of the form factor discontinuity

is reported for the generic state of angular momentum
Upon inserting the discontinuity in the

q2

where for
2\A L

_1_ _1_ _

'K~~1Z(\.fo\) I"nfl'\
The normalization condition

F Ch (O)=1

and mass

M.

dispersion relation we obtain

for the charge form factor :

i-9'Z

leads to

UNDERSTANDING HADRONIC MATTER

133

(3.6 )

For low partial waves

[j;S

(m p/MLJ

asymptotically one has

Co

If we go back to Section 1, we see that (3.7) wrecks the nice concept of

LJ,

the "primeval Pomeron" IEee Eq. (1.3

and with it the successful phenomenology of the MQM. This result is therefore unacceptable, and the only
way for us to salvage the situation is to admit that (3.2) does not represent the whole isospin current.
The vanishing of

Zt(M)

when

M becomes very large can be

understood if we think that the currents can measure the isospin content
of our hadrons only through their hadronic fluctuations, the vector mesons.

Any massive vector meson can sample the isospin charge only in space region
of the order of its Compton wavelength

1/mv

On the other hand the space volumes of our states increase

linearly with their mass, and massive vector mesons can only "see" a
fraction of the charge of high mass states which decrease with the mass
in agreement in (3.7).

B. - Adding a long range piece to the currents

How can we cure the pathological result (3.7)?

The only way

out of this difficulty seems to admit that the vector dominated piece is
only one contribution to the isospin currents, and that the full current
contains a direct coupling to the quarks comprising a hadronic state.
The idea that currents can couple directly to quarks is certainly not new and arises quite naturally

*).

However, it is not widely

appreciated that it is far from an innocent and harmless hypothesis.

In

fact if we try to implement it, we can see right away that all our (unproven) ideas about form factors and their analyticity properties are
ihcorrect.

*) For instance in the parton approach of Feynman and Bjorken 7)

**)

See for example the discussion in Ref. 8).

G. PREPARATA

134

Suppose, in fact, we know how to introduce this additional

direct coupling of the current to the quarks, we will end up with calculating diagrams of the type in Fig. 13.

Such diagrams are quite fami-

liar to nuclear physicists, it suffices to recall the calculation of the

deuteron form factor.

There the

q2

dependence arises from the well-

known anomalous thresholds which are sensitive to the structure of the

wave function.

Also there is no trouble with dtspersion relations due to

the fact that anomalous singularities do arise from the configuration

where all lines are on the "mass shell" **)

When we carry this picture

Figure 13

The direct current quark coupling

contribution to the hadron form factor

over to quarks, this does not make sense any more ; and we must face the
serious possibility that our ideas about form factors and their analyticity
have to be revised.
I am aware that this step is a momentous one and may lead the
whole theory into disaster, but I have found no other way out.
other hand this new piece of the currents might be just

wh~t

On the

we need to

account for the seemingly abundant production of photons in high energy

collision 9)
I propose that in addition to (3.2) the isospin current has
a , and the matrix element between any
a "direct" quark contribution,

:r

two bag states is given by

(3.8)

where

Zo

is a constant carrying the dimension of a mass squared.

In

order to ascertain that the new current piece is indeed conserved we must
utilize (1.5a) which is equivalent to

UNDERSTANDING HADRONIC MATTER

and the vanishing of

135

at the boundary.

From (3.8) we can readily evaluate the charge form factor for
any state of mass

t,

and angular momentum

in particular we get

Putting (3.6) and (3.10) together one gets the normalization condition

Z\l'
a
'2;

l' '"r

~[ ~1
~+~ . .

t>< "a
This result not only reinstates the "primeval Pomeron" behaviour, in fact

for

< (M/m)

at large

~o
2
but for

M/mp

Zt(M)

increases like

(1/M)e+ t (mp /M),

and this implie

that the high angular momenta decouple from the two-quark Green's function
exactly in the way which was discussed in Section 1.

One can estimate that

(3.11) corresponds to the approximation made in Section 1

...

(3.12)

with

10
~.I..
where

Mo

-e:
:: -

~ 1<.1.0"')
2

Mp

{ .,.-z. '\
tc,'t \ 'Mi}
<T

is a mass whose explicit expression depends on the constants

appearing in (3.11).
It is worth noticing that the "primeval Pomeron" acquires
[see (3.13) and (1.30LJ a logarithmic behaviour as well as a shrinkage.
A discussion of how this behaviour affects scaling and its possible
breaking will be presented in Section 6.

G. PREPARATA

136

Thus we see that while, on one hand, the original MQM picture 5)
is basically confirmed, on the other, one sees emerging new interesting
structures which get the whole approach closer to Nature.

We end this

section by adding an item to our list of calculational rules in Section 2

viii)

currents are given by the sum of two contributions which in the

fictitious

theory are given by the diagram in Fig. 14.

The

"vector dominated" piece is given by the quark diagram in Fig. 15,

while the "direct" piece is described by the diagram in Fig. 13,
multiplied by the normalization constant

Figure 14

Zoe

The current coupling ll" the fictitious

theory

Figure 15

The "vector dominated" piece of the

current

4. - HADRON-HADRON SCATTERING TO LOWEST ORDER

We begin in this section a series of calculations of several
interesting hadronic processes.

As emphasized in the Introduction the

importance of these calculations lies in the fact that from them we should
get a clear idea about the real meaning of the rules of Section 2.

Even

though the phenomenologically crucial quark spin has no place in all this,
the reader should appreciate that many of these results, when properly
transposed,

indeed relevant for the actual hadronic world.

A. - The production of resonances and of a single FS

The

diagram and its quark correspondent is reported in

UNDERSTANDING HADRONIC MATTER

137

Fig. 16, where the external hadron shall be always taken as the ground
qq

state.

Figure 16

The hadron-hadron scattering

amplitude to lowest order

The scattering amplitude at low energy will have poles in

to the positions of low mass resonances.

At high

corresponding

we calculate (see

Appendix 3)

which leads to a resonance cross-section decreasing like

1/s2.

Thus

single FS production in high energy hadron-hadron collisions is strongly

suppressed.

If, instead of FS production, we are interested in the product-

ion of Regge states, a calculation similar to the previous one

such production cross-section goes as s-5/2.

B. - Bag exchange.

a~ows

The question of Reggeization

The amplitude we want to compute now is the one shown in

Fig. 17.

..
Figure 17

The bag exchange contribution

to lowest order

that

138

G. PREPARATA

Proceeding in a standard way we write

gn' is the coupling constant of the excited state

states, and

n'

where

Q)Sel:; =

1+

Equation (4.2) is nothing but the model of Van Hove

known to lead to Regge behaviour.
for

cut?

(4.3)
10)

and it is well

Furthermore the amplitude

positive and bigger than a certain

vergence of the series defined by (4.2).

So

to the ground

at

exhibits

a cut coming from the di-

But what is the meaning of this

From Fig. 17, it is immediately clear that it cannot correspond to

a physical state, but rather to a state consisting of a Regge state and

a pair of non-interacting (spectators) quarks.

Thus (4.2) and (4.3) lead

us into the fatal difficulty of yielding a discontinuity which cannot correspond to an observable physical state.
let us go back to (4.3),

cosg t

Is there any way out?

Well,

is unambiguously defined only for

(i.e., on-shell) ; when we go off-shell

extrapolation, but far from unique.

t = m!,

(t/mn ,)
(4.3) is a possible
We shall show that if rather than

via (4.3) we extrapolate according to

1 ... 2s
m~e
our trouble will disappear and the amplitude defined by (4.2) and (4.4) is
actually analytic in

s.

To see this we need only recall that our particles

lie on Regge trajectories

does

m!"

m2n~a
+a'',
n,,- n

and that when

which makes the convergence of (4.2) for

increases so

fixed a trivial

matter to prove.
In Appendix 4 the behaviour of

at

for

->

(J)

and

fixed

and negative is estimated to be :

Q\:; - +

'...,+00

--

&_-<10

a (\0)

(4.5)

s~

bll:)

(4.6)

S2

which makes it legitimate to neglect this contribution at high energy,

except in some special cases for
may play an important role.

t~O

where non-Reggeized

exchange

139

UNDERSTANDING HADRONIC MATTER

To conclude, the main point of this section is that, in defining

the rules to compute "off-shell" behaviours, we should be guided only by the
criterion of physical interpretability of the amplitude discontinuities arising from our rules.

This criterion excludes the interpretation of the

diagram in Fig. 17 as a Van Hove model, and forces us to find an alternative

solution.

It is utterly non-trivial that such solution exists in the defi-

nition (4.4) of the meaning of cosQt' and that the amplitude so defined
does not appreciably contribute to the high energy behaviour of hadron-hadron
scattering~ee (4.5) and (4.6L].

In the language of dispersion relations

(4.4) corresponds to writing a

dispersion relation keeping

cosQt

and not

fixed; this strongly suggests that in calculating "off-shell" beha-

viours the correct procedure to follow will always be to disperse in the

variable we want to continue, while keeping all other independent Lorentz
scalar fixed.

5. - HIGH ENERGY HADRON-HADRON SCATTERING

THE POMERON AND

REGGE BEHAVIOUR
We have seen in the preceding section that the lowest order
diagram contribution to high energy hadron-hadron scattering is negligible.
Here we want to discuss how Pomeranchuk and Regge behaviour may emerge
from our approach.

Once we have disposed of single FS production, as an

unlikely high energy process ~ee Eq. (4.1

Ll,

the perturbative approach

described in Section 2 leads us to consider the production of two FS's

as the dominant high energy process.

This is actually what will be shown

in this section ; but before entering in the midst of calculations I would

like to recall that in previous work 3) the belief was held that a diagram
like the one reported in Fig. 18 would lead to the experimentally observed
Pomeron and Regge behaviours.

P,

Figure 18

FS 1

The two FS production diagram generated

by the iteration of the ~3 coupling

140

G. PREPARATA

However, carrying out a calculation similar to the one reported in Appendix 4,

this belief turns out to be unfounded, and one must conclude that one cannot
get an acceptable high energy behaviour for the two FS's production processes
through the "bag exchange" of Fig. 18.
As anticipated in Section 2, we must have a bag coupling scheme
which in addition to the three-bag vertex allows also a direct four-bag
coupling (see Fig. 19).

A. - The four-bag coupling and the Pomeron

The

discontinuity of the tWO-body scattering amplitude is

given by the diagram in Fig. 19.

P, :::IIcC)-_--{

r----fl=l-= p',

k,-k2=k; -k;
5= (p,+p/ =(p; +p~)2

)----0::-= p'

k :(p,-p/ =(pi _p',)2

Figure 19

The Pomeron diagram

The calculation is completely standard: we write down the wf's and the
Green's functions in the approximate forms (1.13) and (1.30), respectively.

We parametrize all momenta

a la

Sudakov :

and carry out all the integrals in the longitudinal variables

by means of the "fat"

Ii

functions and obtain :

(a,~,

... )

UNDERSTANDING HADRONIC MATTER

dic

where

a.

k1 ,

141

l~.....') c..1l-~)

1\1

~iOl)

are two-dimensional vectors and

t =

-"J?

C.V,,)

By using the

approximations (A3.5) and (A3.6) we can perform the three two-dimensional

ingegrals, and get :

d\'scs

ct ~ ~

l ~~f)4 C['t 1t~~S;

~ l~t)

(5.2)

We have indeed obtained a diffractive amplitude.

From (3.13) we

see that the total hadron-hadron cross-section goes asymptotically like

(log s)2

and that the diffraction peak shrinks with energy.

Our calcu-

lation, therefore, confirms that the origin of diffraction in hadron collisions is the production of two fire sausages through the four-bag coupling
mechanism.

Such a diffractive behaviour lends itself to a geometrical in-

terpretation in terms of a grey disc with a radius expanding logarithmically with energy.
To get the full amplitude we follow our general strategy and
write a dispersion relation in
imaginary part.

with

fixed (5.2) being twice the

The fact that the amplitude thus obtained is already a

quite good approximation to the real high energy amplitude, does nothing
but confirm that the expansion of the

matrix envisaged in Section 2 is

a meaningful one.

B. - The Regge behaviour

To obtain the observed non-leading Regge behaviour, we need
simply identify one of the blobs in Fig. 19 with the discontinuity of the

142

G. PREPARATA

part of the Green's function

Gt , which according to (1.34) gives rise

asymptotically to Regge behaviour. Setting

and performing a calculation analogous to the one reported above, we

easily obtain the non-leading piece of disc s a as

This result may appear surprising; we have in fact obtained unadulterated

Regge behaviour in spite of the complicated structure of the diagram leading
to (5.3).

We would have rather expected a Regge cut type behaviour.

reason for this is that at very high energy the

shrinks faster

(~log2s)

than that of

tion between the two Green's functions.

discG t ,

distribution of

The
discG s

thus dominating the covolu-

Putting (5.3) into a fixed

dispersion relation reproduces the correct Regge phase in the full amplitude.

We can therefore conclude that in our approach Regge behaviour does

not arise from the diagram in Fig. 17 but from the one in Fig. 19.

However,

amusingly enough, due to the peculiar structure of the Green's function

Gs '
everything goes as if we were calculating precisely the incriminated diagram

of Fig. 17.

6. - THE SCALING PHENOMENON

After having successfully passed the (qualitative) test of
hadronic scattering at high energy and recovered all essential experimental features, we must ask ourselves whether such a nice picture survives
another crucial test : the deep inelastic phenomena.

In this section our

theory will demonstrate its healthy status in this respect also.

In order

not to encumber the notation the case of scalar currents and scalar targets
shall be considered throughout.

A full account of the spin structure will

be possible only after we have introduced the quark spin degree of freedom.

A. - "e+e-

annihilation" into hadrons

From the discussion of the currents' structure carried out in

Section 3 the coupling of the current to a generic hadronic system is made
up of two distinct contributions : the "vector dominated" one and the

UNDERSTANDING HADRONIC MATTER

"direct" contribution.

143

The scalar current, whose effects will be calculated

here, will accordingly be given by

(6.1 )
In calculating the

e+e-

annihilation process into hadrons we must therefore

consider the two diagrams in Fig. 20 ; notice that we are neglecting interference terms between the two current pieces.

That this is legitimate

shall become apparent in the following

q/2+ k

q q

q
-q/2+k

-q/2+k'
(0)

(b)

Figure 20

The diagrams contributing to

"e+e- annihilation"

(a) - The vector dominated contribution

---------------------------------

The high energy behaviour of the imaginary part

Pa

of the

vacuum polarization function is given by

fa. t~'2.)

.1.
'2

r d\.

rd\"

J ~r')" J (2':':/"

clib<;' Gs

(6.2)

According to (1.28) and (1.30), (6.2) becomes

J.,l~.l~Z)
Q..L ~ (k-'''!)&

'h~1 \J--+\C.:t) ~IU (:+~L) - (6.3)

G. PREPARATA

144

The

k'

integration is easily made by use of the identity (in the sense

of distribution theory) 5)

Thus we get

o (_, ) ,

(6.5)

\~

where the non-leading term originates from the Regge behaviour of the
piece of the Green's function.

Gt

In order to appreciate the result (6.5)

it is necessary to calculate the "canonical contribution" reported in

Fig. 21 ; a trivial calculation yields immediately :

'\

(6.6)

1l,It

q/2+k
I

I
I

q--(I)--q
I

-q/2+k
Figure 21

The canonical contribution to

e+e- annihilation

i.e., the imaginary part of the vacuum polarization function for large
goes to a constant limit given numerically by (6.6).
notation we can define
annihilation, and obtain

Ra'

q2

Introducing a familiar

the ratio between the hadronic and the leptonic

UNDERSTANDING HADRONIC MATTER

145

(6.7)
Thus we see that the hadronic production cross-section has a pointlike,
scaling behaviour and the ratio

Ra

tends to a constant which is in prin-

ciple different from one, being determined by two crucial hadronic parameters like R2 and Zoo
this ratio should be one.

This is at variance with the general belief that

We now must turn to the contribution of the direct current

coupling.

Recalling the discussion of Section 2B, we need consider only

the configuration where a low mass hadron recoils against a high mass FS.
Proceeding in a, by now, standard fashion we find that also
this term scales and for Rb we obtain the unexpectedly large numerical
result:
(6.8)

The fact that ~ does not depend on the hadronic parameters can be understood by the cancellation between the Z2 terms arising from the current
coupling and the
normalization.

-2

Zo

term appearing in the denominator coming from the

No particular weight, however, should be given to the par-

ticular numerical results (6.7) and (6.8), what is quite important is

that both contributions (a) and (b) give rise to a scaling behaviour as
observed at high energy in the SPEAR experiments.
What about the one-particle distributions?

If you recall the

discussion in Section 1B, you can convince yourself that Bjorken scaling
holds for the "vector dominated" contribution due to the scaling properties
of the decay mechanism of the FS.

As for the "direct" contribution we also

get a Bjorken scaling as can be checked without any particular trouble.

G. PREPARATA

146

B. - "Deep inelastic scattering"

To lowest order we have the two contributions reported in

Fig. 22,

(p+q) =s
V=(pq}
p

(o)
Figure 22

( b)

The two lowest order contributions

to "deep inelastic scattering"

The evaluation of (a) presents the problem of defining the off-shell extrapolation of the vector mesons

m1 and m2 , The procedure we shall follow,

as previously discussed, is to evaluate first the discontinuity wa (sjm 12 ,m 2)
2
of our diagram by keeping

discontinuity Wa(SjQ2)
for fixed, i.e.,

m1 and m2 fixed. To get then the physical

we shall disperse in m; and m~ by keeping s

'tJ (Si<\Z)" 1: ~;rm, llYn.,.

..

m.1II.a

W(Siwn~.~ )
(m~-,I-) C.m:-Cf&)

(6.9)

where

is the current.-hadron coupling constant.

Appendix 5, we get [see (A5.6L]

From the calculation reported in

UNDERSTANDING HADRONIC MATTER

147

(6.11)

fl.:

-+

~~L\1C

lYIr
...mh
2.,$ )

The very interesting feature of this result consists in the peaking of

wa (s;m 12 ,m 2)
for m12 =m 22 , exhibited by the "fat 6". Thus in this approach
2
the so-called "diagonal" approximation of generalized vector dominance 11)
arises in a very natural way.

Moreover, this same mechanism will be always

operative in suppressing the interferences between the "direct" and the

"hadron dominated" pieces of the currents ; a fact already used in the
preceding paragraph.
Inserting (6.10) and (6.11) into (6.9) we have

(x=Q2/2v)

(6.12)

where

(6.13)

The features of

F (x)
a

worth notiCing are :

(6.14)
and

FA. tX) _

x.-..

What does all this mean?

We immediately calculate

\.i-X}.

(6.15)

Let us look at the free field diagram (Fig. 23).

148

G. PREPARATA

q
p

Figure 23

The free field contribution to

"deep inelastic scattering"

(6.16)
where

(6.17)

Thus our results (6.12) and (6.13) show scaling with a "structure function"
F (x) which is highly non-trivial. Equation (6.12) is the leading behaa
viour ; if we add the Hegge behaved piece discsG t we obtain a non-scaling
contribution of the form

(6.18)
with
F~()C.)

)(. .. 1

(6.19)

This piece could play an important r8le 5) in the observed pattern of

breaking of scaling behaviour near

x = 1.

We now turn to the contribution (b) ; the relevant calculation

is reported in the Appendix

6, which shows that we have a behaviour which

simulates scaling

(6.20)

UNDERSTANDING HADRONIC MATTER

149

with

""'"

r-",

12 i1-~

lll.,") '"

~-X)
~

In fact, due to the logarithmic dependence of

peculiar scaling breaking behaviour near
tends like

F (x)
a

(6. 21 )

~1ll~'X.)+.i. RZ
x = 1.

on

R~

For

s,
x

(6.21) exhibits a
small

Fb (x,v)

to a constant.

We want to ask the question whether

and

are enough

to take care of the scaling phenomenon in deep inelastic scattering.

Phenomenologically we know that the data are compatible with a "Regge behaviour"
of scaling functions near

x = 0.

mean for the scaling function

In our scalar case such a behaviour would

F(x)

(6.22)

and setting

cx( 0)

=1

for the leading Pomeron behaviour we get

Fb(x)

F(x) ~ x -2.

The constant limit reached by both

Fa(x)

and

obviously shows that

such terms must be unimportant for

x--> 0,

and it behoves upon us to show

that we do have a contribution which exhibits a behaviour such as (6.22).

The lowest order candidate for this is the Pomeron diagram in Fig. 24, which
we shall compute by following a procedure analogous to the one employed
for Wa (s,Q2). The calculation of Wp(v,Q 2 ) is reported in Appendix 7.
The result exhibits again scaling behaviour but this time

Figure 24

The Pomeron contribution to "deep

inelastic scattering"

150

G. PREPARATA

(6.23)

in complete agreement with our expectations and (5.2).

when

On the other hand,

clearly vanishes much faster than either

and

This concludes our analysis of deep inelastic phenomena.

It

is almost superfluous to stress that in this approach we could in principle

discuss several other topics like form factors, structure of final states
and massive

pairs production.

However, carrying out such an extensive

programme will be really meaningful only after we have introduced in a

realistic way the quark internal degrees of freedom:

SU(3), SU(4)?, ,

and spin.

It is by turning our attention to this very interesting piece of

physics, which has been unravelled for the first time at the CERN ISR that
we conclude our analysis.

Again no extensive discussion will be carried out

of these processes, and we shall simply concentrate on the leading contribution to large

PT

particle production in the kinematical region for which

x T = (2P';~S) 1.
As already discussed in the framework of the MQM we can isolate
two distinct contributions : A) the totally inclusive FS decay, and B) the
semi-inclusive decay at large angles.

We shall see that the results previous-

ly obtained in the MQM framework will be naturally reproduced.

A. - The totally inclusive FS decay at large angle

At very high energy production of large
according to the following mechanism : a

qq

hadrons can occur

pair produced along the colli-

sion direction according to the usual mechanism (see Fig. 19) can propagate
and emerge at large angle and yield along such direction a corresponding FS
which by its canonical decay (see Section 1B) produces hadrons at large
angle and consequently large

PT'

To evaluate the amplitude for the pair to emerge at large angle

we must sort out the large
in (1.29).

t
behaviour of the Green's function Gs given
Going to the impact parameter representation and making use of

the results of Section 3 (!;ee (3.11

D we

write :

UNDERSTANDING HADRONIC MATTER

151

and defining

')

Im tts c.s.~ =

1<.~s

2~')

00

'X

cb.

1+ ~
.fS

Expanding the exponential for large

(7.2)

:fo lx')
Y!f.lL

eo 'Z

c.-t'')~

we obtain :

IIn Gs(s.~) ~ 1tzs [-4Y'ne

toot')

.J-~~

-+

Dr ..l.j]

\....r~1'

(7.3)

and feeding

1m G (s,t) in a dispersion relation we get asymptotically a

s
*)
purely imaginary amplitude

To compute the production of large

PT

hadrons through this mechanism, use

will be made of the following formula :

,
where

d 4cr/dk 4

the vector
x= 2!p!IM

is the production cross-section of the FS as a function of

which can be taken as the four-momentum of one of its quarks,

is the usual Feynman variable, and

M is the FS mass.

From the results of Section 5 and (7.4) we can write the following expression for the FS production cross-section

d~
d~4

I'l. <llnr

J1
41'1
S

dVl'L
$

trS

J [,l"'"cit
0

~"L

l~

l1- 111-

-1'-

..

,
*)

~'2.)

(7.6)

Actually a more careful evaluation yields in addition a real term going

like 1/s ; but this will not change any of the following conclusions.

152

G. PREPARATA

and inserting it into (7.5), we get

Thus asymptotically we get the "scaling"

1/P~

behaviour.

The meaning of

this result and its experimental evidence are fully discussed elsewhere 5)-13).
However here we must recall that such a behaviour should be valid only at
very high energies and very small

x T j in an intermediate situation like

the one at ISR the leading role should be played by the semi-inclusive decay

of the FS, which will be now calculated.

B. - The semi-inclusive FS decay at large angle

We must now compute the diagram in Fig. 25, which describes a
mechanism for producing large

PT

hadrons which at not too high energies

is much more efficient than the canonical fire sausage decay.

Figure 25

t ~ (k 1 -p)
and
(k 1 +k2)2 ~ M2. Using

Fixing
dispersing in

The semi-inclusive FS decay

~2
1m Gs

we calculate our diagram by

as given by (7.3), we have

(7.8)

153

UNDERSTANDING HADRONIC MATTER

where the last factor comes from the loop integration ~ee, for instance,
(A5.6LJ.

Scaling the variables by

M2,

:J\ ()t~ "~I t ") -

we have:

<:4Ist

<:: .. )ls a~

1 _ J(....
M-

To get the final answer we must take (7.9), square it, sum over all states
.).t

the

~hich gives a factor

q

Rf()(.2/4LJ

and finally multiply everything by

pair distribution in a way analogous to the one employed in the

preceding paragraph.

Thus:

(7.10)

which reproduces the result of the MQM, and the experimentally observed
p;8

behaviour.

We do not expect that the introduction of spin will change

these results.
CONCLUSIONS
It seems appropriate to conclude by trying to assess where does
the proposed approach stand.

Let me emphasize once again that any conclusion

one can extract from this work can only be qualitative due to the neglect of
a realistic set of internal quantum numbers.

What has then been achieved ?

Starting from a minimally simple description of meson states in

terms of the quark co-ordinates we have obtained an infinite number of meson
states having well-defined space-time properties.

The spectrum is quite clo-

sely described by a set of parallel linear and exchange degenerate Regge trajectories.
Having constructed our meson states we have attacked the problem
of introducing an interaction among them following the ideas of the massive

154

G. PREPARATA

quark model.

The picture emerging is very simple and quite appealing, however,

it does not give us yet the possibility to calculate physical processes.

In

order to do this we must introduce current operators which will give a physical normalization to our states, i.e., through the correct specification of
the electric charge.

At this point, something very distressing happens ; if

we want to define the current operators through vector dominance the normalization, which requires an extension of the computational rules to deal flith
complex momenta (see Appendix 1), is found to be totally unacceptable.

This

prompts us to reconsider the definition of current operators, and to propose

a new definition which takes also into account the possibility for current
operators to couple directly to quark lines.

This proposal is a very moment-

uous one, it entails a drastic change of our views about the analytic structure of form factors, which may help us in explaining some puzzling photon
behaviours in high energy collisions.
The previous distress now turns itself into gratification,
because not only do we recover the nice properties of the massive quark
model, but also we obtain gratis the damping of the states with angular
momentum exceeding

Zo = (M/2)R.L'

where

R.L

increases logarithmically with

M.
At this stage the framework is precise enough so that we can
carry out calculations of several high energy processes.

The question of

Reggeization is taken up in Section 4 and it is shown that the diagram

usually associated with Regge behaviour (Fig. 17) does not Reggeize but
rather falls quite fast at high energy, thus presenting no problem with
unitarity.

The high energy behaviour of two-body scattering, discussed

in Section 5, shows the relation of the Pomeron with the production of

two fire sausages, as well as its peculiar geometrical nature.

A similar

behaviour is also shown by the normal Regge term.

Deep inelastic physics is amenable to a similar kind of analysis, which yields Bjorken scaling behaviour, and a physically appealing
distinction between "valence" quarks and "sea" quarks.
Finally, large

PT

physics is analyzed.

model picture led to the expectation that the "scaling"

The massive quark

p;4

behaviour

is delayed by kinematical effects, so that even at ISR energies the domi-8

PT' This is completely confirmed by the results

nant behaviour is still

of Section 7.

UNDERSTANDING HADRONIC MATTER

155

I believe that the quark theory of hadronic phenomena recently

proposed is showing a remarkable fitness for attacking the basic problems
of hadrodynamics.

It will not take long before we shall be able to answer

the question whether these ideas can lead us to the long sought mastering
of strong interactions.

156

G. PREPARATA

APPENDIX

integral

In this Appendix we shall calculate the simplified overlap

2
2
(Pl = P2 + P3 ; mi = Pi)

(A 1.1 )
where

In the physical region

to lowest order in

we can immediately calculate

R2

(A 1.2)
-(

We are now going to calculate (A1.1) in a different way which will give us
the possibility to extend our result to the complex momentum case

< O.

We write

~~

teZ

teZ

+et

j cld, Sdd j d~~ .) dk

1

-It"

-Itt.

t-e~

- i

~'"

+~? ...

~ S~I S~ fdd

-R1.

-v

-~

..

(A 1.3)
where the second line has been obtained through Gaussian integration.
changing variables
in

R2,

a 1 ->a=a 1 +a 2 +a 3 ,

On

and neglecting higher order terms

we can perform a Gaussian integration over

a3

to get

UNDERSTANDING HADRONIC MATTER

To leading order in

R2,

157

by integrating over

we obtain

(A 1 .5)
For

A?: 0,

(A1.6)
in agreement with

(A1.2) ;

! t'A<C)

~-

while for

fdcl

A< 0

we have

tt'Z

which diverges logarithmically.

We cure this disease by subtracting the

singular part of the integrand at

a 2 = O. The subtraction term can then

be determined by imposing the natural requirement that at the boundary
of the physical region
of

(A1.2)

(A = 0)

the two functions coincide.

Thus, instead

we get uniquely

1i1,'2.

j de( '2c.o~
o

(II.

I.
2.
z
~)

\.

2.

l1

-e_I(

O(.I'X\~

\)..\ '...

(A1.S)

One can interpret this result by saying that whenever a complex

momentum flows through a quark line the correct extension of the fat
function in this situation is given by

(a,b

real) :

(A 1.9)

158

G. PREPARATA

APPENDIX

In this Appendix we shall carry out an approximate calculation

of the charge form factor of an arbitrary bag of mass

um

and angular moment-

;,.
We begin with the contribution to the discontinuity

disc 2 < M;,IJ (0) 1M;' >1.

of the ith vector meson.
q
~
~
tion given by (1.13) and the rules of Section 3

Using the wave func-

(A2.1 )

where

cosG

is the angle between the vector

(1.14) relative to the wf

*;,(pl,k-(q/2)).

and the vector defined by

Notice that we have chosen to

evaluate (A2.1) in the rest frame of the state

Ip,lm >.

Proceeding in the

usual way we obtain :

elise 1'1.

'2 t+1

(A2. 2)

where

is the triangular function determining

the magnitude of the centre-of-mass momentum

\- ,

t:'

'I)..'"

=~

r~'Z.~'Zrn~)
'- I "

through the relation

(A2.3)
)

UNDERSTANDING HADRONIC MATTER

is the angle between

and
we calculate

159

and the vector

r "~~+I~~,... +.L
L ~

-z.

(lftj'&

the vector meson

m.,

> m2/4,
p

_I~ CD)~__,,40~'

. . -) j

(A2.2)

(A2.4)

with the mass of

(A2.3), ipi beco~es purely ima0R 2 function appearing in (A2.2),

ginary. Upon substituting (2.7) for the

we notice that the integrand is strongly peaked for

(A2.4)

cosQ pk = 0,

and insert-

we obtain approximately:

--L-

(A2.5)

2t+1

where

which for

By using (1.14)

we need only consider the contribution of the first

2
which we call p with mass m. For M >

i= 1),

A is negative and according to

ing this value in

k.

~2....,.

In view of the fast decrease of the integrand of

vector meson (i.e.,

.....

M large becomes

160

G. PREPARATA

APPENDIX

3
Here we shall carry out the calculation of the diagram appearing

in Fig. 16.

Following the steps which lead to (2.5), the imaginary part of

the scattering amplitude is given by

~\2.l~ ) ( ~)' fct2JOC(> Jdz' ~ dC{"

+I

0/

1t+

_I

'2,r

+1

-I

Z1t

"

<>

..to

"L

'b 4 lM~t (1-21)) ~ r~e \I-Z'') '(:'l~If') '(~ l&'J<f') ~r)"p

R,1."4
~ .. \. ~
~
.tmn
~.~ Y'-t- \.t/I'I'\e

We use the well-known addition theorem to perform the sum over

(A3.1)

where

For large

(A3.1) becomes

(A3.3)

where we have used, for small

.("

1:.

,h.

0R2

easily over

~,

and

the asymptotic formula

(?.e+i) ~(CD~tl-) ... ~ Jjl~r..t-)

The steep peaking of the

z',~'

t,

function allows us to integrate quite

and obtain:

161

UNDERSTANDING HADRONIC MATTER

(A3.5)

The integral in (3.5) can be easily carried out by use of the

following approximations (valid for small

z)

(A3. 6 )
and

(A3.7)
We obtain finally

(A3.S)

162

G. PREPARATA

APPENDIX

4
We want to calculate the high energy behaviour of

at

given

by (4.2) according to the prescription (4.4) :

(A4.1 )
where the coupling

is easily calculated from (2.5) ; one gets

gn~

c
Y\'\~e
Keeping

fixed, the sum over n

m2 , according to

(A4.2)

can be converted approximately into an

integral over

where the approximate spectrum equation

has been used.

Thus we write

.z

be)

cO

o.~~
which, defining

x=

C'2.

(2e . . i') 9-'2. C:I)(

becomes

""i&.'"

'2Ir

t=o

~ Islim,

'(i-

dt(l2.

m<4-

~ f. ~\ .... ~ \
m'-t" ~

~1.

[Is.! ~.L

1
IS1~

where in the

P~

lower for negative

L.t:.. C"Zeti,) C:-i)

be

(.

Jdx.
~

..

:3

I_too.,!.

Pt r1%'2,() (

-"
l'il

argument the upper sign stands for positive

s.

\.

s,

A4.5

and the

UNDER~TANDING

163

HADRONIC MATTER
2

To perform the L sum more eaSilY we introduce in the m

integration the smooth cut-off e-2Lx(SRf)rather than the sharp

O~R~/2L)-x)

; and we write

Ui

..
tj~
~~+i)~ ~

'l.

drill ~c

~~,sltl"'o

where

\-xa-

_x:!!

e -4ii~~

\1\

is the lowest mass in the hadron spectrum.

To perform the

(M.6)
L

sum we would like to use the identity

...

L ('t;t') = Z (~+-i) r:i) rt ~ (..iJ =

t ....

which is valid, however, only for

is

and analytic in

we extrapolate

lsi>

lsi

~ 1.

In view of the fact that

at

due to the convergence of the series, for

(A4.9)

by taking the average

(M.S)
On setting

,
and inserting

(A4.S)

into

(A4.6),

we get

(M. 9)

..L

lSI'

(M.10)

164
APPENDIX

G. PREPARATA
5

By means of the identity (6.4) we can dispose of one of the

two loops and reduce our diagram to a simple four-bag overlap :

(A5.1 )

2
where the factor R~/... has the same origin as in (A3.7), and
222
iJ.O = P1 = P2 is the mass of the target. In the centre-of-mass frame the
momenta of the targets

P1

and

P2

such that

(P1 - P 2 )2 = t = 0

are

given by :

b ...
11

r\:1'
~""'-'!'J

,0)

I~l)
1""1
(A5.2)

(A5.3)

where

(A5.4)
notice that for

m1 ;im 2

cosQ> 1

i.e.,

=0

is in the unphysical

region.
Integrating over
in (A5.1), we obtain

ko

and

/-->k/

by means of the

0R2

functions

UNDERSTANDING HADRONIC MATTER

For

165

Is-m~1 ~ 1/R 2 the integral can be approximately evaluated as

(A5.6)

166

G. PREPARATA

APPENDIX

The diagram in Fig. 22b is given by the expression

~~'Z(S~W~_z.Y) f~Z'~q>1 'f''\~~cr') 8:~ lS+4i C.1_z.)) .

-,

(A6.1 )

Using the addition theorem for spherical harmonics we have

(A6.2)

where

We now sum over all partial waves up to

get :

R~(~S/2)

and using (1.30), we

167

UNDER STAND ING HADRO NIC MATTE R

after integra ting over

By using the approx imation s (A3.5) and (A3.6), and
we obtain
Zl,

(A6.4)

and making the last trivial integra tion we get

(A6.5)

ity of introdu cing

This fills the gap left in (A5.1), by showing the necess
there the factor

168

G. PREPARATA

APPENDIX

Using the identity (6.4) we can reduce the calculation of

(S,Q 2 )
p

to the evaluation of the diagram in Fig. 26.

P= (q,+p,)= (q2+ P)

p2= 5

q-~.J

V, =(q,q)
V2 = (p, q)

Figure 26

The diagram equivalent to Fig. 24

We shall proceed as follows : first compute

.
d lsperse
In

.
m12 , m22 , b y k eeplng
s,
4
finally integrate over d q to get

i)

we write:

(A 7.1 )

169

UNDERSTANDING HADRONIC MATTER

Defining

P1

and

P2

as in (A5.2) and (A5.3), we notice the

kinematical relations

(A7.2)

We integrate over

and

k~,

obtain

and integrating over

ko

by means of

6R2~+q2_2V1-2V2-2ko(~S-qo-q3LJ

we obtain :

'ltE~!

'1b

'\'&.

,,'l

$-mr- 21lz.

"
...

~~

v!"

(('-.,1 ')cf.. 1'\.1'1~ - 4-.\~

f.

-"If-'N"

\
J

(A7.6)

170

ii)

G. PREPARATA

We now disperse over

mf

and

m~

and multiply by the current-

hadron couplings (6.10), and obtain

iii)

We finally integrate over

neglecting

d 4q

to obtain

Wp (s,Q2).

We have,

Q2/s

(A7.S)

171

UNDERSTANDING HADRONIC MATTER

Thus we have obtained for

W (s,Q 2 )
p

a scaling behaviour

with

(A7.10)

172

G. PREPARATA

REFERENCES
1)

G. Preparata and N. Craigie - Nuclear Phys.

2)

N. Craigie and G. Preparata - Nuclear Phys. B102, 497 (1976).

3)

G. Preparata - in Proceedings 1975 Erice Summer School (A. Zichichi, ed. )

to be published.

~,

478 (1976).

A. Chodos, R.L. Jaffe, K. Johnson, C.B. Thorn and V.F. Weisskopf Phys.Rev. D9, 3471 (1974).

5)

G. Preparata - in Proceedings 1974 Erice Summer School (A. Zichichi, ed.)

Academic Press, New York (1975), p. 54.

6)

A. Bramon, E. Etim and M. Greco - Phys.Letters ~, 507 (1972).

7)

R.P. Feynman - "The Photon Hadron Interaction", W.A. Benjamin, New York
(1972) ;
J. Bjorken - Proceedings Summer Institute on Particle PhYSics, SLAC
Report 167, Stanford (1967), Vol. 1, p. 1.

8)

R.J. Eden, P.V. Landshoff, D.l. Olive and J.C. Polkinghorne - "The
Analytic S Matrix", Cambridge University Press (1966).

9)

P. Darriulat et al. - "Large Transverse Momentum Photons from High

Energy Proton-Proton Collisions", submitted to Nuclear Phys.

10)

L. Van Hove - Phys.Letters 24B, 183 (1967).

11 )

M. Greco - Nuclear Phys. B63, 398 (1973) ;

H. Fraas, B. Read and D. Schildknecht - Nuclear Phys. B86, 346 (1975).

12)

R.E. Taylor - Proceedings of the 1975 Lepton-Photon Symposium,

Stanford University, SLAC (1976).

13)

G. Preparata and G. Rossi - "Looking at Final States : 1. The Large

PT Phenomena" CERN Preprint TH. 2163 (1976).

173

UNDERSTANDING HADRONIC MATTER

DIS C U S S ION
CHAIRMAN:

Prof. G. Preparata

Scientific Secretary:

M. Falcioni

DISCUSSION
BERLAD:
How is the baryon constructed?

PREPARATA:
The baryon has not yet been constructed. In this case, one
has two relative internal coordinates and if. one imposes appropriate
boundary conditions and differential equations on each of the three
coordinates -- quarks -- that compose it, one might hope to get a
unique description of the baryon.

BERLAD:
How could one get away with SU(3) triality states without introducing colour and glue?

PREPARATA:
It is of equivalent theoretical reliability to introduce triality
zero ad hoc and assume unseen colour and glue to produce it. In the
present approach, I try to avoid as much of unobserved quantities as
possible; thus colour and glue will not be used.

DE LA TORRE:
You told us that quarks are not fields or elementary constitutents
of hadrons; however, in your diagrams the lines look very much like
propagators of quark fields. I would like to ask you what is the
exact meaning of your diagrams?

G. Pfl!:PARATA

174

PREPARATA:

The diagrams drawn represent the Fourier transform of the overlap of the wave functions which, in turn, correspond to the transition probability as indicated ~n the diagram.

KLEINERT:

Why do you only include three- and four-bag couplings and not
higher ones?
PREPARATA:

It is a question of power counting in this perturbation theory:

If I do not go to five- six- etc.- bag couplings, I get a renormalizable theory, and I do not get bad behaviour at very high frequencies
just for the same reason you take 3, 4 and not 5, 6, and so on.
PAULI:

Could you please compare and contrast your work on fire sausages
and strong interaction dynamics with that of the CIM (Constituent
Interchange Model) dimensional counting rules of Brodsky, Blankenbecler,
Farrar and Gunion.
PREPARATA:

We do not need to appeal immediately to fire sausages and can

go back just to the massive quark model. CIM is good in fitting data
but lacks physical basis. CIM has some diseases as other models with
the production of objects with zero triality. With bags, we avoid
their intepretational troubles. Similar behaviour comes again from
free quarks, both are related to the primeval Pomeron which gives
power counting rules and one obtains similar but slightly different
results to CIM. This is more physical and ties in a nice way to
the bare Pomeron.
PAULI:

Can your theory be applied to explain the prompt production of

leptons in hadron-hadron collisions, such as pp ~ ~X, pN ~ eX?
If so, what sort of results do you obtain and how do they agree with
data?

UNDERSTANDING HADRONIC MATTER

175

PREPARATA:
Prompt production of leptons is still mysterious. The combination of many effects can fit data as suggested by Lederman and
others:
i)
ii)
iii)
iv)

v)

hyperon + strange particle leptonic decay

vector meson-photon couplings
new particle (J)
electromagnetic lepton production:

new lepton or other new particles .

Looking simply at the lepton-pion ratio may not be the best

way of understanding the subject. Small PT leptons may come from
direct production by photons coupling to quark lines in the fire
sausage model -- fire sausages sometimes look like Hertz dipoles -which can radiate high-energy photons which, in turn, can give rise
to prompt leptons.

ETIM:
Would you elaborate on the compelling reasons for introducing two
kinds of couplings of hadrons to the photon?

PREPARATA:
First, the sampling of the charge content of the hadron by the
photon through its coupling via vector mesons involves space time
volumes of order (limp) and that is much less than the volume of a
fire sausage. The direct photon coupling to a quark pair can probe
unlimited volumes. The two couplings together therefore probe the
whole fire sausage and can be normalized to give the correct charge
of the firesausage at zero momentum transfer.

FREEDMAN:
Should not any extended model of hadrons enjoy the same success
in explaining Zweig's rule, as long as over-lap between bags 1S necessary for hadron-hadron interaction?

PREPARATA:
I do not know of any model formulated in this way which can
"explain" the Zweig rule. Let me explain how my approach can do this.
First, for the case of the ~, the corrections to the Zweig rule come
from unitarity through the self energy diagram

176

G. PREPARATA

and the fact that this is small -- ~ 10- 1 the direct amplitude -~s supporting evidence for the "weakness" of strong interactions.
In the case of the ~, supposing, as many do, that it is a cc system,
we have the analogous diagram

o
Normal
hadrons
and we must see whether we get an extra factor 10- 1 needed to explain the extraordinarily small r tot ~ 50 eV. It is quite straightforward to show that this factor comes out quite naturally and is
approximately given by

where me ~ 1.5 GeV and mx ~ 0.3 GeV. The reason for this factor stems
from the fact that the wave functions are quite strongly peaked around
the "mass" of the quarks. In the last loop in the right, the integration runs, for kinematical reasons, over a negative value of the
"mass" of the c quark; and one is only drawing on the small tail.

ETIM:
Is the perturbation coupling constant in your model small because
effectively, a large part of the interaction has been used in manufacturing your bags?

PREPARATA:
Yes.

You can look at the situation that way.

PARSONS:
If strong interactions are weak, would you not expect the indirect contribution via vector mesons to be small compared with the
direct one in the deep inelastic processes?

UNDERSTANDING HADRONIC MATTER

177

PREPARATA:
In the indirect type of scattering, one is scattering off an
extended fire sausage type of structure, which for simple geometrical
reasons has a large overlap with the incoming photon or neutrino.

CAN PEDESTRIANS UNDERSTAND THE NEW PARTICLES7 *

Harry J. Lipkin
Weizmann Institute of Science, Rehovot, Israel
Argonne National Laboratory, Argonne, Illinois 60439
Fermi National Accelerator Laboratory, Batavia,
Illinois 60510
I. INTRODUCTION

What is so special about the new particles and why is

there such excitement? So many old particles have been known for
a long time that the discovery of an additional particle has
created very little excitement. An unbiased observer from another
field would conclude "seen one particle, seen them all. ,.
However, the new particles were immediately seen to be peculiar
and interesting because they are very narrow states at high
excitation and nobody still understands why these states are so
narrow. The only argument given supporting the narrowness is
based on the Okubo-Zweig-lizuka rule, but nobody understands the
OZI rule even for the old particles, where many interesting open
questions still remain. There must be interesting physics in this
rule worthy of further theoretical and experimental investigation.
The major part of these talks is devoted to interesting questions
regarding the theoretical validity and possible experimental tasks
of the OZI rule.
In trying to explain to some of my nuclear colleagues
why these new particles are so exciting and interesting, I first
tell them they are narrow resonances that appear at much too high
excitation for their narrowness. The nuclear physicist says,
'We know about those things. We have them too; isobaric analog l
states". Then I explain that they were completely unexpected.
No one thought that they were going to be there. He says, "Of
course. Theorists didn't expect isobaric analog resonances
either. But as soon as they were found, they had the explanation ".
*Supported in part by the Israel Commission for Basic Research and
the U.S.E.R.D.A., Division of Physical Research.
179

180

H. J. LIPKIN

Then I say, "That's the difference. As soon as the new particles

were found, all the theorists came out with explanations, but all
of them are wrong. And they still don't understand why the
particles are so narrow." As soon as the new particles were found
theorists dug into their old files and tried to show that their
old theories really predicted these particles. One theorist
actually quoted the reference to an old paper where he claimed he
had predicted these particles. One of our nasty graduate students
actually looked up the reference and gleefully circulated the
abstract around the department. The abstract said that this was
the only paper explaining the new weak neutral currents without
requiring the existence of new particles for which there was no
experimental evidence. The status of the new particles is well
described by the following quotation:
'T have no data as yet. It is a capital mistake
to theorize before one has data. Insensibly one
begins to twist facts to suit theories, instead
of theories to suit facts."
... . A. Conan Doyle
'The Adventures of Sherlock Holmes'
Niels Bohr developed his model of the atom on the basis
of the experimental data of the Balmer series. Much hard work by
many people was then needed to get modern quantum mechanics
started. But today's theorists are trying to develop a theory for
the new particles as beautiful as modern quantum mechanics when
they don't even have the Balmer series.
During the time that this talk was being prepared and
even during the Erice school itself, new data were being
accumulated indtcating that the new particles were those theoretically predicted by the charm scheme in which an additional
fourth "charmed" quark 2 is added to the three conventional
members of the quark triplet. All this charm spectroscopy was
known long before the discovery of the new particles. Searches
for charm have been suggested for a long time. Yet nobody
suggested that SLAC search for very narrow resonances in electronpositron annihilation in the 3-GeV range. It is instructive to
examine why nobody suggested such a search.
Vector meson states constructed from a charmed quarkanti quark pair were predicted long before the discovery of the new
particles but they were not expected to have narrow widths. Since
their decays into ordinary uncharmed' states was known to violate
the Okubo3-Zweig4-lizuka5 rule, these states were expected to be
narrower than conventional uncharmed states at this mass. But
there was no reliable theory underlying the OZI rule and its
breaking and no theoretical calculation predicting the strength of
OZI violation. The only clue was experimental OZI violating

UNDERSTANDING THE NEW PARTICLES

181

decays into non strange hadrons of mesons consisting of a strange

quark-anti quark pair e.g., the ~ pTI decay. These indicated that
OZI suppression factors were one order of magnitude, possibly two
but certainly not more. This would still leave a large width for
a state at 3 GeV with many open channels. Such a state would not
easily be seen as a resonance in electron-positron annihilation.
Thus even if the charm model is correct, one crucial step
is missing in the description of the new particles and responsible
for the failure to predict their discovery. This missing link is
understanding the OZI rule and why the suppression factor is very
much larger for the new particles than for forbidden old-particle
transitions. This question is still open and considered in detail
in these talks. Some indications, but no conclusive answers, are
given, but answers should not be expected from this talk. They
say that ,{hen .one asks a Jew a question he answers by answering
with another question and when he is asked why he always answers
a question by asking another question, he answers "why not? ".
I shall raise many questions in this talk but I shall not answer
them. Instead I will raise more questions. I hope that pursuing
the answers to these questions will lead to even more interesting
questions and to a better understanding of hadron physics.

II. THE SUe 6) BANDWAGON

How can pedestrians understand the new particles when
we still have so much trouble understanding the old particles?
We still don't understand why the old hadron 9pectrum has been fit
very successfully by an su(6) symmetry scheme 6 which suggests that
hadrons are built from elementary objects called quarks with spin
1/2 and three flavors. If hadrons are made of quarks, and the
forces are independent of charge, strangeness and spin, all of
these six states are equivalent and transformations among them
generate an su(6) symmetry. Particles can then be classified into
su(6) multiplets. The lowest-lying mesons and baryons fit very
beautifully into two su(6) supermultiplets, the baryons in a
56-plet, the mesons in a 35-plet and a singlet. Since the su(6)
scheme was proposed more new evidence has been found for additional
56-plets, 35-plets and 70-plets. Rosner's review7 at the 1974
London conference listed the known su(6) multiplets as a
''Michelin Guide" in which he gave four stars, three stars, and
two stars to the multiplets, depending on how well they were
established experimentally. But where are the quarks?
Now we have a new su(6), which I call the Sicilian su(6)
because it was invented by a Sicilian (with some help from another
island). Arima and Iachell0 8 suggest that we should not stop with
the 56-plet obtained by putting three basic building blocks having
six states in a totally symmetric configuration. Why not try the

182

H.J. LIPKIN

252-plet Or 1~87-plet; obtained. by using five or eight building

blocks? Arima and Iachello have played the standard game of
building an 8u(6) supermultiplet degenerate in the symmetry limit,
removing the degeneracy by using a simple ansatz for symmetry
breaking and obtaining a mass formula which they compare with
experiment. Figure 2.1 shows a typical hadron spectrum obtained
in this way from the 1287-plet. Figure 2.2 shows a comparison
with experiment of the lowest states in the 252-plet and the
spectrum of hadrons with baryon number 170 and electric charge 68.
Further spectra and comparisons with experiment are shown in
Figs. 2.3 and 2.4.
This 8u(6) bandwagon is very amusing: Particle physicists
build particles from a fundamental building block with six possible
states and introduce an 8u(6) symmetry. But nobody has found any
quarks, and more data are needed to see whether this symmetry is
really there and particles are really made out of quarks.
Now the nuclear physicists have jumped on the 8u(6)
bandwagon. If particles can be made out of a sextet of objects
that are not really there, maybe nuclei are too. But the ArimaIachello nuclear 8u(6) model is not based on elementary fermion
quarks. Their building blocks have six states, but they are
bosons, one with spin zero and one with spin two. Everyone knows
that there are no elementary bosons in the nucleus, but the
agreement with experiment shown in Figs. 2.2, 2.3 and 2.4 is just
as impressive as the quark model fits to particle data. Perhaps
the bosons in the nucleus are just as real or unreal as the
quarks in the particles.
Why do nuclei look like composite systems of 8 = 0 and
8 = 2 bosons which nobody has seen? Why do hadrons look like
composite systems of spin-1/2-three-flavored quarks which nobody
has seen? Are these bosons or quarks confined? Are they in a bag?
Or are they simply not there? Perhaps there is an underlying
substructure which makes nuclei behave as if they were made of
bosons in certain experiments and makes hadrons behave as if they
were made out of quarks.
But bosons in the nucleus is not really as crazy as it
sounds. We know that nuclei are made of neutrons and protons. If
they pair in some fashion they may behave somehow like bosons.
Before the BC8 theory there were suggestions that Bose
condensation of electron pairs was responsible for superconductivity.
Then BC8 showed how a proper treatment gave not only the properties
that look like bosons but also the important differences between
fermion-pairs and real bosons. 80 maybe there is something in
this boson model of the nucleus and we are waiting for the right
theoretical description.

183

UNDERSTANDING THE NEW PARTICLES

2+0'-

10~

6~

4~

8+-

13'_
16+-

(0,2)

9+-

14'-

K'O

10~

12+-

2~

0'-

12~12'-

6~

K'O

4+-

(0,8)

5:=

>

K'O

9+-

14 -+:-

Q)

(10,0)

8~8+-

7'-

w
12+-

6~6+-

5'-

4~4:_

10'-

a'-

6C~-=
K'O

K'2

(12,2)

6+4'-

2~

0'~'O

(16.0)

Fig. 2.1.

Typical hadron spectrum from 1287-plet.

E (keV)

Exp

Exp

Th

{3

Exp

Th
+

~ 8-1798.4

1500

8+--913-8+-- 899.3

Th

7+--1598.5
6+--1423.7
5+--1273.8

4+--1123-4+-- 1148.9

1000

6+--1423.7

10+--1374 -10+-- 1373.9

2+--959.8-2+-- 974.0
0+-890.9-0+-- 899.1

K=O

K=O

+
~ 4+-- 1148.9
\ - - 1101 3+--1049.0
3 --1010.5-2+-- 9740
2+--932.

K=2

K=2

(6.2)

500 6+--541.1-6+-- 524.6

4+-- 260.1-4+-- 249.8

+
+
2+--78.59-2+-0-0-0-

K=O

170
68 Er 102

74.9
0

K=O

(10.0)

Fig. 2.2. Comparison of 252-plet spectrum with experiment.

184

H. J. LIPKIN

Exp
E(keV)

Exp

Th

Th

Exp

Exp

Th

Th

(3

8+ 3482
6+ 3180

3+ 2162_ 3+ 2160

Th

b.

4+ 3195

5+ 3056
5+ 2927/ - -

3000

Exp

2+ 2817

2+ 2287 -2+ 228~

2000
4+ 1542 -4+ 1542

0+ 1473_ 0 + 1473

2+ 1476 -2+ 1402

1000
2+ 658 __ 2+ 722

O+~-O+~

Fig. 2.3. Boson model fit to the 110Cd vibrational spectrum.

><1l
~

Exp

Th

19

5
17

Exp
4

Th
14 +

15

14+
13
3

12+

12+
II

10+

10+
9

8+

8+

6 '

6+

5--3

4+

4+

2+

2+

0+

0+

150
62 Sm SS

Fig. 2.4. Boson model fit to the l50sm vibrational spectrum.

UNDERSTANDING THE NEW PARTICLES

185

Let us now discuss the nuclear su(6) and Figs. 2.1-2.4

in slightly more detail. Figure 2.1 shows a typical su(6) supermultiplet with the energy spectrum obtained when su(6) is broken
by a particular type of boson-boson interaction that preserves the
subgroup SU(3). The spectrum looks very much like that found in
the nuclear collective model with a ground-state rotational band
and beta and gamma vibrations. The states are labelled by the
quantum numbers of the SU(3) classification. Data are fitted in a
large group of nuclei in the rare earth and transuranium regions
where these rotational and vibrational spectra occur. The example
of 170Er shown in Fig. 2.2 has experimental energy levels which fit
two SU(3) multiplets which are just the lowest two multiplets
found in a single su(6) supermultiplet.
In particle physics, the SU(3) subgroup of su(6) gives
a good classification for baryon states, but the su(4) group works
better for mesons. For nuclei the SU(3) subgroup of SU(6) gives a
good description of rotational nuclei; while the SU(5) subgroup
works better for vibrational nuclei. Figures 2.3 and 2.4 show
typical vibrational spectra and some fits to vibrational nuclei.
III. THE SEARCH FOR NEW DEGREES OF FREEDOM
What are these new particles? They indicate some new
degree of freedom, but what is it? At the Palermo conference,
Cabbib09 presented the charm approach as analogous to the search
for the planet Neptune where other data on irregularities in the
orbit of Uranus had indicated something must be there. MY guide
to any search is two key questions lO : 1) who needs it? and 2)
who cares if it is not found? In the case of the planet Neptune
anybody who believed Newton's descriptions of the motions of the
planets knew that something had to be there. It would have been
very serious if nothing was found to produce the observed
irregularities in the orbit of Uranus. But if all the new
additional particles that are suggested by new theories are not
found the theorists will find new excuses for their absence and
change the theory a bit to expJain it. As a guide to the search
for new degrees of freedom, it is instructive to recall the search
for a higher symmetryll,12 that eventually turned out to be SU(3).
It begins with isospin, which is SU(2) and strangeness, which is
U(l). The correct higher symmetry SU(3) which included SU(2) and
U(l), was found by an eight-yedr journey, in which all possible
wrong symmetries were tried first. When they finally found SU(3)
they called it the eight-fold way because it took them eight years
to find it.
Why did it take so long? Why did they try everything
else because before they tried SU(3) instead of noting that
SU(2) + U(l) C SU(3) is as simple as 2+1 = 3. Physicists are not

186

H. J. LIPKIN

stupid; the reason that they could not see that 2 + I = 3 was
because they did not know that they had two. In those days
isospin was believed to be a rotation in a three-dimensional space
like ordinary spin, described by isospin operators similar to
angular momentum operators. The natural candidates for a higher
symmetry to include isospin rotations were rotations in four, five,
six, seven and eight dimensions. None of these worked because the
algebra of the group of three dimensional rotations is accidentally
isomorphic to the algebra of two-dimensional unitary transformations
and isospin is really SU(2), not 0(3). The two-dimensional Hilbert
space of proton and neutron states and the transformations of
protons and neutrons into one another have no relation to any
physical three-dimensional space.
Beyond three dimensions there is no longer this
isomorphism between rotations and unitary transformations. Thus
theorists could not get anywhere by extrapolating what they
already had. They had to learn something new, but they did not
realize it. At the Princeton Institute for Advanced Study many
now famous theoretical particle physicists did not bother going to
Guilio Racah's famous lectures on Group Theory and Spectroscopy 13
because they did not think unitary groups had anything to do with
particle physics.

Now the pendulum has swung to the other direction. We

know all about SU(n). We begin with n basic building blocks and
define unitary transformation among these objects to make an
SU(n) symmetry. Now that something beyond SU(3) is needed,
theorists play the same games with SU(n) instead of rotations:
SU(4), SU(5), su(6) and so on. If we keep it up, we will get to
SU(ROSENFELD) where n is the number of entries in the Rosenfeld
table of particles and all particles are classified in the
fundamental representation.
This reminds me of an explanation I heard from a
physicist who works in the field of controlled thermonuclear
reactions about the difference between CTR and particle
accelerators. The particle physicist builds an accelerator. It
works and he is happy and does some physics with it. After a
while he realizes that to progress further he needs a bigger
accelerator. He gets money to build it, builds it, it works and
he is happy and does some physics. After a while he realizes that
he needs an even bigger accelerator, etc. etc. The CTR man builds
a machine which does not work and he is unhappy. Then he decides
that if he had more money he could build a bigger machine that
might work. He gets the money, builds it, it does not work and
he is unhappy. He then decides again that if he had more money he
could build a bigger machine which might work etc. etc.

UNDERSTANDING THE NEW PARTICLES

187

The quark model14 started with the idea that everything

is made from three fundamental building blocks. The experimentalists looked for the quarks and did not find them. So the theorists
said maybe there are more quarks. It started from three, it has
gone up to four, nine, twelve, etc. The current popular colored
six-quark model has 18 quarks. But still nobody is finding them.
So, perhaps what we need is not to keep adding more of the things
that we know, maybe we need to learn something new. In other
words, maybe the quark is something like three-dimensional
isospace, a useful realization of the symmetry at a certain stage
which enables us to do calculations very nicely. But it freezes
our intuition in the wrong direction and thus hides the new things
that we may have to learn to advance to the next stage.
IV. THE PRESENT STATUS OF THE NEW PARTICLES AND THE OZI RULE
Let me now briefly review the status of the new
particles. They are narrow resonances discovered at SLAC as peaks
in the cross section for e+ - e- annihilation into hadrons and at
Brookhaven as peaks in the mass spectrum of e+ - e- pairs in
nucleon-nucleon collisions produced with hadrons. In the SLAC
experiment the mass of the is about 3000 MeV, the instrumental
resolution is about 1 MeV, and the width of the is even smaller.
The peak is already smeared by the instrumental resolution by an
order of magnitude, and would be smeared further by magnet drift
unless the magnetic field is kept stable to better than 1 part in
3000. Such a narrow resonance was not expected by theory and there
is no point in getting that much stability in all the apparatus if
it is not needed. So the particles were discovered by accident.
The first J or that was discovered had a mass of 3100, the same
quantum numbers as the photon (spin 1, odd parity and odd charge
conjugation), and decays into hadrons. Soon afterwards, a whole
family of particles were found having the same quantum numbers as
the photon. This does not mean that most of the new particles
have the same quantum numbers as the photon; it is just that an
e+ - e- colliding beam experiment excites most strongly those
states having the same quantum numbers as the photon.
Once these are found one asks what they might be;
something completely new, a new kind of weak or semi-weak boson,
a new kind of quark-anti quark pair, having a new quantum number
like charm, or some hadron with a new internal degree of freedom.
With too many possibilities and not enough experimental data, we
can try to look for general properties and draw conclusions which
are not too model dependent. It seems fairly clear that the
production is electromagnetic, because it all fits together with
what is known of electromagnetic production via a single virtual
photon. The particles have the quantum numbers of the photon-i~ they are produced by some other mechanism there is no reason
to pick out 1- states in particular.

H. J. LIPKIN

188

The most peculiar property of these particles is the

inconsistency between their production and decay. They are
produced electromagnetically by e+ - e- collisions presumably
through a photon, and with a production cross section comparable
to that for the ordinary vector mesons P, Wand . But they
decay into hadrons with a very narrow width. The characteristic
widths for vector mesons are 100 MeV for strong decays like P ~ 2IT
and 1 MeV for electromagnetic decays like 1tI ~ TTY. This is
consistent with the picture that 100 MeV is a strong width, and
electromagnetic widths are first order in a and down by a factor
of 100. The J/I/J decay into hadrons is of the order of 100
kilovolts while decays into hadrons +y are very small. Thus both
hadronic and electromagnetic decays are down at least three orders
of magnitude from the expected decays of an ordinary hadron.
Relatively large electromagnetic production and suppressed
decays into hadrons suggest that the J/I/J might not be a hadron at
all, but a composite state of some new kind of fermion-anti fermion
pair, coupled to the photon because it has electric charge, but
not coupled to hadrons. That does not work either, because the
coupling to hadrons is actually too strong. This coupling is
conveniently measured by a quantity called R defined by the
relati on1 5
R

cr(e+e- ~ hadrons)
cr(e+e-

U+fl-)

(4.1 )

Production of P+~l- pairs is known from quantum electrodynamics and

experiments agree with QED predictions. Thus the u+u- cross
section provides the scale for measuring other processes. In
parton models where hadron production occurs via the production of
an intermediate state of a quark-anti quark pair by the virtual
photon, the ratio R depends on the 6number of different kinds of
quarks and their electric charges l
R =

(4.2)

where Q. is the charge of the quark of type i and the sum is over
all typ~s of quarks. Experiments show that R in the vicinity of
the J/I/J but off resonance is of the order of two to five, but
inside the resonance it rises to 25.
The J/I/J is produced electromagnetically with a strength
comparable to the P, W or . It is more strongly coupled than the
photon to the hadrons but decays much more slowly than it should.
So in some sense its coupling to hadrons is too small but it is
also too big.

UNDERSTANDING THE NEW PARTICLES

189

I conclude this review with a more detailed picture of

why this narrowness problem is such a nuisance. There are two
problems. The widths for both electromagnetic and strong decays
are too narrow. We first consider the electromagnetic problem.
The paradox is that no simple selection rule can forbid the
decay of J!~ into something else plus a photon. Any selection
rule based on a conservation law that forbids electromagnetic
decay also forbids electromagnetic production. We know that in
e+ - e- annilihation both the J/~ and ware produced with
comparable cross sections. Thus the transition matrix elements
for the two processes are comparable

<~IJ

em

10> rv < wlJ em 10>

(4.3)

where J
denotes the electromagnetic current. Thus the J!~
cannot em have a peculiar eignevalue of a new quantum number,
which is conserved in electromagnetic interactions. Therefore, we
cannot forbid by symmetry the decay of a J!W to a photon and some
hadron states having the same quantum numbers as the vacuum for all
conserved quantities. Since 3100 MeV can make many pions there
are many such states. But experimentally we know that the sum of
the squares of these matrix elements over all possible states is
still very much smaller than the squared matrix element for the
w ~ ny decay. That is the electromagnetic trouble.
For the strong decays, a selection rule is possible.
There are "generalized color models", wite color symmetry in which
all ordinary hadrons are color singlets. l ,17 If the Wis not a
color singlet, its decay into ordinary singlet hadrons is
forbidden, but the photon can excite it because the photon need
not be a color singlet. But in all such models the electromagnetic
trouble is still there because the Wis allowed to decay into
color singlet hadrons plus a photon. Thus generalized color
models solve the strong decay trouble but cannot explain the
electromagnetic trouble.
There is a dynamical selection rule which may be
relevant known as the Okubo-Zweig-lizuka rule. We know from
ordinary hadron physics that the ~ pn and fl ~ 2rr decays are
suppressed. The initial states of both decays contain a strange
quark-anti quark pair and the final states contain no strange
quarks. Some dynamical principle suppresses the transition in
which a strange quark-anti quark pair annihilates and only
non strange quarks are produced. This suggests that if the ~ is a
new kind of quark-antiquark pair like a charmed pair there might
be a similar principle preventing a charmed pair from
disappearing. But this selection rule can only hold in Born
approximation. It cannot be rigorous in higher order because a

H. J. LIPKIN

190

succession o~ transitions which satisfy the selection rule can

produce a ~orbidden transition. For example both the which
consists only o~ strange quarks and the W which consists only o~
nonstrange quarks are coupled to the KK state. So the can go
to a nonstrange ~inal state via the intermediate KK and til states.
Thus the basic question o~ why the new particles are so
narrow is still not understood. There are all kinds o~ hand
waving explanations that may be right but are still unconvincing.
It may very well be that the real structure in the new particles
is completely di~~erent ~rom anything that is being considered
today.

v.

THE OKUBO-ZWEIG-IIZUKA RULE

One o~ the principal open problems in trying to understand the old as well as the new particles is the Zweig rule, or
the Okubo-Zweig-lizuka rule, as it is now co~only called. The
Okubo ansatz,3 which antedated not only Zweig 4 and Iizuka 5 but
even the quark model, applied to nonet couplings and gave all
results ~or the three-meson vertex later obtained by Zweig and
Iizuka ~rom quark diagrams. The quark-line rules of Zweig and
Iizuka define one possible generalization of the Okubo ansatz for
four-point functions and more complicated vertices, but this
generalization is not unigue. Okubo has pointed out other
possible generalizationsl~ that may be relevant to experiment.
Previous papers have separated the "cookbook rules" of Zweig and
Iizuka and the Okubo ansatz, since the validity of the Okubo
ansatz for three-meson couplings is experimentally well established, whereas the particular ZI generalization to more complicated
vertices has not yet been conclusively tested. These notes follow
the present common usage of giving Okubo proper credit for his
pioneering work by using the name OZI rule. This leaves some
ambiguity in its definition for four-point and higher couplings as
discussed in detail below.

5.1.

Pedagogical Examples and Some Basic Questions

The OZI rule has entered the folklore of particle

physics without any clear theoretical understanding or justification.
At the present time nobody really understands it, and anyone who
claims to should not be believed. Investigating the OZI rule for
the old particles raises many interesting questions 1 9,20 which may
lead to a better understanding of strong interactions as well as
giving additional insight into the experimentally observed
suppression of new particle decays attributed to the OZI rule. We
follow an iconoclastic approach emphasizing embarrassing questions
with no simple answers which might lead to fruitful lines of
investigation.

UNDERSTANDING THE NEW PARTICLES

191

The most common applications of the OZI rule are selection

rules forbidding the couplings of the ~ and fl mesons to non strange
mesons and nucleons

gf

'TTTT

(5 .la )

ISwprr

gWTT

gNN~

(5 .lb )

gf[lTT

(5. lc )

~Nw

(5 .ld)

~f' gNNf

where the couplings on the left-hand sides of the inequalities are

forbidden and those on the right-hand side are allowed. Since the
selection rule is not exact, the degree of suppression is
expressed quantitatively by comparing the corresponding forbidden
and allowed couplings appearing in these inequalities.
The selection rules can also be formulated in terms of
two-body reaction cross sections,
O'(TT - P -+ fu)

O'(TT-p -+ llln)

O'(TT-p) -+ fIn)

a(TT-p -+ fn)

(5.1e J

(5.lf)

Although these agree very well with experiment, no

consistent theoretical or phenomenological model explains them
without raising paradoxes and contradictions. There is also no
theoretical indication of how good the selection rule should be in
different processes; i.e. no description of the breaking mechanism.
A principal difficulty to be overcome in any theoretical
formulation is that a succession of transitions all allowed by the
OZI rule can lead to one which is forbidden. For example, all
forbidden couplings (5.1) can proceed through an intermediate KK
state via the following transition amplitudes observed experimentally and allowed by the OZI rule.
T(~ -+ KK)

(5. 2a )

T(fl -+ KK)

(5.2b )

T(KK -+ OTT)

(5. 2c )

T(KK -+ mr)

(5.2d)

T(KK -+ l'iN) 1= o

(5. 2e)

H. J. LIPKIN

192

All the selection rules can thus be broken by the

following allowed higher-order transitions

(5. 3a )
f' ... Id("'TTTT

(5. 3b )
(5. 3c )

f, ... Id("'NN.

(5. 3d)

If the OZI rule holds only to first order in strong

interactions, much greater violations are expected than those
experimentally observed. Some mechanism for reducing these
violations seems to be present. One possibility is a cancellation
of the violating amplitudes (5.3) by other amplitudes, as occurs
in the case of symmetry selection rules. This is described in
detail below.
The essential features of many problems arlslng in
applications of the OZI rule are illustrated in the following
examples.
Consider the decays of vector mesons into two pseudoscalar mesons. The following decays are all allowed by the OZI
rule and observed experimentally.

reo ... 2fT) ~ 0

f(K* ... KTT) -t 0

(5. 4b )

r(cp ... Id() -t

(5. 4c )

(5. 4a )

The decay

f(CP ... 2fT) "'" 0

is forbidden by the OZI rule and experiments are consistent with
zero decay rate. Thus, the VPP decays (5.4) all agree with the
OZI rule.
But the decay
f(w ... 2fT) "'" 0

is allowed by the OZI rule and is also observed experimentally to

be very weak. This apparent contradiction is resolved by noting
that the decays (5.4d) and (5.4e) are forbidden by G parity. The
OZI rule is thus completely irrelevant to VPP decays. A similar
situation obtains for tensor-vector-pseudoscalar decays, where the

UNDERSTANDING THE NEW PARTICLES

193

f' 4 orr decay forbidden by the OZI rule is also forbidden by

G parity.
This example shows that experimentally-observed
suppression of a transition forbidden by the OZI rule does not
necessarily provide evidence for the validity of the rule. The
transitions may be forbidden for other reasons.
The VPP example also raises two questions with interesting
implications for the general case.
1. The Doubly Forbidden Question. Since both the
W 4 2rr and 4 2IT decays are forbidden by G parity but the ~ 4 2rr
decay is also forbidden by OZI rule, is the doubly forbidden
4 2rr decay weaker than the W 4 2IT decay?

+ _
2. The Higher Order Paradox. The scattering amplitude
T(K K 4 n+n-) is allowed by both G parity and the OZI rule. Thus
the 4 2rr decay could take place as a two-step transition in which
both steps are allowed
(5. 5a)
How is this transition inhibited in a theory of strong
interactions where there is no small parameter to make secondorder transitions weaker than first order?
The answers to these questions are simple in this trivial
case and very illuminating for more interesting non-trivial cases.
1. Double Forbiddeness. There is no simple answer to
this question. A transition already otherwise forbidden can be
additionally suppressed by the OZI rule only if the dynamical
process which breaks the other selection rule also respects the
OZI rule. In the case of the VPP decays the G-parity selection
rule is broken by electromagnetic transitions which violate the
OZI rule. The OZI violating transition

and the OZI conserving transition

(5.4e)
have couplings of the same order of magnitude. The OZI conserving
transition (5.4e) may be favored by ki,nematic factors if the r 4 21f
transition is dominated by the P pole, but this has no simple
connection to the OZI rule.

194

H. J. LIPKIN

2. The Higher-Order Paradox. The answer is that the

allowed transition (5.5a) is exactly cancelled by the transition

This cancellation is characteristic of transitions

rigorously forbidden by a conservation law. The conservation law,
in this case G conservation, must hold to all orders. G-violating
contributions to the transition amplitude can arise from particular
intermediate states which are not eigenstates of G. These
contributions must be cancelled by other contributions in the sum
over all intermediate states related to these states by the G
transformation.
Note that this cancellation (5.5) does not occur in the
non-trivial selection rules (5.1) allowed by G parity. The
contributions to the higher-order transitions (5.3) from the
K~- and KOKO intermediate states have the same phase and cannot
cancel. Thus, any cancellation must come from some other state.
An instructive example of a symmetry selection rule
broken in higher order is the SU(3) and charge conjugation selection
rule forbidding the transition 21
(5.6)
However, this reaction is observed experimentally near the mass of
the ,
(5.7a)
All the individual transitions are allowed by charge conjugation
and SU(3). If charge conjugation and SU(3) are exact symmetries
this contribution (5.7a) to the amplitude (5.6) must be cancelled
by other contributions. In the SU(3) limit the 0 and ware
degenerate with the and the amplitude for the reaction (5.7a) is
exactly cancelled by the amplitudes for the reactions
(5.7b)

(5. 7c)
In the real world the 0, Wand are not degenerate and no such
cancellation occurs at the peak. The SU(3)-violating reaction
(5.6) thus occurs just as strongly as the corresponding SU(3)
conserving transition to the ~- final state.

UNDERSTANDING THE NEW PARTICLES

195

The implications of these examples for the OZI rule are

clear. Higher-order OZI violations can be suppressed by cancellations from different immediate states. But such cancellations
require a degeneracy of the relevant intermediate states. Without
exact degeneracy the OZI rule will be broken like the SU(3)
selection rule forbidding the transition (5.6). But no physical
state is degenerate with the KK intermediate state occurring in the
transitions (5.3). Thus these cancellations can at best be only
approximate.
We now pose a number of interesting questions which have
no simple answers today.20
1. What is the theoretical basis of the OZI rule? Can
it be formulated with predictive power to give strengths of
forbidden transitions at least on the phenomenological level?
Could there be a description analogous to the Cabibbo description
of strangeness violation in weak interactions where one or more
parameters analogous to Cabibbo angles describe the relative
strengths of OZI-conserving and OZI-violating transitions?
2. What is the experimental evidence for the OZI rule?
How many of the so-called OZI selection rules also follow from
other considerations like G-parity conservation and therefore do
not really test the OZI rule? How much unprocessed or easilyavailable data could be used to test the OZI rule?

3.

like SU(3)?

Can the OZI rule be exact in some symmetry limit

Can it be formulated as a conservation law?

4. How is the rule broken and by how much? Can the rule
be kept exact for vertices with all the breaking introduced in
properties of external particles and propagators for virtual states,
as in the conventional SU(3) phenomenology?

5. Where does the rule apply? To baryons as well as

mesons? To the not ideally mixed pseudoscalar nonet as well as
ideally mixed nonets? To the new as well as the old particles?
Is it better in some cases and worse in others? How is it
formulated for multiparticle vertices?
6. What is the relation of the OZI rule to ideally
mixed nonets? Can the breaking be described as entirely due to
deviations from ideal mixing? This would allow the deviation angle
to play the role of a Cabibbo angle in a phenomenological
description with symmetric vertices and breaking only in mixing
angles for physical particles.

196

H. J. LIPKIN

7. A cascade of OZI-allowed transitions can produce a

forbidden transition. How are these higher-order transitions
suppressed?
S. How are the large suppression factors observed in
the new particles explained? If they are entirely due to the OZI
rule then the breaking must be much smaller than SU(3) breaking.

9. Are thresholds important? Higher-order transitions

which violate the OZI rule are possible for the and f' decays
via the physical KK state with both kaons on shell. But no such
physical states exist for the ~ and ~, decays, which are below the
threshold of the analogous DD state. Will the OZI rule be better
for the ~ and W' which cannot proceed by a cascade of allowed
on-mass-shell transitions than it is for the old particles or the
higher ~'s above the DD threshold?
5.2. Symmetry, Dynamics, Mixing and the KI -K 2 Analogy
There are two possible approaches to explaining the OZI
selection rules (5.1), symmetry and dynamics. All these rules
apply to processes involving mixed meson nonets, where the SU(3)
breaking and mixing of singlet and octet states plays a crucial
role. When the couplings and amplitudes for the processes (5.1)
are expressed in terms of the unbroken SU(3) singlet and octet
states, the two corresponding quantities are of the same order of
magnitude and nothing vanishes.

fof~rm

(5. Sa)

gf

nTT

(5.Sb)

gNNW 1

g]'fflWS

(5. Sc )

gNNf

gl'llif

(5. Sd)

gw OTT
1

9:- l nTT

cr(TT-p-+uJl n )

f of

cr(TT-p~ln)

f 0

cr(TT-P-+WSn)

(5.Se)

o"(TT-p~Sn)

(5.Sf)

Thus the selection rules (5.1) apparently arise from a

mysterious mixing mechanism which chooses the physical states to
be just the right linear combinations of the SU(3) singlet and

UNDERSTANDING THE NEW PARTICLES

197

octet eigenstates so that the singlet and octet contributions to

the processes (5.1) exactly cancel for one of the physical states.
There must be a better way to understand this.
Each cancellation of singlet-octet contributions required
for the selection rule depends upon two parameters not constrained
by SU(3): (1) the ratio of the singlet and octet couplings, (2)
the mixing angle of the meson nonet which determines the ratio of
singlet to octet in the physical meson. These two parameters
appear very different in character. A specific ratio of singletto-octet couplings suggests a higher symmetry beyond SU(3) which
relates amplitudes unrelated by SU(3). A mixing of singlet and
octet states to give physical states having different masses
requires SU(3) breaking.
In the symmetry approach one looks for a higher symmetry
which classifies the full nonet of mesons in a single supermultiplet and describes singlet and octet couplings by a single
coupling. The mixing angles of the physical states could have a
simple description if the symmetry-breaking mechanism breaks SU(3)
but conserves another subgroup of the higher symmetry which does
not commute with SU(3) and mixes SU(3) eigenstates. This other
subgroup would then predict new conservation laws not found in
SU(3) and could give the selection rules (5.1). One example is an
su(6) description6 which breaks both su(6) and SU(3) while conserving the subgroup SU(4~ X SU(2). But such symmetry approaches
have all had troubles. 14 , 2 The basic difficulty is that a
conservation law rigorously forbidding the transitions (5.1) must
either forbid the higher order transitions (5.3) or cancel them
exactly with other transitions. They cannot be forbidden without
also forbidding some of the experimentally observed amplitudes
(5.2). They cannot be cancelled exactly without introducing
additional channels whose intermediate states are exactly degenerate
with the kaon pair states, and such degeneracies do not exist.
The dynamical approach looks for dynamical models which
naturally give the mixing that decouples one eigenstate.
The neutral kaon system provides an instructive analogy
for describing the mixing and selection rule problem and illustrates
both the symmetry and dynamical approaches. We begin with the
eigenstates KO and KO of a symmetry, strangeness, that is broken
in the decay process. Symmetry breaking determines new eigenstates
which are linear combinations of the strangeness eigenstates and
could be described by a mixing angle
(5.10a)

H. J. LIPKIN

198

IK >

(5.1Ob)

Both the KO and EO states are coupled to the 2rr decay

channel. Consider the particular value of' the mix:i ng anp;le def'ined
by the relation
tan

<2rrITI KO)
(2rrITI KO)

(5.U)

Then we can write

<2rr IT IK O )

T cos

(2rrITI KO)

T sin

e
e

('i .12a)

(5.12b )

where T is def'ined by the relation

(5 .12c )

Substituting Eqs. (5.10) into Eqs. (5.12) we obtain the transition

matrix elements f'or the 2rr decay in the KA,KB basis

(2rrITIKA)

(2rrITI1S3)

(5.13a)

Thus if' both the KO and KO are coupled to the 2rr decay
mode, a mixing angle can always be f'ound which decouples one
state f'rom the 2rr system. The physical mixing angle is chosen by
diagonalization of' the mass matrix. If' experimentally one of' the
two kaon eigenstates is decoupled f'rom the 2rr system, one can turn
the question around. Instead of' asking why one of' the eigenstates
is decoupled f'rom the 2rr system one can ask why diagonalizing the
mass matrix chooses the particular state decoupled f'rom the 2rr
system to be an eigenstate, or why nature chooses the mixing
angle given by Eq. (5.11). We examine two possible answers to
this question, one based on symmetry and one based on dynamics.
1. Symmetry. If' an additional symmetry remains
unbroken it can give rise to the selection rule. For example, if'
CP is conserved in the kaon decays, then the eigenstates of' the
mass matrix must be eigenstates of' CP and the mixing angle A must

UNDERSTANDING THE NEW PARTICLES

199

be equal to 45. Since the 2n state is an eigenstate o~ CP the

2n decay is forbidden by CP conservation for the 1 kaon
eigenstate of CP with the wrong eigenvalue.
2. Dynamics. If the symmetry is broken by the decay
itself and there is no other symmetry breaking, the decoupling is
automatic. Assume, for example, that only the K 4 2n coupling
breaks strangeness conservation. Then all breaking and mixing
come from the loop diagram shown in Fig. 5.1a. The mass matrix is
a 2 y 2 matrix whose elements are given by computing this loop
diagram. But Eqs. (5.13) show that the loop diagram is diagonal
in the KA'~ basis, where all matrix elements vanish except the
diagonal ma~rix element for the state KA.
Thus we see that two completely independent mechanisms
naturally decouple one of the eigenstates of the neutral kaon
system from the 2n decay mode. If the conditions required for
either mechanism hold exactly the decay of one neutral kaon state
into two pions is forbidden. In the real world, both conditions
are very good approximations but not exact. CP is conserved to a
good approximation, but is still violated. The 2n decay mode is
the dominant strangness-violating decay mode of the neutral kaon
system but it is not the only decay mode. Additional loop
diagrams involving other states contribute to the mass matrix.
Thus, both conditions are slighbly violated and the neutral kaon
system contains a long-lived state which decays weakly into two
pions.
5.3

A Dynamical Model for Selection Rules at the SU(3) Level

The first attempt to explain the selection rule (5.1a)

took the point of view 23 of the Kl-K2 analogy and searched for a
dynamical symmetry breaking mechanism which naturally chose the
decoupled state as an eigenstate of the mass matrix. The approach
begins in the nonet symmetry limit where all nine states of the
vector or tensor nonet are degenerate and breaks the symmetry by
coupling to decay channels. This removes the degeneracy and mixes
the two states via transitions through decay cha~nels as shown by
the loop diagrams of Fig. 5.1. If a single loop diagram gives the
7T

f~
7T

Fig. 5.1.

Loop diagrams.

200

H. J. LIPKIN

only contribution to the symmetry breaking the diagonalization of

the mass matrix is trivial and leaves one of the two eigenstates
decoupled from that decay channel. Thus, just as the K2 is
decoupled from the 2n channel the m and f' are decoupled from the
on and nn channels if these are the only decay modes. The presence
of other decay modes complicates the diagonalization but their
effect is small if a single mode is dominant as in the case of the
2n mode in KO decay. The dominant decay mode then remains
decoupled from one of the two eigenstates to a very good
approximation.
For the ~ and f' decays the final states involving the
other members of the pseudoscalar and vector octets must be
considered together with the On and nn modes. However, in the
SU(3) symmetry limit these loop diagrams conserve SU(3) and cannot
mix singlet and octet states. Mixing can arise only from
symmetry breaking. The conventional description of symmetry
breaking assumes that the vertices in loop diagrams satisfy SU(3)
symmetry and breaks SU(3) by USing physical non-degenerate masses
for particles in the propagators. Because the dominant breaking
effect in the masses is the low mass of the pion relative to the
K and n, the dominant symmetry breaking effects in the loop
diagrams come from the On and nn channels. Thus, a dynamical
model in which the symmetry is broken by loop diagrams gives the
meson selection rules (5.la) and (5.lb). The particular model
considered for the vector mesons gave a natural suppression of the
m ~ On decay and a mass formula which fit the vector meson masses.
The loop diagram model does not give the baryon selection
rules (5.lc) and (5.ld) in any simple way. It also gives no
indication why the particular linear combinations of singlet and
octet states which satisfy the selection rules (5.la) and (5.lb)
should also happen to satisfy the baryon coupling selection rules
(5.lc) and (5.ld). Furthermore although the mass formula obtained
gave a non-trivial fit to the data, the very simple experimental
mass spectrum with p-w and f-A2 degeneracy and equal 0 - K* - ~
and A2 - K* - f' spacings is not obtained naturally in this model
and is fit by adjusting a free parameter. Thus, despite its
initial promise the loop diagram model does not provide a
satisfactory description of the selection rules.
5.4

The Quark Line Selection Rules

A crucial mystery at the SU(3) level is the difference

between the meson couplings (5.la) and (5.lb) and the baryon
couplings (5.lc) and (5.ld). Baryon selection rules having the
same SU(3) couplings as the meson selection rules (5.la) and (5.lb)
would decouple the ~ rather than the nucleon from the ~ and f',

UNDERSTANDING THE NEW PARTICLES

201

since the Land n both have zero hypercharge and occupy corresponding positions in the baryon and meson octets. The simple
unified statement of the selection rules (5.1) is in terms of
strangeness, rather than hypercharge.
But hyper charge is simple
at the SU(3) level, while strangeness is not, since it depends on
baryon number which is outside SU(3).
In the quark model the meson and baryon octets are very
different because mesons are quark-anti quark pairs while baryons
are three-quark states, and the very different SU(3) couplings
arise naturally. There does not seem to be any simple description
of the baryon selection rules (5.1c) and (5.1d) without invoking a
quark-like structure for the baryons in which they are composed of
three fundamental SU(3) triplets.
The Zweig-lizuka formulation
with quark diagrams provides a simple unified description of the
meson vertex selection rules (5.1a) and (5.1b), the baryon vertex
selection rules (5.1c) and (5.1d), the choice as eigenstates of
the mass matrix of just those particular linear combinations of
singlet and octet which satisfy the selection rules, and the
simple nonet mass spectrum.
The quark picture begins with a degenerate meson nonet
and breaks the nonet degeneracy by a mass difference between
strange and nonstrange quarks. This gives the so-called "ideal
mixing" which chooses as eigenstates those mixtures of singlet and
octet states corresponding to a pure strange quark-anti quark pair
and a pure nonstrange quark-anti quark pair, and gives a simple
mass formula with a mass splitting proportional to the number of
strange quarks. The selection rule is simply stated by drawing
quark line diagrams for the three-point vertex functions as in
Figs. 5.2 and postulating that only the connected diagrams Fig. 5.2a
and Fig. 5.2c are allowed while the disconnected diagrams
Figs. 5.2h and Fig. 5.2d are forbidden. All the couplings (5.1)
are forbidden since the ~ and fl both consist only of strange
quarks while the remaining particles consist only of nonstrange
quarks. Since the quantum numbers of the quark remain the same
on a given line, the strange quark lines begin and end on the
~ or fl and are completely disconnected from the nonstrange quark
lines. Thus, the couplings (5.1) are described by forbidden
diagrams, Figs. 5.2h and 5.2d.
However, exactly the same meson select~on rules (5.1a)
and (5.1b) are obtainable from other approaches 14 ,19,20 without
invoking connected and disconnected quark diagrams. A mathematically equivalent description for the three-meson coupling was first
proposed by Okubo as a nonet coupling ~nsatz. The same selection
rule was obtained by Alexander et al. 24 from the Levin-Frankfurt
approach 25 in which any hadron transition involves a change in the
state of only one active quark in the hadron while the remaining
quarks are spectators. If pion emission is described as a single

202

H. J. LIPKIN

)
a.

Allowed

b. Forbidden

<==--

c.

Allowed

Fig. 5.2.

d. Forbidden

Quark diagrams for three-point functions.

quark transition only nonstrange quarks can emit pions and

conserve isospin. Thus, a state like the or f' which contains
only strange quarks cannot decay by pion emission. This argument
can be stated more precisely in the language of PCAC and the
Melosh transformation. 2b The PCAC prescription relates the
amplitude for a pionic decay to the matrix elements of the axial
charge operator Q5 between the initial state and the final state
remaining after pion emission. The Melosh prescription postulates
that Q5 is a single quark operator. Since a single quark operator
cannot change a strange quark-anti quark pair into a nonstrange
pair,
(iQ5ip) == 0

(5.l4a)

(f'iQ-Sin) = 0

(5.l 4b)

The PCAC prescription then gives the selection rules (5.la) and (5.lb).

UNDERSTANDING THE NEW PARTICLES

203

None of these alternative approaches for the meson

selection rules (5.1a) and (5.1b) are applicable to the baryon
selection rules (5.1c) and (5.1d). The baryon couplings are much
more complicated than meson couplings at the phenomenological
SU(3) level. The octet three-meson coupling is constrained by
charge conjugation to be pure F and pure D depending on the charge
conjugation properties of the mesons. The singlet meson coupling
is forbidden when the octet coupling is pure F. For baryons no
such restrictions exist. The D/F ratio is a free parameter and the
singlet coupling is always allowed even if the octet is pure F.
Furthermore, an additional spin degree of freedom is present in
the baryon coupling. There are two independent couplings corresponding to helicity flip and nonflip at the baryon vertex.
For the three meson couplings the OZI rule follows
automatically from G conservation for those channels where charge
conjugation invariance requires the antisymmetric F type SU(3)
coupling. For example, the ~ ~ nn decay is forbidden while
~ ~ K*K is allowed. For cases like (5.1a) and (5.1b) where the
symmetric D type octet coupling is required and the singlet is
allowed one additional constraint on the couplings is needed to
obtain the selection rules; namely a particular value for the
ratio of the singlet octet couplings. The correct value of this
singlet to octet ratio is naturally obtained from the OkuboLevin-Frankfurt and Melosh approaches.
For the baryon case, the singlet to octet coupling ratio
needed to obtain the selection rules (5.1c) and (5.1d) depend upon
the D/F ratio. These ratios are different for the spin flip and
nonflip transitions. Thus very complicated constraints on the
couplings are needed in order to obtain the baryon selection rules.
The only simple description so far has been the Zweig-lizuka
formulation with connected and disconnected quark diagrams.
The baryon selection rules (5.1c) and (5.1d) thus provide
the best unambiguous tests of the ZI rule and its breaking. One
can envision a hierarchy in which the rule holds best when several
mechanisms reinforce one another. The smallest breaking effects
would occur in pionic meson transitions where the PCAC derivation
with eqs. (5.14) still holds while the other mechanisms which give
rise to the ZI rule in the other cases are broken. The strongest
breaking effects would occur in the baryon vertex and meson
vertices not involving pions would lie somewhere between these two
extremes.
The extension to more complicated vertices of the quarkline selection rules for th ee-point functions is not unique, as has
been pointed out by Okubo. l
The forbidden diagrams of Figs. 5.2b
and 5.2d can be characterized either as "disconnected diagrams,"
which allow external particles to be separated without breaking

204

H. J. LIPKIN

quark lines, or as ''hairpin diagrams" in which one external meson

has its two lines connected together rather than joining lines to
other particles. Forbidding all disconnected diagrams and
forbidding all hairpin diagrams are equivalent for three-point
functions. However, for four-point and higher functions there
are disconnected diagrams which are not hairpin diagrams, multiply
disconnected diagrams and diagrams with more than one hairpin, as
shown in Fig. 5.3. Without some fundamental theory for the OZI
rule it is not clear whether disconnected diagrams which are not
hairpin diagrams are forbidden just as much as hairpin diagrams,
and whether multiply disconnected diagrams or multiple hairpin
diagrams are forbidden more than corresponding single diagrams.

a.

Allowed

b.

c.

Crossed Pomeron diagram

d. Double hairpin diagram

Fig. 5.3.

Forbidden

e. Forbidden propagator
Quark diagrams for three-point functions.

UNDERSTANDING THE NEW PARTICLES

205

The four-point function of Fig. 5.3a is clearly allowed

and that of Fig. 5.3b is clearly forbidden by any version of the
OZI rule. But Figs. 5.Yc and 5.3d are ambiguous. The simplest
example of a disconnected diagram which is not a hairpin diagram
is shown in Fig. 5.3c and could describe the process nn ~ , or
elastic n scattering in the cross channel. The elastic amplitude
must have a Pomeron contribution, as in N scattering. This
"crossed Pomeron diagram" is forbidden in the Zweig-lizuka
formalism but is not forbidden if only hairpin diagrams are
forbidden.
Double hairpin diagrams like that of Fig. 5.3d can occur
when charm and strangeness are both present, because the two
hairpins could describe a strange and a charmed quark-antiquark
pair. One example of a process described by such a diagram is the
decay J/~ ~ nn in the charm model for the J/~. In this model all
decays of the J/~ into normal hadrons are forbidden by the OZI rule
in any formulation because one hairpin is required for the J/~.
Whether the two-hairpin decays are more forbidden or not has been
questioned. 1S The most recent experimental results on the nn
decay mode suggest that such decays are indeed more forbidden,27
but further data are necessary before any gendral conclusions can
be drawn.
One possible approach to the generalization of the OZI
rule to more complicated vertices is to build everything from
three-point functions which satisfy the OZI rule. Inconsistencies
arise in this approach because of the higher order paradox in which
combinations of 0ZI-allowed transitions can produce an OZI-forbidden
transition, as in Eqs. (5.3). There is also the question of
possible OZI violations in the propagators of particles appearing
as internal lines in the diagrams, as indicated in Fig. 5.3e.
These points are discussed in detail below.
5.5

The Higher Order and Unitarity Paradoxes

We now examine in more detail the violation of the OZI

rule by the transitions (5.3) in which two OZI-conserving
amplitudes (5.'2) combine to produce an OZI-violating transition.
All these transitions are from an initial state containing only
strange quarks to a final state containing only nonstrange quarks
via the intermediate state of a kaon pair. The kaon plays a
crucial ambivalent role. Since it contains one strange and one
nonstrange quark, it couples equally to strange and non strange
systems and can go either way. A kaon pair state contains one
strange and one non strange quark-anti quark pair. It can therefore
be created from a strange pair by the creation of a non-strange
pair or vice versa. The kaon pair state thus links two kinds of
states between which transitions are forbidden by the OZI rule.

206

H. J. LIPKIN

The quark diagram Fig. 5.4 for the forbidden transition (5.3b)
illustrates the essential features of the paradox. Viewed as a
single topological diagram it is indeed disconnected and can be
separated into two disconnected hairpin diagrams. But when it is
separated into two individual transitions, each half is connected.
Connecting the two diagrams together results in a topological
disconnected diagram because of the twist in the quark and antiquark lines in the kaon intermediate state.

Fig. 5.4. OZI Violation via twisted diagram

Thus, to save the OZI rule the connection of allowed

diagrams by a "twisted propagator" must somehow be forbidden. But
a twisted propagator has physical meaning only if there is additional
information in a kaon pair state to specify "which way it is twisted";
i.e., whether it originally came from a strange or a nonstrange
system. Some memory of the origin of the pair is necessary to
prevent the non strange decay of a pair which originated in a strange
system. But a physical kaon pair state has no such memory . A kaon
pair produced from a nonstrange system is indistinguishable from a
pair produced from a strange sy~tem.
The transition (5.3) thus must exist in any consistent
scheme which incorporates the OZI rule. Saving the rule requires
additional transitions via other intermediate states which exactly
cancel these amplitudes. Such cancellations do in fact occur in
dual resonance models where twists in diagrams denote changes of
the relative phase of the contributions of such intermediate
states. But the degeneracy requirement discussed above poses
difficulties. This is discussed in detail below.
The same higher order paradox appears in an S-matrix
formulation as unitarity violation. Consider for example D wave
nn and KK scattering in the vicinity of the f' pole treated as a
system with two coupled channels. The OZI selection rule (5.lb)
requires the f' to appear as a pole only in the KK channel but not
in the nn channel. But the eigenstates of the 2 X 2 S matrix are
not the nn and KK states but mixtures of the two, because the
nn ~ KK transition amplitude (5.2d) does not vanish. Thus, both

207

UNDERSTANDING THE NEW PARTICLES

eigenstates o~ the S matrix contain a rtrt component and any pole

appearing in the S matrix must have a nonvanishing coupling to the
rtrt channel. Complete decouplin o~ a pole ~rom the rtrt channel is
possible only when the rtrt and KK channels are completely decoupled
~rom one another.
equation

This can be seen explicitly by writing the unitarity

Im(KK:ITlrtrt)

(KK:I T t

lKiD (KK:I T Irtrt)

+ (KK:ITtlrtrt)(rtrtITlrtrt)
Since there are only two channels the unitarity sum has only two
terms. I~ the OZI rule holds and the ~, pole is decoupled ~rom
the rtrt channel, the ~, pole appears only in a single term in
Eq. (5.15), namely the ~irst term in the right-hand side. Thus the
OZI rule is inconsistent with unitarity in this two-channel model.
The way out o~ this paradox is to include more than the
rtrt and KK channels. Additional channels can introduce new terms
in the unitarity sum o~ Eq. (5.15) which can cancel the term with
the ~, pole and enable the decoupling o~ the ~, ~rom the rtrt channel
without violating unitarity. Again the paradox is resolved by
canceling the transitions via some other set o~ intermediate states.
Some symmetry scheme or dynamical model is needed to choose
which additional intermediate states cancel the KK: contribution.
There are several possibilities. SU(3) symmetry suggests that the
~ull pseudoscalar octet be included with the additional nn
intermediate state. Nonet symmetry requires the ~'n' and ~,
states as well. su(6) symmetry suggests that vector and pseudoscalar
mesons be treated together with the inclusion o~ intermediate states
involving vector mesons. Duality and exchange degeneracy suggest
that the vector and tensor mesons which lie on degenerate
trajectories must be included together. Exactly how these con~lict
ing suggestions can be resolved is an open question.
5.6

A Simple SU(3) Model

We now show how the higher order and unitarity paradoxes

can be resolved in a simple way in the ~ramework o~ SU(3)
symmetry by including the ~ channel together with the KK: channel.
The model is not relevant to the physical particles, but the manner
in which it avoids the di~~iculty o~ the higher order transition
(5.3b) and the associated problems o~ the unitarity o~ the
S-matrix is instructive. In the nonet symmetry limit there is
always a particular linear combination o~ the two isoscalar tensor

208

H. J. LIPKIN

mesons for which the couplings of the two components to the 2rr
channel cancel one another exactly. We consider a model in which
a small SU(3) symmetry breaking with appropriate properties splits
the masses and leaves the decoupled states as an eigenstate.
We denote by fl and f8 the two isoscalar tensor mesons
classified in the SU(3) symmetry limit in the singlet and octet
representations. We assume an unmixed octet of pseudoscalar
mesons. There are three possible two-pseudoscalar decay modes for
these tensor mesons namely, rrrr, KK and nne The branching ratios
for the fl and f8 into these three decay modes are determined
uniquely by SU(3), but the relative strengths of the fl and f8
couplings are not determined
(5.16a)
(5 .16b)
where 71 and 18 are reduced total widths for the fl and f8 with
phase space factors removed
(5.17a)

'Yl

(5.17b)
If 11 ~ 18' the states fl and f8 have different lifetimes
and cannot be mixed, as mixed states woula not have simple
exponential decays. To allow mixing of fl and f8 in the symmetry
limit we set their widths equal

(5.18 )
The states fl and f8 are now degenerate and any linear combinations
can be chosen to give a basis of states with simple exponential
decays. We choose a basis in which one state is completely
decoupled from the 2rr channel. From Eqs. (5.16) and (5.18) this
basis is
If) =

~ l~

If l ) +

~ l~

If8)

If')=

~ l~

I f l )-

~ l~

If8)

(5.19a)

UNDERSTANDING THE NEW PARTICLES

209

We now assume that the symmetry breaking chooses these

states as eigenstates o~ the mass matrix ~or some mysterious
reason. The partial widths ~or the various decays o~ these states
are given by

<1f1f I~)

"'(39/40)'1

(5.203.)

<nnl~)

=-"'(9/520)'1

(5.2Ob )

(KKI~)

= ",(i/130)'Y

(5. 2Oc)

(1fJ1 I~,)=

(5. 2la)

<nnl~')= "'(4/13)'1

(5.21b)

"'(9/13)'1.

(5.21c)

(KKI~')=

The state denoted by ~, is decoupled ~rom the 2J1 channel

by construction. But the mixing angle described by Eqs. (5.19) is
very di~~erent ~rom the ideal mixing angle o~ the quark model.
Note, however, that Eqs. (5.20) and (5.21) lead to the result

<(3KK + 27111) I~)

(5.22a)

<(2KK - 3Tf1l)' ~

(5.22b )

I )

O.

(5.23)

Equation (5.23) shows that the higher order transition

amplitude (5.3b) is canceled exactly by the analogous transition
via the nn intermediate state. Equations (5.22a) and (5.22b) show
that the ~ and ~, are coupled to two orthogonal linear combinations
of the KK and nn channels. The particular linear combination
(5.20a) which is decoupled from the f is the eigenstate of the
S-matrix whose amplitude has the f' pole. This pole does not
appear in the other eigenstates of the S-matrix, which are
decoupled from the f', namely, the 2J1 channel. Thus the selection
rule (5.la) can be rigorous without dif~iculties from the higher
order transitions (5.3b) or the unitarity of the S matrix.
We now consider the conventional quark model formulation
using ideal mixing and the OZ1 rule. This case is very di~~erent
from the above SU(3) treatment where the cancellation (5.23) of the
higher order transition amplitude (5.3b) depends crucially on the
description o~ the n as a member of an unmixed octet. No such
cancellation occurs in the case o~ an ideally mixed pseudoscalar
nonet with couplings given by the OZ1 rule. We denote the
pseudoscalar state consisting o~ a nonstrange quark-anti quark pair

H. J. LIPKIN

210

by nn and the state of a strange quark-anti quark pair by ns.

OZI rule then immediately gives
<n n If'>
nn
<n n If)
n n

= <n ss
n If> = <n n If> = <n n If'>
sn
sn

The

(5. 24a)

\f2 <n n If'> = ~ <nnlf>

s s

Thus if the coupling of the tensor mesons to the pseudoscalar nonet is described by the OZI rule with ideal mixing, there
is only one pseudoscalar channel coupled to both f and f', the KK
channel, and no additional channel is available to cancel the
transition (5.3b).
For a more realistic treatment we consider pseudoscalars
which are not ideally mixed,
n'
n

sin A
p +n s
p

(5. 25a)

-n n sine p +n s cose p

(5. 25b)

cos e

where ep is the angle of deviation from ideal mixing.

basis Eqs. (5.24a) and (5.24b) become

(n 'n' If>
2

cos e

<nnlf>
2/'1

Sln "

\f2<n'n' If'>
sin2e
p

sinAcose

In this

.J2sin8cost:)

>}.-3

<nn If)

Equations (5.26) give the transition amplitudes for the case where
the couplings to ideally mixed states satisfy SU(3) symmetry and
the OZI rule, but the physical pseudoscalar states are not
ideally mixed.

UNDERSTANDING THE NEW PARTICLES

211

From Eqs. (5.24b)

(5.2'7b)

(5.2'7c)
If the result (5.2'7) is summed over all channels with
equal weighting the total contribution of nand n' decay modes
vanishes, as expected from Eq. (5.24a). A finite contribution
appears under the more realistic assumption that the nn channel is
dominant and others are neglected. However, this contribution has
the same phase as the contribution from the KK intermediate state
and cannot produce a cancellation. It is also much smaller than
the KK contribution.
5.7

The Selection Rule in su(6)w Symmetry

An early derivation of the selection rule forbidding the

This selection rule
holds in any model which satisfies su(6)W. It is therefore of
interest to examine the higher order transition (3a) and see how
su(6)w operates in this case. In SU(6)W the K*(890) is in the same
super-multiplet with the kaon, and higher order transitions via all
possible intermediate states involving one or two K* mesons must
also be considered. In the approximation where K and K* are
degenerate and all vertices are related by Su(6)w the contributions
from the different K and K* intermediate states all cancel and the
decay ~ ~ On is still forbidden. This cancellation is simply
described by a conservation law. In the decay (5.la) the outgoing
P and the initial ~ must all be in the same polarization state with
S = l, since angular momentum and parity conservation forbids the
tansition for the states with S = O. The transition from the
initial state to the intermediat~ state of two strange mesons
conserves separately the total W spins of strange and non strange
quarks. The initial ~ state has strange W-spin I and nonstrange
W spin zero. Thus each strange meson in the intermediate state is
a coherent linear superposition of a K and a K* which is an
eigenstate of the z component of the spin of the strange quark or
anti quark, with the eigenvalue required to conserve the strange W
spin. The spin directions of the two strange quarks are parallel.
Thus, they cannot annihilate and conserve strange quark W spin.
~ ~ On decay was based on su(6)W symmetry.6,28

H. J. LIPKIN

212

The su(6)W selection rule is thus ~ollows simply ~rom the conservation o~ the z component o~ the strange quark spin.
However, the transition (5.3a) is ~orbidden in a much
simpler way by SU(6)W' which also illustrates a basic weakness in
su(6)W' which holds only ~or colinear Erocesses. In su(6)w all
momenta are in the z direction and a KK state must have S = J = O.
Thus the Sz = 0 P2larization state o~ the initial ~
z
z
decays into the KK channel and the Sz = l state decays into the
Orr channel and angular momentum conservation ~orbids transitions
between the two states. But this argument holds ~or the transition
(5.3a) only i~ the momentum o~ the kaon pair in the intermediate
state is in the same direction as the momentum o~ the ~inal Orr
state. Consider ~or example the decay o~ a ~ in the state with
Sx = 0 into a orr ~inal state with momentum in the_z direction. The
transition o~ a ~ into an W via an intermediate KK state with
momenta in the x direction is allowed by Su(6)W because the ~ and
W states with Sx = 0 are both states with W = 0 with respect to the
x axis. However, the state o~ the W with Sx = 0 is a linear
combination o~ states with Sz = l ~or which the Orr decay in the
z direction is allowed.
The Su(6)W argument does not hold ~or the ~, decay,
because the outgoing pions are spinless and there can be no
correlation between the quark spins o~ the ~inal state and the
quark spins in the initial state. In the decay (5.lb) a
component in the initial ~, wave ~unction has the quark and
anti quark spins anti parallel , and there~ore has W = O. They are
allowed to annihilate without any angular momentum trans~er, and
the transition ~rom the W = 0 component o~ the ~inal two-pion
state is allowed by su(6)w. Thus, Su(6)W cannot be used to give a
general derivation o~ the OZI rule.
5.8

Ideal Mixing and Symmetry Cancellations

The peculiar role o~ ideal mixing and higher symmetries

in the OZI rule is illustrated by the ~ollowing example:
Consider the decay o~ a high K*- resonance into three
kaons via an intermediate nonstrange resonance MO

*-

-+ K-Mo -+ K - KK,

(5.28 )

where MO is a resonance like the 00, W, ~ or A2 which consists only

o~ nonstrange quarks.
There are two quark line diagrams ~or this decay shown
in Fig. 5.5, using the
and rill components o~ ~. The pp diagram

pp

UNDERSTANDING THE NEW PARTICLES

213

is connected, obeys the OZI rule and leads to the final state
The nn diagram is disconnected, violates the OZI rule and
leads to the final state K-KOKO. Thus if the OZI rule holds,

K-K~-.

(K* ~ K-M o ~ K-K+K-)

is allowed

(K* ~ K-Mo ~ K-K~)

is forbidden.

(5. 29a)

K*--K-MOLKR

Fig. 5.5a. Allowed K*into charged mode.

Fig. 5.5b. Forbidden

decay into
neutral mode.

r--

But if the resonance MO has a definite isospin, the two

transitions (5.29a) and (5.29b) must be equal by isospin invariance.
Contradictions between the OZI rule and isospin invariance are
avoided if the nonstrange meson spectrum consists of degenerate
isospin doublets, like P and W or f and A2. In that case the
transition (5.29a) proceeds via the particular coherent linear
combination of isovector and isoscalar particles which has the
quark constitution pp. The OZI rule is thus intimately related to
the existence of the isospin doublets found in ideally mixed
nonets.

H. J. LIPKIN

214

I~ the MO in the transitions (5.29) is not a member o~ an

isospin doublet, the OZI rule is inconsistent with isospin
invariance. This is the case i~ ~ is a nO, which has no degenerate
isoscalar partner. Although the nO cannot appear as a physical
resonance in the reactions (5.29) because o~ its low mass, it can
appear as an exchanged particle in the analogous two-body
scattering reactions

K+ +K

-+

-*0
K*o +K

(5.3Oa )

+K*-

(5.3Ob)

K*o + K*-

(5.3Oc)

- -+ K*+

K+ +K

KO + K-

-+

The quark diagrams

~or

these reactions are shown in Fig.

DC=

K-

~c=

A K"-

K- A

A K"o

KO
K-

5.6

K"
K"-

Fig. 5.6.

Reactions allowed by pion exchange.

The charge exchange reaction (5.30a) is clearly allowed

by the OZI rule and can go by pion exchange. The amplitudes ~or
the pion exchange contribution to the reactions (5.3Ob) and (5.30c)
are uniquely related to the charge exchange amplitude (5.30a) by
isospin invariance. But the reaction (5.3Ob) is allowed by the
OZI rule and the reaction (5.3Oc) is ~orbidden when only nonstrange
quark exchange is considered. (The reaction (5.30c) is allowed by
AX exchange but this is irrelevant to the present argument). The
OZI rule could be saved ~rom inconsistency with isospin invariance
i~ a contribution ~rom isoscalar exchange degenerate with pion
exchange cancelled the pion exchange contribution to the reaction
(5.3Oc). But no such isoscalar exists. Thus violations o~ the OZI
rule might be expected in procesges where pseudoscalar exchange
plays a dominant role.

UNDERSTANDING THE NEW PARTICLES

215

5.9 Cancellations and Degeneracies in Quark Line Models

Figure 5.7 shows the essential piece of the diagrams of
Figures 5.5 and 5.6 which break the OZI rule, a transition between
a pp and nn pair. This diagram exists in the propagator of any
neutral isovector meson and will lead to OZI breaking unless it is
cancelled by a contribution from a degenerate isoscalar partner.
The diagrams of Fig. 5.4 describe the higher order transitions (5.3)
which violate the OZI rule as the connection of two allowed
diagrams by a twisted pair of lines. The essential piece of this
diagram which breaks the OZI rule is shown in Fig. 5.8. This is a
transition which interchanges the quark and antiquark lines.

IF YOU LIKE p,n,A USE THIS SIDE UP

dn 30lS SIHl 3Sn s'p'n 3>lll no)" .:II

Fig. 5.7. OZI-Violating diagram required by isospin.

Fig. 5.8. OZI-Violating diagram required by charge conjugation.

The two basic OZI-violating diagrams, Fig. 5.7 and
Fig. 5.8 have a very similar structure. Figure 5.7 describes a
transition between two states related by isospin. It occurs
naturally in the propagator of any particle which is an isospin
eigenstate. Figure 5.8 describes a transition which interchanges
quark and anti quark and occurs naturally in the propagator of any
particle which is an eigenstate of charge conjugation. The OZI
violation implied by the diagram of Fig. 5.7 can be avoided by an
additional degeneracy of isoscalar and isovector particles. This
allows the choice of a basis of states which are not isospin
eigenstates and whose propagators do not include the diagram of

216

H. J. LIPKIN

Fig. 5.7. Similarly, the OZI violation implied by Fig. 5.8 can be
avoided by an additional degeneracy of particle states which are
even and odd under charge conjugation. This allows the choice of
a basis which are not eigenstates of charge conjugation and whose
propagators do not include the diagram of Fig. 5.8. For the case
where the quark and anti quark in Fig. 5.8 do not have the same
internal symmetry quantum numbers, the relevant transformation is
not charge conjugation but an appropriate combination like G parity
of charge conjugation and an internal symmetry transformation. The
conclusions are the same.
We can now specify the additional degeneracies essential
for the validity of the OZI rule in higher order in models described
by quark line diagrams. The states described by the left-hand and
right-hand sides of the diagrams of Fig. 5.7 and Fig. 5.8 must be
physical eigenstates which can propagate unchanged and in particular
can avoid undergoing transitions indicated by Fig. 5.7 and Fig. 5.8.
These states are eigenstates of quark number having a well-defined
quark composition (either pp or nn, but not a linear combination of
them) and are linear combinations of states even and odd under
charge conjugation (either pp or pp but not the linear combination
of them which is a charge conjugation eigenstate). These states
required by the OZI rule are not eigenstates of SU(3) and its
isospin subgroup nor of charge conjugation. They can be physical
eigenstates only if additional degeneracies are present beyond
those imposed by these symmetries.
There must be ideal mixing of the SU(3) singlet and octet
states so that the AA state remains an eigenstate and the pp and
nn eigenstates which go into one another under isospin transformations are degenerate. When processes are described in terms of the
isospin eigenstates as in Fig. 5.5 the amplitude for the forbidden
diagram vanishes because of a cancellation between the contributions
involving degenerate isoscalar and isovector states.
Charge conjugation degenerate doublets are required to
eliminate the OZI violation due to twisted diagrams like Fig. 5.8.
In duality and dual resonance models this degeneracy appears in
Regge trajectories, rather than in individual particle states as
exchange degeneracy of trajectories having opposite signature. In
these formulations the two states on the left-hand and right-hand
sides of Fig. 5.8 represent two different linear combinations of
even signature and odd signature trajectories with equal magnitude
and opposite phase. When the OZI-violating higher-order
transitions (5.3) are described in the conventional basis using
states which are eigenstates of charge conjugation the violating
diagrams cancel one another in pairs involving states which behave
oppositely under charge conjugation.

UNDERSTANDING THE NEW PARTICLES

217

The peculiar relation between ideal mixing, exchange

degeneracy and the OZI rule was noted in early treatments o~
duality using ~inite energy sum rUles. 29 Two possible mechanisms
~or the breakdown o~ the necessary cancellations are immediately
evident upon closer examination o~ the diagrams o~ Fig. 5.7 and
Fig. 5.8.
The diagram o~ Fig. 5.7 will occur and break the OZI rule
whenever a propagator appears ~or a state which is not ideally mixed,
such as a pseudoscalar meson. One can expect the OZI rule to be
violated in processes where there is a strong contribution ~rom
pseudoscalar exchange.
The cancellation o~ the diagram o~ Fig. 5.8 must break
down because charge conjugation doubling exists only ~or Regge
trajectories and not ~or individual physical states. 30 Thus, even
i~ exchange degeneracy is exact the OZI-violating diagram o~ Fig.
5.8 is cancelled only in the kinematic region where Reggeization
is a good approximation; i.e., where the scattering amplitude is
well described by a Regge exchange rather than by one or two
resonances. 31 This is clearly not the case ~or the higher-order
transitions (5.3a) and (5.3b) where the mass is above the
threshold ~or the intermediate KK state but below threshold ~or all
other states on the kaon trajectory and its exchange degenerate
partner. In general, one might say that contributions ~rom high
momentum intermediate states could be described in the Regge
approximation and the desired cancellations ~rom exchange degenerate
pairs could occur. But at low momenta, where the large mass
di~~erence between individual resonances on exchange degenerate
trajectories is signi~icant, such cancellations should not be
expected.
5.10

Quantitative Estimate

o~

OZI Violation

Let us now attempt to estimate the violation o~ the OZI

selection rule (5.1b) resulting ~rom the higher order transition
(5.3b). We consider the ~-~, system analogous to the Kl-K2 system
and diagonalize a 2 X 2 mass matrix. We assume that the dominant
portion o~ the mass splitting comes not ~rom the loop diagram but
~rom a quark mass term.
We there~ore use the ideally mixed basis
in which the quark mass term is diagonal. The loop diagram
contributes both a real part and an imaginary part to the mass
matrix.
The real part is dominated by high momenta. There is no
considering only th contribution o~ the KK state
rather than all states on the K and K Regge trajectories.
Exchange degeneracy and Regge arguments suggest some kind o~
cancellation in these diagrams. But it is very di~~icult to
justi~ication ~or

218

H. J. LIPKIN

estimate these cancellations quantitatively. One possibility is to

consider the entire set of states on Regge trajectories and use
duality and dual resonance models in order to calculate the
contributions. Such calculations are beyond the scope of this paper.
The imaginary part, however, is dominated by the f and f'
poles. It is therefore reasonable as a first approximation to
consider only the KK intermediate state and neglect the contributions
of higher strange meson resonances. Estimates of the imaginary part
of the mass matrix are obtained by using the experimental partial
widths for the decays. We therefore neglect the real part, which
we cannot calculate anyway, and calculate the contribution of the
imaginary part to the OZI violation. This gives a lower bound
since the contribution of the real part cannot cancel that of the
imaginary part.
In this formulation the violation of the OZI rule comes
about as a result of a deviation from ideal mixing produced by the
loop diagram. The relevant parameter which characterizes the mixing
is the ratio of half the width of the f' to the f-f' mass splitting.
For r(f I ~ KK) = 40 MeV and
I = 240 MeV this ratio is small
and consistent with the small observed violation. This also implies
that we can treat the deviation from ideal mixing as a first-order
perturbation. Let us write

mr

mr

If I) = cosSf,lfi> + Sinef,lfI)
where the subscript I denotes the ideally mixed states and ef' is
the deviation of the f' mixing angle from ideal mixing. This is
not necessarily equal to the f mixing angle because the eigenstates
of a complex mass matrix are not necessarily orthogonal. However,
only 9f, is needed to calculate the OZI rule violation. In firstorder perturbation theory sinAf' is given by

sine f,

~<fITIKK)<KKITlf')PF(f')

Ii(Mf

, -

Mf

(5.3la)

where PF(f ' ) denotes the density of final KK states at the mass of
the fl. This can be expressed in terms of the experimental width
of the f' and the ratio of f and f' transition matrix elements
r(f' ~ KK)
2{Mf, - Mf )

(5.3lb)

UNDERSTANDING THE NEW PARTICLES

219

The amplitude of the OZI violating transition is then given by

r(f' ... KK)
2(Mf'- Mf )

(fiTIKK)(2niTif)
(f'ITIKK)

To evaluate the expressions (5.31) we introduce the

SU(3) relation between the transition amplitudes

We now obtain
r(f' ... KK)
2V2(Mf

Mf )

where the d-wave phase space factor (p /Pk)5 is introduced to

account for the difference between pioA and kaon momenta in the
decay of the fl. Substituting the experimental values
r(f' ... KK) = 40 MeV, Mf, = 1514 MeV, Mf = 1270 MeV we obtain
sine f I = 0.06 and

0.02.

This result is in qualitative agreement with experiment. 32

Obtaining a better approximation is difficult because there are too
many uncertainties in the additional effects which must be taken
into account. In addition to the contribution from the real part
there are additional contributions from channels like nn, nln l
which have large uncertainties because of the deviation from ideal
mlxlng. In the limit of ideal mixing these channels give no
contribution as shown by Eqs. (5.24) since there is no state which
is coupled both to the f and to the fl. However, the physical
pseudoscalar mesons are approximately equal mixtures of nand n
n
s

220

H. J. LIPKIN

and all three physical states contribute to the transition. The

three contributions cancel exactly in the nonet symmetry limit
where ~ and ~, are degenerate and the ratio of the singlet and
octet couplings is given by the OZI rule. For the physical
nondegenerate and nonideally mixed states the cancellation no longer
occurs and the magnitude of the contribution is very sensitive to
the mixing angle and to possible changes in the singlet to octet
ratio. Thus, there does not seem to be a serious possibility of
obtaining a better approximation than Eq. (5.34).
An alternative approach to including the higher order
transition (5.3b) is to use an S-matrix formalism and calculate
unitarity corrections. In the most naive approximation results
similar to (5.34) are obtained. However, attempts to include more
channels run into the same difficulties discussed above for the
~ and ~'.
Our analysis of the f' ~ nn decay suggests that the
observed violation of the OZI rule is due to the higher order
transition via the open KK channel, and that there is no effective
mechanism available for cancelling the contribution of this channel.
The effect is small because it is characterized by a small parameter,
the ratio of half the width of the f' to the f-f' mass difference.
Whether the smallness of this parameter has any deep theoretical
significance is not clear at this point. Two fundamental quantities
having the dimensions of mass appear in this ratio, the characteristic width of strong decays and the mass difference between strange
and nonstrance particles or between the strange and nonstrange
quark. Both these quantities are generally considered to be of the
order of 100 MeV and the ratio of the two is then of order unity.
However, in the particular case of the f' decay there are
several factors of 2 present which conspire to provide a factor which
is an order of magnitude. Two factors of 2 arise because the
relevant parameters are half of the width and twice the energy
difference between strange and non strange quarks. The width of the
f' is 40 MeV rather than 100 MeV. As long as there is no
fundamental theory which predicts the ratio of the strangeness mass
difference to strong decay widths, there can be no explanation for
why the numerical value of the parameter which characterizes OZI
violations is small in this particular case.
It is amusing that the selection rule holds both in the
limit of very small and very large values for this ratio of the
widths to the mass splitting. In the other extreme case of the large
value the mixing of the two states is completely dominated by the
decay process and leads to the decoupling of one state via the
Kl-K2 mechanism.

UNDERSTANDING THE NEW PARTICLES

221

In intermediate cases, where the mlxlng is not dominated

by the decay process the KI -K2 mechanism still has some effect in
reducing OZI violation from hlgher-order transitions. In the
above analysis of mixing produced via the intermediate KK state in
f' decay contributions from the intermediate nn state were
neglected. In first order these contributions vanish because the
unperturbed f' does not couple to the nn channel. This
approximation is justified in this case by the small magnitude of
the results (3.31-3.34). But when the mixing is larger and higher
order effects must be considered the contribution of the nn
intermediate state introduces an effective "restoring force"
opposing the mixing by the KK intermediate state and opposing the
increased violation of the OZI rule.
A qualitative estimate of this restoring force effect is
obtainable by introducing all the two-pseudoscalar meson intermediate
states into the calculation. We assume that the 2 X 2 mass matrix
M has the form:
M=t\m+L

+L

where Am is a quark mass term diagonal in the ideal mlxlng basis as

in the naive mixing model, Ls represents the contribution of the
loop diagram of Fig. 5.1c but with all octet pseudoscalar mesons
included in the intermediate state and full SU(3) symmetry, and
Ln denotes an additional contribution from the nn intermediate
state resulting from SU(3) symmetry breaking. We do not have
values for the strengths of these terms from first principles,
but give each one a strength parameter and see how the OZI rule is
affected by their variation. In particular, we shall see that the
Ln term indeed has the effect of a restoring force reducing OZI
violation.
If Ls = Ln = 0, the only contribution to the mass matrix
comes from 6m and gives ideal mixing and the OZI rule. If Ls
0
but
is kept zero; i.e. SU(3) symmetry in the loop diagram,
there is no longer ideal mixing, and the OZI violating decay
f' 4 nn occurs. However, if 6m and Ls are fixed and the symmetry
breaking term'Ln is turned on, the OZI violation is reduced by
this "restoring force". Table 5.1 shows values of the OZI
violation, expressed as the ratio of the forbidden and allowed
decay rates, r(f' 4 nn)/(r(f 4 nn) for different values of Ln and
Ls. The value of Ls is given in arbitrary units, and
is given
in units of the nn contribution to Ls; i.e. Ln = 1 means that the
contributions from the nn intermediate state in Ln and Ls are equal.
Table 5.1 shows that restoring force effects can be quite
appreciable. The significant quantity is the change in
r(f' 4 nn)/(r(f 4 nn) with increasing Ln. For example, when the
loop diagram contribution is sufficiently large to give a 31%

Ln

Ln

222

H. J. LIPKIN

violation of the OZI rule the addition of a symmetry breaking nn

loop of equal magnitude to theSU(3) symmetric nn contribution
reduces the violation to 7%.
Table 5.1
Effect of Kl - K2 Restoring Force Mechanism
Values of 1'(1" -+ nn )/1"'(1' -+ nn) as Functions of Ls and Ln

1/2

3/2

.04-

.03

.02

.016

.012

.11

.04

.03

.02

.18

.06
.09

.05

.035

.02

.31

.14

.r:J7

.04-

.03

5.11

Experimental Tests of the OZI Rule

The OZI rule has been found to be in qualitative

agreement with experiment for selection rules forbidding the
production of the ~ and f' mesons.
Two types of further experimental information are needed
as a guide to theoretical understanding of the underlying
dynamics: 1) Quantitative results on the magnitude of the OZIviolating transitions with systematic comparisons of different
processes, 2) Tests of the OZI rule for nonets like the pseucoscalar no net which are not ideally mixed.
If production of the and f' from nonstrange systems
arises primarily from a small admixture of a nonstrange quarkanti quark pair into the wave function, as in the example of
section 5.10, then all OZI-violating transitions will be
expressible 1 9 in terms of the angle ~f' and the analogous angle
e~

2
gN'N

-2-=

1mw

0'( nN -+
O'(nN -+

X)
WX)

O'(NN
O'(NN

-+
-+

X)
WX)

223

UNDERSTANDING THE NEW PARTICLES

gfl:rr:rr
-2-gf:rr:rr

9Wfl

= -2-9Wf

O'(:rrN -+ fIX)
O'(:rrN -+ fX)

O'(NN-+f'X)

a(NN-+fX)

O'(K-p-+Yf I )back
O'(K p-+Yf )back

tan 9 f

where X denotes any single or multi particle hadron state which

contains no strange particles, Y denotes any neutral hyperon or
hyperon resonance and the subscript back denotes backward neutral
meson production by baryon exchange.
If on the other hand, the violation in a given process
comes from a specific higher order transition appropriate for that
particular process, the relations (5.36) will not hold. One example
of such a higher order transition which has been considered in
detail is 30,5 l
(5.37a)

The unitarity paradox has been formulated for the box diagrams
corresponding to these reactions and has been shown to be related
by SU(3) symmetry to the reactions
(5.38a)

The transition (5.37a) is forbidden by the OZI rule; the

transition (5.38a) is forbidden in standard models because it
involves an exotic exchange. Yet both steps in these second-order
processes are allowed and are simply related respectively to the
corresponding allowed processes (5.37b) and (5.38b). The relations
between corresponding allowed and forbidden processes have been
written as a unitarity sum involving the intermediate states in the
box diagram

224

H. J. LIPKIN

where A, Band i denote initial, final and intermediate states for

any of the reactions (5.37) and (5.38). Note that for any
intermediate state i, the right-hand sides of Eq. (5.39) are either
equal for corresponding allowed and forbidden processes or differ
only by a symmetry coefficient. The paradox is resolved by
requiring cancelations in the unitarity sum for the forbidden
processes, which are seen in quark-line formations to be represented
by twisted diagrams analogous to Fig. 5.4.
Without a detailed model for all possible intermediate
states, it is impossible to estimate how good the cancellations
are. However, SU(3) symmetry has been used to relate the two
forbidden processes, so that OZ1 breaking can be predicted with
experimental data on exotic exchange amplitudes used an input. 30
Estimates of OZ1 violation have also been made by this box-diagram
mechanism for low energies where the lowest mass intermediate state
can be assumed to be dominant without cancellations from other
channels.
Experimental results indicate that in some cases the mixing
mechanism giving rise to Eqs. (5.36) are valid, while in other cases
the box diagram description (5.37) may hold. 32 ,33 Further
experiments would be of great interest.
Interference experiments have been useful in detecting
OZI-violating transitions, since these measure a small amplitude
directly, rather than the square. The forbidden f' production
amplitude has been observed as interference with the tail of the
allowed f amplitude in the reaction3 2

(5. 4 0a)
where MO denotes either the f or f'. Similar f-f' interference
effects between the tail of an allowed f peak and a forbidden f'
amplitude could be seen in the reactions

(5. 4 Ob )
(5.40c)
Comparison of the reactions (5.40a) and (5.4Ob) would be interesting
since they are very similar, with the roles of the production and
decay of the f' interchanged. The production of the f' is forbidden
by the OZI rule in reaction (5.40a) and the decay allowed, while
the decay is forbidden and the production allowed in (5.4Ob). The
reaction (5.4Oc) tests the forbidden baryon vertex, rather than
the forbidden meson vertex and would give insight on the relation
between OZ1 violations for the two cases.

225

UNDERSTANDING THE NEW PARTICLES

Further tests of the OZI rule are now possiblell,18 in

the decays of the new particles into final states containing the
and f'. These also enable tests of the generalizations of the
OZI rule to multi particle vertices and can settle questions
regarding the crossed Pomeron and multiply forbidden diagrams of
Fig. 5.3.
Tests of the OZI rule are more complicated for production
of mesons which are not ideally mixed like the pseudoscalars
because there is no clear selection rule forbidding the production
of any physical particle. The pseudoscalar state analogous to the
or f' whose production is forbidden by the OZI rule in np
reactions is a linear combination of the nand n'. However, the
OZI rule for neutral meson production can also be tested by other
relations which are not seleption rules. Such relations were first
derived by Alexander et al. 24 together with the selection rule
forbidding production in np reactions. An SU(3) rotation of the
selection rule forbidding production with incident pions leads to
the observation that an incident K- which contains no n-type quarks
or anti quarks cannot produce the (nn) vector meson state. Since
this state is not a physical meson but a linear combination of the
po and W states, the selection rule is expressible as an equality
between po and W production amplitudes, and similarly for the
tensor mesons

An additional relation is also obtained from the additive quark

model for each case,
(5. 42a)
O'(K - P --+ f 'Y)

*0 )
0' ( n - p --+ KY.

In the quark model derivation, these relations follow from the

requirement that the meson transition be described as a single
quark transition with the other quark being a spectator. All meson
transitions in Eqs. (5.42) have the same quark transition p --+ 1
and differ only in the quantum numbers of the spectator quark,
which is a A on the left-hand side and an n on the right-hand side.
The same relations (5.42) arise in Regge exchange models which
assume SU(3) symmetry, no exotic exchanges and the OZI rule.

H. J. LIPKIN

226

The relations (5.41) and (5.42) hold for any meson nonet
but are not directly applicable to the pseudoscalar mesons because
they assume ideal mixing. However, they can be combined to give a
sum rule which holds independent of mixing angle and can be applied
to the pseudoscalars as well

a(K-p ~

A2 0y) +

a(n-p ~ K*Oy).
(5.43b)

Additional sum rules independent of mixing angle were also

24
obtained by Alexander et ale for pseudoscalar meson production,

(5 .44a)

(5. 44b)

Analysis of experimental data at 3.9 GeV/c shows striking

agreement with experiment for the relations (5.41a) and (5.42a) for
fector meson production but strong disagreement with experiment 34
for the sum rule (5.43c) for pseudoscalar meson production where
the left-hand side is 705 91 ~b and the right-hand side is
1121 59 ~b. There are also troubles with the sum rule (5.44a).
The validity of the OZI rule for the pseudoscalar mesons thus
remains unsettled. There is also the possibility that the
conventional nonet classification does not apply to pseudoscalars.
One possible explanation for the disagreement of the sum rules with
experiment is to assume that the OZI rule holds, that the n is well

227

UNDERSTANDING THE NEW PARTICLES

represented by a state which is mainly a member of the same octet

that contains the pion and kaon, but that the n' is a more
complicated mixture involving radially excited octet configurations~5
Since the selection rule forbidding the production of the
(nn) states leads to equality of amplitudes for po and W production
and similarly for f and A2, the equalities (S.41) include the phase
detectable in pw and fA2 interference experiments and found to agree
with the predictions. 1 9 This interference provides another test of
the OZI rule in reactions like (S.4l) where the production of the
(nn) state is forbidden; e.g.
e +e -

K+K- n +n - ,

(qq)

K+K- n+n - ,

./, .J,

'1",'1'"

K+K- n +n - ,

where (qq) denotes any quark-a~tiquark meson state, including new

particles. In (S.4Sa-c) the n n- mass spectrum should show the
characteristic "peak-dip" OW-interference pattern, constructive on
the low-energy side of the W peak and destructive on the high-energy
side. In the reactions (S.4Sd-f) the KK spectrum in the fA2 region
should show interference constructive in the K+K- decay mode and
destructive in the KOj( mode. ll Since the relative magnitudes and
phases of the amplitudes follow from the OZI rule indepe.ndent of
kinematics, data at different energies and from several reactions
can be combined to improve statistics.
5.12

Applications of the OZI Rule to the New Particles

The narrow widths of the new particles are attributed to

the OZI rule, but there has been no reliable quantitative estimate
of these widths from any theoretical model. There have been
attempts to estimate the OZI violation for the new particles by
using experimental data from the old particles as input. However,
the analysis of the f' ~ nn decay in section S.lO shows that this
is unjustified. There is no simple way to apply these results to new
particles which are states of new heavy quark-antiquark pairs.
For these states the channel analogous to the KK channel involves
pairs of mesons each containing one new heavy and one ordinary light
quark. These channels are closed for the decays of the lowestlying new meson states made of heavy quark-antiquark pairs. Thus,
there is no possible higher order decay to hadron states made of
light quarks via an intermediate state on the mass shell. The

228

H. J. LIPKIN

higher-order paradox arises only in virtual transitions to states of

pairs of mesons carrying charm or some new quantum number and
which must be off their mass shell. The calculation of these effects
requires some underlying field theory as well as a knowledge of the
spectrum and of the couplings of all these new particles in order to
determine the effectiveness of various cancellation mechanisms.
There is no simple estimate analogous to the one above for the
f' ~ nn decay which is dominated by the on shell transitions.
There is no way to extrapolate the observed violations in the
f' and decays to these other states by assuming a dependence on
masses, 'since the effect observed in the <P and f' cases can be
entirely attributed to open channels on shell which do not exist
for the new particles. Such mass extrapolations could give upper
limits for OZI violation under the reasonable assumption that the
effects of the off shell transitions must be smaller than the
observed violations in the f' and cases. However, such
extrapolations are very risky since the contributions of these
off shell transitions are probably very sensitive to subtle
cancellation mechanisms which may be very different for the old
and the new particles.
Thus the experimental observation that the OZI rule must
be much better for the new particles than for the old particles
leads to no contradictions. But there is also no simple way to
estimate the widths of the new particles from theoretical models
for OZI breaking.
Another application of the OZI rule to the new particles
has been in estimates of charmed particle production in
experiments searching for these particles. It has been suggested
that charmed particles might be found in associated production
with the J/~ because the production of the J/~ without charmed
particles violates the OZI rule. 3b This argument is incorrect, as
can be seen from the analogous case of the ID and strangeness.
Kaons are not frequently found in associated production with the
, even though the production of the without accompanying kaons
in pp collisions is forbidden by the OZI rule. That t production
is dominated by OZI violating mechanisms is still consistent with
the OZI rule .19
The OZI rule can be tested only by comparing corresponding
pairs of processes as in Eq. (5.36). other comparisons are
misleading, such as comparing production with and without kaon
pairs in NN reactions. A proper comparison of these processes
involves other dynamical considerations. This can be seen by
writing

O"(NN) ~ + K + K + X)
O"(NN ~ ih + X)

KS

(5.46)

UNDERSTANDING THE NEW PARTICLES

229

where X contains no strange particles and the parameters K, Sand

Z are defined as
Z == O'(NN -+ c!> + X)
O'(NN -+ w + X) ,

(5.47a)

S _ O'(NN -+ UJ + K + K + X)
O'(NN -+ UJ + X)

= O'(NN -+ c!> + K + K + X)
K - O'(NN -+ W + K + K + X)
The parameter Z is the ratio of a corresponding pair of
OZI-violating and OZI-conserving processes and is small in any
model which suppresses OZI-violating transitions. The parameter S
is also small because experiment shows that it is hard to produce
kaon pairs. The quantity K is of the order unity since it relates
two processes allowed by the OZI rule and differing only by the
interchange of two members of the same vector nonet. The value of
the ratio (5.40) is thus not determined by the OZI rule and
requires additional dynamical input. It depends upon which of the
two small quantities Sand Z, is smaller, i.e. whether it is harder
to violate the OZI rule or to produce a pair of strange particles.
The available data indicate that the strange-particle-production
factor S overwhelms the OZI-violation factor Z and that c!> groduction
is dominated by the OZI-violating transition without kaons 5 ( up to
energies of at least 24 GeV.

A similar argument holds for the production of J/~

particles with and without pairs of charmed particles in a
charmonium model. By analogy with Eqs. (5.46) and (5.47)
O'(NN -+ J/~ + D + D + X)
O'(NN -+ J/iJ) + X)

(5.48)

where X can now contain strange particles but no charmed particles

and the parameters KcZc and C are defined by
Z

O'(NN -+ J/~ + X)
O'(NN -+ W + X)
O'(NN -+ W + D + 15 + X)
O'(NN -+ w + X)

(S.49b)

H. J. LIPKIN

230

Kc

a(NN

J/ + D + D + X)

=a(NN ~ W + D + TI + X)

Again the ratio (5.48) depends on which is the smaller

of two small quantities Zc and C; i.e. whether it is harder to
violate the OZI rule or to produce a pair of charmed particles.
In addition the factor Kc which is of order unity in the SU(4)
symmetry is probably also small. Present experiments suggest here
also that charmed particle production is suppressed more than OZI
violating transitions and that the J/ is produced primarily
without charmed particles. This can also be seen in a model like
Eq. (5.36) with Zc a universal factor describing the suppression of
all OZI-violating transitions inv~lving a charmed quark-anti quark
pair. A value of about 10- 3 --10- for Zc is obtained by comparing
the width of the OZI-violating decay of the J into normal hadrons
with expected widths of about 100 MeV for OZI-conserving decays.
The observed total cross sections for J/ and W production are
consistent with this value of Zc and Eq. (5.49a).
If the OZI rule were exact and all OZI-violating
processes had zero cross section, all production of particles like
the <P and the J/ would be via OZI-conserving reactions. However,
once the OZI rule is broken the OZI-violating transitions are
proportional to small but finite suppression factors as in Eqs.
(5.47a) and 5.49a). Whether a given production process is
dominated by OZI-conserving or OZI-violating transitions depends
upon whether the OZI-conserving transition is suppressed more
strongly than the OZI-suppression factor by a different dynamical
mechanism, as in the reactions (5.46) and (5.48).

VI. WHY ARE NARROW CONTINUUM STATES INTERESTING?

The new particles promise new exciting opportunities for
interesting research. To understand why they are so interesting
it is instructive to compare them with the isobaric analog states.
There are many common features. The isobaric-analog states were
understood very shortly after their discovery, but they are still
very interesting and have opened a new field in nuclear physics.
For the same reason the new particles will still be very
interesting even after we understand their structure. There is
much discussion about these new particles and whether they should
be called or J. I see that Italians know all about the Psi
(Italian socialist party).

UNDERSTANDING THE NEW PARTICLES

231

Narrow states in the continuum are interesting because

they combine the best features of the ordinary low-lying states
and the ordinary continuum states. In any non-trivial
spectroscopy there are models which are not exact but useful
because they are reasonably good for the low-lying states where
configurations are simple. At higher excitation the models get
worse; the configurations are much more complicated, the density
of states is higher, and there is much more mixing. The
interactions neglected in the models become more unimportant at
higher excitation, where the level density becomes greater and
there are many configurations close together and easily mixed.
Thus there is no simple theoretical model for the wave functions
high in the continuum. On the other hand, high continuum states
are very convenient for experiment because they have many open
channels, can be excited in many ways, and provide a very rich
source of experimental data. The isobaric analog states give
the best features of both worlds. They are up in the continuum
and have many possible open channels, many ways of excitation,
and many things to study. But they also have a very simple
structure very much like the low-lying states and simple
theoretical models can be used. Any model good for the low-lying
states is also just about as good for the narrow states in the
continuum. This then opens a wide field for experimental tests
and investigations of the various models.
How do we know that narrow states in the continuum have
a simple structure? We know the structure of the isobaric analog
resonances and can answer this question in detail. But the answer
is similar for the new particles even though we do not understand
their structure. In both cases we know that the structure of the
states must be simple because they are simply produced. The
isobaric analog state is strongly produced by a nuclear chargeexchange reaction at zero momentum transfer on a nuclear ground
state; i.e. by changing a neutron into a proton without changing
its ~omentum.
Any state strongly produced by this very simple
operation must differ from the ground state only by a simple
elementary excitation and cannot be a state with twelve particles
excited.
The new particles are made very strongly by electronpositron annihilation through one virtual photon. The transition
matrix element connects the particle state with a vacuum by the
operator of the electromagnetic current and is appreciable only if
the particle has a very simple structure. For example, in simple
quark or parton models, the electromagnetic current operator creates
only single quark-anti quark pair. This tells us that the single
pair part of the wave function must be appreciable in any strongly
produced state. If it were mainly a state of thirty quarks and
thirty anti quarks it would not have a very large matrix element for
a single photon transition from the vacuum.

232

H. J. LIPKIN

VII.

WHAT ARE STRANGENESS AND BARYON NUMBER?

The new charm degree of freedom 2 provides a new quantum

number like electric charge, strangeness and baryon number. But
understanding charm is difficult when we still do not understand
the old internal degrees of freedom. 38 We have some understanding
of the role of electric charge in particle interactions and dynamics
even though we do not understand why electric charge is quantized
and universal. But our understanding of baryon number and
strangeness is much weaker. There is no theory like quantum
electrodynamics in which baryon number or strangeness appear as
coupling constants defining the strengths of interactions. There
is no formula analogous to the Rutherford formula for Coulomb
scattering describing the dependence of strong interaction
scattering on baryon number and strange~ess.

A few phenomenological models and symmetries like the

quark model and SU(3) symmetry give rough descriptions of the
dependence of total cross sections on baryon number and strangeness.
But these descriptions are highly inadequate and the difference
between mesons and baryons and between strange and non-strange
hadrons are not really understood. Furthermore, many of the
models developed work in only one area of hadron physics and are
incompatible with models used in other areas. For example, the
quark model used in describing hadrons strong interactions is not
the same as the quark model used in weak interactions.
Consider, for example, the description by conventional
models of the difference between pion and kaon wave functions. The
quark model says that both are made from a quark-anti quark pair. 14
But weak interaction quarkists explain the ratio of the 11 -+ U + \)
and K -+ ~l + \) decay requiring the wave functions at the origin to
be very different as described by Weisskopf-Van-Royen 39 formula

III>K( 0) 12
III>11 (0)1 2

(7.1 )

Strong interaction quarkists say that the difference

between pion and kaon wave functions is measured by the difference
between their scattering cross sections on nucleons. These differ
by less than 20%. Recent data at high energies show that 11p and
Kp differential cross sections approach equality with increasing
momentum transfer. This suggests equality within 20% of the mean
square radii of pion and kaon wave functions and nearly identical
short distance behavior, in sharp contrast with the weak quarkist
Eq. (7.1)

233

UNDERSTANDING THE NEW PARTICLES

The very precise experimental data 40 now available on

pion, kaon and nucleon total cross sections give us some information
about the difference between the interactions of strange and
nonstrange particles with matter. Careful examination of the data
show that this difference ~s very interesting but also very puzzling
and not really understood. 41 Instead of the conventional plot of
total cross sections versus laboratory momentum on a logarithmic
scale, we show the systematics in a more interesting plot (Fi g . 7.1)
with a square root scale rather than a logarithmic scale for Plab
and with the total cross section multiplied by ~lab~ This is
equivalent at thes e high energies to a plot against center-of-mass
momentum of the imaginary part of the forward amplitude obtained
from the total cross section by the optical theorem. Theoretical
reasons why the curve of 7.1 is so much simpler than the standard
plot follow from a two-component descr.7i Ption of the cross sections
with a Regge component varying as s-l 2 and a pomeron component
varying slowly as a function of energy. A more detailed discussion

--

100

"-

>

80

Q)

C)
)C

.c
E

60

40

C\I

'-....

.c

0
a...- 20
)C

a
a

Fig . 7 . 1. atot~/20 vs . ~.
multiplied by 2/3 .

Nucleon cross s ections

234

H. J. LIPKIN

is given elsewhere. 42 For our purposes this particular plot shows

very clearly that there is a di~~erence between strange and nonstrange particles and that there are puzzles not explained by the
quark model.
In Fig. 7.1 the nucleon-nucleon and nucleon-antinucleon
cross sections are multiplied by a ~actor 213. The six quantities
plotted are just those predicted to be equal asymptotically in the
simple quark model with the pomeron component an SU(3) singlet
coupled equally to pions and kaons and coupled to mesons and
baryons by simple quark counting prescriptions. Figure 7.1 shows
that these cross sections are indeed all equal at the 20% level.
However, beyond this approximation o~ "seen one hadron, seen them
all" the di~~erence between the np and the pp cross sections is
seen to be strangely similar to the di~~erence between the np and
Kp cross sections. The di~~erence between mesons and baryons
seems to be similar to the di~~erence between nonstrange and strange
mesons.
This regularity is shown more precisely by examlnlng
linear combinations o~ cross sections which have no Regge component
and are there~ore conventionally assumed to be pure pomeron. The
K+p and pp channels are exotic and have no contribution ~rom the
leading Regge exchanges under the common assumption o~ exchange
degeneracy. The ~ollowing linear combinations o~ meson-nucleon
cross sections are constructed to cancel the contributions o~ the
leading Regge trajectories

(7.12.)
(7 .lb)

Figure 7.2 shows these two quantities on the conventional plot

cross section versus Plab on a log scale.

o~

~(p) as de~ined by Eq. (7.1a) is the quark model

expression ~or ~(p); i.e., the cross section ~or the scattering
o~ a strange quark-anti quark pair on a proton.
The very simple
energy behavior o~ this quantity as seen in Fig. 7.2 is striking.
It shows a monotonic rise beginning already at 2 GeV/c That total
cross sections rise at high energies was ~irst noticed by Serpukhov
data ~rom 20- 50 GeV/ c, but the older data at lower energies already
show this rising behavior in ~(p). I~ anyone has suggested
something particularly ~undamental about this cross section ~or
strange quarks on a nucleon be~ore the Serpukhov data were
available and concluded that its rising cross section indicated
that all cross sections would eventually rise he would naturally

235

UNDERSTANDING THE NEW PARTICLES

20~----.-----~,---------r-------.------'

.0

E
b

- - ---- ---- ...

o~

__

~~

10

____

~~

20

______

~~

50

____

~~

100

____

200

P'ab (GeV Ie)

Fig .

7. 2. Plots of Eqs . (7. 1) and (7. 2).

have been disregarded as crazy. But now that the whole picture up
to 200 GeV/c is available we may conclude that there is indeed
something simpler and more fundamental about the cross sections
for strange quarks on a proton target. Understanding this simpler
behavior may help us to understand the more complicated energy
behavior of the other cross sections.
The quantity 6(nK) defined by Eq. (7.1b) represents the
difference in the scattering of a strange particle and a nonstrange
particle on a proton target. In the quark model this is the
difference between the scattering of a strange quark and a non strange
quark on a proton target after the leading Regge contributions have
been removed. This difference between strange and nonstrange also
has a very simple energy behavior, decreasing constantly and very
slowly (less than a factor of 2 over a range Plab of two orders of
magnitude). So far there is no good explanation for why strange
and nonstrange mesons b.ehave differently in just this way.

236

H. J. LIPKIN

Since the two quantities (7.1) have no contribution ~rom

the leading Regge trajectories they represent something loosely
called the pomeron. However, their energy behaviors are di~~erent
~rom one another and also ~rom that o~ the quantities a(~p) and
a( pp) which should also be "pure pomeron." However the ~ollowing
linear combinations o~ a(K+p) and a(pp) have exactly the same
energy behavior as the meson-baryon linear combinations (7.1)
a l (pK)
~(MB)

"23 a(K+p)

"31

a(pp)

(7 .2a)

= "31 a(pp ) -"21 a(K+)

p

(7.2b )

These quantities are also plotted in Fig.

7.2.

The equality o~ the quantities (7.2) and the corresponding

quantities (7.1) suggest that the pomeron, de~ined as what is le~
in the total cross sections a~er the leading Regge contributions
are removed by the standard prescription, consists o~ two
components, one rising slowly with energy and the other decreasing
slowly. The coe~~icients in Eq. (7.2) were not picked arbitrarily
but were chosen by a particular model. In this model the rising
component o~ the total cross section is assumed to satis~ the
standard quark model recipe exactly.
(7 .3a)

where Y denotes a A or
assumed to satis~ the

hyperon.

~ollowing

The ~alling component has been

relation

This particular behavior is suggested by a model in which the

correction to a simple quark-counting recipe comes ~rom a double
exchange diagram i~volving a pomeron and an ~ coupled to the
incident particle. 1
We thus see unresolved problems in the total crosssection data associated with the questions o~ what is the
di~~erence between strange and nonstrange particles and what is the
nature o~ the pomeron. Note that Eq. (3.lb) de~ines the di~~erence
between the scattering o~ a nonstrange quark and a strange quark
while Eq. (7.2b) can be interpreted as the di~~erence between the

UNDERSTANDING THE NEW PARTICLES

237

scattering of a quark in a baryon and a quark in a meson. The

fact that the strange-nonstrange difference and the meson-baryon
difference are equal and have the same energy behavior over such
a wide range is a puzzle which may be explained by pomeron-f
double exchange but may also indicate something deeper.
The cross section differences ~(rtK) and ~(MB) are both
predicted to vanish in the simple model where the pomeron is an
SU(3) singlet which is coupled to the quark number and all the
curves of Fig. 7.1 are equal. The deviation from the additive
quark model ratio of 2/3 for meson to baryon scattering appears
as a finite value of ~(MB), the deviation of the pomeron coupling
from an SU(3) singlet appears as a finite value of ~(rtK). One
might ask whether both these decreasing quantities approach zero
at high energies, so that the simple model would be valid in
asymptopia.

5~----.-------,,----------.-------.-------~

-.0

E
b

o~----~------~--------~~----~~------~

10

20

~ab

Fig. 7.3.

50
(GeV Ie)

100

200

Plots of 6(rtK) and ~(MB) on an expanded scale.

H. J. LIPKIN

238

A close look at the experimental plots on an expanded

scale of 6(nK) and 6(MB) in Fig. 7.3 reveals a small difference
in the behavior at the highest energies. The curve for 6(nK)
seems to be leveling off above 50 GeV/c, while that for 6(MB)
continues decreasing monotonically. This trend seems to continue
in the one or two additional points available up to 280 GeV/c which
are not plotted. Additional data up to 400 GeV/c should determine
whether 6(nK) has definitely stopped decreasing and is approaching
a constant, while ~(MB) is decreasing. If this is the case, then
the additive quark model becomes good at high energies while pions
and kaons continue to look different even at asymptopia and the
SU(3) relation never becomes good. The equality observed between
these two differences over the 6- 200 GeV/ c range and described
by a two-component Pomeron ultimately breaks down at higher energies.
A search for similar systematics in elastic hadron
scattering diff~rential cross section data has led to new surprises
and paradoxes.3~ With only differential cross section data available and no det~iled amplitude analysis, it is convenient to define
the quantity4l,43
S(Hp) =

[~:

(lip) +

~:

(HP)J l / 2

(7.4)

where H is any hadron. This quantity S(Hp) is assumed to give a

good approximation for the Pomeron contribution to the Hp
scattering amplitude.
With this assumption the simple additive
quark model prediction that 6(MB)= 0 becomes
S(np) = (2/3)S(pp)
when we use S(np) to represent a typical meson baryon cross section.
The assumption that the Pomeron is a SU(3) singlet predicts
A(nK) = 0 and
S(np)

S (Kp).

The two relations (7.5a) and (7.5b) describe the dependence of the
scattering amplitude on baryon number and strangeness, respectively.
The two component Pomeron model which relates the deviations from
the two predictions (7.5a) and (7.5b) predicts the weaker sum rule
S(np) =

21

S(Kp) +

31

S(pp).

(7.6)

UNDERSTANDING THE NEW PARTICLES

239

The experimental data 44 show that the weaker sum rule

(7.6) is in much better agreement with experiment than the
additive quark model prediction (7.5a). However, the SU(3)
prediction which is not very good at t = 0 becomes better at larger
values of t and becomes much better than the two component Pomeron
prediction (7.6) or the additive quark model prediction (7.5a).
Two examples of this comparison with experiment are given in
Table 7.1. The same qualitative features are present in all the
data.
Table 7.1
Tests of Additive Quark Model (AQM), Two-Component Pomeron (P2)
and SU(3) Relations Between Differential Cross Sections.
RHS/LHS of Eqs. (7.5a), (7.5b) and (7.6).
P
t
(GeV/c)

100 GeV/c
P = 175 GeV/c
P2
AQM
AQM
P2
SU(3 )
t
(7.6) (7.5b) (GeV/c) (7.5a) (7.6)
(7. 5a)

SU(3)
(7. 5b)

0.0

1.2

1.0

0.84

0.0

1.1

-0.08
-0.16

1.0

0.86

-0.08

0.98

0.94

0.95
0.91

0.88

-0.16

0.88

-0.24

0.85

0.87

0.90

-0.24

0.89
0.81

0.85
0.86

0.84

0.88

-0.32

0.78

0.85

0.92

0.74

0.81

0.89

-0.40

0.71
0.66

0.83

0.94

-0.32
-0.40

G.68

0.79

0.90

-0.48

0.63

0.77

0.92

-0.56
-0.64

0.58

0.76

0.93

0.56

0.97
0.80 1.0
0.80 1.0

0.54

0.74

0.95

-0.72

0.53

0.80 1.1

-0.72

0.50

0.73

0.96

-0.80

0.50

0.80 1.1

-0.80

0.47

0.72

0.98

-0.48
-0.56
-0.64

0.61

0.81

0.97
0.92

0.84

The comparison with experiment of relations (7.5a)

and (7.6) does not really add any new qualitative information.
It is s.ummed up by the observation that at the optical point the
relation (7.5a) is not very good and the relation (7.6) is much
better and that baryon-baryon cross sections decrease much more
rapidly with t than meson-baryon cross sections. The behavior
at the optical pOint is expected from the similar behavior of total
cross sections. The high t behavior is expected since naive
additive quark model predictions (7.5a) and (7.6) neglect

240

H. J. LIPKIN

differences between meson and baryon wave functions. These

differences introduce additional form factors into the scattering
amplitudes, which cause baryon amplitudes to decrease more rapidly
with increasing t than meson amplitudes.
However, the improvement of the relation (7.5b) with
increasing t comes as a complete surprise. One can ask why pions
and kaons should look more alike 4 5 at high t than at low t. One
might also ask whether the two are really approaching equality or
whether there will be a cross over and that still at higher t the
amplitude will differ in the opposite direction.
We thus seem to see a peculiar systematics in which the
additive quark model becomes good at t = 0 and high s but not at
high t, the SU(3)-symmetric pomeron becomes good at high t, but
not at t = 0, even at high s, and the two-component pomeron
description holds at t = 0 and s between 6 and 200 GeV/ c, where
there are discrepancies in both the additive quark model and the
SU(3)-symmetric pomeron. Further data on total cross sections at
higher energy, differential cross sections at higher momentum
transfer, and hyperon total and differential cross sections
everywhere will show whether these puzzling features are really in
the data, and will provide clues for our understanding of the
differences between strange and nonstrange particles and between
mesons and baryons.
REFERENCES
1

J. D. Anderson, C. Wong and J. W. McClure, Phys. Rev. 126

(1962) 2170; A. K Kerman, in "Nuclear Isospin' (J. D. Anderson,
S. D. Bloom, J. Cerny and W. W. True, Eds.), p. 315, Academic
Press, New York (1969).
2For a review of charm, see M. K. Gaillard, B. W. Lee and J.
Rosner, Rev. Mod. Phys. 47 (1975) 277.
3S Okubo, Pbys. Letters 5 (1963) 165.
4G Zweig, unpublished, 1964; and in Symmetries in Elementary
Particle Physics (Academic Press, New York, 1965) p. 192.
5J Iizuka, Supplement to Progress of Theoretical Phys. 37-38
(1966) 2l.
6B. Sakita, Phys. Rev. 136 (1964) B1756; F. Gursey and L. A.
Radicati, Phys. Rev. Letters-13 (1964) 173; A. Pais, Phys. Rev.
Letters 13 (1964) 175.
---

7J . L~Rosner, in Proc. XVII Intern. Conf. on High Energy Physics,

London (1974), ed. J. R. Smith, p. 11-171, Physics Reports llC
(1974) 190.

UNDERSTANDING THE NEW PARTICLES

241

8A Arirna and F. Iachello, Phys. Rev. Letters 35 (1975) 1069.

9N Cabibbo, in Proc. of Intern. Conf. on High Energy Physics,
Palermo, Italy, 1975.
loa. J. Lipkin, Symmetries and Resonances (Theoretical), in
Particle Physics (Irvine Conference, 1971), edited by M. Bander,
G.L. Shaw and D.Y. Wong (American Institute of Physics, New York,
1972), p. 30.
l~. J. Lipkin, Spectroscopy after the New Particles, in Froc.
Intern. Conf. on High Energy Physics, Palermo, Italy, 1975.
12y. Ne'ernan, in Spectroscopic and Group Theoretical Methods in
Physics, edited by F. Bloch et al., Wiley, New York (1968), p. 337.
13G Racah, Group Theory and~pectroscopy, Ergebnisse der
Naturwissenschaffen, 37 (1965) 28.
14H J. Lipkin, Physics Reports 8c (1973) 173.

15S J. Brodsky, Proc. of the 16th Int. Conf. on High Energy

Physics, National Accelerator Laboratory, Batavia, Ill. (1972).
16J ElliS, in Proc. XVII Intern. Conf. on High Energy Physics,
London (1974) ed. J. R. Smith, p. 11-171.
17 0 W. Greenberg and C. A. Nelson, Color Models for Hadrons,
to be published in Physics Reports.
18S Okubo, Phys. Rev. D 13 (1976) 1994 and D 14 (1976) 108.
19H J. Lipkin, Physics Letters 608 (1976) 371.-2oa. J. Lipkin, Who Understands the Zweig-lizuka Rule, to be
published in the Proceedings of the 1975 Rencontre de Moriond.
21H J. Lipkin, Phys. Rev. Letters 31 (1973) 656.

2~. J. Lipkin, Phys. Letters 9 (1964) 203; H. J. Lipkin and

S. Me shkov , Phys. Rev. 143 (1966)-1269.
2 3A Katz and H. J. Lipkin, Physics Letters 7 (1963) 44.
24G Alexander, H. J. Lipkin and F. Scheck, ;hYS. Rev. Lett.
17 (1966) 412.
25E M. Levin and L. L. Frankfurt, Zh. Eksp. i. Theor. Fiz-Pis'rna
Redakt 2 (1965) 105 [JETP Lett. 2 (1965) 65J.
26F.-Gilrnan, M. Kugler and S. Meshkov, Phys. Rev. D9 (1974) 715;
Phys. Letters 45B (1973) 481.
27B Richte~private communication.
28H J. Lipkin and S. Meshkov, Phys. Rev. Letters 14 (1965) 670.
29C B. Chiu and J. Finkelstein, Phys. Letters 27B (1968) 576.

H. J. LIPKIN

242

3C. Schmid, D. M. Webber and C. Sorensen, Nucl. Phys. BIll

(1976) 317.

3~. L. Berger and C.Sorensen, Phys. Letters 62B (1976) 303.

32A J. Pawlicki et al., Phys. Rev. Lett. 37 (1976) 971.
33D Cohen et al., Argonne Preprint ANL-HEP-PR-76-64.
3 4J Mott et al., Phys. Rev. Lett. 18 (1967) 355j M. AnguilarBenitez et al., Phys. Rev. Letters 26 11972) 574j Phys. Rev. D 6
(1972) 29.
35H J. Lipkin, Fermilab Preprint-FERMILAB-Pub-76194-THY.
36M Binckley et al., Phys. Rev. Lett. 37 (1976) 58.
37V Blobel et al., Pbys. Letters 59B (1975) 88; H. Harari,
Theoretical Implications of the New Particles, rapporteur talk at
the Intern. Symp. on Lepton and Photon Interactions at High Energies,
Stanford, California, August, 1975.
38H J. Lipkin, "Why are Hyperons Interesting and Different from
Nonstrange Baryons?", Particles and Fields 1975, Edited by H. J.
Lubatti and P. M Mockett, University of Washington (Seattle) p. 352.
39R Van Royen and V. F. Weisskopf, Nuovo Cimento 50A (1967) 617.
40E Flaminio, J. D. Hansen, D. R. 0. Morrison and~ Tovey,
Compilation of Cross Sections. I. Proton Induced Reactions. II.
Antiproton Induced Reactions, CERN/HERA 70-2 and 70-3j E. Bracci,
J. P. Droulez, E. Flaminio, J. D. Hansen and D. R. O. Morrison,
Compilation of Cross Sections. I. n- and n+ Induced Reactions. II.
K- and ~ Induced Reactions, CERN/HERA 72-1 and 72-3j J B. Allaby,
Yu. B. Bushnin, Yu. P. Gorin et al., Yad. Fiz. 12 (1970) 538j Phys.
Lett. 30B (1969) 500j Yu. P. Gorin S. P. Denisov, S. V. Donskov
et al.:-Yad. Fiz. 14 (1971)j Phys. Lett. 36B (1971) 415j S. P.
Denisov, S. V. Donskov, Yu. P. Gorin et al., Nucl. Phys. B65
(1973) Ij A. S. Carroll et al., Pbys. Rev. Lett. 33 (1974l127 and
FERMILAB-Pub-75151-EXP.
-41H J.
(1975) 76j
42H J.
Tell Us?,"
23-28 June

Lipkin, Nucl. Phys. B78 (1974) 381j Phys. Lett 56B

Pbys. Rev. D 11 (197~1827.
Lipkin, "What--:-re Total Cross Section Data Trying to
Proc. Int. Conf. on High Henergy Physics, Palermo, Italy,
1975.

43H Harari', Ann. Phys. 63 (1971) 432j 1. Ambats et al., Phys.

Rev: D 9 (1974) 1179j V. Barger, K. Geer and F. Halzen, Nucl.
Phys. B44 (1972) 475.
44C. W. Akerlof et al., Phys. Rev. Lett. 12 (1975) 631j
Fermilab Single Arm Spectrometer Group, Phys-.-Rev. Lett. 35 (1975)
1195j G. Brandenburg et al., SLAC-PUB-1607 (1975). The data in
Table I are taken from the Fermilab Single Arm Spectrometer group,

UNDERSTANDING THE NEW PARTICLES

243

but data from all three experiments have been used to test relations
(7.5) and (7.6) with similar results. The help of all three
groups in making data available before publication and computing
preliminary tests of relations (7.5) and (7.6) is gratefully
acknowledged, and in particular correspondence and discusslons
with D.D. Yovanovitch, R. Diebold, D. Leith and J. Mlkenberg.

45 G F. Chew and C. Rosenzweig, Nucl. Phys. Bl04 (1976) 290,

propose a model in which SU(3) symmetry improves with increasing t.

244

H. J. LIPKIN

DIS C U S S ION S

Prof. H. Lipkin

CHAIRMAN:

Scientific Secretary:

M. Pauli

DISCUSSION 1
FALCIONI:
Do you not think that deep inelastic scattering and neutrino
experiments may support the idea of realistic spin-1 quarks?

LIPKIN:
The behaviour to which you refer is limited to a definite low
region of energy. He know nothing about the scaling at high energy.
There is no indication that quarks are elementary and are not composite objects of still more fundamental particles.

FREEDMAN:
\\Then + PTI via ~ + K+K- + pTI and + KOKo + PTI, the strange
quarks in the K's must annihilate to go into pTI. Does this part of
the decay not become suppressed by Zweig's rule?

LIPKIN:

,-uord
or a forbidden

' - - ii

but
O

~ ord}

~.~C,~a
d

..

is allowed

It is called a "duality diagram".

You can compose this into 3 point functions

UNDERSTANDING THE NEW PARTICLES

245

No. The ZI rule does not suppress the annihilation of strange

quarks from two different hadrons such as a kaon pair. Only the
annihilation of a quark-antiquark pair from the same meson is
suppressed.
MARCIANO:

What does the SU(6) you discussed have to do with the SU(6)
traditionally used in particle physics?
LIPKIN:

Nothing. I thought the nuclear physicists can build beautiful

models out of six building blocks -- yet no one takes these building
blocks seriously. There may be a moral in that.
WICK:

Is there a sharp formulation of the ZI rule that applies in

general, rather than specific examples?
LIPKIN:

The formulation of the ZI rule for three point functions is very

sharp and clear. The dual resonance model provides a framework for
extending the rule to n-point functions. One way to formulate the ZI
rule for a general n-point function without drawing diagrams is the
following: if the external particles can be divided into two groups
such that there is no flavour of quark which appears in both groups,
then the vertex is forbidden by the ZI rule. In the three-flavour
model -- no charm -- this reduces to the following statement: if
the external particles can be divided into two non-trivial groups
such that one group contains only strange quarks and the other contains only non-strange quarks, the vertex is forbidden. It is easy
to formulate this in the Born approximation -- but it is harder to
formulate the ZI rule for higher order terms. Since these are strong
interactions, we need to do this.
WICK:

Could you give us an example?

LIPKIN:

KK

The simplest example is +

+ PTI because you know the ,K,K_
coupling. You could make this KK + wand (;j + pTI. + Ki and w + KK
are related by SU(3). If you know how to calculate this loop, then
you can calculate the relationship between + pTI and w + pTI. There
is no reason for this to be small unless there is anot~er diagram

246

H. J. LIPKIN

<P -+

KK -+

-+

pTI

showing loop diagram

that cancels this one. There are a number of possibilities and this
will be discussed in detail in my next talk.

FREEDMAN:
Are there any other cases similar to that of <P -+ KK -+ PTI where
the ZI rule is violated? Specifically, I mean where the two quarks
annihilating from different hadrons in the intermediate state originated from the same initial hadron and are not a G-parity forbidden
decay.

LIPKIN:
There is the ft

-+

KK

-+

TITI

decay, and perhaps also ~

-+

DD

-+

pions.

PAULI:
How do you put G parity into an interacting theory of quarks
and gluons so as to obtain the selection rules implied by G parity?
Could the notions of G parity be generalized in such a way to give
a dynamical explanation of Zweig's rule?

LIPKIN:
G parity is a combination of isospin and charge conjugation.
Any theory of quarks and gluons which is invariant under isospin and
charge conjugation transformation automatically conserves G parity.
There is no transformation or symmetry which can be put into a theory
to give the ZI rule. The ZI rule is not formulated as a conservation
law.

McPHERSON:
Is there any reason why the non-planar diagrams must be cancelled
in the same order? Could it not be cancelled by the same conspiracy
of higher order diagrams?

LIPKIN:
Not obviously.

But all known formulations which provide such

247

UNDERSTANDING THE NEW PARTICLES

cancellations, in particular, dual resonance models with exchange

degeneracy, have cancellations in the same order.
DISCUSSlor~

McPHERSON:
In the discussion of the KK* ~ KK* Zweig rule situation as to
the pseudoscalar intermediate exchanged state, how does the A~ state
behave?

LIPKIN:
K - - - - - . - - - - - KIt

or other eccha1ge

K-----~----

The AA state is always split from the pp and nn states by SU(3)

symmetry breaking. Thus, A~ exchange is always distinguishable from
the other exchanges and does not affect the argument. There is no
degenerate partner of the pion to cancel ZI rule violating transitions produced by pion exchange.

KLEINERT:
pp

Can the ZI rule explain the observation that the new narrow
pp resonance at 1930 HeV is not seen in pp ~ nn?

LIPKIN:
This has no direct connection with the ZI rule. Any resonance
with definite isospin which appears in the pp ~ pp (elastic) amplitude must also appear in the pp ~ nn (charge exchange) amplitude.
The experiment indicates that there must be degenerate isoscalar and
isovector resonances like pw or fA 2 The pp system is too complicated
to provide a simple explanation of why these two resonances interfere
in such a way as to exactly cancel the charge exchange reaction.
There are just too many possibilities and too many free parameters to
adjust. This makes experimental predictions meaningless.

p---------p

248

H. J. LIPKIN

KLEINERT:
Is not the re-arrangement collision the most important one where
you do not have anything go across but just re-shuffle -- the six
quarks corne in, re-arrange, and go out again.

LIPKIN:
Not necessarily.

FERBEL:
You are so little above threshold that re-arrangement collisions
are dominant.

LIPKIN:
This would probably be true for elastic scattering which is
commonly associated with a Pomeron exchange, but here you are dealing
with an s-channel resonance. These are described in dual resonance
models by quark-antiquark annihilations and not by a mere re-arrangement

WIGNER:
Could you give the energy of the resonance?

KLEINERT:
The energy above threshold

~s

of order 100 MeV.

LITTENBERG:
The mass is 1930 MeV, and the width, as deter~ined by Carroll et al
is about 9 9 MeV. Montanet got a similar width with better errors
which made it clear that it is not infinitely narrow. These two experiments had completely different systematic errors; and I think
this is to be regarded as a well-founded object by now.

KLEINERT:
It could be some exotic object which could have three quarks
and three antiquarks in one bag.

ZICHICHI:
I think we should not discuss this too much because the effect
three standard deviations. I would like to see some more convincing data.
~s

UNDERSTANDING THE NEW PARTICLES

249

PARSONS:
Am I correct in assuming that the ZI rule combined with the
assumption of exchange degeneracy produces all the inhibitions that
the dual resonance model of Chan-Hong Mo et al. produces?
LIPKIN:
One does not usually assume both the ZI rule and exchange
degeneracy. One follows as a result of the other. Thus, for example,
from the assumption of exchange degeneracy and the absence of exotics
in all channels, Chiu and Finkelstein obtained the results of ideal
mixing and the ZI rule.

FERBEL:
At what level would the breaking of exchange degeneracy affect
the ZI rule and the nice duality results? For example, 25% violation
of, let us say a(O) with that observed at FNAL; would that hurt exchange degeneracy, exotics, and all those other nice ideas?

LIPKIN:
This is a complicated and interesting problem. Different ways
of breaking exchange degeneracy can have different effects on the
ZI rule in different processes. The violations you mentioned involve
non-strange trajectories. Their consequences would probably show up
in other processes rather than in the violations of the ZI rule where
the observed processes always involve strange quarks; e.g. the common
selection rules for the and fl which contain only strange quarks.
The relevant trajectories whose exchange degeneracy are necessary
to prevent higher order violations of the ZI rule are the strange
K, K*, and Q trajectories. Experimental information is still weak
on these.

KLEINERT:
Is not the partner of the pion the B trajectory or the corresponding strange partner?

LIPKIN:
The Q trajectories are the strange members of the SU(3) octets
containing the B and the Al -- these are the trajectories required
to be degenerate with the K trajectory.

250

H. J. LIPKIN

DISCUSSION 3

WILKIE:
Can you clarify for me the difference between the naive Zweig
rule and the dual resonance model Zweig rule and what they have to
say about pp + ~~? Snnce pp + ~~ is the t channel of the ~p + ~p
scattering process which is dominated by Pomeron exchange experimentally, can one predict how it is going to behave?

LIPKIN:
The naive Zweig rule says nothing about pp + ~~ since it only
concerns the annihilation of a quark with an antiquark from the same
external particle. The generalized ZI rule for states with more than
three particles forbids reactions like pp + ~~ because the diagram
describing it is disconnected. It is true that this is related by
crossing to ~p elastic scattering which is dominated by the Pomeron,
but this seems to be irrelevant. The Pomeron gives a strong amplitude when it is exchanged in the t-channel, not the s-channel.
Other examples of reactions, which are related to elastic scattering
by crossing, but are found experimentally to be suppressed are:
pp + 6-6-, pp + I-I-, pp +~, and pp + ~~. These reactions are
all suppressed because they cannot proceed via simple t-channel exchange. The fact that they have vacuum quantum numbers in the schannel is irrelevant.

Diagram for pp + ~~, forbidden by the generalized

Zweig rule since it is disconnected

ETIM:
You did not say much about radiative decays of mesons in connection with the Zweig rule. Would you briefly comment on this and on
the discrepancies of these decays as predicted by the quark model
with the results using the Primakoff effect, such as p + TIY, KO* +
+ KOY, and ~ + nY, and all these decays that show discrepancies with
the quark model?

LIPKIN:
All that the ZI rule says about radiative meson decays is to
forbid dJ + ny, ,.,hich seems to be all right. The discrepancies you

UNDERSTANDING THE NEW PARTICLES

251

mention involve difficulties of fitting the latest data with quark

model or SU(3) relations which are not directly related to the ZI
rule; in particular, the ratio of r(p + TIY) to r(w + TIY). This
disagreement is very serious, since the relation between p + TIY and
w + TIy is very much model independent. This can be seen by noting
that the same vertex appears in both transitions in the vector dominance description, and that the mass difference between the p and
the w is negligible.

Vector meson dominance description of vector meson decay

WEILL:
Is the ~weig rule applicable to multiparticle production? For
an example, one could give pp + many pions since when one draws the
quark lines, there do not seem to be enough quark lines to produce
many particles.

LIPKIN:
Yes, it is very easy. You see that you cannot make a without
having at least one additional kaon pair.
It

=:J

(
(
(

41
K

non -strange

baryon

Possible graph in multiparticle production.

Hote that a kaon pair must accompany a meson

KLEINERT:
Consider NN + TITI or NN + KK at low energies. The production
in the final state is suppressed more strongly than one would expect
from SU(3) by a factor of ten.

LIPKIN:
If you start with a non-strange initial state, it is always
harder to make strange particles. TIp + pTI is greater than TIp + K*A.

252

H. J. LIPKIN

ZICHICHI:
The TITI is 10- 3 and the KK is how much?

LEADER:
It is about a five- or ten-to-one ratio but it
multiparticles.

~s

also true for

KLEINERT:
Yes, this is just an example.

LIPKIN:
Quasi two-body annihilation seems to be limited by baryon exchange. The intercept difference between the A trajectory exchanged
in KK production and the N trajectory exchanged in TITI production is
sufficient to expalin this.

FERBEL:
Is there any reason why the K production cross-section should
be about a factor of ten smaller than that for pion production?

LIPKIN:
In any Regge description where pions are produced by exchange
of p or N trajectories while kaons are produced by exchange of strange
trajectories, there is a natural suppression of kaon production because of the lower intercept of strange trajectories.

BERLAD:
If one uses ideas of analyticity and absence of exotics in
comparing TIp -+ KL: and TIp charge exchange., one can explain quite well
the strangeness exchange in TIp -+ KL: by SU(3) relations, relating it to
TIp charge exchange. There are no special numerical factors aside from
SU(3) factors. I do not know about this for vector mesons, but I do
know it for the pseudoscalars.

LIPKIN:
If you look at this as a simple peripheral reaction, you always
get a suppression simply by the intercept of the strange trajectory.
Thus, if you look at high-energy inclusive things in any kinematic
region, where you could use some kind of Regge formulation, you
always find that producing a strange particle involves a strange
trajectory somewhere; and the intercept will always knock down the
amplitude compared to non-strange production.

UNDERSTANDING THE NEW PARTICLES

253

ZICHICHI:
This is one of the basic mysterious points which will be discussed next year at this school on the topics "The Whys of Subnuclear
Physics" -- to say that strange quarks are coupled less than non-strange strange quarks, means to translate into the quark language
the mystery of strangeness.

KLEINERT:
The asymptotic freedom people have something to say about that.

ZICHICHI:
This is very strange that they have anything to say. {Laughter
Asymptotic freedom exists independent of strangeness.

&n audience.}

KLEINERT:
Asymptotic freedom people always say that it is harder to make
a pair with high masses such as a strange quark-antiquark pair with
respect to a non-strange quark-antiquark pair because you need higher
q2 to make it, roughly speaking.

ZICHICHI:
Do they know the strange quark mass?

LIPKIN:
When I talk to the asymptotic freedom people, they say it is
very nice for charmed things, but these light quarks, that includes
strange and non-strange quarks, are all too light for them.

ZICHICHI:
The mass difference between strange and non-strange quarks is
very small if you believe in Kleinert - who knows everything about
quarks.

BUCCELLA:
You can co~pute the imaginary part for f' + TITI through unitarity,
inserting the KK state. You also say that the off-shell contribution
can give you the real part. Can you explain this?

LIPKIN:
The real part has contributions from high momenta where there
is no reason to include only the KK state. Once higher K*'s are in-

254

H. J. LIPKIN

cluded, cancellations can occur which are model dependent. It is

very difficult to make any definite statement. You could use some
dual resonance model to sum over all trajectories. You need a definite model to tell you which things are going to contribute. Even
the im~ginary part is not so definite. There are transitions like
f' + KK + feR) + nn, where feR) is the Regge recurrence of the f.
Veneziano has pointed out that the tail of the feR) can give an
appreciable contribution at the mass of the f'.

FREEDMAN:
Recent spectroscopy calculations seem to give a strange quark
mass of roughly 300 MeV and a non-strange quark mass of roughly
50 MeV. Threshold effects for model field theory calculations typically become important for q2 greater than four times the mass squared.
When strange quarks are absent in the initial state, would it not be
more difficult to produce strange hadrons as compared to the production
of non-strange hadrons?

LIPKIN:
Experimentally, there is no evidence for such an effect. The
simplest case is the A2 + pn, KK, or nn where you have SU(3) and
SU(6) relations and you put in phase space and that works. There
are no suppressions for making a kaon pair out of an initial nonstrange pair other than phase space.

KLEINERT:
What energies -- what values of q2 -- are you referring to?

LIPKIN:
Well this, of course, is at very low q2. The production of
p, w, and by e+e- annihilation agrees with SU(3) predictions with
no observed suppression of creation of strange quarks.

ARE STRONG INTERACTIONS STILL WITHIN THE REGGE FRAMEWORK?

ELLIOT LEADER
WESTFIELD COLLEGE, UNIVERSITY OF LONDON
LONDON NW3 7ST
0)

PROLOGUE

This is the first time that I have had the pleasure of attending an Erice school. On many previous occasions, however, I have
studied the circulars announcing the schools, when deciding whether
or not to send a particular student, and I have to confess that
there is one part of these posters that has always aroused my
suspicions; and that is the little section headed "Poetic Touch",
wherein the beauties and attractions of Erice are extolled, with
references to Homer and other lustrous figures of Classical Antiquity.
You will appreciate that in this age of package holidays and
advertising pressure, my reaction of suspicion is not unreasonable.
However, now that I am here I must freely admit that that
paragraph is not only justified, it is, if anything, an understatement, and ever since setting foot in this enchanted vi Ilage
I have found myself overcome by a poetic urge, an irresistable
desire to mimic Homer. You will forgive me, therefore, if I
begin my lecture by showing you the fruits of this urge. I have
to warn you, though, that there are two problems. Fi rstly my
Classical Greek is simply dreadful and secondly my poetic abi lity
is even worse. Nevertheless, here it is:POETIC TOUCH
HERE FOLLOWS A SAGA UN- GARNI
OF A STRANGE YOUNG LAD OF TRAPA(R)NI
WHOSE CHARM AND DETERMINATION
LED OFT' TO THE FUN IVA-STATION.
255

256

E. LEADER

YET ONCE ON THE MOUNTAIN PICHCHI

OUR YOUTH, ANTONINO ZICHICHI,
TURNED HIS THOUGHTS FROM LOVEIS INTERACTIONS
TO BASIC HADRONIC REACTIONS.
IN THE HEART OF ERICE CELESTIAL
CAME AN IDEA SOME THOUGHT QUITE BESTIAL TURN THE CONVENT OF THE HOLIEST VISION
INTO A CENTRE FOR NUCLEAR FISSION.
AND LO! OUR PRESENCE HERE TODAY
GOES TO SHOW IN A DEFINITE WAY
WHAT RESULTS FROM A THOUGHT ICONOCLASTIC
WHEN CONCEIVED IN A MIND SO ELASTIC!
No sooner was my poetic urge satisfied than I found my
thoughts turning to Philosophy. What would be needed to give
a definitive answer to the question raised in the title of my
lecture? How would one really decide whether "Regge Poles" are
right or wrong? Historians and Philosophers of Science spend
a great deal of time discussing how a theory is accepted or
rejected or disproved. Their discussion is dominated by the
rather nai ve concept of the "cruci al experiment". Just how
naive this is, is well illustrated by Regge Theory.
In truth:We do not wish to give up a theory.
We come to love it.
It becomes a part of our language
even if it is wrong.
Moreover: Theories are alive (and slippery).
They are Da rw in i an .
They adapt themselves to the experimental situation.
I shall deal today with a classic case of the "failed crucial
experiment ll , the pion nucleon charge exchange reaction TI-P + TIon.
Not that it was a failed experiment. It was a superb experiment.
But it failed to be crucial. However, since the concept of the
crucial experiment is so unrealistic it is not surprising that the
experiment has turned out to be exceedingly important and interesting.
Because of the great spectrum of interests and experience in
the audience I should like to give a brief recapitulation of Regge
Theory before getting down to the serious analysis.
I shall consider Regge Theory on two levels. About the first,
the fundamental one, there will be no discussion. It contains ideas
and concepts that have already appeared in Text Books on Quantum
Mechan i cs.

STRONG INTERACTIONS WITHIN THE REGGE FRAMEWORK

257

The principal elements are:i) Continuation into the complex J-plane.

ii) Existence of poles at J = a(t).
iii) Relationship of these poles to the bound states and resonances
of the system. A bound state or resonance of mass m and spin J occurs
when a(mZ) = J.
iv) Signature: Continuation from physical values of J must be done
separately for even and odd values. Labelled by = l.
These ideas are as solid as Quantum Mechanics itself. At an
altogether different level are what I shall call the "derived"
properties. which allow contact to be made between the above and
the world of high energy scattering.
The ingredients are:
Ana lyt i ci ty
Crossing
Kinematic Accidents.
The result is that for a process
A+B-+C+D
with (C.M. energy)2 = s and (four-momentum transfer)2
the scattering amplitude has the asymptotic behaviour

s-+oo

f(s.t) t fixed
where
a(t) = Trajectory of Regge Pole in the crossed channel
i.e. in the process AC-+ SD
S(t)

Residue function whose form is not specified by

the theory.
-ina
l+.e
F;(t) = Signature factor

What are the a(t)7 They are labelled by internal Quantum

Numbers corresponding to known particles or resonances e.g. p.
w. A2. f etc. The most infamous of them is the Pomeron. with
Quantum Numbers of the vacuum. invented specifically to "explain"
elastic diffractive scattering. Historically. in the early 1960s
the shrinkage of the pp differential cross section. da/dt.and its
energy dependence led to the need for a trajectory

a IP (t) = 1 + t
where the value a fJ (0) = 1 is of course correlated via the
Optical Theorem with the then believed behaviour aTOT -+ constant.
Looking at a present day picture of aTOT vs energy. one is tempted
to ask what the~ would be like if Regge Theory had been invented
in 1975~

258

E. LEADER

Historically the TIN charge exchange reaction has been

especially significant. Because the Quantum Numbers of the
exchange are so restricted, namely to mesons with 1=1 and G=~,
and since the only particle with such Quantum Numbers was the
p, it was expected that n-p + nOn would display the classic
features of Regge Pole exchange:
i)

Shrinkage of do/dt: Since one has

and since for small t

one expects

a{ t) = t + a l (t)
p

do 0: e talP logs
dt
which, bearing in mind that a p > 0 and that t < 0, shrinks
as s increases.
_

ii} Both the spin non-fl ip amplitude f++ and the spin-fl ip
amp 1 i tude f+_ should have the same phase. One has

f++

0:

f+_

0:

~p{t)

S++{t) sap{t)

~p{t) S+_{t) sap{t)

and the only non-real quantity, ~ (t), is common to both. An

immediate consequence is that thePpolarisation, which depends
on interference between f++ and f+_, will be zero. i.e.
P 0: 1m (f+_* f++) = 0
As is well known (i) was beautifully confirmed by experiment,
whereas for (ii) a significant non-zero polarisation was found!
How could one escape the di lemma?
A secondary trajectory called pi was postulated. It was to
be a daughter of the p i.e. its trajectory would satisfy
al{t)=a{t)-l
p
p
Unfortunately this implies the existence of a vector meson at a
mass of 1250 MeV, something which Nature has failed to provide us
wi tho
A more attractive solution physically was provided by
2
Nicolescu and myself. The polarisation has a zero at t : -1 (GeV/c)
A natural explanation of this would be that at this point apl = a ,
thus making all phases equal at this point. Further, a study of p
the energy dependence of 0n-p - on+p indicated that at t=O one

STRONG INTERACTIONS WITHIN THE REGGE FRAMEWORK

259

should have apl (0) : o. With two points on its trajectory fixed,
assuming as usual a straight line, one found that one had a
rather flat pi trajectory

a pl (t) : 0 +
t
so that the implied vector meson would have quite a large mass,
about 1700 MeV, and this is not ruled out experimentally. Thus
points (I) and (ii) could be reconciled without grossly contradicting
the known particle spectrum.
To summarize the situation on n-p ~ nOn in the pre-FERMI LAB
days: A very good description of all data could be achieved using
p and pi trajectories as specified above. We shall see, however,
that the situation has changed as a result of the recent FERMI LAB
experiments. Indeed we shall suggest that once again this
reaction may be playing a historical role and that it may be
teaching us about entirely unforseen effects in exchange type
reactions.
1) I NTRODUCT ION
The concept of Reggeisation of particle exchanges in
dynamical calculations has been of enormous importance both from
a phenomenological point of view and from the deeper point of view
of the fundamental structure of strong interaction dynamics.
On the phenomenological side a major weakness of the theory
has always been its virtually unconstrained flexibility, so that
most attempts to pin it down or to subject it to critical tests
have ended in abject failure, defeated by the sheer elusiveness
of the theory. Despite this it cannot be denied that the Regge
theory has provided a very convenient and useful framework in
which to analyse reaction amplitudes both for elastic scattering
and for inclusive multi particle reactions. Moreover it has
succeeded, on a qualitative level, in "explaining" and correlating
a great many features of the data.
Regge theory, in its full generality, presumably requires
the consideration of both poles and their accompanying cuts and
the precise predictions emerging naturally reflect the level of
sophistication achieved in handling these complicated effects.
Thus many arguments have raged as to what precisely Regge theory
does imply in a particular reaction, and failure to agree with
experimental data is often only a reflection of the specific
approach adopted in grappling with the cut complications.
In the present lectures we wish to suggest tentatively
that there is growing evidence that in a rather absolute sense
conventional Regge theory does not correctly describe the asymptotic
behaviour of scattering amplitudes. By "absolute sense" we mean

260

E. LEADER

in a way transcending the details of a particular parametrisation

of the poles or a specific approach to the handling of the
associated cuts.
If this is correct then Regge theory is somehow or other
an intermediate energy theory and not a truly asymptotic theory.
It is easy enough to say this, but it is not at all clear how
such a situation could arise at a fundamental dynamical level.
Possibly it could reflect a new type of behaviour of matter at
the very short distances involved in ultra-high energy experiments.
We shall argue that data emerging from Fermi lab and from
Serpukhov for exchange type reactions i.e. reactions in which
there is an exchange of definite, non-vacuum quantum numbers,
seem to show an energy dependence that is completely at variance
with that expected from Regge theory with conventional poles and
cuts. In particular there appear to exist exchange type amplitudes
e.g. some with the quantum numbers of p A2 that give rise to physical
effects that do not vanish as PL ~ 00. Although this is very
startling and unexpected it should be remembered that recently we
have learnt from the behaviour of opp at the ISR that the symmetric,
vacuum exchange, amplitude grows like slogs or s(logs)2 and is thus
not Regge-behaved in the simple fashion expected of Pomeron exchange.
Perhaps it is not too surprising then if we discover that anti-symmetric
amplitudes are also non-Regge behaved at very high energies.
In the following we shall discuss in turn the reactions

np~nn

and

np ~ pn
np ~ pn

We shall not go into the technical points but restrict

ourselves to a qualitative description of the essential points.
Full details and a complete set of references can be found refs. 1-3.
2)

PION-NUCLEON CHARGE EXCHANGE SCATTERING

We consider the differential cross section do/dt for small

angle charge exchange scattering, n-p ~ nOn, and the difference
of the total cross-sections for n+p and n-p. We define

- a np
+
np
and use the standard notation AI, B for the
amp 1 i tudes. The optical theorem then reads
1m AI (t=O) = PL
'2 t:.o.
t:.o

=. a -

np~nn

(1)

STRONG INTERACTIONS WITHIN THE REGGE FRAMEWORK

261

To begin with, we shall study to what extent the data on ~o

and on dO/dtlt=o are compatible with various versions of standard
Regge-pole theory.
i)

Single, pure Regge pole.

a) If one parametrizes the imaginary part of A' at t=O in the

usual way
~

1m A' (t=O) = i3s ,

(2)

and attempts to fit the data on ~o of all energies above

PL = 10 GeV/c, a very good fit can be achieved when the parameters
take on the values

i3 = 106.6 ~b2, ~ = 0.54.

Using the standard formula for the Regge signature factor, one
takes the real part of A' at t=O to be given by

(~~J

Re A' (t=O) = tg

1m A' (t=O).

(3)

One then computes dO/dtlt=o and compares it with the data.

The fit is extremely poor. The X2 for the four Fermilab points
alone is 122. The fit is shown by the solid curve in Fig. 1.
b) If one reverses the procedure, and first attempts to adjust
the parameters so as to obtain a good fit to dO/dtlt=o one now
finds as best fit values
i3 = 184.5

.1

~b2,

= 0.42,

and not surprisingly, when one compares the expressions with the
data on ~o one finds a poor fit both at low and high energies.
The value of X2 is 76 for 16 points for the ~o fit. The fit is
shown by the dashed curves in Fig. 1.
The results of the two methods of choosing best parameters
indicate that, while either ~o or dO/dtlt=o can be very well fitted
by a single, pure Regge pole, the parameters required for the pole
are significantly different in the two cases. Thus a single pole
cannot provide a complete description of the physics at t=O.
ii) Pure p+p' Regge poles. Since the p alone is inadequate
we attempt tQ fit to ~o and dO/dtl t=O using a p and p' pole model.
Old models based on p and p' do not fit the new data, so a reanalysis
of the parameters is called for. We put
1m A' (t=O) = S s~l + S s~2,
1

(4)

and try to fit all the data above 5 GeV/c, keeping ~p close to
0.5 as required for a 1 inear trajectory to pass through the p and
g mesons.

262

E. LEADER
~ 300~---------------------------,

,
'I

'i,

,,

1000

,,

500
.... .....

100~

10

__~~~~~wu~__~__~~
SO
100
500

PL

(GeV/c)

Fig. 1 - Indication of the mismatch between the ~ pole in the first

case optimized to fit the energy variation of ~O (solid curve) and
in the second case optimized to fit the energy variation of dO/dt!t=O
(dashed curve). (Data: ~ from ref. (4) from ref. (5).

from ref. (6).

values

The

X~

is quite good (: 1.3 per point) for the parameter

l
124.5 1J b.

0. 1

-86.1 loibl,

0. 2 =

= 0.49,

-0.16,

but the shape of the curves is unsatisfactory at Fermilab energies.

The bad shape at Fermilab energies is reflected in a bad X2
(: 2 per point) in the Fermilab region.
A good way to see the difficulty is to look at the behaviour
of the experimental ratio

STRONG INTERACTIONS WITHIN THE REGGE FRAMEWORK

263

Re AI (t=O)
1m AI (t=O)

-~....;,;.,...;.;-.....,..;~

As seen in Fig. 2 neither p nor p + pi are able to follow

the experimental trend.

Fig. 2. - The experimental values of the ratio Re AI/1m AI at t=O

compared with the value of this ratio for i) a single p pole and
i i) a p pole plus a pi pole (---). in both cases the pole parameters
being constrained to fit the energy variation of ~o.

We conclude that the p + pi model with a conventional p

intercept cannot explain the data and that a new term whose real
part is negative at Fermilab energies is required.
We have also tried fits in which ~p is left completely free.
A good fit to the t=O data can be achieved (~2 : l.l/point) with
~p = 0.44.
However.
a) the fit is sstematically
100 ~ PL ~ 240 GeV/c (X = 2/point).
b) the pi in this description
It is in no sense secondary to the p
physical interpretation based on the

bad for ~o in the region

plays an exaggerated r;le.
and one loses much of the
p as the dominant mechanism.

As a result. an attempt to use these t=O parameter values

in a study of the t-dependence leads to wildly unacceptable do/dt
and polarizations.
We should also mention that other models such as p+ soft
Regge cuts are very similar to the p + pi model and will suffer

264

E. LEADER

from the same deficiencies at Fermilab energies.

iii) Can dispersion relations influence the phase? Since
all the above models fail as a result of a mismatch between the
energy variation of their real and imaginary parts, or equivalently,
as a result of their unacceptable phases, we wish to examine whether
the whole problem has not arisen because of an injudicious use of
asymptotic phases.
Remember that the usual IIRegge phase ll is in fact an asymptotic
one, and that a Regge-behaved imaginary ampl itude when fed into a
dispersion relation will yield a real part differing from the IIRegge
real partll at low energies, but tending towards the Regge value as
the energy increases.
For all the above models we assume a Regqe-like behaviour in
the w~ole complex plane, and then set up unsubtracted dispersion
relations into which are fed the imaginary part of AI as given by
each model. In no case does the calculated real part of AI differ
significantly from the IIRegge real partll beyond PL = 10 GeV/c.
We conclude that the non-Regge behaviour observed is
genuine and is not due to a misuse of the asymptotic phase formula.
iv) Qualitative properties of the new term at t=O. We
have seen from the above analysis that, if we wish to retain the
standard Regge terms, then we require in addition a new term with
the p rope rt i es
(5)

{a)
b)

Re (new term) < 0 at Fermi lab energies,

1m (new term) > 0 and increasing with energy at Fermilab.

In fact it is also possible to have 1m (new term) = 0

since the standard Regge-pole model can be optimized to provide an
excellent fit to ~o. It is important to note that analyticity does
not allow us to use an analogous argument about fitting dO/dtlt=o
with a standard Regge form, and thereby to conclude that Re (new
term) could be zero.
Thus, finally, our condition on the imaginary part is
(5 I)
b l ) 1m ( new term ) ~ O.
The new term cannot be of the form of a soft Regge cut
since the relative signs of the real and imaginary parts would
be wrong and also since the cut correction would decrease with
energy, in contrast to the empirical requirements.
We could try an assumption that 1m AI (t=O) has a term of
the form sa(log s)~ with a > t, 6 > 0 in it and with IAII bounded
by sa(1og s)6 for s complex-and lsi + 00. In that case one would
have asymptotically
AI : (i + t g ~a) sa ( 1og s) 6 ,
(6 )

STRONG INTERACTIONS WITHIN THE REGGE FRAMEWORK

265

.provided strictly ~ < 1. We see that this is unacceptable since

its real part has the same sign as its imaginary part. This
conclusion is of course only exact asymptotically, but a detailed
analysis using dispersion relations supports this conclusion at
finite energies.
It is interesting to note that, if ~ = I, then with B at its
maximum allowed value, i.e. B = 1, the asymptotic behaviour becomes
AI (t=O) ; s [log2s - i'lr log s],
(7)
so that in this case Re AI and 1m AI have the desired opposite
sign as s + 00. It is noteworthy that this behaviour corresponds
to a singularity at J = 1 in the complex-angular-momentum plane
and has the maximum growth possible for an anti-symmetric
ampl itude.
It would also be possible to interpret the new term as
arising from an odd signatured analogue of the pomeron, i.e. a
Regge pole, which we shall refer to as the "odderon" and which
has ~(O) = 1. This term would be purely real at t = 0 and would
thus only playa role in dO/dtlt=o'
The above qualitative statements are supported by a careful
and detailed study at t=O.
Thus we are led to contemplate the possibility of an exchange
ampl itude that behaves asymptotically something 1 ike s~ with
~ = 1.
Such a behaviour is quite outside the conventional picture
of Regge exchanges. It seems to represent a totally new type of
dynamical effect which only begins to show itself at ultra high
energies.
If such an amplitude is present at t=O one should ask what
ro 1e it plays away from the forward direct i on. I t turns out
that there is not enough data to really pin down the properties
for t < O. However, an exceedingly good fit to the data can be
obtained by making a very simple ansatz for the t-dependence of
the new term. For example, taking a new term in AI of the form
B(t) s~(t) with ~(O) = 1 (the odderon picture) and an exponential
t-dependence for B(t) one finds the near perfect fits shown in
Figs. 3, 4, 5.
In summary there is evidence in TI-p + 'iron for something quite
unexpected at high energies. However, the evidence is flimsy at
this stage and more persuasive consequences must be sought before
a truly convincing case can be established.
It is quite remarkable that despite the smallness of the new
term a dramatic effect of its existence can be demonstrated.
Although the magnitude of the new term is minute its phase is
significantly different from the phases of other standard Regge
contributions. Thus one might expect noticeable effects in an
experimental quantity which is sensitive to interference. Indeed

300~-------------------------,

'l..
~

->III

C>

:::::

~ 100

..

....0
II

50

dcr/dtl t = 0

1000
500

-..

."

."

-2

10

100~10~--L-~~5~0~~I~OO~--~~~5~OO~
PL (GcV/c)

Fig. 3 - P+p'+odderon fit to ~cr and dcr/dtlt=o

Fig. 4
P+P'+od

STRONG INTERACTIONS WITHIN THE REGGE FRAMEWORK

267

ex (0)=05e

..
i.
o
II

10

';c

!oe

o
10'

PL (GcV/c)

102.

Fig. 5 - The ratio Re A'/Im AI at t=O given by the p+pl odderon

mode I ( - ) compared wi th the data. The hatched band is the same
as in Fig. 2.

one finds that the new term generates a huge and totally unexpected
polarization at smal I t values in TI-p + TIon at Fermilab energies.
The precise form of the polarization is not predictable far from
t=O, but in the region of small t, say -t < 0.5, it is quite well
determined. Just how surprising the predicted behaviour is can
be seen in Fig. 6. The discovery of such an effect in the charge
exchange polarization would constitute a most convincing confirmation
of the unexpected asymptotic behaviour of the antisymmetric I = I
exchange ampl itude.
If the new asymptotic behaviour is a real effect one mayor
may not see a change in the energy dependence of ~o = on-p -on+p'
This is model dependent. However, one would certainly see a
continuation of the strange behaviour of
Re AI (0)
r = Im A' (0)
shown in Fig. 2, with r + -00 ultimately as s + 00, and as a
consequence the value of do/dt at t=O would, for example, have
dropped to about 2/3 of the value expected in conventional Regge
theories at PL = 300 GeV/c.

268

E. LEADER

50

5 GeV/c

(Argonne data)

-30

20 GeV/c

~o~====-~~--=====
.250
.....

50 GeV!c

L..

o
o

a...

50
200 GeV/c

-50

Figure 6

STRONG INTERACTIONS WITHIN THE REGGE FRAMEWORK

3)

269

NEUTRON-PROTON CHARGE EXCHANGE SCATTERING

In an analogous fashion we consider the variation with energy

of the difference of total cross-sections 0pp-opn and of the
forward differential cross-section for np + pn.
The nucleon-nucleon case differs from the meson nucleon case
in two major aspects:i) It is extremely difficult to obtain a precise measurement
of the cross-section difference. Experiments either use deuterium
targets, in which case there is some uncertainty in the exact
magnitude of the Glauber correction needed in order to extract
0np' or they use neutron beams, in which case the statistical
accuracy of the data has not been sufficiently accurate - at
least not up to the present.

To make matters worse, different experimental groups have used

different values of the "deuteron radius" when making their
Glauber corrections, so that raw deuterium data on which there is
perfect agreement between different groups ends up yielding
significantly different values for 0np. Clearly, a detailed
study of the energy dependence of 0pp - 0np will be very sensitive
to such anomal ies. An essential requirement therefore is a
re-analysis of the raw deuterium data ensuring a compatible
treatment of the Glauber correction in different experiments.
We have carried out such an analysis and all statements in
the following wil I refer to the results of this analysis.
Also in the neutron beam case there is an unreasonable
difficulty. The targets used for the np and pp measurements
have been very different in length. In comparing 0pp with 0pn
there is thus a danger of some uncontrollable systematic error.
Despite all these problems, some confidence in our conclusions
is inspired by the fact that the results using neutron beam data
are in fairly good agreement with those obtained from the re-analysed
deuterium data.
ii) There is no unique relationship between 0pp - 0np and
dO/dt (t=O) for np + pn. This is because of the existence of
three independent non-zero ampl itudes at t=O in nucleon-nucleon
scattering. qO/dt (t=O) is proportional to I~ll 2+1~212+1~31 '.
But 0pp. - 0np directly determines only 1m (~1+~3) whereas ~2 which
is dominated by ~-exchange, is independent of the difference of
total cross-sections. Thus a comparison of the energy variation
of 0pp - 0np and do/dt (t=O) for np + pn might not be so conclusive
as for the analogous meson-nucleon case. We shall nevertheless
see that fai rly significant conclusions can be reached.
Let us begin by recall ing the expected behaviour of opp - 0np
in a conventional model. Only exchanges with the quantum numbers

E. LEADER

270

of p and A2 can contribute, so that one would have, at high

energies,
~p

= ~(opp

- onp) ---

IS'

(8)

In fact ~p. depends on (P-A2) so that in exchange degenerate

models there is actually a cancellation between p and A2 in (8)
and one has
~p =

at reasonably high energies.

Fig. 7 shows the pub1 ished data on ~p using deuterium data
and Fig. 8 shows the results of re-ana1ysing the new data.
In Fig. 9 the neutron beam data can be seen to be very
similar to the re-ana1ysed deuterium data.
Looking at Figs. 8 and 9 we see that there appears to be a
small non-vanishing component to ~p. If this effect is genuine
then we are again facing the possibi1 ity of a most unexpected
asymptotic behaviour for a quantum number exchange type amplitude.

Analysis of ~p
~(Opp - Q- ), which depends upon (P+A 2 ),
suggests that the new effect r~~ides also in the A2 type amp1 itude.
Taking these effects at face value and bearing in mind our
study of meson-nucleon scattering, we are forced to conclude that
both the symmetric (A 2 ) and anti-symmetric (p) 1=1 exchange
amp1 itudes are showing unexpected, non-Regge asymptotic behaviour.
However, even more so than in the meson-nucleon case, it is
important to try to find some independent check on the existence
of such behaviour which does not rely upon extremely precise
measurements and the dangers inherent in the subtraction of
similar large numbers. Again, as in the meson-nucleon case,
despite the smallness of the new term one can find an experimental
quantity wherein it would playa significant role and wherein it
would thus produce quite unexpected and dramatic effects.
We have already mentioned that the forward differential
cross-section for neutron proton charge exchange scattering depends
upon three of the five nucleon-nucleon he1icity amp1 itudes. One
has
(10)

The imaginary part of (~ ~ ) is fed directly by (f : 0np)'

while ~2' (and only ~2 at t=d) 'is fed by ~ exchange. ~~ IS a
remarkable fact that up to the present, that is up to pre-Fermi lab
energies, the extraordinary spiked shape and the energy dependence
of do/dt for very small t have been in perfect agreement with
~ exchange coupled to a smooth coherent background having the same

STRONG INTERACTIONS WITHIN THE REGGE FRAMEWORK

271

o
oN

>w

C)

Ci...J
~

...

.Il

.2

.." i...
J\
""
Q.

Io--()-I

r--.

~
~

:::>

..0

a:

UJ

:::>
UJ

loa.

)(

a.
a.
10
""--IN

X
)(

)(
)(

II

<J

Lt'I

)II

a.

)(

)(
U"I

If'I

6I

Q)
L-

:J
0'1
LL.

272

E. LEADER

a
aN

i...

III

...c

....... i..
~

III

~
t-O-f

>

"

C>

a...J

aa

00
(\)
L-

:J

en

ll..

.Q

c:

c:

a..

\0

)(
)(
)(

)(

- IN
II

)(

a..

<l

)(

a..
a..

III

aI

273

STRONG INTERACTIONS WITHIN THE REGGE FRAMEWORK

0
0

--

>

-"
C>

a...J

~
.Q

.!

....

0"\

Q)

L.

::J
OJ

\.L

-.0

c:

0..

to

<
lLI

II)

o~
IX 'I(

~a
j

lLI

-0..
0..

lO

1Il

-IN
II

0..

<l

1Il

I()

6I

274

E. LEADER

energy dependence as the ~ itself. No need for any contribution

from ~1+$3 has ever been found. At first sight this might suggest
that do/dt is not the place to look for consequences of the new
asymptotic ter~ However, it must be remembered that ~2 as given
by ~ plus background is constant in energy so that do/dt drops
rapidly 1 ike l/s2. At Fermilab energies this contribution has
dropped to such a small value (less than 10 micro-barns per (GeV/c)-2)
that even a small magnitude of Ap can give a significant additr.onal
boost to do/dt. The exact numerical connection is as follows.
The contribution to do/dt (t=O) from a non-zero Ap is given by

~ (t=O) in ~b. (GeV/c)-2 = 51.09 [Opp - 0pn in mb]2

(11)

-2
Thus even ~ mb produces about 12 ~b (GeV/c) ! - a 100% effect!
Hence, if our interpretation is correct, if there really is a
new asymptotic term present, then we must expect quite a dramatic
change in the energy dependence of dO/dt (t=O) in np + pn. There
is no escape.
The situation with the data on neutron-proton charge exchange
is tantalising. Fig. (10) shows an extrapolation of n + background
contribution, which is in effect an extrapolat[on of the data
below 50 GeV/c. A few experimental points below 50 GeV/c-are
shown as we 11 as some pre 1 i mi nary data from Fermi 1abo It is
remarkable that the Fermi lab points fall significantly above the
extrapolated curve.
However, the interpretation is compl icated by the following.
If op~ - 0pn # 0 then no matter what happens to any other
contributions one would have an exact mathematical lower bound,

do ( t=O )
dt

51.09 [opp - opn] 2

(12 )

The value of the R.H.S. of eqn. (12) depends on whether we use

the neutron beam data or the deuterium data. In Fig. 10 we see
that the Fermilab data falls right between the two lower bounds shown.
It violates the neutron beam lower bound and one is forced to
conclude either that the neutron beam data is wrong or that the
preliminary Fermi lab charge exchange data is incorrect.
Thus despite the huge amplification which charge exchange
produces in any new asymptotic term of Ap we cannot at present
reach a clean conclusion. Suffice it to say that the data is
pregnant with implication and that one awaits experimental
confirmation with much exictement.
4)

BACKWARD np SCATTERING

We come now to what may well be the most startl ing example of
unexpected asymptotic behaviour - the exchange amplitude with
baryon number one that controls backward' (u ~ 0) scattering in
7f p co IIi s i on s .

STRONG INTERACTIONS WITHIN THE REGGE FRAMEWORK

275

Extrapolation of data below 30 GeV/c

. . - - - - uSing 11'
background

100

np--.pn

N
I

10

APr~liminary

TF~rmilab Data

ISO
Fi gure 10

200

250

E. LEADER

276

Before going on to considering the new results emerging

from very high energy studies we must recall that baryon exchange
reactions have generally been less well understood than meson
exchange processes at typical medium energies (5-30 GeV/c). Thus
in backward np scattering where the dominant exchanges are the
N (nucleon) and ~ (Nj3) trajectories there have always been
certain difficulties of interpretation. The major features of the
reactions, which are illustrated for PL = 10 GeV/c in Fig. 11 are:
i) TI+P + pTI+ has an extremely sharp backward peak and a deep
minimum at u: -0.15 (GeV/c)2.
i i)
iii)

TI-p

pTI- has neither a sharp peak nor a minimum.

In the backward direction one has

dOl
5 xdu
TI- P

iv) The above behaviour remain qual itatively true up to

17 GeV/c.

-- 1T~P}

- - -1f P

til

Eye fits

>w
C>

.c

=t..

"-

~~

"-

0-1

Fig. 11 - Data on backward TI-p scattering at 10 GeV/c (Taken from

ref.

n.

STRONG INTERACTIONS WITHIN THE REGGE FRAMEWORK

277

All these remarkable features appear to have a natural

explanation in terms of the exchange of the N and ~ trajectories:i) Both N and ~ contribute to TI+p ~ pn+. N exchange dominantes,
is responsible for the sharp peak, and the dip at u : -0.15 is
connected to the wrong-signature nonsense zero occurring at
~N(u) = -~.
This is in agreement with the linear trajectory
~N(u)

= -0.36 + u

suggested by the Chew-Frautschi plot.

ii) Only ~ can contribute to TI-p
sharp peak or dip.

pTI-; hence the absence of

ii i) The dominance of TI+P over TI-p reflects the dominance in

the coupling strength of N over~.
iv) Although ~ exchange drops off more slowly with energy
than N exchange does (one has ~~(u) ~ 0 + U from the Chew-Frautschi
plot) N continues to dominate up to 17 GeV/c and the main features
are unaltered.
Unfortunately this very pretty picture is somewhat marred by the
polarisation measurements which for ~+p ~ pTI+ show a maximum at
u : -0.15 where the theory would predict a zero. There are also
problems with the strength of the coupl ings needed to fit the
cross-section data, but these may simply reflect inadequate
methods of parametrisation.
However, the properties we wish to examine transcend these
detailed questions. We shall argue that the very high energy data
on np ~ pTI seem to be totally incompatible with this conventional
picture.
"-

In the first place as we go to higher energies the role of ~

exchange must become more important simply because ~~(O) > ~N(O).
Ultimately it has to dominate completely, and at this point one
must find several extraordinary results:i)

do
dU

( TI - P ~ pn-) :

do
9 dU

(TI +P ~ pTI + )

because of the isospin properties.

ii)
iii)

The peak in TI+P ~ pTI+ becomes negl igible.

There is no longer a dip in n+p

pTI+ at u = -0.15 (GeV/c) 2 .

All these results are absolutely contrary to the experimental

data below 17 GeV/c and therefore stand as remarkable predictions
of the conventional theory. In quantitative terms one expects the
TI-p ~ pTI- cross-section to have caught up with n+p ~ pTI+ for u=O
at an energy of about 50 GeV.
Fermi lab data is not yet available but data at 25 and 38 GeV/c
from Serpukhov show no sign of the expected behaviour. The peak in

278

E. LEADER

n+p is just as sharp. and do/du for TI+p is sti 11 about five times
bigger than TI-p in the backward direction.
The overall decrease with energy also shows signs of slowing
down. Fig. 12. but it is perhaps premature to try to estimate
what the asymptotic trend will be.
However. a suggestion of a quite extraordinary behaviour has
been inferred by Diu and Tchang
using an indirect method of
analysis. Basically they use partial wave dispersion relations
to evaluate the TIN ~=1 partial waves in the region of s=O.
where they are known from measurements of the TIN phase-shifts
and crossing. In particular the PI 1 phase shift varies very
rapidly near s=O and it is shown that in order to reproduce this
behaviour it is necessary that the I=t exchange ampl itude increase
strongly above PI = 40 GeV/c. Interpreted in the language of

104

->

'"
I

~
C\J

~~
=- ~f
-

- C)

~ 10!

~~

It
I

i-

10

20

PL (GeV Ie)

30

40

5060

Fig. 12 - Energy variation of backward differential cross section

for TI+p scattering. (Data from ref. 7)

STRONG INTERACTIONS WITHIN THE REGGE FRAMEWORK

d (J / d U (}J b/ (Ge V/ C) 2 )

05

OI~--------~----------~----~
a
50 PLob(GeV/c) 100
Fig. 13 - Possible high energy behaviour of backward TI+p
differential cross section. (From ref. 3)

279

E. LEADER

280

trajectories it requires
( 14)

~I=~ (0) ::: 1

instead of ~I=~ = ~N = -0.36.

If (14) is correct, and the isospin one half baryon exchange
ampl itude is dominated by an ~ ::: 1, we shall have the unprecedented
spectacle at Fermilab of constant backward differential cross
sections for TIP + pTI. Fig. 13 indicates how the slowing down
of the energy variation shown in Fig. 12 might continue to take
place at higher energies.
5)

CONCLUSIONS

With the advent of ultra-high-energy machines we may be

entering a realm in hadronic physics where the general ideas of
Regge pole theory no longer hold. The ISR has already taught us
that the pp total cross-section is growing at a rate which is
difficult, and perhaps impossible, to accommodate within the
Regge framework (though strong coupling Reggeon field theories
may sti 11 hold the key). The early Fermi lab results on TI-p + TIon
and on cross-section differences, opp - 0pn in particular, suggest
that also amplitudes with quantum number exchange, e.g. p and A2
type, may be beginning to depart from the expectations based on
the conventional Regge picture. Even baryon exchange, as shown by
the energy trends in backward TIp scattering, may be behaving in a
completely anomalous fashion as compared with the traditional Regge
structure.
We thus face the disquieting, but exciting possibi lity that
p, A2 and N exchange at ultra high energies is ::: 1;
at least in two body reactions. The more conventional Regge
description would then have to be seen somehow as appropriate only
to an intermediate energy range and the "true" asymptotic behaviour
awaiting our detai led investigation, at Fermi lab and at SPS, may
then be quite different.

~effective for

All the experimental effects discussed above, and which are

the basis for our suggestions, are small and difficult to measure.
The consequences for our interpretation, on the other hand, are not.
Herein, we have played the Devi 1 IS advocate but time will soon tell
whether or not it is all just "Much Ado About Nothing".

STRONG INTERACTIONS WITHIN THE REGGE FRAMEWORK

281

REFERENCES
1) D. Joynson, E. Leader, B. Nicolescu and C. Lopez, Nuovo Cimento
30A, 345 (1975).
2) B. Diu and E. Leader, Nuovo Cimento 28A, 137 (1975); A. Bouquet,
B. Diu, E. Leadel\and B. Nicolescu, Nuovo Cimento 31A, 411 (1976).
3)
4)
O.
C.
P.
P.
M.

B. Diu and P. Tchang, Nuovo Cimento 32A, 185 (1976).

A. V. Stirling, P. Sonderegger, J. Kirz, P. Falk-Vairant,
Guisan, C. Bruneton, P. Borgeaud, M. Yvert, J. P. Guillaud,
Caverzasio and B. Amblard: Phys. Lett., 14, 763 (1965);
Sonderegger, J. Ki rz, O. Guisan, P. Falk'::\Tai rant, C. Bruneton,
Borgeaud, A. V. Sti rl ing, C. Caverzas io, J. P. Gui llaud,
Yvert and B. Amblard: Phys. Lett., 3., 75 (1966).

5) A. V. Barnes, D. J. Mellema, A. V. Tollestrup, R. L. Walker,

O. L. Dahl, R. A. Johnson, R. W. Keeney and M. Pripstein:
California Institute of Technology preprint CALT-68-465 (1974);
w. F. Baker, A. S. Carroll, I-H. Chiang, R. L. Cool, D. P. Eartly,
O. Fackler, G. Gi acomell i, T. F. Kycia, P. F. M. Koehler,
K. K. Li, P. O. Mazur, P. Mockett, K. P. Pretzl, D. C. Rahm,
R. Rubinstein and A. A. Wehmann: Phys. Rev. Lett., 33, 932 (1974).
6) K. J. Foley, R. S. Jones, S. J. Lindenbaum, W. A. Love, S. Ozaki,
E. D. Platner, C. A. Quarles and E. H. Wi lIen: Phys. Rev. Lett.,
J;J., 330 (1967).
7) w. F. Baker, D. P. Eatly, P. Koehler, K. P. Pretzl, S. M. Pruss,
R. Rubinstein, R. M. Kalbach and S. Mukhin, NAL Proposal No. 290,
(1974).

E. LEADER

282

DIS -C U S So I -0 N
CHAIRMAN:

Prof. E. Leader

Scientific Secretary:

T. Wilkie

DISCUSSION I
PARSONS:
To remove the discrepancy between Regge theory and experiment,
you have suggested a singularity at J = 1 for t = O. Does this not
imply the existence of a new zero-mass strongly interacting vector
meson?

LEADER:
It would be fun to have a new strongly-coupled photon but,
unfortunately, experiment denies us this pleasure. We have to arrange
that the contribution is non-singular at t = 0 by a decoupling me chanism analogous to a ghost-killing factor.

PARSONS:
Is it obvious how to derive the energy dependence of these new
terms? What happens to signature for instance?

LEADER:
We show that the imaginary part of the new term is given by
s(log s)B, with 0 ~ S ~ 1. The two most interesting cases are the
extremes S = 0 and B = 1. To show that S = 1 is the maximum possible
value is tantamount to re-deriving the Froissart bound, but now for
an antisyrnrnetric amplitude. The phase then follows from the
Phragmen-Lindelhof theorem. In the other extreme case S = 0, the
singularity behaves like an odd signatured pomeron. We would have
a signature factor ~ = [1 - exp (-i~a)J/sin ~a, (T = -1), and this
would give the phase in the usual fashion. Thus the new term is
real for t = 0, ANEW = cs, with creal.

STRONG INTERACTIONS WITHIN THE REGGE FRAMEWORK

283

BERLAD:

Is there a non-relativistic potential scattering example of

crossing trajectories?
LEADER:

Yes, there are examples. There is a classic paper which discusses this problem. I really would not worry about colliding
trajectories -- there is no theoretical difficulty.
BERLAD:

In np ~ np there is a Coulomb contribution in the near forward

direction. In the subtraction 0 n + p - 0 n - p one adds these respective
contributions. If they are not properly removed, one ends up with
an energy independent term. Are you sure this possible contribution
does not affect the high-energy tail of the reaction?
LEADER:

The experimentalists always attempt to include the Coulomb corrections, but you are right to suggest that this could be a source
of danger.
FERBEL:

The correction can never be done too carefully.

LEADER:

This problem was looked at by Carlitz at Chicago.

explain the effects with which I am dealing.

It cannot

TARNOPOLSKY:

Some people think that the double Pomeron exchange process

with snaIl mOIaentum transfers where a state of lm;7 rapidity is formed

provides a stringent test of the Re8ge ideas? lfuat is your opinion
on this idea and the experimental situation?

284

E. LEADER

LEADER:

This is outside the topics which I covered in my talk. I am

not an expert on Reggeon field theory, so I will say very little.
There were two approaches to this problem. In the weak field
theory approach, one drew perturbative diagrams which just look like
the eikonalization of a fundamental potential due to Pomeron exchange. This was never really successful. The alternative strong
coupling approach started off with unphysical energy dependence
~ sl+C, C > 0, which could then renormalize itself. One did get a
rise in the total cross-section with energy, but at energies beyond
the ISR.
The point is that Reggeon field theory is the study of the high
energy behaviour of complicated Feynman graphs. At ISR energies,
not all the sub-energies are large, and so one cannot just blindly
Reggeize. Capella et al in Paris have adopted a more phenomenological approach and claim to have obtained a rise in 0tot in accord
with the data, but I would not like to comment, as I have not
studied their work closely.
I have been so surprised at the diffraction data shown by
Ferbel that I would not really like to comment on double Pomeron
exchange processes. What I would like to know from Ferbel, however,
is what the latest story is on double diffractive processes of the
type

where HI, n 2 , and t are kept small

FERBEL:

I cannot answer for Fermilab since there is no data for fixed

MI and M2. Because of poor statistics, what is done is to integrate
over the masses. This is, I think, a dangerous procedure. There
are some comments on Pomeron factorization from Schlein's group at
IRS, again integrating over the masses. I do not know how to interpret this.
LEADER:

What would be interesting would be to plot (d 3 0)/(dH l dH 2 dt) at

each energy, with MI, H2 in narrow bins, as a function
A exp [b(H I , Jf 2 )t] and to check factorization by comparing with the
elastic and with the single diffraction processes.

STRONG INTERACTIONS WITHIN THE REGGE FRAMEWORK

285

If the Pomeron factorizes, then for HI '" H2, we should expect

:X:I~
p

=bpp

2b (M,)

b(M .M,)

FERBEL:

Using something similar to this, I get factorization to within

the error bars, integrating over the masses -- but I think that due
to this, the test would not be better than to 30% if the data were
better.
LEADER:

One should remember that all tests of Pomeron factorization

work only to within 30%.
BUCCELLA:

That you find this J = I singularity in both I = I meson exchange and in baryon exchange processes is perhaps a sign that the
Regge language, which deals with the exchange of definite quantum
numbers, is no longer appropriate. If you write the s-channel
unitarity equation for elastic processes, then you have a coherent
sum of contributions. For charge exchange, however, the contributions have different signs which tend to cancel. But there is no
reason why the cancellations should switch off the J = I contribution.
LEADER:

Yes, but the

at high energies.
at the beginning,
~ s (log S)2 than
lations in charge

unitarity sum in the s-channel is very difficult

I stated, as part of my philosophical motivation
that the coherent summation gives a larger answer
naive Regge theory. Perhaps, indeed, the cancelexchange only reduce this to O(s) behaviour.

LIPKIN:
You have not included KN data in your analysis of isovector
exchange. Nearly all models predict that the total cross-sections
are related by
O(K+p) - o(K+n)
o(pp) - o(pn)
and
O(K-p) - o(K-n)
o(pp) - o(pn)

286

E. LEADER

Thus, any inconsistency in o(pp) - o(pn) could be checked by looking

at O(K+p) - o(K+n) where the ambiguities in the Glauber corrections
are much smaller. Are the KN data not consistent with

Is this consistent with the differences in o(pp) - o(pn) which you

find In the nucleon data?
LEADER:

Our new result is not directly checked by looking at the KN

system. The couplings could be different and we have no way to
judge them except by appealing to SU(3). However, as long as the
quark relations you mention continue to hold roughly, th~re cannot
be any contradiction.

DISCUSSION 2
GOURDIN:

Nucleon pole dominance in nucleon-deuterium inelastic scattering

implies the well-known expression
o(pd)

o(pn) + o(pp) - (corrections) ,

where the dominant contributions are shown by the diagrams

(0)

The difficulty is to extrapolate correctly at the nucleon pole, taking

into account the deuteron structure. This procedure, which is suitable to isolate diagrams (a), can be checked by looking at diagram
(b) which we can compare with the result given by a free proton
target. The corresponding kinematical situation is a spectator
neutron of very small kinetic energy; but if this were experimentally
feasible, it would give more confidence on the way to compute the
Glauber corrections.
LEADER:

It is certainly a good idea to test one's ideas on Glauber cor-

STRONG INTERACTIONS WITHIN THE REGGE FRAMEWORK

287

rections by attempting to derive the pp total cross-section from pd

measurements. The diagrams, however, refer to specific channels and
not to total cross-section measurements.
GOURDIN:

This question concerns (da/dt)t=O in np + pn. At low energy,

+ background seems to agree very will with experiment, but the
contribution from 1~1+~312 due to ~a # 0 is still present. Did you
check that it is small compared to the previous contribution?

TI

LEADER:

Yes, the contribution to 1m (~1+~3) via the optical theorem is

always present and cannot be cancelled. However, at lower energies,
as at high energy, ~a is small. Compared with the contributions of
TI + background at these lower energies, its effect through 1m (~1+~3)
is completely negligible. Do not forget that the contribution to
da/dt from TI + background is decreasing as 1/s2. At PL= 10 GeV/c,
da/dt ~ 700 ~b (GeV/c)-2. At 200 GeV/c, TI + background gives about
10 ~b (GeV/c)-2, and now the contribution from ~a is comparable and
therefore very important.
BERLAD:

What is the nature of the pion pole?

LEADER:

Phenomenologically, since one is at such small t when talking

about pion exchange, it does not matter whether one considers it
as a fixed pole or a moving trajectory. On the other hand, your
quastion at a fundamental level is an important one. One does not
really know what kind of object the pion is. We once attempted to
answer the question by looking at high order Feynman diagrams in
photoproduction, but the results were so peculiar that we simply
could not take them seriously.
PAULI:

(A philosophical question from a non-Reggeist.)

Suppose the data show that Regge theory cannot work without a
combination of baryonic Pomerons, and an obscene number of trajectories, and maybe some Ptolomeic epicycles, yet we cannot discard
the fundamental idea of the complex J-plane. What should theoretical
physicists do? Shoud we stick with S-matrix theory, wait for "the
unified field theory", give up trying to systematize strong interactions?

288

E. LEADER
LEADER:

I have never been a fanatical Reggeist. In 1966, Slansky and I

wrote a paper in Physics Review, "Critical tests for Regge pole theory",
suggesting 13 ways of testing the theory critically in NN ~ NN.
Unfortunately, the tests involved very difficult and esoteric spin
measurements which were quite impossible at that time. What is very
exciting is that the polarized beam and target at Argonne now make
the tests experimentally feasible.
No-one has come up with any alternative to Regge theory in the
last 15 years, nor has anyone produced a decent calculation of a
residue function. People should try to get serious and reliable
results in the theory -- then the tests would become really meaningful.
On the other hand, it may turn out that the situation in the
complex J-plane really is exceedingly complicated. We cannot forget
the J-plane. It is an established and unassailable concept in
quantum mechanics, but nothing guarantees simplicity in that plane.
That will be a pity, but we shall have to learn to live with it.

HADRONIZATION OF QUARK THEORIES

Hagen Kleinert
Institut fUr Theoretische Physik
Freie Universitat Berlin, Berlin, Germany
ABSTRACT
Local quark gluon theories are converted into bilocal
field theories via functional

techniques.

The new field

quanta consist of all quark antiquark bound states in

the ladder approximation.

They are called "bare hadrons".

Hadronic Feynman graphs are developed which strongly

resemble dual diagrams.

QED is a special case with the

"bare hadrons" being positronium atoms. Photons couple

to hadrons via intermediate vector mesons in a currentfield identity

The new theory accommodates naturally

bilocal currents measured in deep-inelastic e p scattering

Also these couple via intermediate mesons.
In the limit of heavy gluon masses,

the hadron

fields become local and describe 9(, s> ,AI ,0'" mesons in a
chirally invariant Lagrangian (the "Q""model"). Many

289

290

H. KLEINERT

interesting new relations are found between meson and

2

/'

a.

quark properties such as m.S' ~~M where M is the "true"

non-strange quark mass after spontaneous breakdown of
chira1 symmetry. There is a simple formula linking these
quark masses with the small "bare masses" of the Lagrangian. The quark masses also determine the vacuum expectations of scalar densities,
in the vacuum of
I.

These show an SU(3) breaking

16 %.

INTRODUCTION
In the attempt to understand strong interaction,

two basically different theoretical approaches have been

developed in the past years. One of them, the dual approach, is based on complete democracy among all strongly interacting particles. Within this approach, an elaborate set of rules assures the construction of certain
lowest order vertex functions for any number of hadrons l ).
The other approach assumes the existence of a local field
equation involving fundamental quarks bound together by
vector gluons 2 ) Here strong interaction effects on e1ectomagnetic and weak currents of hadrons can be analyzed in a
straight-forward fashion without detailed dynamical computations 3 ). Either approach has its weakness where the
other is powerful. Dual models have, until now, given no
access to currents while quark theories have left the
problem of hadronic vertex function intractable. Not
even an approximate bound state calculation is avaiab1e
(except in 1 + 1 dimensions 4 ) or by substituting the field
couplings by simple ad-hoc forces 5

At present there is hope that the problems connected with quark models are of a purely technical nature.
A Lagrangian field theory of Yang-Mills type seems to

HADRONIZATION OF QUARK THEORIES

291

have a good chance of defining a true fundamental theory

of elementary particles. Dual models, on the other hand,
seem to be of a more phenomenological character. Once the
fundamental vertices are determined, it is difficult to
find next corrections and to extend the prescriptions
to what might be called a complete theory. If this could
be done it would certainly have to be phrased in terms
of local infinite-component or multi-local fields 6 ).
It would be very pleasing if both models were, in
fact,essentially equivalent both being different languages for one and the same underlying dynamics. In this
case one could use one or the other 3epending on whether
one wants to answer short-or long-distance questions
concerning quarks.
In order to learn how a translation between the different languages might operate we shall consider, in these
lectures, the simplified field theory in which quarks are
colorless, have N flavours,and are held together by vector gluons of arbitrary

mass~.

This theory incorporates

several realistic features of strong interactions, for

example current algebra and peACe Moreover, the case

N-l and

r-o

includes ordinary quantum electrodynamics

(Q.E.D.). This will provide a good deal of intuition as

well as the possibility of a detailed test of our results.
We shall demonstrate how functional methods can be
employed to transform the local quark gluon theory into
a new completely equivalent field theory involving only
bilocal fields. The new free field quanta coincide with
quark-antiquark bound states when calculated by ladder
exchanges only. They may be considered as "bare hadrons".
Accordingly, the transition from the local quark-to the

292

H. KLEINERT

bilocal hadron-theory will be named "hadronization". In

the special case of Q.E.D., "bare hadrons" are positronium atoms in ladder approximation.
The functional technique will ensure that bare hadrons have exactly the correct interactions among each
other in order that hadronization preserves the equivalence to the original quark gluon theory. It is simple to
establish the connection between classes of Feynman graphs
involving quarks and gluons with single graphs involving
hadrons. The topology of hadron graphs is the same as that
of dual diagrams.

It is interesting to observe the appea-

renee of a current-hadron field identity for photons

just as employed in phenomenological discussions of vector meson dominance. Moreover, since the theory is bilocal,
this identity can be extended to bilocal currents which
are measured in deeply inelastic electromagnetic and
weak interactions.
The limit of a very heavy gluon mass can be hadronized most simply. Here the bilocal fields become local
and describe only a few hadrons with the quantum numbers
of

0" , 3t ,~ ,Ai mesons. The Lagrangian co inc ides wi th

the standard chirally invariant

0'" model which is known to

account quite well for the low-energy aspects of meson

physics. Here hadronization renders additional connection
between quark and meson properties. It also makes transparent the connection between the very small bare quark
masses (which describe the explicit breakdown of chiral
symmetry) and the mechanical quark masses (which include
the dynamic effects due to spontaneous symmetry violations).

293

HADRONIZATION OF QUARK THEORIES

II TECHNICAL PRELIMINARIES
Before we embark in our program we have to recall
certain functional techniques 7 ). They are generalizations
of the basic Fresnel integrals, valid for all real A-O

=- A

-1

(2. 1 )

(2.2)

where the complex integral

5~ ~
rJ.. ~
fi' J< "Ii' \
GO

for

stands symbolic

GO

Quadratic completion in the

-f/)

exponent yields:

t A-I.
-i. ) .1

(2.3)

, then for non-singular symmetric or

hermitian matrices A these formulas become

. (l \A \ + j

S.D~ e \

-i _~jA~lj

=~tA)

e
(2.5)

294

H. KLEINERT

i(\tA\-tj+~;~1J)

e
respectivel~

_lj1'A-'j

-1

-~A)

(2.6)

as can be seen immediately by diagonalizing

~via an orthogonal or unitary transformation (which

leaves the measures

I f finally !(J<) is a

1)"

or

function

lex)

8\ cf>!-t

invariant).

and !l'A~

is under-

stood in the functional sense as

(2.7)

these integrals may still be defined by grating the X

axis into finer and finer lattices of points
wi th

X,.,

= It.

k,s;O.-llt2.r ... and reducing the problem to the previous

case via the vec tor components

For large matrices

A,

\k!!

'()C. .) .

the calculation of the determinant

is performed most simply by expanding

d4.t A = exp ~ kt A == eX'~ tr R..n. A

(2.8)

This formula is directly applicable in the functional

case if all sums over intermediate indices are replaced
by the corresponding integrals, for example:

HADRONIZATION OF QUARK THEORIES

-h- A : ~
~

b- A

2.

The integrals
functions

-"/

S~

AIt.(. A,tk --7

A(><J>()

(2.9)

~e;(x~ A(xl~)A(~/l()

(2.5) and (2.6) can also be extended to

~6c) with values in an anticommuting algebra

{~6c)/\(~)l=O).

(Le.

equs.

LIc.L

-=

AleE

295

(2.5~(2.6)appears

In this case the (det A)

in the inverse forms

(det A)

in

1l~

and

(det A)I, respectively.

With these preparations consider now a Lagrangian composed of fermion and boson fields and split into free and
interac ti ng par t

. ("\',"+1 ce) .1:0 -t.t'14t.

All time ordered

Green's functions can be obtained from the derivatives

with respect to the external sources of the generating
func tiona1

{Stbc (~("t+''4~''t, -t}~)

1, [tt"

j]

=Gonst

<olT e

0>

(2.10. )

The fields in the exponent follow free equations of motion and

10>

is the free-field vacuum. The constant is conventionally

chosen to make Z (9,0,0]

1, i.e.

const -

This normalization may always be enforced at the very

end of any calculation such that Z

l"L,\.,j]

is only

296

H. KLEINERT

interesting as far as its functional dependence is concerned, modulo the irrelevant constant in front.
It is then straight-forward to show that

1 ['1.,,, ,j]

can alternatively be computed via the Feynman path integral formula

i~tA'C (~("t/f, ftJ"~~t . . ,"t-t.q;'(. .j~

1.. [,,';:.il 0<. jt1't.e~.ece e

(2.12)

Here the fields are no more operators but classical functions (with the mental reservation that classical Fermi
fields are anticommuting objects). Notice that contrary
to the operator formula (2.10) the full action appears
in the exponent.
For simplicity, we demonstrate the equivalence only
for one real scalar field ~N} The extension to other
" 1d s 1S
"
..
7,8) F"1rst note t h
"
f 1e
1mmed1ate
at"1t 1S
sufficient to givethe proof for the free field case, i.e.

(2.13-)

For if it holds there a simple multiplication on both sides

of (2.13) by the differential operator

(2.14)

297

HADRONIZATION OF QUARK THEORIES

would extend it to the interacting functionals

or

(2.1Q)

(2.12). But (2.13) follows directly from Wick's

theorem

according to which any time ordered product of a free field

can be expanded into a sum of normal products with all
possible time ordered contractions. This statemen can be
summarized in operator form valid for any functional~r:'fJ

Ce('t():

of a free field

.1.

F[C(J1 -

(~)(tA'1.l D(~- ~) ~

2. )

SCf(>c)

~'Ef<j)

: F[CeJ:(2.15)

is the free-field propagator

(2.16)

Applying this to (2.13) gives

:=-

-i

~~ j (,r'IlY,,:!)V:1 )

<"ol."e

Lfw j6c)W-

:/0>

-~ yl<'~ jMlX><"j) {~)

(2.17)

298

H. KLEINERT

The last part of the equation follows from the vanishing

of all normal products of

Gp(k)

between vacuum states.

Now exactly the same result is obtained by performing

the functional integral in (2.13) and using (2.5). The
matrix A is in this case

A(k, '1) =-

EOIC. -ra)b(><-j)

such that its inverse becomes the propagator

I:tk-~)

(2. 18)

yielding again (2.17).

Notice that it is Wick's expansion which supplies
the free part of the Lagrangian when going from the operator form (2.15)

to the functional version (2.IZ).

III QUARK GLUON THEORY

Consider now a system of N quarks ~(x) held together by one gluon field GV(x) of mass rvia a Lagrangian

(3. 1 )

Here

~:v is the usual curl

special case in which

N"

O"....G.. - d v G.J'"

I,? ..

0, and

In the

<t a_ LfrroC

this Lagrangian describes quantum electrodynamics. In

other cases it may be considered as a model field theory which carries many interesting properties of strong
interactions, for example approximate SU(3) symmetry,

HADRONIZATION OF QUARK THEORIES

299

chiral SU(3) xSU(3) current algebra, PCAC, and scaling

up to small corrections. Certainly this model will never
be able to confine quarks, give symmetric baryon wave
functions, and explain infinitely rising hadron trajectories. For this it would have to contain an additional,
exactly conserved, color symmetry with

G"llr)

being its

non-abelian gauge mesons. Before attempting to deal with

this far more complicated situation we shall developlO)
our tools for the less realistic but much simpler model
(3.t) without color.
The generating functional of all time ordered
Green's function is

Z[" ;L(] ex: ~lJ"tl)"fcbG e

C:)b (~+"f, -t~I\f-t, ('bJ

(3.2)

G~Jc), such that the functio7,8)

nal integration over the gluon field can be performed
The exponent is quadratic in
(using (2,5

leaving:

l
~

where the action is

.A t"f ,"I', "{, 't, i" J =

y..t~

cA (~/t/"(/~Jj~J
(3.3)

(e(xl-t"fM~M+'t"1'\'M)

)(~(x-~) - ~ ~ z D(~-~)(~)C)-V~(X)t(V(k~(~)r-.,1\f{1)-tJ'/~]
(3.4)

H. KLEINERT

300

By employing the Fierz identity:

:~

'(

b ""

1""

1~p ~ ~ '(5

hs ).~
(3.5)

- i (\ 011 rl" -~&"y.\~~v'"\~

the quartic quark interaction term can be written in a
different fashion

~ 't 2. D(x-j ){ ~()<')/\f~) ~~ )''t-(k) -t-'\f (x) l'(s-"f(~Y\t(~)()'f(~

-t .:q (Xl'6""tf'l\ "t(~ \y""it",) - ~ "'lO<)'("~..'t('\),'ic,)r."'ffj

==

~~ ~D(x-j) Af(x)"Y(~) t
'"

) oC 0)

~(b

I\f(U)"+ (X)
(J

fb

(3.6)

This is the point where our elimination of quark fields

in favor of new bilocal fields can set in.
Let S()c/~)J W~I!j).I V~l~~
an auxiliary fields, i.e.

,AV(l("ICY

be a set of hermiti-

(3.7)

With these fields one can certainly construct the

.

functional 1dent1t1es

9)

follo~ing

301

HADRONIZATION OF QUARK THEORIES

('

Jb~~e

-i (S{X,Ij)-t t 6"1..lXt_~ ~(\1) '1(~) \ fi~

.~
-i IP(XI~) -tl~ ~D~-'j )"tC~)irs"f(k)\ 2.!(11(YX_~

5d;j I{'"XI~) e -

C l VV()(11 1-

.J ~ V (k 1) e

'L [XX-Ij}

: : ClJII6t

i d D(x-'1\"-t{'j\lt'f(~} IYt1 1Dr;c1)

&

=-(IIt'lSt

S(JO A(k\ 'j\

Av()(\ ~ )

-i ~ 2.rxX_~) ;qr~)~"rs"f[x) I /f12Q:X~~}

-~.
(3.8)

which are independent of the fields ~(X). If we now multiply Zr1t'\J\"V]in (3.3) by these constants and make use
of (3.6), all quartic quark terms are seen to cancel. The
generating functional becomes

1.. h,i( J"]

0<

l~

1)"I-,e;q- .E>s aJ PS VJ:)A e

where the new actionJr

(3.9)

denotes the integral

with the bilocal Lagrangian

~()(I 'j) == {t\f()c)({~- m.) "f(>c) -+'41IC)"t(IC)+~If) "I'k) J~)

-

4()()rntK,~) ~(:1) - ~t~~-1J)jM }..,t~)' (3. 11)

-t~ \S l

'-t ~lP

l2.-(v l~-~\

'J(1 'L[X><-jj

302

H. KLEINERT

Here m(x,y) has been introduced as an abbreviation for the

combined field

= S()(I<j)

w\'(X,j)

1-

(V "('<I':\)

Due to (3.7), the

~ R)(',~)t~
-t

(3.12)

t/'''-1) ~h l~ 2.Dl<-.)r{~)~. -tA~~~~

ma~rix

m(x,y) is self-adjoint in the sen-,

At this place it is worth remarking that the Lagrangian

(3.11) shows its equivalence to the previous form (3.4)

also quite directly. By virtue of the Euler Lagrange
equations the fields

S/P,V/A

are seen to be dependent

fields coinciding with the corres-ponding bi1oca1 quark

expressions

SCX '1)::: -l:~ ~ fr)(-1j}"f(~)"f()()

I

p()(,~) ::::

-Ca-

2.

ty)(-1)i\fc1)~~S"tl')()

Vl1()c, '1) = ~ %a Ctl(-,\ i\f('j) Q11 "f()( }

A "(X', 'i)'"

(3.14)

i ~ ~ 1}><-1) tCf(~) rVrs "t(x)

Inserting these relations back into (3.11) reproduces (3.4).

In the action (3.10), quark

fie1~s

enter only in quadratic

form such that they can be integrated according to formula (2.6)

(in its fermionic version). The matrix A is in

HADRONIZATION OF QUARK THEORIES

303

this case

Hence

A-'(X'I'j)::-\6(lcj'f) becomes simply the Green's function

of the equation

With this notation, the quark integration brings the

functional

(3.9) to the form

Z [ tt.<~, f"] '" ~ cb m (>'j) e

cACm, 'l1~lr~J
(3. 16)

with

A em ,'{ i'\d "] :: ~(.A"o'd -dr(.en l G-I)(",~)6("1)

,-I

(3.17)

- ~ tr( nl()(I'1) m'-'i1)c))Jl1~ [)()(-~) + lytt~)G{ICI':S) 1{(j)

-

~O) V"'(..)QI(-1)j"{~)-'3 ~fShh' j)(a-.)Dl~-')j{'lj"l

Here we have introduced, for brevity, the notation

cern(I(I'1)=~s80
the matrix

~-1

PJ;JVi:JA .

Notice that the effect of

defined in equ. (3.6) is simply to devi-

de the projections into

S, P, V, A by 4, -4, -2, 2,

respectively+). The trace refers only to Dirac indices.

it

+) Since ~ ~e1
f~det~fi) I ~
(Iv I -k'tl'~ oV~S"
are th~-eorrecponding pro-jeetion operators.

304

H. KLEINERT

The new functional

one in equ.

(3.16) is identical to the original

(3.2). As a consequence, a quantum theory

based on the action (3.17) must be completely equivalent

to the original quantized quark gluon theory.
A word is in order concerning the internal symmetry
SU(N) among the N quarks (i-I, N) under consideration.
Since the gluon is an SU(N) ainglet, the interaction in

'l- t\f;:t"~ Gv~)

In the Fierz transformed

version (3.6) the indices i and j appear separated
equ.

(3. I) is

\=1

.
~ t ~ D{)(-Ij) r\ftX) "tJ~) ~ ~ \(~) "t{><)

Hence in the presence of N quarks, the fields m(x,y) have

to be thought of a matrices in SU(N) space

m(xly\ S.

This carries over to the action with the traces including

Dirac as well as SU(N) indices.
Let us now develop a quantum theory for the new
action. In general, the field m(x,y) may oscillate around
some constant non-zero vacuum expectation value

mo~<!tf/('-cp.

It is convenient to subtract such a value from m(x,y)

and introduce the field
(3.18)
With this and the definition
(3.19)
e q u ( 3 I 5) can be r e wr itt e n as

HADRONIZATION OF QUARK Ti-IEORIES

305

Now let us assume that the oscillations m'(x,y) are sufficiently small as to permit a perturbation expansion

G(k\i\ :

for

where

Gt-\(l\1)

'

..

(3.21)

are the usual propagators of a free fer-

mion of mass M.

Using this expansion, the action (3.17) takes the form+)

(3.22)

with

u\1 denoting

the term linear in the field m'(x,y)

(3.23)
+)

A trivial additive constant has been dropped.

H. KLEINERT

306

and

cr\2. being

The term

quadratic in m'

~Uttrm1

(3.24)
collects all remaining powers in m'

(3.25)

The last piece

A,
.0<t

finally, contains all interactions

with the externaL sources

(3.26)

For the quantization we shall adopt an interaction picture.

As usual, the quadratic part of the action,

cA&otM'J J

serves

for the construction of free-particle Hilbert space.

According to the least action principle, the free equation
of motion are obtained from

~cA~tA'J/61f\~~.( rendering

(3.27)

HADRONIZATION OF QUARK THEORIES

tln If
In ..

307

) _

P:al Pl :

and introducing relative and total momenta

together with the notation

the field equation becomes

(3.28)
In this form we easily recognize the Bethe Salpeter equa.
I I ) .1n 1 a dd er approx1mat10n
..
f or t h e vertex f

t10n
unct10ns
of quark-antiquark bound states

['1I(p ~ )~ Nit G.., (Pt\. ) ~~~ e

CPt

(olT ~l~\~c~lo,,"(lil)
(3.29)

where

t-J It

is some normalization factor.As a consequance

our free field

m'(x,y) can be expanded in a complete set

of ladder bound state solutions. These are the bare quanta

spanning the Hilbert space of the interaction picture.
Because of their bound

quar~-antiquark

nature, they will

be called "bare hadrons". In the special case of QED,

"quarkstl are electrons and the bare hadronsll are positronium
a toms.

308

H. KLEINERT

erE. (0, 4tv\2.)

For mathematical reasons it is convenient to solve

(3.28) for fixed

ling constants g2, to be called

and all possible coup-

~I\~(~a.}

i. e.

(3.30)
A useful normalization condition is

Here we have allowed for a sign factor

not be absorbed in the normalization

i~~)

NQ of

which can-

(3.29).

It may take the values +1, -lor zero.Then the expansion

of the free field m'(x,y) in terms of hadron creation
and annihilation operators

~--t(i)jClt\(~)

can be written

as

(3.32)

where

n",

are appropriate factors giving

standard normalization

ai(~)

the

HADRONIZATION OF QUARK THEORIES

hadronic state

~~(~)

0.: lo)

Now the sign factor

309

appears at the norm of the

== I H
In general there will

>.

be many states with unphysical norms since the 'bare

hadrons" are produced by ladder diagrams only and may
not be directly related to physical particles. This
situation presents no fundamental difficulty. There are
many interactions among bare hadrons which are capable
of excluding unphysical states from the S-matrix. In
fact, the equivalence of the hadronized theory to the
healthy

original quark gluon version is a guarantee

for physical results (on shell).

The propagator of the free field m'(x,y) can be found
most directly by adding an external disturbance to the
free action

This current enters the equation of motion as

(3.35)

The propagator

r-a,

OJ .......... _'~/(lc~.'t'~')~",'(~'S)
r
II., II\{JII:"~
-c',.. ('"

fined as the solution of (3.35) for the

is then de-

.-function

disturbance

(3.36)

It satisfies the inhomogeneous Bethe-Salpeter equation

310

H. KLEINERT

(3.37)
This is immediately recognized as the equation for the
two-quark trasition matrix in ladder approximation (see
equ.

(A. 19) in App. A

An explicite representation of the Green's function

in terms of the solutions

~H(P11)

of the homogene-

ous equation (3.30) can now be given.

denotes the Fourier transform

If

(3.1[)4 8('(q -'i')

~"I",,,,,~, (p, pIt, )

rJdxJ.'.f.~X '~' e
J

--J

l [PC<--<j) -+~ (){i~2.. -

(3.38)

prlC!cf)-1~!t~{)lz]

~
(x,'1;r:.;)
~ do{t>,.c'~'

it can be written as the sum over all hadron solutions:

HADRONIZATION OF QUARK THEORIES

311

where the sum comprises possible integrals over a continuous set of solutions. If quarks and gluons were seal a r s, the sum wo u 1 d bed is ere t e for

,,2. e

(0 J 4 M'2 )

since the kernel of the integral equation (3.30) would

be of the Fredholm type. A more detailed discussion is
given in Appendix A. Here we only note that a power
series expansion of the denominator

~",..".~, (P,P'\4 )=-< In-, IH (~.J~(~)r(PI~(~P'i-V(3.4o)

"Q\t"'\

rJ.~

f:>ri.

renders explicit the exchange of one, two, three etc.

gluons. Hence one additional gluon can be inserted (or
removed) by mUltiplying (or dividing)

't '&( ~ H'Z('{Z.)

(3.39) by a factor

This fact will be of use later on.

Seen microscopically in terms of quarks and gluons,

the free hadron propagator (3.39) is given by the sum
of ladders (see Fig.

I)

p!...~

P4-~

Figure 1

Graphically it will be represented by a wide band. In

the last term of Fig I we have also given a visualisation
of the expansion (3.39). Here the fat line denotes the
propagator

312

H. KLEINERT

-r

7-

"(1) . . . S~.
~rt~~a~_~l.

(3.41)

while upper and lower bubbles stand for the Bethe Salpeter vertices

r"CPl,)

and

ftt\(p 1i-1)jres p ectivel y

This picture suggests another way of representing the new

bilocal theory in terms of an infinite component hadron
fie ld depend ing only on the average position

X:=(K-t~)/~

For this we simply expand the interacting field

rn'(p\1)

in terms of the complete set of free vertex function

=L
Ii

(3.42)

Inserting this expansion into (3.22), the free action

becomes directly

implying the free propagator (3.41) for the field n~(~)

With this understanding of the free part of the action
we are now prepared to interprete the remaining pieces.
Consider first the linear part

~\Cm~.

The first

term in it can graphically be represented as whown in

Fig.

2 When attached to other hadrons it produces a

tadpole correction.

+ ..

Figure 2

HADRONIZATION OF QUARK THEORIES

When interpreted

313

the underlying quark gluon

wi~hin

picture, such a correction sums up all rainbow contributions to the quark propagator. Also the second term in

~I[.'] has a straight-forward interpretation. First

of all, the division by ,~,. D(x-~

has the eff ec t

of removing one rung from the ladder sum (such that the
ladder starts with no rung, one rung, etc.) and creating
two open quark legs. This can be seen directly from
(3.30) and (3.39): Suppose a hadron line ends at the
interaction

Then the factor

rL:l"''\
~ . a. D

):1-1

l)(-Ij~

gives (leaving out irrelevant indices)

[\j'" DJ-' r' tt f

'tit a('''}-3 ~

(3.43)

Using (3.30) this yields

(3.44)

As discussed before, the factor

il\-C,,2)(1 z.

amounts to

the removal of one rung. Multiplication by -lito

gtation over

SdPf!.1t')'t

hadron graph

and inte-

yields the total contribution of this

314

H. KLEINERT

~~ tc [C9./P1\)rrpi,foJP-i~

2ltt~(q2-)
~ \( r'fZ)-~ 1..

'<E~(~) r"(p'll)

(3.45)

As far as quarks and gluons are concerned, this amounts

to the insertion of a mass term

Ir\o on top of a ladder

graph with one rung removed (this being indicated by a

slash in Fig. 3

Figure 3

The quark gluon picture leads us to expect that m

be a cutoff dependent quantity

must

cancelling the logarith-

mic divergence in every upper loop of the ladder sum of

Fig. 2 Numerically, m

is most easily calculated by

cancelling the infinite contributed by

c:AI Cia~ to

the equa-

tion of motion (3.28). If we include<A,(IIl'l, this equation

reads

315

HADRONIZATION OF QUARK THEORIES

The first term on the right-band side is exactly the usual

l:(")(~)'"

self energy

i:Olp(p)
Normalizing

-,

JCIC(>,~,

l:(P)

e't') (~t>

'5 ~

in second order

'tp'

(?:tr)'t (r-p/)'=-r&

j"- M

(3. 47)

on mass shell one finds the usual

expression
(3.48)
is the regularized self-energy.
where ~Il
The cutoff dependent term

(3.49)
must be ba lanced by choosing

moe -.lo on the left had

side of (3.46). Also the second term

pendent:

;E

is cutoff de~

(3.50)

and a renormalization is necessary to cancel this infinity.

Most economic

is the introduction of an appropriate wave

function counter term (l~\yqF~~-~)

in the original
Lagrangian (3.1). Such a term would enter equ. (3.15) as

316

H. KLEINERT

Instead of (3.18), m(x,y) should now be assumed to oscil-

[01.

late around
By defining a new

....

(Z;'-\)(\4-M) ] 6(/f-~).

m~x,y) via

the full action (3.22) is obtained exactly as before

except for the linear partJlin which the new wave function renorma1ization term enters together with mo :

tA: r1'1\'1 = ~tJI<~ tr{ G", (><-'))m 1")('1 )

1

(3.53)

- , By choosing

~ 11\ '(Ir , ~) [rn. -t (2;'- IXi)'-f1 )] bt'' 'Il''i (}..,)

-1

~-I=-E,
the cutoff dependent term

l:,

(3.54)

is exactly compensated in the

equation of motion (3.46). After this renorma1ization

procedure, on 1y the f i ni te term

~(I.rJ;1)

is left. The

regularized action is

Using the expansion (3.42), this can be rewritten as

(3.56)

317

HADRONIZATION OF QUARK THEORIES

with

:TH(q'L) = i

S[i.f n-[~;,.c~p.\)r,p~J;Jp-1Ui;~'l
(3.57)

By momentum conservation, the tadpole momentum always

vanishes such that only

5" (0 )

is needed eventually.

Let us now proceed to the discussion of the interaction part

~tntrll\'J of equ.

(3.25). Take as an example

the term of the third order in m'. If a hadron line ends

at every ml, it can be represented graphically as shown
in Fig. 4

+A'"
Figure 4

Employing the expansion (3.42), this interaction term

can be rewritten as

318

H. KLEINERT

l tn f] =- t L

A >k.k

c)\: ltl.t

7l

~
If
~~ ~
~, (21\") \fb(q ~ ~)

~,.J"

ttl .... H,

'..Jq

(2r)"t (z.if)lf

~
r"fR.:r,t ~ 1t(~](3~
S~'tP
(zr)Lt tr [./1+
i fP-1f/'l; )G MlP+11"*1a.)

8)

G M(P-t"l,) rll(PT\: \'1, )G",,(p)1 M.f,. }111.{'{a)IIU/q,1

=

~L
J+I~ ~ r
JiA
tt

'0",

~ ('d~ /~ ;',i'l;' )M* (X')"jf~(!~}'"t (~)

~ IIa.

with a vertex function

derivatives

~~t~

"""'"'\... It.... , (-'llt~

-~";L -""\ ~,\
l~l\vJtll"lt)

I'

whose

are to be applied only to the

argument of the corresponding field

Olth,(x')

A cor-

responding formula holds for every power of mi.

Notice that the flow of the quark lines in every
interaction is anticlockwise. When drawing up hadronic
Feynman graphs it may sometimes be more convenient to
draw a clockwise flow. A simple identity helps to write
down directly the corresponding Feynman rules. Consider
a graph for a three hadron interaction and cross the
upper band downwards (s ee Fig. 5 ). The inter ac t ion appear s
now with the hadronic band~ in anticyclic order, and the
fermion lines in the hadron vertex flowing clockwise.
This is topologically compensated by twisting every
band once. Mathematically, this deformation displays the

HADRONIZATION OF QUARK THEORIES

319

Figure 5

following identity of the vertex functions

where the phase

flu

denotes the charge parity of the

hadron H. This phase may be absorbed in the propagator

characterizing the twisted band.
The proof of this identity (3.59) is quite simple.
Let C be the charge conjugation matrix. Then the vertices
satisfy:

CrHCpl~)C-I=\.H f1H~P(9)
Inserting now
and observing

CC -\
C ~)AC-'

=-<

T
(3.60)

between a 11 factors in (3.58)

T

one has

320

H. KLEINERT

Taking the transpose inside the trace and changing the

dummy variable

to -

P,

the vertices appear in anti-

cyclic order and the right hand side coincides indeed

wi th
+'\... ~ "ttl"&..'til 'VH Il-ia. H~C'1"'1.&/~) Twi s ted propaga tor s
are physically very important. They describe the strong
rearrangement collisions of quarks and certain classes
of cross-over gluon lines. Fig. 6 shows some twisted
graphs together with thei

quark gluon contents.In meson

scattering rearrangement collisions (Fig 6a) have roughly

the same coupling strength as direct (untwisted) exchanges.
In QED, on the other hand, they provide for the main
molecular binding forces.
The exchange of two twisted hadron lines (Fig 6b)
seems to be an important part of diffraction scattering
(Pomeron).
Two more examples are shown in Fig.

7. Notice that in

the pseudoscalar channel these graphs incorporate the effect of the Adler triangle anomaly.

In this connection it is worth pointing out that

all fundamental hadron vertices are planar graphs as
far as the quark lines are concerned. Non-planar graphs
are generated by building up loops involving twisted

HADRONIZATION OF QUARK THEORIES

321

a)

+ ...

b)
+

Figure 6

H. KLEINERT

322

+...

+ ...

Figure 7

propagators.
With propagator bands, their twisted modifications
and planar fundamental couplings hadron graphs are seen
to possess exactly the same topology as the graphs used
in dual models l2 ) except for the stringent dynamical property of duality itself: In the present hadronized theory
one still must sum sand t channel exchanges and they
are by no means the same. Only after introduction of
color and the ensuing linearly rising mass spectra one
can hope to account also for this particular aspect of
strong interactions.
The similarity in topology should be exploited for a
model study of an important phenomenon of strong interactions: the Okubo" Zweig" and Iizuka rule. Obviously all
hadron couplings derived by hadronization exactly respect
this rule. All violations have to come from graphs of the

HADRONIZATION OF QUARK THEORIES

323

.
d er type 13) ( for example F1g
so ca 11 e d cy 1 1n
6b). If it is

true that the topological expansion l2 ) is the correct

.
+ ),1t
.
. f
b aS1S
or l
expa1n1ng
t h1S rule
may also provide the
appropriate systematics for nrganizing the hadronized
perturbation expansion.
Let us finally discuss the external souces. From

~~~

in (3.36) we see that external fermion lines

are connected via the full propagator G wich after expansion in powers of m' amounts to radiation of any number of hadrons (see Fig. 8)

J1JlE M-U
I

Figure 8
These hadrons then interact among each other as quantum
fields. Diagrammatically, every bubble carries again
a factor

T'if(Pl~).

It has to be watched out that hadrons are always

emitted to the rigbt of each line. For example, the
lowest order quark-quark scattering amplitude should
initially be drawn as shown in Fig 9 in order to avoid
phase errors due to twisted bands. Then the graphical
rules yield directly the expression (3.39) as they should.
Afterwards, arbitrary deformations can be performed if all
twisted

+)

factors

\H

are respected.

See tne forth of Ref. 14).

H. KLEINERT

324

External gluons interact with hadrons according to

the third term in equ.

(3.26)

(3.63)
Hence every external gluon enters the hadronic world only
via an intermediate vector particle and there is a current field identity as has been postulated in phenomenological treatments of vector mesons (VMD). Here one
finds a non-trivial coupling between the gluon and the
vector mesons: As discussed before, the division by

'A2.0

amounts to a removal of one rung from the ladder

of the incoming hadron propagator and takes care of the

direct coupling of the gluon to the quarks without the
ladder corrections. This effect was shown to be accounted
for a factor

~it"1('{a.)I~"1

in the propagator sum (3.39).

Thus the direct coupling of the vector meson field trlj-f(X)

to an external gluon field

C;c~t

(X)

can be written as:

In a hadronic graph, the removal of one rung will be indicated by a slash. As an example, the lowest order contribution to the quark gluon form factor is illustrated
in Fig 10. The slash guarantees the presence of the direct coupling. The free propagator of external gluon is
given by the second term of equ.

(3.26). The lowest

radiative corrections consist in an intermediate slashed

vector mesons (see Fig 11).

325

HADRONIZATION OF QUARK THEORIES

Figure 9

+ ...

Figure 10

+ ...

Figure 11
Here the slash is important to ensure the presence of
one single quark loop.
The divergent last term in the external action
(3.26) has no physical significance since it contributes
only to the external gluon mass and can be cancelled by
an appropriate counter term.
A final remark concerns the bilocal currents as
measured in deep inelastic electron and neutrino scattering. These are vector currents of the type

326

H. KLEINERT

(3.65)

It is obvious, that also for bi10ca1 currents there is

a current-field identity with the bi10ca1 field

\I~,~

In fact, if one would have added an external source

term

C(}(.'1}

"

in the quark action:

(3.66)

this would appear in the hadronized version in the form

which proves our statement. Again, a rung has to be removed in order to allow for the pure quark contribution
(see Fig 12)

+...

Figure 12

327

HADRONIZATION OF QUARK THEORIES

Bilocal currents carry direct information on the

.
.
.
15) Th ere f ore t h e present
propert1es
0 f Regge traJector1es
bilocal field theory seems to be the appropriate tool for
the construction of a complete field theory of Reggeons l6 ),
which is again equivalent to the original quark gluon
theory. Technically, such a construction would proceed
via analytic continuation of the propagators (3.39) in
the angular momentum (and the principal quantum number)
of the hadrons H. The result would be a "reggeonized"
quark sluon theory. The corresponding Feynman graphs
would guarantee unitary in all channels. Present attempts
at such a theory &nforces

t channel unitarity onlyI7).

Also, they are asymptotically valid by construction and apparently have a chance of approximating nature only at
energies unaccessible in the near future+~
IV. THE LIMIT OF HEAVY GLUONS
As an illustration of the hadronization procedure
we now discuss
ons

18 34)

'

in detail the limit of very heavy

glu~

.Apart from its simplicity, this limit is quite

attractive on physical grounds since it may yield a reasonable approximation to low energy meson interactions.
This is suggested by the following arguments:
Suppose hadrodynamics follows a colored quark gluon
theory. In this theory the color degree of freedom is
very important for generating a potential between quarks
rising at long distances which can explain the observed
great number of high mass resonances. However, as far as
lpw-energy interactions among the lowest lying mesons

+)

See, for example, D. Amati and R. Jengo,

Physics Letters B 54 (1974).

328

H. KLEINERT

are concerned, color seems to be a rather superfluous

1 uxury:
First, many fundamental aspects of strong interaction dynamics such as chira1 SU(3) x SU(3) current algebra, PCAC (together with the low-energy theorems derived
from both) and the approximate light cone algebra are
independent of color.
Second, there is no statistics argument concerning
the symmetry of the meson wave functions as there is for
baryons I9 ).
Third, high.1ying resonances are known to contribute very little in most dispersion relations of 10wenergy amplitudes. For example, the low-energy value of
the isospin odd

3(~

scattering amplitude is given by

a dispersion integral over the mesons

';:190% accuracy20).

Similarly,

'It'j

and

<J'"

with

scattering is satu-

rated by the intermediate mesons 9r and

Ai

By looking

at all scattering combinations one can easily convince

oneself that the resonances

1r:, ~ , V' ,

Ai'

form an

approximately closed "subworld" of hadrons as far as dispersion relations are concerned. As a consequence, it
would not at all be astonishing if the neglect of color
in a quark gluon theory would not change the dynamics
when restricting the attention to this hadronic "subwor1d"+)

+) There is one estimate concerning the electromagnetic

decay of ~o_ "''(
which is based on short distance
arguments and therefore depends on co10r 21 ). However, the
same decay can be estimated also via intermediate distance
arguments, namely by using the coupling ~~~ and vector
meson dominance such that color does not come in.

HADRONIZATION OF QUARK THEORIES

329

The point is now that in the limit of a large gluon

mass

,exactly this restricted set of mesons

~OO

appears as particles in the hadronized quark gluon theory

(3.1) without color.

Thus it might be considered as some

approximation to the low-energy aspects of the colored

version. Indeed, we shall see that the hadronized theory
coincides exactly with the well known chirally invariant

cr

mod~l.

This model has proven in the past to be an

appropriate tool for the rough description of low-energy

meson physics 23 ). Our derivation of the cr model via hadronization will render several new relations between
meson and quark properties I8 ). We shall at first confine
ourselves to SU(2) quarks only, such that symmetry breaking may be neglected. The extension to broken SU(3) will
be performed afterwards.
In order to start with the derivation observe that
in the limit
a

<5

P--~OO

, the gluon propagator approaches

-function:

(4. 1 )

The equation of motion (3.27) forces m'(x,y) to become a

local field m'(x):
(4.2)

which satisfies the free field equation

(4.3)

In the local limit, the action without external sources

takes the form

330

H. KLEINERT

(4.4)

.,. 'Gf\I\rn.'~.-) stands short for

(GM

where

ltA.'i GM(X-If)CI\(('i\GM~-Jt)

etc. As before in the general

discussion, the constant
01 0 is determined by the vanishing of the tadpole parts in (4.4) which amounts to
balancing the constant contributions in the wave equation.
Due to the singularity of

G", (X-'j)

for

X->'J

this con-

dition has a meaning only if a cutoff is introduced such

that

is finite:

[G",(o)J - (

"'jJ

~[ ~

(2.11")4 P-M

""\ - M

JCl~ -

s~~
"~M;I
-2.w-'/Pe ~~
0

l")

(4.5)

M ~ 1j\:L_MJ.11'M I\'t.M ~ = M Q
(?-rr)* lJ
~ " ) a~
'Z.

:=

Here the

~pO

lL \

integration has been Wick-rotated by 90

such that the momentum

with

P" E: (-~I GtI)

real momentum

P"=(P~E)becomes ((p't.)

along the integration path. The new

(P"*, P )

HEJA and
pAt i "1. :=r _ P 2-

has been denoted by

euclidean sca 1ar product

by

n 2=
t'e

its

&.

The tadpoles can now be cancelled by setting mo equal

(4.6)

Remembering the relation to the bare quark mass

mo = M-lr(

this determines the connection between the "true" quark

.J

HADRONIZATION OF QUARK THEORIES

1Jt

mass M and the bare mass

M :=

on. -+

331

contained in the Lagrangian:

1-

4 ,:.. Q t-1

(4.7)

Equation (4.8) is often called "gap equation" because of

its analoguous appearance in the theory of superconduc. .
24)
tl.Vl.ty

Cons ider now the free par t

cA.2,. em']

of the ac t ion.

Performing again a decomposition of type (3.12) but with

the local field m'(x), it can be written in the form

~~

Lm I] = (tJ1<
:1\\ (lG) m.(k}
J +rsu,C2.) {!-In\'f,<)
1
.I

- ~ (5~-t p~) Zt~

where

S(X)

m!(x)
l.

-2A~)()) 'lJ

A(~)

and the trace runs only over

SU(2) indices. The coefficients

are given by the integrals

J ri (q)., - 4 ~ ~~..
where

ttj (P(~)

~j(1)

(f'~ ~)\M'"
(P-~t
.. M~ t.j(PI'{,l
e
e

(4.9)

denotes the Dirac traces

t.J (Pl'l)\ = 4:L -Irt)r&' lJ

l

+)

(4.8)

(i=I,2,3,4) stands short for the fields+)

p()() ( V(s<),

internal

2Vtx}

d'l.ces
The Lorentz l.n

(7 (?-+t +M)r.(~-a: i-M)~(4.IO)

l

V..", and A"'

fl.elds

are suppressed.

332

with

H. KLEINERT

"

(i=I,2,3,4) abbreviating the standard Dirac

'

covariants A , ~r.
~"
("'(5 . The traces are displayed in Appendix A equ. (A.33). Some of them grow quadratically in

P .

The corresponding integrals

J~(,)

are quadratically divergent for large cutoffs. The others

diverge logarithmically. If one introduces the basic integral

L - (d~PE

'J( 2.

J @r)'t (Pe'--tM -"Y - ~Lt

the divergent parts of

8'i (,)

(t.

o~

1\ 2.. )
M& -,

(4.11)

are (see App. A)

(4.12)

with all other integrals vanishing.

If we neglect the finite contributions as compared
with these divergent ones, the action

cA ... Cm.']

is seen

to corresponds to the local Lagrangian+)

+) Since P\(k) and ",~Jr) differ only by a Dirac scalar constant
P\o-4.~
there is no difference between primed and unprimed
fields except for S/(x)=~")-1'flo

HADRONIZATION OF QUARK THEORIES

333

~ l>< ) = 1r-$..oW { ~ S(.:) [4Q - 2J .. 4 Hz-I L -

.... i
1"

-+

P()()[ 4:Q -20

L - ~:

l' JS'('<)

1 R'x)

~ vJl>c)[\(O~""-d~~V)L -t f.~JY,,~

1 A~()()[ 1,( 0 ~J&'1_ o"''')L -t ~ M1a--L +z1~lA~(/fj

-t 2 ML ('O,...R"')A"'(~) + "jt(~)?~ R.lC) )

(4.13)

If we respect the gap equation (4.7) in this Lagrangian,

the quadratically divergent terms Q can be eliminated.
The mixed terms can be removed by introducing a new field

~)via

(4.14)

and fixing

as
(4.15)

where

2..

0\"

mA

stands short for

2..

=- my 7--t

f\1 ~

with

O\.v

l.

= 6JA ')../~~z

LJ

(4.16)
(4.17)

This substitution producas additional kinetic terms for

the pseudoscalar fields which now appears with a factor

fr-

&L(2)

(R')()O P(x))(\-t ~WttA2.~2~ 4M)..)L

~ tr5tA(~l,l~~)( ydf R'><)) ~p -I L

Using (4.15), this renormalization factor becomes

( 4. 1 8 )

334

H. KLEINERT

- I .:

(4.19)

After this diagonalization, the Lagrangian reads

y:.(X):::. 'it-Su.(2) ~ d}-'-t~/ ~S'

- (4M 2...t ~ rvtV"a rt/M js'<l.

-\<),..P~P ~pl- \mV4.(Jr({M)P'

(4.20)

-~
where

~v_

...~

are the usual field tensoIS of vector -and

~'"
axial vector fields. The particle content of this free

Lagrangian is now obvious. There are vector mesons of

mass

M"z. , axial-vee tor mesons of mass

tn.;4 'f..

and scalar

and pseudo scalar mesons of mass

(4.21)

(4.22)
With (4.17), the constant (4.19) can also be written as

~p

-I

(4.23)

As we have argued before, there is a good chance that

the fields

P , V, S ,A

describe approximately the

lowest lying mesons ?t , ~ , ~

,AI

Let us test this

hypothesis as far as the masses are concerned. Since

experimentally

mA~::;; lm~'1..

the factor

'Zt-

becomes ~

2.

HADRONIZATION OF QUARK THEORIES

Forthermore, equ.

335

(4.16) determines the quark mass as:

~ ~lO MeV

(4.24)

in good agreement with other estimatei~)The small pion

mass yields via (4.22)

lrt

15 Me..V

(4.25)

Thus the bare quark mass has to be extremely small. As10

this result has been obtained by many authors 26 ). It is
common to all models in which the smallness of the pion
mass is related to the approximate conservation of the
axial current (PCAC).
The scalar meson finally is predicted from (4.21)
to have a mass

(4.l6)
This agrees well with the observed broad resonance in
tjt"1t scattering. 2 7)20)
One disagreement with experiment appears in connection with the SU(2) singlet pseudoscalar mass (the, meson).
According to

(4~22)

it should be degenerate with the pion.

The resolution of this proble'm wi-l1 he discussed later

when the theory has been extended to SU(3).
After these first encouraging results we shall rename the fields ~ ,

V , S ,A

by the corresponding par-

ticle symbols

(4.27)

H. KLEINERT

336

where a normalization factor has been introduced in order to bring the kinetic terms in the Lagrangian to a
conventional form.
A comment is in order concerning the appearance of
a quadratic divergence in equations (4.7), (4.12). Such a
strong divergence indicates, that the limiting procedure

JA-' a:J

of equ.

(4. J) has been performed too care-

lessly. In fact, if one inserts (4.1) into the action

(3.4), the theory becomes of the

(~~)2

type and thus

non-renormalizable. In order to keep the renormalizability while dealing with a large gluon mass

)A--2.>")-

Mao

we actually have to watch out that the gluon mass stays

a.
A:a. . Then the quaalways far below the cutoff: }A
dratic divergence becomes actually of the logarithmic
type (compare (3.47~:

(which in the careless limit

"&.
fA;..p
GO

reduces again to

(4.5. The logarithmic divergence (4.11) on the other

hand becomes in this more careful treatment independent
of the cutoff which is replaced by the large gluon mass

(4.29)

337

HADRONIZATION OF QUARK THEORIES

Hence all our results refer to a renormalizable theory if

one reads both Q and L as logarithmic expression once in
the cutoff and once in the gluon mass, respectively.
Let us now proceed to study the interaction terms.
The n'th order contribution to the action is given by
(4.30)
In momentum space this can be written as the one loop
integral

-1

(P+qa'T .."t" tTM~

"(0P.:t-'\. e + M'
e

\,/p'''.'''. J'l,)

where
-t", ...
the tensor (4.10)

(4.31)

t, ;\0, (Plo .._n'

/111
'"n .

In.-.'

is the generalization of

t.'A."",. (p tQ".,
ir [rt.. (-?-i",-.....-&1 T~,-tM)P.'.,... (4.32)
-, ... tA,)!!.L
It
J

'-I

The result is hard to evaluate in general (except in a

I

+ I dimensional space). With the approximation of a

large cutoff

one may

howeve~

neglect again all contri-

butions which do not diverge. This considerably simplifies

338

H. KLEINERT

the results. Since

mials in
for

>

. (Pia-1"-1/',-"
a)

,,, 1. 1

are polyno-

of order n) the integral is seen to converge

4 For

there is a logarithmic divergen-

ce with only the leading momentum behaviour of

contributing. For

of

~II.". ~I

n~ 3

(Pia,"-1,' ",~I
a)

ti... i,

also lower powers in momentum

diverge logarithmically. A

simple but somewhat tedious calculation of all the integrals (see App. B) yields the remaining terms in the
Lagrangian. They can be written down in a most symmetric
fashion by employing the unshifted fields +) S(K)=

M+S~lIl)

rather than S', or in renormalized form

(4.33)

Then the Lagrangian reads

Here ~~
tives:

and

are the usual covariant deriva-

Dp. q-:: 4-q"" -l '( [v,..<r1- '( {A,..7t 1

(4.35)

Dp. 7t
and
+)

r- Y

rJ'-'t1

:: Opwlr
J

-t'( [Vr'it ] -+ '( {

are the covariant curls

Notice that with this notation

:a

A,... Q'""l

(tv\ -"lrt ) ~ mf()(') = - i1

~5

-t

O\(x)==m.o-+m'(x)

P.: '(5 -to VI ,,1'- of AlA ~,. '(s

HADRONIZATION OF QUARK THEORIES

The constant

l'

339

denotes

(4.37)
It describes the direct coupling of the vector mesons to
the currents, i.e. it coincides with the coupling conventionally denoted by

~g

Here

has its origin in the

renormalization of the fields. The mass term

stands short for

(4.38)

Actually, the so defined mass quantity has an intrinsic significance. This can be seen by deriving the Lagrangian in a different fashion from the beginning. consider the tadpole terms of the action

(4.39)
In the former treatment we have eliminated

mo

completely

by giving the quarks a mass M satisfying the gap equation

(4.40)
Instead, we could have introduced an auxiliary mass M
satisfying the equation

H. KLEINERT

340

(4.41)
where Qo is the same function of Mo as Q is of M. The connection between this Mo and the other masses is obtained
by inserting

M::. Mo+&M

into Q:

(4.42)
which holds exactly in S~ with only small corrections
for large cutoffs (notice that at this accuracy Lo - L).
Inserting this into (4.40) we find

(4.43)
and using

from
(4.44)
is split in a different fashion

If now

(4.45)
with a new

mo= Mo-tr(

then the propagator

Gfk,'j)

would

have an expansion

(4.46)
For this reason, the derivation of all Lagrangian terms yields exac tly the same resul ts as before only wi th

Lo '

and

Go

occurring

rather than

I\'l', M ,

ftl",~.,

Land

G.. ,

respectively. There are only two differences: First, due

to the gap equation (4.41), the scalar and pseudoscalar

HADRONIZATION OF QUARK THEORIES

341

I\A01
mass terms become
I~'
and 0 rather than (4.21),(4.22)
second, the tadpole terms in this derivation do not cancel

completely. Instead one finds from (4.39)

cA 4 r ,.'J = ~ i1- t (4Q"t-1 o=:

i:,;)m

"t"J

""f.: -fr\m lI\'r... )}: ~ "'V").I><H1!'(..~y7)

These tad pole terms provide exactly the necessary additional shifts in the fields which are needed in order to
bring the scalar and pseudoscalar masses from

~~& and

The sym-

to their correct values

~~2

and

~~~

metric form (4.34) of the Lagrangian is again reached by

introducing the original unprimed fields

Then the mass term appears as an SU(3) x SU(3) invariant

+)

With this substitution, the unprimed field

coincides with the formerly introduced field

while before

m.()() &::

~ -lTt) ~

-m

Sl(X)-+ p()(') {rS- --t .'

S(K\

~Pb(') C'(5"+ .. '

really

since now

H. KLEINERT

342

Notice now that this coincides exactly with the former calculation which rendered (see {4.38)

(2M2.Inserting here

rn.~ lnlrv{ X~~-t1tj

M= MoTbM

and (4.44) gives

2..iV\ ~- ~ rny'2..V1./Vl =2Mo\4M oOM-t 2.~M t-4t1lh~~~

= .f. Mo

(4.48)

1.

Hence the SU(3) symmetric mass Mo defined by the gap

equations (4.41) coincides with the mass M introduced
o
as an abbreviation to the mass combination (4.38).
The Lagrangian (4.34) is recognized as the standard
chirally invariant ~model. Its symmetry transformations
are for isospin

~Q"" = l

[o(/crJ

cS 1t

S V~:=& l [0<, V"'" J -t .~ arci..

For axial transformations the fields change according to

(4.50)

pte: l [0( V'J ~l.~ O""~

/

The only term in the Lagrangian which is not invariant

is the last linear term. In fact from

- r~~/~

(4.51)

HADRONIZATION OF QUARK THEORIES

343

one finds

(4.52)
Introducing the conventional pion decay constant via

(4.53)
one can read off

(4.54)
Inserting

rY\Jf:"l-

-5'1':

from (4.22) this gives

-'lJa.

~ ~

2M lL:""'

By squaring this and using

5_',.. a.=
which for

/"U'

Q~ -

~I"'.L

e..

'4

~'Z.ft:,~1

~a.

\\ =
&.

1i..
~

MA - IllS

MA~

(4.55)
"J

one obtains

(4.56)

renders the well known KSFR relation.

The model has the usual predictions

(4.57)
and

344

H. KLEINERT

When compared with experimen; the only real defect consists in the d-wave

A,~7t coupling) ~lq1(

being absent.

additional chirally invariant terms are needed in the

Lagrangian, for example, the so called

6-term:

~ if" [( t-,...~ -t FI'-~ ) rI"(;r+lrt) U"\?~-,..)- (y -.-V) 'It"""~1l:'~

(4.59)

Such terms appear in our derivation if the approximation of

large

"&

mically in

is improved by terms which do not grow lagarith-

Let us now determine the couplings of

7t ,

g ,A

~~ to external quark fields. The external propagation

proceeds via

If one defines the couplings by

)::. 0{

~JtI5l& t'f l("~ 1:"(4 "f

1\ ~

"'t

~crQG2 ~ 1:"4 "t.Q"" c:;..

__

~ ~ V(t1Q ~'(r ~ "t Vr

and can read off

-t'Adtl~ \''(\; ~"f

(4.61)

A;

345

HADRONIZATION OF QUARK THEORIES

We see the vector coupling to quarks agree with vectormeson dominance. Due to PCAC also the Goldberger Treiman
relation is respected

(4.63)
since the axial charge of the quark is
the quark mass is

Since

, the pionic coupling

to quarks is considerably smaller than to nucleons. Numerically

(4.64)
The V- meson couples even weaker
~

~Q"(IQ./4r ~

. \.+ 3

Vector- and axial-vector mesons, on the other hand, couple

as strongly as to

nucleons which is an expression of uni-

versality:

-1

2.

41r

(4.65)

We are now ready to extend our consideration to SU(3)

(and higher groups). In this case the explicit symmetry
breaking in the Lagrangian is too large to be neglected.
Thus the bare masses

lr(

of the quarks have to be con-

sidered as a matrix

(4.66)

H. KLEINERT

346

The derivation of the Lagrangian presented above (via the gal

equation (4.41

has shown the complete SU(3) symmetry of

~06 Hence when extending from SU(2) to SU(3~ no

change occurs except in the last symmetry breaking term
of (4.34). As a consequence, the mass expressions for
~

ntp

and

M.::,L

rema1n as they are only that the

0

renormalization constan~ ~p

become more complicated

SU(3) dependent quantities due to the involved mixing

of pseudoscalar and axial-vector mesons. For a complete
discussion of this SU(3) x SU(3) invariant chiral La.
grang1an t h e rea d er 1S
re f erre d to t h e reV1ew art1c 1 es 23)
o

Here we only give a few results:

A best fit to 1\ and

meson masses requires +)

(4.67)

Thus the explicit symmetry breakdown of SU(3) caused by ~

the bare masses is quite large. The standard parameter 28 )

characterizes this:

c.:
Inserting into (4.44) we find the shifts in the quark
masses caused by dynamics

1Z9

+) For other determinations of

) MeV
see Ref. 26).

(4.69)

347

HADRONIZATION OF QUARK THEORIES

and hence for the "physical" quark masses

(4.70)

Thus contrary to the large explicit SU(3) violation the

bare masses

ll1 ,

the physical quark masses M show only

the moderate violation

:: ~(M""+~.l-2.MS) ~
Il'(M ...... MA~ H!)

(4.71)

Since the quark masses M are produced almost completely by dynamical effects we expect some symmetry breakdown to appear also in the vacuum. A measure of this is
provided by the expectation values of the scalar quark
densities

(4.72)
In the hadronized

theor~

the scalar densities are identi-

cal with the scalar fields

up to a factor:

5'r t"t&) ""/)

c -

(4.73)

as can be seen most easily by considering the equations

of constraint (3.14) following from the Lagrangian (3.11)
in the 1ar g e-JA

limit. Hence

(4.74)

H. KLEINERT

348

Inserting ( 4.17) and (4.55) the factor becomes simply

such that

Lo{ tAo lo7 R::---frca.. Mel

=-f.t ~(M "'-tM~...M') ~ -8~lo~l

(4.75)

This shows that the SU(3) violation in the vacuum equals

~ - (Co e(0

t ha t i n the qua r k ma sse s

three results (4.22 )

+). Not ice t hat the

(4.44) and (4.73) are in complete

agreement with what one obtains by very general considerations using only chiral symmetry and PCAC (see App. C).
The extension of the Lagrangian to SU(3) produces
additional defects which are well known from general discussions of chiral SU(3) x SU(3) symmetry23). For example
the vector mesons

W"

ly as they should but

'f

C.f

are not mixed (almost) ideal-

remains close to an SU(3) singlet.

In general discussions, additional terms have been added

to chiral Lagrangian in order to account for this. There
are the so called "current mixing terms":

as well as "mass mixing terms"

+)

In Ref. 30),SU(3) breaking in the vacuum was neglected.

For a more general discussion and earlier references
see Ref. 29).

HADRONIZATION OF QUARK THEORIES

349

tr &!4-t PI"y-(q-Hlf )(11" -<It ) -+N J'-N')'"&-iJrYIf.bt)]

(4.77)
In our derivation these arise as a next correction to the

}L&~~

limit. Another problem is the degeneracy of the

ideally mixed isosinglet pseudoscalar meson ~(~I with the

pion. In order to account for the fact that the

~ (~)

meson is almost a pure SU(3) singlet and much heavier

than the other pseudoscalar mesons one needs some chirally
symmetric term

(4.78)

Such a term breaks PCAC for the ninth axial current. It

is well known 30 ) that the quark gluon triangle anomaly
operates in the singlet channel and might be capable of producing such a PCAC violation. In fact, if this was not
true, quantum electrodynamics would possess an exactly
massless Goldstone boson 31) with

'l

Also the term (4.78) will appear when

quantum numbers.

J4

is not any

more very large.

It is obvious that correct10ns to the ~ ~ ~

approximation will become even more important if one tries
to extend the consideration to SU(4) since then vector
and pseudoscalar masses are quite heavy.In addition, the
narrow width of the SU(4) vector meson""

1::1

seems to

indicate that short-distance parts of the gluon propagator are being probed. Thus the colorless quark gluon theory itself cannot be considered any more a realistic approximation to the colored theory.

H. KLEINERT

350

At this place we should remark that present explanations of electromagnetic mass differences require also an
32) Tho1S 1S convent10brea kd own 0 f SU (2) symmetry 1n U7
~~
0

nally parametrized by

(4.79)

From meson masses (as well as from the electromagnetic

tat.., O~

decay) one finds 33)

(4.80)
This amounts to the bare quark masses

10 2.0

4!1S

Me.'-/

(4.81)

giving the "true" masses

(4.82)

Thus the SU(2) breaking of the vacuum is very small

(4.83)

With all parameters fixed numerically we should finally check whether the approximation of a large gluon mass

HADRONIZATION OF QUARK THEORIES

351

is self consistent. From (4.55) we have

(4.84)

Inserting this into (4.29) we calculate

(4.85)
and hence

(SOO

M'to

(4.86)

or
(4.87)
It is gratifying to note that this value is much larger
than the mass of the vector mesons. In this way it is
assured that higher powers of

c:ra.I(Pe~M&)

which were

neglected in the derivation of the Lagrangian remain

really small as compared to unity for all mesons of the
theory (see app. B).
We should point out that the quark gluon theory in
the limit

~&-tDa coincides with the well-known Nambu-

Jona Lasinio 24 ) model which has proven in the past to be

a convenient tool of studying the spontaneous breakdown
of chiral symmetry and the dynamical generation of PCAC.
Those authors have demonstrated the close analogy of the
dynamic structure of this model with that of superconductivity.

As

we have mentioned before the

equation (4.40) removing the tadpoles

analogous

in the action is

to the gap equation for superconductors.

352

H. KLEINERT

A similar analogy to super-conductors exists also

for the hadronized theory. The classical version of it
corresponds exactly to the classical Ginzburg-Landau
equation for type II super-conductors in which the gap
is allowed to be space time dependent. In fact, the classical hadronized theory can be derived alternatively by
.
.
.
34,18) .
assum1ng
suc h a d epen d ence 1n
t h e gap equat10n
The advantage of our functional derivation is that the
hadronized theory is not merely some classical approximation
but becomes upon quantization completely equivalent to
t-he original quark gluon theory.

A final comment concerns the Okubo-Zweig-lizuka rule.

As argued in the general section, the meson Lagrangian
exactly respects this rule. This can be checked directly
for all interaction terms in (4.34). Violations of this
rule are all coming from meson loops. The calculation of
some important loop diagrams leads to straight-forward
estimates for the size of such violations.
V

OUTLOOK
We have shown that in the absence of color,

gluon theories can successfully be hadronized.

quark
The re-

sulting quantum field theory incorporates correctly many

features of strong interactions.

It's basic fields are

bilocal and the Feynman rules are topologically similar

to dual diagrams.Our considerations have taken place at
a rather formal level. Certainly,

there are many problems

which have been left open. For example,

there is need

for an understanding of the non-trivial gauge properties of

the bilocal theory.

Also,

a consistent renormalization

procedure will have to be developed in future investigations.

HADRONIZATION OF QUARK THEORIES

353

The inclusion of color is the challenging problem

left open by this investigation.

If color quark gluon

theory is really equivalent to some kind of dual model

the corresponding hadronization program should not produce bilocal but multilocal fields which are characterized
by the position of a whole string rather that just its
end points. A field theory should by constructed for
gauge invariant objects like

i\.y (X) ~p(l~ ~ 'j G"(~)tA~~~';f)

which depend on the whole path from x to y.
The difficulty in a direct generalization of the previous procedure is the self-interaction of the gluons.
Only after the infrared behaviour of gluon propagators
will be known, bare hadrons can be constructed inside
the corresponding potential well and the "hadronization"
methods can serve for the determination of the complete
residual interactions.
It is hoped that hadronic Feynman rules in the
presence of color will follow a pattern similar to
that found here for the non-abelian theory.
Let us finally mention that an interesting field
of applications of our methods lies in solid-state
physics. Semi-conductors in which conduction and
valence band have only small separations may show a
phasetransition to what is called excitonic insulator.
The critical phenomena taking place inside such an
exciton system will find their most appropriate description by studying the scaling properties of the bilocal field theory.

354

H. KLEINERT

APPENDIX A:

Remarks on the Fermion Bose-Salpeter Equation

Consider the four Fermion Green's function

Expanding the exponential and keeping only the ladder exchanges corresponding to the Feynman graph in Fig 13,

Xa

Xa

Y{j

Xa

Y{j

+
y'{j/~-==:::r~ X'a"

Figure 13

Yf3

355

HADRONIZATION OF QUARK THEORIES

we obtain

,,-(&t)

"=''''P', "''',()(~ )(I~')- Got_, (K-)(')G~.~ ('1 '-,)

I

..,~h' tA~, G....

(A. 3)

.c.....,)~~". ~(" ,(><,-",)'%'i".({~,) KJ<,.-p.~~~lk-,J

~ ~ J.>c..I~, .I""J.~ <i..(,...'")f.:... ~:"I"-'~ )"~""~L'""-~'J

+ ..

...'~-~')~Ift\~1
, G ..,i~-J}
,-a. ~/_'j&)lJ.a.{ll&f'aGA.
~.-.
,..,-

O(l(,-ja) [X'lC.&-'1a.) G,6'.'

+ ...

The series can be summed to the integral equation

With the abbreviation

~.~., f4'",,' = a':..-.., ~I'(o'p. = -i ",~, 1,..... ~(s~fJ r~}~;

-~ 'lI'",,~, y,. ,. '..' -i. G"'c1S ~/~r. ~,,,,.

This can be written as

356

H. KLEINERT

or, symbolically:

(A.7)

The transition matrix T is defined by removing the external particle'poles in the connected part of

g'f)

G~)",,,,,/k'1')c''j') - GoCa<'u.-.,)~.,.. 1'1'-'1)

+ 5~'C,Dl~ ,tJl)(.t:l~L Go(,c (k'-lC,)G,
,

p~

(c..'-CS J
J

TrI.'fu-a"a.~'i~tr...~(A.

8)

~rl.l)(a-)t')G' ~,~)
which may be abbreviated by

"',.

(A.9)

From (A.6) and (A.8) the transition matrix satisfies the

integral equation

(A. 10)

+ f" ...

x~ ~L D(l(,~,)\t!k{~/~
J
'.c'{JC,-lfl')T
, fl.: ('#.~".&,.(,...''i:~~~t\'f'\g.iJ

~ 'l~J r~

which is seen to coincide with the equation (3.37) for

the propagator of the bilocal field.
this equation can be written as

In a short notation,

357

HADRONIZATION OF QUARK THEORIES

(A. 1 1 )

The perturbation expansion

(A. 12)

reveals the one, two etc. photon exchanges of the ladder

diagrams. In momentum space the four particle Green's function is defined by

~v-)'+ 6 '+Cq'-~) QLtrp. P 'I,)

o/.fj"-f'

..

~b~tA.l('~1 e.

[(?+\}r -+(p~~)~ '-~~ -fit}!]

(A.13)

where the momenta are indieated in Fig. 14.

p~9.

Figure 14

358

H. KLEINERT

The corresponding scattering matrix

satisfies the integral equation

T(p, P 'I" )-hLD(p-p') ~~~ ~~ Qt1-P")G(p!.I.)

(A.14)

T(P-J P' 1,\ )b' (P!!i)

The ladder exchange is in general expected to produce quark
(H-C~

anti-quark bound states. Suppose

> is

one of them.

Inserting it into (A.I) as an intermediate state gives for

)(."

~o

">

}<:o',

'i.'

a contribution

(A. 15)

KLo
where

\T{~o'(I()~(~))'HCV")<.lf(~) \rf:t~I)"t~,(~'))lo>

G2. ~ ~'1}('l.

and

~ = ><-j

non-zero if

Using the integral representation

we have

The

funct ion is

HADRONIZATION OF QUARK THEORIES

359

(A.16)

Introducing Bethe-Salpeter wave functions

q>~~,C'j'lr' M).. <H(~) IT("+~,('1) ~("')) \6)

(A. 17)

and their momentum space versions

the four-particle Green's function in momentum space is

seen to exhibits a pole at

360

H. KLEINERT

)(0,<1_ >C:1'-JO'

The opposite time ordering

'to. - E"

a pole at

contributes

Both poles can be collected by

writing in (A.19) the factor

This factorization is consistent with the integral equation only for a specific normalization of the Bethe-Salpeter wave functions.

In order to see this write (A.7)

in

the form

Gf'+) ,.
:a.

GG T + G"G T '~2.D <:;CLt)

(( -

Q; T ~ %,LD)

-, G G; T

(A.20)

= GGr('_'~LDG;GT)-1
Suppose now that a solution is found for different values
of the coupling constant
G(Cf)

for small changes of

dG(4 )
- '0%
.&.

Then the variation of

~&.

is

(1- GGT~'tL Dyt-G T ~ 0

)C

G(Lf)

G Gr((_~~ 2.0 )-\

~ D <;(4)

(A. 2 1)

361

HADRONIZATION OF QUARK THEORIES

If one goes in the vicinity of the pole

this becomes

if

<C>~ ... s-M*Y~"') cPiI (pl'f)~~(ph)

(A.22)

== t

s- Mtt~a \

~ff, (p (,) ('

c;f PeW A,,. /1'11~) f D(Hf)~(P'lq)
~) ~}a I loft Y I )
--r;;

This can be true at the double pole at

.
cpl(PI(~ )S-~*~~)
'\ 2..z:: MIt~(3a)

only if

If we go over to the Bethe Salpeter vertex function (3.29)

(A.24)

H. KLEINERT

362

this amounts to

a~')
'dH tt2(j'1 =l[N \~ r~P~ ~1r:: (p~~)prph)
o~
'0 <3
It) ~Y)'t~r,'t -{ ~ ~
r,
l.

Gk (P-i. hl1") lXp-p'J<;;.. (P!-\ W?PI~;;'~~j

Using the integral equation (3.30) this reduces to the normalization

'a"Z~a.) = -l [NK ( -Z (' ~

~:((J

~~)L+

')

~'I~

(A.26)

,. iT- [~(P+1.)f7t(J~)G'.lP-4.)f5~f}~
This determines

)Nttl"

as

~*Z.(12) (-q~H~('~)

(A. 2 7)

-a42.
Notice that this normalization is defined for all
with some

Nf4(~l.)

N~ (".. )

For real

P(PIC()

1~

one may choose

real such that

'0(P1 4) ==

f7(P\-,)

(Both satisfy the same integral equation).

The orthogonality of

1'''(Pl,,)

and

I'tt(Pb{)

for dif-

ferent hadrons is proved as usual by considering (3.30)

once for

(1 ft D firtH

and once for

(~~&D ) -I

fl at (

HADRONIZATION OF QUARK THEORIES

363

mUltiplying the first by

r~

and the second by

, taking the trace and subtracting the results

from each other (assuming no degeneracy of

~flr('i"')

and

~,r,&)

The normalization (A.27) is seen to be con-

sistent with the expansion of the T matrix given in (3.39)

If

'13.

runs into a pole

Mila.

this expression is singu-

lar as

r"

,
. ~ 1
r~
'tl~
fl.lfII' (~P ('1 ) AI -l8'( .. M.. )'a~} ct..lI'I,) "4,1"1-,)
~.,-,
1q -. H
-1'
...
(A.29)
~'1a.

According to (A.9) this produces a singularity in

<=f~)

(in short notation)

(A.30)

which coincides with (A.19) by virtue of (A.27).

For completeness we now give the Bethe-Salpeter
equation (3.27) the form projected into the different
covariants:

H. KLEINERT

364

(A.31)

If

nl,(Pl~)(i-I,2,3,4) abbreviates S,P,V,A, one has

m.(P( )--4i
~ '( :>~

Lt

at I

2~ ~

%T-. [2~" (fl-P~;+rl (P'"t~~"M' (P'-1)~~&

(A.32)

Ie

-C'i (Pi,,)

with
and

~~ -

t'l ( P",) rrll (Pi \~ )

being the traces defined in (4.10)

(4, +4, -2, -2). Explicitly one finds

with all other traces vanishing. Notice that in the

Bethe-Salpeter equation for p\there is no tensor contribution due to the absence of such a term in the Fierz transform of

'(,.(!>~,.

. The integrals in (A.31) go directly

over into (4.9) for large gluon mass

HADRONIZATION OF QUARK THEORIES

APPENDIX B:

365

The vertices for Heavy Gluons

Here we present the calculation of the vertices

Jc .. ,

~3

,cA't

for large

)A-. As discussed in the text,

all higher vertices remain finite when the cutoff and

go to infinity in the order

};.2.

1\

~~

AA&'

,~,

and will conse-

quently be neglected.

c.A.a.

Consider first
The integrals

:J~l('1)

as described in (4.10) and (A. 33).

are evaluated by expanding

-I

JjP-t~):TM~] l(P-\.)~~M~1

-':q(P(,)

Since

the terms

-\

grow at most like

O(M'i1 'Elf J~"/Pe't)

Pe

'l.

(see (A.33

contribute finite amounts upon

integration and will be neglected. At this place we have

assumed

q~ to remain of the same order of M2. Actually

this is not true for vector and axial-vector meson field:)

but since numerically

rrt~"I. I m.A~ ~

fuoJA ~ the

neglected

terms are indeed very small.

The following integrals are needed in addition to
(4.5),

+)

(4.11)

(neglecting finite amounts)

H. KLEINERT

366

There is one subtlety connected with gauge invariance

when evaluating the integrals !JW('\) and ::JAo (,) . In

factI the first of these integrals coincides with the standard photon self-energy graph in quantum electro-dynamics.
There the cutoff procedure is known to produce a non-gauge
invariant result. The cutoff calculation yields:

(B.3)

There are many equivalent ways to enforce gauge invarianceo The simplest one proceeds via dimensional regularization
If one evaluates the integrals Q and L in D - 4
dimensions with a small

E '"70 , then

-~

367

HADRONIZATION OF QUARK THEORIES

Hence at the pole

that

6(iM2.L =0,

:::0) Q

and L become related such

cancelling the last terms in (B.3).

Notice that when dealing with the renormalizable theory

with large gluon mass ~~ ~~ )\~, this cancellation is
still present while the other Q integrals in :7~(~} become
unrelated with the L integrals, the first being essentially

JL% ~I'Y~J

the other

~~a./'1a. .

Consider now the interaction terms cA~

traces grow at most as

~E3.

Here the

Thus as far as the divergent

contributions are concerned, the denominators in the integrals (4.27) can be approximated as

1
(B.4)

368

H. KLEINERT

Since this expression decreases at least as

1/ Pe'

the traces have to be known only with respect to their

leading

Pe3

and

PSl

~~(PI1"1.)Q' ~P~

behaviours. These are

+Spp

*SPA~ PI,..,.)~ \ p2 p" + 2[ p~ [R''I.-tCfa)1

-, P1~

t~A"p(PI'~11)Q- _l'pZpp. -l P 3.(.f.+t{ a.)J4

(B.5)

t ~SV.... (P,,&,.)S' p2pl"- -P:2et~-t2P "CR","t-(a)l =tppv.J'~)

t ~V""S (p ~-a''') Q- p;&.pr-t pZ.(~ .-fct:a.)"'" = t:-PV~P (PI,." )

-t SV"V" (Plt{a.1') Q- 41-1 P""P'"t5A"-i\.,(pr"t1,)Q

M p~J--"

4 M P""P"-~MP:a.a'Y

t v,..v~v~(PIl(('a.) ~ 4 prp~p"'_ pa.(p"'~~1I:-tp.,).~}A~P~)

+2

P~P.)..q (i/t:.. + 2.. p).. Pd<:..('fl-t~,) -t" 2p}-l..p~(~,-t"'{.t

2 P"lCp(1,"T14) q/C->- - 2 P" PGlI ".)0.7;.. pz{-~(~~ of 1-";a.+'~,,,,

Using (B.2) one obtains exactly the third order terms in
the Lagrangian equ.

0-

(4.29)

(if this is written in the

'form),
The fourth order couplings in

to evaluate. Here only the leading

t'f(P(~3Iq~I'II)

~I.{ are the simplest

p+

behaviour of

contributes proportional to L
and the propagator can diractly be used in the form

-'+

[pe2-+~3.J

HADRONIZATION OF QUARK THEORIES

369

(B.6)

370

H. KLEINERT

APPENDIX C:

Some Algebraic Derivations

Here we want to compare some of our results with

1 d er1vat10ns
.
.
28,29) 0 b ta1ne
.
d b y pure 1 y a 1 ge b ratra d 1t10na
ic considerations together with PCAC.
The vector and axial-vector currents

generate chira1 SU(3) x SU(3) under which the quark gluon

Lagrangian transforms as

l.A.0-CU~-~

U3

chira1 invariant
where

u..D of c.lA.8' "1" J.. ~-eo.. rtf In.. "'t

=- L m ;q- ~ 'f
A.

== r~(m+m~'lYl')t\f~"f
+ ~ ( 1Tl"'"-+ m ~ -2\1tS)'\t" .,)..: "t
Hence

-t ~~ -1JI\. ~ )

(C.2)

"t ~: "t
(C.3)

Jd..

(C.4)

Defining also the pseudosca1ar densities

(C.S)

HADRONIZATION OF QUARK THEORIES

then l4 ~ and

OQ. form the

371

d3)

(J3) representation of

SU(3) x SUO):

(C.6)
From the equation of motion one finds the conservation
law+)

.=:

-'

\. ~

~""c

ft1 I. "f
t ,(, '\., "f

A=' I, ..., ,

(C.S)
Let us neglect SU(2) breaking in

ltt. .

By taking

(C.$

between vacuum and pseudoscalar meson states one finds

(C.9)

etc. for the other members of the multiplet, where one has
used (see the pseudoscalar version of (4.73:

372

H. KLEINERT

By writing

OUT

as

(C.IO)

equ.

(C.9) takes the form

-51t Z.t'\r-z.

-= (

~\ /

- 1/10.2..

2 i71
(l-

btt" of lfl. )/ 2- i: ~ ~"'v

(C.II)

which agrees with (4.54)

By evaluating (C.6) between vacuum states and saturating
the commutator with pseudoscalar intermediate state one
finds

and similar for the other partners of the multiplet.

Inserting the result of equ.( 4.73)

-<ol u,ct.lo") = m. O~o( C(4(o7= -\/!." mOMo.

1-

and writing M in the same way as

(C.12) to the form

1ft

in (C.IO) brings

(C.13)

373

HADRONIZATION OF QUARK THEORIES

(C.14)

which agrees exactly with (4.54 ) (written there in SU(3)

matrix form). Considerations of this type have led to
.
.
28,29,33)
the d eterm1nat10n

R:

-1.2.~
(C. 15)

or

.1-

29
Including also SU(2) violation in such a consideration
.
33)
g1ves

or

(C.16)

There are numerous extensions to SU(4)35) but they have

to be viewed with great caution since it is hard to see
how the large pseudosca1ar and vector masses occuring
there can dominate the divergence of the axial current
and the vector current, respectively.

H.KLEINERT

374

REFERENCES
I.

G. VENEZIANO; Nuovo Cimento 57 A, 190 (1968)

K. BARDAKCI and H. RUEGG; Phys. Letters,

~,

342

(1968)
M.A. VIRASORO; Phys. Rev. Letters
C.J. GOEBEL and B. SAKITA; ibid.

!!,
~,

37 (1969)
257 (1969)

H.M. CHAN; Phys. Letters 28 B, 485 (1969)

Z. KOBA and H.B. NIELSEN; Nucl. Phys.

~,

512 (1969)

Y. NAMBU; Proc. Int. Conf. on Symmetries and Quark

Models, Wayne State University 1969
H.NIELSEN; 15th Int. Conf. on High Energy Physics,
Kiev 1970
L. SUSSKIND; Nuovo Cimento
2.

~,

457 (1970)

Y. NAMBU; in Preludes in Theoretical Physics, North

Holland (1966)
H. FRITZSCH and M. GELL-MANN; Proc. XVI Intern Conf.
on High Energy Physics, Chicago 1972, Vol. 2, p.

135

W. BARDEEN, H. FRITZSCH, M. GELL-MANN in Scale and

Conformal Invariance in Hadron Physics, Wiley, New
York (1973)
D.J. GROSS and F. WILCZEK; Phys. Rev. Letters 30,
1343 (1973), Phys. Rev.

~,

3633 (1973)

H. FRITZSCH, M. GELL-MANN and H. LEUTWYLER; Phys. Lett.

~,

G.
3.

365 (1973)

't HOOFT; Erice Lectures 1975 (to be published)

J.M. CORNWALL and R. JACKIW; Phys. Rev.

~,

367 (1971)

H. FRITZSCH and M. GELL-MANN; Proc. Intern. Conference

on Duality and Symmetry in Hadron Physics (Weizman
ACKNOWLEDGEMENT
The author

1S

grateful to Dr.Annemarie Kleinert for

many inspiring discussions.

HADRONIZATION OF QUARK THEORIES

375

Science PRSS, 1971) and Proc. XVI Conf. on High Energy

Physics, Chicago, 1972, Vol. 2, p.

135

See also: R.A. BRANDT and G. PREPARATA, Nucl. Phys.

B 27, 541 (1971) and the review by
R.A. BRANDT, Erice Lectures 1972 in Highlights in
Particle Physics, ed. by A. ZICHICHI
4.

G.

't HOOFT; Nucl. Physics B 75, 461 (1974)

C.R. HAGEN; Nucl. Phys. B 95, 477 (1975)

C.G. CALLAN, N. COOTE, D.J.-GROSS; Phys. Rev.

~,

1649 (1976)
T. APPELQUIST and H.D. POLITZER; Phys. Rev. Letters
34, 43 (1974)
5.

G. MORPURGO; Physics

!,

95 (1975) and Erice Lectures

1968, 1971, 1974 ed. by A. ZICHICHI

R.H. DALITZ; Proc. Xlllth International Conference on
High Energy Physics (Univ. of Calif. Press) Berkeley,
(1967) p. 215
See also the book by J.J.J. KOKKEDEE, The Quark Model,
Benjamin, New York 1969
H. LIPKIN; Phys. Rep.

~,

173 (1973)

M. BHHM, H. JOOS and M. KRAMMER; CERN Preprint TH 1949

(1974)
6.

H. KLEINERT; Lettere Nu6vo Cimento, ~, 285 (1970)

M. KAKU and K.K. KIKKAWA; Phys. Rev.

LlQ., 1823 (1974),

~,

~,

1110 (1974),

3943 (1974)

E. CREMMER and J.L. SHERK; Nucl. Phys. B 90, 410 (1975)

7.

For a very detailed introduction see:

J. RZEWUSKI; Quantum Field Theory II, Hefner, New York,
1968
S. COLEMAN; Erice Lectures 1974, in Laws of Hadronic
Matter, ed. by A. ZICHICHI, p.

172

376

H. KLEINERT

8.

S. HORI, Nuclear Physics 30, 644 (1962)

9.

This is a generalization to bilocal auxiliary fields

of an old method of P.T.Mathews and A.Salam; Nuovo
Cimento, ll,563 (1954);

(1955), reviewed by

~.120

D.J. Gross and A.Neveu, Phys.Rev. DIO, 3235 (1974).

10.

H. KLEINERT; Phys. Letters 62B, 429 (1976)

II.

For a thorough review to

N. NAKANISHI; Progr. Theor. Phys. Suppl. 43, I

(1969)

The transition matrix T is discussed for scalar particles by

H. ZUR LINDEN; Nuovo Cimento 65 A, 197 (1970),
Phys. Rev.

~,

1335 (1971)

In the non-relativistic limit:

J. SCHWINGER; Journ. Math. Phys.

~,

1606 (1964)

Recent discussions on fermion-fermion Bethe-Salpeter

equation:
W. KUMMER; Nuovo Cimento

21,

219 (1964), 34, 1840

(1964)
K. SETO; Progr. Theor. Phys.

~,

1394, (1969)

!l, 1035 (1970)

Phys. !!, 1578 (1971)

H. ITO; Progr. Theor. Phys.

N. NAKANISHI; J. Math.
12.

K. KIKKAWA et al.; Phya. Rev. 184. 1701 (1969),

187, 2249 (1970), Phys. Rev.

~,

3258 (1970)

C. LOVELACE; Phys. Lett. 32 B, 703 (1970)

V.A. LESSANDRINI; Nuovo Cimento 2 A, 321 (1971)
G. VENEZIANO; Nucl. Phya. B 74, 365 (1974),
Phys. Letters 52 B, 220 (1974)
13.

See Ref. 2., the review articles by

H.D. POLITZER, Phys. Reports

~,

129 (1974) and

the one dimensional colored quark gluon model in

Ref. 4.

HADRONIZATION OF QUARK THEORIES

14.

377

S. OKUBO; Phys. Letters 5, 165 (1963)

G. ZWEIG; (unpublished)
J. IIZUKA; Supplement to Progress of Theor. Phys.

11,

21 (1966)
P.G.O. FREUND and Y. NAMBU; Phys. Rev. Letters 34,
1645 (1975)
G.F. CHEW and C. ROSENZWEIG; Phys. Letters 58 B, 93
(1975), Phys. Rev.

~,

3907 (1975), Nuc1. Phys.

B 104, 290 (1976)

C. ROSENZWEIG; Phys. Rev.

~,

3080 (1976)

15. H. KLEINERT; Nucl. Phys. B 65, 77 (1973), Erice Lectures 1974, in

Le~ton

A. ZICHICHI, P. 681,

and Hadron Structure, ed. by

!...l!,

526 (1974)

16. R.A. BRANDT; Nuc1. Phys. B 83, 60 (1974), Phys. Rev.

E....J..Q., 3509 (1974)

17. V.N. GRIBOV, I. Ya. POMERANCHUK and K.A. MARTIROSYAN;
Yad. Fiz. ,!, 361 (1975), Sov. J. Nucl. Phys. ,!, 258
(1966)
18. H. KLEINERT; Phys. Letters 62 B, 77 (1976)
19. O.W. GREENBERG; Phys. Rev. Letters

11,

598 (1964)

M.Y. HAN and Y. NAMBU; Phys. Rev. 139, B 1006 (1965)

M. GELL-MANN; Acta Physica Austriaca Supp1. 9, 733
(1972)
20. See any book on current algebra or
H. KLEINERT; Fortschr. Physik ,!!, 8 (1973)
21. W.A. BARDEEN, H. FRITZSCH and M. GELL-MANN;in Scale
Conformal Symmetry in Hadron Physics, ed. by R. Gatto
(John Wiley and Sons, 1973) p. 139
22. G. PREPARATA; Erice Lectures 1972, Highlights in Particle Physics, Editrice Compositori (Bologna), ed. by
A. Zichichi, p. 247

378

H. KLEINERT

23. For a detailed review see

S. GASIOROWICZ and D.A. GEFFEN; Rev. Mod. Phys.

il,

531 (1969)
Extension to see (4):
M. SINGER; Wisconsin Preprint COO-521
24. Y.NAMBU and G.JONA LASINIO; Phys.Rev.
124,

(1976)
~,345

(1961)

246 (1961)

V.G.VAKS and A.I.LARKIN: JETP iQ,282

(..!.l,192)

(1961)

Further developments:
J.D.BVJORKEN: Ann.Phys.

174 (1963)

~,

I.BIALYNICKE-BIRULA, Phys.Rev.
G.S.GURALNIK; Phys.Rev.

130,465 (1963)

136B, 1404, 1417 (1963)

H.UMEZAWA; Nuovo Cimento XL, 4S0 (1965)

Y.FREUNDLICH and D.LURIE; Phys.Rev. D8,2386 (1974)
D.J.GROSS and A.NEVEU; Phys.Rev. DIO, 3235
H.PAGELS;

Phys.Rev.

K.LANE; Phys.Rev.

~,

(1974)

3689 (1973).

DIO,2605

(1974)

R.JACKIW and K.JOHNSON; Phys.Rev. D8, 2386

P.LANGACKER and H.PAGELS;

Phys.Rev.~,

P.LANGACKER; Phys.Rev.Letters

~,

(1974)

3413 (1974)

1592 (1975)

H.PAGELS; Rockefeller University preprint COO-2232B-

102
See also:
H.MATSUMOTO, H.UMEZAWA, N.J.PAPASTAMATION;
B68,236

(1974),

Nucl.Phys.

B82,45 (1974)

L.LEPLAE, H.UMEZAWA, F.MANCINI; Phys.Rev. 10C (1974)

25. For different estimates see Ref.S and
M.K.GAILLARD, B.W.LEE, J.L.ROSNER; Rev.Mod.Physics
!!.J...,277

(1975)

G.PREPARATA; Erice Lectures 1974, 1n Lepton and Hadron

Structure, ed. by A.Zichichi
and

re f e rence s 1n

K.S.SOH; Phys.Rev.

D13,2954

(1967)

HADRONIZATION OF QUARK THEORIES

379

26. H.LEUTWYLER; Phys.Letters 48B, 45, 431, NuclPhys.

B16, 413 (1974)
See also: R.t.HEIMANN; NuclPhys. B78, 525 (1974)
27.

Compilation of Coupling Constants; Springer Tracts

in Modern Physics, Vol.55, ed. by G.HBhler

28.

M.GELL-MANN, R.OAKES and B.RENNER; Phys.Rev.

175,

(1968)
29.

P.R.AUVIL and N.G.DESHPANDE; Phys.Rev.

183, 1463

(1969)
30.

S.L.ADLER; Phys.Rev.

177,2426 (1969)

P.LANGACKER and H.PAGELS; Phys.Rev. D9,3413 (1974)

31.

Y.NAMBU; Phys.Letters

~,214

(1964)

See also:
A.D. MARIS, V.E.HERSCOVITZ and G.JACOB; Phys.Rev.
Letters 1l,313 (1964)
V. G. YAKS and A. I. LARKIN, JETP 40,792 (1961) <11,556(1961))
32.

J.GASSER and H.LEUTWYLER; Nucl.Phys. B94, 269

(1975)

33.

G.CICOGNA, F.STROCCHI and R.CAFFARELLI; Phys.Rev.

D6,301

34.

(1972)

T.EGUCHI and H.SUGAWARA; Phys.Rev. DIO, 4257 (1974)

H.SUGAWARA; Phys.Rev. D12, 3212 (1975)
H.KLEINERT; Phys.Letters 59B,

163 (1975)

A.CHAKRABARTI And B.HU; Phys.Rev. D13, 2347 (1976)

G.KONISI, T.SAITO, and K.SHIGEMOTO, Phys. Rev. DI5
(1976)
These authors prove the operatorial validity of the
"gap wave equation" by using equs. of motion rather
than the functional method employed in Ref.18 and

380

H. KLEINERT

extended to bilocal form in Ref.IO.

T.EGUCHI, Chicago Preprint,

1976.

This author studies in detail the local version

of hadronization via functional methods which was
outlined in Ref.18 on p.78.
H.PAGELS,

Rockefeller Preprints,

COO-2232 B-I02

Here a bilocal generalization of the classical gap

wave equation for colored gluon theories is proposed.
However, no bilocal field theory equivalent to the
quark-gluon theory is constructed.
35. For a typical calculation and earlier references see
Y.UEDA;

ICTP Preprint,

1976.

HADRONIZATION OF QUARK THEORIES

381

DIS C U S S ION S

CHAIRMAN:

Prof. H. Kleinert

Scientific Secretaries:

B. Freedman and B. r1athis

DISCUSSION 1

GOURDIN:
Why is colour so difficult to introduce into your formalism?

KLEINERT:
When you have colour, a pluon self-interaction coming from
l;/2
- av~ - g(~,Gv)_ will be present which cannot be
directly integrated out in the generating functional.

GOv
2

= _a~Gv

GOURDIN:
Is what you have done just a formal transformation?

KLEINERT:
No, this is a re-summation of the perturbation series that
yields results which cannot be obtained by a perturbation expansion in
the coupling. This we know from solid-state physics where one reaches
a new phase by such a technique; to describe superconductivity, one
sums bubblewise, while for a collective effect, such as plasmons, one
sums ringwise. Here, in hadron physics, we sum ladder by ladder.
It is important to note that one does not expand in powers of
the quark-gluon coupling constant. Let me remind you that in
electron-positron scattering with two-photon exchange, the parameter
in the expansion is not g, but becomes something like g4 log (s - 4m 2 ).
When s is near threshold, the effective expansion parameter becomes
large. Thus, we have to sum the whole ladder in order to get finite
results, and the ladder can be a good approximation even close to the
threshold.

382

H. KLEINERT

ROSSI:
Turning the heavy lines in your diagrams into Preparata's
double lines, one obtains the same graphs describing the same phenomena. Did you find a theoretical framework into which Preparata's
point of view can be inserted and justified, at least as far as
quark-antiquark bound states are concerned? Did you find anything
like vector-meson dominance and the direct-coupling term? What
about multihadron production in fire sausages?

KLEINERT:
My rules contain all of Preparata's rules but the reverse is
not true. For simplicity, Preparata only keeps three- and fourpoint couplings among hadrons. However, in order to have a hadronic
theory that is equivalent to the original quark-gluon theory, one must
must keep all n-point couplings. The photon couples via vector mesons, and there is an additional direct-coupling term as Preparata
wants it. The direct coupling is taken care of by slashing the propagator as I have explained in the lecture.

BUCCELLA:
Preparata's unconventional model incorporates many hadronic features, such as confinement; these are put in so as to agree with
known experimental facts. Your formulation provides a natural framework for explaining current algebra, PCAC, etc. How is confinement
and the colour-singlet nature of hadrons introduced into your functional integration method?

KLEINERT:
The quark theory that I have hadronized is not yet unconventional
because there is no confinement in it. The whole structure that I
have shown works for QED for electrons and positrons. The success
of Preparata's model indicates that this new structure can be generalized to the confined situation more easily than the original quarkgluon theory. The graphical rules for hadrons may be independent of
whether or not there is confinement in the original quark theory.

GARCIA:
In connection with the introduction of colour, you would like
to have infrared slavery? What happens to the gluon mass at long
distances?

HADRONIZATION OF QUARK THEORIES

383

KLEINERT:
This is difficult to answer within my framework because the
gluon field has disappeared from the theory upon hadronization. In
the colour quark-gluon version, there will be a massless gluon providing the long-range force to confine the quarks. However, this
will be true only within the confinement region, the hadron. Outside,
the gluon should acquire a dynamically generated effective mass, ~ +
+ 00, to prevent the quark-gluon system from escaping.

FREEDMAN:
Ultimately, when you look for the bound states in the SU(3) x
x SU(3) theory, you want to ensure that the poles correspond to hadrons and not to quarks since that is what one means by confinement.
Would you care to speculate about how you want to achieve this in
the context of your program?

KLEINERT:
There has been a very exc~t~ng talk by Migdal at the Tbilisi
conference which will be very helpful here. 11igdal has assumed
confinement and has calculated the hadronic spectrum following from
quantum chromodynamics. He looked at the vacuum expectation value
of operators like T-~(x)~(x)~(O)~(O)-, etc., Fourier-transformed
to momentum space, ~nd used the beha;iour in the far space-like
region known from asymptotic freedom. He then assumed a simple
pole structure for q2 positive and found a unique meromorphic function that fits the required asymptotic behaviour: -log (_q2)-Y,
with exponential accuracy.
- -

FREEDMAN:
This new development seems very interesting; however, the absence
of coloured states from the pole structure as used by Migdal is still
an input to the calculation.

KLEINERT:
He never looked at coloured currents.

JONES:
Could you amplify your claim that a colourless theory forms a
reasonable description of meson interactions at intermediate energies?

384

H. KLEINERT

KLEINERT:
I would claim that the low-mass mesons, the TI, p, 0, and AI,
and their low-energy, S I GeV, scattering amplitudes form a closed
set of physical phenomena. This approximation may work to within
20% accuracy. The fact that higher mass resonances do not interfere
here shows that colour, which is responsible for the existence of
these higher-mass states, can play no role at this level.

ORZALESI:
Your hadronization corresponds roughly to treating hadrons as
bound states in a ladder approximation. l1y first question is whether
your approximation is a first step in an iterative approximation
scheme and, if so, what is the prescription for calculating hadronic
amplitudes to arbitrary order? Hy second question has to do with
gauge invariance: in QED, the ladder approximation is not gauge
invariant, and the position of bound-state poles depends on the
gauge chosen; furthermore, if one only keeps ladders and rainbow
type graphs, the theory is not even renormalizable. How do you deal
with such difficulties?

KLEINERT:
My bare hadrons are certainly a first approximation, as ~s any
bare field in an interacting field theory. However, after taking into
account all hadronic interactions specified in my Lagrangian, there
is no more approximation but a complete equivalence to the original
quark theory. This also answers your second question: gauge invariance ~n the original theory has a counterpart in the bilocal theory
causing relations among hadron graphs.

FREEDMAN:
I propose that, for gauge theories, it would be useful to work
in an arbitrary covariant gauge and let the gauge parameter keep
track of the cancellation between gauge-dependent term when calculating
physical quantities. I think this is especially useful here where one
does not have an intuitive picture of the cancellations occurring ~n
the perturbation series in the effective hadronic coupling.

KLEINERT:
Yes, you will always find the correct family of hadronic diagrams
which throws out the gauge dependence.

HADRONIZATION OF QUARK THEORIES

385

PAULI:
Today, you wrote down an equation: m' = ~g2DGom'Go , which is
essentially a Bethe-Salpeter equation for quark-hadronic vertices.
Can this, or should this, give results similar to the constitutentinterchange model of Blancenbecler, Brodsky, Farrar, and Gunion for
hadron-hadron scattering?

KLEINERT:
I think you are referring to those rules that apply to rearrangement collisions. An example of this is electron exchange in molecular
collisions which give rise to the Van der Waals forces. Certainly the
rules given by Brodsky et al. can be rephrased in terms of my ladder
re-summation of scattering graphs.

POSNER:
Could you please reassure us that the expansion of
log (1 + iGom') converges?

KLEINERT:
The convergence has not been studied on rigorous grounds. I
can only assure you that my re-summation will converge better near
thresholds and bound states where normal perturbation theory certainly
fails.

POSNER:
In your process of functional integration, you first eliminated
the gluons and then the quarks. If you integrate out the fields in
the reverse order, another equivalent theory is obtained, which is
very different in appearance. What is the theory like? Is it worth
studying?

KLEINERT:
If you do it in the other order, you will get what may be called
"plasmonization". By first integrating out the fermion fields, you
will leave only a dressed photon field, which is now a very complicated
object. Plasmonization occurs by eating up all the fermion degrees of
freedom bubblewise. Although this theory plays an important role in
solid-state physics and in some two-dimensional field theories, e.g.
the Schwinger model, I have not explored it further because it seems
uninteresting for hadron physics.

H. KLEINERT

386

DISCUSSION 2
MARCIANO:

Can we understand how infrared divergences emerge from your

hadronized version of QED? Their role will be very important in
providing confinement in QCD.
KLEINER'I':

I have not examined the infrared properties of my formalism.

PHAM QUANG HUNG:

This morning you mentioned the n + 3n problem and the neutronproton mass difference. Do you have any idea how to solve these
problems?
KLEINER'I':

For these two problems one usually introduced an explicit

SU(2) breaking term into t:le quark mass matrix. Taking the d quark
to be 10 MeV heavier than the u-quark yields the correct (IDu - ~)
value. This choice also gives the correct n + 3n decay rate. I do
find this an unsatisfactory procedure, and I hope a more natural
explanation will be found.
ALVAREZ:

How do you propose to recover the sigma model in the colour

non-Abelian theory?
KLEINERT:

The sigma model can be recovered by neglecting the effect of

higher resonances. Within quantum flavour dynamics, the same thing
was achieved by sending ~2 to infinity. By this trick, quantum
flavour dynamics reduces exactly to the sigma model.
FREEDMAN:

One cannot introduce a gluon mass ~ due to renormalization requirements. What mass parameter do you plan to use to implement
confinement?

HADRONIZATION OF QUARK THEORIES

387

KLEINERT:
The only mass parameter available in the colour gauge gluonquark model, other than the quark mass, is the infrared cut-off that
one needs to define the Green's functions. Many people believe that
this parameter is related to the momentum cut-off of the quarks 1nside the confinement region.

FREEDMAN:
But this parameter is arbitrary.
hadron masses?

How can it determine the

KLEINERT:
I think that eventually one will get an infinite family of solutions, and that the infrared cut-off will be fixed by the lowest
hadron mass in the spectrum. All other masses are then determined.

WILKIE:
Are you going to have difficulties forming Regge poles, which
are long-distance effects in a ~ + 00 limit?

KLEINERT:
The ~ + 00 limit of quantum flavour dynamics gives the sigma
model which corresponds to the zero-slope approximation in Regge
theory where all higher-mass states have moved to infinity. I
assume that when colour is introduced and we make contact with the
dual aspect of hadron dynamics, the correct Regge behaviour will
be regained.

WILKIE:
1S a

Is it obvious in your treatment of the sigma model that the p

resonance?

TITI

KLEINERT:
The p appears here as a fundamental field as well as a resonance.
This is similar to the situation encountered in HID calculations;
one gets only a narrow contribution to the p when treating it as a
pure TITI system. One has to put in an elementary p as well.

BERLAD:
In performing the functional integration over the gluon field,
there should appear not only ladder exchanges but also crossed ladders. Where do these appear?

388

H. KLEINERT

KLEINERT:
This is a point I was trying to explain by showing some examples.
The cross-ladder diagrams can be obtained by topological rearrangements
of ladder diagrams.

PAULI:
In order to derive the sigma model, the mass of the gluon must
be ~ 12 GeV. This would result in point-like hadrons, yet we know
that the hadronic size is roughly the Compton wavelength of the pion.
Could you elaborate?

KLEINERT:
In this model, mesons are fundamental fields after the large
w-mass limit has been taken. Obviously, the form factor for the
mesons will be point-like. Radiative corrections due to heavy gluons
should broaden the form factor by a small amount only. However,
for this purpose, this is not a realistic model of hadrons.

VON DARDEL:
What does your model give the for magnetic moments of the baryons? Do small bare quark masses make it difficult to obtain small
enough magnetic moments?

KLEINERT:
No. The magnetic moments are determined by the dynamical quark
masses 11 ::: 312 MeV. Hence
1

WQ ::: 2H '" 32M~

becomes thrice the nucleon Bohr magnetic moment, thus coinciding with
the magnetic moment of the proton, as it should.

POSNER:
When you consider the three-point hadronic interaction, a mass
relationship involving m' is obtained. Would you give a physical
explanation of why the four-point function depends only on m', m'2,
and m' 3?

KLEINERT:
I only took leading order in W ~ 00 into account. The higher
powers of m'n (n ~ 4) are lower order corrections. They certainly
exist, but remember: }12/W 2 is very small 1%,).

HADRONIZATION OF QUARK THEORIES

389

POSNER:
The quantity C' = HelMa (:::: -16%) plays a role in TIN and KN
scattering. Please elaborate on its significance in these interactions.

KLEINERT:
In the exact SU(3) x SU(3) symmetry limit, the TIN and KN crosssections would be equal, but experiment gives (oTIN - 0KN) :::: 6 mb.
One would expect this difference in the elastic scattering amplitude,
governed by the Pomeron, to be caused by the breakdown of the symmetry
in the vacuum of the underlying quark dynamics, specifically in the
mass matrix. This is only a rough argument, not a quantitative one.

PHENOMENOLOGY OF NEUTRAL CURRENT INTERACTIONS t

J. J. Sakurai
Department of Physics, University of California, Los
Angeles, * and CERN, Geneva
1.

INTRODUCTION

It was three years ago this summer that the Gargamelle

Collaboration discovered neutral-current phenomena1--perhaps the
most important high-energy physics discovery made on this side of
the Atlantic since the discoveries of pions and strange particles
in the late 1940s. The initial phase of this three-year period
may be called the "discovery era," during which we witnessed
breathtaking "alternating neutral currents." It appears that this
first era, which may be characterized as allegro con brio,z is now
over, and we have entered a new era during which we are confronted
with the extraordinarily difficult task of determining the detailed
space-time and internal properties of the hadronic and leptonic
neutral currents. The tempo of the present era may perhaps be
described as lento.
Going through the current literature, one finds that most
discussions on neutral currents are motivated by gauge-theory considerations. Typical theoretical review papers on this subject
start by showing that neutral currents are needed because, otherwise, the cross section for
(1.1)

would grow as s.

Experimental papers on neutral-current reactions,

t Supported in part by the National Science Foundation and

the John Simon Guggenheim Memorial Foundation.
* Permanent address.
391

392

J. J. SAKURAI

after lengthy discussions of background subtractions, etc., all

end up with the conclusion
(1. 2)

It appears that theorists and experimentalists are interested only

in comparing the available data with gauge models, with the oneparameter Salam-Weinberg model,3 in particular.
I believe there is some danger in taking one-parameter models
too seriously. As I'll show in Section 5, the inclusive cross
sections for

(~) +

P -+ (-)
\i

+ any ,
(1. 3)

are potentially sensitive to as many as four coupling parameters.

When we analyze the data using a one-parameter model, e.g., the
Salam-Weinberg model, we are implicitly assuming that the remaining
three parameters either have pre-assigned values or are related in
a definite manner to the single parameter we are trying to determine. I hope this point will become transparent as we proceed.
My approach in this set of lectures is somewhat different. I
would like to propose a rather general phenomenological framework
for discussing the data without commitment to any particular theoretical model. The main objective of my lectures may be summarized
by asking three questions as follows:
(i)
What do various experiments really measure?
(ii)
How can we perform "complete" experiments to determine the
neutral-current couplings?
(iii) How can we test any model in an objective, emotionally uninvolved manner?
There are two reasons for following this rather general approach. First, let us note that the observed neutral-current phenomena are genuinely new and therefore interesting regardless of
whether gauge-theory considerations are on the right track. Until
1973 essentially the whole of weak interaction physics could be
visualized as being based on Fermi's 1934 interaction with only
minor modifications:
(i)
Change V to V-A.
(ii)
Change p and n to u and d C (= d cos8 C + s sin8 C).
(iii) Double the number of leptons:
e -+ e,

]J,

NEUTRAL CURRENT INTERACTIONS

393

With the advent of neutral-current interactions some major modifications are needed! After 40 years of physics with neutrinos,
something qualitatively new is finally happening. Whenever we have
a genuinely new phenomenon, it is profitable to study it in its
own right, without recourse to any particular theoretical framework. We may also keep in mind the possibility that renormalizable
gauge theories of weak interactions look attractive now only because theorists don't know anything better at this present moment.
Second, even if the basic philosophy of gauge models is to
triumph ultimately, there are now so many models, all within the
general gauge-theory framework, that make many different predictions on the neutral-current interactions. If we have learned anything in the past year and a half, it is that the range of model
builders' imaginations is unbounded! Once we have a general phenomenological framework, it becomes easier to compare your favorite
model with your competitors' in an objective manner.
2.

UNORTHODOX POSSIBILITIES

The original Gargamelle paperl on the neutral-current discovery is entitled "Observation of Neutrino-like Interactions without
Muon or Electron in the Gargamelle Neutrino Experiment." This
title is appropriate because the reactions they observed actually
looked like
"invisible particle" + N

-+

possible "invisible particle(s)"

+ hadrons .

(2.1)

Most physicists assume from the beginning that the invisible particles in the initial and final states are ordinary neutrinos arising
from TI and K decays and that the basic couplings involve the
neutrino current iVYA(l+yS)V interacting with hadronic or leptonic
densities of the V and/or A type. It is worth spending some time
checking these assumptions. More complete discussions of the various topics covered in this section can be found in my DESY lectures.

There is now good experimental evidence to believe that the

initial invisible particle in (2.1) is indeed the neutrino that appears in TI or K decay. First of all, all experimental papers on
neutral currents show in some detail why observed events of the type
(2.1) cannot all be interpreted as being induced by known hadrons
such as the neutron and the ~ meson. Furthermore, the observed neutral-to-charged-current ratio

0((V) + N -+ (v) + any)

0((V) + N -+ ~-(+) + any)

394

J. J. SAKURAI

is independent of the very different manners in which the neutrino

beam is prepared from experiment to experiment--wide-band, narrowband, horn-focused, unfocused--a feature not expected if the
neutral-current events were due to some yet unknown background
particles that survive a tremendous amount of shielding material.
The assumption that the outgoing invisible particle is also an
ordinary neutrino is more difficult to prove experimentally. The
fact that the neutral-to-charged-current ratio appears to be constant between CERN PS energies (Gargamelle) and Fermilab energies
(Harvard-Pennsylvania-Wisconsin-Fermilab 5 ) , as shown in Fig. 2.1,
suggests that there is no obvious threshold effect in the neutralcurrent interactions, so the outgoing invisible particle in (2.1)
is unlikely to be a neutral heavy lepton of appreciable mass. It
is, of course, difficult to rule out the possibility that a very
light neutral heavy lepton is produced in (2.1) with V + V + V as
its major decay mode.
Even if the final invisible particle in (2.1) is a massless
(or nearly massless) "neutrino," it is still meaningful to ask
whether the neutrino in the final state is the same as the initial
neutrino. This can be checked in V e scattering by examining
e

whether the s channel diagram due to the charged-current interaction

can interfere with the t channel diagram due to the neutral-current
interaction, as will be discussed in Section 3. 6 In neutrinoinduced hadronic reactions we can test the question of neutrino
identity in a more indirect manner as follows. First, it can be
shown that the equality
do
lim -- (v+N
O dy

+ any)

lim do (V+N + V + any) ,

O dy

(2.2)

where y stands for viE (the energy transfer divided by the incident
energy), must hold as long as the hadronic current is Hermitian.
[At y
0 the two sides of (2.2) are in general different because
of VA interference, but VA interference vanishes as y + 0.] Now
the Hermiticity of the hadronic current is required by the CPT
theorem in all models where the incident and outgoing neutrinos are
the same. The equality (2.2) can therefore be regarded as a test
of V identity.7

Some theorists 8 have speculated that the so-called neutralcurrent reactions are actually due to the electromagnetic interactions. Suppose the neutrino had, for some reason, an anomalously
large charge radius. It can be scattered elastically or inelastically by a nucleon via single-photon exchange. Since the neutrino
is strictly neutral, the Dirac form factor of the neutrino must
start as

(2.3)

395

NEUTRAL-CURRENT INTERACTIONS

>c::

0.4

';::1..

+;:::.

-~

0.3

f---

9- -------

>-

c::
c

;:::.

;:::.

0.1

20

40

Gargamelle

HPWF

60

80

100

<E'I/> GeV
Fig. 2.1. Neutra1-to-charged current ratio as a function of the
mean neutrino energy.

396

J. J. SAKURAI

Now, in scattering processes the q2 dependence of the leading term

in (2.3) just cancels the 1/q2 coming from the propagator of the
exchanged photon. The net result is exactly what you would expect
from a current-current interaction of the standard form. There
are, however, two difficulties with this interesting proposal.
First, with the charge radius we may compute on the basis of conventional mechanisms, the resulting cross section is a factor of
~ a 2 too small to account for the observed cross section. 9 What is
even more serious, the form factor F(q2) of V (v) must be the
same by the CPT theorem as that of

fl
e
(v) except in sign.
fl
e

So the

cross section for any v induced reaction, exclusive or inclusive,

mediated by photon exchange is predicted to be equal to the cross
section for the corresponding V induced reaction:

o(v +

A ~

v +

B)

= o(V +

A ~

V+

(2.4)

B)

In Sections 5 and 6 I'll present a substantial amount of evidence

against (2.4) in both inclusive and exclusive reactions.
In review talks on neutral currents it has become almost traditional to report on the status of "SPT heresy." A number of
authors lO speculated on the possibility that covariants other than
V and A are involved in the neutral-current (or better "neutraldensity") interactions. It is relatively easy to test pure S and/
or P by studying the y distributions in the inclusive reactions
(V)

+ N~

(V)

+ any .

(2.5)

If the total cross section for (2.5) scales, i.e., if it rises

linearly with the incident neutrino energy, as seems to be the case
experimentally, the hypothesis of pure Sand/or P predicts a y
distribution that goes as y2; in contrast, with a linear combination of V and A, the y distribution is not predicted to rise with
increasing y. The data of the Caltech ll and HPWF 5 groups who
studied the hadron energy distributions in (2.5) are in serious
disagreement with the pure Sand/or P hypothesis. It can be shown,
however, that any inclusive distribution obtained by V and A can
also be simulated by suitable linear combinations of SPT, which is
known as the "Confusion Theorem. "I 0 As a result, it is impossible
to rule out SPT by studying the inclusive distributions.
Recently an astrophysical argument has been advanced against
a sizable tensor component in the neutral-current interactions. 12
A tensor coupling between neutrinos and charged fermions of the
form vaAL V FOAlF where F may stand for a charged lepton or a quark
would lead to a finite magnetic moment for the neutrino via charged
fermion pairs which couple to the photon. The resulting magnetic
moment interaction leads to copious production of
pairs by plasma

vv

NEUTRAL-CURRENT INTERACTIONS

397

photons in stars:
Yplasma + V +

(2.6)

This has interesting implications on stellar evolution. Too high

a production rate for neutrino pair creation implies that the cooling of a white dwarf would be too rapid to be compatible with the
observed distribution of white dwarfs. To calculate the neutrino
magnetic moment we must know what kind of charged fermion pairs
couple to neutrinos but with a reasonable estimate based on the
known (?) spectrum of leptons and quarks we may deduce from the
observed astrophysical data a limit on the tensor coupling constant
at least an order of magnitude lower than the Fermi's coupling
constant G, hence insufficient to explain the observed neutralcurrent phenomena.
I may also mention that, if atomic physics parity-violation
experiments, to be discussed in Section 7, are to give a positive
result, models with SPT only will be conclusively ruled out. This
is because parity violation with SPT implies that time reversal
invariance must also be violated, but there is already a very
stringent limit on the validity of time reversal invariance from
the apparent absence of diagonal electric dipoles in atoms. 13 Furthermore parity violation with SPT does not give rise to parityviolating effects in atomic physics experiments of the kind to be
discussed in Section 7 because of the pseudo scalar term measured in
such experiments is consistent with time reversal invariance.
3.

NEUTRINO-ELECTRON SCATTERING

. Being purely leptonic, neutrino-electron scattering is the

simplest neutrino-induced reaction. It serves as a pedagogical
laboratory for exhibiting many of our theoretical concepts--the
helicity rule, VA interference, scaling, etc.--in a clean manner.
There are four reactions that can be studied:
V

+ e

+.ve + e

(3.1a)

e + e

+ V

+ e

(3.1b)

+ e

+ V

+ e

(3.1c)

+ e +

+ e

(3.1d)

The last two reactions (3.1c) and (3.1d) induced by V

and V

are

398

J. J. SAKURAI

of particular interest because they are forbidden

charged-current V-A theory; they proceed only via
current interactions. In contrast, the first two
by V and V are actually allowed also by the old
e

by the old
the neutra1reactions induced
V-A charged-

current coupling because the charged current can appear in the u

and s channels, respectively, without violating muon number conservation. All this should be clear from Fig. 3.1.
Since this is a summer school, not a conference for experts,
let me begin by dwelling on extremely simple considerations based
on angular momentum conservation. First, I define a scaling
variable y as follows:
y

= [(p"
v

where p ,p'

- p' ).p
V

lip

(3.2)

.p

eve

are the four momenta of the incident V, the

e
outgoing V and the initial electron, respectively. In the laboratory system in which the target electron is at rest, we clearly have
V

and p

viE

(3.3)

where V is the energy transfer, and E, the incident neutrino energy.

If we go to the center-of-mass system, we get, in the approximation
m
0,
e

= l2

(1 - cos8

cm

sin 2 (8

cm

(3.4)

12)

Note that the range of y is from 0 to 1.

For definiteness let us consider

]J

scattering (3.1c) and

assume that the basic neutral-current coupling is V-A, which is

probably not the case experimentally. The usual V-A (1 + YS) rule
says that at relativistic energies the interaction takes
place only when fermions are left-handed (negative he1icity) and
antifermions, right-handed (positive he1icity). It is easily seen
that the rule forbids neither forward scattering (y = 0) nor backward scattering (y = 1). See Fig. 3.2. One can, in fact, show that
the reaction proceeds only for a J = 0 spin-singlet state; the resulting angular distribution is therefore isotropic. The situation
is very different for V e- scattering. This time forward scattering
(y

]J

0) is fully allowed as in the V e- case but backward scattering

]J

(y = 1) is impossible because of angular momentum conservation. So

the configuration y = 1 is forbidden for V e- scattering with V-A.
]J

Furthermore, if you are familiar with the Jacob-Wick formalism,

you'll readily see that backward scattering is "doubly forbidden"
at 8
= TI with an angular distribution going like
cm

NEUTRAL-CURRENT INTERACTIONS

399

lie

~/
x

(lIJ e- scattering

/x~

(lI: e- scatteri ng

allowed by CC and NC

x
x
x
x

lie

CC in u channel

allowed only by CC

e-

CC ins channel
r-I

~.~ ~

~
xxxxxx

I-I

lie, I'

~e-

_c_

NC in t channel

Fig. 3.1.(v)e- scattering.

J. J. SAKURAI

400

----,;~~

"VI-'

---e

in i tia I

..

-..

forward scattering(y=O)

"VI-'

allowed

allowed

backward scattering (y =1 )

"Vp.

e-

--

allowed
Fig. 3.2. Helicity rule

in(v~e-

e
forbidden .I

lip.

scattering.

NEUTRAL-CURRENT INTERACTIONS

[dii)(S)]2

= cos4(Scm/2)

401

or, equivalently, (1_y)2.

A similar

exercise may be worked out for V+A, and we can construct a simple
table for the y distributions as follows:
V-A

V+A
(l-y)

\! e
)J

scattering

scattering

(l-y)

-\!

)J

2
(3.5)

The most general coupling with V and A can be regarded as a linear

combination of V-A and V+A. For the cross section there is no
interference between V-A and V+A--states of opposite helicities do
not interfere--and we can just add the two contributions incoherently to obtain
dO' __ 1C 12 + 1C 12 (l-y) 2 for {\!)Je}
_ _
scattering
(3.6)
dy
+\! e
)J
where C are the amplitudes for V + A. Note in particular that

for pure V or pure A we have

(3.7)
for both \! e- and ve-. Quite generally, any difference between
)J
)J
the \! and V cross section is due to VA interference; this can be
traced to the fact that the V and the A current behave oppositely
under charge conjugation.
All these results, which we have obtained using pictures, can,
of course, be more rigorously derived by explicit calculations. We
start with an effective Lagrangian

(3.8)

~C

which is the most general current-current interaction between \! and

e involving V and A as the incident \! (v) from nand K decays is
left-handed (right-handed). Throughout these lectures G stands for
Fermi's coupling constant

~ lO-5/ mp 2.

My metric is such that 1+Y5

appears with a left-handed coupling, hence gv

Using (3.8), we can derive 14
dO' = (2G 2m E/n) ~l
-dy
e
4

+gA for V minus A.

} (3.9)

402

J. J. SAKURAI

where E is the incident neutrino energy in the laboratory system,

and E m has been assumed.
e

There are two points worth mentioning in connection with (3.9).

First, both the differential cross section at fixed y and the total
cross section (integrated over y) rises linearly with E. This is
the simplest form of "scale invariance"
la Bjorken. 15 If no
masses appear explicitly at high energies, the energy dependence of
the cross section (assumed to be nontrivial) must be given by
dimensional considerations. Now daldy or a has the dimension of
11M2 in natural units; Fermi's constant G also has the dimension of
11M2. Since the cross section is proportional to G2 , the only way
to match the dimensions is to have a proportional to s, the centerof-mass energy squared. Note, in this connection,

(3.10)
for E

m
e

- to Ve cross section ratio is bounded by

Second, the Ve
1
3

<

a(ve)
a(ve)

<

(3.11)

as can be seen by using

1
(l-y)
J
0

dy
1
3

(3.12)

1 dy

The lower bound, 1/3, is obtained for pure V-A while the upper
bound, 3, is obtained for pure V+A. It may be mentioned that in
the SPT case the allowed range for the ve to ve ratio is considerably wider, between 0.139 and 7.20. 16
The scattering of V

11

and

11

by electrons was first studied in

a heavy liquid bubble chamber by the Gargamlle Collaboration using

the CERN neutrino and antineutrino beams. 17 To date there are three
likely events for V e and no candidate for v e I n terms of the
11

cross sections we have

a(v e-) < 2.6 x 10- 42 cm2 x E/GeV ,

11

11

(3.13a)

NEUTRAL-CURRENT INTERACTIONS

403

-1

-1

m>\] allowed region

Fig.3.3. Coupling constant plane for

(v0e-

scattering.

404

J. J. SAKURAI

42
1 . 0 { +2.1
-0.9 } x 10cm 2 x E / GeV

(3.l3b)

More recently, there is a counter-spark chamber experiment by an

Aachen-Padua Collaboration 18 which has reported a somewhat large
value of o(v e-) but I understand that the value depends sensitively
jJ

on the energy cuts one imposes to reduce nO background. (See

Signora Baldo-Ceolin's talk and also discussion remarks at the end
of this session.) The Aachen-Padua Collaboration also has positive
evidence for V e- scattering.
jJ

The most objective way to compare experimental data with the

Lagrangian (3.9) is to display the range of gv and gA compatible
with the data in a two-dimensional gV-gA plane.
The upper limit (3.l3a) for

jJ

See Fig. 3.3.

19

e- scattering obtained by the

Gargamelle Collaboration implies that the allowed values of gv and

gA must lie inside the ellipse denoted by vjJe.

holds for
(gv

VjJ e

0, gA

A similar remark

scattering except that here the middle region

0) is excluded because we have positive evidence for

this process.
So far I have tried to present a general phenomenological
framework for discussing (v)e- scattering.
jJ

If you like, you can

always specialize to some particular model. For example, let us

look at the one-parameter Salam-Weinberg mode1 3 which predicts
1

gA
gv

+ 2 sin 2 ew

(3.14 )

We see from Fig. 3.3 that the Gargamelle data are compatible with
sin 2 ew between 0.1 and 0.4. As will be shown in Section 5, this
range of sin 2 ew is just what we deduce by analyzing the hadronic

data. It is important to keep in mind, however, that when one

expresses the experimental result in terms of the Weinberg angle,
one is implicitly assuming that gA is equal to -1/2. It is just as
important to check this assumption as to compare the various ways
of obtaining the Weinberg angle.
Let us now consider V e- and V e
Lagrangian is taken to be e
e

scattering.

The effective
(3.15)

NEUTRAL-CURRENT INTERACTIONS

405

This Lagrangian contains the charged-current contribution as well;

even though the charged-current coupling is usually written as
(ve)(ev), it can be rearranged into the (vv)(ee) form using a Fierz
transformation. If the charged-current interaction is correctly
given by the usual V-A coupling, GV and GA can be written as
(3.16)
where the "1" stands for the charged-current contribution, and gv
and gA arise from the neutral-current contribution. Furthermore, if

~e

universality holds, the gv and gA appearing in (3.16)

(v~e

must be the same as the gv and gA appearing in the

scattering

Lagrangian (3.8). The expression for d0/dy is the same as (3.9)

apart from the substitutions gv ~ GV ' gA ~ GA
Experimentally studies of (v)e- scattering are difficult at ace

celerator energies because the

than the

(v~

flux due to

TI

and

(v~
K~2

flux due to Ke3 decay is lower

decay by about two orders of

magnitude.

However, V e scattering can be detected using antie

neutrinos from fission reactors where the antineutrino energy is
typically a few MeV. This scattering process has been, and is still
being, studied by Reines and collaborators for the past several
years at the Savannah River reactor, which is a very copious source
of v's. Their most recent results are shown in Fig. 3.4. 20
e

Notice that to compare Fig. 3.3 and Fig.

be shifted in accordance with (3.16).

3.4~

the

ellipse must

It is clear from (3.9) and its (v)e- analog that the observable
e

cross sections are invariant under gv ++ gA' GV ++ GA In terms of

the ellipses of Fig. 3.3 and 3.4, the axes of each ellipse make
angles of 45 0 and 135 0 with the gv (GV) and the gA (GA) axis. More
physically, given the observable cross section, there is still a
two-fold ambiguity; we cannot tell an axial-vector dominant solution
from a vector. dominant solution. Actually, in writing down (3.9),
we have ignored nonasymptotic terms like m IE. The coefficient of
e
meylE to be added to (3.9) is given by
(gv 2- gA2) or
(G V2-G A2)
in the (v)e case. 21 So, by measuring this term, it is, in princi-

pIe, possible to resolve the ambiguity. In addition, the magnitude

of this term in the case of V e- scattering can settle the question
e
of whether the s channel diagram indeed interferes with the t channel diagram (see Fig. 3.1), as required by models in which the
final neutral particle in the neutral current interaction is the

J. J. SAKURAI

406

1.0
1.5 - 3,0 MeV

""""""""~M""'T""-r-~""'M"~F"""rT"T""T""T'T"'II""'T""T"T""T""T""'T"""I'""T""'I GV
.. 1.0 -0.5
0.5
1.0
0.5
2.0
2.5
Fig, 3.4. Coupling constant plane for vee- scattering.

NEUTRAL-CURRENT INTERACTIONS
same as the ordinary

6
v.
e

407

In practice, however, detection of this

nonasymptotic term appears difficult.

4.

HADRONIC CURRENTS AND MODELS

Leaving the purely leptonic world, I now start discussing the

space-time and internal properties of the hadronic part of the weak
neutral currents. 22 The first question we may ask is: How many
independent coupling constants are needed to specify completely the
interaction Lagrangian responsible for neutrino-induced hadronic
reactions? If SU(2) were the ultimate symmetry of the hadronic
world, we would naturally consider four constants--isovector vector,
isovector axial-vector, isoscalar vector, and isoscalar axial-vector,
which, for later purposes, we may call a, S, y, and 0 respectively.
We are prejudiced against currents with isospin equal to two units
or higher; such objects cannot be readily constructed out of quark
fields. Besides, there is no evidence for isotensor in the electromagnetic or in the charged-current neutrino interactions.
With SU(3) we must distinguish two kinds of isoscalars--the 8th
component of an octet (hypercharge-like) and a unitary singlet
(baryon-like). It is actually simpler to use the quark model notation and decompose the isoscalar current into two parts--w like
(uu+ dd) and ~ like (5S). In either way of counting there are now
6 independent constants altogether. If you like charm and other
flavors, you may speculate on how cc, tt, bb like components enter.
So there are as many as 8, 10, 12 ... constants to be determined-even more, if you start considering flavor-changing neutral currents.
To set up the normalization convention, I write down the effective Lagrangian appropriate for reactions induced by left-handed
(right-handed) neutrinos (antineutrinos) using the quark model notation:

G - -- vyA(l+yS)V

[UyA(y+oyS)u + dYA(Y+OYS)d] + 5(Y'+0'Y S)s

+ possible cc, tt, bb terms}

(4.1)

As in the ve Lagrangian, a = +S, Y = +0 for V minus A. I may mention that every theorist--or even every paper written by the same
theorist--uses a different normalization convention for isoscalar
currents. 23 I would like to urge every author to specify how his or
her currents are normalized.

Don't just write down J~O) without ex-

plaining what it means, say, in terms of quark fields.

The only

408

J. J. SAKURAI

practical way of punishing violators of this rule is, in my oplnl0n,

not to quote those papers which do not explain clearly how the currents are normalized!
Various models can be classified by specifying a, S, y, ....
For example, in the one-parameter Salam-Weinberg mode1 3 we have

1 ,

(4.2)

When this model is supplemented by the GIM 24 mechanism for eliminating the stangeness-changing neutral currents, we get predictions for
y' and 0' as follows:

- -1 + -2

y'

. 28

Sln

(4.3)

W '

When we compare experimental data with the Salam-Weinberg model, we

are implicitly assuming that S is unity, 0 is zero, and a and yare
related by

a-I

= -

(4.4)

Perhaps some experimentalists or theorists prefer to analyze the

data without such constraints. It is to such physicists that my
approach to the neutral currents is primarily addressed.
If you are interested in other models, you may look at Table
4.1. I don't expect you to digest this Table in detail; the important point is that various theoretical models predict very different sets of coupling constants. In gauge models there is a
strength parameter A that depends on the details of the Higgs mechanism; it is fixed once we specify how mZ and ~ are related. For
historical reasons, in the Salam-Weinberg model that strength parameter is usually set equal to unity so that we have a simple relation
(4.5)

In addition, when the data are analyzed using the two-parameter

version of the Salam-Weinberg model, the value of the strength
parameter indeed appears to be roughly unity after all.
In Table 4.1 I did not list the predictions for y' and 0'.
With the exception of the isovector V-A model of Adler and Tuan,29
which gives y' = 0' = 0, all the models in the Table predict

y'

y-a
2

0'

o-S
2

(4.6)

Isov. V_A29

G_S(C)28

G_S(B)28

AKW27

t coS2~ 2 sin2 ew)

(t - 2 sin 2ewJ
1..(1 + t cos 2 i - 2 sin 2ew)

1..(1 +
cos2~)

-1 A cos 2 -a
2
2

1:..1..
2

1
1..(1 - "2

Vectorlike 26 1..(1 - sin2ev)

1 - 2 sin 2e

a
sin2e
W

sin 2ew +

t COS2~)

. 28 )
1..(1 - -1 cos 2 -a - -2 Sln
W
2
2
3

2
1..(1:.
2 - 13 sin eWJ

1..(-

1
- - A sin 2e
3
V

1:. A
3

_1

Classification of models.

EM 25

S-W3

Model

Table 4.1

A
- 1:.
2 A cos

- "2

'2a

- "21 A COS2~

-0

en

oZ

::0

-I

-I

::0
::0

()

::0

-I

J. J. SAKURAI

410

This is a simple consequence of the fact that in all these models

dd and 5S appear in a definite combination:
(4.7)

which is invariant under a Cabibbo rotation (a particular form of U

spin rotations)
d

-+

dC

-+

Sc

d cos8 C + s sin8 C

d sin8 C + s cos8 C

(4.8)

The requirement (4.7) is automatically satisfied in gauge-type

models which eliminate the strangeness-changing neutral currents
via the GIM trick. 3 a
5.

NEUTRINO-INDUCED INCLUSIVE HADRONIC REACTIONS

We are now in a position to discuss various hadronic reactions.

Let us begin with the inclusive reactions
(-)

p -+ (-V)

+ any ,

(-)

+ any

(5.1)

The basic kinematical variables are: (i) E, the incident neutrino

energy, (ii) q2, the invariant momentum transfer squared, and
(iii) V, the energy transfer. Alternatively, we may use the two
scaling variables

(5.2)

and q2. If scaling holds, (l/E)do/dxdy is a universal function of

x and y, independent of q2.
Can we completely determine a, S, y, and 0 by studying these
inclusive reactions? The answer turns out to be: almost yes, but
we need data on protons, not just on isoscalar targets. I say
"almost" because there are quadratic ambiguities that can be
settled only by studying exclusive reactions, as will be discussed
shortly.
Provided that we are below "new thresholds" in the charged as
well as in the neutral current reactions, the constants a, S, y, and
o can be extrac ted from the following "master equations" 3h 32:

NEUTRAL-CURRENT INTERACTIONS

-vp
Vp
Vn
\in
(daNC + daNC) + (daNC + daNC)
vp
\in
(dacc + dacc) +

\in

(da~ + dacc)

vp
Vn
vp
\in
(daNC + daNC) - (daNC + daNC)

411

(a 2+S 2+y2+02) + -ss corrections

1
4

(S.3a)

-vp

-Vn =z

vp
Vn
(dacc + dacc) - (da cc + dacc)

(as+yo)

(S.3b)

vp
Vn
Vn
(da vP + daNC) - (daNC + daNC)
NC

Vn

vp

vp
Vn
(dacc + dacc) - (dacc + dacc)

Vn
Vn
vp
(davP - daNC) - (do
- daNC)
NC
NC
vp
Vn
Vn
vp
(dacc - dacc) - (dacc - dacc)

-z

(ay+So) ,

(S.3c)

(aMSo) ,

(S.3d)

-z

In (S.3a)-(S.3d) NC and CC refer to the neutral-current and chargedcurrent inclusive reactions, respectively, and do may stand for the
differential cross sections da/dxdy, da/dx (y integrated out) or
even the total cross section; in the last case we must be a little
careful with the "ss corrections," which are x dependent, as will be
seen later.
The proof of these relations is extremely simple within the
framework of the valence quark mode1 33 that ignores the sea of
quark-antiquark pairs as well as strange quarks and antiquarks. It
is sketched in Appendix. Even with the sea, as long as the sea is
SU(2) symmetric, there is no correction due to the ordinary (nonstrange) qq pairs which affect the neutral- and charged-current processes in the same way. The only corrections we get are due to S8
pairs which affect just (S.3a); when explicitly written the correction term reads
ss corrections

(y,2 + 0,2) S(x)

Q(x) + Q(x)

(S.4)

where S(x), Q(x), and Q(x) are, respectively, the distribution functions for the strange quark (or antiquark), the quark distribution
function for u and d averaged, and the antiquark distribution function for
and d averaged. There is reason to believe that the ss
corrections are important only for x ~ O.lS and, in any case, affects the total cross section by at most ~ 3% in most models. The

412

J. J. SAKURAI

actual contribution can even be smaller particularly because a comparison between electroproduction and the charged-current reactions
reveals that the strange quark distribution function S(x) is smaller than is predicted by the hypothesis of an SU(3) symmetric sea.
The original quark parton model is constructed to accommodate
(or explain?) scaling. There is now some experimental evidence
against exact scaling in both inelastic electron-nucleon scattering 34
and the charged-current inclusive reactions. 35 So we may naturally
ask: How would (S.3a)-(S.3d) be affected? If scaling is violated
in such a way that the relative importance of the sea and the
valence quarks varies at high.q2 or high E, then (S.3a)-(S.3d) may
still hold provided, of course, that the differential cross sections
for the charged- and neutral-current reactions are compared at the
same values of q2 or E. On the other hand, if scaling violation is
due to new threshold channels opening up, then (S.3a)-(S.3d) must
clearly be modified; so it is safer to work at energies not too
high for the purpose of extracting a, S, y, and o. It is also worth
mentioning that the isovector part of (S.3a) and (S.3b) follows from
charge independence and chiral symmetry alone, and to fix the scale
of the isoscalar contributions we do not really need the full
machinery of the quark parton model even though some quark model
arguments are needed--of the kind that leads to a(pN) = a(wN).36 In
summary I believe that the results (S.3a)-(S.3d) are more general
than the derivations based on the quark parton model.
Physically the four ratios (S.3a)-(S.3d) measure. respectively,
the overall strength, VA (but not isoscalar-isovector) interference,
isoscalar-isovector (but not VA) interference, and VA and isoscalarisovector interference. These relations are intuitively reasonable
if we recall that the difference between the neutrino and antineutrino cross sections is a measure of VA interference while the
difference between the proton and neutron target cross sections isolates isoscalar-isovector interference.
Our information is quite incomplete when only data with isoscalar targets (targets with equal numbers of protons and neutrons)
are available; we can then determine just the first two of the four
ratios. Clearly no matter how skillful an experimentalist you may
be, you cannot extract 4 constants from 2 ratios! At the end of this
section I'll mention what we currently know experimentally about the
two ratios.
When data on proton targets are available, we'll have four
measured ratios and four coupling parameters to be determined. So,
apart from the quadratic ambiguities to be discussed in a moment, it
is, in principle, possible to determine the four constants. It is
very important to note that each of the four relations is supposed
to hold at every value of x and y; we must therefore have four "universal" ratios independent of x and y everywhere except for a small

NEUTRAL-CURRENT INTERACTIONS

413

<
correction in the region x ~ 0.15 for (5.3a). This is actually a
very powerful consistency check of our basic assumption that the
neutral and charged currents are related in the manner indicated by
the usual quark field currents. If the neutral current phenomena
are due to SPT or second class V, A, or if there are "new thresholds"
opening up, universality of this kind is not expected.
I now discuss the "ambiguities." Apart from the overall sign
ambiguity, the right-hand sides of (5.3a)-(5.3d) are invariant under
(i)

VA ambiguity:

(ii) isoscalar-isovector ambiguity:

-+-+

8, y

-+-+

8,

(5.5)

-+-+

y, 8

-+-+

8.

(5.6)

From the inclusive distributions alone it is impossible to resolve

these two ambiguities. When one compares data with the SalamWeinberg model, one is implicitly assuming that the 8 > a, 8 = 0
solution is the right one.
Equation (5.3a) defines a sphere in a four-dimensional a, 8, y,
8 space while (5.3b)-(5.3d) define three hypersurfaces. Our objective then is to obtain the common intersection points of the sphere
and the three hypersurfaces. Unfortunately our human mind is not accustomed to visualizing hypersurfaces in four dimensions. For pedagogical purposes let us look at a more simplified (if unrealistic)
situation where the isoscalar constants y and 8 are set equal to
zero by assumption--"pure isovector models." Equation (5.3a) then
defines a circle in a two-dimensional 0.-8 plane while (5.3b) defines
hyperboles. We look for the intersection points of the circle
(0. 2 +8 2 = constant) and the hyperbolas (0.8 = constant). This is
shown in Fig. 5.1. The existence of the VA ambiguity is evident from
the figure; it amounts to the question of choosing between the intersection point near the a. axis (vector dominant solution) and the one
near the 8 axis (axial-vector dominant solution). The ambiguity
disappears when the hyperbolas just touch the circle; VA interference
then becomes maximal, and we have pure V-A (or pure V+A if the
hyperbolas turn out to lie in the 2nd and 4th quadrants).
One of the most important questions in this field is: How large
is isoscalar-isovector interference or, equivalently, the p-n difference in the V and V induced neutral-current reactions? Within
the framework of the valence quark model it is easy to relate the
p-n difference in the neutral-current reactions to the p-n difference
in electroproduction. The reason is that in both cases the difference is proportional to U(x)-D(x) where U and D stand for the quark
distribution functions for the u type and the d type quark, respectively, inside the proton. We obtain 31

----

"

"

"-

/ \\

',/

'"
\
\
\

",,4

, t--)(

'"
"

c..

:::J

Q)

'\. ,

\.,/

'\. ,

"--

""

"
,
,

'"

pure V

,"I::?-/ ",

\
\
\

4 [ cr(v-v) +0- (V"- v)

1
-----

a~ =

cr (V-'fL-)-O-(V--fL~

2[cr(v-v)-0-(v-v)]

cr (V-I.e) tcr (V-fLt)

----a~~2=~----------

--

Fig. 5.1. Coupling constant determination in pure isovector models.

"

en..

:0

en

c....

!-

.....

.....

415

NEUTRAL-CURRENT INTERACTIONS

Vn
dONe
~dx
Vp
Vn
dONe
dONe
--+-dx
dx
Vp
dONC

Vp
dONe

=3

[ F;P(X) - F;n(x) ]
( 2aX+2So+a8+SX)
(a 2+S 2+y2+o2+aS+yo)
F;P(x) + F;n(X)

(5.7)

Vn
dONe

~-~

Vp
Vn
dONe
dONe
--+-dx
dx

=3

(2ay+2So-ao-Sx)
(a 2+S 2+y2+o2-aS_yo)

[Fi(X)

- Fen(x) ]

F;P (x) + Fin(x)

Even though these results can be most straightforwardly derived

using the valence quark model, much of the above relations can be
shown to follow also from the assumptions of generalized eve and
chiral symmetry.
Armed with these relations, we can solve for the n-to-p ratio
as a function of x. Model predictions are shown in Fig. 5.2 where
the electroproduction n-to-p ratio needed was taken from the SLACMIT data. 37 ,38 In the electromagnetic current (EM) mode1 25 the
n-to-p ratio is, of course, the same as in electroproduction. In
models with pure isovector currents the ratio is obviously equal to
unity at all x. Isoscalar-isovector interference in the SalamWeinberg mode1 3 is not very large if sin 28w ~ 0.35. It arises primarily from the dominant isovector axial amplitude (S term) interfering with the small isoscalar vector amplitude (y term); it therefore changes sign as we go from V to
If the neutral current is
of the pure dd form, as in vectorlike models 26 with sin 2 8V = 3/4,

v.

then the n-to-p ratio is predicted to become very large, as indicated by the top curve of Fig. 5.2. To see the physical origin of
this spectacular behavior, let us recall that experimentally the
n-to-p ratio in electroproduction appears to approach 1/4 as x approaches 1, which, in the quark model, is possible only if D(x) + O.
So, in a model in which the neutral current interaction explores
only the d type quark distribution inside the proton, the n-to-p
ratio goes to 00 as x + 1.
One can also compute the total cross section ratios, as shown
in Table 5.1. It is seen that, when we integrate over x, the isoscalar-isovector interference effect is less spectacular, which illustrates the importance of studying the x distribution.
VP/ 0ee
vp with
Neutrino physicists often talk of comparing R
- ONe
vN vN
vp
- 0Ne/OeC where N stands for the average of proton and neutron.

J. J. SAKURAI

416

,....

eIII

"

I\I:II~

-- --

"0"0

"0"0

...

,....

"

.z..

III

--

if

~I:I~
"0"0

.:::J

"0"0

s-w
(v)

ISOVECTOR

=(v)

sin 2 9 w =0.35

S-w

(v)

E-M

0.5

0.6

0.7

0.8

Fig. 5.2. Cross section ratios in the inclusive reactions (5.1).

417

NEUTRAL-CURRENT INTERACTIONS

Table 5.1.
Cross section ratios in the inclusive reactions (5.1).
vn / Vp
NC NC

Pure isovector (pure isoscalar)29

vn vp
NC/ONC
1

Salam-Weinberg (sin2ew = 0.35)3

1.l3

0.955

Vectorlike (sin 2e v

0.5)26

1. 30

1.30

Vectorlike (sin 2e v

0.75)26

1. 56

1.56

0.75

0.75

Electromagnetic current (EM) 2 5

For this comparison, the charged-current n-to-p ratio also enters.

Even if there is no isoscalar-isovector interference in the neutralcurrent interactions, we expect
R

vp

1.3 RV N

(5.8)

where the factor 1.3 arises from the fact that the n-to-p ratio for
the charged-current interaction is about 1.6. We have, quite generally,
R

> R

vp -

VN

(5.9)

Only in models where the neutral current is of the pure dd form does
R become equal to R N; this readily follows because in such models
vp
V
the neutral- and charged-current interactions both explore the same
quark distributions, viz. D(x) on proton targets, U(x) on neutron
targets.
I have discussed at some length how we might extract the coupling constants a, 8, y, and 0 given good data on the inclusive distributions. Unfortunately with the data currently available to us,
what we can do is rather limited. First of all, most of our
knowledge on the inclusive distributions comes from experiments in
which the target material is made up of complex nuclei, nearly
"isoscalar targets." This means that we have no information on
isovector-isoscalar interference from the inclusive reactions.
Second, we have no experimental information on the x distributions
because to determine q2 it is necessary to study the direction of
the hadronic showers, which has not been done by any of the experiments performed so far. In both the Gargamelle Collaboration

418

J. J. SAKURAI

experiment 1 ,39 and the two counter experiments at Fermilab

(Caltech 11 and HPWF 5 ) what has been measured is the hadronic energy
distribution or the V distribution. So knowing the incident neutrino spectrum, we have some information on the y distribution even
though the variable y is not directly determined for each event.
What is usually done is to assume scaling, spin 1/2 constituents,
and Hermiticity to write the cross sections as
(n/2G 2 m..E) do
N
dy

+ A (l-y)

{ for )!N
for vN

(S.lO)

and determine A+ and A_, taking into account the effect of the energy cut.

Now, if we had no antiquarks, the coefficients A and A+

would directly correspond to V-A and V+A, respectively. Because the

antiquark content is known to be nonvanishing, there are corrections to this naive valence quark approach, which, however, can be
estimated by studying the analogous coefficients for the chargedcurrent reactions.
(We, of course, assume that the charged-current
interaction is pure V-A.) In any case, knowing A and their charged
current analogs, it is possible to compute the left-hand sides of
(S.3a) and (S.3b) so that we can determine a 2 + S2 + y2 + 0 2 and

as + yo.

Using the experimental data reported at the Aachen Conference

(June 1976), I have deduced the following quantities 22 :
a 2 + S2 + y2 + 0 2 =

2(aS + yo)

ColS

0.16
1.18 0.19
1. 24 0.17

(Gargamelle 39 )
(HPWF 5 )
(Caltech 11)

(S.lla)

0.S7 0.18

(Gargamelle 39 )
(HPWF 5 )
(Caltechll)

(S.llb)

0.93 0.38
a 2 + S2 + y2 + 0' "{ 0.44 0.2S

In obtaining these numbers I assumed: (i) the strange quarkantiquark pairs (y' and 0' terms) contribute negligibly to the total
cross sections, an assumption justifiable to an accuracy of a few %
in most models, and (ii) the observed deviations from the standardone-to-three ratio in the charged-current reactions are due to the
presence of quark-antiquark pairs rather than new particle production with right-handed currents.
There are a few remarks to be made in connection with (S.lla)
and (S.llb). First, from (S.lla) we can conclude that we now know
the overall strength of the neutral-current interactions to an accuracy of about 20%. The three different groups with very different
tastes in physics, using very different experimental techniques and
neutrino beams (different in both the mean energy and the spectrum

NEUTRAL-CURRENT INTERACTIONS

419

shape) came up with the same number within errors.

opinion, is a remarkable achievement.

This, in my

Let us now look at (5.llb). This quantity, which measures the

amount of VA interference, is +1 for V-A, 0 for pure V/pure A, and
-1 for V+A. Experimentally, even though the three groups do not
completely agree on the central values, it appears that this quantity isnonvanishing and positive. So we are led to conclude that
both V and A coexist in the hadronic part of the weak neutral currents. Models with pure V (or pure A) are ruled out to the extent
that the quantity a8+Yo does not seem to vanish.
Given (5.lla) and (5.llb) we can determine the parameter(s) of
any of the models listed in Table 4.1. For example, if we specialize
to the one-parameter Salam-Weinberg model,3 we see
",2
u.

Q2
I-'

y2

2(a8 + yo)

~2 = .
1
u

2.

. 26 W
Sl.n

{ O. 5 } . .

0.9

Sl.n

= 0 30
26

W=

(5.l2a)
{

O. 35 }

0.21

(5.l2b)

We note that the two ways of determining 8W yields consistent results within errors. A similar exercise can be done with any of
your favorite models.
6.

NEUTRINO-INDUCED EXCLUSIVE HADRONIC REACTIONS

We have seen in the previous section that the inclusive distributions do not completely determine the coupling constants because
of the VA and isoscalar-isovector ambiguities. To settle this
question let us now turn to exclusive reactions.
As in beta decay physics, low energy nuclear transitions between states of definite isospin and spin-parity are sensitive to
specific pieces of the neutral currents. A classical example along
this line is 40

(6.1)
which may be studied using reactor antineutrinos with 7Li* identified by its subsequent y transition. This is a typical Gamow-Teller
transition which isolates the isovector axial-vector part (8 term)
of the neutral current interactions. Another reaction of interest
is

(6.2)
which is the weak analog of the photodisintegration of the deuteron.

420

J. J. SAKURAI

At reactor energies the vector matrix element is just the overlap

integral between the initial and final wave functions, which vanishes by orthogonality. The reaction is allowed with isovector axia13
vector (8 term again) because the deuteron ( Sl' I = 0) can disintegrate into a 1 s0 , I = 1 state of the n-p system. A preliminary
feasibility study by Reines and co-workers q1 shows that 82 , which
should be 1 in the Salam-Weinberg model. is given by

82

= -

(6.3)

0.9 2.2

It is expected, however, that eventually the errors will shrink to

0.2.
Coherent elastic scattering of complex nuclei

(6.4)

v+A+v+A

first considered by Freedman,q2,q3 is another exclusive process of

considerable importance. This scattering reaction received much attention because of its possible relevance to supernova explosion.
On isosca1ar targets it is sensitive to only the isosca1ar vector
part of the current (y term). Unfortunately astrophysical calculations on this subject appear to be rather uncertain at this moment,
and the experts in this field do not seem to agree on whether supernova explosion is possible, for example, with y ~ -0.25, a value
predicted by Salam-Weinberg model with the currently accepted
Weinberg angle. qq
I now turn to diffractive production of vector and axial-vector
mesons. We first recall that in photoproduction and also in low q2
e1ectroproduction, the reactions
y(rea1 or virtual) + N + (po, w, ) + N
show up conspicuously.

(6.5)

Furthermore, these vector meson states,

having the same quantum numbers as the photon (JPC = 1 -- ), are produced with characteristic features we expect from "diffraction"-energy-independent cross sections, diffractive slopes typical of
elastic scattering, sharp coherent peaks when nuclear targets are
used, etc. Likewise we expect that in neutrino reactions
V

+ N

+ V

~r

,f,

w,~,

A1 0 )

+ N

(6.6)

the meason states with the same quantum numbers as the current are
copiously produced via diffractive mechanism:
po for isovector vector
A1 for isovector axial-vector

(a term)
(3 term)

421

NEUTRAL-CURRENT INTERACTIONS

w and

(y and y' terms)

for isoscalar vector

There have been many calculations on these diffractive processes

based on vector (and axial-vector) meson dominance. 45 Such estimates are somewhat model dependent and may not be too reliable in
quantitative details. Fortunately, for the purpose of estimating
the coupling constants a and S, it is sufficient to compare the
neutral-current diffractive processes with the changed-current diffractive processes as follows:
do
dq2
do
dq2

0
(vN -+ vp N)

"2 a

(VN -+ )J - P+ N)

(6.7a)

diffractive

do
(VN -+ VAl oN)
dq2
do (VN -+ - +
)J Al N)
dq2

1. S2
2

(6.7b)

diffractive

In this way much of the model uncertainties cancels out. To resolve

the ambiguities in inclusive data, only crude knowled~e of diffractive reactions is needed, e.g., whether lal > lsi or lsi> lal;
whether a 2+S 2 is a sizable fraction of, or much smaller than,
a 2+S 2+y2+o2. It is also interesting to note that the elusive coupling constant y', which measures the strength of the vector part
of as ( like), can be estimated by comparing the -to-w ratio in
the neutral-current reactions to the corresponding ratio in the
electroproduction reactions:
do
dq2
do
dq2

do

(VN -+ VN)

(eN -+ eN)

~ _d~q_2___________

y2

(vN -+ vwN)
diffractive

do (eN -+ ewN)
dq2

(6.8)
diffractive

at the same value of q2.

Elastic scattering
v+p-+v+p

(6.9)

is, without doubt, one of the most important reactions in highenergy physics. It is, in a certain sense, as fundamental as neutron beta decay. At q2 = 0 we have a clean prediction 46

422

J. J. SAKURAI

dO'

(Vp + vp)

(a+3y)

(q2=0) = dq2
ReI
- ~d~O'~---------dq2

(vn + ~-p)

(6.10)
Possibly the only uncertainty in the q2 = 0 prediction lies in the
isoscalar axial-vector renormalization factor gls), which is the
analog of the gA/gV ratio in neutron beta decay.

According to the

way I normalized the coupling constants, it assumes the value 1 in

the naive SU(6) limit; however, the SU(6) prediction for gA/gV in
neutron decay is wrong by a factor of 1.25/(5/3), so a reasonable
attitude to take may be to assume the same correction factor to obtain
gls) = 1.25/(5/3) = 0.75 ,

(6.11)

a result derived by Adler and collaborators~7 using a more involved

argument. ~8
At q2 = 0 this elastic reaction is unfortunately unobservable;
recall

q2
where T

(6.12)

2m T
p P

is the kinetic energy of the recoil proton, so both parti-

cles in the final state are invisible as is the incoming particle!

There are two groups which have reported experimental studies of
this elastic reaction, a Harvard-Pennsylvania-Wisconsin Collaboration~9 and a Columbia-Illinois-Rockefeller Collaboration,50 both
working at Brookhaven where the mean neutrino energy is about 1 GeV.
Both groups study elastic events with q2 > 0.3 (GeV)2;

={

0.17 0.05
0.23 0.09

HPW
CIR

(6.13)

Unfortunately it is not straightforward to compare (6.13) with the

q2 = 0 prediction (6.10). As we go away from q2 = 0, there are uncertainties with the axial-vector form factors, and furthermore,
some form factors which are dormant at q2 = 0 could contribute. In
(el)
(el)
any case, the correction factor needed to get R 2 0 from R 2 0 3
q =
q >
depends on the relative ratios of a, S, y, and 0, the very quantities we are trying to determine!

In typical models R(el)

q2=0 at

NEUTRAL-CURRENT INTERACTIONS

423

10- 37

Data: HPW
2_{1.32GeV 2 --MA - 0.71 GeV 2 IIn-- fL- P

200

10- 38

100

~
.......

........

(!)

........

50
N

Q.)

(!)

C"
"'C

.......
b

"'C

>

0
en

-Lfi
.......

10- 39

10
5
lip ---liP

10-40~

__~____~____~__~__~~1
0.2

0.4

0.6

0.8

Fig. 6.1. vp elastic scattering

1.0

cQ.):

424

J. J. SAKURAI

R~~;~.3

Brookhaven energies is larger than

by about a factor of 1.5

to 2. Ideally it would be nice if we could extrapolate dO/dq2 observed in the q2 > 0.3 GeV 2 range to q2 = 0 in a relatively model
independent manner. But this appears to be difficult, as you can
see from some typical model calculations shown in Fig. 6.1, taken
from a recent paper of Albright et al. 51 Notice, in particular, how
sensitive the predicted curves are to the mass of the axial-vector
form factor assumed to be of the dipole form. The situation is presumably more favorable with accurate data down to q2 ~ 0.2 GeV 2
The HPW Collaboration 52 has also studied vp scattering.
nearly the same kinematical conditions they r~port
o(vp -+ \ip)

o(vp -+ vp)

Iq2>0.3

0.4 0.2

Under

(6.14)

This provides further evidence against models based on pure V or

pure A.
and N*
l / 2 's. There are
many papers on this topic, using static model calculations with O(q)
corrections, taking into account nuclear charge-exchange effects,
etc. Here, however, I wish to concentrate on results that follow
from simple isospin considerations. There are clean predictions we
can make in models with pure isovector--or almost pure isovector
as in SW with sin 2ew ~ 0.35--models 53 :
I now turn to single pion

R(~+'O)

= o(vp-+v~+)
20(vn -+

production--~

+ o(vn-+v~ )

(6.l5a)

- +

]J ~ )

o(vp-+vN *+) + o(vn-+vN *0 )

20(vn -+ ]J - N*+ )

1
4

a 20V + S2 0A + aSo I

v + A +

01
(6.l5b)

where 0V' 0A' and 01 are the vector, the axial-vector, and the
interference contributions to the cross section for the charged-

and N*
l / 2 stand,
respectively, for any I = 3/2 and I = 1/2 TIN (or any B = 1, S = 0)
system, not necessarily nucleon isobars of definite spin-parity. To
make use of these equations we must know the relative size of 0v and
0A; it turns out that most model calculations indicate 0v ~ 0A
current reaction appearing in the denominator, and

for almost any nucleon isobar, and this can simply be understood by
applying Bloom-Gilman duality54 to the neutrino and antineutrino
cross sections at resonance energies. The relations (6.l5a) and

NEUTRAL-CURRENT INTERACTIONS

425

(6.l5b) mean that when the same charge states are selected, there
is a factor of ~ 4 decrease in the neutral-to-charged current ratio
as we go from the

* dominant region.
dominant region to the Ni/2

In

reality the situation is not so simple because of nonresonant backgrounds, but we can conclude that: (i) a possible I = 1/2 contamination in the ~ region should be less serious in the neutral-current
than in the charged-current data, and (ii) a possible distortion in
the shape of the ~ peak is also expected to present a less serious
problem in the neutral-current data. All these simple conclusions
based just on isospin invariance are borne out by much more elaborate
calculations performed by Adler and collaborators 55 and by Paschos. 56
Even though data available so far on the TIN mass distribution
in the ~ region are still inconclusive as to the presence or absence
of ~, there is one piece of information relevant to the isospin
properties of the neutral currents. Recently the Gargamelle Collaboration 57 has examined the pion charge ratio in

(v) + N + (v) + TI'o + N'

(6.16)

If the current is pure isoscalar and if the target is also isoscalar,

we must have
a(TI+) : a(TI 0 )

1 : 1 : 1

(6.17)

which is clear because there is no preferred direction in isospin

space in the initial state. Work of the Gargamelle Collaboration
shows that the TIo-to-TI- ratio is more like 1.4 0.4 for the neutrino reaction and 2.1 0.4 for the antineutrino reaction. These
numbers should be compared to the isoscalar-current prediction 0.9,
a value expected when we take into account the fact that freon is
not quite pure isoscalar. Provided the experimental group knows how
to estimate the relative efficiencies of TIo and TI-, we must conclude
that pure isoscalar models are ruled out, i.e., a and/or 8 must be
nonvanishing.
Instead of single pion production we may also examine pion
semi-inclusive reactions
(6.18)
Any difference between TI+ and TI- in this set of reactions is a
measure of isoscalar-isovector interference when isoscalar targets
are used. However, to extract the relevant coupling constants we
must rely on the details of parton fragmentation models 58 which are
only now being critically tested in electroproduction.

J. J. SAKURAI

426

7.

NEUTRAL-CURRENT PHENOMENA WITHOUT NEUTRINOS

The total neutral currents to which the Z boson may be coupled

are made up of three pieces, (vv) , (It), and (qq), where t stands
for e or~. The interaction of (vv) with (It) gives rise to

(v)e scattering discussed in Section 3 while the neutrino-induced

hadronic reactions treated in Sections 5 and 6 take place via
(vv) (qq). As examples of (t) interacting with itself, I may mention weak interaction effects in Bhabha scattering
e+ +e- -+e+ +e -

(7.1)

and muon pair production in electron-positron collisions

-+

+ + ~-

(7.2)

There are also phenomena due to (It)(qq): parity violation effects

in inelastic muon-nucleon scattering, electron-positron annihilations
into hadrons, and atomic radiative transitions.
I would like to emphasize that it is of vital importance to
study at least one neutral-current reaction not involving neutrinos.
We recall that

V~ e

scattering and

(V)

induced inclusive hadronic

reactions are both characterized by cross sections of order G2 s/n.

This means that the strength of (vv)(t) is similar to that of
(vv)(qq) so that we have rough "universality." However, we can still
conceive of an unconventional model in which the coefficients in
front of t and qq are both enhanced by a factor r compared to normal models with universality while the coefficient in front of VV is
down by 1/2. Such a model, of course, would give large neutralcurrent effects in processes not involving neutrinos. 5'
Among the various neutrinoless neutral-current reactions mentioned earlier, I would like to concentrate on just two topics:
electron-positron annihilation into muon pairs and parity violation
in atomic radiative transitions. As you'll see, both topics are of
high current interest.
We begin with muon pair production (7.2). First, before starting to perform detailed computations, let us look at the order of
magnitude of weak neutral-current effects. At low (e.g., SPEAR or
DORIS) energies muon production is dominated by QED one-photon
exchange. Its amplitude (the Dirac spinors omitted) goes like e 2 /s
where l/s comes from the photon propagator in the s channel. The
process can also proceed via the neutral-current interactions where
the current is again in the s channel. In any model with rough
universality, the weak amplitude must be of order G [or Gm z2 /(mz2 -s)

if the center-of-mass energy IS is not so small compared to the boson

NEUTRAL-CURRENT INTERACTIONS
mass].
as

427

So the ratio of the weak to electromagnetic amplitude goes

G/(e 2 /s)

10-4 s

(7.3)

where s is in GeV 2 If the weak interactions continue to be pointlike (i.e., of the Fermi-type) at very high energies, the weak
amplitude is going to dominate the electromagnetic amplitude at
center-of-mass energies of several hundred GeV. Actually most
models which attempt to unify the electromagnetic and weak interactions promise us a spectacular resonance peak due to the Z boson
before reaching such an energy region.
Even at IS as low as tens of GeV, the weak amplitude could be
several per cent of the electromagnetic amplitude, and weak interaction effects may well be detectable via weak-electromagnetic interference. This is of immediate interest because there are now
colliding beam machines--PETRA and PEP--under construction precisely
in this energy range.
To be quantitative let us start with the most general V, A
couples compatible with ~e universality 6o:

+ 2hyA( eY Ae + ~YA~)(eYAYSe + ~YAYS~)

+ hAA(eYAYSe + ~YAYS~)(eYAYSe + ~YAYS~)]

(7.4)

In writing (7.4) I have assumed that the center-of-mass energy IS

is much lower than the Z boson mass; otherwise just multiply the
whole expression by mz 2 /(m z 2 + s - imzr z ). Even though models with
a single Z boson predict

(7.S)
from a general phenomenological point of view the three constants
hAA' and ~A must be treated as independent parameters. It is

~V'

therefore desirable to plan experiments that measure the three terms

separately.
In 1961 Cabibbo and Gatto,61 in one of the earliest papers discussing the raison-d'etre of e+e- colliding beam machines, pointed
out that if there are weak neutral currents, weak-electromagnetic
interference may be studied in electron-positron annihilation into
muons. In particular they proposed to look for neutral-current

J. J. SAKURAI

428

effects in the following three places:

(i)
the magnitude and s dependence of the cross section,
(ii)
forward-backward asymmetry,
(iii)
the longitudinal muon polarization.
Just to appreciate the time scale involved in weak interaction
physics, I may mention that the first positive results along this
line will be obtained probably in the early 1980s when PETRA and
PEP come into operation--twenty years after the Cabibbo-Gatto paper!
According to QED, the cross section for (7.2) to order a 2 is
given by (m 2 s)
11
0QED

=j

na 2 /s

(7.6)

With neutral currents added, the muon pair cross section is predicted to deviate from the QED prediction (7.6) both in the magnitude and in the s dependence. For s not too large the deviation is
sensitive only to the ~v term 61

~O/OQED

(7.7)

(G/1:2 na)hvVs

Let us now turn to the angular distribution for which the QED prediction is
(7.8)
forward-backward symmetric in the one-photon exchange approximation.
With an axial-vector piece in the 1eptonic current (hAA term),
forward-backward asymmetry is possible via weak-electromagnetic interference. The relevant formula iS 61
0(8) - 0(n-8)
A(8) - 0(8) + 0(n-8)

1:2
(G/ 2 na)hAAs

2 cos8

(7.9)

1 + cos 2 8

Finally, there is a genuinely parity-violating effect, muon polarization, due to the ~A term. The formula for the longitudinal
polarization with unpo1arized initial beam is 61
P(l1+)

-P(l1-)

(G/1:2 na)hVVs [1 +

2 cos8
1 + cos 2 8

(7.10)

Notice that the three different experiments are sensitive to the

three different terms of the Lagrangian (7.4).
It is also worth
keeping in mind that the formulas (7.7), (7.9), and (7.10) are valid
only if the weak amplitude is still small compared to the electromagnetic amplitude; in particular, they are not valid if there is a
Z boson with mass close to IS.

Cb

Cb

Cb

I:::

I:::

--- --

Cb

--

Cb

II

I PETRA
PEP

-.5

2000

4000

Fig. 7.1. Forward-backward asymmetry in e+ + e-

oW

+.5

S-W model (sin 2 8w= 0.35)

v+ +

v-.

8000
GeV 2

-0

..,.

C/l

o2

l>

:xl

2
-I

2
-I

:xl
:xl

(")

l>

:xl

C
-I

J. J. SAKURAI

430

To obtain numerical results expected in experiments of this

type, let us now specialize to the Salam-Weinberg model. 3 The coupling constants in this model are given by 62
hvv

~A

(1 - 4 sin 2e)2

(1 - 4 sin 2e)

(7.11)

In Fig. 7.1 I show forward-backward asymmetry expected for sin2ew

0.35, taken from a recent review talk by Gourdin. 63 At low
values of s the asymmetry is linear in s in agreement with (7.9) but
there is a spectacular deviation from the linear extrapolation as s
approaches the squared boson mass, at which the asymmetry changes
sign. At a typical PETRA or PEP energy of s = 900 GeV 2 (15 GeV beam
energy) the asymmetry A(e) evaluated at e = 0 is predicted to be
as large as 8%, which should be comfortably measurable. 64
In models in which the charged-leptonic part of the neutral
currents is purely or predominantly vector, a measurable angular
asymmetry is not expected. However, in such models the deviation
from the QED prediction given by (7.7) may be sizable. The coefficient (G/12 na) is 3.5 x 10-4/GeV 2, so in any model with hVV ~ 1/4,
the muon pair cross section may deviate from the QED prediction by
as much as 8% at s = 900 GeV 2 . So regardless of whether nature
chooses hAA to be sizable (as in the Salam-Weinberg model) or ~ to
be sizable, something interesting is bound to happen at PETRA or PEP.
We now turn to the subject of parity violation in atomic
physics. Consider the axial part of the electronic current interacting with the vector part of the hadronic current. The interaction is clearly parity-violating, and, moreover, its effect in complex atoms is coherent in the sense that the contributions from the
various nucleons in the nucleus are additive as the charges add up.
In principle, we can also consider a parity-violating effect due to
the vector part of the electronic current interacting with the
axial-vector part of the hadronic current but this latter effect is
not so easy to detect because the spin orientations inside the
nucleus don't add up to give rise to a huge coherence effect.
Let us look more closely at the V
A
interaction. For the
had lept
hadronic side we must consider the time component of the vector current because for nonrelativistic nucleons YA is large only for the
time component. The leptonic side is then given by eY4Y5e = e t y 5 e,
-+

-+

-+

which reduces in the plane-wave representation to -oo(P.+Pf)/2m .

1

NEUTRAL-CURRENT INTERACTIONS

431

The resulting interaction can be represented by an effective parity

violation potential between the electron and the nucleus (treated
to be pointlike) as follows 65
H
p.v.

(G/4i:2m) QW ~.{;,8(3)(~)}
e

(7.12)

with QW given by

QW

-(ae/av)[a(Z-N) + 3y(Z+N)]

(7.13)

In (7.13) Z and N are the number of protons and the number of neutrons, respectively, and a and a are the coefficients in front of
e
V
ieYAYSe and ivyA(l+yS)V that enter in the leptonic neutral current
to which the Z boson is coupled. Notice, how the contributions from
the individual nucleons add up coherently for the isoscalar piece.
Just as an example, I may mention that a /a is -1 in all
e v
SU(2) (9U(l) models where the 3rd components of left-handed weak
isospin have opposite signs between v and e, and that in the
Salam-Weinberg model with sin 2 Sw ~ 0.3S, QW is -160 for bismuth
atoms.
Because of the potential (7.12) a given atomic state ia> is no
longer an exact parity eigenstate but contains an opposite parity
admixture with amplitude going like <biH
ia>/(Eb-E ) where ib> is
p.v.
a
a nearby state of opposite parity. As a result, a radiative transition between a pair of atomic levels acquires a parity-violating
component; a dominantly Ml transition, for instance, contains a very
small amount of El as well. This means that photons emitted by
excited atomic states are, in general, circularly polarized.
There are now two 3xperimental groups--Fortson and collaborators 66 at Seattle (the University of Washington) and Sandars and
collaborators 67 at Oxford--who are carrying out experiments of sufficient sensitivity to detect parity violation in atoms. In both
experiments 209Bi (Z = 83) atoms are used. A linearly polarized
laser beam whose wavelength corresponds to the level difference in
question--between 4S3/2 (the ground state) and 2D3/2 (the first excited state) in the Seattle experiment 66 and between 2D3/2 and 2DS/2
(the second excited state) in the Oxford experiment 67 --is allowed to
pass through an oven containing bismuth vapor. The basic principle
of the experiments is to detect parity violation by observing an
optical rotation of the linear polarization expected when right- and
left-circularly polarized beams have different indices of refraction.
Theoretical calculations which, incidentally, require good knowledge
of the atomic wave functions, show that with a realistic experimental arrangement, the expected rotation is in the neighborhood of

432

J. J. SAKU RAI

10- 7

10- 6 radians with QW given by the Salam-Weinberg model. 68

At the time of this summer school, despite some earlier rumors,
there is no firm number that can be announced publicly by either
group.
It is worth remarking that these atomic physics experiments can
be very powerful in discriminating various theoretical models. It
is clear from (7.13) that by varying the neutron-to-proton ratio,
one can, in principle, measure different linear combinations of the
isovector and isoscalar coupling constants. Furthermore, models
that give very similar predictions on neutrino-induced hadronic
reactions can be shown to give very different values of Q in both
sign and magnitude. 69
W
8.

CONCLUSION

Three years have passed since the Gargamelle Collaboration announced the great discovery of neutral currents. Having spent tens
of millions of Swiss francs (or U.S. dollars) on both sides of the
Atlantic, what have we actually learned since then? First, the
existence of neutral currents was confirmed by several independent
experiments. It has been demonstrated to everybody's satisfaction
that neutral currents indeed exist and are here to stay. The currents are no longer alternating.
Second, some extreme models have been ruled out; the casualty
list goes as follows:
(i)
models with pure Sand/or P hadronic densities,
(ii)
models with pure V or pure A hadronic currents,
(iii) models with pure isoscalar (I = 0) hadronic currents,
(iv)
models with very different strengths between (vv)(qq)
and (vv)(ee),
(v)
models that violate scaling violently.
Considering, however, that the history of weak-interaction physics
is marred by wrong experiments and bad theoretical models inspired
by the wrong experiments, we should not regard this casualty list
as absolute final.

l'

In terms of the coupling constants a, 8,

and 0 defined by
(4.1) we know that the overall strength a 2 + 8 + y2 + 0 2 is equal
to 1.2 to an accuracy of about 20% and that the quantity a8 + yo
is positive and nonvanishing, indicating that the hadronic current
is somewhere between pure V-A and pure V/pure A. In addition,
from single pion reactions we must have

and/or

8 I- 0

(8.1)

Assiduous model builders can still construct an infinite variety of models subject to the constraints above!

NEUTRAL-CURRENT INTERACTIONS

433

What further experiments are needed to completely determine the

properties of the neutral-current interactions?
As far as the inclusive hadronic reactions are concerned,
clearly the next major progress will be made when we obtain accurate
data on proton targets, not just on isosca1ar targets. Furthermore,
it is desirable to measure the neutron-to-proton-target ratio as a
function of x because isosca1ar-isovector interference, if present,
is predicted to show up more conspicuously for x > 0.5; the effect
'V

tends to get diluted when we integrate over x. In new generation

bubble chambers with track-sensitive targets, it is not beyond the
realm of feasibility to study the x distributions of the neutra1current reactions separately for protons and neutrons.
I have already emphasized in Section 6 the importance of studying diffractive vector and axial-vector production in resolving the
two "ambiguities." A careful study of <P and w diffractive production
will determine the elusive constant y' and test the details of the
GIM mechanism or U spin invariance. As for elastic vp scattering,
it is desirable to design experiments to study do/dq2 down to q2
values as low as possible.
Turning now to neutrino-electron scattering, we would like to
see how the three ellipses for V e, V e, and V e with negligible
)l)l
e
error bands intersect in a gV-gA plane. Very accurate data for Vee
scattering at reactor energies are needed to see whether the vector
dominant or axial-vector dominant solution is the right one.
The prospects for detecting weak-electromagnetic interference
in electron-positron collisions at PETRA and PEP energies are extremely promising. We look forward to hearing about the first experimental report on weak neutral-current effects in e1ectronpositron annihilations into muon pairs at the 1981 Erice School.
Of more immediate interest is the status of parity violation
experiments with atoms. Even the sign and crude order of magnitude
of QW are of enormous value. Measurements of QW with different atoms
should be very powerful in eliminating various theoretical models
(the experimentalists' delight!).
Some of us theorists are often discouraged by the slow rate of
progress made in neutral-current physics--s10w, when compared to the
fantastic progress made in the field of new-particle spectroscopy
since that fateful Sunday, November 11, 1974. But we have to be
patient. It took 25 years to demonstrate V, A structure in beta
decays where the number of invisible particles is only one. You may
speculate how long it will take to determine completely the structure
of the 1eptonic and hadronic parts of weak neutral currents by studying reactions with two invisible particles.

434

J. J. SAKURAI

ACKNOWLEDGMENT
Much of the lecture material was prepared whilp 1 was a scientific associate at CERN holding a John Simon Guggel
. 'u Memorial
Foundation Fellowship. I wish to take this opportunity to acknowledge the hospitality of the Theory Division at CERN.
APPENDIX
In this appendix I present a simple derivation of the master
equations (5.3a) - (5.3d) within the framework of the valence quark
model.
In the quark language the charged-current reactions are visualized to be due to the "fundamental processes"
\I

+ d + 11

\I

+
+ u + 11 + d

+ u

(AI)

To the extent that the antiquark content within the nucleon can be
ignored, the \I induced inclusive charged-current reaction on protons
explores D(x), the momentum distribution of the d type quark within
the proton. Likewise the V induced reaction on protons explores
U(x), the distribution function for the u type quark within the proton. Now, because of the charge symmetry of the strong interactions,
D(x) is also the momentum distribution of the u type quark within
the neutron; so the V induced reaction on neutrons measures D(x).
Similarly the \I induced reaction on neutrons is sensitive to U(x).
Recalling now the helicity rule for V-A--a flat y distribution for
fermion-fermion scattering, (1_y)2 for antifermion-fermion scattering--discussed in Section 3, we obtain
\lp
dCJ CC

--=

dxdy

(A2)

and similar expressions for the neutron target with U and D interchanged.
If we now go to the neutral-current case, we must keep in mind
that both V-A and V+A are in general present. Now the coupling conA, and y a and
stant combinations a 8 and y 8 go with V

8 8 go with the {~} type quark [see (4.1)].

Combining all this

NEUTRAL-CURRENT INTERACTIONS

435

information with the helicity rule, we obtain

(A3)

where a L Rand b L R are given by

,
,

i (a+S+y+o)
=i (-a-S+y+o)

aR

aL bL

-1 (a-S+y-o) ,
4

(A4)

bR

= !4

(-a+S+y-o)

The expressions for the neutron target reactions can again be obtained by interchanging U and D.
Notice that in this naive valence quark approximation, all the
differential distributions are completely determined once U(x) and
D(x) are given together with the coupling constants. Conversely, by
taking the ratios of suitable linear combinations of (A2) and (A3)
and their neutron analogs, it is possible to eliminate U(x) and D(x).
In this manner we readily derive the master equations (5.3a) - (5.3d)
which do not involve U(x) and D(x).
REFERENCES AND FOOTNOTES
1)
2)
3)
4)
5)
6)
7)
8)
9)

F. J. Hasert et al., Phys. Lett. 46B, 138 (1973).

I am following Professor H. Faissner who compared the Neutrino
Conference (Aachen, June 1976) he organized to a well-structured
Viennese symphony.
A. Salam, Proceedings 8th Nobel Symposium, ed. N. Svartholm
(Almquist and Wiksells, Stockholm, 1968), p. 367; S. Weinberg,
Phys. Rev. Lett. 19, 1264 (1967).
J. J. Sakurai, CERN TH-2099 (to be published in Proceedings
International Summer Institute of Theoretical Physics, DESY,
Hamburg, September 1975).
T. Y. Ling, Proceedings Neutrino Conference (Aachen, June 1976).
L. M. Sehgal, Phys. Lett. 55B, 205 (1975).
L. Wolfenstein, Nucl. Phys. B9l, 95 (1975).
J. E. Kim, V. S. Mathur, and S. Okubo, Phys. Rev. D9, 3050
(1974) .
J. Bernstein and T. D. Lee, Phys. Rev. Lett. 11, 512 (1963);
C. Bouchiat, J. Iliopoulos, and Ph. Meyer, Phys. Lett. 42B, 91
(1972).

436

10)
11)

12)
13)

14)
15)
16)

17)
18)
19)
20)

21)

J. J. SAKURAI

B. Kayser et a1., Phys. Lett. 52B, 385 (1974); R. L. Kingsley,

F. Wilczek, and A. Zee, Phys. Rev. D10, 2216 (1974).
D. Buchholz, Proceedings Neutrino Conference (Aachen, June 1976).
P. Sutherland et a1., Phys. Rev. D13, 2700 (1976).
C. Bouchiat, Phys. Lett. 57B, 284-zl975); E. A. Hinds et a1.,
Phys. Lett. 62B, 97 (1976r:G. t'Hooft, Phys. Lett. 37B, 195 (1971).
J. D. Bjorken, Phys. Rev:-i79, 1547 (1969).
S. Pakvasa and G. Rajasekaran, Phys. Rev. D12, 113 (1975);
M. Gronau, Nuovo Cimento Lett. 14, 204 (1975).
F. J. Hasert et a1., Phys. Rev. Lett. 46B, 121 (1973);
J. B1ietschau et a1., CERN/EP/PHYS 76-~(submitted to Nuc1.
Phys. B).
F. Bobisut, Proceedings Neutrino Conference (Aachen, June 1976).
A plot of this kind was first proposed in H. H. Chen and G. W.
Lee, Phys. Rev. D5, 1874 (1974).
F. Reines, H. S. Gurr, and H. W. Sobel, Phys. Rev. Lett. 37,
315 (1976).
When the electron mass is not ignored. the variable y in the
formula for do/dy should be understood as E /E where E is the
e

22)
23)

24.
25)
26)

27)
28)
29)
30)

31)
32)

33)

final electron energy in the laboratory system.

Sections 4 - 6 are largely based on an invited talk I gave at
the Aachen Conference; J. J. Sakurai, Proceedings Neutrino
Conference (Aachen, June 1976).
For example, the constant y in this paper is related to the constant Vs of Reference 4) by Y = 3 vs
S. L. G1ashow, J. I1iopou1os, and L. Maiani, Phys. Rev. D2,
1285 (1970).
M. A. Beg and A. Zee, Phys. Rev. Lett. 30, 675 (1973); v. S.
Mathur, S. Okubo, and J. E. Kim, Phys. Rev. D11, 1059 (1975).
A. De RUju1a, H. Georgi, and S. L. G1ashow, Phys. Rev. Lett. 35,
65 (1975); H. Fritzsch, M. Ge11-Mann, and P. Minkowski, Phys.
Lett. 59B, 256 (1975).
Y. Achiman, K. Koller, and T. F. Walsh, Phys. Lett. 59B, 261
(1975).
F. Gursey and P. Sikivie, Phys. Rev. Lett. 36, 775 (1976).
S. L. Adler and S. F. Tuan, Phys. Rev. D11,-r29 (1975).
The importance of U spin invariance in understanding the absence
of the strangeness-changing neutral currents was emphasized in:
J. J. Sakurai, Phys. Rev. D9, 250 (1974); V. S. Mathur, S. Okubo,
and J. E. Kim, Phys. Rev. D10, 3648 (1974).
Pham Quang Hung and J. J. Sakurai, Phys. Lett. 63B, 295 (1976).
Special cases of these equations have been derived by a number
of authors: E. A. Paschos and L. Wo1fenstein, Phys. Rev. D7,
91 (1973); G. Rajasekaran and K. V. L. Sarma, Pramana 1, 62
(1974); A. De Ruju1a et a1., Rev. Mod. Phys. 46, 391 (1974).
R. P. Feynman, "Photon Hadron Interactions" (W"-:- A. Benjamin,
New York, 1972), p. 132; J. D. Bjorken and E. A. Paschos, Phys.
Rev. D1, 3151 (1970).

NEUTRAL-CURRENT INTERACTIONS
34)
35)
36)
37)
38)

39)
40)
41)
42)
43)
44)
45)
46)
47)
48)
49)
50)
51)
52)
53)
54)
55)
56)
57)
58)
59)
60)

437

C. Chang et a1., Phys. Rev. Lett. ~, 901 (1975);

H. L. Anderson et a1., Phys. Rev. Lett. 37,4 (1976).
A. Benvenuti et a1., Phys. Rev. Lett. 36~1478 (1976);
A. Benvenuti et a1., Phys. Rev. Lett. 12, 189 (1976).
H. J. Lipkin, Phys. Reports~, 175 (1973).
W. B. Atwood, SLAC Report 185 (1975).
After Fig. 5.2 [taken from Reference 31)] was prepared, we have
learned that the EM and vector1ike models considered in the
figure are ruled out by recent data. However, variants of the
models where we add axial-vector currents with the same
isovector-isosca1ar ratios (S/8 = a/y) give the same predictions
for neutron-to-proton ratios, and they are not yet ruled out by
the data.
W. Van Doninck, Proceedings Neutrino Conference (Aachen, June
1976).
T. W. Donnelly et a1., Phys. Lett. 49B, 8 (1974).
H. S. Gurr, F. Reines, and H. W. Sobel, Phys. Rev. Lett. ll,
179 (1974).
D. Z. Freedman, Phys. Rev. D9, 1389 (1974).
3
The constant a in Reference 42) is related to our y by a o = 2 y.
o

J. R. Wilson, Phys. Rev. Lett. ~, 849 (1974); D. N. Schramm

and W. D. Arnett, Phys. Rev. Lett. 34, 113 (1975).
M. K. Gaillard, S. Jackson, and D. Nanopou1os, Nuc1. Phys. B102,
326 (1976).
Equation (6.10) is a generalization of a relation first obtained
in: S. Weinberg, Phys. Rev. 2, 1412 (1972).
S. L. Adler et a1., Phys. Rev. D11, 3309 (1975);
S. L. Adler, Phys. Rev. D12, 26~(1975).
A somewhat smaller value of gls) is obtained if SU(3) is used
together with the D/F ratio inferred from hyperon decay.
D. Cline et a1., Phys. Rev. Lett. 37, 252 (1976).
W. Lee et a1., Phys. Rev. Lett. 37-,-186 (1976).
C. H. Albright et a1., FERMlLAB-PUB-76/45-TH.
D. Cline et a1., Phys. Rev. Lett. 37, 648 (1976).
J. J. Sakurai, Phys. Rev. Lett. 35~1037 (1975).
E. D. Bloom and F. J. Gilman, Phys. Rev. Lett. 12, 1140 (1970).
S. L. Adler et a1., Phys. Rev. D13, 1216 (1976).
E. A. Paschos, Proceedings 10th Rencontre de Moriond, ed.
Tran Thanh Van (CNRS, Paris), p. 341.
C. H. Bertrand Coremans et a1., Phys. Lett. 61B, 207 (1976).
C. H. Albright and J. C1eymans, Nuc1. Phys. B76, 48 (1974);
L. M. Sehgal, Nuc1. Phys. B90, 471 (1975).
C. H. Llewellyn Smith and n:-V. Nanopou1os, Nuc1. Phys. B78,
205 (1974).
The spirit of my approach to this subject closely parallels that
of L. Wo1fenstein [AlP Proceedings No. 23, Division of Particles
and Fields, Williamsburg, Sept. 1974 (ed. C. E. Carlson), p. 84]
who discussed the reaction (7.2) without reference to particular
models.

438

61)

62)
63)
64)
65)
66)
67)
68)

69)

J. J. SAKURAI

N. Cabibbo and R. Gatto, Phys. Rev. 124, 1577 (1961); A. Love,

Nuovo Cimento Lett. 5, 113 (1972); J:-Godine and A. Hankey,
Phys. Rev. D6, 3301 (1972); V. K. Cung, A. K. Mann, and E. A.
Paschos, Phys. Lett. 4lB, 355 (1972); R. Bundy, Phys. Lett.
45B, 340 (1973).
Equation (8.12) of my DESY paper [Reference 4)] is wrong.
M. Gourdin, Proceedings Neutrino Conference (Aachen, June 1976).
Note from (7.9) and (7.11) that at low values of s the asymmetry
prediction in the Salam-Weinberg model is independent of 8W.
M. A. Bouchiat and C. C. Bouchiat, Phys. Lett. 46B, III (1974).
P. C. Soreide et al., Phys. Rev. Lett. 36, 352 (1976).
P. G. H. Sandars, "Atomic Physics" 4, e~ G. Zu Putlitz, E. W.
Weber, and A. Winnacker (Plenum, Ne; York, 1975).
E. M. Henley, Proceedings 9th International Conference on the
Physics of Electronic and Atomic Collisions, Seattle, Washington
(1975); E. M. Henley and L. Wilets, Phys. Rev. A (to be published).
Some examples of this can be found in a recent paper by J.
Bernabeu and C. Jarlskog, CERN TH-2206.

NEUTRAL-CURRENT INTERACTIONS

439

DIS C U S S ION S
CHAIRMAN:

Prof. J.J. Sakurai

Scientific Secretaries:

H. Pham Quang and M.R. Pauli

DISCUSSION 1
MARCIANO:

In your local leptonic Lagrangian, you assumed ~e universality

for the neutral currents. Is there any experimental evidence that
this is indeed the case?
SAKURAI:

At present, there is no experimental evidence for ~e universality

in the neutral currents. Even in the charged current inclusive reactions, ~e universality is only crudely tested.
MARCIANO:

Will the neutrino beam experiments allow us to examine interesting questions about the nature of the neutrinos, i.e. vev~ mixing,
whether they are massless, and, if not, what is the decay rate for,
say, v~ + ve + y?
SAKURAI:

Starting with a beam of neutrinos, from a reactor for example,

we can examine the distance dependence of the cross-section and infer
various properties from the data. The neutrino from a reactor is
to start with, but if there is a finite mass difference between the
muon neutrino and the electron neutrino, the beam may acquire a v~
component. Thus, there might be a reduction in the cross-section
for inverse beta decay.

ve

440

J. J. SAKURAI
ZICHICHI:

In this context, I must tell you something which has already

been brought up on several occasions concerning the neutral currents
generated by the muon neutrino. One has an incoming muon neutrino
and no muon in the final state. The question is what would happen
if an electron were found in the final state? In this case, it would
not be the neutral currents, but it would be a violation of the leptonic selection rule. How can one exclude this? It would be very
poor to use the energy distribution of the calorimater.
SAKURAI:

Comparison of the visible energies in the final state shows

that in the neutral current case, there 1S a sizeable fraction of
the total energy missing.
BUCCELLA:

What is the evidence of the neutral leptonic currents?

SAKURAI:

The Gargamelle evidence is based on three events for v~e + v~e;

however, each of the events has some small (0.5%) probability for being
due to ve + n + e- + p where the proton is not seen. The Aachen-Padua
evidence, based on a much larger sample of candidates, is considered
to be a 2.5 standard deviation effect for both v~e and v~e. However,
since this is a spark chamber experiment, the background due to single
TI o production is serious.
KLEINERT:

Can the experts explain what made the elastic

sections drop by a factor of two?

ve and -Ve cross-

SAKURAI:

I believe you are referring to the data of the Aachen-Padua

Collaboration. As I understand the situation, the data itself has
not changed since the Aachen Conference, but if we use a subsample
of data with a large Ee to reduce the neutral pion background, the
cross-section becomes smaller, perhaps more in agreement with the
Gargamelle data.
JONES:

Could you explain the significance of the various ellipses on

the gV/gA allowed region plot?

NEUTRAL-CURRENT INTERACTIONS

441

SAKURAI:
(See lecture notes for the ellipse diagram.)
The outer ellipse denoted by v~e is the 90~ confidence level
upper limit for v~e + ~e; the inner ellipse is the 90% confidence
level lower limit. The ellipse denoted by v~e is the 90% confidence
level upper limit for v~e + v~e.

SOHNIUS:
How good is the evidence that no S or P contribution
In neutral currents?

lS

present

SAKURAI:
It is very good for the inclusive hadron reactions, provided
that T is not simultaneously present, but it is very poor for
neutrino-electron scattering.

SOHNIUS:
Are possible contributions included in your ellipse graph?

SAKURAI:
No.

SOHNIUS:
How would the ellipses change using the new data presented here
earlier by Baldo-Ceolin?

SAKURAI:
The ellipse of ~e would become larger. In terms of the Weinberg
angle, the Gargamelle ellipse corresponds to 0.1 < sin 2 8w < 0.4 -see the diagram in the lecture notes -- while the Aachen-Padua ellipse,
as presented at the time of the Aachen Conference, corresponds to
0.4 < sin 2 8w < 0.7. However, I understand from Baldo-Ceolin that
the Aachen ellipse becomes smaller when they select events with high
electron energies.

GOURDIN:
Let me make a general comment concerning the SPT situation. As
it is well known, both in the leptonic and hadronic cases, we have
the so-called confusion theorem which tells us that to a (V,A) solution
corresponds an (S,P,T) solution and an infinity of (S,P,T,V,A) solutions.

J. J. SAKURAI

442

Therefore, we must add some information to eliminate (S,P,T) components. In the leptonic case, we need to study the low-energy correction and to study the charged current-neutral current interference
in electron neutrino-electron scattering by comparing with muon
neutrino-electron scattering. In the hadronic case, we need to study
a very complicated transfer of polarization measurements at the
hadronic vertex, for instance, in neutrino-proton scattering.
SAKURAI:

That is quite correct.

DISCUSSION 2
GOURDIN:

You showed a graph for R(V) as a function of energy where

Gargamelle data and HPNF results are in agreement. Do you have an
analogous graph for R(V) , because if charged current data coming from
HPWF are right, we must have an increase of charged'current antineutrino cross-sections, and therefore a decrease of R(V) with energy?
SAKURAI:

The mean energy in the HPWF experiment is 40 GeV for the antineutrino. This is in the energy range where the famous "anomalies"
in the charged current antineutrino data start showing up. I believe
that the ratio a(v)/a(v) for the charged current reactions is about
0.45. You are, of course, quite right in saying that R(v)should decrease if the neutral current data does not exhibit anomalies.
GOURDIN:

as

Your plot in the

plane with a pure isovector neutral current
can be used in a more complex way, independently of any valence quark
model. This was dine in my 1975 Schladming lectures and it is possible to define, by positivity, a domain of allowed values for a and
S. The limit of that domain obviously corresponds to a pure isovector
neutral current.
SAKURAI:

Yes, other plots are possible to exhibit the constraints imposed

by the data on the various coupling constants. For pedagogical reasons,
I have shown the simplest plot.

NEUTRAL-CURRENT INTERACTIONS

443

MARCIANO:

Concerning the mass of the neutral ZO boson, we are all influenced

by the Salam-Weinberg model into thinking that the mass of the ZO must
be very large. However, merely from experiment, what bound can we put
on the mass?
SAKURAI:

One can probably infer from the linear rise of the cross-section
as a function of E -- assuming scaling -- that the mass of the ZO is
rather large. The data still has large errors but we can say that at
least the mass of the ZO is greater than 10 GeV. We can also rule out
a low mass ZO boson using the fact that the QED calculations for
e+ + e- + ~+ + ~- fit the data at SPEAR energies.
MARCIANO:

Assuming a coupling constant of the order of the electric charge,

what does the mass of the ZO seem to be?
SAKURAI:

From purely dimensional arguments, the assumption that the dimensionless coupling constant is of the order of e would infer a very
large mass for the Z0 Remember that Fermi' sG has dimensions of
M""~i the squared dimensionlesoS coupling constant is of the order of
10
-- where the reference mass is identified with the proton mass.
In any model where the dimensionless squared coupling constant is as
large as e 2 , the reference mass must be as large as the mass of a
heavy nucleus. All this is completely independent of whether the
SUeZ) x U(l) gauge model is right.
PAULI:

You had written down data for neutrino and antineutrino in inclusive reactions and obtained 90% V-A and 10% V+A. How does this agree
with the data f!om HPWF =a(vp + vp)/a(vp + vp) = 0.4 :!: O.Z= and the
data from CERN ~a(vp + vTIp)/a(vp + VTIp) = 0.49 O.lZ_for exclusive
reactions? Are the experiments that are being done to determine
pure V versus pure A versus V+A versus V-A consistent with each other?
SAKURAI:

The inclusive reactions and the exclusive reactions agree to

the extent that they both indicate the presence of sizable V,A.interference. However, to extract quantitatively the relative amounts of
V and A from the exclusive reactions, more work is needed. It is very
important to note, in this connection, that the ratio

444

J. J. SAKURAI

O(Vp + Vp)/O(Vp + vp) is dependent on the beam energy. On general

considerations, it is possible to show that this ratio must approach
unity for Ev rep, even if both V and A are present. A similar
remark applies to single pion production.

PUGLIERIN:
For (do/dy)
0.9 0.1 (1_y)2, the values of 0.9 and 0.1 are
derived from the R value in the HPWF experiment. For a direct
measurement, it is probably better to refer to the Cal tech data since
they have a narrow-band beam. Do you know if there are some new
data?

SAKURAI:
It is true that there is a considerable advantage of using a
narrow-band beam if we are to measure the y distribution. In fact,
for a strictly monoenergetic beam, the y distribution is essentially
the same as the Ehad distribution, where Ehad is the total hadronic
energy deposited in the calorimeter. There were some new Ehad plots
by the Caltech Group presented at the Aachen-Conference. The coupling
constants deduced implicitly depend on the Ehad distributions as well
as the neutral current/charged current ratios with certain energy
cuts.

WEAK CURRENTS AND NEW QUARKS

Michel GOURDIN
Laboratoire de Physique Theorique et Hautes Energies
Tour 16 - ler etage - 4, place Jussieu
75230 PARIS cedex 05
The aim of these lectures is to give a review of the situation concerning neutrino and antineutrino inelastic scattering
and the implication of recent high energy data concerning the
number and the properties of quarks.
The obvious reference for a comparison is the set of experiments performed at Gargamelle which together with SLAC electroproduction experiments can receive a simple interpretation in terms
of a quark parton model. We therefore first study low energy data
using the two component quark parton model and we show how well it
fits the data by making predictions which are ~n agreement with
experiments.
Anomalies with respect to the low energy description have
been claimed to appear at Fermilab energies but the experimental
situation is still unclear. Nevertheless, we discuss two possible
schemes for weak currents and we show how various data can be
explained. The first class of models use only left handed charged
current and it is the natural extension of the low energy situation
where the Cabibbo current describes nicely experimental data.
Variation be~ween low energy and high energy of the quark parton
model parameters are discussed. The second class of models involves
new types of currents with right handed helicity and consequently
new quarks with high masses. It follows that the quark parton model
formalism has to be modified in order to take into account threshold
effects for heavy quark production. We discuss this point and the
slow rescaling in the Bjorken variable x implies a continuous variation with energy of these new contributions.
Finally in part III, we look at the implications for neutral
currents of the previous considerations using the simplest
SU(2) 8 U(I) gauge group. The mixing angle
is then determined

eVol

445

446

M. GOURDIN

after comparison with low and high energy data. Again consistency
with experiment is found.
PART I
GARGAMELLE DATA
In this part we briefly review the main features of neutrino
and antineutrino inclusive reactions as observed in the Gargamelle
experiment made at CERN. The electromagnetic counterpart is the set
of inclusive electroproduction experiments performed at SLAC and
DESY. They will not be discussed here and we only assume they are
well known.
I - BJORKEN CANONICAL SCALING
1) High Energy Cross-Sections
High energy incident beams in the laboratory frame mean
E

In that case the fixed x, y double differential cross-sections

have a very simple form. As a consequence of spin one exchange in
the t channel they are described by three Lorentz invariant structure functions F? (x, q2) where 1\ = tl ,0 is the helicity of the
weak current. We simply obtain T

d'26:

Go\\ - ~
1. _21
- ~
}
=
_c (let to. (~'4 2) t- (i-~.) '"'+ (~,1'~) 1-~(1.-~) ro (<<I'~
d~dl.;i
,~\
_:.1

11

j~~:S

IllM'=

_q~o~= "'!--~

r-"
~
ZF
2
'")JA
11 ~3
t C3:,'P +(1.-~.) _ (~,=t) +~(.~-~) 0 (11:,,1('

'V(}

qOCdtj
Tt
l. "tAs usual G is the Fermi coupling constant as measured in ~
decay. Experimentally, effects due to an intermediate boson propagator have not been observed at present energies and we shall work
in the Fermi local approximation.
2) Bjorken Canonical Scaling
With this choice of structure functions, Bjorken scaling is a
very simple limiting statement at large q2

..

The structure functions usually used, F I , F 2 , F3 are related

to F,.' s by

447

WEAK CURRENTS AND NEW QUARKS

l~m

q=) 00

~ L~, 4 Z)

=~ (Q(..)

(2)

x fixed
Such a property has been observed at SLAC for electroproduction
and it is expected to hold for weak processes in the same range of
values for q2 and W : q2> 1 GeV2
W > 2.6 GeV.
lloreover small breakdown of exact scaling seem to be present
both at SLAC T and Fermilab energies. They have a natural explanation in the framework of asymptotically f~ee gauge theories. We
shall not discuss this problem in that part, being essentially
interested to the dominant features of the data.
We then assume exact scaling in what follows and the structure
functions F ~ being q2 independent, the y and E dependence of the
double differential cross-sections (I) is entirely known and the
two important features are
i) the cross sections rise linearly with E ;
ii) the cross sections are second order polynomials ~n y
Therefore all the dynamics is concentrated in the x dependence
of the scaling functions F~ (x).
3) Energy Distributions
The fixed y distributions are obtained integrating over x the
double differential cross-sections. Such an integration defines the
first moments I? of the scaling functions

T
and we get

d~~

-~

=:

f1 rae "F",(~) ~a:.

0

"

(.3)

~dG
J-;

Properties i) and ii) obviously hold for the energy distributions (3)
4) Total Cross-Sections
Integrating now over y the energy distributions, we obtain the

At SLAC energies, it is not completely clear what type of

scaling variable: x,x', etc and of scaling function we must
use and quantitative statements depend crucially of these choices.

448

M.GOURDIN

total cross-sections

G" "" G~M G

1t.

.A.

(4)

where the slope parameters A" and A ware linear combinations of the
first moments I~ given by

A' - ir~ r~
= ..l:. + ~ -t:....-~
c1~ _ ,
:i .J'V 't" T
- ""+

1'.s -

(5)

"'0

Again, property i) of linear rising with incident energy

emerges as a consequence of scaling.
5) Distributions in x
The x distribution are obtained integrating the double differential cross-sections over y. They contain the complete information
over the dynamics but they are relatively difficult to measure with
accuracy at low energy ( E~I - 10 GeV) because of poor statistics.
6) Constituents
It is convenient to introduce two constant parameters A and B
measuring the relative importance of the three helicity contributions.
The usual definition is
Il
_1.1

1"_ -

t .. 1" 1 ... _
=____

r. .

(6)

and we immediately deduce the positivity constraints

The energy distributions (3) can be equivalently written in

the form

~ G~E !.-t~".'I~{ (1.:!:~) + (-i~la(~ :OB) - \I' (t-lIl}

From electroproduction experiments at SLAC we know that the
longitudinal contribution is small as compared to the transverse
one. It means, in this language, that A is close to unity. In fact,
A = 1 is the integrated form of the Callan-Gross relation.

The upper sign refers to

reactions.

reactions and the lower sign to

449

WEAK CURRENTS AND NEW QUARKS

In what follows, we shall neglect the longitudinal part for

weak processes replacing in equation (7) A by unity. The only
remaining parameter B describes the shape of the y distribution.
II - GARGAHELLE DATA
The experiments were made using neutrino and antineutrino
beams with energy E in the range 1 - 10 GeV. The Gargamelle bubble
chamber was filled- with freon and the main results are the following.
1) Total Cross-Sections
They are compatible with a linear rising with energy

6"'::r (O/81.~O/o3)

and their

'oa~ ~~"I./Ge'f'

cc = (0 , 2.6::t 0, 0 1) e -:10-a~ Q'11~GeV"

ratio
"eo = 1:"'/6: remains constant with energy
Q

0,35:! 0,03

2) Energy Distributions
For neutrino the y distribution is essentially constant and
for antineutrino it behaves like (1 - y)2 to a good approximation.
Using equation (7) with A = 1 we immediately conclude that the two
parameters Band B must be close to unity. To a first approximation
we have the simple relation

-.a.

oLJ

=u.
__
U'JI

"to

-i _

'(\c

:1+ ('c.

From equations (6) it follows that the contribution due to

helicity?\
-1 strongly dominate over the contribution to
helicity 1\ = +1.
3) Conclusion
The data are consistent with the consequences of Bjorken
scaling but we do not have a direct and unambiguous proof of
scaling from these experiments because of the poor statistics
at large values of q2 and W2 .

M.GOURDIN

450

III - TWO COMPONENT QUARK PARTON MODEL

How electroproduction, neutrino and antineutrino data at these
energies can be interpreted with a parton language is now folklore.
We only wish to show in this section how a particle quark parton
model can reproduce the existing data and make predictions.
1) Basic Ingredients
For a quark parton model we need two types of ingredients.
Firstly, the distribution functions for each type of quark and
antiquark in the hadron. We call it Dj (x) taking in mind that
these quantities are probability distributions and therefore they
are positive. Secondly, we have point-like coupling constants
between the current and the quarks or antiquarks. These coupling
constants depend on the internal symmetry group used for strong
interaction via the definition of currents.
2) Two-Component Model
In the so-called two component model, we have two types of
quarks.
a) the valence quarks, u, d, giving rise to non-diffractive
distribution Vu (x) and Vd (x). The integral over x of these distributions is just the number of valence quark in the nucleon and
it is therefore finite.
b) the quark antiquark pairs building an isoscalar sea,
uu + dd, ss, ce, etc. The integral of these diffractive distributions
Sq(x)can be infinite if the number of pairs in the sea is infinite.
Such a quantity is related in the quark parton model to the limit
at x = 0 of the scaling function F~(x). Recent electroproduction
data at Fermilab indicate a continuous decreasing of F2 (x) when
~ =I/x increases. The limit at x = 0 requires at infinite incident
energy but in the present situation the extrapolation of existing
data suggests the possibility for F~ (0) to be zero or something
finite. Therefore such a question remains open.
3) SU(3) Quark Parton Model
In this range of energies, it is reasonable to work with a
parton model based on SU(3) symmetry. We have three basic quarks,
u, d, s and the weak charged current is the usual Cabibbo lefthanded current.
The first moments dJ of the quark distributions Dj(x) being
defined by

d ~.

1
\

":It.

::ild (-.:.) ~oc.

451

WEAK CURRENTS AND NEW QUARKS

We have only four independent parameters in the two component

quark parton model.
a) Valence parameters Vu , Vd
b) Sea parameters

Su

Sd, Ss

For simplicity we shall choose a SU(3) symmetric sea

Su = Sd = Ss = S and we get the following expressions for the
quantities I~ which govern the energy distributions and the total
cross-sections
Proton target

-.

::: S 40

-< V"cl ~.}. eC.

.e

= .s +

~ V"""

-~'p

T J~
:a

.59

.... + :.
Neutron target

- :;;1'>

_ OlIn

T;tl\.

J_=

- 1- = S
T';,y\'
-to =oS

- n

..L

.s + ~ \f"'" (k~Qc:

= $+~"'~

IV - ANALYSIS OF GARGAMELLE DATA

The small number of free parameter of the simple two. component
quark parton model allows us to take into account in a consistent
way two small corrections.
i) the difference between the number of protons and neutrons in
the freon Nn / Np = 1,19 ;
ii) The

\A Sl

contributions of the Cabibbo current

The input quantities are

i) the neutrino and antineutrino total cross-sections in freon
S'~ and fj 11
ii) the difference between proton and neutron electroproduction
scaling functions integrated over x

-e"

-e."l

1. -.1

Si l__ do
0

_e~] _l
tt2" (:lC) - ';, (~~ a.Q:.

-=

0, 04 ~O,O 2

which fixes the isovector component of the valence parameters

Vu - Vd

0,12:! 0,06

The output of the calculations are the value of the isoscalar

component V = Vu + Vd of the valence parameter
V

0,53.!" 0,05

(\0)

452

M. GOURDIN

and an inaccurate estimate of the sea parameter

s = 0, 004

~ 0,

19

( 1 1)

Let us notice that the origin of the large error for S is

closely -if not entirely- related to the large uncertainty entering
in the electroproduction difference.
1) Main Features
Let us first remark that the model is consistent with neutrino
and antineutrino experiments, the non trivial positivity constraint
being satisfied S~ 0. Moreover in this framework the dominance
of the helicity ~ = -I contribution implies the dominance of the
valence quark contribution and in the one standard deviation limit
the sea amount S/V is less than 4,4%.
The shape parameter for the y distributions is easily computed
and its value
B = B
(J 2)
0,98.!" 0,07
is close to the estimate made directly from the data neglecting
small corrections.
Finally, it is well known that non interacting partons called
gluons are needed in this class of model in order to insure the
conservation of energy momentum . From

cS=I--r

a large amount of gluons is needed

e.

0,46.:t 0,02

(J 3)

2) Predictions
We are now able to make predictions for other quantities a part
of them being already measured.

JL

i) Isoscalar electroproduction integral

- ep

J.

1 e"l.::

-i.

eta
et\
\'2., (:r.) 1'~ (-:t.)] d~.: 0 / 2.98 O "Oil

(J 4)

The value measured at SLAC : 0,30 ~ 0,02 is remarkably close

to the computed value and we see that in addition electroproduction
data are compatible with this simple model.

We do not discuss the dynamical role of gluons in this paper.

WEAK CURRENTS AND NEW QUARKS

453

ii) Neutrino and antineutrino total cross-sections on proton and

neutron separately. The predictions are
(15)

Again the large error is to be attributed to the large uncertainty on proton-neutron difference in electroproduction.
A preliminary result from Argonne
=

is in

2,08:r 0;23

good agreement with the model prediction

iii) Strangeness violating cross-sections can also be evaluated

and the result is

G~(A'3=\ )

~I:o'
OS;
\0 (AS=-I)

CO ~o"

<

fV

0,4%
(J 6)

5,3%

No reliable measurement of these quantl.tl.es has been obtained

because of the difficulty of separating 1A Sl = 1 events from
associated production of strange particles. Nevertheless they
have been observed in both neutrino and antineutrino scattering
showing, in particular, that the sea amount is not strictly zero.

PART II
FERMI LAB DATA
We now study the data obtained at Fermilab by various groups
for charged current reactions. We can classify the experiments into
two categories.
1 - Counter experiments;
2 - Bubble chamber experiments.
It is not yet absolutely clear what is happening at these
energies. Anomalies in the y distribution for antineutrino scattering
have been observed from a long time by the HPWF collaboration. On
the other hand, all groups agree about the existence of dimuon events
not coming from the decay of a heavy neutral lepton. But the precise

454

M. GOURDIN

magnitude of these new effects is not clear for a theorist and it

is always very delicate to choose between various experiments.
On the theoretical side we can classify the various types of
models into two categories
I - without right handed currents: standard models;
2 - with right handed currents and new quarks.
We discuss successively the two possibilities using a simple
two component quark parton model and we compare the result of the
analysis with what we have obtained at Gargamelle energies in
Part I. A lot of interesting experimental informations are missing,
in particular absolute differential and total cross-sections. Nevertheless some conclusions can be reached but they strongly depend
on what data we use.
I - LEFT HANDED CURRENT MODELS
1) General Structure
We assume, in this section, that only left handed quark
doublets are present and the charged current coupling constants
are chosen in such a way that the associated neutral current is
diagonal in the quark basis. (see Appendix).
It then follows for the parameters I + and I - , a general
structure independently of the precise number of quarks doublets.
For an isoscalar target we get
_ W"
\
--

(17)

where V is the isoscalar valence first moment and SI and S2 two

linear combinations of sea first moments given by
(18)

In these models, we implicitely assume that only quarks with

electric charge +2/3 (set Q+) and -1/3 (set Q-) are present.
Let us consider as the simplest example the SU(4) case with the
Glashow-Iliopoulos-Maiani mechanism. We have two doublets

where the index c means here the usual Cabibbo rotation for d and
S quarks. The expressions of SI and S2 are
SI = Su + Sc

S2 = Sd + Sc

455

WEAK CURRENTS AND NEW QUARKS

with the relation Su = Sd implied by the isoscalar character of the

sea.
In general, the above current is not charge symmetric.
We only have the weaker relation
_

l"'(

J.-t

io

'))f'I'

_;;H' _ 5N"

:: 1+ ... ..1

= \f +5

+.s~

(I 9)

and violations of charge symmetry occur via the sea excepted if the
sea is SU(4) symmetric.
2) Energy Distributions and Total Cross-Sections

The shape parameters Band

and the ratio of total crosssections 1r~
are easily computed from the previous expressions of
1+ and 1-

~: V-SITSa

V-+S,+Sz.

I'c.~

'b:

'4+5, -Sa.

(20)

V~S,+Sl.

V+ $,;- 3Sl.

3'1"+S,,,,,,3S2,

It the amount of sea is small as it is at Gargamelle energies

Sj <.<. V we get

and we obtain the following qualitative features

i) Band
are both close to unity;
ii) Charge symmetry violation is small;
iii) The ratio -r.. is close to 1/3
Conversely, a dramatic change between low and high energy for
one of the shape parameter Band B would imply a dramatic change
for r~ and it can only be explained, in this framework, by a
considerable increasing of the sea amount.
In fact in this class of models, where the first equality(1,)
is satisfied, the ratio 1C is related to the shape parameters
Band B by

Such a relation has already been used in Part I to obtain a

first estimate of the shape parameter B using only the ratio
of total cross-section given by the Gargamelle experiment.

M. GOURDIN

456

(21 )
3) Harvard-Pennsylvania-Wisconsin-Fermilab Collaboration Results.
When the data for neutrino and antineutrino y distributionsintegrated over x for x < 0.6- are represented with the shape
parameters Band B, they exhibit noticeable differences with the
Gargamelle resutls.
B exp

0.75 0.15

B exp

0.55 0.13

(22)

the incident beam energy being larger than 50 GeV.

From the previous considerations, the only possible fit of
these data, in the framework of left handed current models, implies
important sea contributions.
Absolute cross-sections being not available, we can estimate
only the ratio S/V. Neglecting new quarks contributions and using
as previously an SU(3) symmetric sea
2S
a typical value for the ratio S/V 1S
S/V = 0.17
within 40%. The corresponding values of Band B
B = 0.77

-B = 0.56

fit very well the data and the corresponding ratio of total crosssection
~ = 0.52

is also in agreement with HPWF results.

Fits of the antineutrino y distributions integrated over x 1n
the two ranges x < 0.15 and 0.15 <x< 0.60 are shown on Fig. 2
for three values of the ratio S/V: 0.10 - 0.17 - 0.24. The y
distributions integrated over x with x < 0.6 are less sensitive
to the value of this ratio as it is clear from Fig. 3 for antineutrino data and Fig 4 for neutrino data.
4) The main feature of this analysis of HPWF data with left handed

WEAK CURRENTS AND NEW QUARKS

457

--------------------------~r_--~~_,~

\0

V
H

.0
N

tn

>.

Q)

::I
00

-...4
~

,?o

,
I

,
I

,
y

/
I~

oP)

458

M. GOURDIN

x<.6

30
c
..0
'-..

2
c
CII
>
<I>

20

10

y
Figure 3

1/

x<.6

y
Figure 4

459

WEAK CURRENTS AND NEW QUARKS

current model is acensiderable increasing of the sea amount by at

least a factor of 4 with respect to its value at Gargamelle energies.
It is obvious that such an important change associated to the
nucleon structure will have observable consequences in electroproduction. In the same two component quark parton model, the isoscalar electroproduction integral is given by
(23)

or using the same sea as previously

.,. e~ Ie" _ ~ "

Using our previous estimate of S/V at Fermilab energies
~

the
bracket becomes 1.41 and it looks very unlikely to compensate such
a large increasing by an equivalent decreasing of V. Therefore, a
large effects would appear in ~ inelastic scattering experiments
performed at Fermilab. Up to now, there is no indication of such
an effect.
On the theoretical side, the only consistent explanation of
such a modification of the sea amount is a breakdown of canonical
Bjorken scaling. In asymptotically free gauge theories such a
breakdown exists with logarithmic factors of the form Log ~~l'" where
~is the group renormalization unknown mass parameter. In this
framework, the valence moment V decreases and the sea moment S
increases when q2 becomes larger and larger. H~wever the magnitude
of this effect is expected to be considerably smaller than what
seems to be needed to accomodate HPWF data in their present form
with left handed current model.
5) Conclusions
i) The standard models with only left handed charged currents
are not excluded by existing experiments.
ii) The counter experiments performed by the HPWF and CTF groups
require a very large value of the sea over valence ratio S/V.
iii) The bubble chamber experiment E 180 does not need a large
value of S/V and
with

S
V

0.06

we get B

0.8

which is the best fit of the data at all energies.

iv) The production of new quarks is always very small in these
models. Consider, as an example, the production of charmed quark~
It can occur either on the sea quarks or antiquarks - for~ and ~
scattering - or on the d valence quark with a damping due to the
Cabibbo factor tg 2 c - for'" scattering - and we have the constraint

460

M. GOURDIN

c:: H means charmed hadrons.

Therefore in order to account for dimuon experiments we need
a large leptonic branching ratio for the decay of charmed particles
which has to be between 10% and 15%.
where

II - MODELS WITH RIGHT HANDED CURRENTS

The terminology means that in addition to the previously
studied left handed currents, we allow the possibility of having
right handed currents.
If one believes the HPWF and CTF data, and if we refuse the
considerable change in the sea amount found in the previous section
as unphysical, we are obliged to introduce new currents with right
handed helicity as an alternative way for analyzing the data. The
sea amount will remain small if these new contributions involve
valence quarks. We then can have two doublets of this type

where t and b are new heavy quarks of respective charges +2/3 and
-1/3. The first doublet will give valence contribution to neutrino
scattering and the second doublet valences contributions to antineutrino scattering. A comparison of Gargamelle and HPWF results
shows clearly that the sensible point is antineutrino scattering.
Therefore, we begin the discussion by considering the second
handed doublet.
1) General Structure
The two parameters 1+ and 1- for antineutrino scattering off
an isoscalar nucleon taget are now given by
-~N
I
_

-+

- -;;.,

~_::'J""+S&

W+S2,.

(24)

where the u-quark valence moment W has to be equal to V if universality holds and if x scaling is valid.
The shape parameter B of the y distribution becomes

V+S, - \N

-S,

V+S, +W' .... S,z.

and using the HPWF value B = 0.451' 0.13 we get

461

WEAK CURRENTS AND NEW QUARKS

0.38! 0.14

We then get, in that way, an ad hoc explanation for the large

violation of charge symmetry without the necessity to modify the
sea amount by a large factor.
Clearly the W/V is far to unity and one can think to a
breaking of universality. That is deeply unaesthetic and we must
keep in mind that the new b quark we introduce in our doublet
cannot be a light quark and its mass is expected larger than the
charmed quark one. Therefore, thresholds effects are very important
and they give the explanation why the ratio W/V obtained assuming
x scaling is not unity even if all doublets enter in the current
with the same coupling constant.
2) Modified Quark Parton Model
In their analysis of operator product expansion near the
light-cone for color gauge theories, Georgi and Politzer have
shown that the correct scaling variable to use for the transition
of a light quark to a heavy quark is not Bjorken variable x but
the quantity !. defined by

'2.

~ 1-m~
2M

=re +

(25)

where 'm., is the heavy quark mass. Of course, for light quark
light quark transi tions we put ~ = 0 and we recover the x
variable.
The consequences of this observation have been studied by
Barnett and many authors where we refer for details. We only
give here the rule of the game. In units of G2ME/1C the quark
model expressions for the contribution to the double differential
cross-section of the transition
light quark j ~ heavy quark k
is given by

ll,

reactions with ~

v-A\.

weak current

\.V+,,/~

jl~ J]4<tc) (-i - .'t~:!.1

" ...... reactions with

",,+A '\

(26a)

weak current

\. '1-~ /

(26b)

M.GOUROIN

462

For antiquark - antiquark transitions the two expressions are

exchanged.
The main effect of the change of scaling variable from x to
is a suppression of the production of heavy quarks off light
quarks as compared to what is expected in the x scaling region.
Kinematics is responsible of this fact

(27)

~O

by the suppression of the low

region. Explicit calculations
using realistic valence and sea quark distributions show that x
rescaling is very slow and it is far to be achieved at Fermilab
energies even for the production of charmed quark. We then understand better why the effective valence quark ratio W/V is only
0.3 at 100 GeV.
3) Analysis of experiments
It is clear that the magnitude of the threshold effects depend
crucially on the heavy quark masses. Fixing the charmed quark at
h1~ = 1.65 GeV, the b quark mass is considered as a free parameter
one has to determine from experiment.
We show, in Fig. 5, a fit of HPWF y distributions integrated
over x in the two ranges x < 0.15 and 0.15 <a< 0.60. The sea
over valence ratio is fixed to 10% and the curves correspond to
quark parton model predictions with ttlb = 3 GeV, 5 GeV and 6 GeV.
The lowest value 1nb
3 GeV seems to be excluded by the present
data and acceptable fits of the data are obtained with 11lb in
the range 5 - 6 GeV.
An analysis of CTF data gives essentially the same result for

4) Other doublets
The second right handed doublet involving the valence quark
u and a Q = 2/3 charged quark t will give analogous effects in
neutrino scattering. But in that case experimental data do not
seem to require such a doublet and standard models with only left
handed currents fit easily HPWF neutrino y distributions with
S/V(!: 10%.
However, if one insits for the presence of such a doublet,
the data imply a t quark mass as large as 5 or 6 GeV. This means
that the hypothetical t quark cannot be identified with the
charmed quark.
Ohter possible right handed doublets not involving valence
quarks can be introduced. They will not destroy the agreement

WEAK CURRENTS AND NEW QUARKS

463

.x <.15

30
,,

----

.............. ,

,,

20

,,

,,

ell

<II

>

<II

10

Figure Sb

. ..
c

n 20
...........
ell

C
<II

10

y
Figure Sa

M. GOURDIN

464

between theory and experiment as long as the sea over valence ratio
remalns small. The simplest example of such a possibility is the
doublet.

~ IR

From the point of view of anomalies we expect each new doublet

of quark to be accompanied by a doublet of leptons of same helicity.
There is some evidence for a new heavy lepton at SPEAR of mass
around 2 GeV. It can be the leptonic mirror of the new b quark
considered here.
In that case the observation of the charged lepton spectrum
in the decay mode

must exhibit the characteristics of a V + A interaction at the

heavy lepton vertex.
5) Conclusion
i) Reasonable fits of HPWF and CTF data are obtained using
only one right handed doublet J ~l R in addition to the usual left
handed one of the standard model. The mass of the new heavy quark
b has to be as large as 5 - 6 GeV.
ii) Such fits use a sea over valence ratio of 10%, e.g. the
variation of S/V between Gargamelle and Fermilab energies is
considerably smaller than what is needed in the standard model.
Moreover quantities like the ratio lrc of total cross-sections
and the averaged value of y for antineutrino energy distributions
exhibit in the 30 - 50 GeV region a variation with energy which
cannot be reproduce by threshold effects only. If correct, this
feature of the data implies an increasing with energy of the
ratio S/V and now asymptotic freedom can account for.
iii) The introduction of the right handed doublet

I I R provides
~

a new source of dimuon events in antineutrino scattering. Such a

contribution turns out to be important because the transition
u ~) b occurs off a valence quark without Cabibbo suppression
factor. But there is a damping due to heavy quark mass effect
which decreases when energy increases. We therefore expect more
dimuon events induced by antineutrinos than by neutrino and
the associated ratio will increase with energy.
iv) We now have two types of new hadrons, those contalnlng
c or-C quarks and those containing b or b quarks. The branching
ratio for the leptonic decay of these new hadrons is now estimated
between 5% and 10% from the dimuon rates and it is compatible with
the same value for both types of new hadrons.

465

WEAK CURRENTS AND NEW QUARKS

v) The rescaling in the x variable is very slow and it is far

to be reached at Fermilab energies even for c- quark production.
Therefore, we expect the shape parameter Bito continue to decrease
and the averaged value
to continue to increase when the
incident energy increases. In a minimal model with two left handed
and one right handed quark doublets, we get

:Y);

64

S.-B.
V

PART III
IMPLICATIONS FOR HADRONIC NEUTRAL CURRENTS
It seems reasonable to r.eTate weak charged, weak neutral and
electromagnetic currents in a unified scheme and gauge theories
provide a nice framework to that purpose. The aim of this part is
to make predictions for neutral current cross-sections using the
results of the analysis made in Parts I and II.
We make the usual choice for the gauge group SU(2)~U(I) and
we have the mixing angle ~ w as a free parameter. Therefore the
neutral current cross-sections are functions of&w. A comparison
with experiment determines an allowed range of values for 9w and
the consistency of the scheme is obtained from positivity constraints.
I - LEFT HANDED CURRENT MODELS
1) In the quark representation, the neutral current is assumed to be diagonal. When the charged current is purely left handed
we have the following structure

wher x = Sin29w. As previously Q+ (Q~) is the set of quarks with

electric charge + 2/3 (- 1/3). The coupling constants bj and bk
are 1 or o depending of the details of the model.
2) It is now straightforward to compute the scaling moments
I + and I _ entering in the y distributions and in the total
cross-sections. Because of the hermeticity of the neutral current,
neutrino and antineutrino parameters are simply related

466

,.. -

M. GOURDIN

' ).)

....
_\

.~

:to

and we get the following results

Proton target

I +'P =.

:I_P .:
Neutron target

.2 (Si: _

.... S",c:..

~.)2Vu..,"t-.i (i- ~)2 "Q -+.! N'C

.e~c:c. a (.I.\..1- ...../. V~) -t- oS .,~

Ii"n. =

.1~ ~

..::: (A Vu.+ V.. )

-2

'-1- ~Ja\f,u.:-l.(i.- ~Ja tel +

where the sea contribution

S""c.

is given by

bL

StYe:

S~.. =~~+ ~ (~- ~)\~~J S~ "'I~~" t (A~- ~~+ S 3S~

~

3) In order to analyze the Gargamelle data, we use the

SU(4) symmetry group. In that case
L

--

b c: .....
- Ii

:a

b~ ~ -i

The neutral current sea is simply a linear combination of the

parameters SI and S2 already introduced for charged currents.

.5N'C:':

.\
(L

~
..... t3'2.,S . . ~ (~_ $0 +~~S ..
~
.s J . .
4.a
.,
..
T"

with
Neglecting charmed quark and antiquark distributions in the nucleon
and using as previously an SU(3) symmetric sea

we get

It is now very easy to compute neutral current total crosssections in Freon and we obtain quadratic functions of x. But the
usual procedure for presenting the data is to compare inclusive
neutral current and charged current cross-sections.

467

WEAK CURRENTS AND NEW QUARKS

Again these ratios R>l and R ~ eire quadratic functions of

x. They have been represented in Fig 6 and 7 using for the
parameters Vu. Vd and S the values obtained in Part I from the
quark parton model analysis of electroproduction and charged
current data.
The most recent Gargamelle data are

~))

0.25 0.04

'A,-

0.39 0.06

and we can easily deduce

for )l scattering
for ;:i scattering

the values of x allowed by experiment.

0:..

r.II:.,.

o,a2.:!:

~'.1~

0,08 _ O"G6

From these results we see that the quark parton model with
SU(4) symmetry is compatible with neutral current experiments when
the parameter x is chosen in the range

4) For high energy experiments we use a model of the type

previously described and we restrict the calculations to an isoscalar target N. The slope parameters for the total cross-sections
are of the form

A~~_

- We::. -

(i

_~

~O 0!.'2.)y- + (1.- ,~+ -1~~~ 5

:l..1l
S?
A
'a: ~O _'2.)...,. .... (-\ _~+ -iGa:. '.5
(,-a+~""'"
-.5
~J

(28)

h
We have computed t h e tree
r a.t 1 R~
0 s , Rii"and"""N
'

,....

:.,)

"'r..,:: '0 HCo /G NC

and x = 0.40 which are values in the range

for x = 0.3 x = 0.35

suggested by low energy data. The three ratios depend only on the
sea amount S/V and we consider the high value S/V = 0.17 implied
by HPWF and CTF experiments and the value S/V = 0.06 which fits the
E 180 bubble chamber experiment. The results are given in Table 1

468

M. GOURDIN

0.1

sin 2 w

0.4

02

Figure 6

R\J

0.5

0.4

0.3

0,2L.......-_ _ _-=-'=-_ _ _ _......,..r.._ _ _ _ _..L-_ _ _. .

0.2

0.4

Figure 7

0.6

. 2 w
Sin

WEAK CURRENTS AND NEW QUARKS

0.30

0.35

0.40

S/V = 0.17

0.27

0.24

0.23

S/V = 0.06

0.27

0.24

0.22

S/V = 0.17

0.33

0.33

0.34

S/V = 0.06

0.36

0.37

0.39

S/V = 0.17

0.64

0.7I

0.88

s/V = 0.06

0.56

0.64

0.74

469

.,.-

TN

Table 1
Interesting qualitative features can be read on Table 1
i) the ratio R)J is rather insensitive to the parameter S/V.
We then expect R)J to be energy independent.
ii) the ratio R;3 decreases when the sea amount S/V increases.
On the other hand, it varies slowly with x.
iii) the ratio lrN increases when S/V increases and it varies
rapidly with x.
The presently available experimental data at Fermilab energies
come from counter experiments.
HPWF collaboration
~~

= 0.29 0.04
'rN

CTF collaboration
~~

= 0.25 0.04

1"",

'R~=

0.39 0.10

0.48 0.20

~:a

0.36 0.11

0.75 0.14

470

M. GOURDIN

"

The results of the three experiments for Rand R

do not
exhibit any dependence with energy but they do not exclude such a
dependence as that shown on Table 1.
Concerning the quantity TN' we must keep in mind that it
has not been directly measured and the quoted numbers are the
results of computations involving R, RSi and r c the ratio of
charged current total cross-sections.

Y'~=

'Rii
_
1=1

JC

1"c:

In particular, the HPWF and CTF groups take different values

forT c
II - RIGHT HANDED CURRENT CONTRIBUTION

1) Let us discuss only for simplicity the minimal case of

two left handed doublets as in SU(4) and one right handed doublet
~ JR The moment integrals 1+ and 1_ become for an isoscalar target

_N"
I
-+ ::.

(If

;I

(i
- .2'iJ1t
\..4

(1 -

.3

-5

+ ~;) V-

.. 5":9~

+ 8t(~

,-V "1- .stJ'~

As previously, we retain only, in the sea, contributions

coming from Ull, dd, S8 pairs in an SU(3) symmetric way

S ....... _ (~ - ~ T ~1Ie.~.).s
.,,'- -

.3

.9

The slope parameters for the total cross-sections follow

immediately from these expressions
~
6 1b.))tr
~
i{~-+~()CI!. ) V T ( ~ -~.;~~S
n tic. ::
9
(29)

A:DN"
~~

(i~
s..ua
8~
I~ _ ~ +- ~~) V + c.S _.2~ +~.)S

~~

~1-

~.9

2) By comparing equations (28) and (29), we see that the

variation with x of the slope parameters At-r
and A S;""
are
very different in the two models. In the range 0.3< x < 0.4
previously considered the ratio 1rN of neutral current total
cross-sections is essentially independent of x and it increases
substancially with the sea amount S/V as shown on table 2.

471

WEAK CURRENTS AND NEW QUARKS

0.30

0.35

0.40

0.65

0.64

0.63

1%

0.66

0.65

0.65

10%

0.73

0.73

0.73

Table 2

Because of the large experimental uncertainties these numbers

agree with low and high energy data but the comparison is unconclusive.
As noticed by Barger and Nanopoulos the valence contributions
are the same in the two models for the particular value x = 0.375
and this fact makes more difficult a clear choice between the
two types of models if nature choose a value of x close to that
value.
3) Let us consider the ratios R ~ and R ; which compare
charged and neutral current total cross-sections. At fixed values
of x threshold effects are present via charged current cross-sections
and we expect a variation with energy of these quantities. We give
numerical estimates in Table 3 using an effective quark model where
right handed current effects are effective only at'3Q% at ,Fermilab
energies

1i~

"R

0.30

0.35

0.40

LE

0.28

0.25

0.22

HE

0.29

0.26

0.23

LE

0.54

0.46

0.40

HE

0.29

0.25

0.22

Table 3

472

M. GOURDIN

L.E

means low energy with s/v : 1%

H.E

means high energy with S/v

10%

Again we observe that the neutrino ratio R ~ remains essentially i~ependent of the incident energy E whereas the antineutrino
ratio R~ decreases rapidly with E. The effect is clearly more
important than with left handed current models but the experimental
data are not accurate enough to allow any definite conclusion.

473

WEAK CURRENTS AND NEW QUARKS

APPENDIX A
KINEMATICS
The relevant diagram for inelastic lepton or antilepton
scattering is shown on Fig 1
Three Lorentz scalar variables can be defined

w2 = _

,.M

(p + q)2

-p.q

and they are not independent

W2 = M2 + 2M)/ _ q2
AS usual M is the nucleon mass.
In the laboratory frame, we call as E the incident beam
energy
-p.k = M E and in this frame ~ represents the transfert
of energy between lepton e.g. the energy carried out by the current.

- -.

Finally, we introduce two dimensionless scalar variables

~

and we get the useful relations

q2

= 2 ME

xy

M2 + 2M E Y (1 - x)

~E=-W

Figure 1

474

M. GOURDIN

APPENDIX B

ELEMENTARY QUARK TRANSITIONS

Neutrino Scattering with SU(4)

Ac

4S

...
d

AC

:,I) u

VALENCE

ud

Sin

ftc

SEA

uu

. . ':!u

Sin

8c

SEA

- -

6s

= I

ss

us

Sin

8c

SEA

uu

~ su

Sin

Ic

SEA

..- 0

Sin8c

VALENCE

dd

=II>

cd

Sin8c

SEA

cc

~ dc

Sin8c

SEA

- -

AC

8C

dd

6S =

A C

COS

L\ S =
ss

~ cs

cc

sc

ec
Cos e c
Cos

SEA
SEA

Antineutrino Scattering with SU(4)

Ac

As =
u

at>

- -

at) dt!
uu

dd

::!) ud

cosec

VALENCE

Coslc

SEA

Cos

SEA

475

WEAK CURRENTS AND NEW QUARKS

AC=O

~S =

-1

If>

..

Sin

su

Sin

::!>

~s

Sin

..

-
uu
ss

As

-1

AC

VALENCE

ec

SEA

ec

SEA

..

cc

11> dc

Sin 'c

VALENCE

dd

cd

Sin8 c

SEA

Cos, c

SEA

Cos

SEA

..
AC

= 0

- ., -

AS= -1
ss

cs

cc

. . -sc

I IR Doublet added
~

New quantum number T associated to the b quark. All transitions

Ac=o

As=o

- - -

Neutrino Scattering

bb

-'>

..-

uu

AT

ub

SEA

bu

SEA

Antineutrino Scattering

+1

AT

VALENCE

- -

-1

uu

b~

SEA

bb

ub

SEA

Doublet added

All transitions

6S

476

M. GOURDIN

- => - => -

Neutrino Scattering

AC

+]

4s

88

C8

SEA

CC

8C

SEA

Antineutrino Scattering

AC

-]

As

cca

=b sc

SEA

ss

::JlI, cs

SEA

-]

477

WEAK CURRENTS AND NEW QUARKS

APPENDIX C
NEUTRAL CURRENTS

We study models with only two types of quarks with electric

charge + 2/3 (Q+) and - 1/3 (Q-).
m quarks
n quarks
The electromagnetic current has then the following quark
representation

For the weak charged current, only quarks of the two sets are
related and we introduce two m x n matrices respectively associated to left handed and right handed couplings.

J ~c =: ~ q ~4 (1i-t.6 )

\:

~~ \ 't

-+ i

<\ 1'(4 (1.- Pes) \ ~ ':" J Cf

1) Neutral currents in SU(2)QP U (~) models

In this class of models, the weak charged current is simply
related to the weak isotopic spin current.
-

etC!

/""t" i.

.,.2.

\"

.jt'-;a-2LVt"-+~...Jt"'")

:e.=.

\J r

(J~
_i
,-

:1:' )
,-

The third component of the weak isotopic spin current is

obtained from the commutation relations

[J',..;.ja. > J'~-i..ja.J = ~:r3

All these expressions are valid for both the left handed and
right handed parts using the projection operators i %~S

In the class of models with a SU(2)QO U(~) structure the neutral current is given by
a.
_~

oJ -:::: .l J ~ - ~ A In. 9 w J

tI-

Using the previous notations, we obtain

'l\ 0toe. (1. ... 0$) 1A~~ _ A~A~ \ '1

... :i. ~

~~ (d - ~~ )

~
At\A~
0

A~ Aft ,

478

M.GOURDIN

2) Diagonal neutral current

Starting with the Cabibbo current constructed from only three
quarks u, d, s, we obtain a neutral current having a strangness
violating component. From Experiment such a component is not present
at the same level as the strangness conserving one. We have to
cancel the unwanted part and the simplest way is to introduce a
new quark c as proposed by Glashow, Iliopoulos and Maiani. In this
model the neutral current is diagonal in the quark representation
and the matrix AL is orthogonal
A

=l .. Cos
&c
Sin e

Sin
Cos

ec I

&c

Bc being the Cabibbo angle as usual.

We now restrict ourselves to models where the neutral current
is diagonal in the quark representation. From the previous calculation, the constraints on the matrices A are

AA and A

A are diagonal matrices

for both the left handed and the right handed parts.
In the particular case where A is a squared matrix
get the following properties

i) AA

n, we

ii) The irreducible parts of A are proportional to unitary

matrices

I\~

U,: .
I

:
,

~.l. U~I

- -,..--

-~---.
(

'l?Ut>

If time reversal holds, A can be chosen as real and the

hermitian conjugate matrix A~ reduces to the transposed matrix A T.
Consider now a

m x m matrix A which satisfies the constraints

AA

=A *

= Diagonal

Let us add a new quark of electric charge + 2/3

B = \ b 1 b 2 . .. bm \

It is straightforward to prove that the constraints BB* and

479

WEAK CURRENTS AND NEW QUARKS

B B diagonal matrices imply

bl

b2

= ... =

bm

=0

The physical consequence of this result is the following

a theory with ( m + I, m ) quarks or ( m, m + I) quarks is equivalent
to a theory with (m, m) quarks. Obviously the equivalence can be
different for left handed and right handed current and we finally
get a new model.
Let us give an example with five quarks u, d, s, c, b. For the
left handed current in the usual SU(4) model, the matrix AL is that
previously given. The introduction of the quark b does not change
this part of the current.
A

1-Sin
Cos & c
c

Sin 9 c
Cos 6 c

o
o

But the new quark can be used for a right handed current.
If it is coupled to the u quark the two possible solutions,
assuming universality, are

o
o

corresponding to right handed doublets

or
With six quarks, u, d, s, c, b, t, three of charge + 2/3 and
three of charge -1/3 the situation is entirely different. The matrix
AL is chosen as unitary and we allow in that way a small violation
of time reversal invariance constrained by the structure of the
Cabibbo current. The matrix AR is free and only experiment will
determine its precise structure. From parity violation in neutral
current reactions we know that it cannot be an unitary matrix but
only a part of it, the matrices 4RA~ and A.f AR being only diagonal
matrices with some eigenvalues zero. The other eigenvalues can be
unity if universality holds.

480

M. GOURDIN

REFERENCES
A - EXPERIMENTS
1 ) Gargamelle Collaboration
V. BRISSON
~.

Talk at the Moriond Meeting Flaine (march 1976)

Talk at the Aachen Conference (June 1976)

KRENZ

Talk at the Aachen Conference (June 1976)

W. Van DONNINCK

2) Harvard Pennsylvania-Wisconsin-Fermilab Collaboration

A. BENVENUTI et al

HPWF Reports 76/1

76/4

C. RUBBIA

Talk at the ENS meeting Paris (January 1976)

A. BENVENUTI

Talk at the Vanderbilt Conference

D.K. MANN

Talk at the Storage Ring Meeting Flaine (Feb. 1976)

T.Y. LING

Talk at the Madison Conference (1976)

P. WANDERER

Talk at the Madison Conference (1976)

A. BENVENUTI

Talk at the Aachen Conference (June 1976)

T.Y. LING

Talk at the Aachen Conference (June 1976)

3) Caltech-Fermilab Collaboration

B.C. BARISH et al. Phys. Rev. Lett. 36 939 (1976)

A. BODEK

Talk at the Moriond Meeting Flaine (March 1976)

L. STUTTE

Talk at the Madison Conference

B.C. BARISH

Talk at the Aachen Conference (June 1976)

D. BUCKHOLZ

Talk at the Aachen Conference (June 1976)

(1976)

4) ANL - CMU - Purdue collaboration

M. DERRICK

Talk at the Aachen Conference (June 1976)

481

WEAK CURRENTS AND NEW QUARKS

5) American Russian Collaboration Hydrogen-neon Bubble Chamber E 180

F.A NEZRICK

Talk at the Aachen Conference (June 1976)

B Parameter at Fermi Lab Energies

HPWF

25

0,95

78

0,45

50

0,64

150

0,36

20

0,76

40

0,90

62

0,73

CTF

E 180

Fitted

Bvalues

for antineutrino Scattering

B - THEORY
1) Models for weak currents
S. WEINBERG

Phys. Rev. Lett 19

1364 (1967)

Phys. Rev. Lett 27

1688 (1971)

0, 10

+ 0,15

- 0,10
+ 0,22

- 0,26
+ 0,30

- 0,36
+ 0,08

- 0,10
+ 0,08

- 0,10
+ 0,12

- 0,18

M.GOURDIN

482

A. SALAMI

Proceedings of the XVlll th Nobel Symposium

Aspenasgarden (1968)

S.L. GLASHOW, J. ILIOPOULOS and L. MAIANI

Phys. Rev. D 2 1285 (1970)
S. PAKVASA, W.A. SIMMONS and S.F. JUAN
Phys. Rev. Lett. 35

70(1975)

H. FRITZSCH, M. GELL-MANN and P. MINKOWSKI

Phys. Lett. 59 B 256 (1975)
F. WILCZEK, A. ZEE, R.L. KINGSLEY and S.B. FREI~~N
Phys. Rev. D 12 2768 (1975)
A. de RUJULA, H. GEORGI and S.L. GLASHOW
Phys. Rev. D 12 3589 (1975)
H. HARARI

Phys. Lett. 57 B 265 (1975)

F.E. CLOSE, R.L. KINGSLEY, G.A. RINGLAND, D. SIVERS

Preprint Rutherford (March 1976)
2) Quark Positon model
M. GOURD IN

Lectures given at the Bonn Summer Institute

(August 1974) and references

V. BARGER, T. WEILER and R.J.N. PHILLIPS

Nuclear Physics B 102 433 (1976)
University of Wisconsin Preprint (1976)
3) Asymptotic Freedom in non abelian Gauge Theories
H. GEORGI and H.D. POLITZER

Phys. Rev. D 9 416 (1974)

D. GROSS and F. WILCZEK

Phys. Rev. D 10 2141 (1974)

D. GROSS

Phys. Rev. Lett. 32 1071 (1974)

Lectures given at the Ettore Majorana
Center (July 1975)

G. ALTARELLI, G. PARISI and R. PETRONZIO

University of Rome preprint (Feb.1976)
P.W. JOHNSON and WU-KI TUNG

Illinois Institute of Technology

preprint (1976)

483

WEAK CURRENTS AND NEW QUARKS

4) Threshold Effects
H. GEORGI and H.D. POLITZER

Harvard Preprints (1975)

H.D. POLITZER

Talk at the Coral Gable Conference

(Jan. 1976)

R. BARNETT

Phys. Rev. Lett. 36

A. de RUJULA

Talk at the Coral Gable Conference

(Jan. 1976)

E. DERMAN

Nucl. Phys. B 110 40 (1976)

J. KAPLAN and F. MARTIN

University of Paris Preprint (May 1976)

C. ALBRIGHT and R. SHROCK

Fermilab-Conference 76/50 THY

1163 (1976)

REVIEW OF MASSIVE DILEPTON PRODUCTION IN PROTON NUCLEUS COLLISIONS

J. W. Cronin
The Enrico Fermi Institute
University of Chicago, Chicago, 11 60637 U.S.A.
In this lecture we will review recent experiments on production
of lepton pairs in proton-nucleus collisions at large effective
dilepton mass. For the purpose of this talk we will consider the
production o~ lepton pairs with masses above the J/~ and ~' region
(m ~ 4 GeV/c). This range appears to be above the region dominated
by resonant production l and we will discuss it in terms of continuum
dilepton production. Dilepton production for m ~ 4 GeV/c 2 will be
discussed by Prof. A.J.S. Smith of Princeton University.
It should be noted that the study of dilepton production in
hadronic collisions was begun about eight years ago with the
pioneering experiment of Christenson et al. at Brookhaven National
Laboratory. 2 In this experiment whic~was characterized by high
sensitivity, but poor resolution, a significant dimuon continuum
was established. In retrospect a major part of the signal in this
experiment was due to the J/~ which was subsequently discovered at
BNL and SLAC.3,4 Nevertheless, when these resonances were subtracted, a substantial dimuon continuum remained with dimuon masses
which were a significant fraction of the total available center of
mass energy. 5
Within the framework of our current understanding, dilepton
production experiments are important for two reasons. First,
Yamaguchi 6 pointed out that the production cross section in nucleonnucleon collisions for the charged intermediate vector boson which
mediates the weak interaction can be related by the conserved
vector current hypothesis (CVC) to the production cross section of
virtual photons and hence to the production of di1epton pairs. It
is common to make the assumption 7 that the dilepton production
485

486

cross section

J. W. CRONIN

has the scaling form

2

do = m- 3 f(!!!. )
dm
s
where s is the square of the center of mass energy and f is a
universal function. Then one can use measurements of dilepton
production at the existing low energy machines to predict the
production of intermediate vector bosons in colliding beam
machines of the future.
A second reason for the importance of these experiments is the
connection pointed out by Drell and Vans of the dimuon production
process with the constituent structure of the nucleon which has
emerged during the past eight years. In this picture the production
of the dileptons is due to the annihilation of a constituent (quark)
in one nucleon with its corresponding anti-constituent in the other
nucleon. Recently a number of authors 9 ,lO,ll,l2 have made quantitative predictions of the cross section for dilepton production.
They use as input to their calculations the distributions of
fractional momentum of the nucleon carried by the quarks and antiquarks as deduced from deep inelastic neutrino and electron
scattering. As we shall see later, the predicted cross sections
depend sensttively on the assumed anti quark distributions and exact
quantitative agreement between the predictions and experiment cannot
be expected. Another important consequence of the Drell-Van process
is the scaling prediction referred to above. This latter prediction
is a necessary condition for the validity of quark-anti quark
annihilation as the dilepton source.
Recently two experiments have been completed at Fermilab which
have measured dilepton production in proton-nucleus collisions. The
first of these has been carried out by a Columbia-Fermilab-Stony
Brook (CFS) group. Part of the results on dielectron production
have been published. l Preliminary results on dimuon production are
also available. l3
The CFS experiment has observed both electron pairs and muon
pairs produced.in collisions of 400 GeV protons with beryllium and
copper targets respectively. Figure 1 shows a sketch of their
apparatus which is a symmetric double arm magnetic spectrometer.
Each arm views the production target at an angle of -90 in the
c.m. system of the nucleon-nucleon system. The electrons are
identified by lead glass detectors. Discrimination against hadrons
is achieved by the requirement that the momentum measured in the
magnet agree with the energy measured in the lead glass. Background
consists of accidental coincidences between the electrons detected
in each arm and real coincidences between an electron and a hadron
misidentified as an electron. The backgrounds become small for
m > 5.5 GeV/c 2 .
0

MASSIVE DILEPTON PRODUCTION

(I ~ ,:;\ JI rra,

TARGET

PRIMARY!
PROTON ,~
BEAM

J
/// _. . --

UP-ARM

MAGNET

BEAM
DUMP

....'

.I

~/
-

./

U3

U2

UI

Pb -GLASS

........

__ 01

............ ,_

10

02

' - ' - ' - :~ 1__

DOWN-ARM

Fig. 1.

/I~/

...,

95

.................. - '-......
~ ................... -

COLUMATOR

487

20
meters

'

'-',

30

--~'

40

Apparatus of the Columbia-Fermilab-Stony Brook group.

Data on dimuon production was also taken when beryllium hadron

absorbers were inserted in each arm near the target, and muon
identifiers were installed behind the lead glass.
Figures 2 and 3 show the measured cross section dcr/dmdyly=o for
dielectrons and dimuons respectively. Here y is the c.m. rapidity
of the dilepton system. This cross section is quoted per nucleon
assuming a linear A dependence of the cross section. The authors
quote this cross section rather than dcr/dm because the apparatus
is sensitive to dileptons produced nearly at rest in the c.m.
system (i.e., y - 0).
Another important characteristic of their observations is that
the vector sum of the two muon momenta has a net transverse momentum.
The mean transverse momentum of the dilepton system is observed to
be - 1.5 GeV/c in the mass range 4.5 to 8 GeV/c 2 .
A second experiment 14 by a Chicago-Princeton (CP) group has
recently been completed. They have observed the production of
dimuons in collisions of 400 GeV protons with a copper target.
Figure 4 shows a schematic view of the apparatus, which consists
of a highly asymmetric double arm spectrometer. One arm is a

488

J. W. CRONIN

do
dmdy

=0

p + Be

10- 35

\
c

Q)

u
::l
c
L..
Q)

10- 36

a.
>Q)

f
\

c.!l

........

e+e- + Anything
400 GeV

E
U

10- 37

10

Fig. 2. Dielectron production cross section measured by CFS group.

magnetic focusing spectrometer of small solid angle which can be

adj us ted to select, with hi gh purity, di rect s i ngl e muons produced
at _90 in the c.m. system. The other arm consists of an array of
10 large scintillation detectors shielded from the copper target
by earth and steel. The shielding thickness and configuration is
such that a muon must have a transverse momentum, (Pi) ~ 3.2 GeV
in order to register in the detector. The large arm covers in the
c.m. system a polar angle of 60 < 8 < 126, and an azimuthal angle
of _8 < ~ < 25. The angle ~ is defined with respect to the plane
containing the beam and the magnetic spectrometer.

489

MASSIVE DILEPTON PRODUCTION

do
dmdy

Iy =0

p + Be + ~+~- + Anything

400Ge.V

!
I

c
0

Q)

10- 36

::l

'-

Q)

0.

::-

Q)

......

(',j

10- 37

II

10

Fig. 3. Dimuon production cross section measured by CFS group.

Data is taken as a function of pI, the transverse momentum

selected for the muon observed in the magnetic spectrometer. The
rate of events in which a second coincident muon is detected in the
large arm is recorded. In the circumstance that the dimuon system
has very small trans verse momentum, then the p1 of the other muon
which is detected in the large arm must balance the PI of the
selected muon and the dimuon mass is approximately m - 2 PI. If
the parent of the dimuon is produced with significant transverse
momentum then the mass acceptance of the apparatus is broadened
but is still centered about 2 PI.

490

J. W. CRONIN
MUlTIHOlE SPECTROMETER
----------(J)-(D-(D-~_<D_-(J)_-~--(J)---~---_<D_

PROTON BEAM

~ARGET

_ -_ _ _.-.J}m
100m

Fig. 4.

Schematic view of dimuon apparatus of CP experiment.

Knowledge of the mean transverse momentum of the dimuon parent

is necessary to relate the coincidence rate at a given pl to the
average mass being observed and the cross section. In thlS
experiment the mean transverse momentum can be inferred from the
azimuthal angular distribution of the muon detected in the large
arm with respect to the plane of the magnetic spectrometer and the
incident beam. For m ~ 9 GeV where an adequate number of events
was available the mean transverse momentum of the dimuon parent was
found to be - 1.25 GeV/c. This is about the same value observed by
the CFS experiment. In order to evaluate the mean masses and the
cross sections observed in the CP experiment a mean transverse
momentum equal to 1.25 GeV/c independent of m was assumed. For a
given PI' the mass acceptance of the spectrometer was about 2 GeV/c 2
FWHM. Thus the CP experiment cannot resolve narrow resonances
whose total yield is less than or equal to the continuum yield in
a 2 GeV/c 2 wide mass band. For what follows we assume we are dealing
with a continuum.
Figure 5 shows the results quoted as da/dmdyly=o. Like the
CFS experiment, the CP apparatus was only sensitive to dimuons
produced in the neighborhood of y = 0 (at rest) in the c.m. system.
Also the cross sections are quoted per nucleon assuming that the A
dependence is linear. Measurements of dimuon production at lower
m give an A dependence which is consistent with a linear A
dependence. IS
In the CP experiment measurements were also made for 300 GeV
incident protons. These results will be discussed at a later point

MASSIVE DILEPTON PRODUCTION

491

o
II

~1637
>.

-0

E
~ 1038
b

C\J

-0

10 II

12

m (GeV/c 2 )
Fig. 5.

Dimuon production cross section measured by CP group.

in the lecture when we will discuss the evidence relevant to scaling.

Another observation of the CP experiment bears directly on the
question of the source of direct sigg1e lepton production. 16 For
single direct muons selected with pI ~ 4.5 GeV/c, the fraction that
is accompanied by a second muonsin the large solid angle arm is
significant. For example, for PI = 5.25 GeV/c, (132)% of the
single muons are accompanied by a second muon. Using only up-down
symmetry with respect to the production plane and a conservative
extrapolation of the observed distribution beyond the solid angle
subtended, one can estimate in a model independent way that ~ 30%
of the single direct muons have their origin as one member of a
high mass muon pair.
In fact if one parametrizes the dimuon production cross section
as Edo/dmd 3p ~ e- m e- 1. 6Pl(l-x F)4 one predicts a ratio of double

492

J. W. CRONIN

muons to single direct muons of - 10%. Thus it is possible to

account for all of the single direct leptons at large Pl
(Pl ~ 5 GeV/CY-by the production of massive dilepton palrs. In
the conventional Drell-Yan model calculations it is not possible
to account for the observed yield of single direct leptons. The
reason for this result is that the Drell-Yan model in its simplest
form does not provide any transverse momentum for the dilepton
system. If a substantial transverse momentum (-1 GeV/c) is
arbitrarily added to the model, then the single lepton production
can be explained.. This fact was pointed out by M. Duong-Van 17
before the current observations were made. It is important to
stress, however, that such a large transverse momentum does not fit
into the Drell-Yan model in any natural way.

..--...
N

:;;

1035

Q)

(9

"N

1036

"-

~j;<akvasa.ParaShar

""f'" ~

and Tuan

u
..........

"'1~

II

_~
>.

1037

-0

-0

"b

-0

1038

~t,

(fJ-fJ-)} CFS
(e e)
o (fJ-fJ-)
CP

4 5 6 7 8 9 10 II 12

MfLfL
Fig. 6. Plot of all dilepton cross sections measured in the CP
and CFS experiments. For comparison two Drell-Yan calculations
for the cross section are also plotted.

MASSIVE DILEPTON PRODUCTION

493

In Fig. 6 we have plotted the results of both experiments on

the same graph. As one can see the agreement is accep.table given
the statistical quality of the data. In the range 5 ~ m ~ 11 the
cross section da/dmdylyco can be fit by an expression _e-.72m.
Also plotted are the predictions of the cross section predicted by
the Drell-Yan process as calculated by two different authors.
The two models differ principally in the parametrization of
the anti quark distributions. The model of Pakvasa, para,~~r, and
Tuan 10 parametrizes the antiguark distribution as (l-x)
while
the parametrization of Farrar 9 is (l-x)/. Here x is the fraction
of the nucleon momentum carried by the quark. Since the latter
model is less rich in anti quarks at large x, the predicted cross
section for massive dimuon production is smaller and falls more
rapidly with increasing mass than the former model. Both models
assume that the quarks possess a color variable.
The data fall in between the two calculations which indicate
that the Drell-Yan process can be easily adjusted to fit the
observations. It is interesting to compare the dimuon cross section
predicted by a two photon process. This two photon process has been
calculated in detail by Chen, Muzinich, Terazawa and Cheng 18 and
must be observable if no stronger process intervenes.

Fig. 7. Diagram of dilepton production by the two photon process.

494

J. W. CRONIN

Figure 7 shows the diagram of the two photon process which can
be related to measured elastic and inelastic electron-nucleon
scattering. The kinematics of the process are such that the
dilepton system produced by the two photons has a large rapidity
in the nucleon-nucleon c.m. system. The two experiments described
here are insensitive to a dilepton system produced with large
rapidity. Nevertheless it is interesting to compare the total
cross section dcr/dm measured in the experiments with the predictions
of the two photon process.
To calculate the total dimuon cross section at a given mass
we assume that the dependence of the cross section on xF is (1-XF)4
(xF is the Feynman x of the dilepton system.) This dependence was
observed for J/W production 19 and is assumed somewhat arbitrarily
to hold for all dilepton production. In Fig. 8 we plot the total

..--..

C\J

Q)

(9
..........
C\J

tt ft2
t

u
-..-

it f

21

tt

-0

..........

-0

10 II

m (GeV/c 2 )
Fig. 8. Plot of total dilepton production cross section per unit
mass. Expected cross section for two photon process is also plotted.

MASSIVE DILEPTON PRODUCTION

495

cross section so calculated and compare with the predicted dilepton

yield for the two photon process. What is clearly demonstrated in
Fig. 8 is that the yield of dimuons is at least 20 times larger
than the yield predicted on the basis of the two photon process,
and hence must come from some other process which is much stronger.
The most attractive interpretation is the Drell-Yan process since
the observed rate is consistent with the predictions of that process.
There is, however, the one major reservation to the above
conclusion and that is the fact that the average transverse momentum
of the parent of the dimuon is 1.2 - 1.5 GeV/c. This is a large
amount since one conventionally believes that the mean transverse
momentum of the quarks should be comparable to the typical average
tranverse momentum in hadronic processes, 300 MeV/c. We regard
this observation of large transverse momentum as an important fact
which must be understood before one can accept the Drell-Yan process
as being demonstrated.
Regardless of the source of the dimuon continuum, one can apply
the CVC argument to obtain lower limits on the production of the
charged vector bosons (W) as a function of their mass. Since the
recent gauge theories which unite the weak and electromagnetic
interactions predict the mass of the W and heavy neutral bosons to
be in the range from 50 to 100 GeV/c 2 , it is necessary to establish
the energy dependence of the dimuon production. If the dimuon
production process is indeed Drell-Yan, then the predicted scaling
behavior gives a precise definition of the energy dependence. Thus
it is of great interest to investigate the scaling behavior of the
dimuon process. The most comprehensive and recent review of scaling
has been given by Palmer, Paschos, Samios, and Wang in the 1975
Isabelle Summer Study.2o
At present the data on the scaling question are rather sparse.
In order to give as complete a view as possible we will include
continuum muon data at lower masses. These data come from the
Columbia-Hawaii-Illinois-Fermilab group (CHIF)21 and a second
Chicago-Princeton Group22,23 (CP-II) using the Chicago Cyclotron
Facility at the Fermilab.
The former group has measured continuum dimuon produc~ion by
-320 GeV neutrons in the mass region from 1.2 to 2.6 GeV/c , while
the latter has measured the production by 150 and 225 GeV protons
of dimuons in the mass region from 1.4 to 2.6 GeV/c 2. All the
measurements were made on nuclei. The cross sections per nucleon
were extracted using the measured A dependence in the case of the
CHIF group. The crost sections/nucleon for the CP-II group were
deduced assuming an A dependence. Finally we have also used the
original data of the Columbia group at BNL for 29 GeV protons
incident on uranium.2~

J. W. CRONIN

496

The Chicago-Princeton experiment has measured two points at

m = 7.6 and m = 8.6 GeV/c 2 for 300 GeV incident protons. This is
to be compared with the measurements at m = 7.7, 8.8, 10.0 and 11.2
GeV previously discussed. These measurements are the only measurements of the energy dependence of massive dimuon production with
the same apparatus. These measurements unfortunately lack the
statistical precision required to make a good check of scaling.
With all the qualifications discussed above, we present the
results in Fig. 9. Here we plot m3dcr/dm vs m2/s. If scaling holds
all these points should lie on a universal curve. The solid diamond
points from the BNL experiment are assigned no errors because of the
many uncertain operations performed on the original data. One sees
that despite all the corrections the 29 GeV data lies about a factor
3 above the mean of the points measured for 300 and 400 GeV incident
protons.
The most reliable comparison of two energies are the open and

1030
(\J

-31

N' 10

>
I

225
150
" N320
29
300
c 400
+ 400
x 400
0

Q)

(9
(\J

1032

E 1033

""0
'-..

fJ~
f
)k
f~

""0
r0

0,04

Lab

Energy Group

I~

10 34

CP (II)
CP (II)
CHIF
Columbia
CP
CP
CFS (f-L)
CFS (e)

G}"-+
I
1 Ij
1-'--r~
1 ~

0,08

0,12

0,16

0.20

m 2 /s
Fig. 9. Plot of m3dcr/dm as a function of m2/s for available
dilepton production data.

497

MASSIVE DILEPTON PRODUCTION

1031

........
E 1(532

.
J

>

Q)

(.9

1(533

---E

"'0
.........

"'0

rt)

1036

----

//,

~/J/
I I
I

,.
~,bj
' I
I

10-35

.... -

//
,t /

1034

,,j,,'"
I

10

103

51m 2
Fig. 10. Comparison of scaling curve obtained from recent data
with scaling curve of 1975 Isabelle Summer Study (Ref. 20).

closed squares at 300 and 400 GeV incident protons. In one case
there is good agreement with scaling, while in the other there is
a three standard deviation discrepancy. Clearly the evidence for
or against scaling is not established. Nevertheless, if we assume
that scaling will eventually be established, then the universal
curve should lie close to the solid curve drawn by eye on Fig. 9.
We will use this curve to make some predictions for rates of dilepton
production and W production as a function of its assumed mass.

498

J. W. CRONIN

In Fig. 10 we plot the solid curve of Fig. 9 (now shown as a

dot-dash curve) on a similar curve obtained from the Isabelle study.
The solid curve is the result of a Drell-Yan calculation by the
Isabelle study group which neglects the color variable for the
quarks. The dashed curve is the result of the Isabelle study when
divided by three to account for color. The solid circles are the
29 GeV data from BNL, while the dashed circles are the BNL data
assuming an Al dependence. One sees that the scaling curve
obtained from the data is ~ualitatively different from that of the
Isabelle study. For low m /s the data is higher, while for high
mZ/s the data is lower. As one can see below,
studies of
dilepton production at the ISR will be most valuable in verifying
the scaling curve at low m2/s.

>

Q)

1000

<..9
..........

(/)
~

:::J

0
0

100

10

..........
(/)

+-

c
Q)
>

10

15

m (GeV/c 2 )
Fig. 11. Expected dilepton yields at ISR assuming ~he scaling
curve of Fig. 9. Dashed lines are lines of constant m Is. Solid
lines correspond to the indicated energy in GeV/c.

MASSIVE DILEPTON PRODUCTION

499

We now assume that the solid curve of Fig. 9 is indeed a

scaling curve and calculate the dilepton production rates that can
be expected for the CERN ISR. We assume a luminosity of
2.S x 10 31 sec- l cm- 2 which corresponds to an interaction rate of
106/sec . We neglect the small difference that may arise because
the data here is derived from nuclei with roughly equal mixtures
of neutrons and protons extrapolated to A = 1, with a case of pure
proton-proton collisions. We assume an ISR detector that has a
product of detection efficiency times solid angle which is 2S% of
4~.
In Fig. 11 we plot the expected yields for 100 hours of
operation at each of the standard ISR operating energies as a
function of the dimuon mass. We also show lines of constant m2/s.
One can see that the ISR should be a good device to investigate
scaling for m2/s ~ O.OS. For a dilepton mass of 6 GeV/c 2 the yield
is predicted to rise by two orders of magnitude between the lowest
and highest ISR energies.
We can now discuss the CVC relation between the dilepton
production and the production cross section for charged intermediate
vector bosons. If one uses an isoscalar target with an incident
proton beam it can be shown' that
Gcos 2S m 3
+
- > 3
w do ( )
Ow + Ow - 2/2
2
dm mw
a.
isovector
where G = 1.16 x 10-S GeV-2 is the Fermi coupling constant, mw is
the mass of the intermediate boson, cosS is the Cabbibo angle,
ow+ + ow- is the sum of the cross section for the production of W
bosons of both charges, a. is the fine structure constant and
do { )1
is the continuum isovector dilepton production
dm lmw isovector
cross section evaluated where the effective mass of the dilepton
system is equal to the vector boson mass. If we assume that the
isoscalar part is small, then the above relation becomes
numeri cally:

where all masses are given in GeV/c 2 .

We can use the smooth curve of Fig. 9 to estimate the cross
sections for Wproduction as a function of mass. The expression
above gives lower limits on the production of Wbosons on an
isoscalar target. The smooth curve of Fig. 9 is the result of
measurements of incident protons on nearly isoscalar targets and
is reduced to a cross section per nucleon. We use these cross
sections, the scaling assumption, and the inequality above to find
the lower limits for Wboson production for collisions at 200, 400

500

J. W. CRONIN

tt
1036~~~--~~--~~~

20 40 60 80 100 120
mW (GeV/c 2 )

Fig. 12. Production cross sections for Wbosons as a function

of mass for various total c.m. energies.

and 1000 GeV c.m. energy. While these are lower limits for incident
protons on an isoscalar target, they should be very close to the
values expected for proton proton collisions which would be obtained
with colliding beams. These results are plotted in Fig. 12. The
striking feature of these curves is the increase with c.m. energy
in cross section for large Wmass. For a 100 GeV/c 2 mass, the
increase between 200 GeV and 1000 GeV is greater than three orders
of magnitude. In Fig. 13 we plot the estimated r~tes for W
detection assuming a luminosity of 2.5 x 10 31 cm- sec- l (10 6
interactions/sec) and a detector with a solid angle efficiency
product of 0.25 of 4TI. The branching ratio for W + ~ + v
was assumed to be 0.10. These curves give one some idea of what
can be expected from various colliding beam devices that one might
contemplate.

MASSIVE DILEPTON PRODUCTION

501

(j)
L

:::J

o
o

"

(j)

+-

Q)

>

20 40 60 80 100120
mW

(GeV /C 2 )

Fig. 13. Yield per hour of W + ~ + v events detected in a

colliding beam experiment. See text for details.
Some cautionary remarks should be made concerning the W cross
section estimates used above. The predictions for 1000 GeV c.m.
energy come from very low values of m2/s. The data come in fact
from measurements at Fermilab where the actual masses are - 2 GeV/c 2
which may not be in the scaling region. Also the evidence for
scaling in m2/s is at present not very strong. The estimates depend
totally in the assumption of scaling. Finally all the cross sections
per nucleon are deduced assuming an atomic number dependence Al
which has not been verified for dilepton masses larger than 3 GeV/c 2 .
I hope this review has succeeded in demonstrating that a great
deal of progress has been made in the study of dilepton production
in proton-hadron collisions. These studies aim to verify that the
hadron structure, as deduced from the study of a single hadron with

502

J. W. CRONIN

a point lepton probe, is sufficient to understand a more complicated

process involving the collision of two hadrons. The important fact,
that the mean transverse momentum of the dilepton parent is
-1.5 GeV/c,is not understood in terms of the present picture of
hadron structure. These studies also serve the practical purpose
of defining the cross section for production of a W boson, given
knowledge of its mass.
Over the next few years we can expect further progress. Th2
question of how well the dilepton production process scales in m /s
should be answered. Also one will begin to measure dilepton
production by incident pions at larger m2/s. This process should
provide a richer source of anti quarks than one has in proton-nucleon
collisions. One can in fact anticipate the penultimate experiment
in which colliding beam experiments can be made between protons and
anti protons. 2.5
The author would like to acknowledge the hospitality of the
Aspen Center for Physics where this report was written.

REFERENCES

1.
2.
3.

4.
5.
6.

7.
8.

9.
10.
11.
12.

At the pres 2nt time the existence of a possible resonance at

m = 6 GeV/c remains unconfirmed. See O. C. Hom et ~., Phys.
Rev. Lett. 36,1236 (1976).
J. H. Christenson et al., Phys. Rev. Lett. 25,1523 (1970);
ibid, Phys. Rev. 0 ~,:2016 (1973).
-J. J. Aubert et ~., Phys. Rev. Lett. 33, 1404 (1974).
J. E. Augustin et ~., Phys. Rev. Lett. 33,1406,1453 (1974).
L. M. Lederman, Columbia University preprint (1974).
Y. Yamaguchi, Nuovo Cimento 43,193 (1966).
L. M. Lederman and B. G. Pope, Phys. Rev. Lett. l, 765 (1971).
S. o. Orell and T.-M. Yan, Phys. Rev. Lett. ~, 316 (1970),
and Ann. Phys. 66, 578 (1971).
G. Farrar, Nucl. Phys. B ZZ, 429 (1974).
S. Pakvasa, O. Parashar, and S. F. Tuan, Phys. Rev. 0 lQ, 2124
(1974).
H. P. Paar and E. A. Paschos, Phys. Rev. 0 lQ, 1502 (1974).
J. Finjord and F. Ravndal, Phys. Lett. 62 B, 438 (1976).

MASSIVE DILEPTON PRODUCTION

13.
14.
15.
16.

17.
18.
19.
20.
21.
22.
23.
24.

25.

503

The author is most grateful to Prof. Lederman for permission

to incorporate these results in this review.
L. Kluberg et ~., Phys. Rev. Lett. to be published.
M. Binkley et al., "Dimuon Production on Nuclear Targets,"
Fermilab preprTnt (1976).
Recent reviews of this subject have been given by J. W. Cronin,
Lectures of the School of Subnuclear Physics (Erice, 1975),
and L. Lederman, Proceedings of the International Symposium
on Lepton and Photon Interactions at High Energies (SLAC, 1975).
M. Duong-Van, SLAC preprint (SLAC-PUB-1604, 1975).
M.-S. Chen et ~., Phys. Rev. Z D, 3485 (1973).
See for example: H. D. Snyder et al., Phys. Rev. Lett. 36,1415
(1976).
-R. B. Palmer et ~., B.N.L. preprint (BNL-20634, 1975).
M. Binkley et ~., Phys. Rev. Lett. E, 574 (1976).
K. J. Anderson et al., "Inclusive ]..I-Pair Production at 150 GeV
by 7[+ Mesons andProtons," (submitted to XVIII Conference on
High Energy Physics, Tbilisi, USSR, 1976).
K. J. Anderson et al., "Production of Continuum Muon Pairs at
225 GeV by Pi onsand Protons," (s ubmitted to XVII I Conference
on High Energy Physics, Tbilisi, USSR, 1976).
We use these data in a rather indirect fashion. We take the
points from Ref. 20. The data have been corrected for the
fact that the laboratory momentum of the dimuon was required
to be ~ 12 GeV/c. In Ref. 2~/~he cross sections per nucleon
were extracted assumi ng an A
dependence. l l~e have corrected
these cross sections assuming instead an A dependence. This
is equivalent to dividing the BNL data obtained from Ref. 20
by a factor 6.2.
C. Rubbia, P. McIntyre, and D. Cline, Harvard University
preprint (March, 1976).

504

J. W.CRONIN

DIS C U S S ION
CHAIRMAN:

Scientific Secretaries:

Prof. J.W. Cronin

Y. Afek and G.J. Tarnopolsky

DISCUSSION
YFSILANTIS:
Except for your two points at 300 and 400 GeV/c, there seems
to be no experiment at the same values of U2 /s to check scaling.
Would you comment on this.

CRONIN:
In addition to our points, some measurements overlap at the
lowest ~12/s values; but the curve is rather steep there. For W
boson production you want to be in the range of M2/s of 0.1 to 0.2.

vlIGNER:
Is your relation M~lJ da/dUlJlJ = f(M0lJ/s) compatible with the
existence of singularities at all thresholds or is it not to be
considered such an exact relation?

CRONIN:
The scaling relationship would not hold if new thresholds open
in H~lJ'

ETIM:
If the lJ-pairs came from the decay of a series of vector mesons,
perhaps infinite in number, could not the large PI events be easily
understood?

505

MASSIVE DILEPTON PRODUCTION

CRONIN:
The experiment of CFS has good resolution and extends to
If the ~ pairs are a series of vector mesons then their
decay widths and spacing would have to be such that the result is an
apparent continuum. Above 7 GeV/c 2 the CP experiment has poor resolution and could not detect a narrow resonance. Hence there is certainly the possibility that the dimuon continuum is a series of
vector mesons.
~ 7 GeV/c 2

CLEYMANS:
Will you measure this process with pion beams?

CRONIN:
We have an approved experiment at the Fermilab in which we plan
to study dimuon production by incident pions. We expect to begin in
about two years.

KLEINERT:
In the Chicago-Princeton experiment, what is the background due
to secondary muons from weak decays of ~'s and K's? Or of charmed
D mesons?

CRONIN:
~ and K meson decays are experimentally removed.
We observe
highly correlated pairs. It is kinematically unlikely that the ~'s
come from different particles. We will check this correlation by
triggering on electrons in the precision spectrometer and looking
for muons in the M.ll.S.

ZICHICHI:
In the Chicago-Princeton experiment, what are the solid angles of
the spectrometers? The acceptance really matters for uncorrelated
pairs.

CRONIN:
The solid angles are
~1AG

2 x 10

-3

sr,

~ffiS

(i~)

sr .

In addition to the solid angle constraint we also demand at

least 3 GeV/c on each side. This requires a rather stringent correlation.

506

J. W. CRONIN

YPSILANTIS:
Assuming perfect energy and angular resolution on the ~'s detected in the M.R.S., what would be the ~~ mass resolution du only
to the mUltiple scattering in the magnetic spectrometer?

CRONIN:
It becomes clearly much better and perhaps comes to 5-7%.

LIPKIN:
When a muon pair has a high transverse momentum, there must be
something recoiling against the pair with opposite transverse momentum.
Could this be the result of a higher mass particle decaying into a
pion, or something else, and a vector meson with high PT and the
vector vector then decaying into ~+~-?

CRONIN:
It is possible.

It would be a very heavy particle since the

~+~- pair consumes around 40% of the centre-of-mass energy, at the

highest mass measured.

ARE JETS REALLY THERE?

Endre Lillethun
Department of Physics, University of Bergen
N-5014 Bergen-Univ., Norway
INTRODUCTION
In introducing the concept of jets I shall go back to some
results from high energy proton-proton collisions obtained at the
CERN Intersecting Storage Rings (ISR) in 1972 - 1973 1,2
The data are shown in Fig. 1. The initial work measured the
differential production cross section for pions only in the region
of transverse momentum, PT' up to about 1 GeV/c. In this interval a strai~ht line fit to the logarithm of the invariant croSS
section Ed a/d 3 p as function of PT appeared very good, indicating the functional relationship
=

This function was indeed expected from the ideas of Feynman scaling 3
which also predicted that the constants A and B should be independent of the centre of mass (c.m.) energy, Is, of the protonproton system at very high energies.
A simple picture of the collision process is shown in Fig. 2.a.
On colliding, the two protons (Lorent~-c0ntracted) break into two
fragments each. The two parts which really hit each other remain
almost at rest in the c.m. system, forming a lump of very excited
energy, boiling off the energy as particles of relatively low momentum. The other two parts continue almost in their original
direction with rather high momenta, "decaying" into a few particles
with low transverse momentum.
507

508

E. LI LLETHUN

J
10 0

10- 1

"'i.I.
I.
I

~D

Xl'

0.
LJLJ

I'
I'
I .

10- 2

'"

10- 3

"

10- 4

I
I

10- 5
I

1.0

2.0

3.0

4.0

TRANSVERSE MOMENTUM,Pr(GEV/c)
+

Fig. 1. The mean invariant cross section for production of TI and

TI
in proton-proton collisions with centre of mass energy 52.8 GeV
plotted as function of the transverse momentum of the n.
The
straight line is a fit to early data (see Ref. 1).

When the measurements at the ISR were continued to higher

PT 4-6 , it turned out that the cross section did not follow the
slmple exponential PT-dependence. The measured values were very
much higher, as can be seen from Fig. 1.
The explanation for this behaviour may be that there are hard
grains (quarks., ~ partons) wi thin the protons and when the grain from
the two protons meet they enter in a hard collision, in a sense like
the semi-elastic scattering of protons in nucleus-nucleus collisions.
The idea is pictured in Fig. 2b. The grains that undergo the hard
collision may leave the interaction region with high transverse momenta, carrying along with them some part of, or the entire central
lump of energy, "decaying" as they flyaway. It is the "decay" products of such hard collisions, indicated by the double arrows at
the bottom of Fig. 2b, which we call jets.
It seems to me that the hard scattering also could occur

ARE JETS REALLY THERE?

509

BEFORE COLLISION

BEFORE COLLISION

JUST AFTER COLLISION

JUST AFTER COLLISION

0--0

~rtJ

-rs~
SOME TI'- AFTER COLLISION

--\1/-

-------1--...... a

SOME TI'- AFTER COLLISION

--\1/---;1--"""b

Fig. 2a. Sketch indicating a proton-proton collision resulting In

particle production following Feynman scaling.
Fig. 2b. Sketch indicating a proton-proton collision including
hard scattering of constituents of the protons.

between two regions of very high energy density (due to statistical fluctuations) in the kinetic part of the energy associated
with the protons. Therefore I shall in this talk frequently use
the word "flucton" to describe any part of the energy, associated
with the protons, that takes part ih the hard collision leading to
jets, including quarks, partons, gluons and constituents. Although
it is not necessary for an energy density fluctuation to have fixed
quantum numbers, the jets resulting from the final state interactions ("decays") must have specific quantum numbers, but their
invariant mass distribution may be continuous.
The inclusion of hard scattering within the protons in high
energy collisions has led to the description of many models and
parametrizations. See Ref. 7 for a review of Large Transverse Momentum Processes as of June 1975 and Ref. 8 for parametrization of
data on inclusive production in terms of such models.
The properties of jets have been discussed in several papers 9- 13
beginning with Bjorken and Brodsky's9 article in 1970 about hadron

E. LI LLETHUN

510
BEFORE COLLI SION

UNOER COLLISION

PARTON
PARTON

-----d-

SOME TIME AFTER COLLI SION

Fig. 3. Sketch indicating an electron-positron collision resulting

in the production of hadrons.

production in electron-positron collisions. In the following I

shall describe as an experimentalist the different main features
of jets, leaving you to study the more mathematical approaches in
the above references.
The simplest picture (theoretical and experimental) of jets may
actually come from the production of hadrons in electron-positron
(e+e-) scattering, since when these leptons collide they annihilate,
leaving a very highly condensed amount of energy (a virtual photon)
which in some cases will appear as a single parton-antiparton pair
as shown in Fig. 3. The pair of partons may then fly in any direction, back to back in their c.m. system, "decaying" very quickly
into two jets of hadrons with rather low momentum transverse to the
direction of the parent partons.
+ It is reasonable to assume that these jets, produced in e e
collisions are similar to the jets produced in proton-proton collisions. Actually the structure of the jets in their c.m. system may
look like the structure of a proton-proton inelastic event without
high transverse momentum particles. See bottom part of Fig. 2a.
The e+e- events frequently have a single set of jets, while the
proton-proton, pion-proton or any hadron-hadron event with jets
will include also the particles from the forward-backward flying
fragments, thereby confusing the structure: which particle belongs
to the jets, which belongs to the forward-backward fragments?

511

ARE JETS REALLY THERE?

An extra complication in proton-proton events comes from the

(assumed) fact that the protons consist of at least two types of
quarks and a number of quark-anti quark pairs of different kinds in
addition to other types of fluctons. The jets can therefore be the
result of hard scatterings of several types of pairs of fluctons
and/or antifluctons, while in the case of e+e- scattering the colliding fluctons must be expected 'to be pairs of flucton-antiflucton
of the same kind (in quark. language (uti), (dd) or (S8) )
JETS IN

+ e e

COLLISIONS

At SPEAR, the e+e- colliding beam facility at Stanford Linear

Accelerator Centre, a study has been made of the hadronic jets produced in e+e- collisions at several c.m. emergies. 14 ,lS
The property measured is called sphericity, s , and is defined as
3
2

(~ PT~) min
2
L. -+
p.
1

where

A1

are the eigenvalues of the tensor

0.40

0.39
0.38

>- 0.37

I-

~ 0.36

'"rnif

0.35

:i

0.34

'"~

'"~
rn

0.33
0.32

~ 0.31
0.30
0.29
0.28
0

,
,,
,,

,,

, , -'

~
\

02.0 3.0 4.0 5.0 6.0 7.0 8.0

E c.m, (GeV)

Fig. 4. Observed mean sphericity versus centre of mass energy for

events of electron-positron collisions producing hadrons. The
solid (dashed) curve shows a mean sphericity obtained from a Monte
Carlo computation based on a jet model (phase space model). See
Ref. 15.

512

E. LI LLETHUN

as)

p. p.
l

The Ai are the sums of the squares of the transverse momenta with
respect to the corresponding ei genvector directions. A3 lS the
minimum eigenvalue and the eigenvector corresponding to A3 is defined as the jet aXlS.
Events with high multiplicity and a uniform distribution are
characterized by S ~ I , for events where all particles follow
the jet-axis the value of S = O. The value o f S therefore is
a measure of the jet structure of the events.
The group has found evidence for jets as we s hall see In the
following figures. However, because of the non-uniformity of the
acceptance of the particles produced, the measured results had to
be compared to Monte Carlo simulations of the same distributions,
based on either an isotropic phase s pac e model or a jet model with
a mean transverse momentum of 315 MeV/c with respect to the jet
axis. Note that the computations assume production of p i ons onl y
(neutral and charged) and that all charged tracks found experimentally are assumed to be pions. (We shall come back to th e importance of really identifying the part i cles later).

Fig. 5. Observed sphe r ic ity distributions for events of electronpositron collisions producing hadrons. The curves were obtained
from computations a s explained for Fig. 4 and were normalized to
the number of events in the data. The data were recorded a t centre
of mass energies 3. 0 ,
6.2 and 7.4 GeV for particles of
x = 2p/E
< 0.4
(p i s the particle momentum).
C.m.

ARE JETS REALLY THERE?

513

Fig. 4 shows the mean spericity, <S>, versus c.m. energy.

The measured values of <S> vary from 0.35 at Is = 3.0 GeV to
0.29 at Is = 7.4 GeV. The energy dependence is in good agreement with the jet model but disagrees with the phase space model.
The distribution of events in the sphericity space is shown in
Fig. 5 for c.m. energies 3.0, 6.2 and 7.4 GeV. At the lowest
energy both model calculations agree very well with the measured
values. For the higher energies the discrepancy between experimental data and the phase space model increases with energy while
there is always good agreement between data and the jet model distributions. From observations it therefore seems clear that hadrons in e+e- collisions are produced in jets. However, there
could be other effects that might simulate jets, e.g. statistical
fluctuations, kinematical correlations and production of resonances.
These effects have been studied theoretically by Grassberger and
De Groot 16 who introduce a somewhat different definition of
sphericity, or rather asphericity, which eliminates the effect of
statistical fluctuations. From the above mentioned effects they
estimate contributions to the asphericity of less than 20 %while
they expect the value of about 40 %at the highest SPEAR energies
if the hadrons are produced in jets with transverse momentum distributions similar to that of hadronic reactions.
A definitive measurement and analysis following this method
has not yet been carried out, so, although jet-like production of
hadrons in e+e- collisions seems very likely, it is not quite
proved to be so.

700
>

600

vi

500

...0

400

"on"
>-

w
~

0::

w 300

<D

::!:

~ 200
100
~

~--~~--~--~~~~~=-4~
0

0.50

1.00

1.50

2.00

MA55.GeV

Fig. 6. Invariant mass distribution for neutral two-particle combinations of hadrons produced in electron-positron collisions (all
hadrons interpreted as pions).

E. LI LLETHUN

514

It is natural to ask whether these jets are something new or

just known resonances. The SPEAR-group has studied the invariant
mass distribution for two-particle combinations in different categories. One of the distributions is shown in Fig. 6 for events
with at least three tracks, c.m. energy 7.4 GeV and for neutral
combinations of particles, all assUmed to be pions. The masses of
some resonances are indicated in the diagram, and it is clear that
no large production of resonances is found. However, the nonidentification of particles, if there is a fair fraction of particles heavier than the pions, washes out the possible resonance structures both because a too large number of combinations are formed
and because possible resonances of pion and a heavy particle spread
over relatively wide mass ranges due to the misidentification.
JETS IN PROTON-PROTON COLLISIONS
In dealing with the jet concept in proton-proton collisions it
has so far not been found practical to use sphericity as the observable. Instead one has concentrated on the study of other correlation
effects which will be discussed in the following. Several experiments 17-23 have been performed in order to study correlation effects
in high energy proton-proton collisions. I believe that the clearest
information about jets will come from experiments involving particles
in the large-angle region and with the identification of as many
particles as possible. Therefore I shall in this talk mainly use
PROTON

~I
~I

/
PROTON

Fig. 7. Sketch of an experimental set-up used at the proton-proton

intersecting storage rings (ISR) for the study of events with high
transverse momentum. SFM and BM indicate bending magnets, single
straight lines show scintillator hodoscopes,
shaded rectangles
show multi wire proportional chambersvand the open rectangles spark
chambers. The round structures are Cerenkov counters.

ARE JETS REALLY THERE?

515

some new preliminary data from the British-Scandinavian-Orsay Group

(BSO)22 and results from a CERN group (Darriulat et al.)23 which
have been published and already reviewed several times.
The sets of data have both been obtained at t~e ISR, using a
general experiment device, the Split Field Magnet (SFM), and spesific trigger systems. The set-up for the BSO group is shown in
Fig. 7. The SFM-system contains a large number of multiwire proportional chambers to record the tracks of "all" the particles from
an event. The BSO trigger system consists of a magnetic spectrometer placed at 90 0 to the bisector of the 15 0 angle between
the proton beams and almost in the horizontal plane, capable of
identifying charged particles of momenta up to 5 GeV/c. The CERN
group used an array of lead glass blocks, also placed at about 90 0
to the bisector and centered on the horizontal plane, but on the
side opposite to the direction of motion of the c.m. system of the
colliding protons.
The coordinate system used in the presentation of the data is
described in Appendix A and indicated in Fig. 8, where the protons
come In from left and right along the y-axis.
In the present description I have neglected small differences
between the systems used by the two groups in order to give a coherent, simple picture of the data. The simplifications can cause only
very minor changes to the distribution, well within present accuraCles.
In Appendix B a description is given of the difficulties in
presenting the data because of the non-uniform acceptances. In
particular a ratio R = (413/min.bias) is defined, normalizing event
rates with a high PT trigger to those with t\. socalled minimum bias
trigger.

Fig. 8. The coordinate system used to describe the events. The

protons are coming in along the y-axis, and the positive x-aXlS
is defined by the momentum of the trigger particle, p .
See
.
.
w
Appendlx A for further explanatlons.

E. LI LLETHUN

516

Pw

Pw

"~l
.v-

(0)

Pw

Fig. 9. Sketch indicating production of jets of particles in

proton-proton collisions, the protons entering from top and bottom
for each figure.
PW' PJ and Ps represent respectively the
momenta of the trigger particle, the jet and the secondary particles making up the jet. Figure a may be substituted for the left
side of the Figures band c.

General Jet-Features
We shall assume that a jet is characterized by a set of particles within a cone of relatively small opening angle (~45), and
that in general two jets are produced through a hard collision between two fluctons, the fluctons flying apart back to back in their
own c .m. system.
Fig. 9 shows jet configurations where the total momentum of
the jet (or the initial flucton) is drawn as a double lined arrow
PJ. The momenta, Ps of the final jet particles (the secondary
particles) are drawn as single line arrows. Pw represents the
trigger particle, usually of high transverse momentum, PT.
The
jet on the trigger side may consist of the trigger particle only
or a jet o~ the type shown in ~ig. 9a. As indica~ed, PJ ~ Pw .
The full plcture may look as Flg. 9b or 9c, or Wl th Pw
In those
drawings replaced by Fig. 9a. The jet opposite to the trigger
may of course also consist of only one particle. The angles between
the particles are chosen smaller than expected on the average 24 ,
In order to bring out the features more clearly.
It should be noted that the polar angle of the jet opposite
Pw depends on the fraction of the proton momenta carried by
~he initial fluctons.
The azimuthal ~ngle betwee~ Pw and PJ(o )
lS expected to depend only on a relatlvely small lnternal
pp
motion of the fluctons in the protons (similar to Fermi motion in
the nuclei) so that the two jet-axes 3hould be approximately coplanar. The experimental study of jets is complicated by the fact
that only a fraction of the jet-particles are detected, e.g. only
to

ARE JETS REALLY THERE?

517

r-

413

ALL p

MIN. BIAS

~-tc
I

< 1 GeV Ie
T

-tC

~I

~I

RAPIDITY,Y

Fig. 10. Rapidity distribution of charged particles produced in

proton-proton collisions with centre of mass energy 52.8 GeV, in
events with a charged trigger particle of transverse momentum,
Pw < 1 GeV/c. The vertical scale (for experiment 413) is normalized
to a distribution obtained with a minimum bias trigger. Secondary
charged particles of any transverse momentum are included. Three
distributions are shown for particles in the following ranges of
azimuthal angle with respect to the plane defined by the trigger
particle and the incoming protons:
I cp I < 30 0 (~ ) , 1180 0 - cp I < 30 0 H<) and 75 0 < I cp I < 105 0 (

I ).

the charged ones and not even all of them.

Turning now to experimental data, Fig. 10, displays R, the
ratio of the rate of pw-trigger events to that of minimum bias
events (see Appendix B) as a function of rapidity, y of the particles. Rapidity lS defined as

where E is the total energy of the particle and Py its longitudinal momentum. In this case Pw < 1 GeV/c, no cholce of
trigger particle type has been made, and all charged particle tracks
found in the following regions are included:
Region
I ("same side")
defined by \CPI < 30 0
180 0 . - cpl < 30
Region II ("opposite side") defined by
defined by 75 0 < cp < 105 0
Region III ("up")
Region IV ("down")
defined by 75 0 < (-cp) < 105 0
Regions III and IV are added in the plots.

518

E. LlLLETHUN

p > 3 GeV Ie

-1

413
MIN. BIAS

-J

ALL p

f
I

-------------~---- -11----)---- 1------------------I. I

7j

->f
2

->I

7j

RAPIDITY,y

Fig. 11. Same type of distribution as In Fig. 10, but with

Pw > 3 GeV / c .

Even for these low values of Pw there is a clear trend for the particles to accumulate in the neighbourhood of the trigger particle,
R decreasing from about 1.4 at y = 0 to about 0.1 at y = 4 .
For low values of y the ratio R is largest for the opposite side.

-5

5 Rapidity

TRIGGER PARTlCLE,PW>2 GeV/e

CHARGED PARTICLES,P r > 0.8 GeV Ie

SCALES SHaw RATES DIVIDED BY RATES IN
MEASUREMENT WITH MINIMUM BIAS TRIGGER.

Fig. 12. Same type of distribution as in Fig. 10, but with the
different azimuthal ranges represented "three-dimensionally" by
the respective coordinate planes.
In this case p > 2 GeV/c and
secondary particles are required to have transvers momentum,
PT > o. 8 Ge V/ c .

519

ARE JETS REALLY THERE?

Fig. 11 shows a similar plot of R, but now Pw > 3 GeV/c

The most striking difference between Figures 10 and 11 is the increased number of particles near y = 0 for the opposite side as
would be expected according to the hard scattering model. Also the
multiplicities of regions I, III and IV have increased, but only by
about 30 %. In addition one notes that the multiplicity in the
forward direction y ~ 4 is very low (R~ 0.3) for the same side.
(The data for y ~ 4 are lacking for the opposite side.)
It is interesting to notice the similarity between these distributions for the up, down and same side regions (the trigger particle is not counted on the same side). This property changes
drastically when one requires PT > 0.8 GeV for the particles, as
shown in Fig. 12 where Pw > 2 GeV/c.
The figure is plotted in a
3-dimensional way to bring out the picture of coplanarity. The
values of R around y = 0 are R~3 for the same side, R ~ 1.5
for the up-down regions and R ~ 7 on the opposite side, in good
agreement with a jet-picture.
Note that the mUltiplicity on the same side is lower than on
the opposite side due (at least In part) to the requirement of a
large value of pw. 11 ,24
The distribution on the opposite side lS

mu~h

broader than that

OPPOSITE SIDE
1q>1> 150

P >0.8 GeVlc
T
p >2.0 GeVlc

tj

zl>'

~1

4 11-

II

u<l
u

i t
1
I

II

tfI flI,

ft*,
4
2
3
RAPIDITY,y

t
5

Fig. 13. Distribution of two-particle combinations as function of

the rapidity difference 6Y between the two particles( *). The
two particles are on the opposite side to the trigger particle
(1180 0 - ~I < 30 0 ) and are each required to have PT > 0.8 GeV/c
Pw > 2 GeV/c.
For comparison is shown the distribution of
Fig. 12 of single particles with 1180 0 -~ 1 < 30 0 (+) .

520

E. LI LLETHUN

on the same side. This, as stated above, is expected because of the

motion of the c.m. system of the fluctons. A selection of events
with two particles on the opposite side, each with PT > 0.8 GeV/c,
was made in order to look for the widths of the jets themselves.
In Fig. 13 we plot the rates of such events as function of the difference between the rapidities of these two particles, and for comparison the R as function of y as also shown in Fig. 12. (The
two rates are not to the same scale.)
The 6y distribution is clearly much narrower, with a width
approximatelY like the one for the same side in Fig. 12. Although
the 6y measurement by no means is a correct measure of the 6y
distribution of the whole jet, the result at least seems consistent
with the expectations outlined above for the jet model.
The last figure concerned with these general features of jets
is due to the CERN group. Using a nO trigger with Pw > 2 GeV/c
they select the particle on the opposite side balancing the largest
fraction of the Pw'
(xE) , with a PT > 1.2 GeV/c and lying in
a particular rapidity interval. The rapidity distribution of the
remaining particles (without any momentum cut) with respect to the
chosen particle is shown in Fig. 14. As the rapidity interval for
the chosen particle moves away from y = 0 of the proton-proton
c.m. system, an enhancement in the distribution follows it, again
dN Largest

a)

xE:

-2

.j--.-.---

b)

-3

-2

-1

d)

06

335 events

--------.

02

Largest

x,

t-~------I

-3

f~ '.

dY

253 events

02

o
Y
91:! I

-1

Largest

08

343 events

06

,.-._-t-----3

dN

av

c)

dY

0
y

-2

-1

0
Y

91:!

dY

06
--------

Low P,
>--~

- - - ~-

02

~_1~~-~~~--1_-L-o~~+-~~-L~
Y

Fig. 14. Rapidity distributions of charged particles,of any transverse momentum in the hemisphere opposite to a nO trigger particle
with Pw > 2 GeV/c.
In Figures a, band c events have been chosen
with a particle of PT > 1.2 GeV/c (not included in the plot) In
the rapidity range indicated by double arrows
In Fig. d
similarly a particle of PT < 0.6 GeV/c has been chosen. The dashed
lines show the rapidity distribution observed in minimum bias events
and the solid curve is the rapidity distribution observed in average
events with Pw > 2 GeV/c .

---.

ARE JETS REALLY THERE?

521

consistent with the picture of a jet of particles following the

high PT particle.
Note that an enhancement of the particle multiplicity with
respect to the minimum bias values also remains at y = 0, l.e.
opposi te to the nO trigger particle, independent of the y-region
of the selected high PT particle on the opposite side.
So far we have mostly looked at the very general features of
the jets, mixing the results from recordings with triggers of all
charged particle types. In looking for a dependence on the particle
type we show in Fig. 15 the ratio of the multiplicities associated
with a K or p trigger to that with a n trigger, with
Pw > 2 GeV/c. The ratio is given as function of rapidity for particles on the same side and on the opposite side. The ratio is
consistent with lover the entire rapidity interval available,
maybe showing a slight trend toward higher mUltiplicities at large
y for a K-trigger with associated particles of PT > 0.4 GeV/c .
In the jet interval, y~ 1 , there is no difference within the
present statistics, but it shall be interesting to get more accurate results and separate data for the different charges of the
trigger particles.

OPP.SIDE

K/TT
PITT

,_~_.,IjJ
't' 1_.> ..

GoVI

-.f_ -t- -f-",- - Pr c0.4

GeVlc

1.

2.

O.

RAPI DITY

1.

)2
2.

Fig. 15. The multiplicity associated with K (+) or p (t)

triggers divided by those associated with n as function of
rapidi ty, y.
Trigger momentum Pw > 2 GeV / c.
In the left
figure 1<P 1 < 30 0 in the right one 1180 0 - <p 1 < 30 0 Three sets
of data are shown for each side, viz. for associated particles with
iransvel'se momentum PT < 0.4 GeV/c,
0.4 < PT < 0.8 GeV/c and
PI> 0.8 ?eV/c a~ indicated. The points shown at y = 2.25 include
a 1 partlcles Wl th Y > 2 .

522

E. LI LLETHUN
MASS SPECTRA

Pw > 2GeVIc

..

:>-

~ 10

'"
iii
iii
::E

10

0.5

Fig. 16 . Invariant mass spectra of two-particle combinations of

trigger particles n,
K or p of Pw > 2 GeV/c with other
particles in the event with Px > 0.8 GeV/c and all assumed to be
pions. Both neutral and doubly charged combinations (shaded areas)
are shown .

INVARIANT MASS DISTRIBUTI ONS IN HIGH

EVENTS

If there are jets produced in hard collisions between constituents of the protons, the really interesting measurements would be
th e determination of the masses and quantum numbers of these constituents. This is possible only if we can identify and measure
the energy or momentum of all the particles (charged and neutral)
in the events and in addition have a clear separation o f the jets
from the other partic les produced in the co llision. We are far
from this goal, and have at pres ent only identification of a trigger particle together with momentum measurements of particles over
a fairly large solid angle. (The BSO group has recorded information which in their fut ure analysis wi ll enable t he identification
of a few additi onal charged particles per event.)
While waiting for bett er data the BSO group has produced invariant mass (M) plots of two -particle combinations of an identified trigger part icle wi th another part icle assumed to be a n
Fig. 16 shows histograms of the number of s uch combinations per
~M = 50 MeV as function of M when
Pw > 2 GeV/c, the particles
app ear on the same side with Px > 0.8 GeV/c and their momentum
is measured with an accuracy better than 20 %

ARE JETS REALLY THERE?

523

Combinations (nn)o and (nn) are plotted in the same diagram, showing that in the high mass region they are almost equal.
In the low mass region the (nn)o distribution shows a strong,
broad peak around the mass of the p. Similarly the (Kn)O and
(Kn) distributions show a marked difference only in the region
of the K* (890) and the (pn) distributions show a significant
difference only in the region of the N* (1500-1700 MeV) where
again the frequency of neutral combinations exceeds that of the
charged ones.
The neutral combinations stand out more clearly in Fig. 17
which shows the difference between the number of neutral and the
charged combinations.
The finer details of these invariant mass distributions can
only be studied with more statistics together with a clear identification also of the particles associated with the trigger. Measurements of inclusive spectra show a decreasing fractional component
of pions among the particles produced as PT lncreases (Fig. 18).
The component of heavier particles increases from about 25 % to

TTTT

20
10
0
KTT

~ 20
~

0
0

10

f/)

al
~

u
20
10
0

O.

1.

MASS

2.

3.

GeV

Fig. 17. The invariant mass spectrum obtained as the difference

between the spectra of the neutral and the doubly charged combinations shown in Fig. 16.

524

E. LlLLETHUN

Fig. 18. The fraction of the charged particle production cross

section in proton-proton collisions at a = 90 0 going into pions,
kaons and protons (summed over both charge types) as a function of
transverse momentum. The curves indicate a smoothed average of
data for proton-proton centre of mass energies between 23 GeV and
63 GeV. Note that the curve for protons plus antiprotons is
shifted two units upwards.

about 40 %in the range 0.8 < PT < 1.6 GeV/c so the contamination
of heavy particles in the combinations of Fig. 16 may be expected
to be large.
I also want to present some of the invariant mass plots from
the CERN group measurements. Fig. 19 shows to the left the number
of particles on the same side as the trigger 1T O , per unit rapidity
for 4 different regions of the momentum component Px of the particle (p~o = Pw > 2 GeV/c). There is a clear enhancement of the
distributlon around the 1T O (y ~ 0) which sharpens as the required
Px is increased. For values of Iyl > 1 the distribution is
roughly equal to that obtained with a minimum bias trigger as indicated by the dashed lines (compare also with Figures 10 and 11.)

To the right in Fig. 19 are shown the invariant mass plots for
the two-particle combination 1T o together with any other particle
in the given px-range, assuming the latter to be a pion. The
distributions are compared to background curves calculated for uncorrelated minimum bias particle distributions and show deviations
from the latter only in the low mass region. The deviation increases
with increasing Px and is more or less centered on the mass of
the p , but with a width too large for a p

525

ARE JETS REALLY THERE?

0.08

004

++
+
+
-,--------,-,

160

80

-2

BO~
>

>

'tl

'tl

002

001

-.-+- ------+-.-.
v

Fig. 19. Rapidity distributions (left) for particles associated

with a nO of momentum Pw > 2 GeV/c and invariant mass distributions (right) of the two-particle combination of the high transverse momentum nO with another particle in the event assumed to
be a n+ or n
The particles accepted were wi thin I cp I < 27 0
and the data have been grouped in the following intervals of horizontal transve,rse momentum component, Px:
a)
b)
c)
d)

0.4 < Px <

0.6 < Px <
0.8 < Px <
1.1 < Px <

0.6
0.8
1.1
1.7

GeV/c
GeV/c
GeV/c
GeV/c

The mass distributions include particles with Iyl < 2 only. The
dashed lines represent the minimum bias rapidity distributions and
the dashed curves the mass distributions expected for uncorrelated
particles.

526

E. LI LLETHUN

In order to bring out the mass peak features more clearly the
CERN group has added up all the data for 0.7 < Px < 1.7 GeV/c
which show a mass distribution as seen in Fig. 20.
The dashed
curve again represents a computed distribution expected for uncorrelated particles.
The mass distribution shows a clear peak centered at the p
mass, with a width consistent with that of the p and the accuracy of the measurements. From the p peak and the assumed background rate shown as a dot-dash line, the group estimates the p
production cross section and finds a value
1
2

a +
p

a+ P

0.9

0.2

+
for the ratio between the average of p
and p
production and
the direct nO production (nO from p-decay removed).

However, the distribution in Fig. 20 shows an indication of

a peak also around M ~ 400 MeV. This could come from K* (890)
production, which we already have seen is not unimportant (Fig. 17)
because if the K in the (Kn) decay of K* is wrongly labelled
n, the invariant mass of the (Kn) system from K* will appear
close to M = 400 MeV for the particle momentum intervals used.
If this interpretation is correct the cross section for p may
have been underestimated.
70
~

60
50

'5 40

E
~

30
20
10

,-

04

-0.8

12

m(Ttn!), GeV

1.6

Fig. 20. The invariant mass distribution for two-particle combinations of nO and n+ or n- as explained for Fig. 19, but with
0.7 < Px < 1.7 GeV/c. The dashed curve is the mass distribution
expected for uncorrelated particles and the dash-dot-line represents the background used when estimating the p production cross
section.

527

ARE JETS REALLY THERE?

~-+2'TT

90

90

"PK=2GeVlc
\

"

'"

Fig. 21. The opening angle, ex, for a p decaying into two pions
(left) or a K* (890) decaying into a pion and a kaon (right), as
function of pion momentum for fixed values of the momentum of the
other particle.

Having now found that a very large part of the particle combinations are due to p we again turn to the left part of Fig. 19.
The distribution was found to become more and more peaked as Px
was increased, with a width of the peak with 1.1 < Px < 1.7 GeV/c
corresponding to an opening angle of 25.23 This is very reasonable if the pions come from the decay of p as is shown in Fig. 21,
left, where three curves are drawn showing the opening angle between
the pions for different pion momenta. It is seen that if
PTII = 2 GeV/c and Pn = 1.5 GeV/c , the opening angle is 25 0
Similarly the opening an~le for the Kn system of the K* is shown
In Fig. 21, right. Here the angle is found to be even smaller.
The trend towards more peaked distributions of the associated
particles when their momenta are required to be higher, could therefore just reflect the production of low mass resonances with high
transverse momentum in the proton-proton system.
The same jet-like trend as described above for particles on the
same side as the trigger particle, holds also for particles on the
opposite side. Fig. 22 shows the number of particle combinations on
the opposite side where each particle has PT > 0.8 GeV/c, per unit
rapidity difference between the two particles. The trigger particle
is required to have Pw > 2 GeV /c. For comparison is drawn in the
distributions obtained when a particle from one high PT event is
combined with particles from different high PT events. It is clear
that the proper combinations show a distribution more peaked around
l1y = 0 than does the "background" distribution. Within the poor
statistics there is .also here an indication of a larger number of
neutral than doubly charged combinations in this region.

528
50
40

E. LlLLETHUN

OPP. SIDE

30
20
10
0
(!+)

~I-S' 30
20

If!

10
0
30

o.

f t71Lf-n
(!!)

1.

2.

RAPIDITY DIFFERENCE

Fig. 22. Distribution of two-particle combinations of particles on

the opposite side to a trigger particle with Pw > 2 GeV/c, as function of the difference between the rapidities of the two particles.
Both particles are required to have PT > 0.8 GeV/c. The histograms
represent combinations obtained with one particle taken from one
event, the other from another high pw-event. No distinction between
different types of particles. Data are shown for neutral and doubly
charged combinations and the sum of these.

In order to get a measure of the mean transverse momentum, q ,

of a secondary particle in a jet with respect to the jet axis, the
BSO group has chosen to look at two-particle combinations once more.
The momenta of the trigger particle and the other particle are added
vectorially and the components of momentum of each particle perpendicular to this direction is called q (see upper right corner of
Fig. 23). Pw is chosen in three different intervals as indicated
in Fig. 23 and the other particle on the same side is required to
have PT > 1 GeV.
The distribution of number of particle combinations as function of q shows a double peaked structure. One peak
stays near q = 0 GeV/c and shows a tendency of shrinking as Pw
is increased. Due to experimental effects the first bin in the
diagram is nearly empty so a quanti tati ve value of a mean q cannot
be given but a value around q = 0.4 GeV/c for Pw > 3 GeV/c is a

ARE JETS REALLY THERE?

529

I<Pw <2 GeV/c

~PT~1
....... :

".

-II

O.OOI----+----t------L+--+__+_--;

dou ble charged

pairs only

Fig. 23. Distribution of two-particle combinations as function of

the particle momentum (q) transverse to the vector sum of the
momenta of the two particles. The trigger particle is chosen with
Pw in the intervals indicated. The other particle is required to
have PT > 1 GeV/c and II < 90 0

reasonable estimate.
In the figure for Pw > 3 GeV/c the shaded areas show the
distribution for doubly charged combinations. The peaking is by
no means as clear as for the rest, and the peaking may therefore
still b e due mainly to neutral resonances.
The second peak in Fig. 23, moving towards higher q-values
as Pw is increased, reflects only the fact that there are highmomentum particles in the forward direction which, when combined
with the trigger particle, whose momentum spectrum is very steeply
falling (see Fig. 1), produce the q-values observed.

530

E, LlLLETHUN

CONCLUSION
We have used a simple definition of jets as the result of
a hard collision between two fluctons (quarks, partons, constituents, gluons or statistical fluctuation giving regions of very
high energy density) of the colliding particles. This leads
to a picture of coplanar jets of particles (one or more) coming
out back to back in the centre' of mass system of these constituents. The jets studied here have been defined by having at least
one particle of high transverse momentum. If jets frequently
break u~ into many,particles no high PT pa:ticle,may be present,
and a dlfferent trlgger system for the experlment lS needed.
The experiments c}early bring out jet-like events both in
electron-positron and in proton-proton collisions as measured by
sphericity in electron-positron production and by the jet opening
angle and coplanarity in the proton-proton production.
It is not clear whether the jets represent new physics or lS
just another way of stating that resonances (p, K',j:, /::', N*,
etc.) are produced with high p
and that in such productions
the high transverse momentum must be balanced essentially locally
In the collision.
In order to obtain a clear picture of what happens in high
PT events it is necessary to continue experimentation with large
solid angle for recording identified high PT particles (low
cross section) and for identifying as many as possible also of
the other particles (both charged and neutral). Only in this way
can we hope to separate out jets and study the quantum numbers of
the jet-parents available in the process of two particles colliding
with high kinetic energi.
REFERENCES
1.

B. Alper et aI, Proceedings of the 4th Int. Conf. on High

Energy Collisions, Oxford, 2 (1972) 197 and Phys.Letters
47B (1973) 75.

2.

M. Banner et al, Phys.Letters 41B (1972) 547.

3.

R.P. Feynman, Phys.Rev.Letters 23 (1969) 1415.

4.

B. Alper et aI, Phys.Letters 44B (1973) 521 and 527 and

Nuclear Physics B87 (1975) 19-.-

5.

M. Banner et al, Phys.Letters 44B (1973) 537.

6.

F.W. Busser et aI, Phys.Letters 46B (1973) 471.

ARE JETS REALLY THERE?

531

7.

D.Sivers, S.J. Brodsky and R. Blankenbecler, Phys.Reports

23C (1976) 1.

8.

B. Alper etal, Nuclear Physics B100 (1975) 237.

9.

J.D. Bjorken and S.J. Brodsky, Phys.Rev. Dl (1970) 1416.

10.

J.D. Bjorken, Proc. of Summer Institute on Particle Physics,

SLAC-191 (1975) 85.

11.

S.D. Ellis, M. Jacob and P.V. Landshoff, CERN preprint

TH.2109 (1975)

12.

P.V. Landshoff, Jets in High PT Reactions.

XI Rencontre de Moriond, March 1~76.

13.

G. Preparata and G. Rossi, CERN preprint TH.2163 (1976).

14.

G. Hanson, Proc. of Summer Institute on Particle Physics,

SLAC-191 (1975) 237.

15.

G. Hanson et al, Phys.Rev.Letters l2 (1975) 1609.

16.

P. Grassberger and E.H. de Groot, Nuclear Physics B102 (1976)

297.

17.

S.R. Amendolia et al, Phys.Letters 48B (1974) 359.

18.

S.R. Amendolia et al,

19.

F.W. Busser et al, Phys.Letters 51B (1974) 306 and 311.

20.

F.W. Busser et al, Nuclear Physics B106 (1976) 1.

21.

K. Eggert et al, Nuclear Physics B98 (1975) 49 and 73.

22.

British-Scandinavian-Orsay Collaboration, data presented by

R. Mller at the 7th Int. Colloqium on Multiparticle
Reactions, Tutzing, June 1976, and private communication.

23.

P. Darriulat et al, Structure of Final States with a High

Transverse Momentum nO in Proton-Proton Collisions, to be
published in Nuclear Physics.

24.

M. Jacob and P.V. Landshoff, CERN preprint TH.2182 (1976).

Talk at the

11 Nuovo Cimento 31A (1976) 17.

532

E. LI LLETHUN

APPENDIX

A.

Coordinate Systems.
The maln coordinate system used in the description of thE reults is shown in Fig. 8. The x-axis is taken to be at 90 0 to the
incoming protons in their centre-of-mass system with positive direction along the trigger particle. The x - y plane is defined by the
trigger particle and the incoming protons and does not usually coincide with the horizontal plane in the laboratory system (The trigger particle of the BSO system is at ~ 40
and that of the CERN
system within 8 0 of the horizontal plane in the laboratory).
The z-axis of this righthanded system points upwards as shown in
Fig. 8.
The polar angle (8) in the system is the angle made with the
y-axis, the azimuthal angle () is measured in the x - z plane,
the positive x-axis at = 0 .
In addition to this system there is a coordinate system
attached to the jets, with an axis along the directions of the
parents of the jets in the c.m. system of these particles. We
have at the moment no possibility to find this system because we
have no means of knowing that we have found all the particles belonging to a jet (e.g. the SFM detector records charged particles
only).
In some cases we approximate this jet system with an axis
along the trigger particle on the one side and that of the particle
with the highest transverse momentum, PT, on the other side.
In
other cases the axis is taken to be along the vector sum of the
momenta of the particles included in the set defined as a jet.
The opening angle in the jet system is called a, often taken to
be the angle between two particles of high PT. The transverse
momentum of a particle with respect to the jet axis is called q.

APPENDIX

B.

Acceptance and Normalization.

The acceptances of the equipment and the normalizations should
be considered before a discussion of the measured jet-relevant data
lS started.
So far no other particle than the trigger particle has been
identified.
In principle one should want a detector with uniform
acceptance over the total solid angle of 4~. The non-uniformity
of acceptance (for charged particles only) is indicated in Fig. 24

ARE JETS REALLY THERE?

533

3
0
2

."

<!>.

&

~
-1

"

-2
-3

-3

Fig. 24. Diagram representing the particle distribution in the

Split Field Magnet detector associated with a trigger particle
in the spectrometer of the British-Scandinavian-Orsay group. The
horizontal and vertical axes represent the horizontal transverse
(px) and the vertical (pz) momentum component respectively (see
Figures 7 and 8). The different shadings represent the number of
particles recorded as indicated in the upper right corner. The
centre of mass energy of the proton-proton system was 52.8 GeV
and only particles with longitudinal momentum, Py < 1 GeV/c
were included in the plot.

which shows the spatial distribution in terms of Px and Pz .

The "four-leaf clover" has large openings between its leaves where
there is either no detector, or the detector is shaded by obs tacles.
The computing of acceptance for such a system requires a very large
amount of Monte Carlo simulations. The data of the BSO group22
(experiment 413) is instead normalized by dividing the rates
observed per event with a high PT trigger by the rates obtained
with a "minimum bias" trigger which requires only that minimum two
tracks are found in the SFM detector. A similar normalization has
also been used by the CERN group.23
From Fig. 23 one can see that the distribution is skew, with
more particles in the posi ti ve than in the negative x -direction.
This is due mainly to the motion of the c.m. system of the two incoming protons, moving with a velocity 0.13 c in the direction
of the trigger particle, coupled with the steeply falling production cro s s section as function of PT (Fig. 1).

E. LlLLETHUN

534

DIS C U S S ION
CHAIRMAN:

Prof. E. Lillethun

Scientific Secretary:

W. Marciano

DISCUSSION
McPHERSON:
Do you have a model which explains the relationship between
the high mass PT trigger particle and the extension to higher y
of the associated pions?

LILLETHUN:
We have no model for this at the moment.

McPHERSON:
Are the "jets" opposite to the trigger particle better collimated on an event-by-event basis than by the sum of all events which
includes an averaging of any angle between the "jets" and
trigger?

-Ii

LILLETHUN:
In ,some events we have observed rather well collimated jets;
but no real analysis of this has been performed. They could, for
example, be pure statistical fluctuations. However, as I showed in
one of the transparencies, the Yl-Y2 distribution for two particles
on the side opposite to the trigger with high PT is much narrower
than the y distribution of the same particles, indicating a yes to
your question.

CLEYMANS:
What is known about the variation of the average multiplicity
of particles in either the trigger side jet or the opposite side
jet?

ARE JETS REALLY THERE?

535

LILLETHUN:
We have not tried to look into that as yet.

CLEYMANS:
Do you have any hope of obtaining information on the mass of
the jets?

LILLETHUN:
In order to obtain the jet mass, we have to identify, get the
masses, of the particles in addition to their momenta. With the
present experimental set-up, this information is available for some
particles and we hope to study the mass of some particle combinations. We do not, however, detect neutral particles. A new experiment at the ISR with particle identification over much larger solid
angles is being planned. Here, jet mass information should become
very much better.

CLEYMANS:
Have you corrected your data so as to take into account the
centre-of-mass motion at ISR?

LILLETHUN:
Yes. There are, however, some uncertainties since at the
present state of analysis, only the trigger particle is identified.
All the others are taken to be pions. The data includes information
on the identitiy of the particles within some solid angle region;
therefore, in the future, better analysis will be possible.

FERBEL:
Are there any new results on the mUltiplicity of particles
associated with any particle as a function of the transverse momentum of that particle? About two years ago, there was evidence for
an increase in multiplicity for triggers with PT > I GeV/c. Has
this been confirmed?

LILLETHUN:
(Added after the discussion session.)
There is new data from the British-Scandinavian-Pisa-Stony Brook
Collaboration at the ISR presented at the Tbilisi Conference. The
total mUltiplicity of the event clearly rises with the transverse
momentum.

E. LlLLETHUN

536

SMITH:
Why do people studying jet structure in pp interactions not use
the sphericity parameter that is so famous in the analysis of jet
structure at SPEAR?

LILLETHUN:
The sphericity concept is less useful for the following reason:
as I indicated under the discussion of Fig. 2b, a proton-proton
collision will include a set of particles in the forward-backward
direction, i.e. a jet-like structure in a fairly well-defined direction. This comes in addition to a possible jet of the type we look
for -- it will both confuse the picture and reduce the sphericity
effect. Perhaps, in the future, we may find a way to obtain these
two axes and use some more complicated sphericity analysis.

BASILE:
Is the "over-all" charged multiplicity of minimum bias events
the same as that of events with jets; high PT events?

LILLETHUN:
No. The jets have a higher mean charge multiplicity than the
minimum bias events. In comparing data with a high PT trigger with
minimum bias trigger data, we shall look into the necessity of
choosing minimum bias events with the same mUltiplicity distribution.

CHARACTERISTICS OF ~-e+Ko EVENTS PRODUCED BY A NEUTRINO BEAM

W. F. Fry
Department of Physics, University of Wisconsin-Madison
1150 Universitv Avenue. Madison. Wisconsin

53706

INTRODUCTION
A search for new phenomena in neutrino interactions has been
made in film from the 15 foot bubble chamber exposed to the neutrino horn beam at FNAL. The chamber was filled with 20 percent
neon. An external muon identifier (EMI) which surrounded a part of
the chamber was very effective in separating muons from strongly interacing particles. About 80,000 pictures were taken which yielded
approximately one neutrino interaction in every 10 frames.
One goal of the early analysis was motivated by the observation by the Harvard-Pennsylvania-Wisconsin-FNAL collaboratjon of
neutrino induced events with two muons in the final state. I A
search for events with a muon and an electron in the final state
was made in the bubble chamber film and a sample of events has been
isolat d. Early results from the study have been previously reported and more detailed results on a significantly larger sample
will be given in this publication. These results are to be compared to results from similar studies in the BNL3 and Gargamelle 4
chambers.

A surprising characteristic of the events is the large number

of K mesons produced. The confirmation of dilepton events is these
data strengthens the interpretation as production of a new family
of hadrons characterized by a new quantum number.
This talk will report in detail the procedures and techniques
used to select the events; the characteristics of the events; the
associated kaon phenomena, and the systematic errors pertinent to
the results.
537

538

W. F. FRY

EXPERIMENTAL DETAILS
The 300 GeV proton beam was extracted from the accelerator and
directed on a Berylium target one interaction length long. The resultant spray of positive mesons was focussed by two horns. A typical burst of 6 x 10 12 protons produced v neutrinos. This beam produced about one neutrino event per ten f~ames. The electron neutrino flux (ve) was estimated by Monte Carlo calculation to be
about 2% of the neutrinos. This study also found the electron antineutrino (ve) flux to be about one order of magnitude smaller
than Ve. Although the absolute flux is difficult to determine by
calculation, the flux is probably known to about 20%.
For this exposure the chamber was filled with a mixture of neon
and hydrogen (20% neon by volume). The density of this mixture is
0.281 gm/cm3 , the radiation length (AR) is 116. cm., and the characteristic interaction length is about 3 m.
An important feature of this chamber is the addition of apparatus to distinguish muons from charged hadrons. This External Muon
Identifier (EMI) consists of 24 proportional wire chambers each 1
meter square in area located behind and outside the vacuum vessel of
the chamber. Metalic zinc has been placed between the coils such
that about four interaction lengths is provided by the zinc in the
central region. A track is extrapolated from the sensitive volume
of the bubble chamber through the absorber and the magnetic field to
the position of the wire chambers. The track is called a muon if
there is an isolated "hit" in the wire chambers near the position of
the extrapolated track.
A hadron will interact in the absorber and the shower of
secondary particles will record multiple hits in the wire chambers
which have little correlation with the extrapolated parent track.
The misidentification of a true muon as a hadron is due mainly to
chamber inefficiency and is about 1%. The misidentification of a
hadron as a muon is due chiefly to the "punch through" of the energetic shower into the wire chambers and the accidently correspondence of one hit to the parent track. The probability of incorrectly
identifying a muon is thus dependent on the number and energy of the
hadrons produced in the v interaction.
For the study of ~-e+ events to be discussed here the background of punch through is sufficiently low as to be unimportant.
SCANNING
In the study of rare event type phenomena it is imperative that
the scanning and selection process be well understood in order that
the results will be quantitatively meaningful. In the experiment

CHARACTERISTICS OF J.t"e+Ko EVENTS

539

reported here, the procedure involved scanning of a segment of the

film by physicists and subsequent scanning and rescanning by professionalscanners. All candidates for neutrino interactions of all
types were studied by physicists, looking for events where an identifiable positron or electron was associated with the neutrino interaction. An electron is said to be identified if it exhibited one
or more of the electromagnetic phenomena: (1) visible bremmstrahlung
where the change in momentum was clearly discernible by visual inspection of the curvature (abbreviated as V.B.S.), (2) visible
bremmstrahlung accompanied by the conversion of the y ray into an
electron pair or compton electron (abbreviated as V.B.S.+e.p.),
(3) non-visible bremmstrahlung plus electron pair where the electron
pair did not point to the vertex but tangentially to the electron
track or (4) the superposition of an electron pair on the track in
question (triplet) which comes from the bremmstrahlung and conversion of the y ray in a short distance so that the electron pair is
superimposed on the incident electron.
Because there is a finite probability that the electron pair,
for some of the above processes could come directly from the main
vertex, it was necessary to apply an origin-electron pair distrance
criteria such that this background would be less than 10 percent of
the signal.
ELECTRON DETECTION EFFICIENCY
The detection efficiency for electrons and its energy dependence
was determined from a study of electron pairs. In this study an
electron pair must satisfy the following criteria: (1) zero opening angle and (2) point to a vertex in a chamber. The following
quantities were recorded: (1) the type and number of electromagnetic processes which lead to its identification as an electron (if it
was identified), (2) its position in the chamber and (3) its momentum. The chamber was divided into five zones along the length of
the chamber. The detection efficiency was determined for each zone.
To set the average detection efficiency each zone was weighted by
its volume and the average taken. The results are tabulated in
Tabl e I.
Although a quantitative study of the detection efficiency
above 5 GeV/c has not been made, the fact that electron neutrino
events have been found which have electrons up to about 100 GeV/c
and about at the correct rate is considered to show that this detection efficiency cannot be much lower than at 5 GeV/c.
All events which had an electron of either sign were taken as
electron candidates and treated separately from all of the other
neutrino interactions. Each of these events were then examined
under higher maqnification to see if there was any evidence for a

w.

540

F. FRY

VO or a chanqed K meson. Since the most strikinq characteristics

of the dilepton events is the number of VO decays, the question
arises as to whether the higher percentage of V's comes from a more
detailed study under high magnification of the electron events. It
was found that an additional VO was found by this more detailed
study of the vertex in only about 5-10 percent of the cases and
hence their effect is small.
CHARACTERISTICS OF e+w- EVENTS
Some of the more obvious characteristics of the events are
given in the following tables of those events where a e+ is clearly
identified and a w- is clearly identified by the EMI or where there
is a w- candidate.
Table 1
Electron Detection Efficiency
Detection Efficiency
100
300
800
3000

<
<
<
<

P
P
P
P

<
<
<
<

300
800
3000
5000

MeV/c
MeV/c
MeV/c
MeV/c

0.29 0.08
0.47 0.01
0.52
0.41 ~:~~}O .48 0.07

RAT~ OF w-e+/CC
.
The oroductlon
rate of w- e+ events for the cuts E + > .8 GeV
and Evis > 5 GeV can be estimated by correcting for e the electron
detectlon efficiency which has been given previously. The contribution of neutral currents in the denominator has been subtracted if
the partial cross section is to be compared to the charged current
cross section. To obtain the true production rate, however, two
more corrections have to be made.

Loss of Events Due to Simulated Dalitz Pairs

True w-e+ events will be lost in our selection process if the
positron, combined with one of the negative non-interacting hadrons
in the reaction, gives a mass m(e+L-) < m(nO). This loss was determined empirically by studying events with Dalitz pairs, where both
e were identified. For these events, the positron simulated a
Dalitz pair with a leaving negative track in approximately 20% of
the times.

541

CHARA CTERIS TICS OF p-e+Ko EVENTS

Howe
Ident.

IO#

Table II
Topology of the Events
11

Ident.

KO
S

KO
L

11

1NVB+PR
lVB+PR

EMI

lNVB+PR
2VB+PR

Geom miss + of EMI

2(VB+PR)
ANNIH+PR EMI

1DEL TA
lVB
lTRID

Geom miss
+L+(2.8)
in EMI

1(VB+PR)
1NVB+C
o(-2)at
VTX

EMI

+-,00 ld(l 11, 11+?)

1NVB+C
1NVB+PR
1NVB+C?

EMI

+-

1NVB+CPR
lVB+CPR

1VB+PR

SHWR+NO
EMI

1VB+PR

Geom miss
EMI

+-

1 +)
ld(l1?,l

l+d

A(+-)
A(+)
1I+A
1I-TKO? 1 d
11

EMI

+-

11

lVB+PR

EMI

00

12

1NVB+PR
lVB+PR

EMI

or
A

13

2(VB+PR)

EMI

14

1NVB+PR
1VB+PR

EMI

NVB
VB
TRIO
PR
CPR
C
n
y

LIS

+-

1NVB+PR
1VB
KINK or
VB

10

31+
2L+,6y
'-CH l-(K)

11,ln

l+d
or
A(+-)

Non-visible brems.
Visible bremi'
e pair on e track
Electron-positron pair
Close electron-positron pair
Compton electron
Nlutron
e- pair from prim. VTX

L
I
S
d
CH

Leaving track
Interacting track
Stopping track
Decay
Charge exchange

542

W. F. FRY

Table III
Lepton and Stranqe Particle Kinematics
10#

Evis

33

14

12

Momenta
+
O
O
e
KS
KL

Visible
K

Q2

W2

18

.004

.56

35

1.6 6.1

.13

3.4

1.1

3.1

1.6 8.1

.10

.71

15

26

21.8

2.2

1.6

6.2

3.1

1.

.12

.4

28

9.3

5.2

5.8

7.2

18

.21

.66 28

98

58

2.1

+-00 2 1
4.4
.

.21

39

.003

.41

75

24

15

.4

2.2

5.5

7.8

.37

.34

10

15

SS
2.0

5.6

1.211.

.16

12

.007

.86 24

29

(5.4)

.64

1111.

1.0 23

.024

.81

43

( .80)

3.9

16

.49

.96

18

9.4 87

.058

.90 154

1.2 5.7

.11

.29

10

2.7

.49

.10

3.6

.70 42

.009

.70 78

3.8 7.3

.29

.17

20

3.6
4.6

10

97

9.3

5.7

11

21

14.1

12

30

26.2

1.3 00
1.1
1.4 1.6

13

61

SS
17.9

4.0

14

42

33.6

.80 1.5
orA

SS

3.3

38

.6

1.0

orA

"same side" muon

+
.25

18

2.9

10

12
13
14

11

10

7
8

1
2
3
4
5

10#

.70
.81
1.04
.70
.60
.69
.66
.67
2.6(A)
(1.5(K))
1. 3(A)
.85
.75
1.4K1.3
2.1
1.08
1. 38( K)
1.89
.95K
(1. 5A)

M
e+K

.081
.929
.116
.219
.580
.562
.491
.587
.451

.067

-.390 -.843
.079
.085
-.213

-.047

-.052
.054 -.577
.379 -.416
-.397 .288
-.334 .482
.420 -.164

.019
.065
.597
.531
.257

-.006
-.040
-.503
-.499
-.256

-.018
- .051
-.321
- .187
.007

-.552
.277
.180
-.279
-.089
.272
.143
.024
.320
-.490
-.082
-.371
.026
.249
.049
-.002
-.194
.210
- .104

.776
-.924
-.862
-.550
-.205
.556
-.219
-.573
- .221
.592
-.629
- .391
- .241
1.296
-.254
.327
-.925
.068
-.024

.952
.965
.881
.617
.229
.619
.261
.573
.389
.769
.634
.539
.242
1.320
.259
.327
.945
.221
.107
.581
1.341
1.129

.033 - .581
-.237 1.320
1.052
-1. 251

.407
-.019
-.740
.678
.062
-.240

-.487
-.990
-.593
-.814

.812

-.807

-.091

1.252
.886
1.200
.596
.849

1.964
.289
.957
.279
.437

1.897 .512
.170
.234
.108 -.951
.260
.102
-.369 -.234

Table IV
Mass, Momentum Correlations lto ~inc
Positron
. Event .Tota 1
K StraRge
pe
I~issl
p~
IpKI p~lSS p~lSS
Ipel
PA
PA
ft
n
n
II
II
II
1..
.1
ol-

::I:

.co..
Co)

1.11

en

-I

<
m
z

"m

;.

-n

n
en

:::!

en

::D

-I

::D

544

W. F. FRY

Loss of Events at Energies E +

e

<

.8 GeV

Unfortunately, this loss cannot be determined experimentally

since it depends on the mechanism by which the positrons are being
produced. If the positron originates indeed from the decay of a
particle with a mass of several GeV, the momentum spectrum of the
positron will extend to quite small values and the effect of the
E + > .8 GeV cut will be substantial. It has been estimated using
3~body decay of a 2 GeV mass particle, that as much as 50% of the
events will be lost with our cut.
The best estimate for the true rate is then
jl - e+ICC = 1.5%.
ENERGY DEPENDENCE OF THE PRODUCTION RATE
The sample of (e+jl-) events have been divided into two groups,
those where the total visible is below 30 GeV and those above. This
ratio is then to be compared to the same ratio of number of events
in a partial sample of charqed current events where in each event a
jl- was identified by a hit in the EMI. These data are summarized
in the following table.
Table V
e+jl - Events
Number of events Evis > 30 GeV*
Number of events EV < 30 GeV*
Number E > 3D/Number E < 30 GeV

6
9

.66

.18

CC Events
31. 5%
68.5%
0.46 .05

*The average energy for Ev < 30 GeV is 21 GeV while the average
for Evis > 30 GeV is 62 GeV for e+jl- events.
There is no meaningful evidence for an energy dependent in the
cross section for e+jl- events from these relatively poor data.
STRANGE PARTICLES
In order that the ~ multiplicity mo that has been found in
e+jl- events be meaningful it must be compared to that found in "normal" charged current interactions. A sample of charqed current
interactions have been studied looking for associated V's which

CHARACTERISTICS OF J.fe+Ko EVENTS

545

come from decays of AO, KO and KO. Very preliminary results of

this study in a partial s~mp1e i~ summarized in the following table
(VI).
The VO events listed in this tables comes from a sample of
about 970 v interactions which have a visible energy above about 5
GeV. An attempt was made to eliminate neutral currents in this sample (see Table VI).
Since about 85% of the V's in the ~-e+ events are K~ decays it
is interesting to compare the+ratio of various K~ decays to charge
current interactions in the e ~- events with that in the charge currents. This comparison is shown in Table VII.
The total VO/(interaction) in (e+~-) compared to VO/(CC int.)
is a ratio of
12/14 x .05 = 17.
A more detailed analysis of the stranqe K particle production
in (e+~-) now follows.
KO PHENOMENA
Although the available data on strange particle production in
the e+~- events is somewhat meager, it is enough to give some important results. The most quantitative results come from the KO
phenomena while the evidence for charged K mesons is at best only
collaborative. A listing of the events along with the strange particle phenomena is given in Table VIII.
The most definitive mode of identification of the KO is its
two body decay; KO + TI+ + TI-. Typically it is a 3C fit to a vertex.
The mean error inSthe invariant mass is about 25 MeV in this bubble
chamber so that the chance coincidence of a neutron background
event is negligible.
It is important to note that in two cases the VO has been identified as a AO and not a KO. In two other cases there is a KO-Ao
ambiguity.__ One possible source for the AO's is the production of
AO's by a KO meson which was produced in the initial neutrino interaction. A crude estimate of this probability gives a number
which is consistent with the 2 AO events observed. (The two A-K
events which are ambiguous are more likely to be KO than AO from a
statistical consideration of probabilities.)
Perhaps the simplest way to estimate the mean number of neutral K mesons produced per event is to use only the K~TI+TI- mode and
correct for the branching ratio. It will be assumed that the numbers of K~ and Kr are equal and that the branching ratio of
K~(TI+TI-) to K~ total is 2/3, this gives the efficiency for
detecting a K; as 1/3. In Table VIII the total number of events

w.

546

F. FRY

with one KO or other ViS in addition is 10 out of a total of 14 e+~ events. T~is gives for the average multiplicity of KO production
m(KSL )

= (10/14)

x 3

= 2.1

0.7.

The error is just the statistical error in 10 events.

A more sophisticated method to evaluate this same multiplicity
is to use all of the KO phenomena (reactions 1 through 5) and to
weight the value of the mean given by the various reactions by the
statistical weight of each reaction. The reactions considered are:

KO + 'IT + + 'IT S
KO + 'ITo + 'ITo
s
KL Decay + charged modes
KO interactions

KO or A
s
KO interactions in the \! target nucleus before escaping
+

(1)

(2 )
(3)

(4)
(5)

The evaluation of the detection efficiency for each reaction is involved and a breakdown of the probabilities for each reaction is
given in Table IX.
Not only is the mean multiplicity mO important but also the
second moment of the distribution. Clearly the data are insufficient to give a quantitative evaluation of shape of the distribution
of the multiplicity. However, in events 4 and 6 there are three
charged V decays. Even though one of the VOIS in each of the two
events may have come from a charged exchange of a K+ it does show
that at least three or more strange particles were produced.
event 4 there is good evidence for four strange particles, three of
which were neutral KOIS.
The expected numbers of events using the detection efficiencies
given in Table IX for estimating the various types of events are
given in Table X. There is reasonable agreement of the number of
events of the various types with what is expected.
OTHER ELECTRON PHENOMENA
In the scanning, all tracks were examined to see if there was
an electromagnetic process which would identify it as an electron
track; either as an electron or a positron. Five processes lead to
the production of electrons or positrons; namely:
1)

CHARACTERISTICS OF /[e+Ko EVENTS

547

Table VI
Number and Type of V Events in "Normal" Charged Current
v Interactions in 20% Neon
]l
Decay Mode
Number
Number of V's/CC
KsO

19

19/970

.02

11.0

28
2

28/970
2/970

.03
.002

2/970

.002

KO
L
KOA o
s

~~

Total
(lota 1 sample CL = 97)

0
51

5%

Table VII
Comparison of V Event Rates in e+]l- Events with CC Events
VO's in e+]l Decay Mode
VO in CC
in VO Events
Number
N/Event
~('T/'IT-)

6/14 = 42%

2%

11. (P'IT-)

2/14
1/14

14%
7%

3%
0.2%

1/14

7%

KO

L
KO ('IT +'IT -) KO
s
L
*K~ ('IT+'IT- ), Kt:, Kt:
One or more V's/Event

l(Ko int)
L
2

2/14 14%
12/14 = 85%

0-+.:0;.0.1%
0-+.::;.0.1%
5%

*The Kt: decay is taken in both (e+]l-) events and (CC) events even
thouqh it may come from a charge exchange of a K.

548

W. F. FRY

Table VIII
Event

T,n~e

of V

Ks (+-)

Ks (++)

Ks (+-)
Ks (+-)' Ks(OO?), KL, KL
Ks (+-)' KL, KL
AO(p'IT-)
Ks (+-)

4
5
6
7

2)
3)
4)
5)

Event
8
9

10
11

12

TtQe of V
AO
(K L?)
Ks KL(int-rAO)
Ks(OO?)
Ks (+-)

13

14

Ks (+-)

ve interactions and hence an absence of a muon.

Dalitz pairs from 'ITo or n.
High energy knock knock-on electrons along a track very
close to the neutrino vertex, and
Electron pairs where the conversion of the y occurred so
close to the v interaction that the gap could not be observed. If the pair is very asymmetric the lower energy
electron or positron may be unobserved and hence only one
electron of either sign seen.

Background classes 3, 4 and 5 can be reduced to a level below the

signal by a suitable choice of electron energy and asymmetry
parameters as was discussed in the section on backgrounds.

~-e+

The number of electrons and positrons which are consistent with

or identified as Dalitz pairs are estimated from the average number
of pions in normal neutrino interactions. There are about 4-5
Dalitz pair candidates for each ide~tified e+~- events.

CHARACTERISTICS OF J[e+Ko EVENTS

549

Table IX
Evaluation of Detection Efficiency for Reactions
Reaction
K
S

-+ 1T

+ +

1T

Overall Det.
Efficiency 1;
1;

= 0.95

Comments
If the decay occurs within
cm of origin it may be
missed. This percentage is

~1.0

~0.5%.

1; - .38

KL Decay

1; - .03

KL

1; - .12

Interacti on

Difficult to estimate. The

probability of observing 3 out
of 4 y's is about 0.5. The decay must occur at a distance
>5 cm 1; = 0.5 x 0.75 = .38
The mean usuable decay length
in the chamber is about 1.8
meters giving a probability of
decay into charged modes of 3%.
The detectable reactions are
KOL + n -+ KOs + n'
(1 )
KO + n

KO interaction in
target nucleus
giving AO

.07

-+

A + n'

(2 )

Reaction (1) gives 0.6 for probability of KO interacting

times 0.2 for ~ regeneration
making the total of 6%. Reaction (2) gives ~30% for K interacting' in KD mode times 20%
for A production making total
of 6%.
KO + n -+ A + n' in the v target
nucl eus. (Assume that the number of KO's available for this
reaction is one less than <m>.
Probability of interaction is
KO/Ko = 1/2 times probability
of interaction giving AO = 20%
from ko mode. 1; = m>-l/<m
x 1/3 x .2 = .07.

w.

550

F. FRY

Table X
Expected KO Phenomena
Based on m(O)

observed
expected

2 (for 13 ~-e+ events)

Ks (+-)

Ks(OO)

KL
Decays

8
8

1.6

4
0.4

\
i

into
1i q.
1
0.8

KL i nt. in target
nucleus -+ AO
2

1.8

There is a non-negligible contamination of electron neutrinos

in the muon neutrino beam. No quantitative measure of the ratio of
electron to muon neutrinos has been made, but an estimate can be obtained from the ratio of K/IT from the proton horn target. This
gives a ratio of about 4 percent. In the sample of filmed scanned
for e+~- events there are with the number estimated from the ve
flux, the number of ve candidates is consistent. These events which
have all the characteristics to be expected from v interactions;
namely the x and y distributions are reasonable ana they do not have
a muon associated with the interaction. Not only are these events
of interest in the study of electron neutrinos, but they clearly
show that the detection efficiency for high energy electrons does
not become small. The electron energies go to as high as about 90
GeV. In summary, these ve interactions look like the normal v~ interactions expected for the lepton and they both differ in the ways
from the e+~- events in that the normal ve and v~ charged current
interactions have a low VO multiplicity and the electron energy in
the e+~- events peaks at a much lower value.
SUMMARY AND CONCLUSIONS
The most important properties of the ~-e+ events are summarized
in the following paragraphs.
a) The most striking characteristics of the ~-e+ events is
the hi gh ~ multi pl i city. A search for decays of charged k mesons
is clearly less conclusive but the results are consistent with the
charged multiplicity being equal to the neutral multiplicity. The
best e~timate for the average neutral K multiplicity is about 2
KO/~-e events.
This is to be compared with an observed KO multiplicity of 0.06 for the charged current neutrino interactions.
b) The AO production is much lower than the KO production.
It is possible that the observed AO'S come entirely from secondary
KO interactions in the target nucleus. The observed KO/Ao ratio
associated with the ~-e+ events corrected for the probability of observing each, is found to be 8 while in v CC interaction this ratio

CHARACTERISTICS OF J.fe+Ko EVENTS

551

is found to be abot 1.3. Clearly the st~ange production is different for the ~-e events than in the v~ CC events.
c)

The rate of production of e+~-/CC is estimated to be 1.5%.

d) There is no meanin~ful evidence for any energy dependence

in the production of the ~-e events, although the statistics are
poor.
e) The values of x and yare not significantly different for
the ~-e+ events than for CC v~ interactions.
f)
The ratio of the momenta of the two leptons is a significantly large number <P~>/<Pe> = 6.28 suggesting that the two leptons
do not come from the same vertex.
g) There has been observed one event of the type ~+e- which
is consistent with an interaction of an anti-mu-neutrino giving
~+ + e- + (hadrons).
h) The remaining v interactions which have an electron and
not a muon are consistent with interactions of electron neutrinos.
The high KO multiplicity suggests that there could be the
strong decay of a charmed particle into another lower state which
subsequently decay leptonically into a positron, a neutral K meson
and a neutrino. Several such schemes have been proposed.
REFERENCES AND FOOTNOTES
1.

A. Benvenuti, D. Cline, W. T. Ford, R. Imlay, T. Y. Ling, A.

K. Mann, R. Orr, D. D. Reeder, C. Rubbia, R. Stefanski, L.
Sulak and P. Wanderer, Phys. Rev. Letters 35, 1199 (1975), 35,
1249 (1975).
- -

2.

J. von Krogh, W. F. Fry, U. Camerini, D. Cline, R. J. Loveless,

J. Mapp, R. H. March, D. D. Reeder, A. Barbaro-Galtieri, P.
Bosetti, G. Lynch, J. Marriner, F. Solmitz, M. L. Stevenson,
D. Haidt, G. Harigel, H. Wachsmuth, R. Cence, F. Harris, S. I.
Parker, M. Peters, V. Peterson and V. Stenger, Phys. Rev. Letters 36, 710 (1976).

3.

E. Cazzoli et al., Phys. Rev. Letters 34, 1125 (1975).

4.

H. Deden et al., Phys. Letters 58B, 361 (1975).

J. Blietschau et al., Phys. Letters 60B, 207 (1976).

w. F. FRY

552

DIS C U S S ION
CHAIRMAN:

Scientific Secretaries:

Prof. W.F. Fry

J. Shigemitsu and A. Bisell0

DISCUSSION
BALDO CEOLIN:
You said that you can check the hypothesis of two KO,s for each
charmed event by comparing the interactions of the naturally produced
KO in the v and
cases, measuring, for example, AO production,
K~-K~ mass difference, etc. Do you not always get v, V producing
dimuons and dikaons in the two cases with the same couple of KO and
iO, the only difference being the exchange between the naturally
produced and the other?

FRY:

There is a large difference between v~ and v~. For neutrinos,

the D is positively charged and decays into a iO; while the KO produced in the nucleus is KO. In antineutrinos~ the D is negative,
giving a KO, while the K in the nucleus is a KO which has a high
probability of producing AO hyperons. For this reason, antineutrino
production may lead to a high number of AO hyperons.

WIGNER:
Three questions: What is the order of magnitude of the nuclear
cross-section in these energies? What is the chance that a hadron
interacts? and What is the neutrino energy involved in these reactions?
FRY:

The answer to your first question is of the order of 10- 38 cm 2

per nucleon. It is quite likely that a hadron interacts because the
geometrical mean free path is about 2.8 m, while the diameter of the
chamber is a little more than 4 m. The mean neutrino energy is about
30 GeV, but the spectrum goes beyong 150 GeV.

553

CHARACTERISTICS OF J.Ce+Ko EVENTS

MARCIANO:

Does the rate (e + ~-)/(charged currents v) = 1.3-l.5~ include

the ~+~- mode? If not, is this not rather high for the charm picture?
And could you elaborate a bit on how the contamination of the beam
occurs?

FRY:
The rate 1.3-1.5% did not include the ~+~- mode. If one naively
doubles this percentage, it is high and could, under certain assumptions, present somewhat of a problem only if the non-leptonic decay
is high; however, I will show in a moment that the non-leptonic mode
does not contribute a large percentage.
Most of the V 's come from TI~ decays and Ke decays, which strongly
favour v~ and are ~ightly peaked in energy. The main source of electron neutrinos is Ke 3 decay which produce more energetic vets than
the v~ from TI decays.

SMITH:
Why are there so many K's?
only one kaon?

Does the charm picture not predict

FRY:
The extra K can result from the neutrino acting on a strange
quark from the sea. This would give a KO or K+ from production,
along with a KO from the leptonic decay of the charmed particle.
For example

v + N + D+ KON ~
lKOev
For antineutrino beams, there will be a KO or K- produced in the Ne
nucleus. This should result in greater AO production from V beams
than from V beams.

FERBEL:
Would you not expect a lot of charged K production?

FRY:
Yes. This is a harder question to evaluate; we need more data
than exists at the present time in order to quantitative evaluate
the charged K production.

HADRON PHYSICS AT FERMILAB *

T. Ferbel
Department of Physics and Astronomy
University of Rochester, Rochester, New York, 14627 USA
ABSTRACT
I survey recent experimental results from studies of hadron
interactions at Fermilab. Elastic, total and charge-exchange cross
section measurements, diffractive phenomena, and inclusive production, using nuclear as well as hydrogen targets, are discussed
in these lectures.
INTRODUCTION
The momentous J/~ discoveries of almost two years ago [1]
have put an unusually depressing damper on the entire field of
"old-fashioned" hadron physics. Subsequent to the J/~ announcements, every able-bodied experimentalist who had equipment installed on the experimental floor of Fermilab abandoned all his
previous well-planned efforts to join in the mad search for charm,
color or what have you. (Unless your experiment involves a prompt
lepton trigger or has such key words as "search for narrow . " on
the proposal, you are still regarded in some ~ircles of Fermilab
as a myopic pariah.) Despite all the frenetic J/~ related activity,
several groups have nevertheless managed to generate some outstanding results on "hard-core" hadron physics. In these lectures I
will summarize some recent experimental and phenomenological investigations which I find particularly stimulating. I will try,
where possible, to present those results which tend to raise new
questions rather than answer old ones.

*Research supported by the United States Energy Research and

Development Administration.

555

T.FERBEL

I. TOTAL, ELASTIC AND CHARGE-EXCHANGE CROSS SECTIONS

Total Cross Sections
Precision measurements of total and of elastic scattering
cross sections are now available for K, n, n, p and p projectiles.
The latest data for total cross sections have recently been published [2]. I reproduce the results of these measurements on
hydrogen and deuterium in Fig. 1. It is clear that all hadronic
total cross sections appear to fall from values measured at lower
energies; after passing broad minima, which occur near ~10 GeV/c
for K+p and perhaps as high as ~300 GeV/c for pp (no pp measurements are as yet available above 200 GeV/c!), the cross sections
start to rise again. The differences between antiparticle-nucleon
and particle-nucleon cross sections shown in Fig. 2 can be parameterized as a function of the square of the energy in the center
of mass (s) in terms of a power-law dependence of the form Asa-l.
The values of a obtained from a fit to the difference in the pp
and pp cross-section (~p=0.390.02) and for the K-p and K+p
difference (aK = 0.40 - 0.04) are the same within error. From
Regge phenomenBlogy we expect these differences in the particleantiparticle cross sections to isolate the dominant contribution
of the w Regge trajectory (although the differences also contain
p and A2 terms, these mesons do not couple nearly as strongly to
the proton as the w does). Recent measurements of K~ regeneration
on carbon [3] yield a value for the w trajectory at t=O of
a w(t=0)=0.430.01, a result in good agreement with the above given
app and aKp' Similarly, the value of ~=0.550.03, obtained from
the data in Fig. 2, is reasonably consistent with that expected
from the contribution of the p trajectory to the difference between the n-p and the n+p cross section. (A value of a p (0)=0.48
0.04 has recently been extracted from measurements of n-p+~n at
Fermilab [4].)
+

The data in Fig. 2 indicate that ~aT(p-p) is about 2-3 times

larger than ~aT(Kp);alSO, ~aT(pn) is about 4 times larger than
~aT(Kn). These kind of differences in the hadron-hadron cross
sections can be used to test various symmetry relations. In particular, the ratio 3[~a(Kp)+~a(Kn)]/[~a(pp)+~a(pn)] measures
the contribution of w-exchange to kaon-nucleon relative to baryonnucleon total cross sections. The constancy observed for this
ratio from ~15 GeV/c through 200 GeV/c speaks for an energy-independent universal coupling of the w meson to kaons and baryons. I
will not elaborate further on this subject, I will note only in
passing that,besides w universality, p-w universality and the
Johnson-Treiman relations are all in agreement with the data, to
~(10-20)% accuracy, in the 30-250 GeV/c momentum range.
(These
agreements may be short-lived if the trend observed for the

HADRON PHYSICS AT FERMILAB

557

44

., ... f
t"tnqp ....... ~ I ... tt
ti
: !~rl... ,,:.
41

III' t

iVf\t- ......,.

'I .............":' ..... :

t<+p
."... . eo1
'IT'p

I ..

ON

'1
10

20

30

70 100

200 300

IlOO

1000

ZOOO

PuelG&YA:)

88
84
80

t+ t

eo

01>.0" t

".. ,,-- - .

... :

38

30

50 70 100

PLAB (GeVlc)

Figure 1:

Total cross sections for hadron interactions with

hydrogen and deuterium target particles.

T.FERBEL

558

K- p-K +P difference to decrease

more rapidly for momenta above
100 GeV/c persists with further
increase in energy or in data!)

+ GALBRAITIl 01 01 196~
a FOLEY 0' aI 1967

o OENISOV 0' 01 1973

TIllS EIIPAIM[NT

The last item I will dis cuss in this section is a compilation by Whitmore [5] of the
total pp meson-annihilation
cross section as a function of
laboratory momentum. Figure 3
displays the energy dependence
of this cross section in comparison to the difference of
the pp and pp total cross sections. It is interesting that
Ct.- -1

LO

05

02

02

Figure 2: Differences in hadronnucleon and antihadron-nucleon

total cross sections as a
function of incident momentum
in the laboratory.

the s pp
form given in Fig.2
also appears to describe the
meson annihilation cross section. An energy variation of
just this sort was, in fact,
predicted recently by Eylon
and Harari [6] on the basis of
a duality-diagram model. The
implication of the constancy
of ~~/~(pp + mesons) according
to Eylon and Harari, is that pp
annihilations contribute through
the unitarity relation to the
Pomeranchukon rather than to
the meson exchange terms in the
pp total cross section. The
surprising (and perhaps accidental) result of Fig. 3 is that
the meson annihilation cross
section agrees in magnitude with
~~ (Eylon and Harari predict
that as s increases
~~/~(pp + mesons) approaches
a constant ~ 1). To put
these results in some perspective, I remind you that at
~5 GeV/c, for example, the
difference in the pp and pp
elastic cross section is
~5 mb, which means that the
pp inelastic non-annihilation
cross section is smaller than

559

HADRON PHYSICS AT FERMILAB

x .jip ANNIHILATION

CTT(ppl-CTTlppl

triO
,.!..

0.
.0.
~

tr

"
~ 2
2

10

20

lab

50

100

200

(GeVlc)

Figure 3: Comparison of the difference in pp and pp total cross

sections with the cross section for pp annihilation into mesons.
the inelastic pp cross section, and so it is clearly not proper
to. regard the pp interaction as simply the super-position of the
pp interaction and the additional annihilation channel which is
not available to the pp system. Absorption effects in the final
states certainly play an important role in the generation of ~cr;
this is a theme I will periodically return to in the course of
this lecture.
Elastic Scattering
+

Data on the elastic scattering of n-, K- and p on hydrogen

in the 50-200 GeV/c momentum range have become available during
this past year. Figure.4 provides the character of the data obtained at small momentum transfers [7]. The differential cross
sections for the scattering of particles on hydrogen are observed
to have shapes similar to those found for the scattering of their
respective antiparticles on hydrogen. All the differential cross
sections appear to have curvature in t, and all except the pp
channel tend to become steeper with increasing momentum (only the
100 GeV/c data are shown in Fig. 4). The latter result is emphasized in Fig. 5, where the slopes of the elastic spectra
B(t) = d(~n dcr/dt)/dt at t=0.2 GeV 2 are plotted as a function of
s. The difference in the steepness of the K+p and K-p diffraction
peaks, as well as of the pp and pp data, is slowly disappearing
with increasing energy. All meson channels appear to be approaching a common value for B, a value which is substantially smaller
than that describing the pp and pp slopes. At the smallest values

T. FERBEL

560

102
102

~
~

102

(!)

102

E
..........

102

bl"

"1:1"1:1

10 1
IrfJ
10- 1
10-2

0.2

0.4
0.6
It 1 (GeV)2

0.8

Figure 4: Elastic scattering differential cross section for

hadrons incident on protons at 100 GeV/c.
12

10
8
6

v+p
0

10

CD

10
8
20

50

100 200
$ (GeV2)

sao 1000 2000

Figure 5: Slopes of the t-distributions at Itl=0.2 GeV 2 for hadronproton and antihadron-proton elastic scattering as a
function of s (see ref. 7).

HADRON PHYSICS AT FERMILA8

561

of t, negative-particle cross sections tend to be larger than positive-particle cross sections, and vice versa at larger t. This
effect, referred to as the crossover phenomenon, has been examined
extensively at low energies [8].
In terms of a Regge-exchange picture, the observed difference
between antiparticle and particle elastic scattering can be understood as follows. At small values of t the Pomeranchukon is the
dominant exchange trajectory. The Pomeranchukon contribution to
the elastic amplitude is mainly imaginary and non-flip. As in the
case of total cross sections, the difference between antiparticle
and particle elastic scattering can be attributed to the exchange
of meson trajectories with negative charge conjugation. Consequently, the difference in elastic scattering must correspond to
an interference term between the Pomeranchukon and the non-flip
imaginary part of the relevant vector meson trajectories (1m V~A=O).
In the case of Kp and pip the strongly coupled w trajectory
dominates the C=-l exchange and the difference in K-p and K+p, as
well as pp and pp~ elastic scattering is therefore expected to be
large. For the ~-p channels, wherein w cannot be exchanged because of the constraint of G-parity, only p contributes and consequently, because of the weak p-nucleon coupling, the (~-p)-(~+p)
difference in elastic cross sections is expected to be small. The
t-values at the positions of the cross overs are t K=0.190.04 GeV 2
for Kp and t P=0.110.02 GeV 2 for pp data, in theCmomentum range
50-175 GeV/c [9]. There appears to be a weak energy dependence'
in t P (t P was measured to be 0.1620.004 GeV 2 near 5 GeV/c [8]).
The Clatgst data [9] pertaining to the crOSS-Qver phenomenon for
Kp and pip channels are shown in Fig. 6 in terms of the ratios
[(da/dt)-/(da/dt);] and [(da/dt)+/(da/dt);], where (da/dt); are
the fitted (smootfied) data for positively charged projectiles.
The fact that t is so small means that ImV~A=O goes through
zero for t ~ 0.2, whtch implies that the exchanged C=-l contribution must be exceedingly peripheral. Davier and Harari [10] have,
in fact, shown that within the framework of a dual absorption
model the non-Pomeranchukon part of the elastic scattering amplitude is strongly dominated by the largest partial waves. (The
Pomeranchuk contribution is central while the other Regge terms
peak near an impact parameter of ~l fermi.) Consequently, these
results suggest that a simple Regge-pole description of the crossover phenomenon cannot be adequate - strong absorption, cuts, or
the like, are required to parameterize the effect.
For s-values above ~25
cross sections appear to be
accuracy) for antiparticles
0.15/0.12/0.18. Similarly,
to total cross sections are

GeV2 the ratios of elastic to total

constant and same (to within ~5%
and particles [11]: ~p/Kp/pp
the ratios of forward elastic slopes
energy independent [12]:0.34/0.39/0.26.

T . FERBEL

562

1.5

50G.V

1.5

0.5~--~--~--~--~

0.2

0.4

0.6

0.8 0

-t

0.2

0.4

0.6

0.8

(G.V 2 )

Figure 6: Study of the crossover phenomenon in elastic scattering

between 50 GeV/c and 175 GeV/c.

The energy independence of these ratios can be realized in a model

invoking geometrical scaling (GS) of the scattering amplitude in
impact parameter (A(b,s)) [13]. The s-dependence in this model
is contained entirely in the effective radius of the interaction
(the scale), and the amplitude is energy independent when expressed
in terms of b/R(s).[12,13] A pictorial way of viewing the GS
assumption is that the distribution of nuclear matter does not
change with s when it is parameterized in terms of the dimensionless
quantity /rrbL/a lne
. 1as t'lC . An immediate consequence of GS is that
inelastic, elastic and total cross sections are proportional to
R2, as is the slope of the elastic differential cross sections at
small t . Furthermore, because da/dt can also be equated to
R~lfel(R2t) 12, the quantity (l/a:ot)(da/dt) should be s-independent

563

HADRON PHYSICS AT FERMILAB

when plotted as a function of atotot. [14] All the above features

of the GS hypothesis are in remarkable agreement with the data.
Parameterizing a logarithmic growth of R2 with increasing s also
provides the observed rise in ael' a tot and B for the FNAL-ISR
energy regime. Independent of the success of the GS hypothesis,
it is clear that, phenomenologically, elastic scattering at small
t is simplifying with increasing s-values.
A very surprising result in the pp elastic cross section at
large t has been the rapid onset ~t 200 GeV/c of the2Chou~Yang
dip, observed previously at the ISR near t = 1.5 GeV [15].
Figure 7 displays 100 GeV/c and 200 GeV/c pp data from Fermilab
illustrating this unusual effect. The dip at 200 GeV/c occurs at
to=1.5 GeV2, which is to be compared with a value of tD~1.3 GeV2
at 1500 GeV/c (ISR). An energy dependence of this kina for to is,
again, consistent with the simple GS hypothesis.

~\
~\

".
II
I

+
-t-

100

Gov/_

\ + +++++
t

t+i-+t

ff
o

011

LO

L5
2.0
ItIIGeV/c)o_

Figure 7: Elastic scattering cross sections for pp collisions at

100 GeV/c and 200 GeV/c.

Finally, preliminary
measurements of the ratio
of the real to imaginary
part of the foward
scattering amplitude (p),
obtained using nuclearcoulomb interference
at small t, for TIp,
Kp and pp in the 70
GeV/c to 150 GeV/c range
of incident momenta [16]
appear to be in general
agreement with the
latest calculations [17]
using dispersion relations and the new
measurements of total
cross sections [2]. The
values of p are within
~5% of zero near 100
GeV/c momentum, except
for Ppp' which is
~ -0.1, consistent
with previous measurements [18] at Fermilab.
The only sizeable discrepancy between the

T.FERBEL

564

new data and the calculations may be for the K-p channel; however,
the experimenters go out of their way to warn of the preliminary
nature of their results.
Two-Body Charge-Exchange Reactions
Pion charge-exchange scattering has been measured at Fermilab
in the TI- momentum range of 20 GeV/c to 200 GeV/c [4]. The differential cross sections and phenomenological fits to these spectra
are shown in Fig. 8. The data display the well known helicityflip turn-over at t=O and the dip near t=0.5 GeV2; these features
have been attributed to the dominance of the p trajectory in the
production. Although the authors point out that it is likely that
the charge exchange amplitude has small contributions from sources
other than just the p-trajectory (an asymmetry has been observed
in the scattering from polarized targets near 5 GeV/c), they have
nevertheless attemgt~d to fit their data to the simple Regge-pole
form da/dt=S(t)v2a1tJ-2, where v=(s-u)/4M, u is the square of the
four-momentum transfer between the incident TI- and the neutron,
and M is the nucleon mass. The authors have performed a 7 parameter fit to S~t) and a 3 parameter fit to the effective trajectory
a =a +alt+a 2t . The result of their fit to ap(t) is displayed in
F~g.o9. A straight line through points corresponding to p and g
mesons falls remarkably close to the extracted aCt) for Itl<0.3
GeV 2 . The value of a (0) is significantly lower than the value
of 0.580.03 obtainedPat lower energies [19]. The relatively poor
agreement of the present fit with data at lower energies is displayed in the comparison with the 5.9 GeV/c data in Fig. 8.
(There is, in addition, an apparent systematic discrepancy with
the data from Serpukhov.)
The authors have also compared their fits to the charge exchange reaction with the difference in the TI-P and TI+P cross sections. Assuming that the forward charge exchange cross section
continues to fall with increasing s in the same manner as observed
up to 200 GeV/c, the authors determined the ratio of the real to
the imaginary part of the forward amplitude using a dispersion
relation. This, along with the recognition that the imaginary
part of the forward charge exchange amplitude is related to the
difference in the TI-P and TI+P total cross section enabled them to
check the consistency of their data with that of reference (2).
The successful result of this check (and of dispersion relations,
isospin invariance, and the optical theorem) is displayed in
Fig. 9.
Just as the pion charge exchange reaction is believed to be
dominated through the exchange of the p trajectory, the reaction
TI-p+n n is expected to proceed through A'),. exchange. This reaction
has now also been measured at Fermilab [4]. Near 100 GeV/c the n

565

HADRON PHYSICS AT FERMILAB

1000.----~--~--~,

.01

.001

o
o

.2

;4

.6
.8
-t (GeV2)

.02

.04

.06

.oS

-I (GeV2)

. to

.t2

1.0

Figure 8: Differential cross sections for n-p + nOn between

20.8 GeV/c and 199.3 GeV/c. The data at 5.9 GeV/c
are from a previous investigation.

production reaction has a cross section of ~0.8 ~b, compared to

the charge exchange cross section of ~3 ~b. The data and the
authors' fit to the A2 trajectory are shown in Fig. 10. The properties of this reaction appear to be quite similar to those noted
for the charge exchange process. Again, except in just a qualitative way, the fit to a A at high energies is not very consistent
with the results at low 2 energy. (In particular, the extracted
value of a A (0)=0.370.01, is low again.) Because the effective
p and A2
2 trajectories do not have the same slopes or intercepts this obviously means that the are not degenerate for t<O.
And since there is no established J =4+ partner to the A2 , it is
not clear what is the situation for t>O. In any case, the concept of exchange degeneracy which has often been taken as an
article of faith must be considered, at best, a poor approximation
to reality.

In contrast to pion charge exchange, neutron-proton charge

exchange cross sections exhibit characteristic sharp peaks (with
widths of ~.02 GeV2) at t=O. Measurements of differential cross

T.FERBEL

566

.4

(I)

.2

/'

or-------------~~~----------~
/

-.2

/ /
/'

-.4
/

Straight line through

points corresponding
to pond g mesons

-1.4

-.8
-.6
I (GeV2)

2
I

2.0

..s

:0

-.4

-.2

Foley et 01. 1967

Denisov el 01.1971
Carroll el 01. 1976

1.0

<J 0.6
0.4
0.2

10

20

40

60

100

200

Plob (GeV)

Figure 9:

Fit to an effective Regge trajectory aCt) for the reaction n-p + nOn. The bottom graph provides a comparison of the measured difference in n-p and n+p
total cross sections with the difference expected
from an analysis of the charge exchange process.

sections for neutron-proton charge exhange for incident momenta

between 60 GeV!c and 300 GeV/c are shown in Fig. 11. [20] The
authors remark that, although the absolute cross section falls
by a factor of ~105, the shape of their spectra is essentially
the same as observed near 1 GeV/c. The total cross section falls
as P~2b between 2 GeV/c and 12 GeV/c, as p~lb5near 50 GeV/c and
as
a pi1b at FNAL energies. The sharp a peak near t=O is
reminiscenr of reactions which are dominated by pion exchange.
The data at low energies do, in fact, agree with the presence of
a large non-flip contribution from pion exchange. The onset of
a shoulder in the cross section at t~O.l at high energies may be
due to the emergence of contributions from higher lying trajectories

567

HADRON PHYSICS AT FERMILAB

10-' 0!:--'.2~-.4~-.6~-.8~-'1"':.0-L..'''''.2----'

-t (Gev',

.4

a(t)

.2

Straight line through

points corresponding
to p. A2 and g
mesons

O~~----+---------~~----------~

-.2
-.4

-1.2

-1.0

-.8

t
Figure 10:

-.6
-.4
(GeV2)

-.2

Phenomenological fits to the data for rr-p + nn. The

effective Regge trajectory aCt) should correspond
to the A2 meson.

T.FERBEL

568

10

~l lij "II

I: I

I"

~~

:rr1'11I

II

"

i'!t'~1 1'1

60 -90 GeVic

-:

-:
I

Ie f.

~IIIIIII'

10~" -

90-120

' ", , I

lit IIII

I I

ItI (GeVld

1-:

I,

.4

.5

J3

tit

~O =-

r
.3

200-240

"
r

I.O=-

.2

,ij~,

ttl'''''''

~.Hllitl , , ~,,,,.240-300

I'

.1

-:10:-

"\.I
120-160
.
It,

I:tl D4 D6 0

, , !Y'.,,'60-200 GeVic

0.1

.1

=I:tl

.()4

D6

.1.2.3;1.5.6.1

Itl (GeV/cf

Figure 11: Data for the charge exchange reaction np

Fermilab energies.

pn at

which were overwhelmed by the pion term at small s. Figure 12

displays the author's fit to an effective exchange trajectory for
the process. Unlike the results at lower energies, the effective
aCt) has more resemblance to a vector-meson than to a TI trajectory.
This is not entirely surprising because the TI-exchange contribution
falls rapidly with increasing energy. But then arguing this way
makes the presence of the steep peak at Itl~0.02 somewhat enigmatic.
From an extrapolation of the fitted np charge exchange data
to t=O, the authors have provided a bound on the absolute difference between the pp and np total cross section as a function
of beam momentum. These bounds are far more stringent than those
available from direct measurements of the individual total cross
sections. At 250 GeV/c, for example, the expected difference can
be no greater than 0.300.03 mb.
I conclude this section with the note that the simple Regge
ideas have fared surprisingly well in their confrontation with
the charge-exchange data. Although exchange degeneracy has been
dealt a severe blow, the essential features of the Regge-pole

HADRON PHYSICS AT FERMILAB

569

0.6 ~
.- 0.5

.::

,+!' It! t

0.4

t t

0.3

o
Figure 12:

0.1

0.2

0.3
0.4
Itl (GeV/c)2

0.5

0.6

Effective Regge trajectory for the data displayed in

Fig. 11.

exchange mechanism have survived the stringent tests of ultra-high

energies. The small changes in the effective-trajectory parameters
can certainly be attributed to absorption effects or to the presence
of contributions from other non-leading exchanges which cannot be
ignored at low energies. For the np charge exchange reaction
matters are somewhat less clear, but here the low-lying ~-trajec
tory is involved and consequently the test of Regge ideas are more
muddled by possible contributions from cuts and strong absorption.
II.

INELASTIC DIFFRACTION PRODUCTION

Selected Exclusive Channels

An extensive investigation of the dissociation of neutrons in

the reaction np + (p~-)p is presently in progress [21]. Preliminary results, which have recently been pnblished, [22] indicate
that the decay properties of the produced p~- system can be described quite well in terms of a diffractive Deck model involving
pion as well as proton exchange graphs. Figure 13 provides the
essential features of the data. The p~- mass eM) spectrum has a
characteristic accumulation at low mass and shows, in addition,
resonance structure near 1500 and 1680 MeV. The t distribution of
the p~- system is sharply peaked at small angles, as is the case
in all diffractive processes. Although the t spectrum for all M
shows substantial curvature near t~.l, it can be approximated
rather well by an exponential for t$0.2 when the data are not integrated over M (i.e., at fixed M). The decay angles of the

570

T.FERBEL

ICXII

....

"

'Itt

~I;;

'00

~\+..

ttt

.-

tt

++

~-+-

.0'---'011,
. ",111""-'11111
_ r-lI'.m--..-.vn
.......,.....-..,.1-1..

n=E-}s.,.
t

Pl..

~,

c.. Be-CUI

-t

_8

ISO

Figure 13: General characteristic of the reaction np

Fermilab energies.

P1T-P at

proton as examined in the Gottfried-Jackson frame of the p1T system

(for M<1.4 GeV) are also shown in Fig. 13. The scatterplot displays the sort of structure expected from the dominance of Deck
processes in the production. The pion-exchange contribution is
dominant for cos8~1 and the proton exchange near cos8~-1. (Because
there are no known I=~ baryon resonances below M=1.4 GeV, I would
regard with suspicion fits to these data which involve only the
super-position of the 1T-exchange diagram and resonance production.)
Figures 14, IS, 16 and 17 display some other striking and
rather unexpected features of these same data. Simple exponential
fits have been performed to the t spectra as a function of M.
Figure 14 displays the slopes in t obtained in these fits. At the
threshold in P1T- mass the slope is about two times larger than observed in elastic pp scattering at comparable t-values. This
immediately suggests that the diffraction process takes place at

HADRON PHYSICS AT FERMILAB

571

larger impact parameters than

elastic scattering. The slope
falls almost linearly to a
value of 5 GeV- 2 at M~1.55 GeV,
above which it is constant. An
interesting shoulder is observed
in the slope at M~1.35 GeV.[23J
There is essentially no energy
+
dependence observed for the dissociation process. In Figure
15 I show the slope in t for
M<1.4
GeV as a function of
1.2
1.11
1.6
I.B
2.0
momentum. There is no apparent
MASS (rpl (GeV/c2J
variation in this parameter:!
Figure 14: Slope of the t-disalthough a change of the
tribution as a function of
order of ~5%, as observed
the mass of the n-p system
in elastic scattering, clearly
produced in the diffractive
cannot be excluded by the data.
reaction np ~ pn-p.
Another characteristic of
diffractive processes is the
lack of any substantial sdependence in the cross section. This feature of the data is shown
in Fig. 16, where cross sections for specific mass intervals are
given as a function of momentum. The absolute values of these
cross sections are typically about a factor of two smaller than
observed at ~20 GeV/c.

Last year at my seminar here [22J I showed the t-distribution

of the data for two regions of cos8. E. L. Berger encouraged us to
examine the t-dependence on cos8
in more detail and the remarkable
result of this effort is shown
in Fig. 17 (for M<1.35 GeV).
30~----~------'-------'
At very forward 8 the t spectra
MASS < 1.11
display a turn over characteristic of the Deck n-exchange
diagram. At large 8 the t dis~ 20
tribution is rather steep (pre8
sumably due to proton exchange).
UJ
~
But for 8 - n/2 there is an
iil 10
unusually steep interference dip
near t-O.2 GeV 2 [24J. It is
not clear at all what causes
:3110
this highly peripheral feature
200
100
p
(GeV/cl
of the data. The results for
higher mass values are
Figure 15: Energy dependence of
the t-slope parameter in the
reaction np ~ pn-p.

T.FERBEL

572

150....----r----r---,
.1. 1.1.
I I
I
100

++

50

M<I.4

oL - _ - ' - _ - - - L _ - - J
1.4<M< 1.55
b

1.55< M < 1.75

1.75< M< 2.5

Figure 16: The variation with incident momentum of the cross section for the production of a prr- system of fixed mass
M in the reaction np + prr-p.

qualitatively similar, but the sharp dip moves out in t and becomes less pronounced with increasing M (akin to the results in
Fig. 14). I would conclude by saying that although the qualitative
aspects of these data can be understood on the basis of a Deck
model (with absorption), the most striking result, namely the unusual correlation between cose and t is still to be explained.
In addition, it has been emphasized by several authors [25] that
when spin is taken into account in the baryon-exchange Deck diagram (spin was ignored in the work of ref. 22), the predicted
distribution for cose~-l tends to peak at II - rr, which would be
in disagreement with the data. Obviously, more work is required
here to establish the nature of the exclusive diffraction production mechanism.
The double diffractive dissociation (~O) of the proton in the
reaction pp + (prr+rr-)+(prr+rr-) has been studied at 200 GeV/c and
300 GeV/c [5]. This reaction has been compared to the single dissociation (SO) of one proton in the reaction pp + (prr+rr-)+p to see
whether there is evidence for factorization in these Pomeranchukondominated processes. At fixed t and M, Pomeranchuk factorization

HADRON PHYSICS AT FERMILAB

573

...

-1<C058 ...<-.9

ALL CO 58"

-t-i ..

1+1 '"

- .9<C058... <- . 6

,,

.....
'tI'tI

< 1. 35

.,,'

'.

.......
..

. 3<C058....<. 6

'.

-.6<C058.. <- . 3

.....

-.3<C058,,< . 3

...

II

....

. itt,

t+t-t-

to .,.+..,...++

#I~ftt++

~tIit

f--

. 9<C058,,< I

.6<C058 ..<.9
+t

ttt
ttt t

-T

it-t-

-1

~.

.......

~-

t(Gev 2)

Figure 17: Differential cross sections for the production of a

PTI- system as a function of the decay angle (8 GJ ) of
the system in the reaction np + PTI-p.
would imply the following relationship between the DD process
into masses Ml and M2 and the SD processes into Ml and into M2 :
da
da
da
I da
.
dt (M l ,M2 ) = dt (Ml,p) dt (M 2 ,p) / dt (elastlc).
Because of poor statistics the comparison has been performed integrated over M and t. In this case, assuming an exponential t dependence (e Bt ) for all the diffractive processes, the relationship
for the inte~rated cross 2sections becomes:
aSD
BSD
aDD = a -

elastlc

DD el

where also BDD=2BSD-Bel must hold to have factorization in t.

Figure 18 shows the actual comparison. The PTI+TI- mass for
the SD data, again, displays the low-mass enhancement characterizing exclusive dissociation phenomena . The PTI+TI- mass in the DD

T.FERBEL

574

pp_

PQ'w ... ..,.-

200.300 G,Wc

M (p,.-.. + I G,V

20

..,~

(b)

2.. +2"""200.300 <low.

pp- pp

10 M(pw+r+)OfId

o'"

MCpw""'w"-lo'

"E 10
~

'

'" 3: 3 dlGrQllucrt'"

"ltn"OlU 2

'\:::t,.,q..,=-l4-- SM" d

"UClI",-r' )
for "1: 3" or '" 3:1"

II (p ..-".+) GoV

IOO'r---------~

10'r-----------~

PCI- H2,+2r""
200, 300 G.V/c

0.'

0.6

0.8

Figure 18: Comparison of single

and double dissociation of
protons into pn+n- systems.

events appears to be considerably broader; however, upon

subtraction of background
expected from choosing wrong
pn+n- triplets (estimated
from the non-diffractive pn+n+
combinations), the shape of
the 00 mass spectrum becomes reasonably consistent
with that observed for the SO
data. The value of 000 is
37lO ~b whereas 237 ~b is
expected on the basis of
factorization. Also, the
value of BOO (integrate~
over M), is 2.0O.S GeV 2,
whereas 2.SO.6 GeV- 2 is
expected (see the distributions in Fig. 18). Although
these results are reasonably
consistent with Pomeranchukon
factorization and with the
presence of double dissociation, an excellent additional
test would involve a comparison of the expected slopes
only for M<l.S GeV. Here
the value of BO~ should be
larger than Bel' (It is, of
course, not
clear to what
accuracy one should expect
factorization to hold because of the presence of nonfactorizable absorption
effects in the final states.
In addition, the strong M-t
correlation noted in Fig. 14
raises questions concerning
the relevance of results on
factorization when data are
examined integrated over M.)

Inclusive Diffraction Production

As discussed earlier in this lecture, elastic-scattering cross
sections are characterized by their very weak energy dependence
and their steep angular distributions which peak sharply at t=O.
These properties are similar to those expected for the diffractive

575

HADRON PHYSICS AT FERMILAB

scattering of light from an absorbing disk. Certain exclusive

channels, such as the above described neutron dissociation into
low-mass PTI- systems, also exhibit characteristic diffractive
properties. In addition to the elastic and to the low-mass dissociation processes which occur at low energies, there exists at
higher energies a substantial cross section for the diffractive
excitation of a particle into what appears to be a large mass
(M;:::2 GeV) continuum. The inclusive channels which are relevant
here are those which contain leading particles in the final state,
namely, reactions of the kind A+B-+A+Anything [26]. When M21 GeV2
and M2/s1, a regime which exists only at or above Ferrnilab energies, the triple-Regge formalism becomes appropriate to this inclusive reaction.
Figure 19 displays the M2 distribution of the Anything system
for the inelastic reactions:
TI P -+ P + Anything

(1)

pp -+ p + Anything

(2)

The data are from bubble chamber measurements at 200 GeV/c. [27]
The dramatic enhancements at M2~25 GeV 2 in Fig. 19 have two contributions: one is from the
dissociation of the projec240
tile particle into low-mass
(a )
resonances or peaks (such as
200
d IT / d M2 inclusive
the 3TI Al enhancement for
pp-p+X
reaction (1), and the
~ 160
t9
---- .".- P- P +X
"N*(1400)" pTI peak for re".a
action (2)); the other con.:: 120
tribution is from the excitation of the projectile
~I~ 80into the multiparticle
40
large-M 2 continuum mentioned above. The total
cross section for the peak
in reaction (1) is "'2mb,
and in reaction (2) it is
"'3 mb (the latter is for
the excitation of only one
proton). An analogous
o
50
200 250
peak is observed in the
reaction ITp7TI+Anything,
also at the 2 mb level.
Figure 19: Inclusive cross sections
Hence, the sum of the
(integrated over t) for the procross sections for the
duction of protons in the regime
excitation of either a
of target fragmentation at 200
OJ

GeV/c.

576

T.FERBEL

target or a beam particle are comparable to elastic scattering cross

sections for the same incident channels (3 mb and 7 mb respectively).
Other properties of the peaks at M2<25 GeV 2 can be summarized
as follows [26]: (1) The average multiplicity of the events in the
enhancement is considerably lower than for events outside of the
peak; the mUltiplicity of the Anything system grows approximately
as Q,n M2. (2) The cross section for the entire low mass peak is
almost energy independent. (3) The t distribution of the M2 system
is sharply peaked at small angles, particularly for the very lowest
M2 (recall the n-dissociation data). (4) These peaks occur only
for systems which have the same charge, strangeness, and baryon
number as the incident particle. From these characteristics it is
reasonable to presume that the low mass enhancement is mainly a
diffractive phenomenon involving Pomeranchukon exchange.
Figure 20 illustrates the triple-Regge pole graph which is
used to describe the break up of a particle B into a massive system (M). R. are any Regge poles which can be exchanged. The sum
oyer M is t6 be interpreted as a sum over all states contributing
to the reaction at fixed M2. The large-M 2 requirement of the model
has to do with the substitution of the R.' Regge pole for the elastic scattering of R on B at t.=O; this i~ only valid at large M2,
above the resonance region. JThe triple-Regge formula can be
written as
dO'
--2 ~

dtdM

s )
I s.. (t) r-2

ij

1J

2[a. (t)-!z a. (0)] [ ()-l]

1
J
a. 0
s J

where S is an unknown function of the vertex couplings. When i=j=P

(Pomeranchukon), then the invariant cross section at fixed M2/s~
I-x becomes essentially proportional to s/M2 and s-independent.
Also, the cross section d 2a/dtdM 2 has a M- 2 form, independent of
energy. When j=V (vector meson) and i=P we obtain another diffractive term which has the approximate form (s/M2) 312 s-Yz for the
invariant cross section. At fixed M2/s this term falls with

R:
~:~

:EM
8

(a)

Figure 20:

A
A
A
R II) Ri II

--+

(b)

RJ

M2 >')1

Ilj 01

8
(e)

Relationship between an inclusive cross section

and the relevant triple-Regge graph.

HADRON PHYSICS AT FERMILAB

577

1:

energy as S-2
Another interesting situation occurs when i=V and
j=P. This non-diffractive R-P-R term scales with energy (i.e., is
independent of s) and consequently d 2(i/dtdM 2 at fixed M2 falls
approximately as s-l. (The above statements are only roughly
correct because we have ignored Regge interference [28] terms as
well as the t-dependence of the aj' and have set aV(O)=O.S and
a~(O)=l.O.)
The P-P-P term is expected to dominate for small
M /s. The P-R-P term should also be important at low M2 and become negligible at very high energy. The R-P-R term should dominate for M2/S beyond the diffractive peak. Data appear to exhibit the gross features of the triple-Regge parameterization.
Figure 21 displays the M- 2 fall off (beyond the resonance
peak which is observed near M2_3) expected from the P-P-P term
for the large-mass diffractive excitation of target protons in
n+p and pp reactions between 140 GeV/e and 170 GeV/c [29]. The
authors have extracted the P-p total eross section and the P-P-P
coupling constant from their np, Kp and pp proton-fragmentation
data. The results from all channels agree; i.e., cross sections
factorize to ~10% accuracy and the P-p cross section can be
written in the factorized form [29]:

cl p

tot

(M 2 , t) = [M 2

~
dtdM 2

/ do
] 116n(do/dt)
dt elastic
pp

the expression in the bracket refers to the ratio of inclusive

and elastic scattering for the same incident channel. The value
of 0pp(t) is independent of M2 (Fig. :21) and equals
2.9 exp(-l.04t+0.:.:!iSt 2)mb; at t=O the P-P-P coupling constant ro
is 0.800.03 GeV 1. (It is interesting that at larger values of
M2, where the PPP term is no longer expected to dominate, factorization is grossly violated in pp data relative to TIp and Kp
channels [30]. This can be attributed, for example, to the
presence of different RPR terms, such as nPn, in the pp as
opposed to the meson-proton reactions.)
The energy dependence of the TIp and Kp data at low M2/S has
been checked and appears to exhibit an s-~ component, particularly
at small t, similar to that observed for older pp results [26].
In summary, there exist at high energy large cross sections
for the diffractive excitation of hadrons into multiparticle systems. The cross section for the inelastic excitation of a proton
in a pp collision is ~3 mb, and that of a pion in a TIp collision
is ~2 mb. Consequently, the total diff~active, essentially energy
independent cross section (elastic, inelastic, and estimated
double diffractive) is ~O. 4 0TOT ' for both TIp and pp data. The
inclusive leading-particle distributions in np, Kp and pp

T. FER BEL

578

reactions at large x values are consistent to ~10% accuracy with

factorization of the Pomeranchukon contribution to the cross
section. The triple Regge formalism appears to yield a surprisingly good description of the data [28].

1rP- ..X
10.0

1.0

Ial

<-t>

0.1 -'~""""""Oll"oQ225~OOQooooooo~
0.4 ~ooooaaaooooo~

OJ

O.~oneo 000000

N
"CI

pp_pX

10.0

~oooo~ooauoa

0.1
1.0

O.2a5~oo ooo_"ooo
~ovoo 0=0 0.0.0 ClOoO

0.4

O. I

IN

(.-t>

O.6~oAQoono .... ~

468
M,.2 [~]

Figure 21:

Test for the presence of the trip1e-Pomeranchukon

contribution to leading-particle reactions.

III. INCLUSIVE PARTICLE PRODUCTION

Nuclear Targets
Mu1tipartic1e production off nuclear targets can, in principle,
be used as a tool for examining the space-time development of
hadronic processes. For example, if the asymptotic hadron final
state in a hadron-nucleon collision evolves in a distance far
shorter than the nuclear dimension, then in a hadron-nucleus
collision we might expect that the final-state particles emitted

579

HADRON PHYSICS AT FERMILAB

in the initial interaction of the hadron with a particular nucleon

will collide again with the downstream nucleons in that same
nucleus, initiating a high-multiplicity cascade. On the other
hand, if the initial collision involves the production of a resonant system which decays outside of the nucleus, there would then
be very little cascading expected. Consequently, if the hadronnucleon interaction time is short we expect, naively, that the
charged-particle multiplicity in a hadron-nucleus collision will
be a strong function of atomic number A and energy. On the contrary, the A dependence and s-dependence should be weak if the
time scale for the development of the hadron final state is long.
Figure 22 presents the ratio (R.'E) of the inelastic chargedparticle multiplicity in nuclear emu1sion (average A~73) to that
in pp collisions as a function of laboratory momentum [31]. After
an initial increase of ~E at energies below 100 GeV/c, the multiplicity in nuclear emulslon is observed to saturate at a value
well below ~E=2. This well-known result from cosmic ray studies,
supplemented oy new data from FNAL, speaks against a cascade model
for production in nuclei and for a rather long time-constant for
the development of the final state.
Busza et al [31J have performed a series of measurements of
multiparticle production in ~, K and p interactions with nuclear

2.0

..

1.5

RNE

1.0

I t

to

f f I
o Buszo et 01

Gurtu et 01
(compilation)

0.5

10
102
10'
Proton Momentum In Laboratory (GeVlc)

Figure 22:

104

Ratio of inelastic multiplicities for particle production in nuclear emulsion relative to production
in hydrogen.

580

T. FERBEL

targets ranging from hydrogen to uranium. The data consist of

angular distributions of secondaries for 50 GeV/c, 100 GeV/c and
200 GeV/c beam particles. Figure 23 provides a global view of the
overall richness of the data. The multiplicity in pA collisions
is plotted for different intervals of the pseudo rapidity n, as a
function of nuclear target material. (n=Q,ncot (8/2), 8 is the production
angle in the laboratory. n is a good approximation to the laboratory rapidity variable y=Yz Q,n[(E+pQ,)/(E-pQ,)], where E and PQ, are,
respectively, the energy and longitudinal momentum of any particle.
n and y become indistinguishable when the transverse momentum is
larger than the particle mass, p~m2.) Because of the difference
in the inelastic TIP, Kp and pp cross sections, and because even
large nuclei are partly transparent, the average nuclear thickness
of a target material depends on the type of incident particle. A
parameter (v) can be defined to take account of this difference
in inelastic cross sections for different channels. The value of
v for any nucleus A is given by Aainelastic/ainelastic, where h
represents any incident hadron.
hp Thus deff~ed, the
parameter
describes the average thickness of the nucleus in units of absorption mean free paths for hadron h. Also, V gives a measure of the
average number of inelastic collisions h would make with the nucleons if subsequent to each collision it could interact anew.

The data presented in Fig. 23 indicate that the produced

multiplicity is strongly dependent on n and on
except at large
n (small angles, or forward production), where the multiplicity
is independent of nuclear size. The authors indicate that although
pion and proton-induced reactions exhibit different behavior for
the same nucleus, the behavior is similar for target nuclei with
identical values of v. (The correction for V is important because,
typically, pions have absorption lengths of ~3 f in nuclear matter
while protons have absorption length of only ~2f.)

v,

Figure 24 presents a plot of RA (average multiplicity relative to hydrogen, integrated over n) as a function of
for ~
and p data. There is no clear systematic variation of R (v) with
energy or with the type of incident beam particle. The ~ependence
of R on V requires a quadratic term in
for an acceptable fit to
the ~ata. (For proton data RA is approximately proportional to

AO. 25

.)

In Fig. 25 I show the n spectra for the proton data at 200

GeV/c. As V (or the nuclear-target size) increases, the multiplicity is observed to grow rapidly everywhere except in the very
forward direction.
Although a variety of theoretical models have been suggested
in order to explain all these remarkable features of the data [32],
the Energy Flux Cascade model [33] appears to be in best overall

HADRON PHYSICS AT FERMILAB

200GoV

1.39

200 GeV

p'

.0.'" c'rJ c t.l'

;;

~!~

(a)

p'
C

~f

2.2'

j'
(b)

c'rJ

J(

581

2000,V

P+

22"~"'o.

/IV
I

J /'

",/C)I

Ailll

Ill'

.0.S7. "1 II 0.'2

20
200 GeV PiIIOTOMl

toO GeV/c p.
1-

7.00 ..

"c ".01

"
(e)

(t)

~
f
J

Figure 23: Inelastic multiplicities for particle production in

proton-nuclear collisions at 200 GeV/c. The data are
displayed for various angular regions (n) as a function
of the parameter v (see text).
agreement with the measurements. In this model it is assumed that
the initial collision of a hadron with one of the nucleons in a
nucleus generates a distribution of energy which has the same
rapidity content as that observed in hadron-nucleon collisions.
This energy flux travels through the nucleus, expanding spatially
with time, and behaves, as far as its interaction with nucleons
in its path, as several systems of hadrons (one large-rapidity
pack and perhaps several small-rapidity packs, any group being
approximately 1 fermi wide, as measured in the rest frame of that
pack). Although all the packs or slices of the original flux can
collide with another nucleon, only the large rapidity component
has enough energy to produce a new flux (of somewhat lower rapidity content). This process repeats V times yielding a weak muliplication of particles. Although the Energy Flux Model provides
excellent qualitative agreement with the data, quantitative predictions, such as the asymptotic prediction RA ",1 + (\5-1) /3, are

1.0

I.e

to

(al

i'

.,
.

2.0

ji

I-

eo GoV I'
100 GoV I'
aoo o.v I'

3.0
4.0

",.... _&4 ..

eo Oov ,,100 'oV ".

100 OOV ,,-

.. aoo 'oV"-

,
..,

.,

,I

Figure 24: Ratio of inelastic multiplicities

for particle production in proton-nuclear
and pion-nuclear collisions relative to
multiplicities on hydrogen for data between 50 GeV/c and 200 GeV/c, displayed
as a function of v.

R..~~

<n=\

.0

l.5

~I _._
i
I

"1

_J

,- _ _ -J

"L

~::"'::"'.....,

--,.; - - --,

~" _"'
~_'---,
I

II

2
345
.". -iii hon
/2)

II

I
I oJ

I
I
I
J
I
"~_-+JI

r
I.

,._p", I
I )
rJ
L _,I
I
~
I
I
""",.-'
I

! r-1

p+

-11

___

-liol
f7 02

'--y

i.

i,

L,

r--"

--1

I:

r.J

...J

!
I
i

,..,

Figure 25: Differential multiplicity

distributions as a function of the
angular variable n for different
target materials for 200 GeV/c
inciden~momentum protons.

-I

AT}

AN

!S

t.n

m
r

to

::0

TI

:-I

...:>

0>

HADRON PHYSICS AT FERMILAB

583

clearly inconsistent with presently available measurements.

The Energy Flux idea certainly appears to corne to grips with
the problem of multiparticle production in nuclei, nevertheless,
there is, perhaps, a more familiar way to understand the above
physics. Figure 26 displays the 2n-n+ and 2n+3n- mass distributions observed in inclusive pp collisions at 400 GeV/c. The data
indicate that the mean value of the mass spectra are 2-3 Gev, depending somewhat on multiplicity. The widths of these distributions are 1.5-2.5 GeV, comparable to the mean values. Now, if
as it is commonly believed, hadron clusters are produced in particle collisions, and the average number of particles in a hadron
cluster is ~3-4 (on the average we have about 4 clusters per pp
collision at 400 GeV) , we expect that the typical cluster of
hadrons produced in high energy collisions will have a mass of
~2.5 GeV and a width of ~2 GeV.
If the clusters are distributed
uniformly in rapidity, then the forward-most cluster will have
essentially all the beam
momentum and the backwardmost cluster will be essentially produced at rest in
40
the laboratory. Because of
the localization of the
I'
I":
cluster mass, using an unI I
I I
certainty-principle argument,
I I
5 Pions,Q=-I,.otOO GaV/e
I L,
the fast forward cluster
I I
I
I
would have an effective
I
;
I :
lifetime of ~!r and a mean
I
I
I
free
path for decay of
,J
3 Pions,Q=-I,400 GaVIe
>Ie _ ytlc/r", (400/2) (2 x IO- 1 L!) /
a::
c(
2 '" 20 fermi! Consequently,
a::
~
if the interaction occurs
iii 20
on a nucleon inside a nucleus,
a::
c(
it is clear that the leading
cluster will behave as a
single massive object as it
travels through nuclear
10
matter. Thus time dilation
will clearly diminish cascading effects within nuclear
matter. If the leading cluster interacts again within
the nucleus,the secondary
246
interaction will produce new
MASS (GeV)
clusters distributed in
rapidity in a manner similar
Figure 26: Mass distributions of
to that in the first colli2 n+ 311 and n+211 systems produced
sion. Thus the cross section
in inclusive pp collisions at
in nuclei at small y will
400 GeV/c.

584

T. FERBEL

be increased as a result of the secondary collisions but the forward y component will not be grossly affected. (The slower secondary clusters will again tend not to have sufficient energy to
multiply in the nucleus and therefore the cascading will not be
severe.) Although the above remarks are rather qualitative, they
may contain the essence of the physics of multiparticle production
in nuclear matter [34].
Hydrogen Targets
Multiplicities. One of the general qualities to emerge from
investigations of hadron collisions at Fermilab is the essential
similarity of all hadronic matter. It appears that the total
s-value of a system, and possibly the charge, determine all the
major features of multiparticle production in hadronic collisions.
Figure 27, taken from reference 5, displays the veracity of these
remarks. The figure shows cross sections as a function of chargedparticle multiplicities (n) observed in pp, ~~ and K+p collisions
at 100 GeV/c. Except for the absolute scales, the distributions
are all quite similar. In detail, however, differences are apparent between the various channels. In particular, there is an unusually large excess of cross section in pp channels relative to
pp reactions for n ~ 12. The opposite may be true for n=2. It
will be very interesting to compare pp and pp values of
at
~ 300 GeV/c, by which time the cross sections for the n=4n multiplicity may also reverse (i.e., opp>opp). ! expect this to happen
because stronger final-state absorption in pp reactions should
always tend to reduce pp cross sections relative to their analogous
pp reactions (such as, for example, the low~multiplicity diffraction channels). In other words, as the pp and pp total cross sections become more equal, their small difference will appear, partially at the expense of low-multiplicity reactions, at ever increasing n-values. (I am assuming that the mean multiplicity in
the annihilation channel, as in all processes, increases with s.)

Conclusions pertaining to the pp - pp difference similar to

those reached above have been advanced previously by Eylon and
Harari [6]. These authors suggest that on the basis of their
model (mentioned earlier in the lecture) the quantity Rn=[on(pP)0n(pp)]/on(pp), at fixed s, will approach the form (3/2)n for
large n; also, for fixed n, -Rn will approach the form (1/2)n at
large s. Figure 28 displays a check of the dependence of R on n
for several s-values [35]. The data have been fitted succe~sfully
to the form Rn=Sns-a, where the s-dependent parameters S and a
appear to approach "asymptotic" values of S=1.24 and a=O. 73 at
100 GeV/c. In the specific model of Eylon and Harari S should
approach 1.5 while a/2 should become equal to the difference between the intercepts of the exchanged meson and baryon trajectories.

o.

t
+

/:),

12

16

f ~

1.

a.

~ ~

p~

pp

20

24

~ tl
t\

C K+p

0 IT+P

/:),lTP

Figure 27: Cross section for producing n

charged-particle inelastic final states
in hadronic collisions at 100 GeV/c.

0.0

b"

:aE

1e

I"

I'

{'

,,/

l'

~'

I:t/~~~~:V/

"

'

/ /

"

10

t'

/'

/"

12

'

'

,,/

'

//

14

22.4 G.Vlc

16

18

6.9 r;.Vlc

6.94 GeVic

14.75 G.Wc
.. 12 G.Wc

"

G.Wc
G.Wc

100

o 32

35/'////'

22'~/j
'

/~/ t//lJ/
f

t' ' /
,f / /~/ ?,/

II

~,

'
A
12/~t,'~l"75
, ,,

Figure 28: Linear fits to the ratios:

[an (pp)-an (pp)]/[an (pp)] at
different momenta.

0.1

1.0

Rn

10.0

,,

6.9/6.94,'"
,
I

,,

til
00
til

to

3:

:tJ

"T1

-t

en

(")

-<
en

:I:

"C

:tJ

:I:

T. FERBEL

586

Both asymptotic predictions are in rough, albeit surprisingly good,

agreement with the data. It is not possible as yet to test
whether -Rn approaches (1/2)n. (This is related to the fact that
~%(pp7ffiesons) essentially saturates the Eylon-Harari bound of 1,
which naively also implies that difference ratio ~ is due entirely
to the annihilation contribution to the total cross section.) Nonetheless, as mentioned in the previous paragraph, higher energies
may provide a confirmation of the essential prediction that, for
fixed n and large s,On(pp) > 0n(pP).
The dependence of multiplicity distributions on target material
is displayed in Figs. 29 and 30. Figure 29(a) compares pp with pn
data at ~15 GeV/c, and Fig. 29(b) compares pp with pn data at 300
GeV/c [36]. (The neutron data were obtained using deuterium targets. The distributions shown in Fig. 29 have been corrected for
secondary rescattering and are somewhat model dependent.) Although
in detail it appears that the inelastic mean charged-particle
multiplicity <n > tends to be smaller in neutron-target data than
in proton-targef data, to an excellent approximation, the distributions in Fig. 29 simply interleave as a function of nc:
On (pp) = !z [on -1 (np) + On +1 (np)] ,
c
c
c
again suggesting the presence of similar, target-independent, production processes.
Figure 30 displays multiplicity data for ~ on Ne at 200 GeV/c
(solid circles), ~- Ne data at 10.5 GeV/c (open squares) and ~-p
data at 205 GeV/c (solid triangles) [37]. The multiplicities are
presented in terms of the Koba-Nielsen-Olesen (KNO) scaling parameters, [38] where n~ for the Ne data refers to pion shower tracks
(i.e., nuclear break-up prongs have been excluded from consideration). Except for the fluctuations in the Ne data, due to enhanced
coherent diffraction production at low multiplicities, the three
distributions are again very similar, indicating the universality
and target independence of multiparticle production processes.
It is well known that the dependence of <nc> on s is not just
simply logarithmic [5,26]. Several years ago I showed [39] that
the dependence of <nc> on s in pp collisions can be represen~ed
using the phenomenological expression <nc>=A+B Q,n s + C Q,n s/s~ .
The fit was excellent for data from ~ 10 GeV/c up to ISR energies.
This form is more ~ppealing than an often used quadratic [5] dependence on Q,n 2 s because it provides a way to check whether,
asymptotically, multiplicities grow as Q,n s and become independent
of the incident channel. If all channels yield the same value for
the parameter B, the result would suggest (although without a
theoretical justification for the above form of <n that the
c

"

12

*
f

*.

nc

'ft

pp 300 GeV/c

"

tXU X.

*+

12

*+

20

1111

16

*'t

X pn 300 GeV/c

(b)

p--neutron inelastic multiplicities.

Figure 29: Comparison of p--hydrogen and

0 .01

C.

be

.....

r;

0 .1

14.75 GeV/c

X pn 14.60 GeVlc

pp

(0)

I)

t+

0.5

1.5

i+~,

.c+"t +,.. +--:

2
2.5
Iln/{nn) -

++'t

+~D

Figure 30: Comparison of rr--Ne data with

rr-p multiplicities in terms of the
KNO scaling parameters (see text).

0.01

0.1

be: ~
.......'" be:
........

I. .

1=

1.0

'I

0.
00

tD

l>

:s:

:0

"

-t

en
l>

c=;

-<
en

'"0

:::c

:::c
l>
o
:0
o

T.FERBEL

SBB

small differences which exist in <n > are only low-energy effects
and that at large-s multiplicities ~row logarithmically and
Pomeranchuk factorization holds. Fi*ure 31 displays the highest
energy <n > data available for yp, p-p, ~p and K~ [40]. I have
tried to festrict the fits to data for ~10 GeV/c incident momenta.
Unfortunately, the only available yp measurements are from a lowenergy SLAC experiment [41].
The KP and ~+p data each consist of only three points - just
enough to determine the three parameters. The pp data have but
four points and extend only up to 100 GeV/c. The best measurements
are for pp and ~-p collisions; these extend from ~20 GeV/c to ~400
GeV/c. The results of all the fits are given in Table I, and
several of the curves are graphed in Fig. 31. Except for the
relatively poor X2 for the fit to the sparse pp data, the fits
and the agreement in the value of B for all the channels is quite
impressive! We see that all data are consistent with Pomeranchuk
factorization at large s-values to perhaps ~5% accuracy. It would
be invaluable to have Kp and pp data points for momenta in excess
of 200 GeV/c (and, of course, any high-energy yp data) to provide
a better check of the form for <n > and of the universality of
c
the B parameter.
9
8

yp

pP
opp

".+p
D ".-p
.. K+p
.. K-p

<nc>
5
4
FITS TO: A+B 109 S+C

151
5"4

10

S (GeV 2 )

100

1000

Figure 31: Inelastic average charged-particle multiplicities as a

function of s for various hadronic channels. Curves represent
fits to the data (see text and Table I).

589

HADRON PHYSICS AT FERMILAB

TABLE I
Fits of Average Charged Particle Multiplicities to the Form:
A + B J/,n s +

C J/,n s
-~

S4

PARAMETERS
CHANNEL

2
X /d.f.

pp

5.35 1.5

1.36 0.05

4.26 0.92

1. 07

pp
+
KP
K-p
+
IT P

5.44 3.2

1. 30 0.07

3.89 2.0

4.18

2.17 5.0

1.29 0.12

1.57 3.7

6.38 3.5

1.34 0.10

--4.73 2.3

3.82 6.3

1.19 0.10

--2.26 4.2

IT P

0.67 1.9

1.40 0.10

--0.82 1.0

1.15

'YP

1.67 0.9

1. 28 0.53

--1. 09 1.6

0.31

Inclusive Production Spectra and Limiting Behavior. Typical

rapidity distributions for pions produced in high energy collisions
are displayed in Fig. 32 for 102 GeV/c and 400 GeV/c pp data [42].
The data have been inte~rated over PT' an~ no corrections have
been applied for ~10% K-,~l% e, and~2% p contamination. (The
estimated corrections required for proton contamination of the IT +
spectra are indicated on the graph.) It is clear from these data
that the cross section does not scale in the central region of
pion production. In particular, for YCM=O (YLAB=2.69 and 3.37 at
102 GeV/c and 400 GeV/c, respectively), the inclusive IT- cross
section rises by 4510%, whi Ie the IT'- cross section rises by
3010% between the two energies. At 400 GeV/c the IT+ cross section is ~15% greater than the IT- cross section near YCM=O, indicating that asyrnptopia is still far away. For small values of
YLAB ' where the IT+ yield is about three times the IT- yield, there
is some indication that the cross section falls with energy [43].
The IT- data at 102 GeV/c and 400 GeV/c are compared in more detail in Fig. 33. The rise in the invariant cross section with increasing s is clearest at small x. Near Ix I~o.l the cross section
appears to scale; while at large Ixl" and especially at small PT'
there is a small violation of the Hnlothesis of Limiting
Fragmentation [43].
Figure 34 displays invariant
observed in ITP collisions [44].
non-leading IT-data are similar to
collisions. Namely, near x=O the

cross sections for pion spectra

The characteristics of these
those just discussed for pp
cross section grows with energy,

T.FERBEL

590

102 GeVlc
o 400 GeVlc

100

.".-

(xIO)

.0.1o__0.00000
0

orolfO

rfJ

0.0

.".+

1.0

2.0

3.0

YLAB
Figure 32: Rapidity distributions (integrated over PT) of pions
produced in pp collisions at 102 GeV/c and 400 GeV/c. Data
have not been corrected for K, e or p contamination
(see text).

while at large lxi-values cross sections tend to fall with increasing s.

Whitmore [5] has recently updated and refined an earlier
study [45] of the energy dependence of inclusive particle-production
cross sections at YCM=O. Figure 35,taken from reference [5],
summarizes the situation concerning pion production for various
reactions. The data are presented as a function of s-~ in terms
of the invariant cross section, integrated over PI' and normalized
by the contributions of the Pomeranchukon to tota cross sections
for each of the incident channels [46]. (The solid line shown on
the graph for n- production represents recent ISR measurements.)

10

.t~

0.!5

f.

1.0

t~t

*f

:ft

tt

400~/c

't

tt

..

t.

-.01<. <.01
102 GoNIc

[0"1

- .20<0<-.10

1**

I\l

+1 t

- .03 <-.01

i,

!~

0.!5

1.0

If!
~~

-.50< '<-.20

{r

-.05<,<- .03

tf~r

It

0 '

,j)

~ O. ,

I I Gtif/c

o.

'lfp.w

10

1. 1

...,

' ,0

..

0"

.GttN/c

,
o

w+p."'-

0 ..

' ,0

I.'

Figure 34: Invariant cross sections, integrated over PT' for the inclusive production of 7T+" and 7T- in 7T - P and TI+P
reactions in figures (a) and (b),
respectively.

;;:

:I--;.~

b.....:-

-i

Figure 33: Invariant cross section for negative pions

produced in pp collisions at 102 GeV/c and 400 GeV/c,
as a function of PT' for different regions of x. Data
have not been corrected for small contamination from
e-, K- and p.

IIJ

't)

't)

",0.

..s

"-

C>

10

+.

01
-0

l>

r
eo

s:

:0

-n

-I

l>

(fl

(')

"::c-<

:0

::c
l>
o

T.FERBEL

592
(b) rr+ production

( a) 11"- production

0.40

j5p

pp

0 .35

oII

bl>-

""

-If

0 .30

+ rr-p

+99

0 .25

...l

9 o K-p

0.20
0 . 15

0 . 10

0 .05

+,.

pp
o pp
pp non-ann
t:. tr+p
tr-p

is)

pp(NonAnn)

o K-p

is)

2000 400 100 40 20

10

is)

:3

2000400 100 40 20

Plob (GeV/cl

0.1

pp (Non Ann)

0 .2

0 .3
$-1/4

0.4
(Gey"1/2)

10

Plob (GeVlcl

0 .5

0 .6

0 .1

0.2

0 .3
$.1/4

0.4

(GeV- 1/21

0 .5

0 .6

Figure 35: Density function for the production of ~- and ~+ mesons

at1yCM=O, for various incident channels, as a function of
s-~.
See text for explanation of the normalization.
Mueller-Regge ideas would suggest that cross sections for ~+ and
~- production should factorize at large s (converge to the same
value independent of the incident chap'nel) , and, furthermore, the
approach to scaling should have an s-~ dependence. We see that
data for all reactions may be converging to the same value of
dcr/dYCM at YCM=O as s -roo. As indicated previously [39], this
value in
Mueller-Regge phenomenology should equal B/2~,
where B is the parameter determining the asymptotic behavior of
<n >. From Table I we see that the best value for B/2~ is
O.~IO.OI, a result in rather poor agreement with extrapolations
suggested in Fig. 35. I am not certain how to interpret this discrepancy. One possibility, of course, is that the agreement for
all B values in Table I is purely accidental. I can, however,
take a rather positive view and say that the discrepancy between
the extrapolation in Fig. 35 and the value of B/2~ is only off by
~30% of the extrapolated value!
Considering all the theoretical

593

HADRON PHYSICS AT FERMILAB

uncertainties regarding asymptopia, this is, in a sense, still a

remarkable result [47].
Independent of the above discrepancy, the data displayed in
Fig. 35, and the K~ production data displayed in Fig. 36, suggest
that, after leading-particle effects subside, the inclusive cross
sections at x=O increase asymptotically with increasing s, grossly
consistent with expectations from Mueller-Regge ideas.
A comprehensive investigation of the approach to limiting behavior of pion production in the target-fragmentation regime has
recently been reported in the literature [48]. Figure 37 displays
the results of that compilation. The data are presented in terms
of an integral of the inclusive cross section over PT' and over
a fixed interval of longitudinal momenta in the laboratory frame
(PL). Specifically, the function G(PL)' which is defined as

Plab (GeV/c)

0.05 r-_....;1'""0..,0=-0-"3T=-0...:.IOr=<--...>3r0'--_TI0'--~5~-'i3~~

0.04

o
10

*,...

b~1 0.03

>.
-0-0

-If

0.02

0.01

0.1

0.2

0.3

0.4

0.5

0.6

s-i/4 (GeV- it2 )

Figure 36: Density function for the production of KO mesons at

YCM=O, for various incident channels, as a function of s-~.

594

T.FERBEL

[n/a p (s)]/E(da/d 3 p)dP?, and is normalized as in Fig. 35 by the

contribution to the total cross section [46], is
examined as a function of s-~ This sort of s-dependence is expected on the basis of application of Mueller-Regge ideas in the
target-fragmentation regime for reactions studied in Fig. 37 [49].
Asymptotically all n- production channels should factorize, and
separately, all n+ production channels should factorize. This
appears to be happening at increasing s. At small s-values different channels can have contributions (in addition to the s-independent Pomeranchuk term) from different exchanges; furthermore,
if Regge trajectories are not degenerate, and if leading trajectories do not have intercepts of a(O)=~, then the approach to
scaling can be fairly complicated and not just of the s-~ form.
Figure 37 suggests that, in fact, the picture is quite complicated.
Nevertheless, the simple exoticity pattern of Chan et al [49]
seems to be in qualitative agreement with the data. That is, reactions A+B~+Anything, where the quantum numbers of (ABC) are
exotic, display least energy dependence in their inclusive cross
sections.
Pome~anchukon

Results similar to those compiled in Fig. 37 have recently

been obtained in a counter ex~eriment at BNL and Fermilab [50].
The data are for K, p and n- projectiles incident on protons, in
the momentum ran~e between 4 GeV/c and 250 GeV/c. The variation
with energy of n-, K and p particle production was examined at
a fixed PT~0.3 GeV/c, and values of YLAB=0.6, 0.4 and 0.2. These
global studies indicate that the approach to asymptopia is, again,
not just of the s-~ form. I reproduce one of their graphs in
Fig. 38, which summarizes their pion-production results. To achieve
factorization of particle-proton and antiparticle-proton inclusive
cross sectiop-s as s~, the authors require in addition to an energy
dependent s-~ term an S-l term (from lower-lying trajectories) at
low energies. Cross sections for meson-proton and baryon-proton
channels here again appear to be in agreement with factorization
at large s (when using normalizations akin to those suggested in
reference [46]).
The SAS Group has also examined inclusive channels involving
the exchange of quantum numbers [51]. The available data, which
are for the region of projectile fragmentation, have been used to
extract the effective Regge trajectories for production at large
x values. The specific reactions studied are given in Fig. 39.
The authors analyze their data using a semi-empirical one Reggepole exchange formula, similar to our triple-Regge expression,
keeping explicitly the total cross section for the scattering of
the exchanged object R(t) on target B (see Fig. 20). The specific
form used is:

HADRON PHYSICS AT FERMILAB

595

0.10

"'\'f+-p-n-

n-p

..

K-p

TT ... p .....

0.20

... K'.
X

~.

..,
~

~
~

N~ 0 .06

0'"

'--J!

I.

f P.

~
.. /A
h

"'+-

x-

,.

I<"'p

~.

'J16

x-.

____

0.08

. . - - - TT+P

,,44;--9____ .-
.- ~:.:--+~o
-<>*~-~_

1/+

-"x-.
-+
'~
x+

..-

~
0 .0'

0 .1

0 .4

0 .3

0 .2

0.1

) V

0 12

/'

PI'

0 .02

//
I

0 .08

0 .04

.. , -.

(bl

lal

0.2

0 .3

o.

'-":I(GeV")

Figure 37: Density function for the production of 1T+ and 1T- mesons
in the region of proton fragmentation in hadron-pr~ton collisions. Data are displayed as a function of
s-~.
Pbeom (GeVlc)

i
~
]
Z

II!
VI

~0

0 .7
0 .6
0 .5
0.4
0 .3
0 .2
0 .1

0 .2
0 .1

0 .3
0 .2
0 .1

0 .1
0 .4
0 .3
0 .2
0 .1

0 .4
0 .3

m
~ 0 .1
0 .1

0 .2

0.3

0 .1

S"'12 (GeV,-'

Figure 38: Cross sections for 1T+ and 'IT production in the region
of proton fragmentation for hadron-proton collisions. The
integration is over the following final-state momenta and
approximate angles in the laboratory frame:
0.3 GeV/c , I"PI , 0.6 GeV/c and SOo ::; 8 ~ 60 (for details
see ref. [50]).

T.FERBEL

596

2
l-2a
2
d CJ _ f(t) (I-x)
eff CJRp(M )
dtdx where f(t) contains all the t-dependence, a ff are the effective
values of the exchanged trajectories at theet-values specified in
Fig. 39. The energy (i.e.,M 2) dependent cross section for Rp
scattering is taken to be the same as for TIp, Kp or pp scattering,
depending on the nature of the exchanged object (for strange-meson
exchange Kp is used, etc.). The data in all but the ~+p+p+Anything
reaction are consistent with scaling in s to ~15% accuracy. Consequently, all data in the SO GeV/c to 175 GeV/c range were combined to study the x-dependence of the cross section. For the
~+p+p+Anything process the cross section decreases by ~30% in the
energy range of the experiment, and therefore only 140 GeV/c and
175 GeV/c data were used to extract a ff' The ordinate in Fig. 39
is defined as an average of the crosse section over sand t, normalized to the total cross section for Rp scattering at 200 GeV,
as follows:
CJ (p. = 200 GeV/c)
[ Rp III

d 2 CJ ]

dtdx average
The values of a ff for exchanges involving K* quantum numbers are
a ff~0.22 at <t~ ~ -0.3 GeV2, amazingly consistent with the K*
tfaJectory. Similarly, the values of aeff for baryon exchange processes typically equal aeff ~ -0.55, agaln in qualitative agreement with expectations from baryon exchange. The difference observed for the cross sections in Fig. 39(a) is presumably due to
u-channel nucleon exchanges, which are expected to fall steeply
with increasing s (as they, in fact, appear to be doing). The
agreement observed between the C-conjugate reactions in Fig. 39(b)
is impressive indeed.
Local Compensation of Quantum Numbers. Recent theoretical investigations by Krzywicki and Weingarten [52] have established a
new industry for the determination of whether quantum numbers,
such as charge, strangeness, baryon number, or even kinematic
quantities such as transverse momentum, are locally compensated
in rapidity space. The LCQN hypothesis asserts that any produced
particle carrying a quantum number q must be accompanied nearby in
rapidity space by a small group of particles carrying a total value
of the quantum number -q. Local compensation of electric charge
had previously been shown to hold [53], and now there is also evidence for the local compensation of transverse momentum in high
energy collisions [54].

HADRON PHYSICS AT FERMILAB

597

The best way to illustrate the essence of the LCQN hypothesis

is through the idea of a zone graph. Let us consider for simplicity a ten-particle final state produced in a neutron-neutron collision. Two somewhat different possibilities for the distribution
of the ten charges in rapidity space are shown in the sketch on
the following page.
A generalized charge-transfer variable Z(y)
can be defined such that for a positive charge located at some
particular y, Z(y) increases by
one unit at that y value, and for
a negative charge, Z(y) decreases
by one unit at the position of
that charge. The step function
Z(y) for a complete event is
known as a zone graph. Z=O
separates different zones.
In
general terms, the LCQM hypothesis
requires that for large s-values
the internal structure of zones
(e.g., mean zone length, number
of charges per zone) become sindependent. Furthermore, the
properties of zones must be such
that it is improbable to have a
large number of particles per
zone or a zone which is long in
rapidity space. Finally, correlations between pairs of zones
must decrease quickly with increasing separation in y. Thus
stated, LCQM implies that the
mean number of zones will grow
with energy as 5/,n s. These conditions are fulfilled in a qualitative way for data in the
Fermilab energy regime. Figure
40 (from Bromberg et aI, ref.[53]),
0.1
displays the correlation function
D(Yl'Y2)=<Z(Yl)Z(Y2 <Z(Yl<Z(Y2 for 102 GeV/c and
400 GeV/c pp data as a function
of ~y=Y2-Yl for Yl=O and Yl=-1.2.
The curves were obtained by recalculating D(Yl'Y2) after ranI-II
domly reassigning the charges to
Figure 39: Cross sections for
inclusive reactions involving the exchange of
quantum numbers.

T.FERBEL

598

+1
q

-1

+1

Z(y) a

Z(y) a

-4

-2

-1

4
2
-2

-4
Two Possible Zone Graphs

the observed particle tracks. Hence the difference between the real
and the randomized-charges data
can be attributed to a mechanism
which prevents large fluctuation in charge transfer. In particular,
the small value of D(Yl'Y2) reflects a tendency for zone multiplicities to be smaller than for the case of a random distribution of
charge, and the rapid fall of D(y l , Y2) for large ~Y, characterized
by a correlation length of ~1.2 units in rapidity, shows a tendency
for real zones to be typically shorter than in the randomized data,
all consistent with expectations from the LCQN hypothesis.
The preliminary results on local compensation of transverse
momentum [54] have been used to extract a lower bound on the value
of the slope of the Pomeranchukon trajectory. The result is somewhat model dependent in its treatment of unobserved neutrals in
the final state, but provides a stringent limit of ~~(o)~O.2 GeV- 2,
which may be compared with phenomenological values of a~(o)~O.25
GeV- 2 The fascinating implication of this result is that the
dynamical mechanism responsible for the shrinkage of the elastic
diffraction peak is the local compensation of transverse momentum [54].
Off-Shell Inclusive Scattering. We are all familiar with
particle-exchange processes and with inclusive-production reactions. In the past few years interesting data have been presented which indicate that inclusive reactions initiated with offshell particles bear a great similarity to real inclusive processes.

599

HADRON PHYSICS AT FERMILAB

(0)

102 GeVlc

. ..
.
..
..

~'A'

1.0

0.5

1.0
0.5

y.:-1.2

0.1

i:
"0-

2.0

400 GeVlc

(b)

YI"O

1.0
0.5

1.0
0.5

YJ'"-1.2

0.1

-3

-2

-I

6y=y.-y,

Figure 40: Zone correlation function D(Yl'Y2) for pp data at 102

GeV/c and 400 GeV/c, evaluated at Yl=O and Yl=-1.2, as a
function of ~Y=Y2-Yl (y. are rapidities in the center of
mass). Curves are the lresults of randomizing the charge
assignments in each event.
As an example of these sort of studies, I will discuss production of ~++(1236) in Pp reactions. This process might be expected to proceed through pion exchange, however, this does not
appear to be the case. In Fig. 41 I display the rr+p mass distribution for the reaction pp~+p+Anything at 102 GeV/c and at 400
GeV/c [55]. A clear, energy independent peak at the ~++(1236)
resonance is apparent in the data. The t-spectrurn of the events
in the ~ region is shown in Fig. 4l(b). In Fig. 4l(c) I show the
x distribution of the rr+p events in the ~ mass region. The cross
section is observed to scale with s. The curve on the data is
from a calculation by R. Field. To fit the x spectrum, Field had
to introduce a large off-shell damping term (form factor in t)
for the pion-exchange contribution to the ~++ process. This form
factor was not required for the pion-exchange term in the

600

T. FERBEL

pp+p+Anything reaction. In addition to pion exchange, a comparable

contribution from p and A2 exchange was required to fit the x distribution. Although it does not appear that pion exchange dominates
the production, I will nevertheless compare the characteristics
of the XO system accompanying the production of ~++ to n-p scattering at a center of mass energy IS = M

Xo

Figure 42 demonstrates that the total charged-particle multiplicity (including "elastic" two-body -rr-p final states) of the XO
system has the same dependence on M2 as real n-p data have on s.
The f2 moment n2>_<n>2_<n for the XO system produced in association with the ~++ also has the same dependence on M2 as real n-p
data have on s. Taking this comparison one step further, in Fig.
42(c) I display the x distribution for the inclusive reaction
n-p+n-+Anything (smooth curve) and the analogous off-shell R-p+n+Anything data for several regions of M2. The distributions in
Fig. 42 (c) have been normalized in such XOa way that the integral
of the data over the invariant phase space yields the values of

...

pp-{1T+p)+..
(t<0.6 GtV2 )

>

-102 G,V/c

400 G.Vlc

10

102 8eVlc

o 400 "VIc

"'x

(0)

I... ~

j,
... 2;

1.0

1.8

1.1

;: ~

,,:t~j
2

4.

10

20

40 to

100

"'x(GeV}

..~

4
102 G,Vlc

::,<3

Ie)

bit 2

t<J
~~~

o "

0L--L--OL.2--L--0.L4--L-~0.6

-tl!. (GoV)

-I!

:;;

(e)

;<J

.~I(

. I'

0.1

!~.

L-____L-____L-__~
~ -1.0
-0.9
-0.8
-0.7

-to

0.0

1.0

Xl!.

Figure 41: Characteristics of

~++(1236) production in
pp collisions at
102 GeV/c and 400 GeV/c.

Figure 42: Comparison of offshell Reggeon-proton

(R--p) scattering with
real n-p production data.

HADRON PHYSICS AT FERMILAB

<n>TOT given in Fig. 42(a).

601

(The definition of x~_ for R-p

scattering in such that positive x corresponds to ~- emission along

the R- "direction", where the usual Gottfried-Jackson t-channel
direction is chosen as the R-p collision axis.) The similarity
between the reaction ~-p~-+Anything and the off-shell process
R-p~-+Anything at comparable s values is remarkable.
This agreement is particularly puzzling considering the fits that Field made
to the ~ production reaction. It appears therefore that the mass
and charge of a hadronic system have the greatest bearing on the
dynamics. The fact that R- has admixtures of objects having different spin (but same isotopic spin) does not grossly affect the
inclusive spectra.
A similar study to that discussed in Figs. 41 and 42 was also
performed for the reaction pp~+Anything [56]. Here off-shell
ROp~-+Anything was compared to ~-+Anything at values of s ~M2.
The inclusive spectra are shown in Fig. 43. The agreement iSIP
again, impressive but quite surprising in view of the fact that
for low values of M2 the R object should mainly have properties
characterizing the Pomeranchuk trajectory and certainly not a
photon. Consequently, here again we see that gross features of
inclusive production appear to be independent of all dynamic
qualities, except for the charge and effective mass of the interacting system.
"If there is a question of importance, look to the masses"
(Mao Tse Tung). This last part of my lecture can be summarized by
the above quotation from Mao's Little Red Book, as paraphrased to
me by Ed Berger. We have made a study of the PT-dependence and
the x-dependence of multiparticle systems produced in pp collisions
[57]. Specifically, we have examined how the mean value, <Pr>'
and the full width at half maximum of the x distribution (r ) vary
with the mass, charge, and particle multiplicity of a produ~ed
system. The results of this investigation are summarized in Figs.
44 through 47. There is essentially no dependence on the charge
of the system being examined (particularly for <PT' At fixed
mass, a weak variation with multiplicity can be
discerned in the
<PT> data. A somewhat stronger variation is observed for r. The
<PT> dependence is essentially s-independent (between 102 G~V/c
and 400 GeV/c) while the r values become smaller with increasing
s (presumably due to approiimate Feynman scaling). The variation
of r and <PT> with mass is very similar to that observed for
stable hadrons (e.g., ~,K,A etc.). In detail, the 3~ systems
appear to coincide in character with the properties of the longlived objects [58]. The latter result may be due to the fact that
clusters, which have typical multiplicities of ~3, have production
properties akin to the more stable particles. (This is not an
entirely consistent picture because of the charge-independence of

602

T. FERBEL

1.00
0.50

N~

0.
~

0.20
0.10

Nb/No.""

~
)(

*1*

IoJ 0.0

'-:..... 0 01

1b~

-1.0

0.0

1.0

X
Figure 43: Comparison of the off-shell reaction RO+p+rr-+Anything
with the inclusive data for y+p+rr-+Anything at s~M2.

the result.) The rise in <PT> with multiplicity at large masses

can be understood, qualitatively, as due to a damped random-walk
process. The unusual reversal at small masses may be due to a
Bose effect (but here again there is inconsistency because of the
lack of dependence of <PT> on charge). The arrows labeled M.C.
are the expected values of <PT> (essentially independent of M)
which our Monte Carlo model provides for the 400 GeV/c data.
(This model contains proper single-particle distributions and
multiplicities observed at 400 GeV/c but no explicit correlations
among particles.) It is interesting that although the Monte Carlo
predictions for <PT> are completely incorrect, the predictions for
r (not shown) appear to be in far better accord with the data.
If is not clear to me whether this new result is important or
whether it is just another way to study particle correlations.
In any event, the mass of the produced system clearly has the
largest effect on the dynamic properties of the hadron bundle,
hence one should "look to the masses."
I thank J. Whitmore for providing me with an early version of
his update of "Multiparticle Studies Utilizing Ferrnilab Bubble
Chambers". In addition, I wish to thank E. L. Berger, D. Duke,
P. Slattery, and C. Quigg for helpful discussions. Several of the
topics in these lectures could not have been prepared without the
excellent assistance I have received from our students J. Biel and

603

HADRON PHYSICS AT FERMILAB

P. Stix. Finally, I wish to express my gratitude to D. Weingarten

for a critical reading of the manuscript and for providing extensive suggestions as to content.

j .-._._._._._._.-

1.0

/0

.....---

_--------

".r- .................... (2.".)

'" 1(.

1C'J.

.~

'\... 0.5
Q.

..

.K

.... _.(M.C.)

(6.".)

........

:t

- - - pp- (3.".) + Anything

400 GeV

(GeV)

M
Figure 44: Comparison of <PT> for the production of long-lived
particles in pp collislons with 400 GeV/c data (smoothed
curves) for the reactions pp+(2~)o+Anything, pp+(3~) +
Anything, and pp+(6~)o+Anything. The values of <p > for
the produced multipion systems are displayed as a function
of the masses of the systems. (Only charged particles were
used for the calculations.)
~

1.0

~ 0.8
(!)

"c:t" 0.6
v

0.4

-- -

OJ(

-~-f\IM-~.x.a-~- - - - -

4
0 4
4
D 4

it"
,

lC

0.2

-(M.C.)
Pions,Qa+2
Pions, Qa-2
Pions, Qa+4
Pions, Q a - 4

4 Pions , Q a 0
(400 GeVle)

MASS (GeV)
Figure 45: The value of <PT> for a (4~) system produced in the reaction pp+(4~)+Anything at 400 GeV/c. This figure displays
the variation of <p > with the mass and the charge (Q) of
the (4~) system. The dashed line represents the variation of
<PI> with mass, expected on the basis of a Monte-Carlo calcu ation. (There is only a very weak dependence on charge
in the Monte Carlo.)

T. FERBEL

604

0.6

0.2
1T

MASS (GeV)
Figure 46: Comparison of the full width at half maximum (fx) of
the x-distribution (da/dx, integrated over p ) for pions and
K~ produced in pp collisions at 102 GeV/c with data (smoothed)
for the reactions pp+(2TI)o+Anything and pp+(4TI)o+Anything,
also at 102 GeV/c. The widths of the x-spectra are plotted
as a function of the masses of the produced multipion
systems.

0.6

0.5

r)(

0.4

.x

0.3

400 GeV/c
4 Pions, Q = + 2

0.2

o 4 Pions, Q = - 2
x 4 Pions, Q = 0

0.1

MASS (GeV)
Figure 47: The dependence of f on the charge of a produced
(4TI) system in pp cOllisiofls at 400 GeV/c.

HADRON PHYSICS AT FERMILAB

605

REFERENCES
1. J.J. Aubert et aI, Phys. Rev. Lett. 33, 1404 (1974); J.E. Augustin et aI, Phys. Rev. Lett. ~, 1406--(1974).
2. A. Carrol et aI, Phys. Lett. 6lB, 303 (1976). Also, private
communication from W. Baker of the BNL-FNAL-Rockefeller
Collaboration.
3. Private communication from H. Kobrak of the ChicagO-San DiegoWisconsin Collaboration.
4. A.V. Barnes et aI, Phys. Rev. Letts. 37, 76 (1976). See also
D.L Dahl et aI, Phys. Rev. Letts. 37:-80 (1976) for a study
of 1T-P + nn.
5. J. Whitmore in Particles and Fields APS/DPF Meeting in Seattle
(1975). H. Lubatti and P. Mockett eds. See also the comprehensive review of bubble chamber data by J. Whitmore, MSU
Report (Feb. 1976) to appear in Physics Reports.
6. Y. Eylon and H. Harari, Nuclear Phys. B80, 349 (1974).
7. D. Ayres et al (FSAS Group), Phys. Rev. Lett. 35, 1195 (1975);
see also the report of D. Cutts which appears in these proceedings.
I have included the results of the Michigan-ANLFNAL-Indiana Group in Fig. 5; see C. Akerlof et aI, Phys. Rev.
Lett. 35, 1406 (1975).
8. I. Ambats et aI, Phys. Rev. D9, 1179 (1974); G. Brandenburg
et aI, Phys. Lett. 58B, 367 (1975).
9. R.L. Anderson et al (FSAS Group) Fermilab Report Pub-76/47EXP/ANL-HEP-PR-76-28.
10. M. Davies and H. Harari, Phys. Lett. 35B, 239 (1971).
also R. Carnegie, Phys. Lett. 58B, 37r-(1975).

See

11. Private communication from J. Butler (FSAS Group). See also

results of Akerlof et al given in ref. (7). The latter
measurements indicate that the ratios 0elastic to 0 0 1
are 0.15/0.14/0.20 for 1Tp/Kp/pp channels respectlv~ty. The
small systematic discrepancies between the two sets of data
have been commented on previously by J. Lach, Fermilab
Report Conf -76/l5-EXP (1976).
12. V. Barger and R. Phillips, Phys. Lett. 60B, 358 (1976).

T.FERBEL

606

13. A. Buras and J. Dias de Deus, Nucl. Phys. B7l, 481 (1974).
14. See reference 13 for a comparison of pp data, and J. Butler,
Ph.D. Thesis, MIT (1975) for a comparison of K+p data.
15. C. Akerlof et aI, Phys. Lett. 59B, 197 (1975). For the initial
prediction of Chou and Yang see Phys. Rev. Letts. 20, 1213
(1968).
16. C. Ankenbrandt et aI, FNAL-Yale Group, Fermilab-Conf-75/6l-EXP
(1975). See also J. Lach in reference (11).
17. R. Hendrick and B. Lautrup, Phys. Rev. 011,529 (1975).
18. V. Bartenev et aI, Phys. Rev. Lett.

il, 1367 (1973).

19. V.N. Bolotov et aI, Nucl. Phys. B73, 365 (1974). See the
lectures of E. Leader for a comprehensive discussion of
charge-exchange data and their phenomenological implications.
20. H.R. Barton et aI, OSU-MSU-Carleton-Alberta Collaboration,
Preprint COO-1545-l96. I thank N.W. Reay for a discussion
of the data.
21. J. Biel, Ph.D. Thesis, University of Rochester (1976).
22. J. Biel, et aI, Phys. Rev. Letters 36, 507 (1976); T. Ferbel,
Proc. of Int'l School of SUbnuclearlPhysics at Erice (1975),
A. Zichichi ed.
23. This has been emphasized by V. Chaloupka in a preprint discussing similar data from SLAC.
24. A similar effect has been reported by a group at the ISR
measuring the reaction pp + nn+p. See H. de Kerret et al
CERN preprint (1976).
25. A. Minaka et aI, Preprint KYUSHU-76-HE-6 (1976).
communication from E. Berger.

Also, private

26. For a review of the phenomenology and earlier data see

H. B~ggild and T. Ferbel, Ann. Rev. Nuc. Sci ~, 451 (1974).
27. F. Winkelmann, Phys. Lett. 48B, 273 (1974).
28. R. Field and G. Fox, Nucl. Phys. B80, 367 (1974).
29. R.L. Anderson et aI, FSAS Preprint (1976). I thank D. Ritson
for providing me with this paper prior to general circulation.

HADRON PHYSICS AT FERMILAB

607

30. M. Sogard, private communication from the FSAS Group.

also the paper of o. Cutts in these proceedings.

See

th
31. The data are from A. Gurtu et aI, Proc. of V Int'l. Symp.
on Many-Particle Hadrodynamics, Eisenach-Leipzig (1974),
G. Ranft and J. Ranft, eds. and W. Busza et aI, MIT/Carleton/
FNAL preprint (1976). See also W. Busza, VIth Int'l Conf. on
Nuclear and Particle Physics, Los Alamos (1975). I wish to
thank J. Elias for a helpful discussion of the Fermilab data.
32. See the review of L. Bertocchi, VI th Int'l Conf. on Nuclear
and Particle Physics, Los Alamos (1975).
33. K. Gottfried, Phys. Rev. Letters

~,

957 (1974).

34. For an alternate description of multiparticle production on

nuclear targets see the report of G. Berlad in these proceedings.
35. R.E. Ansorge et aI, Phys. Letts. 59B, 299 (1975); J.G. Rushbrooke et aI, Phys. Letts. 59B, 3~(1975).
36. The figure is from reference (5). The pn data are from
A. Sheng et aI, Phys. Rev. 012, 1219 (1975), and the pn data
are from A. Fridman et aI, Phys. Rev. 012, 3414 (1975).
37. J. Elliott et aI, Phys. Rev. Letters 34, 607 (1975).
38. Z. Koba, H. Nielsen and P. Olesen, Nucl. Phys. B40. 317 (1972).
39. T. Ferbel, Phys. Rev. 07, 925 (1973).
40. I wish to thank J. Whitmore for providing me with as yet
unpublished results from the latest exposures at the 30-inch
bubble chamber.
41. K. Moffeit et aI, Phys. Rev. OS, 1603 (1972).
42. C. Bromberg et aI, Nucl. Phys. BlD7, 82 (1976).
43. This effect has been discussed previously in R. Schindler
et aI, Phys. Rev. Letters 33, 862 (1974).
44. Reference (5) and P. Bosetti et aI, Nucl. Phys. B54, 141 (1973).
45. T. Ferbel, Phys. Rev. 08, 2321 (1973).
46. C. Quigg and E. Rabinovici, Phys. Rev. 013, 2525 (1976). The
parameterization used by Whitmore for the cross sections are:

608

T.FERBEL

(15.18 mb)p~A~755, a(KN) = (13.09 mb)p~A~755,

(25.33 mb)p~.0755
LAB
47. It is uncertain, for example, how the latest parameterization
of the Pomeranchukon contribution to total cross sections by
Quigg and Rabinovici affects the asymptotic form of the
theoretical expression for <n c >. Preliminary indications are
that <nc> will rise more rapidly than that indicated by the
B tn s rise in Table I.

a (TIN)
a (NN)

48. J. Whitmore et aI, Phys. Lett. 60B, 211 (1976).

49. Chan, H-M., Hsue, C.S., Quigg, C. and Wang J-M, Phys. Rev.
Letts. 26, 672 (1971).
50. E.W. Beier et aI, University of Pennsylvania preprint,
submitted to the Tbilisi Meeting.
51. R.L. Anderson et aI, Inclusive Hadron Exchange Scattering at
50-175 GeV. I thank D. Ritson for providing me an early
version of the preprint.
52. A. Krzywicki and D. Weingarten, Phys. Lett. SOB, 265 (1974).
53. J. Derre et aI, French-Soviet Collaboration, preprint M-12
(1974); C. Bromberg et aI, Phys. Rev. D12, 1224 (1975);
D. Fong et aI, Phys. Lett. 6lB, 99 (1976); J. Lamsa et aI,
Phys. Rev. Lett. ~, 73 (1976).
54. D. Weingarten et aI, Rochester-ISU-MSU-ANL-FNAL-Maryland
Preprint (1976).
55. J.P. DeBrion et aI, Phys. Rev. Letts. 34,910 (1975).
56. J.P. DeBrion et aI, Phys. Letters. 52B, 477 (1974).
57. P. Stix et aI, Rochester Report (in preparation).
58. See the lecture of A.J.S. Smith in this volume for an interesting comparison of the latest results for <Pt> of ~+~
pairs produced in hadronic collisions as a function of the
mass of the ~+~- system. It appears that <Pt> for ~+~
pairs is larger than for produced hadrons of equivalent
invariant mass.

609

HADRON PHYSICS AT FERMILAB

DIS C U S S ION
Prof. T. Ferbel

CHAIRMAN:

Scientific Secretaries:

A. Degre and K.L. Wernhard

DISCUSSION
LEADER:

How can the parameter C

plicity be negative?

~n

your formula for the average multi-

FERBEL:
C has to be negative.
in the following way:

To show this, we can rewrite the formula

A + B In sll +

(n)

.'

r. / B

~'1/4 J

We assume that (n) is proportional to the plateau in the du/dy distribution. The width of the plateau rises proportional to In s. For
s + 00 the height of the plateau is constant. Therefore
(n)

B In s,

for s

00

but for lower energies, we have to correct for the increase in height
with energy. This means that (n) reaches A + B In s from the bottom;
therefore, C < o.
WIGNER:
~oes the Al show up
co llisions?

~n

a variety of reactions or only

~n

TIp

T.FERBEL

610

FERBEL:
To summarize the situation about the AI, one can say the
following: Initially, the Al was found in a Tip experiment as a
peak around 1.15 GeV in the 3TI effective mass distribution.
Secondly, it is also produced coherently on nuclei. Thirdly, the
Al shows a (TIp) substructure. Fourthly, a spin-parity analysis
taking into account the production and the decay of the 3TI-system
Ascoli-analysis -- shows that 0-, 1+ and 2- waves are present, 1+
dominant, but the phase-shift does not go through 90 at the mass
of the AI.

RANFT:
A double Regge model describes the mass distribution in the
3TI (p,TI) system in the Al range. It gives an s-wave but without the
circle in the Argand diagram, thus the Al should not -- at least in
this global manner, with respect to mass distributions -- be called
a resonance.

BUCCELLA:
To find the Al in PTI phase shift analysis is more difficult
than to find the A2. In fact, while JP conservation allows only
D-wave for the A2 , s-wave is also possible for the AI. So, one has
a two-channel problem, i.e., to diagonalize a 2 x 2 matrix.

MARCIANO:
Caldi and Pagels have shown that SU(3) x SU(3) chiral symmetry
does not require the existence of the Al meson. In their scheme,
the B meson, a well-established resonance, is the chiral partner
of the p.

ETIM:
What happens to Weinberg's sum rules if you throwaway the
Al meson?

LIPKIN:
SU(3) does not need the Al or the B, but if the Q states are
there, SU(3) requires the rest of the octet to be there. Chiral
symmetry requires an axial vector meson, but perhaps this could be
the B as ~1arciano suggested. The quark model needs both the Al and
B which are the 3 PI and IpI states.
lIm.rever, if the Al is not there,
the quarkists can probably find a spin-orbit interaction to get rid
of it.

A REVIEW OF THE ISR RESULTS

Giovanni Valenti
Istituto Nazionale di Fisica Nucleare
v. Irnerio 46, Bologna, Italy
I.

INTRODUCTION

Operation of the CERN Intersecting Storage Rings (ISR)

has provided a large variety of new results and insight in the
study of high energy proton-proton interaction.
This lecture
will consist of a review of recent results on total cross-section,
elastic scattering, large transverse momentum phenomena and direct
lepton production.
II.

TOTAL CROSS-SECTION

New measurements of the proton-proton total cross-section

have been performed l
with a precision of + 0.6% over the energy
range 23.5 < ~ < 6207 GeV.
Two different-techniques were
implemented by the two groups, CEP~-RO}m (CR), PISA-STONY BROOK
(PBS): the PBS group derived the cross-section from

where L is the Luminosity measured using the Van der Meer method 2,
and ~OT is the interaction rate recorded on a nearly 4TI scintillation aetector (Fig. 1).
To determi~e ~OT from R b
d' the following corrections
have to be cons1dered:
0 serve

611

612

G. VALENTI

i)

loss of acceptance due to dead regions

ii) elastic events at very small angles escaping into the

beam pipe.
This can be expressed by:
Robserved (l+e:)
L
+ 11 a el
where e: accounts for trigger losses and l1a el corrects for losses
of elastic events.
The CR group used the differential elastic cross-section at
zero momentum transfer (t=O) to compute the total cross-section
via the optical theorem

is the ratio of the real to the imaginary part of the forward

elastic scattering amplitude, which at ISR energies is small and
positive (Fig. 2)3.

The forward elastic differential cross-section is given

by:
(1 - e:

where

do
duJ

coul

) eb / t /

R
= rate of elastic scattering,
el

I1w = solid angle of the apparatus, e:

1 take into account coulomb
scattering effects important at smallo~.
From this one obtains:
( 1 - e:

coul

) eb1tl

-=ReI
I1w L

(_) In determining aTOT ' the necessary values of P and b were input
data.
The CR group has also performed a coulomb nuclear
interference measurement which should allow a simultaneous
determination of aTOT ' b, p.

A REVIEW OF THE ISR RESULTS

613

Fig. 1 ISR intersection 1-8. General layout of the Pisa-Stony

Brook large solid-angle secondary-particle detector, and the CERNRome hodoscope measuring elastic scattering in the vertical plane.

10

100

1000

Fig. 2 Ratio of the real to the imaginary part of the forward

amplitude plotted versus the laboratory momentum.

41.45 0.26
42.38 0.29
43.05 0.33

41.92 0.25

42.73 0.34

43.02 0.40

44.7

52.8

62.7

0.22

40.35 0.34

40.08 0.24

30.6

0.36

39.01 0.29

38.80 0.25

23.5

Scale
error

o(eR)
(mb)

O(PSB)
(mb)

( GeV)

,;,

0.25

43.13 0.65

41.99 0.54

41.01 0.50

40.58 0.68

39.22 0.54

O(L-ind)
(mb)

0.28

43.04 0.31

42.50 0.27

41.70 0.21

40.16 0.22

38.88 0.21

o(weighted average)
(mb)

Total proton-proton cross-sections as a function of the total c.m.

energy, IS. ~he weighted averages of the last column are plotted in fig. 2.

TABLE I

:::!

<

G)

.,..~

615

A REVIEW OF THE ISR RESULTS

Amaldi et at
(1973)
Amendolia et of. (1973)
v Bellettini
(1973)
o Eggert et at
(1976)

44

0.

43
..c

42

41

40
39

50

70

VS/GeV

41

40

39

-----.---- . ....--;- -~---

/:

/.,

--~//

~.".,.,....

/
/

+/

/,1..

T/

/.J,. /

,//+

/1

44

43

42

T/

..c
41

40

This experiment (1976)

39

Carroll et at (1976)

38---

38

10

20

50

100

VS/GeV

200

30
I

Elal>/GeV

500

50
I

1000

70
I

2000

100
I

5000

Fig. 3 The proton-proton total cross-section measurements obtained

by the PBS-CR groups are plotted together with those from FNAL
(Ref. 4). The broken lines represent in both cases the extended scale
error. In the upper left corner, on a reduced scale, the results of
all previous measurements at the ISR are shown (Ref. 5).

The two measurements were performed at the same time in the

same interaction region of the ISRj this ensures that the
luminosity is the same in both experiments so that an absolute
(luminosity independent) measurement of the total cross-section
could be obtained.
Table I gives a comparison of the measurements obtained as well as weighted averages.
The total crosssection compilation shown in Fig. 3 is in the range of energies
above NAL up to 62.7 GeV in the c.m.s. and represent our best
knowledge so far.

616

G. VALENTI

III.

LARGE ANGLE ELASTIC SCATTERING

One of the key features of proton-proton elastic scattering

is the dip in the differential cross-section da/dt occurring at a
It I value of 1.4 (GeV/c)2.
The existence and the characteristics
of such a dip are important in the attempt to understand the shape
of the proton.
A detailed study of the dip position as a function
of the total center-of-mass energy, I~, has been done by the CHOV
collaboration (CERN-HAMBURG-ORSAY-VIENNA) at the Split Field
Magnet 6
The following features have emerged:
i)

the

~IN

decreases with increasing s,

tMIN

(is"

23 GeV)

- (1.44

tMIN

(is"

62 GeV)

- (1.26 ..! 0.03) (GeV/c)2

0.02) (GeV/c)2

The slope parameter b seems to behave (as a function of

1;) differently in different momentum transfer regions,
that is:
ii)

between t = - 0.25 (GeV/c)2 to t = - 0.6 (GeV/c)2

b = (10.3 ..! 0.2) (GeV/c) 2
between t = - 0.6 (GeV/c)2 to t = - 1.1 (GeV/c)2

(I;

23 GeV)

(9.1 ..! 0.1) (GeV/c) 2

b (I;

62 GeV)

(11.8 ~ 0.3) (GeV/c) 2

iii) the peak value observed after the dip increases with s

iv)

I; =

23 GeV

45 + 5 nbarns

/;=

62 GeV

73 + 10 nbarns

for t < -2.0 (GeV/ c)2

da
dt

does not depend on

I;.

These features are shown in Fig. 4.

Fig. 5 gives a compilation
for various values of I;.
Notice that the ratio

(I;

23 GeV)

td.1p (Is

= 62 GeV)

td.

1p

1.14

is in agreement with geometrical scaling; it should in fact be

proportional, as a 0 and a. 1 to the square of the radius of the
T T
1ne
proton.

A REVIEW OF THE ISR RESULTS

0.1

617

~_,......----r----T""--'

pp .. pp

o VS

VS

23 GeV
62 GeV

10
de'

(it

Fig. 4 Differential cross-section for momentum transfer -t > 0.6

GeV 2 The t-resolution in the vicinity of the minimum is O.013 GeV
at 23 GeV and 0.028 GeV 2 at 63 GeV. The solid lines represent the
best fit result of a two-amplitude model (Ref. 8).

618

G. VALENTI

da
The same group has extended --measurements
up to -t
dt
at a center-of-mass energy;;
emerge:
i)

= 53 GeV

= 9 GeV 2

Two features

(Figo 6)7.

the absence of a second dip expected by the optical model

ii)

the shape of

~~ above ~ 2.5 GeV2 agrees with the fourth

power G(t)4 of the nuclear electromagnetic form factor.

IV.

LARGE TRANSVERSE MOMENTUM PHENOMEMA

The discovery of two "PT regimes" in high energy proton-proton

interactions is one of the ma~n accomplishements of the CERN ISR.
The low PT reactions (PT < 0.35 GeV/c) are characterized by
a striking energy independent exponential fall-off of all single
particle inclusive cross-sections with an overwhelming pion
production 9 In large PT configurations. PT > 2.0 GeV/c. the
exponential cut-off no longer holds; there ~s a strong energy
10
10. 1

>

3
c.:> 10
......
Q)

.D

-....J2

-'I

~~

E 10

"0

105

"0

106

......
b

t~

107

100

10 8
0

(x

10-1)

(x

1( 2 )
5

I t I (GeV 2 )

Fig. 5

Compilation of proton-proton differential cross-sections.

A REVIEW OF THE ISR RESULTS

619

10-'

1.0-'

10-'

pp- pp

.'....,..... .
'

'"

'

V5=53 GeV

....

....

",

",

1O- 'OO
!c-----'---!:-2--'----'4--'---6!c-----'---!:-S---'----=1'::-lO
-I [GeV']

Fig. 6 Differential cross-section as a function of t. The error bars

represent statistical errors and bin-to-bin systematical errors.
An overall scale uncertainty of S% has to be added.

dependence in the production cross-sections, and heavier particles

are more and more abundantly produced.
The growing interest in high PT physics relies on the possibility of studying the small transverse distance structure of the
proton. Here I will discuss recent ISR results only; more complete
reviews of earlier experiments can be found in the literature 10

620

G. VALENTI

Single particle inclusive distributions

Invariant cross-sections are observed to vary strongly with
transverse momentum; data are usually parametrized according to
E

do

dt

-n

= PT

f(xT,e)

where x T= 2PT/i; , i; and e are the total energy and production

angle in the center-of-mass system (c.m.s.). Typical sets of data
for ~o, n- inclusive single particle distributions obtained by the
CERN-Columbia-Rockefeller-Saclay (CCRS) Collaboration 11 are shown
in Figs. 7, 8, 9. In Fig. 10 nO inclusive spectra at various
center-of-mass energies are compared. Finally, Fig. 11 shows data
obtained by the Aachen-CERN-Heidelberg-Munich (ACHM) Collaborati~n12
and by the British-Scandinavian (BS) Collaboration 13 on nO and nproduction respectively. Table II summarizes the values of the
exponent n obtained by fits to these data.

The behaviour of f(xT,e) is known at the ISR up to

90 0 only; data are consistent with

0.35 and

TABLE II
Single pion inclusive data
Data
CCRS

CCRS

ACHM

Range of Is

Range of P-r
0

>

2.5 GeV/c

>

3.3 GeV/c

>

2.0 GeV/c

2
2

23.5 to 62.4 GeV

7.2 0.1

44.8 to 62.4 GeV

7.8 0.2

23.6 to 62.9 GeV

7.2 0.2

A REVIEW OF THE ISR RESULTS

621

=44.S GeV
It"

10- 31

u >
Mr
E

'"

10-32

III

<!)

M
IM
'0

Co

'0

10-33

1LI

10-34

Fig. 7 Invariant cross-section for n , n , and nO production at

~ = 90 and ;; = 44.8 GeV.

G. VALENTI

622

\IS =52.7 GeV

J!l.

10- 31

ItIt

Til

I>
I>

~I>

~NU IN>41
E

c>

bl

...

"Q

M0..

"Q

10-34

\~
t t

t tf

10-36

L.........----1___---'---....L....--~--.l...--......,.J_=__----l
~5
9S

Fig. 8 Invariant cross-section for

e = 90 0 and ;; = 52.7 GeV.

TIT,

TI-,

and

TI o

production at

A REVIEW OF THE ISR RESULTS

623

VS= 62.4 GeV

It-

It

ltD

A
A

10-31

~UIN>
E
UC)"

bl~Co

...

"0"0

A~A

'\

'*t#ttt t

.;

t
Pr GeVic
Fig. 9

e=

Invariant cross-section for Tt+, Tt-, and Tt production at

= 62.4 GeV.

90 and ;;

G. VALENTI

624

10- 29

10- 30

P+ P _

...!.
..
+f~.".....

.
0

..

o. :d.
to . .
+to . ;
...

+~

10-31

tl

\IS

It + anything

=23.5
30.6
44.8

GeV

52.7

62.4

,
I

o .u
.6

tt~

~IN~

Cl

..Pr,c.

10- 32

~"tJ

IaJ

. ...!

.. !.

',+

Int:'I ttill!
I

ttit i;ltij\ III

10- 33

fijj IH ~

10- 34

PT

G,Nlc

Fig. 10 Invariant cross-section for TI o production at 90, for five

center-of-mass energies. The extrapolation from data with PT < 1
GeV/c is shown for comparison.

A REVIEW OF THE ISR RESULTS

625

10'

10

10-'

p.p _

,,0 ...

e ,,90
.. e. 53

10-2

:::u
>II
C>

.0

...!

..,
..,

10-3

10-4

Q.

. ..D

10-5

lIJ

10"'

10"'

to"
10-t

.,..,
lO-n

VI 23.64 G.V
(x tOO')

Pr (G.V1c)

Fig. 11 Inclusive invariant cross-section for TI o production for

various values of e and I;. The solid lines indicate the data of the
CERN-Columbia-Rockefeller-Saclay Collaboration (Ref. 11) and the
dashed lines those of the British-Scandinavian Collaboration
(Ref. 13).

G. VALENTI

626

with

= (14.8

0.6) 10- 27

and B

= 12.6

0.2.
m

A fit of the type f(x ,e) T C (1 - x)

can also represent
the data; in this case the Best fit valuesTare:

C = (12.9 0.2) 10- 27 , m = 10.5 0.2.

A compilation of ISR results on the behaviour of f(x ,e) as a
function of x for e = 90 is shown in Fig. 12. The beha~iour of
f(xT,e) for x~ > 0.35 is derived by data obtained by the ChicagoPrinceton group14 at Fermilab. There, pions produced in proton
nucleus collisions, show a dependence of n on x T as shown in Fig. 13.
Information on the e dependence of the invariant cross-section
can be derived from the ACHl1 measurements of 'ITo spectra at e = 53
and e = 90 shown in Fig. 11. No e dependence at fixed p and;;
is found. These results seem to rule out the hypothesis 5f scaling
in terms of the radial variable x R = xT/sine.
Heavy particle production as a function of p increases with
respect to pions at larger PT' This is illustrat~d by the BS data
in Fig. 14.
The ACHl1 and CCRS Collaborations have also observed n production via its yy decayll 15 16
in a range of transverse momentum
between 3 and 5 GeV/c (Figs. 15, 16). The PT dependence of the
production cross-section is the same as that of 'ITo production, the
n to 'ITo cross-section ratio is approximately 1/2 independent of ;;.
Particle correlations in high p

final states

~~~~~--~~~~----~~-T~~~~~=

The basic motivation for studying particle correlations in

high p final states can be found in the hope that high transverse
momentfrm reactions result from hard scattering of hadron's constituents in a way similar to deep inelastic electron scattering.
The study of electron and muon scattering has demonstrated
that hadrons have an effective point-like constituents structure.
The quark parton model interpretation leads one to expect that
hadrons can scatter to large transverse momentum via hard, large
angle scattering processes involving their constituents. The final
state structure that should emerge from such a picture is characterized by the so called jets. One is brought to expect, in high
energy proton-proton interactions, two coplanar jets, collinearity
being spoiled by the internal mition of the hadron const~tuents.
Each hadron of the jet, moreover, should have a limited momentum

627

A REVIEW OF THE ISR RESULTS

10- 27

l'w

vv

SYMBOL

on

23.5
30.6
44.8
52.7
62.'
44.8
52.7
62.'
44.8
52.7
62.'

..

Q,

..

- ..
Q,

lO-n

10- 21

OS

0.10

ito
,,0

II
j1

Ii'

ito
,,0
ft'

,,'
It'

,,,,It-

0.15

0.20

0.25

0.40

0.35

0.30

liT

Fig. 12 Plot of the quantity p~.6 E(d~a/d3p) for

production as a function of x T = 2PT/1s.

"IT

"IT

and

"ITo

n
4

Fig, 13 Plot of the exponent "n" (see text) versus x T for inclusive
pion production at large ;;,

628

G. VALENTI

1.0

FrQction of all charged

rs:S28GeV

SlOb: 89 O

20

i
0

'"
t-

6.
~

0.1 t-t-

l-

t-

t-

9~

~t

At.;'

~'ff A

:~

-.

.
.

II

P
10

the p

(I

K
K-

.
-

n
n-

0.01

Fig. 14

pari icles

20

_L

30

PT

[GeVlcJ

40

dependence of the charged particle composition at

Is = 52.8 GeV,Te = 89.

629

A REVIEW OF THE ISR RESULTS

1000

62.4 GeV

1000

-E 1000
~

100
30.6 GeV

10

Myy MeV

Fig. 15 Invariant mass distribution

center-of-mass energies (CCRS).

of two-photon events at four

630

G. VALENTI

a)

b)

Fig. 16 Invariaht cross-section for inclusive-production of nO's

(Fig. l6a) and n's (Fig. l6b) at e = 90 and Is = 52.7 GeV. The
curve shown in Fig. l6a is the best fit to the nO data; the curve
of Fig. l6b is the best fit to the nO data, multiplied by 0.58.

631

A REVIEW OF THE ISR RESULTS

component perpendicular to the jet axis.

Following P. Darriulat 17 it is possible to separate, ~n high
PT reactions, three components:
i)

a component characteristic of a typical collision with

small transverse momentum having

I;eff <;; -

2 P

ii)

a component including the large transverse momentum

particle itself and few others close in angle to it

iii)

a component produced over a wide angular cone opposite

to the high PT particle.

The data of Fig. 17 from ACHH 12 show the features listed. In

particular one can observe:
a)

an enhancement in the same azimuthal region as the TI o

<i1 < 30) centered at the same pseudorapidity y as
the TI o ( 0.7 units).

b)

in a ~ sector 30 0 around = 180 0 the particle

density peaks at y = -y (TIO).

Two particle correlations are usually given in teDrrlS of the

ratio of the two-particle invariant cross-section integrated over
the TI o transverse momentum, to the TI o invariant cross-section
integrated over the same momentum range, plotted versus the particle
transverse momentum; that is
co

00

f
~Y 1

dY l

PTl

CCRS data on two particle correlations of Figs. 18, 19, 20

show a character analogous to the three-component picture previously
described.
New data obtained by a CERN group 15 at the Split Field Magnet
(SFM) are now available. Their study tries to disentangle the

632

G. VALENTI

0.60
040
>

l-

~
a:

020

Ii.

-----~
o

VI
I-

II>

0.40

::J

060

040

Ci

++
.+ ++

020

....

1200 <1op 1<ISOO

I-

UJ

>

a)

040

0.20

UJ

in

~
~

a:

0.40

040

0.20

020

tt

-3 -2 -1

060

++H

040

+
~

020

15O"<lop 1<180"

-3 -2 -I 0 I
y"-

23

0.60
040
020

::J
II>

Ci

~
~

UJ

040

VI

UJ

020

c
~

040

300<1 op 1< 600

b)

040

. - - - - .

0.20

-3 -2 -I

-3 -2 -I 0 I
y"-

Fig. 17 Charged particle densities from ACHM (ref. 12) with

PT(TIO) > 2 GeV/c; y~ = -In tg(8/2) is the pseudorapidity. The solid
l~nes show charged particle densities in minimum bias (~ inclusive)
triggers: a) 8 = 53; b) 8 0=90.
TI
TI

A REVIEW OF THE ISR RESULTS

633

101r-----r---~r_--_,~----r_----r_----~----~--~

\'5=44.8 GeV

rr > 3.3 GeVlc 6. ~ = 180

3 GeVlc 6.~:: 1800

Tt > 3 GeVlc 6,4>:: 00

... Tt>

o Fully inclusive

10- 2

Fig. 18 The function F (see text) plotted versus the associated

charged particle transverse momentum p , at Is = 44.8 GeV. The probability per inelastic interaction is Ilso shown.

634

G. VALENTI

VS = 52.7 GeV

n! > 3.3 GeVic

...

nO>

3 GeVic 6~= 1800

nO>

3 GeVic

6~=

6~::

1800

00

o Fully inclusive

+Y t

10-3

tt
10- 4

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

PT2 GeVic

Fig. 19 The function F (see text) plotted veLSus the associated

charged particle transverse momentum PT' at is = 52.7 GeV. The probability per inelastic interaction is also shown.

A REVIEW OF THE ISR RESULTS

635

VS

=62.4

GeV

elt1>3.3GeV/c ~4>=1800
.. ltD> 3 GeV Ic ~4> = 1800
b ltD> 3 GeV Ie ~4> ::0 0
o Fully inclusive

10- 2

10- 3

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

PT2 GeV/c
Fig. 20 The function F (see text) plotted veLsus the associated
charged particle transverse momentum p , at Is = 62.4 GeV. The probability per inelastic interaction is Ilso shown.

636

G. VALENTI

"jets" pattern by studying the secondaries produced in high PT

triggered nO events. The kinematical definition used by the authors
in describing the data is shown in Fig. 21.
Charged particles produced in the nO hemisphere
The average single particle distribution function integrated
over ~ in the region I~I < 27 and over p
for four bins of p
between 0.4 and 1.7 GeV/c for 90 data isxshown in Fig. 22. The x
relative minimum bias curve (dashed line) has errors of the order
of 10 ~ 20 %. The excess grows and shrinks in rapidity with increasing p ; for Iyl < 2, I~I < '27, (0.4 < p < 1.7) GeVic the integral
of theXexcess is:
x
0.13 0.02

charged particles/event.
+

The associated invariant mass distribution for the (nO,n-) system,

in Fig. 22, shows a low mass excess over the background for uncorrelated minimum bias particle distribution. These features are
consistent with correlated particles distributed symmetrically
around the triggered nO with an average transverse momentum of
300 100 MeV/c for all p
x

The question whether this low mass excess could be explained

by resonance production is answered by Fig. 23. The 0 peak is
clearly visible but does not exhaust all the excess. Assuming that
o and nO have the same p dependence (2.7 < p < 5 GeV/c) so that
the 0 production cross-s~ction can be express~d as a multiple of
the nO's, and using the following parametrization which has been
found to describe correctly 0 production

with n = 4.0 0.6 (in agreement with n production spectra), one

obtains:
I
+
2 (0 + 0 )

/T[0

0.9 0.2

( ~,)

U,) 0 decays have been substracted from nO cross-section.

637

A REVIEW OF THE ISR RESULTS

PROlON
1

TRANSVERSE
AXIS
Tt

SCATTER
PLANE

PROTON

Fig. 21 Diagram showing the kinematical quant~t~es used to describe

the triggering TIo and the secondary hadrons in the c. m. system.
The hadron momentum, p, is broken into three components. The absolute value of the component along the TIo transverse momentum is called
p , the one out of the TIo scatter plane is called p
The azimuthal
a~gle of the hadron, ~, is referred to the TIo scatt~tplane which
is the plane containing the incoming protons and the TIo.

638

G. VALENTI

++

+
+
-,--------'-1
+

0.08
0.04

-2
Go

>

'c

++

0.04

~
>

0.02

80

C
UJ

160

-.--+-~----------+

::J
"Go

Q.

-2

1)

.,v

Q.

004

>

"0

002

UJ

(e)

80

_+- t _____ t+ +_+_

-2

t
t

0.02
0.01

t+

-.-+- ------+-'.
-1

Fig. 22 Rapidity distributions and nO - n- mass distributions for

particles in the nO hemisphere with \~\ < 27. The data are sliced
into different p intervals between 0.4 and 1.7 GeV/c:
(a) 0.4 < p < oJ~, (b) 0.6 < p < 0.8 GeV/c, (c) 0.8 < p < 1.1 GeV/c,
(d) 1.1 < pX< 1.7 GeV/c. The m~ss distributions are plotted for
\y\ < 2. Th~ dashed curves show the minimum bias rapidity distributions expected for uncorrelated particles.

639

A REVIEW OF THE ISR RESULTS

760 MeV/c 2

70

I....
~

60

"-

50

40

CII

.0

E
::J
z 30
20
10

I ;'

--

......

'" ....... _

0.8

--_

1.2

m(yto yt!) (GeV/c 2 )

1.6

Fig. 23 The n - n mass distribution for particles in the nO hemisphere, within the limits I~I < '27, Iyl < 2 and 0.7 < p < 1.7 GeV/c.
The dashed line is the mass distribution expected for un~orrelated
particles and the dashed-dotted line is the background assumed to
estimate the p signal.

640

G. VALENTI

In other words at Pr
decays.

2 GeV/c only 16% of the 'ITo,S come from

+

The absence of w ~ 'IT 'IT 'ITo and the presence of a strong

three-particle correlation (at least as strong as the two-particle
one), see Fig. 24, excludes that a common parent, if it exists,
decays always into two particles.
Charged particles in the hemisphere opposite to the 'ITo

A distribution of Iyl for charged particles in the hemisphere

opposite to the 'ITo with 1180 - I < 35 is shown in Fig. 25. One
observes a large positive/negative ratio at large p and large y,
fairly independent of PT('IT). Moreover, plot c) inaicates an
increase of particle density for high p restricted to Iyl < 1.5.
This increase is better evidentiated inXFig. 26 where charged multiplicity data having Iyl < 1, 1180 - I < 30 are plotted against
PT(~O); a~ PT('IT) = 2.5 GeV/c the charged multiplicity is twice the
m~n~mum b~as one.
P
distribution
-out'-------

The important point to emphasize, in order to test the jet

structure in single particle distributions, is the coplanarity with
the scatter plane defined by the trigger part~cle and the incoming
protons. Fig. 27 shows dN/dlp
I for various intervals of p
0 fth~s
'
..
h out
x
A f ~t
quant~ty ~n t e torm
d N

gives:
B

= 2.1

0.3

fairly independent of p. Minimum bias data, on the contrary, show

a B = l/<p
> with <p x > ~ 0.6;P-.
To better understand the way
.
.
out
o u t . b a lXance d , one can cons~. der t h e s~ng
. 1e
~n wh~ch transverse momentum ~s
particle distribution in the variable x E = Ip /PT('IT)I This is
shown in Fig. 28 for p > 1.2 GeV/c; for comp~rison are also shown
the distributions for ~ (Feynmann scaling) and z analogously
F
+defined for electron-proton and e - e scattering processes.

A REVIEW OF THE ISR RESULTS

641

v
Fig. 24 Rapidity distribution of charged particles in the TI o hemisphere from the 45 data. Secondaries from events in which a fast
particle (p > 1 GeV/c, y > 0) occurs have been plotted; the fast
particle isxexcluded from the plot.

642

G. VALENTI

Q)

o2 ~ P. ~ 0 6
>

z
-0

04

-oo+_ ........

03

02

...

GelJc

and - summed

., ....
,

01

IVI

b)

posalollE.'s }

u negatives

11 ~P. 4i 17

..

0
IVI
c)

005
>
u

+++

z 003

-0

~~

._._._._.-

001
0

~I

..

IVI

Fig. 25 Distribution of the absolute value of the rapiditr for

charged particles in the hemisphere opposite the nO, with 1180 - ~I
< 35. (a) Summed over +ve and -ve charges for 0.2 < P <
0.6 GeV/c.
The dashed line is drawn to guide the eye. (b) Broken tnto +ve and
-ve charges for 1.1 < p < 1.7 GeV/c. The dashed lines are the same
curve as in (a) but scated down to agree with the data in the Iyl
< 1 region. (c) Broken into p (nO) intervals, 2 < p (nO) < 7..4 GeV/c
and 2.7 < P (nO) < 4.1 Gev/c,T and summed over charg~s for 1.1 < p
< 1.7 GeV/c~ The dashed-dotted line indicates the mean density
x
from minimum bias events.

643

A REVIEW OF THE ISR RESULTS

>
u

to-

21-

~
~
:::>
~

~
~

::I:

1 ..
I

*
+4
+
*
++
I
2

I
1

Tt

Pt
Fig. 26

I
3

(GeV/c)

Number of charged particles per event as a function of

= 0) for
1 80 - ~I < 30 and Iyl < 1 (full dots). Also shown (open triangles)
are data from ref. 12.

PI(nO) including the minimum bias value (shown at PT(nO)

644

G. VALENTI

1.0 ~P.
0.5
0.3

. . . .+....

...... ....

1.7

&

,...., ....

1.7 ~P. ~ 3.2

1.0~IYI~ 2.5

+. .

},?...

0.05

:;

a.

t,

0.1

:>QI

'0

1.4~p.<

0~IYI~2.5

'+-- ....

.......

~1.4

O~ IYI~2.5

0.03

'0

, l-(t
t"'L,

't'l'

0.01

0.005

Ipout I (GeVlc)

Fig. 27 Distribution of Ip
I for different p intervals of the
charged particles. The dist~~utions are integr~ted over Iyl < 2.5;
for the highest p interval the y range is restricted to 1 < Iyl
< 2.5. The dashedxlines correspond to a function
dN/dlp

out

1~

exp(-Zlp

out

I).

645

A REVIEW OF THE ISR RESULTS

(a)

This Experiment

t 41-

{.
o

210 l...w

0.8 -

'0

0.6 l0.4

nO at 90
nO at 450

0.2 I-

-1
0.4

as

(b)

This Experiment

nO at 900

:j

{ xE
x~

i
~

~I~'~
10

ItI

II
Q4

1
QS

xE or

I
0.8
x~

lD

(z)

tf
+?+

+++9

i-

;'1".

'" e'e

.,

'-

I
0.8

nO at 90

I-

o ep (x,)

l-

This Experiment

I-

(e)

l-

tt

"Q4

I
O.S

~~I

0.8

1.0

II[ or x, or z

Fig. 28 a) The x distribution obtained from the 90 0 and from the

45 0 data. b) Comp~rison between the x E and x E ' distributions
(90 0 data). c) The x distribution (90 0 data) is compared to the
Feynmann x F distribu~ion observed in deep-inelastic ~-p ~cattering
(ref. 18) and with !dN/dz (z = p/p
) observed in e - e annihilation at E
= 4.8 GeV/c (ref. 19).max
cm

646

G. VALENTI

V.

DIRECT LEPTON PRODUCTION

The motivation for investigating direct lepton production in

proton-proton interaction lies in the possibility of relating these
leptons to the leptonic decays of new particles. Several measurements have been performed at the ISR in the last few years, contributing to th~ now well established feature of almost constant (e/n)
ratio, fQr Is from a few GeV up to the highest ISR center-of-mass
energy Is = 62.4.
On the electron side two groups have published data on single
electron production. The CCRS 20 Collaboration has measured the
invariant production cross-section for single electrons with transverse momenta from 0.6 to 4.7 GeV/c and for five center-of-mass
energies I~ = 23.5, 3Q.6, 44.8, 52.7, 62.4 GeV. Fig. 29 gives,
for fixed values of Is, the p (center-of-mass momentum) dependence
of the charge averaged invariInt differential cross-section
E d 3a/d 3p. Fit to the charge averaged pion data from the BS Collaboration 13 are superimposed to the data points. The agreement between
pion and electron production seems to be goo~ at all energies considered.
A compilation of the e/n ratio fQr n- and nO (see Sec. III)
data is given in Fig. 30. Finally the Is dependence of the e/n
ratio for the CCRS data is shown in Fig. 31, together with two possible fits.
The AC~lliR21 Collaboration has extended the electron production
data for I~ = 52.7 to small p. While the direct electron producT
-4
tion for p > 1.5 GeV/c seems to behave (except for the 10 factor)
as the pionTs one, below p = 1.5 GeV/c, it increases when PT
decreases something like
p Fig. 32 shows the e/n ratio in
T
. .
excess after all known background sources present ~n the exper~ment
are taken into account. A compilation of e/n ratio versus PT for
I~ = 52.7 is shown in Fig. 33.

l/

In conclusion the e/n ratio emerging from these data appears

to remain constant (or slowly varying) with s at a level of 10- 4
from a few GeV up to 62 GeV (c.m.s). The new measurements reported
at the time of the Palermo conference, furthermore, wash out a
possible threshold effect around 25 GeV.
The recent discovery of the new family of particles (J '1" '1''')23
+ +decaying into lepton pairs (e e , V V ) has suggested the possibility
that the direct single electron yield could be explained as a decay
product. Indeed such decays represent a sizeble contribution to
the single electrons observed, but do not explain completely the
e/ n ratio data.

t>1~Q.

ILl

""

Ne

~'2

10-3'1=

1=

p; (GeV/c)

f=

f=

Fig. 29 The charge-averaged invariant cross-section for electron

production plotted a~ a function of the center-of-mass momentum PT
for five values of Is. The curves represent a fit of the chargeaveraged pion data of the BS Collaboration (ref. 13), and has been
multiplied by 10- 4

1=

'-I

0-

--t
(J)

(J)

:0

:0

(J)

::c
m

--t

o"Tl

<

:0

648

G. VALENTI

~.62.4

.111
I'

'

,~

2 Kr<

~.44.8

VI. 30.6
2 10-<

1 10-<

p; (GeVlc)
Fig. 30 a) The ratio of the charge-averaged invariant electron
cross-section to the BS fit (ref. 13). b) The ratio of electro~s to
neutral pions plotted as a function of PT for four values of Is.

GI

0.5.10- 4

1.0.10- 4

1.5.10- 4

1.3 GeVlc

eo+ e-/2 Tt!

BS
eO+e-/2 Tt

e/Tt> for p.

20
30

40
VS (GeV)

{0.63 In VS -1.46).10- 4

50

60

.I ___

____

Fig. 31 The ratio of invariant cross-sections of electrons to

pions of PT > 1.3 GeV/c plotted as a function of c. m. s. energy
Is. The two curves shown are fits to the full points.

10

0.934.10- 4

.
-.
-.
j
+
____ J__ v/fl/".
t .c:~

<

'0

.j>..

0.

cil

:c
:c
m
Cf)

Cf)

:::c

--l

o
.,

:ii:

<

:c
m

G. VALENTI

650

10-3~_------.-------~------~----~

o
0+-

0::

/--- -- -- W

/
/1
/
/

- -------

P
__ - - - - - _ _ _ _ _ _ _ _ ...../-../
,/'"

/
/

~///

/
/
I06~______~____~/L-_ _ _ _ _ _L -_ _ _ _~

0.5
Tra nsverse

1.0

1.5

2.0

momentum (GeV Ie)

Fig. 32 Observed electron excess compared with the prediction of

vector meson decays according to Bourquin and Gaillard (ref. 22).

10

-4

10

-3

TT

-L

o.

3.

Pr dependence of the e/n rate at ISR energies.

2.

al. at 30

Gav

at 90

et

-t-

C CRS

. latt'T

1.

+ttt+

Fig. 33

Vs = 52.7

0 - Baum

.;t

PT Gev

8:

(I)

-t

r-

(I)

::0
::0

en

-t
:I:
m

"

::E

<

::0

652

G. VALENTI

The J production i~ Eroton-pr~t~n collision at the ISR has

been observed in both e e 24 and ~ ~ 25
The results obtained are:
dG
B

for <p >

T

ee dy

~~

J1

y=O

0.67 GeV/c, and

B

for <p >

dG J /
= (7.2 2.4) 10-33cm2
dy y=1.6

1.0 0.2 GeV/c.

Electron-muon coincidences

Some very recent and preliminary results 26 seem to hint to

the presence of electron-muon production in proton-proton collision.
The existence of e-~ pairs is very relevant in the process of
understanding the source of prompt leptons in hadron interactions.
The method implemented by the ACIThlliR Collaboration is to compare
muon yield in coincidence with identified electrons and hadrons.
Electrons are identified using threshold gas Cherenkov counters,
lead glass calorimeters and magnetic analysis; background due to
electron pairs is detected via dE/dx measurements. Muons are
identified by absence of visible interaction in a calorimeter and
magnetic analysis.
Table III presents measured yield defined as:
Arml Arm2
Yield = ~~~~~~
Arm 1
where Arml and Arm2 are the electron and muon arm of the spectrometer respectively.
The excess of electron-muon with respect to hadron-muon
coincidences is measured to be
(1.61 0.50) 10- 4
A charge analysis of these
excess yield

e-~

events gives for the opposite charge

Hadron absorber
out

Hadron absorber in
visible interaction in calorim.
or wrong t.o.f.

Hadron absorber in
no visible interaction in calorim.
correct t.o.f.

Selection in
Arm 2 (muon arm)

.....

----

-2

-4

-4

--------------

(148 events)

(1.220.17) x 10

(34 events)

(2.000.34) x 10

(63 events)

(3.720.47) x 10

Electrons

(596 events)

-2

-4

-4

--

(1.02O.04) x 10

(123 events)

(1.870.17) x 10

(139 events)

(2.110.18) x 10

Hadrons

-4

(0.200.11) x 10- 2

(0.130.38) x 10

'3.2 s.d. from 0

(1.610.50) x 10- 4

Electrons-Hadrons

Event selection in Arm 1 (electron arm)

I

"

Fraction of events in Arm 1 (electron arm) with a valid count in Arm 2 (muon arm)

0-

<.n
c.>

m
en
C
r
-l
en

:0
:0

en

-l
:I:
m

::E

TABLE III

<

:0

654

G. VALENTI

II

:;:

and for the equal charge configuration

e

}l

= (0.23

0.44) 10- 4 ,

These results hint to an OPFosite charge e-ll pair production.

VI.

DIRECTLY PRODUCED PHOTONS

A CERN group27 has recently reported on a simultaneous measurement of large transverse momentum (1.6 to 3.8 GeV/c) 'yield of single
photons and photon pairs at 90 production angle performed at the
SFH.
The method implemented to extract the single photon production
cross-section is the following:
measurement of the
TI o -+ yy decay)

ii)

calculation of the single photon yield from

TIo

yield

(TI o

i)

is identified from the

TI o

decay

iii) single photon yield measurement.

Several sources of systematic bias had to be considered in
such a measurement; the most important are:
a)

single photons do not have the photon pair

hence higher background

b)

loss of photons by the apparatus affects single and pair

rate differently

C)

accurate knowledge of the energy response is necessary as

one wants to compare rates at different detection energies
(PT fall-off).

(TIO)

signature,

The TI o data, relative to 1.6 < p < 3.8 GeV/c for I~ = 45 and
53 GeV are compared to TI data from r~ference 28 in Fig. 34. Table
IV gives the detailed single photon observed yield as well as the

(*) The data are not corrected for acceptances.

A REVIEW OF THE ISR RESULTS

10- 21
5

2
~

~u

10-29

~
Ne

\!)

~
~a.

"C

"
UJ

2
1~

655

VS=45 GeV

VS =53 GeV

-t

q.

t.
0

Oc

t:l

Cf

10-31

)(

)(

Fig. 34 Inclusive nO production cross-section at 90 for I; = 45

and 53 GeV, as measured from resolved photon pairs (open circl~s).
Also shown are the data of ref. 28 for n+ (full circles) and n
(crosses).
5000

'IS. 45 GeV

VS= 53 GeV

P:," (GeVk).

Fig. 35 Single pho~on yields uncorrected for antineutrons contamination at 90 for Is = 45 and 53 GeV. The observed distributions
(full dots) are compared to those inferred from the resolved photon
pair data (open circles). The latter are also shown after addition
of the contribution from n + yy decays (crosses).

G. VALENTI

656

TABLE IV
Single photon data
Data are integrated over both values of IS (45 and 53
cm
GeV ) and over Pt between 2.8 and 3.8 GeV/c.
All
numbers are given as fractions of the nO production
cm
cross-section in the same range of Pt

Observed single photon yield

0.462 0.008

Single photon events

Conversion losses

+ 0.032 0.009

Events with an independent

second photon

+ 0.037 0.009

Antinucleon contamination

0.077 0.046
0.455 0.053

II

Single photon yield from nO

and Tl decays
nO .. y y

0.192 0.006

Tl .. y Y

0.062 0.011

Uncertainty in acceptance
calculation

0.013
0.255 0.018

III

Excess of single photons

yin ratio
Uncertainty on energy response

0.20

0.06
0.07

657

A REVIEW OF THE ISR RESULTS

o VS

0.4

=53 GeV

VS =45 GeV

0.3
0.2

0.1

O~--~----~----~---L----~-

2.8

3.0

3.2

3.4

p~m

3.6

3.8

(GeV/c)

Fig. 36 The ratio R between the excess of single photons (corrected

for antineutron contamination) not accounted for by TID and n decays
and the TID yield measured.

contamination estimates. The data are integrated over both values

of I; and over 2.8 < PT < 3.8 GeV/c. Fig. 35 shows the PT distribution of the single pfioton yield.
The ratio R between the excess of single photons not accounted
for by TID and n decay and the TI o yield as a function of the transverse momentum is shown in Fig. 36.

658

G. VALENTI

REFERENCES
1) - CERN-PISA-STONY BROOK Collaboration, Physics Letters 62B, 460
(1976).
2) - S. Van der Meer, CERN Internal Report ISR-POj68-31 (1968).
3) - See for instance U. Amaldi et al., Annual Review of Nuclear
Science vol. 26.
4) - A. S. Carrol et al., Physics Letters 61B, 303 (1976):
5) - U. Amaldi et al., Physics Letters 44B, 112 (1973),
S. R. Amendolia et al., Physics Letters 44B, 119 (1973),
S. R. Amendolia et al., Nuovo Cimento 17~735 (1973),
G. Bellettini, 5th International Conference on High Energy
Collision, Stony Brook, 1973 (ed. C. Quigg) AlP Conf. Proc.
No. 15, K. Eggert et al., Nuclear Physics B98, 93 (1975).
6) - CERN-HAMBURG-ORSAY-AUSTRIA Collaboration, Physics Letters
58B, 233 (1975).
7) - CERN-HAMBURG-ORSAY-AUSTRIA Collaboration, Physics Letters
62B, 363 (1976).

8) - R. J. fL Phyllips and V. Berger, Physics Letters 46B, 412 (1972).

9) - B. Alper et al., Physics Letters 44B, 5'21 (1973), ~1. Banner et
al., Physics Letters 44B, 537 (1973), F. \oj. BUsser et al.,
Physics Letters 46B, 471 (1973).
10) - P. V. Landshoff, Large transverse momentum reactions, Plenary
report at the 17th International Conference on High Energy
Physics, London 1974,
O. Sivers, S. J. Brodsky and R. Blakenbecler, Large transverse
momentum processes, Physics Reports 23, (1976).
11) - F. W. BUsser et al., High transverse momentum phenomena
involving TI and n mesons at the CERN ISR, paper presented at
the Int. Symp. on lepton and photons, Stanford University,
21 - 27 August 1975,
F. W. BUsser et al., Physics Letters 51B, 306 (1974).
12) - K. Eggert et al., Nuclear Physics B98, 49 (1975).
13) - B. Alper et al., Nuclear Physics B87, 19 (1975).
14) - J. W. Cronin et al., Physical Review 011,3105 (1975).
15) - P. Oarriulat et al., Nuclear Physics B107, 429 (1976).
16) - F. l~. BUsser et al., Nuclear Physics B106, 1 (1976).

A REVIEW OF THE ISR RESULTS

659

17) - P. Darriulat, Hadronic collisions with large transverse

momentum products, Rapporteur talk at the International
Conference on High Energy Physics, Palermo 1975.
18) - P. Capiluppi et al., Nuclear Physics B79, 189 (1974).
19) - J. T. Dakin et al., Physical Review D10, 1401 (1974).
20) - B. G. Pope, Rapporteur talk at the International Conference
on High Energy Physics, Palermo 1975.
21) - L. Baum et al., Physics Letters 60B, 5 (1976).
22) - M. Bourquin and J. M. Gaillard, Physics Letters 59B, (1975).
23) - See for example S. Ting, Rapporteur talk at the International
Conference on High Energy Physics, Palermo 1975.
24) - F. BUsser et al., Physics Letters 56B, 1044 (1975).
25) - E. Nagy et al., Physics Letters-60B, 96 (1975).
26) - Aachen-CERN-Heidelberg-Munich-Northwestern-Riverside (ACHMNR)
Collaboration private communication.
27) - P. Darriulat et al., Nuclear Physics BllO, 365 (1976).
28) - B. Alper et al., Nuclear Physics B100, 237 (1975).

G. VALENTI

660

DIS C U S S ION
CHAIRMAN:

Scientific Secretaries:

Prof. G. Valenti:
M.A. Ichola and T. Wilkie

DISCUSSION
CRONIN:

Where does the background estimate come from in the

correlation events? From hadronic measurements?

e~

charge

VALENTI:

The estimate for the various e~ charge states is derived from

the comparison of the rates of e~ events identified as background
to the rates of good e~ candidates.
FERBEL:

Can the

e~

events be interpreted as charm?

ZrCHICHI:

There was a seminar at CERN about this. Gaillard has shown that
these events are in the wrong phase-space region for charm decays
since they are in the very forward direction.
What are you conclusions about Feynman scaling?
VALENTI:

The conclusion is that Feynman scaling is broken at x ~ O.

Furthermore, the deviation seems to be stronger at small PT'

A REVIEW OF THE ISR RESULTS

661

WIGNER:
I do not understand the significance of your single photon data.
Could you elaborate please?

VALENTI:
Experimentally, the comparison of the measured single photon
yield to that expected from nO decays is done by first measuring the
yield of nO,s -- identified as photon pairs with correct invariant
mass -- then calculating the single photon yield expected from nO
decays seen by the detectors, and finally by measuring the single
photon yield itself. The result is that there is an excess of single
photons not accounted for by nO decays, suggesting that high transverse momentum photons are directly produced in proton-proton
collisions.

ZICHICHI:
A bit of history: in the past, people estimated the number of
nO,s produced in pp collisions simply by counting the number of
photons observed. It was assumed that all these photons were produced by nO decay. One always found that the rate of neutral pions
was greater than that of n. This is the first time the nO has been
reconstructed from both photons, and this is the first time that the
nO rate coincides with the n rate.

HIGHLIGHTS OF THE TBILISI CONFERENCE

C.W. Fabjan
European Organization for Nuclear Research (CERN)
Geneva, Switzerland
The 18th International Conference on High-Energy Physics in
the series of the so-called 'Rochester' Conferences was held in
Tbilisi, the capital of Georgia, 15-21 July, 1976.
Georgia, which is one of the 15 republics of the Soviet Union,
has produced men as unlike in character as C. Roustaveli, the great
12th century poet with his world-famous work "The Prince in the
Lion's Skin" and Joseph Stalin. The chain of the Caucasus mountains,
which surround and traverse this country, contrasts dramatically with
the fertile plains, which ensured a privileged life to a succession
of invaders over the last three millennia. Christianity found an
early stronghold in this country, to which a number of monuments
testify, illustrating eloquently the varied and highly advanced expressions of the architectural and decorative arts.
In selecting the "highlights" of the Conference I preferred not
to present a catalogue of the many, very interesting results reported
at Tbilisi, but rather to discuss only those issues which, during the
last year, haye most strongly affected our fundamental concepts.
Some of the issues, according to common consensus, appeared to be
settled, while others were still very much debated and promise exciting results in the near future.
LEPTON PRODUCTION IN HADRON COLLISIONS
In 1974 at the London Conference the first measurements of inclusive lepton production at FNAL and ISR energies were reported.
The subsequent two years were marked by intense experimental activity, and work submitted to this conference permits us to draw preliminary conclusions.
663

c. W. FABJAN

664

Observation of a Dilepton Continuum

with High Invariant Mass
Measurements from FNAL are summarized in Fig. 1, including both
dimuon 1 ,2) and dielectron 1 ) data. The data are compared with one of
many models proposed, namely the mechanism of parton-antiparton annihilation, as advocated by Drell and Yan 3 ) . The solid line in Fig. 1
shows the result of such a calculation, assuming a (1-x)7ft distribution for the coloured partons, where x is the ratio of the parton to
nucleon momentum.
For some time it has been noticed that the PT distribution of
the dilepton pairs flattens with increasing mass of the pairs. This
trend is clearly exhibited by the data 4 ) , as shown in Fig. 2. Data
for both proton-induced and pion-induced ~ pairs are shown, permitting 'a comparison of cross-section and slope. This PT behaviour,
which differs considerably from the PT distribution of normal hadronic events, was not well reproduced by Drell-Yan-type calculations;

o electrons}
muons
Ref .

x muons

0
~

Ref. 2

10- 35

>-

"C

"C

"C

+t l

<II

OJ

...

<II

a.

10- 36

l!)

NE
u

j
8

10

11

m,.,_ (GeV)

Fig. 1

Production of massive dilepton continuum. A theoretical

prediction based on a Drell-Yan calculation with colour is
also shown.

665

HIGHLIGHTS OF THE TBILISI CONFERENCE

-Ioc-

0.4 0.8 1.2 1.6 2.0

PT (GeV/c)

Fig. 2

0.4 0.8 1.2 1.6 2.0 2.4

The PT dependence of pion- and proton-induced

function on the invariant mass range.

pairs as a

more recent recalculations of this effect, as discussed at the

ConferenceS), appear to have considerably better agreement with the
PT data.
An important result 3 ) of the Drell-Yan model is its prediction
that for s ~ 00, M ~ 00, but s/M2 finite, the cross-section for pair
production should exhibit a scaling behaviour of the form

The observation of such a scaling behaviour at present machine

energies would be of extreme importance, as, based on CVC arguments 6 )
connection can be made between the production cross-section at fixed
s of a lepton pair of given invariant mass and the charged and neutral
intermediate vector mesons of the same mass. The present status of

c. W.

666

10- 33

,
,,-,
,
r,
,I

10- 34

I
10-35

,t

/-5

//

FABJAN

'"

=760 GeV'

COWMBIA, B.N.L. (1970)

o (LEDERMAN REVISED) (1974)

,I-

... M.I.T., B.N.L. (1974)

'" COLUMBIA,FNAL, U. of 1.(1975)
CHICAGO - PRINCETON [
- (preliminary) (400 GeV)

(300 GeV)

10- 36 '__---'---'---L-LJ.-LJ..l-'--~'__L_l__'_.l...LL.l.L_~_'__'_L_'...J..J....LJ
10

10'

103

5/M'

Fi g. 3

The pair production cross-section M3 (dcr/dr) as a function of

scaling variable s/M2.

the scaling behaviour of the data is summarized in Fig. 3. Whereas

a large scaling violation between the BNL data and FNAL data is observed, there may be a hint in the data (compare the 300 GeV/c and
400 GeV/c points) that at FNAL energies incipient scaling may be
reached. In the near future, new data from the CERN ISR will clarify
this point.
Whereas the existence of a mass-continuum of dileptons has been
clearly established experimentally, its theoretical interpretation
requires further work. The experimental study of this continuum
over a large range of invariant masses, and in particular at very low
masses, and its dependence on IS, xF, and PT should allow a better
distinction to be made between competing models of production 7 ) .

667

HIGHLIGHTS OF THE TBILISI CONFERENCE

Dileptons from Vector Meson Decay

New data on vector meson production were reported at the
Conference. Precise knowledge of these cross-sections is of course
necessary to evaluate the contribution of vector meson decays to
the inclusive single lepton spectrum (see next subsection). Table 1
gives cross-sections for inclusive production, and Fig. 4 shows some
of the relevant data, again for proton- and pion-induced production 4 ).
The dilepton continuum is also shown in Fig. 4, as the data have been
corrected for background sources.
Table 1
Inclusive vector meson production
by 150 GeV/c protons
a/nucleon
< 11 mb

w (assumed)

8.6 2.5 mb
660 220 ].1b

<P

94 31 nb

J/1jJ

In Fig. 5 the measured dielectron spectrum is shown as reported

by the Columbia-FNAL Groupl). Refined analysis of the data resulted
in improved energy resolution of the apparatus: the previously reported structure at 6 GeV 8) persists, containing 14 events. This
structure is, however, not seen in a recently obtained data sample
on ].1+].1- pairs by the same group; but, as the energy resolution for
the ].1 pairs was considerably worse, the two experiments are not in
>cu

:::;:

;;;Q

PROTON INDUCED _104

V>
=>
XF >0.15
~

XF>0.15

<0
N

PION INDUCED
XF >0.15

~103

:;;:cu

(f)

<!l

I-

z
2
>
UJ

~102

UJ

:::;:

:::;:

-0

"-

UJ

lI
~

-0

"-

10

~IO

UJ

;;:

0.6

1.0

1.0

2.0

~~~~
3.0

MJLJL (GeV)

Fig. 4 The production of vector mesons and the low-mass ].1-pair

continuum.

C. W. FABJAN

668

d'a

dmdy

em' IGrN Inucleon

y;O

[Columbia - F N AL ]

10- 33

10-34

10- 36

'-------'-------'-----'---------'------'--------''------'--

10

m (GeV/c')

Fig. 5

The dielectron spectrum observed by the Columbia-FNAL

Collaboration. With the clear evidence for the J/~ and the
~/, structure is also observed at 4.3 GeV/c 2 and 6 GeV/c 2

direct contradiction. The reanalysis of the electron data gave a

somewhat sharper structure at 4.3 GeV and it is tempting to identify
this region with the broad resonance structure observed at SLAC in
0tot of e+e- annihilation. Further evidence for a resonance structure was reported by another group9), which observed through photoproduction not only the J/~ and ~/, but also a clustering of eight
events at M ~ 4.5 GeV/c 2 above an estimated Bethe-Heitler background
of 0.24 events.

HIGHLIGHTS OF THE TBILISI CONFERENCE

669

Inclusive Single Lepton Production

The data obtained during the last two years up to the time of
the Conference show the following global features 10 ):
a)

For PT > 1 GeV/c and large production angle 8 cm , data have been
obtained by the CERN-Columbia-Rockefeller-Saclay (CCRS), ChicagoPrinceton II (CPII), and Columbia-FNAL Collaborations. The ratio
R = lepton/pion is found to be R = 10-~ with an indication that
R may be rising with increasing values of IS and PT. Measurements
at Serpukhov indicate a threshold for production of ~'s around
IS = 8 GeV reaching the plateau value of R~ = muon/pion = 10-~ at
IS = 12 GeV.

b)

For PT < 1 GeV/c and large production angle 8 cm , data have been
reported by the Aachen-CERN-Harvard-Munich-Northwestern-Riverside
Collaboration. The ratio Re = electron/pion rises below
PT = 1 GeV with decreasing PT up to values of Re = 3 x 10-~ at
PT ~ 250 MeV/c (IS = 52 GeV). A similar behaviour is reported
by the Pennsylvania-Stony Brook Collaboration at ISvalues
between 4.5 and 7 GeV, at variance with an old experiment, reanalysed by Winter, which resulted in Re = -(0.2 0.29) x 10-~.

Three contributions dealt with the very important question of

possible sources of this prompt lepton production. Based on the
measured dilepton rield (continuum and vector mesons), the ChicagoPrinceton II Group ) computed the invariant cross-section for single
inclusive lepton production at IS = 17 GeV (Fig. 6a). A parametrization of the data is necessary to extrapolate the measurements outside
Vector Mesons

(b)

(a )

Continuum

3.0
~

Q
><

2.5
2.0

'"
~ 1.5

Vector Mesons
plus Continuum
---- Vector Mesons only

1.0

0.5

Pr (GeVlc)

Fig. 6

2
Pr (GeV/c)

a) Monte Carlo estimate of the single inclusive lepton yield,

based on the measured dilepton spectrum.
b) Comparison of the Monte Carlo estimate of the ~-/TI- ratio
with various experimental results.

c. w.

670

FABJAN

the limited kinematical region, which was covered in the measurements.

In particular, the result for large PT is a sensitive function of the
PT distribution of the high-mass pairs (J/~-particle, continuum). For
comparison, the measured inclusive ~-spectrum, downscaled by x 10- 4 ,
is also shown. The ratio R = win, computed in the above way, is compared with the data in Fig. 6b. Within the rather large errors, reflecting the uncertainties of the kinematical extensions and renormalization errors, the measured dilepton production accounts well for
the observed ratio R. Consequently, electromagnetic processes are
identified as a major, if not the only source of the observed single
lepton spectrum.
The same conclusion 1S reached by the BNL-Yale-FNAL Groupll),
which measured wand w-pair production at PT < 2 GeV/c in the forward
direction. Again, within errors, reflecting the assumptions necessary
for kinematical extension of the data, the measured dimuon yield accounts for the observed single muon production. Furthermore, the
group measured the polarization of the prompt W's and found for the
polarization parameter
P

-0.15 0.2 ,

consistent with zero. This measurement provides additional confirmation of the electromagnetic origin of the prompt leptons.
Perhaps contradictory evidence has been reported from a team
working at Serpukhov l2 ). Inclusive W production was measured in the
transverse momentum range 1.9 ~ PT ~ 3.0 GeV/c and at production
angles Bcm ~ 90. A charge asymmetry was observed of

1.20.1

a measurement of the polarization yielded

+

peW )

= -(0.85

0.36) ,

whereas the W-'s were found to be unpolarized. As different kinematical regions were investigated in the Serpukhov and in the BNL-YaleFNAL experiments, the two results are not necessarily contradictory;
a planned remeasurement by the Serpukhov group should clarify this
very important point.

671

HIGHLIGHTS OF THE TBILISI CONFERENCE

THE

J/~

FAMILY AND ITS INTERPRETATION

AS AN "-ONIUM" SYSTEM

Immediately after the discovery of the J/~ and ~/, an interpretation was suggested which identified these particles as bound states
of a heavy spin 1 particle bound to its antiparticle. The level
scheme 13 ) for such a bound state, closely resembling that of positronium, is shown in Fig. 7. At the conference, improved data on the
intermediate states and their tentative quantum number assignment were
presented. They provide crucial support for the bound-state interpretation; an impressive list of data can now be accounted for, if
charmed quarks with Q = 2he and M ~ 1.6 GeV/c 2 are assumed as the
constituents of the "-onium" system ("charmonium").
Further Evidence on Intermediate States
New data on radiative transitions of the form ~' ~ y + X;
X ~ Y + J/~ were presented by two groups working at SLAC. The SLACLBL Group14) reported evidence for states at 3.45, 3.50, and 3.55 GeV.
3
3

0,

3D,

1--

33 5,

3
3 3

0,

1-- \II' 3700

0-+ 1I~ 3450

D,

23 5,

2' So

2++

3550

1++

3510

3p

0++

3415

3 pO

3p,

1+-

'P,

J3100
--''-'-'=-=-_13 5,

Fig. 7 The level scheme of "charmonium" with a tentative identification of the recently seen intermediate states.

672

C. W. FABJAN

The Pavia-Princeton-Maryland-UC San Diego-Stanford Group15) measured

with considerable statistics the y-lines at 3.415, 3.512, and
3.551 GeV. The information on these intermediate states is supplemented through the observation of the hadronic decay modes of
X states:

Wi

-+

Y + X,

hadrons (2n, 2K2n, 3nTI, ... ) .

-+

Such data were reported by the SLAC-LBL Group14) and give further information on the 3.41, 3.50, and 3.55 states (Fig. 8). The angular
distribution of the emitted photon with respect to the beam was also
measured. For the X(34l5) it is consistent with 1 + a cos 2 8, a = 1,
identifying this state as an S = 0 level. For the other states, the
angular distribution was consistent with a < 1.
1ji'(3684)-y +hadrons Preliminary

40

20

0
20
II>

C
(l)

>

10

lJ.J

0
15

n;+ n;+ n;+ n;- n;- n;-

10
5
0
5
0
30

32

3.4

3.6

Effective mass (GeV Ic 2)

Fig. 8 Preliminary data on the hadronic decay modes of the intermediate states of the J/W family.

673

HIGHLIGHTSOF THE TBILISI CONFERENCE

The Double-Arm Spectrometer (DASP) Collaboration 16 ) at DESY reported

data on the decay channels
J/1jJ

+ m

+ y,

m+y+y

ljJ' + m + y,

y + y,

where m = TID, n, n', X. The group reported 110 events of J/ljJ + 3y;
this sample contained decays to n, a hint of n' at tV 900 HeV, and
29 events clustered at 2.8 GeV above the background, estimated to be
14 events. Further evidence on this X(2800) state is seen in decays
of ljJ' + Y + (2y) , where again five events above background are seen
clustering at 2.8 GeV. The branching ratios as observed by the
DASP Collaboration are given in Table 2.
Table 2
Radiative decays of J/ljJ, ljJ'
Decay
mode

1.0 0.2 x 10- 3

< 4.4 x 10- 2

ny
n'y
X(2800)y

ljJ'

J/1jJ

1.6 x

10- 4

< 1.3

10- 3

< 1.4

10- 3

< 3.7

10- 4

< 1.6 x 10- 4

TIDY

Based on this newly reported experimental evidence, the most

plausible identification 17 ) with the states predicted by the
charmonium model is given in Fig. 7.
J/ljJ Family Interpreted by the

Charmonium Hodel

In these calculations 13 ,17) the bound-state fermion is assumed

to be a charmed quark of mass MC tV 1.6 GeV/c 2 , of charge Q = 2he
and magnetic moment ~C = eh/2MCc. A phenomenological, strongly
confining potential of the form

ar + S/r + y

is assumed; the problem is treated non-relativistically as typical

momentum transfers in the system are of the order of hundreds of MeV.
The existence of the intermediate states provides one of the
strongest suggestions for the validity of the charmonium picture.
However, further high statistics work on these states and the experimental verification of those at present only tentatively assigned

c. W.

674

FABJAN

quantum numbers will be crucial. Besides a rather impressive list

of correct predictions of the charmonium model, there are some open
problems requiring further investigation 13 ) . A sample is given in
Table 3.

Table 3
Some open problems with the charmonium model
Experiment

Quantity

r(JN

->

nc

+ y)

M('P 2 ) - M('P,)

Theory

S 7 keV

20-30 keV

0.4-0.5

1. 0-1. 4

M('P,) - M('P o )

('S-'S) splitting

for n = 1 :
n = 2:

'" 300 MeV

'" 230 MeV

'" 100 MeV

Conments

Relativistic effects may be important;

a recent relativistic treatment gave
0.3-0.5
Only the ratio of (n=l) to (n=2)
splitting is consistent with theory

EVIDENCE FOR CHARM

The mounting evidence for the "Hidden Charm" hypothesis as a
possible explanation for the J/~ family, and the observation of final
states in deep inelastic V scattering consistent with the production
of hadrons carrying a new quantum number, found its culminating
corroboration in the observation of hadrons with properties precisely
predicted by the charm scheme.
Observation of Charmed Mesons by
the SLAC-LBL Group
In a series of two different measurements 14 ) the group observed
the decay of neutral and charged charmed mesons. The neutral state
was observed through the decay modes
MC

-+

K + 'TT+
+

-+ K-

+ 3TI

The data presented at the conference for these two decay modes are
shown in Fig. 9. Particle identification was based on time-of-flight
information.
From the data the mass of MC was determined to be m = 1865
r < 40 MeV/c 2 , compatible with the experimental resolution of the detector. Table 4 summarizes the observed
production cross-sections for various intervals of W.
15 MeV/c 2 with a width

675

HIGHLIGHTS OF THE TBILISI CONFERENCE

.,..+.,..-

(0)

~ 600

~
o

+.

++

+++

........

400

.....

fi

Iii
:::;:

...........

++

.....

ft.

...........

(e)

(f)

(h)

Ii)

100

L-~~~~~~~~~~

1.6

1.8

2.0

1.6

1.8

2.0

INVARIANT MASS

Fig. 9

....,..............

__~.~~~~~
l.6

l.8

2.0

'2.2

(GeVlc 2)

Experimental evidence for decay of a neutral charmed meson .

Graphs (a) to (c) show the invariant mass spectrum for an
arbitrary assignment of particle identification. In graphs
(d) to (f) the measured TOF was used to statistically improve
the particle identification. A clear peak in the nK channel
is observed. Graphs (g) to (i) show the resonance decaying
into K3n.
Table 4
Cross-section for Mc(lS65) production
BR'O
,~

hadronic
(nb)

K1r

3.9-4.6

27 3

0.20 0.05

4.S

IS 2

0.10 0.07

J/IjJ,1jJ'

< 0.02

[nb]
K3n
0 . 69 0.15

< 0.04

676

C. W. FABJAN

In a subsequent running
which corresponds to a local
cross-section around 4.0-4.3
evidence for a charged meson
MI
C

-+

period, data were taken at W = 4.03 GeV,

maximum in the multiply-peaked total
GeV. These data (Fig. 10) provide clear
decay:

K+ + ll + n

Its mass was determined as m = 1875 15 MeV/c 2 with a width of

M ~ 40 MeV/c 2 , again consistent with the experimental resolution.
The decay into the exotic final state and the absence of decays into
non-exotic channels are expected for the weak decay of charmed mesons
and constitute the strongest evidence for the existence of such
states.

80

60 N

40 -

20 -

(f)

K+ TT TT

IIII I \
tit !!!!t t t!
HtH

(0) -

ttHHt

150 -

co
0

u
0
W
lI
t9
[jJ

100 -

Hf

I-

H+ +

+ +++.

1.6

1.8

2.0

2.2

INVARIANT MASS

Fig. 10

(b)

t t TtH+T

:3:

K+ TT + TT -

ttt

t ttH
50

(GeVlc 2 )

Experimental evidence for a charged, charmed meson. The

decay into the exotic channel K+ + n + n is clearly seen,
precisely as predicted by the charm scheme; it is absent
in the non-exotic channel.

HIGHLIGHTS OF THE TBILISI CONFERENCE

677

Evidence from Other Laboratories

Two groups working at DESY reported on the observation of anomalous lepton production 16 ) . One group, using the supraconducting
solenoid "Pluto" for the momentum analysis of charged particles,
studied the reaction
e

+ e

-+

e + KO + x
S

at W 4.1 GeV. They observed 27 such events above an estimated

background of nine events, suggestive of a semileptonic decay of a
charmed meson. They quote BRoo ~ 1 nb for this class of events.
Similar evidence is reported by the DASP group, which used two magnetic spectrometers to study
e+ + e

-+

e + x

at various values of W. A fairly strong signal was found at

W ~ 4.4 GeV. The signal shows a threshold effect, as no events
were observed at W = 3.68 GeV. After careful analysis and subtraction of background sources, a total of 22 events was seen, of which
nine are accompanied by KS'S' These events are characterized by a
fairly high hadron multipl~city n) ~ 5), a relatively soft momentum spectrum of the leptons, and BRoo ~ 1 nb. Again, this observation is consistent with the production and subsequent decay of a
hadron pair carrying a new quantum number, conserved in strong and
electromagnetic interactions.
The Dubna-Berlin-Budapest-Prague-Sofia Collaboration 18 ) reported
very preliminary data on hadron production by neutrons of average
momentum of 50 GeV/c on a carbon target. They observe 14 events in
the invariant mass spectrum of (An+n-) clustering at 2090 MeV/c 2
The measured width of this cluster is r = 23 MeV/c 2 , consistent with
the experimental resolution and to be compared with the A(2l00)
having a width of r = 150-200 MeV/c 2 We should know shortly
whether this effect represents the first evidence for a charmed
baryon or whether it is merely a statistical fluctuation in the
data.
The compelling evidence for charmed hadrons found at SPEAR was
truly the highlight of this Conference; the hints of charm seen at
other laboratories promise feverish activity and a rich harvest.
Iliopoulos in his rapporteur's talk at the 1974 London Conference
showed confidence that this 1976 Conference would be dominated by
the discovery of charm. In a similar talk at Tbilisi, de Rujula
proclaimed that charm will still occupy the centre of the stage at
Tokyo in 1978.

c. W.

678

FABJAN

HAVE WE SEEN A NEW HEAVY LEPTON?

The SLAC-LBL Group was not just content to announce the discovery of charmed hadrons. For a year this group has been observing
anomalous lepton events, and more data permitting a more refined and
suggestive analysis was presented at Tbilisi 14 ) . This group has now
recorded 105 events of the form
e+ + e

~ e + ~+ + (missing energy) ,

where no charged particle or photon was observed in the final state.

Whereas the experimental evidence for this class of events has
been growing over the last year and is now well established, the
interpretation of this effect is still being debated. The general
features were presented, which characterize this phenomenon, and
arguments were given, which, in principle at least, permit a discrimination between the production of a hadron pair carrying a new quantum number and the production of a pair of "sequential heavy leptons",
given by the reaction
e

and

+ e

the subsequent decay

+ V

+ \i

The cross-section for the observation of those events as a

function of the centre-of-mass energy is shown in Fig. 11. The production shows a threshold around 3.5 GeV, rises to a maximum at
4.5 GeV, and decreases slowly at higher W. Indication of the nature
of the decay can be obtained from the momentum distribution of the
detected leptons, as shown in Fig. 12 as a function of the normalized
momentum
p

(p - 0.65)/(p
- 0.65)
max

and three ranges of W. The choice of the parameter p allows the

comparison of the lepton spectra above the cut-off of 0.65 GeV/c
taken at different values of W. In two of the three W-ranges, a
three-body decay of a massive lepton is clearly favoured over a twobody decay of a boson.
The key to the discrimination between a new hadron and a heavy
lepton is provided by the "missing energy" in the observed reaction,
which for lepton decay would be carried by neutrinos. From the
p-distribution of Fig. 12 it can be deduced with 95% confidence
that the mass m of the unobserved particles must be
m < 0.7 GeV/c 2

HIGHLIGHTS OF THE TBILISI CONFERENCE

35

679

.....-i!--r----,r-----..---r----r----..

:::} 2 -body

30

-0-

I.G} Y A
-1.8 -

E
25
u

U>

rtl
I

20

....
.

15

10

'C

,-

,/

.a
0

tf

'

5
0

Ecm

15
3.8 S Ec.m.< 4.8 GeV

10

(GeY)

Fig. 11
The production cross-section
of the "anomalous" lepton
events as a function of Ecm.
The solid and dash-dot
curves are predicted for
heavy lepton production,
the dotted and dashed curves
for a 1.8 GeV/c 2 boson and
subsequent two-body decay.

3V-A
-bod, d.",

MU = 1.8 GfN/c2
Mvu=O

5
___ !2-bOdY decoy
-.-.- Mu = 1.8 GeY/c2

0
In

15

...
0

10

-...

"::I..
In

"(I)

(I)

.c
E

40

::I

30
20
10

0.5

1.0

Fig. 12
The distribution in normalized momentum for the detected lepton. The solid curve
corresponds to a heavy lepton, Mu = 1.8 GeV/c 2 , MvU =
= 0.0, and V-A coupling.
The dashed and dashed-dotted
curves are predictions for
two-body decays of a
1.8 GeV/c 2 boson.

c.w.

680

FABJAN

This excludes the possibility of neutrons, and hence, for example,

the decay of a charmed baryon into e- +
+ n. The possibility of
unobserved K{'s can also be evaluated by studying the class of events

ve

e+ + e

~ e + ~ + K~ + x .

Similarly, the probability for missing TIo,S can be evaluated by

studying events with one or more visible photons. Such a careful
analysis, evaluating also the effect of non-instrumented regions
of the detector, indicated that at most 39% of the observed anomalous
decays could contain undetected photons or charged particles.
The simplest hypothesis compatible with all the data is the
production of a Rair of heavy leptons, with a mass in the range
1.6 to 2.0 GeV/c
CONCLUSION
This conference will be remembered by the announcement of the
discovery of charmed hadrons. The evidence for anomalous lepton
production in e+e- collisions and its interpretation as heavy lepton
production will initiate a varied experimental activity and we may
hope for more definitive conclusions soon. Deep inelastic leptonhadron scattering continues to be a rich source for new information.
The weak neutral currents have been studied by several groups with
increased attention to detail. Several reports were presented at
the conference, which indicate that parity is violated in these
interactions in a way which is consistent with the Weinberg-Salam
theory. With the increasing consensus that the electromagnetic
and weak interactions are unified, the notion that gauge field
theories ("quantum chromodynamics") are also underlying strong interactions seemed to be rapidly becoming accepted. One such
consequence -- systematic scaling violation in deep inelastic leptonhadron scattering -- was observed in ~ scattering at FNAL and reported at the conference 19 ) .
The range of activities in high-energy physics is perhaps best
illustrated by the announcement 20 ) of the observation of a TI~ bound
state ("pionium"), produced in
decays with a branching ratio of
~ 10- 7 Study of such atoms will not only permit a study of TI~
interactions, but may also shed light on po.ssible e-~ differences.
On the more speculative side, an old idea of Pontecorv0 21 ) was discussed with new interest: although direct measurements of the rest
mass of neutrinos slowly yield improved upper limits, a much more
sensitive way to look for differences in the v~ and ve rest masses
may be provided by the possible observation of "neutrino" oscillations. In such a case -- the well-known oscillations in a
KO-io beam may serve as a conceptual guide -- the intensity of, for

Kl

HIGHLIGHTS OF THE TBILISI CONFERENCE

681

example, an initially pure ve beam is predicted to vary in space

with a characteristic length L, which depends on the mass difference
and momentum of the two neutrinos. Observation of low-energy solar
neutrinos may be the ultimate test, but meanwhile experiments at the
highest energy accelerators are actively contemplated.
May this short 'tour d'horizon' have transmitted some of the
excitement and importance of the new discoveries, given an indication
of the experimental program with which some of the pivotal questions
are being addressed and given a flavour of the breadth of the field.
We are justified in our expectation to learn of similarly important
discoveries at the next conference at Tokyo in 1978.

C.W.FABJAN

682

REFERENCES

Unless otherwise
this Conference.

stated~

all references are contributions to

1)

Columbia-FNAL Collaboration;

D.C. Hom et al.

2)

Chicago-Princeton I Collaboration;

L. Klugberg et al.

3)

S.D. Drell and T.M. Yan, Phys. Rev.

Letters~,

4)

Chicago-Princeton II Collaboration;

5)

M. Duong-van, private communication and SLAC-PUB 1726 (1976).

6)

L.M. Lederman and B.G. Pope, Phys. Rev. Letters

7)

li.B. Green et al., Nuovo Cimento 29A, 123 (1975).

8)

D.C. Hom et al., Phys. Rev.

9)

Columbia-FNAL-Univ. Illinois Collaboration;

316 (1970).

K.J. Anderson et al.

Letters~,

765 (1971).

~,

1236 (1976).
B. Knapp et al.

10)

A. Staude, invited contribution.

11)

BNL-Yale-FNAL Collaboration;

12)

MPEI Moscow-IHEP Serpukhov Collaboration;

13)

T. Appelquist, invited contribution.

14)

SLAC-LBL Collaboration;

15)

Pavia-Princeton-Maryland-UC San Diego-Stanford Collaboration.

16)

B. Wiik, invited contribution.

17)

M. Chanowitz, invited contribution.

18)

Dubna-Berlin-Budapest-Prague-Sofia Collaboration;
invited contribution.

19)

V.I. Zakharov, invited contribution.

20)

M. Schwartz et al., BNL preprint (1976).

21)

B. Pontecorvo, invited contribution.

R.K. Adair et al.

N.Z. Anisimova et al.

R. Schwitters, invited contribution.

I.A. Savin,

HADRON NUCLEUS COLLISIONS IN THE COLLECTIVE* TUBE MODEL t

G. Ber1ad
Department of Physics
Technion, Haifa, Israel
Introduction
High energy particle production data measured with nuclear
targets exhibit several outstanding features. These features
appear to contradict the naive picture in which the projectile is
assumed to propagate "on mass shell" through the nuclear medium
by a succession of collisions with single nucleons in the target.
We can distinguish two classes of such phenomena:
(1)

A class of features that look like the corresponding particle nucleon scattering data. Among these we find the
KNO (1) scaling function (2) and the normalized rapidity distribution
dN - 1 da
dy = cr dy

for large rapidity values (3). This would seem to indicate that
particle nucleus scattering looks more like a single step
collision of the projectile with chunks of nuclear matter,
rather than a chain of successive collisions with individual
nucleons inside the target.
(2)

A second class of features exhibit production of particles

with energies much higher than the kinematical limit set by a
process where the projectile collides only with a single nucleon at one time. This is a big effect and cannot be explained by the Fermi motion of the nucleons inside the target
nucleus. Such an energy excess has been observed by
683

684

G. BERLAD

A.M. Baldin (~) in an experiment at Dubna. Using various nuclear targets a large number of fast pions were detected in
the backward direction. Their kinetic energies exceeded by
factors of two and three the kinematical limit set by multistep collisions with single nucleons inside the target.
Baldin called it "the cumulative effect".
On these observations we have based an intuitive picture of
particle nucleus collisions. This plausible picture claims no
deep understanding of the underlying detailed dynamics; however,
when it is applied to the various phases of hadron nucleus interactions, all the main features of the experimental data are correctly reproduced. Some of its predictions will be discussed in
detail hereafter.
The Collective Tube Model (CTM) (2)
Let us consider a high energy projectile incident on a nuclear target of atomic number A
The CTM then assumes that:
(a)

The projectile collides simultaneously with

cleons situated along its path through the
present purposes it is syfficient to assume
on the average, about A ~ nucleons in such
energy available for particle production in
tube COllision, E (A)
is given by
av

a tube of all nunucleus. For our

that ther~ are,
a tube ( ). The
the projectile-

(1)

Here S is the C.M. energy squared, El b is the projectile

lab. energy, and m is the mass of a a nucleon. We shall
refer to eq. (1) as the energy rescaling effect.
(b)

In its respective CM-frame the projectile tube collision resembles a projectile-nucleon collision at the same available
energy.

Assumptions (a) and (b) lead to the following results. Let

us denote by a h1
the cross section for a given final state "i"
in hadron-nucleoR scattering, and by a~A the respective cross
section in hadron-nucleus scattering.
Then,
i

ahA (E lab )
inel
ahA
where

inel

is

'"

a hn (A
inel
a hn

1/3

Elab )

the total inelastic cross section.

(2)

HADRON NUCLEUS COLLISIONS

685

Using the experimental observation that the effective area of the

nuclear target can be parametrized as

(3)

w?ere ~ and ~ are projectile dependent constants (6) , we der1ve two types of sum-rules (7,8) :
a relation connecting "hn" and "hA" processes,
(4)

and a relation,. connecting "hAl" and "hA2" processes,

These sum-rules allow us to compare experimental results from

different nuclear targets at different energies.

Discussion of

the High Transverse Momentum Data (8)

J.W. Cronin and his collaborators have observed that for large
transverse momenta the A dependence of cross sections for inclusive production of hadrons from nuclear targets can be represented
in the form (9)

where a,(Pr L2...!.. Such a strong A-dependence is hard to understand unless collective effects are present. We claim that such
a behaviour is natural in the CTM.
It has been observed both at the ISR (10) and FNAL (11) that
high energy and fixed but large transverse momenta the incluS1ve proton proton cross section for particle production shows a
remarkable increase with energy. This increase is reflected
onto hadron collisions by the energy rescaling effect, and may be
a~

686

G. BERLAD

1.2

0.8

1.2

1D

o.e
1.4

1.2

1.0
0.8

tf

+
2

+ ++ +
p

PT (GeV/c)

Fig. 1. - The effective A-exponent, a(PT)'

Experimental points
are from reference (9). The continuous lines are the
CTM results. Predictions for proton production are not
presented because part of the detected protons in the
final state does originate from the fragmentation of
the recoiling tube for which the CTM does not apply.

687

HADRON NUCLEUS COLLISIONS

.the cause of the steepened A-dependence observed in these

cesses. Let us see how this happens:
We first parametrize the

pp

pro-

data as follows:
(7)

where

B(PT)

is implicitly energy dependent.

Then we insert eq. (7) into eq. (4) and obtain:

The effective A-dependence. aCp). is therefore determined

by two contributions. One is the nucIear target area responsible
for the a
exponent. and the other. the ~(p )/3 exponent.
reflects. p via the energy rescaling effect. the growth of the
corresponding "pp" cross section with energy. The overall Adependence is compared with experimental data of reference (9) in
figure (1).
A most sensitive test for the necessitr of energy-rescaling
is provided by the high-PT particle ratios (2). The trivial target geometry dependence
(be it A2~ or AI) is cancelled in
these ratios. and. in the absence of energy rescaling. particle
ratios from all nuclear targets at a given lab. energy and p
should coincide. If. on the other hand. energy-rescaling iI
present. then. as particle-ratios from different nuclear targets
at fixed PT are plotted as a function of the available energy

all points should follow a single universal line.

The failure of theories with no energy rescaling to explain
the p/TI- ratio is clearly evident in figure (2). (Other particle
ratios are not shown because. within experimental errors. they
seem to be energy independent. thus they cannot distinguish energy
rescaling from non-energy rescaling models.)
We now plot the same data as a function of the available energy (eq. (1)). This is shown in figure (3). In this representatio~
within experimental uncertainties.the data points do indeed show
a tendency to align along a universal line.

G. BERLAD

6BB

0.1

.------~,.--------r------....,

Ie

H2
02

Be
TI
+

Pl7r-

PT = 4.58 GeVle

0.01 ' - - - - - -......1 - - - - - ' - - , - - - - - - - 1

200
300
400

P'ab (GeV Ie)

Fig. 2. - The p/~- ratio (12). In the absence of energy rescaling
all points of equal PI b-values should reduce to a single
point, regardless of a atomic number. This is clearly
incorrect.

The CTM and New Particle Production (13)

The production of very massive particles can occur only above
thresholds of extreme energy. However, if Hydrogen targets are
replaced by nuclear targets and energy rescaling does take place,
these thresholds are significantly lowered. To be explicit, if a
certain threshold is at El b = E
in a hadron proton collision,
the ener~ rescaling tells a us that it should be lowered to
Elab = A ~Eo when a nuclear target of atomic number A is used.
A reductlon of almost an order of magnitude in the required
energy can thus be gained by replacing a Hydrogen target by a
Uranium target. This gain, superimposed on the enhancement due to

HADRON NUCLEUS COLLISIONS

1 t-

I:iII-

689

l-

\ Jit I \

K-l7r

t t

If

Hz
x Oz
A Be

"f t I PI
0.1

+i

TI

f K+l7r+

--

l-

I+ t
I-f ftip

tf

PT
I

P I7r-

=4.58 GeV Ie
J

0.01 ~2..&.0--3..&.0--4""0--5""0--6""'0--7""'0----I

Eav (GeV)
Fig. 3. - Particle production ratios (12) are plotted as a function
of the available energy. Note that the p/~- data
(lowest data points) tend to align along a single line,
in accord with the CTM prediction. All other particle
ratios are energy independent, and thus insensitive to
energy rescaling.

G. BERLAD

690

i-

1---

A'

'"E

10-A

100
-Plab (GeV/c)

10
+ -

1000

Fig. 4. - a(pp + (J/W + ~ ~ ) + X) as a function of Plab extracted

from various nuclear targets (Iq). The circ es correspond
to an AI-dependence with no energy rescaling. The triangles present the CTM predictions. The lines were drawn
to guide the eye. Note the shift of the threshold in the
CTM to higher Plab - values.

HADRON NUCLEUS COLLISIONS

691

the larger nuclear target area. makes it possible to measure processes with high thresholds and small cross section.
The effect of threshold shift for inclusive J/~ production in
the CTM is shown in figurffi (4) and (5). The experimental data of
ref. (1~) has been used as an input to equation (4) and (5) to predict. respectively. inclusive J/~ production off Hydrogen and off
Uranium. At large energies. far above the threshold. the gain is
only marginal. Hence. at El b = 20 GeV/c energy rescaling turns
the "no go" process of J/~ ~roduction on Hydrogen to a "go"
process on Uranium. whereas above El b = 100 GeV/c it only enhances
the effective A dependence from A2h ato Al . This enhancement was
confirmed by recent FNAL data (15) as demonstrated in Fig. (6).
Finally. what does the CTM tell us about the exotic W-boson
production process? If we use as input a theoretical calculation
of W-boson production in proton-proton collisions by E.A. Paschos
and Ling-Lie Wang (16). then for a Uranium target we obtain from
the CTM the cross section that is shown in figure (7). The threshold shift is enormous. and on the level of cr U
~ 10- 33 - 10-3~
cm 2 this cross section should be measurable p ~i~h a 10 3 GeV/c
proton beam (provided ~ ~ 70 GeV).
What About Nucleus Nucleus Collisions?
There has been recently a continued improvement in the heavy
atoms ionization techniques. Therefore. it is of interest to find
out what could be gained by scattering high-energy. highly ionized
nuclei on each other.
The answer is rather surprising. If in particle-nucleus collisions we gain a factor of A1/3 in available energy. in a fully
ionized nucleus (A 1) - nucleus (A 2) collision. the gain is approximately (Al.A2)~3
Moreover. whereas in particle-nucleus
scattering only one tube is active in any given event while the
rest of the nucleus acts as a spectator in nucleus-nucleus
scattering. very many tube-tube collisions occur simultaneously.
and particle production rates are strongly increased.
The results of a preliminary study of the nucleus-nucleus
scattering problem. and the enormous effect of many simultaneous
tube-tube collisions are presented in figures (8) and (9).
Figure (8) shows the charged multiplicity ratio

For comparison one should recall that

p.U(A = 239)

~ 2 (6).

692

G. BERLAO

pU- (JIlJ!- f'+ f' -)+X

1-

i---

A'

eTM

10- 12

100

10

Fig. 5. -

P lob ( GeV Ie)

+ -

1000

cr(pU + (J/~ + ~ ~ ) + X) as a function of p

The notation and input data are that of figu!~b(4).

Note the shift of the threshold in the CTM to lower
Plab - values.

HADRON NUCLEUS COLLISIONS

693

n A- (J 1'/1-",+ ",-)+X
Plab =

300 GeV/c

A 0."
N

(,)

10lIZ

10-IS

I
10

--FNAL
---CTM

102

10'

+ -

Fig. 6. - a(pA + (J/~ + ~ ~ ) + X) as a function of atomic number

for PI b = 300 GeV/c. The triangles represent the predictio~s of the CTM. Error bars reflect the ex~erimental
error of the input data taken from reference (1). The
dashed line is the best fit to Aa (A > 9) of the CTM
results. The solid line is the best fit to Aa of the
experimental data (15). A closer value to the CTM result
has been reported by Professor A.J.S. Smith at this
school.

G. BERLAD

694

pU - W +

-.-

,/

/'

--- CTM
- - CT ....

10- 31

--CT ....

OJ:A

. ........ -

,. .".,/~
.

E
u

;"

//
/

/'

./

10-32

10-33~/__~__~__~__~__~____- L__-L~~__~

10

lOt

s/m~
Fig. 7. - Prediction of the CTM for W-meson production. The
input is a pp + W + X calculation of reference (16).
The L.H.S. curve represen~ the CTM result, the
central curve the results of the assumption

pp

cr A 'V Acr

and the R.H.S. curve the results of

695

HADRON NUCLEUS COLLISIONS

A 1 + 70 - ."c

12

+X

10

..
-

c
c
a: 6

10

20

Fig. 8. - The normalized charged multiplicity for nucleus(Al = 70)

- nucleus(lO < A2 < 240) collisions as predicted by the
CfM.

This tremendous growth in multiplicity that originates from all

over the rapididty range is shown in figure (9) where the normalized
rapidity distribution
dN

dy

1
a

da

dy

is plotted for "pp (1,1)", "pA (1,70)" and "AA(70, 70)" coIl isions.
Evidently, these results are stimulating enough to justify the
channeling of at least a small fraction of the efforts aimed at
constructing more powerful accelerators into advancing the techniques of heavy ion acceleration with existing machines.

696

G. BERLAD

bl

>.
'V 'V

-Ib

1+70

Fig. 9. - The normalized rapidity distribution as predicted by

the CfM:
- lower curve, Al = 1 and Az = 1. (pp) is used as input;
- middle curve, Al = 1 and Az = 70 (p-Emulsion) ;
- upper curve, A1 = 70 and Az = 70 (Emulsion-Emulsion).

697

HADRON NUCLEUS COLLISIONS

A Comment on Inelastic Processes with Deuterium Targets (17)

Deuterium targets have been used at high energies mainly for
two purposes:
(i)
(ii)

to study neutron cross sections

to test theories of particle-nucleus collisions, where
the target nucleus is relatively simple and well
understood.

The Glauber Theory provides us with a good description of

elastic scattering and diffraction dissociation of the Deuterium
target. These processes, originating mainly from grazing collision of the projectile and the Deuterium, probe only the fringe
of the target. In this peripheral region the tubes are "very
short", and the probability of hitting both proton and neutron
collectively is extremely small. CTM effects are, hence,
negligible.
Inelastic collisions,on the other hand, do probe the
central nuclear-matter distribution in the Deuteron. It can be
shown (l~ that for such processes there is a probability of
about 20% to hit a tube containing both neutron and proton.
This non-negligible collective effect has to be properly handled.
There are experimental indications, e.g., that for some final
states the data shows a positive shadowing. Such a behavior is
unexplainable in a naive Glauber model where the projectile propagates "on shell" and the shadowing is always negative. In the
CTM, however~ it is a unique consequence of the model assumptions
(see ref. (1) and fig. (1) therein). More complicated Glauber
type models can be formulated by including inelastic intermediate
states, but these are by no means unique. Different input
assumptions will yield different predictions for the type and size
of the shadowing effect. It has thus been proposed (17) that the
CTM should be used to extract neutron cross sections from high
energy inelastic processes with Deuteron targets.

Summary and Conclusions

There are many other effects which are successfully explained
by the CTM. These, however, require a probabilistic treatment of
the nucleon content of the struck tube (5).
In order of increasing subtlety they are:
(1)

The dependence of the scaled charged multiplicity,

the target atomic number (2)

RA , on

698

G. BERLAD

(2)

The dependence of the KNO scaling function on the target

atomic number (2).

(3)

The dependence of the scaled charged multiplicity and its

dispersion on the number of fast proton (presumed to be
fragments of the recoiling tube) (2).

( 4)

Shapes of rapidity and pseudo-rapidity distributions as

functions of energy, atomic number and fast protons (18).

(5)

Photon-Deuteron interactions (17).

It is rather amazing that such a naive picture of partic1enucleus scattering does explain so wide a field of high energy
phenomena of extreme diversity. It is, therefore, our conviction that whatever the ultimate theory of high energy particle
nucleus scattering will, eventually, turn out to be, collective
interactions will certainly be there.

Acknowledgement
We would like to thank Professor A. Dar and Mr. Y. Afek for
their valuable help in preparing this summary.

References

We thank Professors O. Kofoed-Hansen and H. Lipkin for

suggesting the use of the word "Collective" rather than
"Coherent" in the present context.

This research was performed in collaboration with Prof. A. Dar,

Dr. G. Eilam and Mr. Y. Afek.

1.

Z. Koba, H.B. Nielsen and P. Olesen, Nucl. Phys. B40, 317

2.

G. Berlad, A. Dar and G. Eilam, Phys. Rev. 013, 161 (1976).

3.

W. Busza et al., FNAL preprint, June 1976.

4.

(1972) .

A.M. Baldin et al., Yad. Fiz. 20, 1201 (1974) (Engl. Transl.

lQ, 629 (1975)).

5.

A more sophisticated approach is possible, whereby a probability is calculated to hit a tube of exactly j nucleons.
Physical quantities are then properly weighted with these
probabilities and summed over the j's from 1 to A. See,
e.g., reference (2).

6.

W. Busza, Proceedings of the VI International Conference on

High Energy Physics and Nuclear Structure, Santa Fe, June
1975, and references therein.

HADRON NUCLEUS COLLISIONS

699

7.

Due to the different CM systems of different projectile

tube collisions, variables which are not invariant under
boosts in the beam direction, such as e.g. the rapidity,
must be treated with special care. This treatment is not
displayed in eqs. (4) and (5). For details see reference (8).

8.
9.

Y. Afek, G. Berlad, G. Eilam and A. Dar, "Scaling Laws for

Inclusive Production of Hadrons in High Energy Particle
Nucleus Collisions", Technion Preprint PH-76-l2.
J.W. Cronin et al., Phys. Rev. 011, 3105 (1975).

10.

See, e.g., Professor Valenti's report in these proceedings.

11.

See, e.g., Professor Cronin's report in these proceedings.

12.

We would like to thank Professor Cronin for suggesting this

test, and for providing us with his new data, which has
been presented at this school.

13.

Y. Afek, G. Berlad, G. Eilam and A. Dar, "Cumulative Enhancement of J/1JJ Production in Hadron Nucleus Collisions",
Technion Preprint PH-76-24.

14.

S.C.C. Ting, Proceedings of the International Conference on

High Energy Physics, Palermo, Italy (1975).
J.J. Aubert, Proceedings of the XIth Rencontre de Moriond,
France (1976).

15.

M. Binkley et al., "Dimuon Production on Nuclear Targets",

FNAL Preprint, June 1976.

16.

Emmanuel A. Paschos and Ling-Lie Wang, "The Quest for WtS",

BNL Preprint, BNL-2l048.

17.

A. Dar and Tran Thanh Van, "Inelastic Interactions with

Deuterons at Very High Energies", Technion Preprint PH-76-6l.

18.

Y. Afek, G. Berlad, G. Eilam and A. Dar, "Rapidity

Distributions in Hadron-Nucleus Collisions", Technion
Preprint PH-76-48.

PRODUCTION OF DDIDONS BY PIONS AND PROTONS AT FERMILAB *

A. J. S. Smith
Department of Physics
Princeton University, Princeton, NJ 08540 U.S.A.
INTRODUCTION
In this lecture I shall describe an experiment in progress l
at the Fermi National Accelerator Laboratory (FNAL) to study the
production of ~+~- pairs in hadron-hadron collisions. So far, we
have used beams of 225 - GeV n+, n- and protons incident on carbon
and tin targets for most of the data. In an earlier run, a 150 GeV
positive beam was used with a beryllium target. The results I report of the 150 GeV run are final; those at 225 GeV are preliminary.
The goal of this experiment is to carry out, as completely as
possible, a measurement of ~-pair production over a wide range of
kinematic variables and particle types, using a spectrometer with
good resolution and very large acceptance. Only such a comprehensive experiment can measure differential cross sections without resorting to model-dependent assumptions. There are many interesting
aspects of these measurements: Among them are the study of the production mechanism of the new particles 2 J/~ and ~', as well as a
search for other states decaying to ~-pairs. I shall discuss in
detail the features of J/~ production by n+, n-, and protons, and
compare them with the production of the more common vector mesons
pO, W, ~, and with non-resonant ~-pairs in neighboring mass regions.
The non-resonaQt pairs are also of great interest; for example, in
1970 Drell and Yan 3 proposed parton-antiparton (qij) annihilations
as their source, and since then many predictions of yields have been
made by numerous authors. Virtual bremsstrahlung has been put forth
as another possible source of non-resonant pairs. 4

701

702

A. J. S. SM ITH

To produce ~-pairs with proton beams by quark-antiquark annihilation all antiquarks must come from the sea. On the other hand,
pion beams contain valence antiquarks and hence present a more
straightforward probe with which to test_these models' l Also, because the valence antiquarks of n+ and y have charge 3 and respectively, a cross-section ratio of 4 is predicted for ~-pairs
produced by n+ mesons versus n- on an isospin symmetric target.
Thus by using n+ and n- beams of the same momentum incident on a
carbon target, our experiment can test the Drell-Yan mechanism in a
way not too dependent upon details of a particular model.

The experiment is also sensitive to processes resulting in

"multimuon" final states (states having more than two muons). Such
events could originate in associated production of J/W pairs, or of
a J/W along with a charmed particle decaying promptly into a muonic
final state. We have searched for multimuon events, finding none
above background, as explained in what follows. This negative result puts quite stringent upper limits on associated production
processes.
Finally, one of the most compelling motivations for this experiment was to ascertain the source of the single prompt leptons,
first observed in high-energy p-nucleus collisions at Serpukhov s
and FNAL. 6 (Prompt leptons are those not attributable to n or K
decays.) These measurements were all made at high transverse momenta and near 90 in the center-of-mass system. Since then other
experiments 7 have detected prompt leptons over a wide range of
kinematic variables, and have found the ratio of the yields to be
approximately ~/n z 10- 4 , and quite insensitive to the kinematic
region of the experiment.
Our goal is to see if these single leptons can be accounted
for by lepton pairs. (We actually study only muons -- there could
be different sources of electrons.) We obtain the pair contribution to the inclusive single-muon yield by measuring the differential cross-sections for all ~-pairs, and then integrating over one
muon in the pair. The conclusion, as described below, is that our
measured ~-pairs at 150 GeV are sufficient to explain the prompt
single muon measurements in the kinematic range of our experiment.
When our 225 GeV data have been completely analyzed, they should
give an even more comprehensive measurement of the pair contribution to the single-~ yield, because of greater statistical power
and because of the wider range of acceptance for the higher momentum
beam.
Before discussing the experiment in detail, it seems appropriate to define the variables of ~-pair production. Table I shows
the variables we use, as well as the range of each accepted by the
spectrometer. It is seen that we have virtually complete coverage

703

PRODUCTION OF DIMUONS

for all ~-pairs produced in the forward hemisphere in the center of

mass system.
Table I
Variables to Describe the Production of

- (p+ + p )

p
-L

= Transverse

~-Pairs

.211 GeV < M

~~

momentum of pair

e * - Helicity angle of
decaying state

~'s

from

$ 12 GeV

o<

o<

p-L::: 4 GeV / c

o<

Icose * I < 0.8

< 1

EXPERIMENTAL CONSIDERATIONS
To obtain a large acceptance detector for our experiment, we
have modified the Chicago-Cyclotron spectrometer, constructed for
~-p scattering experiments by a Chicago-Harvard-Oxford group, whom
we specially acknowledge. This detector, located in the muon laboratory at Fermi National Accelerator Laboratory, as shown in Fig.
1, is fed by a secondary beam from the "neutrino" target. The beam
is very long (~ 1 Rl1), a good feature for ~-p scattering, but less
than optimal for hadron experiments, in that even at 225 GeV/c about
43% of the kaon flux is lost by decays.
For this experiment the highest possible beam momentum is
desirable; not only is there more energy available to produce new
objects, but the folding forward in the laboratory of the reaction
products gives one a larger detection efficiency for a given magnet
aperture. Unfortunately, the pion flux decreases drastically once
the beam momentum exceeds roughly half the momentum of the primary
protons. This fact determines the maximum useful beam momentum.
For our first run, the accelerator operated at 1 300 GeV, so we chose
Pb
= 150 GeV, obtaining a n+/p ratio of ~ -3. Last winter in
earn
our second run, when the primary energy was 400 GeV, we put Pb
225, at which n+/p.~}. So far we have used beryllium, carbon;am
and tin targets, with beams of n+, n , and protons, the intensi~ies
being ~ 2 x 106/burst for protons, < 5 x 10 5 burst for n+ and n .

A. J. S. SM ITH

704

,..." "

Booster

.- "

/"
----15'
::: ----Bubble
'-::'::::Muon
Chamber
Counter
v Facility
Spectrometer

--1.---/'

-~~.......:::::::::----=::::=---'-'-':"'::'-.......=::::::!:~-~.:-===l!~
---- -:=:-v

Main
Accelerator

\"
30
Bubble
Chamber

Primary Beams

Interna~

Target
Area

Counter
Facility

Secondary

Beams

Targets

Fig. 1. Plan of the Fermi National Accelerator Laboratory.

detector is shown as the muon spectrometer.

Our

The beam was defined by three scintillation counters upstream

of the detector, two of which were separated by a dipole magnet to
sweep out debris from interactions along the beam. To identify pions
and protons individually, three threshold gas Cerenkov counters
averaging about 70 meters in length were placed in the beam. The
construction was very simple: the existing beam pipes between transport elements were filled with helium gas to a pressure just below
the threshold for Cerenkov radiation from protons. (The operating
pressure was about .2 atmospheres absolute at 225 GeV/c.) In each
counter, light from pions was focussed by a spherical mirror
through a quartz window onto an RCA 3l000-M phototube, which has
excellent ultra-violet sensitivity. Information from the Cerenkov
counters was recorded for each event, allowing us to accumulate n+
and proton data simultaneously. This feature is very desirable in
that many systematic effects, such as changes in magnet currents,
beam intensity, detector efficiencies, to name a few, cancel out in
a comparison of proton-induced and n+-induced events.
The spectrometer is shown in Fig. 2. Interactions took place
when the beam, typically 3cm x 3cm wide, struck the target, usually
~ .25 interaction lengths thick.
(The last beam-defining counter
is shown as T3 in the figure.) To detect the rare ~-pair events in

705

PRODUCTION OF DIMUONS

SPARK CHAMBERS
LEAD

HADRON
ABSORBER

CHICAGO
CYCLOTRON
MAGNET
Fig. 2.

HOD()SCOPE~S

P
HODOSCOPE

The Spectrometer.

the presence of overwhelming hadronic processes, we absorbed the

hadrons before a substantial fraction could decay by placing a 2.2
meter thick steel hadron absorber 1.3 meters downstream from the
target. Thus only muons emerged into a large cylindrical magnet
(the former Chicago-Cyclotron), 2.2 meters in radius and 1.27 meters
in gap height. Downstream from the magnet, the muons had to penetrate an additional 2.5 meters of iron, after which they were detected by a large scintillation-counter hodoscope (shown as P in
Fig. 2). The apparatus is charge symmetric - hence TI- - induced
pairs were measured simply by reversing the polarity of all beam
magnets.
The trigger requirements were as follows: 1) One and only one
beam particle and no "halo" particles within a time interval of
~ 100 nsec.
2) > 2 particles leaving the target, as determined by
pulse-height in counter T4 situated just downstream of the target.
3) > 1 particle in the G hodoscope. 4) > 2 hits in the P hodoscope.
(To-reject single ~'s accompanied by lo~energy electromagnetic
showers, the P hodoscope was mounted flush against the downstream
side of a 20cm-thick Pb wall. Additional rejection was accomplished
by requiring the muon pulses to be in non-adjacent counters.) 5)
No count in TS' a 7.Scm-square counter downstream of the magnet to
veto beam muons. 6) For most of the 22S-GeV/c running, we also required two non-adjacent hits, either vertically or horizontally, in
a hodoscope (shown as J in Fig. 2) placed at the downstream edge of
the hadron absorber. This requirement, while reducing the trigger
rate from background and low-mass pairs, had a completely negligible
effect on the detection efficiency of pairs with masses> 1 GeV.

A. J. S. SM ITH

706

Hadron Absorber

Fig. 3. Details of the upstream configuration of the detector.

Variables are defined in the text.
Approximately 70% of all triggers were clean ~-pair events.
The rest, mainly single muons showering into the P hodoscope, were
easilS eliminated by the analysis. The trigger rate was about 1
in 10 beam particles.
The particle trajectories were measured by multi-wire proportional chambers (Mv.rPC) upstream of the magnet, and by wire
spark chambers downstream, where instantaneous rates were always
lower than IDS/sec. The MWPC configuration is shown in Fig. 3.
The unique feature of our experiment is the chambers in front of
the absorber, which were used to measure the directions of the ~'s
before any significant mUltiple scattering occurred. These chambers
enabled us to obtain good mass resolution, reconstruction efficiency, and background rejection even at the lowest ~-pair masses.
For the data reported here, a 4-plane module SScm x SScm in area
was used, consisting of an X- and a Y-measuring plane, and also a
U and a V plane at 4So to X, Y. For data we are still analysing,
an additional module with 3X and 3Y planes was used to determine
more precisely the vertex in the target. To follow the ~'s as
they scattered through the absorber, another ~lPC module (XYUV) was
placed half way through the steel, where the hadron showers had subsided to an extent that the average hit multiplicity was about 4.
To eliminate the effect of the multiple scattering in the absorber from the momentum determination, we remeasured the muon trajectories before they entered the cyclotron magnet. Eight MWPC
planes (4X, 4Y), each 1m by 1m in area, were used. To complete the
momentum analysis, we measured the trajectories, after they had been
deflected by the magnet, with 12 planes of capacitive read-out wire
spark chambers 2m by 4m in area, and 8 even larger magnetostrictive
planes 2m by 6m. The planes alternated U X X V, with U and V

PRODUCTION OF DIMUONS

707

tilted 15 with respect to X, to provide narrow-angle stero

"views".
Event reconstruction downstream of the absorber was straightforward, since all the chambers had in general better than 90%
efficiency, and only about 10 percent of the events had an extra
track, (usually an accidentaD through the spectrometer. A typical
mass resolution

of

1.6% (standard deviation) was obtained.

Accidental tracks, as well as any hadrons penetrating the absorber,

were eliminated by requiring all ~ tracks to point within resolution to struck counters in the P hodoscope. A small contamination
from beam muons was eliminated, with no loss of good events, by
rejecting any event in which the laboratory momentum of either
muon exceeded 88% of the beam momentum.
The most difficult part of the analysis was reconstructing
the events upstream of the hadron absorber. The procedure by which
the tracks emanating from the target were matched to those downstream of the absorber was tested by a Monte Carlo simulation.
Clean ~-pairs were generated and transported through the spectrometer. Hadron tracks and extra hits from real data were superimposed, and the reconstruction program then was asked to find the
~-tracks.
For masses above 500 MeV a large and uniform efficiency
was obtained for a typical K of 0.3. For a mass of 325 GeV the
efficiency had only dropped y a factor of 2. The rejection of
background from muons produced in the absorber was measured in a
similar manner, this time the ~-pairs being generated within the
first interaction length of the absorber. We found the rejection
power to be greater than 15 at a mass of 300 MeV and to rise sharply
with increasing pair mass to at least 100 by 1 GeV. (This factor
is indpendent of the factor of 50 or more against primary interactions in the absorber obtained by demanding at least two particles
in the counter T4 in order to trigger.)

Dead-time in the readout electronics of the MWPC's upstream

of the absorber limited their use to maximum incident beam fluxes
of ~ 106/pulse. To study ~-pairs with masses ~ 1 GeV, we were agle
to turn these chambers off and run with fluxes as high as 3 x 10 /
burst. This was possible because the MWPC's downstream of the absorber provided good enough direction measurements to separate background ~-pairs produced in the absorber from pairs of 1 GeV mass or
more produced in the target. As shown in Fig. 3, because of multiple scattering the trajectory measured by these M}~C's alone differs
from the true trajectory by an angle 00 and a displacement ox. (A
similar situation pertains to the y-coordinate.) Muon pairs produced in the target were separated from those produced in the steel
absorber by a Xl procedure. Specifically, 00 , ox, 00 and oy were
varied to minimize the X2 of each of the two xmuon tratks for two

A. J. S. SM ITH

708

assumptions: First, that the ~ carne from the center of the target; second, that it carne from a point 1 interaction length into
the absorber. Excellent separation was obtained even for masses
below 1 GeV. In the expression for X2 , the uncertainties in and
correlations between 08 and ox, 08 and oy were calculated for the
multiple scattering of x the absorbet, from well-known formulae in
Rossi's book. s The mass resolution obtained by this method was
typically

~:::

.04, mainly due to the uncertainty in opening

angle from multiple scattering. This is perfectly adequate to

study the interesting features of both continuum and resonant ~
pairs.
FINAL RESULTS OF ~-PAIR PRODUCTION BY 150-GeV/c TI+ AND PROTONS
In this run, a bery~lium target was used with positive beam
to study the reactions (TI ) + Be + ~+~- + X. Figure 4 shows the
mass spectra of all
p ~-pairs with x F ~ 0.15, after correction
has been made for the efficiency of the spectrometer. The acceftance calculation assumed a uniform angular distribution in cos8
for muons from the decaying pair. For ~he p-w mass region, our
analysis gives a distribution dN/d(cos8 ) = 1 - (0.07 0.24)
cos 28*, consistent with the assumption used in the acceptance. In
any case, the only effect of a 1 - standard deviation decay asymmetry would be to produce a uniforma 5% change in the over-all acceptance. The measured width of the p-w peak is 100 MeV, narrow
enough to indicate a substantial w contribution. 9
A striking feature of both pion-induced and proton-induced
spectra is the prominent continuum present at very low masses,
falling steeply by about 500 MeV. The observation of this continuum has important implications for the single direct lepton question, so it is important to be sure that is not caused by hackground
of one form or another. The most obvious source of background, the decays of uncorre1ated TI'S or K's, is eliminated because 99% of the observed pairs have muons of opposite charge
(~+~-).
Secondary processes in the shield are negligible because
of trigger and reconstruction rejections. Other sources which have
been estimated quantitatively include photoproduction of ~-pairs in
target and absorber, secondary p-w production and decay in the absorber, and even the associated production and subsequent decay of
K+ K- pairs. All background sources combined produce less than
10% of the observed low-mass pairs, so we conclude the signal is
real.
There are some very plausible sources of this prompt low-mass
continuum. The two most obvious are the Da1itz decays of nand w,
which,a1though they have not been observed experimentally, have
reliably calculated branching ratios: 10

PRODUCTION OF DIMUONS

>

:e'"
N

;;;...
Q

~103

UJ

>

t:
N.t
~

(J')

I-

PION INDUCED
X F >0.15

PROTON INDUCED _104

XF >0.15
~

XF > 0.15

I/)

709

'"

(!)

~. ,~

2
:e

UJ

"0

"-

UJ

I-

::x:
!:2
UJ

0.6

1.0

10

~_--,-_~Jill
1.0

2.0

MfLfL

3.0
(GeV)

'--'_-L----L-_.L----L-----"_...L.

0.2

1.0

2.0

3.0

Fig. 4. Effective Mass Distributions, 150 GeV beam, corrected for

detector efficiency, for x F > 0.15. a) Combined TI+ and p induced
events with M < 1.2 GeV/c. b) Proton induced events, logarithmic
scale. c) piog~induced events, logarithmic scale.

+ -4
r(n ~ ~ ~ y)/r(n ~ yy) = 7.5 x 10 ; r(w ~ TIo ~+~-)/r(w ~ TIo y) =
5.5 x 10-4 . If we assumed all the low-mass continuum were due to
n decay, the inclusive n production cross-section required would be
11 mb/nucleon, or about 10% of the TI inclusive cross-section. This
is entirely reasonable - although no measurements have been made at
low p~at high pJ-the ratio of inclusive cross-sections 11 of nand
TIois iQ~ 0.5.
(0

If we assumed that Po and w production are equal

0 ), then the process w ~ ~+~-y would explain 13% of the

Po
w
continuum.
The dependence of ~-pair production upon the kinematic variables of Table I was studied by first dividing the data into 6 mass
intervals. (The dependence upon particle type, TI+ or p, is studied
as well.)
In Fig. 5 is shown the dependence upon longitudinal
momentum, where EdO/dxF is plotted vs. x F . In each mass interval
TI+ mesons produce more ~-pairs at large x F . It appears that the
differential cross sections for TI+ and p approach equality at
x F = O. Also, in Fig. 6 is shown the de1endence of the differential
cross sections upon PJL, in the form p~ do/dP-L' Here no significant differences appear between ~- and p-induced spectra. The
data of Fig. 5 have been fitted to the forms EdO/dxF ~ (17x)C and
PJL- l do/dp~ ~ e-beL respectively. The lowest p bin in Fig. 6 has
not been included in the fit. The results of t~e fits are given in
Table II, along with normalization constant obtained from the parametrization E d 30/d 3p = A(l - xF)e-bp~. As described in the pre-

710

A.J.S.SMITH

"'.\:"

10

M<0.45

"

t"

\{\t

:3

:3
C

"

"" t''i.
'"" '-.t,

'\1"

\1--, , t

\!'\1

\'11
~r.y

0.1

Q)

0.65<M<0.93

~,.

-,I\..

\)',

IJ)

>:.

0.45<M<0.65

I~'

~'

..Q

:1.. 10

>Q)

-"-,+~, L.

LL

~I~

1\r'LI

lwlt

IN'

~\

0.1

0.01

0.2

0.4

0.6

0.8

~t:tt f, !" f

0.2

0.4

0.6

XF

0.8

t "'t ",

~~'

1\

Fig. 5. Invariant cross-sections vs. Feynman X.

the fits of Table II.

0.2

0.4

0.6

0.8

1.0

The curves are

vious section, the high-mass data, mainly J/~ production, were obtained without the front t1WPC's. We verified that the normalization of this data and the rest, where the MWPC's were used, agreed
to within 10%. Other results listed in the table include the crosssections per nucleon obtained by using the dependence upon the
atomic mass number A reported by Binkley et al.: 12
a(A) 0:: AO.67 for H < 1.13 GeV, 0:: AO.8S for 1.13 < H
< 2, and
1111
1111
1
0:: A
for H
> 2 GeV. As will be discussed below, we have confirmed
1111

0.93 - 1.13

IV

VI

p-w

0.65 - 0.93

III

2.7 - 3.5

1. 13 - 2.0

Continuum

0.45 - 0.65

II

Conti nuum

Continuum

Cont i nuum

Continuum

2.7 - 3.5

Continuum

Conti nuum

0.21 - 0.45

VI

0.93 - 1.13

IV

1.13 - 2.0

p-w

0.65 - 0.93

III

Continuum

Continuum

0.45 - 0.65

II

Source

Continuum

GeV

flass

0.21 - 0.45

Regi on

16050
16048

x 10 2

8127

8953

1.47.44 x 10 2

2.57.36

3.22.4

3.61.40

9420

4.901.5

370 11 0
22044
960190

6.23.7
3.91.3

3512
1.72.38

113
~024

4814

3711

1 .331.0

1.73.44

6.52.2

106

216.4

7021

370 130

41082

1.34.14

18037

4.31.16

780 160

3.41.70 x 10 3
1. 46. 30 x 10 3

1 . 92. 25

4.64.24

1250500

3.31.1

167

4313

8325

20060

510150

620220

5.07.25

4. 30. 33

nb

> 0

1550620

xF

4.02.80 x 10 3

38376

1.6.6

145

2.94.32

1660330

6.02.5

3916

3.78.80

Pion Production

2.08.26

3.41 .85

155

299

8617

37275
12738

22044

18537

34070

960190

800160

1470300

nb

nb

6620

4.06.40

2. 79. 12

4. 34. 21

6.03.22

x F >0.15

2.20.44 x 10 4

36 12

250100

. 55. 11 x 10 3

3.93.28

3.79.09

4.69.95 x 103
1.83.40 x 10 3

x 10 3

4.58.14

9.031.8 x 10 3

1.06.21

4.63.15

Proton Production

b
(GeV/c)-l

2.67.53 x 10 4

A
nb/GeV 2/c 3

Cross Section/Nucleon

Cross Section
/Nuc1eus
xF >0.15

TABLE II
Results of fitting the Lorentz-invariant cross-section per
Be nucleus to the form Ed 30/d 3 p = A(l - xF)C exp(-bp~) for different mass
regions. The x F and p.projections of the data were fit independently.
Calculation of the integrated cross-sections is described in the text.

"C

'I

en

o
s:
c

"T1

o
z
o

(')

o
c
:::!

:0

712

Fig. 6. Differential cross-section

are the fits of Table II.

A. J. S. SM ITH

dcr/dp~/p~

vs. PJL.

The curves

in our 225 GeV experiment the linear A-dependence for high-mass

Finally we give total cross-sections for x F > 0 for the
various mass regions. Note that we cannot resolve
pO and w, so
have assumed their production to be equal, as suggested by the
width of the peak.
~-pairs.

PRODUCTION OF DIMUONS

713

AN APPARENT SOLUTION OF THE SINGLE LEPTON PUZZLE

As mentioned in the introduction to this lecture, an important goal of our experimental program was to find out whether or
not ~-pair production can account for the anomalous direct muons
observed by numerous authors in p-nucleus collisions. In none of
these previous experiments, however, was there a significant accep'
tance for detecting a second muon, should it have been there. By
gathering up all the pairs, then, we can simply see how much they
contribute to the yield of single ~'s. It has already been estimated by several authors 13 that ~-pairs from the known resonances
pO, w, ~, and J/~ were insufficient, explaining at most about 30%
of the single ~'s. These estimates were made difficult by the fact
that the data was obtained by various experiments differing vastly
in almost every respect. By contrast, in our experiment we calculate the single ~ yield from all pairs over our large kinematic
region, and compare with single-~ measurements from the same target,
beryllium. In this way many systematic uncertainties are made negligible. llt
To include" all significant ~-pairs in our calculation, we had
to extrapolate our cross-sections down to xF = 0 using a parametrization of the measured data for x F ~ 0.15. First, we assumed
the cross-section Eda/dx F ~ da/dy to be flat in the region 0 < x F
< 0.15. Second, we parametrized the data to the form da/dx F =
A(l - xF)C and then extrapolated to x F = O. These two methods
agree to within 10%. If one used the (probably wrong) extrapolation Eda/dxF = A(l - xF)C down to x F = 0 one obtains larger crosssections because the rapidity plateau has been ignored. The ~ara
meters used in these fits are given in a recent publication,l and
are omitted here. The contribution for each of the sources in
Table II was calculated separately. The mass dependence of the
continuum was parametrized as exp(-SM ) within each mass interval.
Two independent calculations, one by ~~onte Carlo techniques, the
other by a numerical integration, agreed perfectly.
The sensitivity of our calculation to the p~dependence of
the pair spectra was studied also. Because our pair measurements
extended out past P-L of 2 GeV/c, and virtually all single ~'s of
P.L~ 2 GeV were found to come from pairs withint this P...L region, we
therefore do not depend on extrapolations.
Figure 7a shows our calculated single-~ yield at xF = 0 as a
function of p~, indicating separately the contributions from
vector mesons and the continuum. The TI- cross-section on beryllium,16 multiplied by 10-4 , is shown for reference. It is seen
that vector mesons canngt account for the observed single ~ yield
(which follows TI- x 10- ), especially as PjLgoes below 2 GeV/c.
However, with the addition of the continuum, quite good agreement

A. J. S. SM ITH

714

Vector Mesons

(0 )

(b)

Continuum

en

3.0

II

't)
::II

c:

iJ'

>

2.5

,M 2.0

~
.D

.....t::

c:

':t,

..,

Q.

1.0

0.5

0.1 0

Pr (GeV/c)

Vector Mesons
plus Continuum
---- Vector Mesons only

1.5

.....

..,b

Fig. 7. a) Invariant cross-sections vs. p~for single ~'s produced at x F = 0 by the ~-pairs measured in 150 GeV p-Be collisions.
The ~- cross-section is also shown, scaled by 10- 4
b) Comparison
of existing single ~ data with the ~-/~- ratio calculated from our
pair measurements. ( Denotes Ref. 5 ; . and 0, Ref. 6; A ,
Bintinger et al., Ref. 7).

is seen. Figure 7b shows a comparison of our calculated single ~'s

from pairs with other experimental measurements of ~+/~+ and ~-/~-.
The agreement is good, indicating that pairs can account for the
observed single ~ yields. The hatch marks show the uncertainties
in the calculation of ~-/~-, which are mainly due to the extrapolation of our pair cross-sections to x = O. (In our 225 GeV data,
the extrapolation, and hence the sysfematic uncertainty, will be
significantly smaller.) The dashed curve in Fig. 7b shows the
single-~ contribution from vector mesons alone.
Single-~ yields
for. other values of x F are shown in Fig'. 8 as a function of P..L'
There have also been several measurements of single ~'s in
the forward direction. 17 ,18 In these experiments, all ~'s are accepted which have PJL ~ 600 MeV/c because of multiple scattering.
To compare with them, we have summed, at a given value of x F ' all
single ~'s with p~ < 600 MeV/c coming from pairs. The results are
shown in Fig. 9. In particular, Fig. 9(b) shows once again that
pairs can account for most of the single ~'s and that for small
values of x F of the single ~, the continuum becomes the dominant
source. An experiment by Kasha et al. 17 has also found that single

715

PRODUCTION OF DIMUONS

Xj?O.60

10

Fig. 8. Cross-section for inclusive single-w production from all

sources, as a function of p~for fixed x F .
w's in the forward direction arise mainly from pairs, but proposes
a strong source of W pairs (in addition to pO, W decays) with a
mass of about 900 MeV. We definitely do not see such a source.
The basic conclusion from our measurements is, then, that Wpairs can account for the yield of single ~'S for PJL ~ 2 GeV at
x F = 0, and also in the forward direction. Particularly in the forward direction, the continuum pairs are a significant source of
single W's. It should be mentioned that there are no single-~ measurements, because of experimental difficulties, in the kinematic
region (x ~ 0, p~ ~ I GeV/c) where the largest contribution from
decays of charmed particles is expected. Therefore it is still possible that a substantial contribution from the decays nO ~ Kwv
exists. 19

A. J. S. SM ITH

716

,,

Pr <600 MeVlc (0)

Vector Mesons
+
Continuum

",,,,

.a
co:

-11-1
><
........
b

-,:J

"",,

1.4
><
I

...

1.2
1.0

~0.8

(b)

0.6

,,

,,

0.4

10

Pr<600 MeV/c

1.6

"
Vector Mesonsf',
",,,
Only

-,:J

1.8

0.2

0.1 0.2 0.3 0.4 0.5 0.6

XF

Fig. 9. a) Single-~ differential cross-section vs x F in the forward direction (integrated over all p~ < 600 MeV/c). b) Comparison
of the ~/'IT ratio measured in Ref. 17 (.) and Ref. 18 C.) with
the yield calculated from ~ pairs. The solid line and hatch marks
show the yield from ~-pairs of all sources. The dashed line is
the yield from vector mesons only.

PRELIMINARY RESULTS:

~-PAIR

PRODUCTION AT 225 GeV/c

At the higher beam momentum of 225 GeV we have a usefully

large acceptance for all pairs produced in the forward hemisphere,
and in our second run, which ended this April, we have collected
about 20 times as much data as in the 150 GeV sample .1 ust described.
In addition, we were able to+use both positive and negative beams,
which allows us to compare 'IT and 'IT- produced pairs. Except for a
one-week run with a tin target to measure the A-dependence, we used
exclusively a carbon target to exploit its isoscalar nature in interpreting the comparison of pairs produced by 'IT+ and 'IT- mesons.
In all the results I shall now present, only the MWPC's downstream
of the hadron absorber have been used in the analysis. Therefore
I shall discuss only those pairs with masses> 1.5 GeV, where as
explained above the upstream MWPC's are not needed. The complete
analysis is in progress and will be reported within the next few
months.
The general features of the high-mass

~-pair

spectra are

717

PRODUCTION OF DIMUONS

400

>

Q)

300

lO

"-

(J)

z
W
>

200

100

4.5

2.0

2.5

3.0

3.5

4.0

MJ.L+J.L- (GeV)

Fig. 10. ~fass Spectrum of all

uncorrected for acceptance.

TI

and p induced events at 225 GeV,

apparent from Fig. 10, in which all the data analyzed so far, about
85% of our exposure, are shown. He observe a sharply falling continuum out to about 2.7 GeV, where the large peak from the J/~ sets
in. The width of this peak shows the mass resolution of the spectrometer. A shoulder is seen around 3.7 GeV, the mass of the ~'. In
what follows, we shall discuss the properties of these events in
detail, comparing and contrasting the resonant and non-resonant ~
pairs. Finally, the results of a search for events having 3 or
more ~'s will be given.
PRODUCTION OF THE

J/~

AND

~'

The sample of J/~ events obtained in our 225 - GeV run is the
most comprehensive of any experiment to date. The more than 2000
events give adequate statistical power to determine precisely the

A. J. S. SM ITH

718
0)

P+(!n)-fC+X

..

>

~ 300

IZ

225 GeV

Q
......

en

(;:)+(~n)-fL+fL-+X

225 GfN

200

1&.1

>
1&.1

3.4 4.0

3" 40

5.0 6.0
~~(GeV)

2.0

I!? >

3.0

4.0

5.0

6.0

2.0

5.0

4.0

3.0

6.0

M~~(GeV)

5.0

9.0

6.0

MfLfL (GeV)

30

ffi:lzo
>8
1&.1-

ALL 225 GeV DATA

10
3.5

4.5

5.5

6.5

7.5

8.5

9.5

MfLfL (GeV)
Fig. 11. Mass distributions uncorrected for efficiency. a) pinduced events, b) n+ and n events combined, c) All events above
3.35 GeV. The inserts in a) and b) magnify the ~I region.

dependence upon kinematical variables of the differential cross

section for J/~ production, over the wide range of spectrometer acceptance. The use of n+, n-, and proton beams gives additional insight into the production dynamics. The raw data for H
> 2 GeV
are shown in Fig. 11. Part a) shows all p-induced evengM, part b)
all n-induced events. In the inserts the scale is expanded for
M > 3.35 GeV, and in part c) of the figure all events from n and
p~~with masses M > 3.35 have been combined. An apparent enhancement appears in ~~all cases at M = 3.7 GeV, which we interpret as
~I production.
Although there are only a small number of ~I candidates, they can be used effectively to get the approximate level of
~I production relative to that of the J/~.
To get as large a sample
as possible we have combined carbon and tin data, assuming that J/~
and ~I production have the same dependence on A. We have performed
a very simple background subtraction assuming a straight-line continuum, obtaining 6 n-induced events, and 4 p-induced events. Using
the same x - and p-L-dependence for ~I as for J/~ we calculate the
average defection efficiency and estimate roughly the ratio of ~I
to J/~ production, finding:

719

PRODUCTION OF DIMUONS

~ 3 x 10-3

Proton-Induced:

Ba{~')

Pion-Induced:

Ba{~') Z 7
10- 3
Ba{J/~)
x
.

Ba(JN)

These values are lower than those reported by Snyder et al.,2o who
guoted, at y = 0 (or x = 0), for protons on beryllium, at 400 GeV,
lBda{~')/dy]/[B.da{J/)/dy] ~ 0.02. This suggests that the ~'
cross section is increasing faster with energy than that of J/~ in
the range 20 <
< 27 GeV, although more work is needed to
settle this issue.

;-s-

Figure 11{c) shows that only a few events were observed in the
mass range from 4 to 10 GeV, where the average detection efficiency
was? 12%. There is thus no evidence for resonances to the level
of this exposure, where each event corresponds to 'V 10-35 cm2/
nucleon.
To obtain differential cross sections for J/~ production, the
events in the mass range 2.7 <}1
< 3.5 GeV were binned in ~, Pl'
and cose*, and then corrected fo~ the spectrometer acceptance.
This efficiency was calculated by a Monte Carlo simulation, which
included effects of energy loss, multiple scattering and bremsstrahlung, as well as the transport of the particles through the
spectrometer to the final hodoscope. To obtain absolute crosssections we divided by the effective luminosity, corrected for ab90rption of the beam in the somewhat thick targets. We collected
approximately half of the positive~beam carbon data with a 7.5 cm
thick target, the other half with a 12.5 cm target. No evidence,
either in absolute rate or in energy distribution of the events,
was seen of secondary production, so we have combined the data from
the two targets in what follows.

The differential cross-sec ions are shown in Fig. 12 as a

function of x F ' in Fig. 13 of PJL' and in Fig. 14, of cose*. The
cross-sections are expressed as nanobarns/nuc1eon, using a linear
dependence upon the atomic mass number A, as observed by Binkley
et a1.12 in neutron-nucleus interactions. (The result of our own
measurement of the A-dependence of total cross-sections is given
below. )
The xF-dependence of the cross-section is, qualitatively, independent
of target type, and much broader for ~-induced events
than for those induced by protons. The data have been fitted to the
form -da = AU-x) b , as suggested by parton models; the results of
dX F

A. J. S. SM ITH

720

(~:) +C-J+X
P

L..fLfL

+
c:

o
G

'" 4.0
~

...c:

th

CI

.a
o

c:
CI
c:

-100
"-

"0

'b
"0

:l

1.0

...,t

4.0

I
0.4 0.6

XF

TT- INDUCED
7T+INDUCED
P INDUCED

0.8

Fig. 12. Differential cross-sections B(J+~~).dcr/dxF for J production. Fits have been made to the form (l-x)b.
the fitting procedure are given in Table IlIa. The fit to the TI - C data is not good, being dominated by one high bin, but aside from
that, the fits give a reasonable description. The values for TI+
and p induced events are in good agreement with our earlier results
at 150 GeV/c.
Next, 'the distributions in p1 were fitted to linear and quadratic forms, respectively:
2
-dt:J...
dcr
-gp...L
Ce
and B(J/1jJ+~~). - 2 = Fe
dP...L
As seen from the table, either parametrization is adequate, and
there are no striking differences among the various reactions.
In Table IIIc are given the results of fitting the relative
distributions in cosG* of Fig. 14 to the form

721

PRODUCTION OF DIMUONS

Table III
Properties of the Differential Cross-section for J/~
Production at 225 GeV. A linear dependence on A has been used to
obtain cross-sections per Nucleon.
a)

Dependence on x F :

Beam

dX F

A(l-x)b nanobarns/nuc1eon.

Target

X /dof

22.31.8

3.980.20

11+

27.73.3

2.620.22

3.5

11

39.15.0

2.150.16

5.2

Sn

25.83.4

3.940.30

4.1

Sn

32.26.7

2.330.32

1.7

p
+
11

b)

3.6

~
2
(units of ---2 are nanobarns/nuc1eon/(GeV/c) )

Dependence on PL:

dP..L

LINEAR

QUADRATIC

B'do !dpi ~ Ce

-dp

.1

B'do!dp2

Fe

1.

2
-gp.i

Beam

Target

//dof

//dof

p
+
1T

9.10.9

1.97.08

1.8

3.50.3

0.800.06

1.2

20.93.3

2.03.15

.5

7.91.2

0.880.12

1.6

1T

13.92.2

1.58 .13

1.5

6.20.8

0.590.06

1.0

Sn

9 .81. 5

1.86 .14

1.3

4.10.6

0.780.09

0.95

Sn

13.24.5

1. 56. 25

0.95

6.21.4

0.640.14

1.0

p
11+

c)

Dependence on cosS * ~~

we have fitted the form do/d(cosS*)

Beam

Target

//dof

p
11+

-0.068+ .24

3.14

0.18H.37

0.64

11

0.36 .44

1.3

Sn

-0.947.31

2.0

Sn

11
P

11

,1T

-0.35 .45

1.45

C+Sn

-0.2670.19

2.2

C+Sn

0.26 . 25

1. 34

(1+acos 2 S*)

722

A. J. S. SM ITH

(~:\+C-J + X
p)

+ +

'-fL+fL10

'"c...
CI

.0

-0

:l
:l
.., 0.1

0.1

.2

Fig. 13.

ducti~n

7r+ INDUCED

P INDUCED

1.0 1.8 2.6 3.4 4.2 5.0

pI

7r+ INDUCED

INDUCED

(GeV/c)2

Differential cross-sections B(J+WW)'da/dpjl2 for J proas a function of

P~.

The lines shown are fits to the form

Fe -gp..L

da

d(cos8 * )

A(l + acos 28 *).

In no case is there compelling evidence of any significant polarization of the J/~.

To find more precise values of a, we have therefore combined
all proton data (Fig. 14) and all TI data (Fig. l4g), obtaining
a = - 0.26 0.19 and a = 0.26 .25, respectively. Further
s~udy is required to seeTIif the small polarization in the proton
events is significant.

PRODUCTldN OF DIMUONS

723

roOr-----------------------,

I
~

r-

20 0

a:i

45 0

400

:::::>

;.~!J.: I I
~~~

LLO_)J--L-L~__L_~~~~

~ 0

"0

IIII I

0.25

0.45

COS

600

0.25

045

COS

e'"

III

400 I

200

I.~~

10 o c)
0.05

0.05

0.65

e*

0.85

L~~

P+C-J+X

20 0

7T-+C-J +X

b)

.......
b30( I

50

III

C::: tO
0111
<[
11

Cb

100 e)

f)

III

II

pt(~n)-J+X
I.~~

400
0.65

085

300

200
100

I IIII

7T +(n)--J + X

L-~~

g)

.05

.25

.45

.65

.85

Fig. 14. Relative differential cross-sections da/(cos8 ) for

production.

J/~

Values of the total cross-sections (times branching ratio)

for J/~ production are given in Table IV for the range 0 ~ x F ~ 1.
The errors have been inflated somewhat arbitrarily because
of
the preliminary nature of the data, and should be about twice as
small in the final analysis. To obtain the cross-section per
nucleon, one must know the dependence on atomic mass number A, which
is found by comparing the actually measured cross-sections per
nucleus for carbon 'and tin. Assuming the form a(A) ex: A(j, \OTe find
a ex: Al . OS 0.10 a ex: Al 03 0.15. This is consistent with a
1T

linear dependence, which we have therefore used in obtaining

Table IV.

A. J. S. SM ITH

724

Table IV
Total Cross-Sections for J/t/J Production, 225 GeV Beam
Beam

Target

a(xF>O)
nb/nucleon

868

4.130.37

390

8.1 l. 2

350

9.8 1.4

Sn

331

4.850.7

Sn

150

9.501.4

p
n

n
Note:

No. of
Events

A linear dependence on atomic mass number has been used.

The dependence of the cross section for J/~ production on

beam energy is quite interesting, as seen in Fig. 15, where
dO',
dXF

and
x =0

dO',
dy

vs::::

for proton-induced events are plotted versus

y=O

I2Hp ELab .

I t would be better to plot the total cross-

section a vs ;-S, but most experiments 21 so far have only been

sensitive near x F = 0, where there is the simple relationship between x F and the center-of-mass rapidity y:
da/dyl ::: (2H/;-S) (da/dxF ) I . One sees a large increase from
o
Q
threshold up to ;-s = 16.8 GeV (E Lab = 150 GeV) and then a virtual
dO'
r-constancy of d
for v s > 20 GeV. One infers from this a similar behaviour y of the total cross section.
The results of Table IV provide a test of charge symmetry in
inclusive J/~ production by n+ and n-, and also a measure of the
relative effectiveness of nand p in producing J/~'s. First, we
find that the ratio of the n+/n- cross-sections is consistent with
unity:

a+(x>O) = 0 83

a (x>O)

-.

16

In our final analysis, the error should be about half that given
here, because in principle there are only very small systematic
errors in this comparison. Also, as will be discussed below" the

PRODUCTION OF DIMUONS

725

10-31
edCTI
dx 0

._L
__
d~~
....---

x-

8 dCTI

__

Q)

,,

::l

"E

I
I

bl~
~~
CD

..

bl- )(
0

10- 35

~~

~~

JI'

P +A-.J +X

'-JLJL

o.

~.

o.
Ot
X.

BNL Aubert et 01.

SERPUKHOV Anti pov et 01.
FNAL THIS EXPT
Snyder et a I.

FNAL
ISR

Busser 8t 01.

CD

10- 36

20

30

50

./S(GeV)
Fig. 15. Energy dependence of J production cross-sections at
x F = O. The open points show (dcr/dxF)O' the solid points show
(dcr/dy)O' where y is the center-of-mass rapidity. For experiments using nuclear targets, a linear A.dependence has been used
to obtain cross-sections/nucleon.

726

A. J. S. SM ITH

continuum is smaller for n+ events than for n , an effect which will

push the above ratio closer to one. The analysis of this correction is not as yet complete. Secondly, the proton-induced events:
averaging n+ and n- results, we have, for carbon,

a (x>O)
p

a (x>O)
n

4.13

37

-=-8-.9-=-7::--C-:-=-9-=-2 = O. 46 063

Again, this error will be reduced somewhat in the final analysis.

To conclude the discussion of J/~ production by hadrons, I
summarize the most important qualitative features which have emerged:
A dependence

+
linear for both nand
p.

x F dependence

independent of target material, much

broader for n beams than proton beam.

p....Ldependence

independent of both beam and target

material. Much broader than that of
lower mass vector mesons.

Beam energy dependence

da/dyl

Cross-section ratios

increases sharply from

y=O
threshold, becoming virtually constant for ;-s- ~ 20 GeV. Dependence
of total cross section has not been
measured, except at Fermilab energy,
and must be inferred.
For x>O, n +In - consistent with unity,
n's give twice as large a cross section as protons.

STUDY OF NON-RESONANT W-PAIRS

It is very important to learn about the process by which nonresonant, or continuum w-pairs are produced. For example, the parton models make quantitative predictions for the cross sections,
but say little about the transverse momenta expected for the pairs.
So far, all previous experiments have used neutron or proton beams,
so all the antiquarks are provided by the sea, leading to rather
model-dependent predictions for the features of the resulting
spectrum. This is the first experiment to use n+ and n- beams on
an isoscalar target, a particularly attractive experimental situation, in that valence antiquarks are present in the beam. Annihilations of fractionally charged valence quarks in the target with the
antiquark in the projectile should then show a charge assymetry

a(n+C)/a(n-C) = 1/4,

PRODUCTION OF DIMUONS

727

independent of details of the model. Unfortunately this large

asymmetry is diluted by the presence of the isospin-symmetric quarkpairs in the sea. Using recent measurements of structure functions,
A1tare11i, Brandt, and Cabibbo,22 and Farrar 23 have estimated the
effect of the sea, finding the charge ratio to be close to 1 at
low pair masses, and then to decrease slowly to the value 1/4 as
M2/s increases. For example, at M : 2.5 GeV the ratio is predicted to be ~ 0.6.
WW

If)

:::J

Q)

4000

:::J

........

>

Q)

<.9

100

........

..c
c

10

"'0

........

"'0

.4

(0)

Fig. 16. Differential cross-sections d0/ dH ww versus MWW

background has been subtracted.

The decay

A. J. S. SM ITH

728

The mass spectra are shown on a logarithmic scale in Fig. 16.

Decay background, as estimated from like-sign (W+W+ or W-W-) pairs
has been subtracted. In four mass intervals below the J/~ and+two
at higher masses, we now compare the production of dimuons by n
and p and by ~+ and n- on carbon. Figures l7a and l7b show these
cross-section ratios and equivalent ratios for the J/~ mass region.
Of particular interest is the n+/n- ratio, which appears to change
from approximately unity at low mass to about .25 at the highest
masses, except for the J/~ mass region, where it returns, within
errors, to unity. Within the limited statistics, this behaviour
is in agreement with the prediction of Ref. 22.

4.0

--

a. 3.0

2.0

b 1.0

a)

t
\ ~f* 4

If

!5

..

tP
4.0

oj>

P\;+

*+ J
I

b .75 ~

........

.5 ~

b
.25

o
Fig. 17.

b)
I

t
I

Ratios of total cross sections for xt>O vs. pair mass.

729

PRODUCTION OF DIMUONS

When TI+ and proton induced pairs are compared, on the other
hand (Fig. l7a), the J/~ and neighboring continuum regions do not
show any significant difference. In both Figs. l7a and l7b the
p -w and ~ mass regions are also shown, where a charge ratio of 1
i~ expected from isospin conservation.
Other features of the data have been studied in a preliminary
manner by selecting three mass regions having significant numbers
of events, in a region relatively free of resonance signal. The
regions are: 1.5-1.9 GeV/c 2 , 1.9-2.3 GeV/c 2 , and 2.3-2.7 GeV/c 2

1000r------------------------------------------------------,

l.9 c Mf'fLc 2.3 to

2.3c~c2.7

_1000
fI)

:J

~
(.)

:J
C

........

.0
C

LL.

"0

to

........

"0

1.0

1.0

O.t

O.t

7T- INDUCED
A 7T+ INDUCED

0.2 0.4 0.6 0.8

1.0

0.2

O.t

p INDUCED
0.4 0.6 118

1.0

02 0.4 0.6 0.8

XF
Fig. 18. Dependence of pair cross-sections on x F for various
intervals of pair mass.

t.O

A. J. S. SM ITH

730

Figure 18 shows the dependence of the cross-section in these

bins on the variable x F

~ , for p,
---

Is

TI

+
- incident on car, and TI

bon. The basic feature of all these distributions, that the TIinduced events have a slower fall-off with increasing x F than those
of p-induced events, is very similar to what we have observed in
production of the resonances p, w, <, and J/l~. The data have been
fitted to the form B'dO/dxf = A(l - xF)b; the best fits, of somewhat limited statistical slgnificance, are summarized in Table Va.
Figure 19 illustrates the dependence of the differential crosssections do4p 2 upon the transverse momentum p of the dimuon. Exponentials boib linear and quadratic in ~ have been fit to these
spectra, the results appearing in Table Vb. The quadratic fits are
shown in the figure, although either hypothesis gives an adequate

fit.

tOO,,--,---,---,--,--,,---,--,--.--,

1T-INDUCED

1T+ INDUCED

P INDUCED

(/)

::::J

CI)

"0
::::J

tOO

c:

~
~

fO

fO

~
..c
c:

1.0

O.t

0.4
2.3<MJLJL<2.7

Py2 (GeV Ic)2

2

Fig. 19. Differential cross-section vs. PL

of pair mass.

for various intervals

PRODUCTION OF DIMUONS

Table V

731

Fits of Differential Cross-sections of Continuum Dimuons.

(P)
:~

Note:

Beam
p

11

11

b)

+_

+ X

~ ~

All cross-sections are per Carbon Nucleus.

B dxp

= A(l~x) b

Mass Region

Dependence on xp:

a)

+ C+

de'

nanobarns/nuc1eon
2
X /dof

1.5-1.9

26032

4.9.3

1.9-2.3

5l21

3.8.7

.82
.86

3.4

2.3-2.7

4511

5.2.6

1.5-1.9

32192

3.5.6

1.9 2.3

3418

1. 7.8

.88

2.3-2.7

2829

2.31.6

.63

1.5-1. 9

17078

2.5.7

1.9-2.3

52:'- 23

2.4.8

2.3 2.7

291l

1. 7. 7

Dependence on PT:

1.4

1.5
.16
1.7

Bdo
2
(units of dpt2 are nanobarns/(GeV/c )/nuc1eus)

Linear
2
B'do/dP T = Ce-dPT

Quadratic
2
B.d~/dpa = Fe-gPT
T

Beam

Mass
Region

2
X /dof

1.5-1.9

38l30

2.9.1

3.9

1.9-2.3

1l033

3.3

.75

2.3-2.7

2514

2.4.5

.85

11011

1.4 .1

.3

+
11

11

2
X /dof

844

1.2.O5

1.9

243

1. 3.1

7.51.2

.98.12

.55
.75

1.5-1.9

586106

3.3.2

.88

1.9-2.3

6630

2.4. 5

.3

216

1.1L25

.001

2.3-2.7

8960

3.5.8

.44

176

1.7.4

.5

1.5-1.9

26960

2.7 .2

2.7

677

1.11.1

1.1

1.9-2.3

2812

1. 6. 4

1.2

2.3-2.7

3218

2.0.5

.5

133

.65~.15

.5

113

.81'.18

.3

732

A. J. S. SM ITH

3.0

Fig. 20. Average p~of muon pairs vs. pair mass, uncorrected for
spectrometer efficiency.

Figure 20 shows an interesting feature of ~-pair production.

We have plotted the avera*e transverse momentum <p~> as a function
of the dimuon mass, for IT and p-induced events. (The analysis of
the IT- sample is not yet completed.) No discontinuity at all is
apparent between <p~> of resonant and non-resonant dimuons. Instead, one sees a slow, smooth increase of <p~> with increasing
dimuon mass. As described in Professor Cronin's lecture, within
rather large uncertainties this smooth increase continues out to
~-pair masses of 10 GeV.
We have also studied the differential cross-section as a *
function of cose*; a hint of polarization of the form 1 + cos 2e
is suggested by the data, (see Fig. 21) but the statistics are insufficient for a firm conclusion to be reached. Finally, in Fig.
22 we show the dependence on M on a log log scale. Good agreement
with a power law is seen.
~~

733

PRODUCTION OF DIMUONS
2.3<M~~

t.5<M~~ < 1.9

10

t.9<M~~<2.3

15
10
5

III f 1!ll _
rr

10
5

10
V'I

"0

II

rr-

tlltII

0
10

IIII!!JI

rr+

iff

! ! I!

rr -

rr+

........

! t

10

b 10

"0

5
0

! f I II

< 2.7

5 l
0

.2

.4

.6

.8

f 1p

iii
.2

.4

COS

.6

eft

.8

Fig. 21.* Decay angular distribution for continuum dimuons. The

angle 8 is the helicity angle of the decay muons in the eM frame.

In an upcoming experiment to run at Fermilab in s~ring '77,

we shall measure these cross-sections to a level of 10- 7 cm 2 /
event, approximately 100 times more sensitively than the present
exposure. Because the results of our experiment and others are so
far very suggestive of quark-ant i-quark annihilation, one shouid
perhaps take seriously another prediction, that the n-/p crosssection ratio for WW production rises sharply with increasing M2 /s,
reaching a value> 100 for H
= 10 GeV. Our new run will test
this prediction conclusively~W

734

A. J. S. SM ITH

fOO
1T- +

C--fL+fL- X

225 GeV

en

t:

:::>

>
0:::
<t

0:::
~

10

0:::

<t

:E
"0

"-

"0

4.0

4.5

3.0

2.5

3.0

M ( GeV/c 2 )
Fig. 22. Differential cross-section dcr/dM
vs. }1~~, plotted on a
log-log scale. An approximate agreement w~~ a power law is
suggested. All of the region xF>O is included in the cross-section.

735

PRODUCTION OF DIMUONS

SEARCH FOR MULTIMUON EVENTS

The efficiency of the spectrometer for the detection of an
extra muon accompanying a ~-pair is very large. Such extra muons
are expected if, for example, the J/~ is produced in pairs, or in
association with charmed mesons, as is required by the Okubo-ZweigIizuka rule.2~ We have searched all our data - about 850,000 ~
pair events in all - and 2100 J/~ events for any having extra muons.
After performing cuts on the data to reduce to a negligible level
the backgrounds from secondary interactions and muons in the halo
around the beam, we were left with 1229 three-muon events, and one
event with 4 muons, the latter being a double p+~~ event. Only 2
of the 2100 J/~ events had an extra~. This level of extra muons
is completely consistent with the contamination expected from pi
decay, namely ~ 2 x 10- 3 . We thus conclude there is no evidence
at this level for extra muons accompanying either high or low mass
pairs.
The computation of cross section limits is made simple by the
fact that the major losses of single muons in the forward hemisphere
are due to the 8 GeV range requirement imposed by the iron muon
identifiers, to the geometrical aperture of the proportional chambers, and to a halo-suppressing requirement in the analysis that
all muon laboratory momenta must be less than 100 GeV/c. The first
eliminates an elliptical segment in the x F - p~ plane with intercepts at Feynman x F = .035 and P.l = 740 ~1eV Ic. The last eliminates
muons with x F > .43. Within these boundaries our acceptance is
more than
50%. The two accompanied J/~ events are shown by .'s
in Fig. 23, which shows that the extra ~'s are low p~, low x F ' as
would be expected from decay muons. The cross section limit for
J/~ events accompanied by an extra muon is a(J + ~) < 2.5 x 10-3
with 90% confidence. This result rules out,
a(J) for example,
pair production of J/~'s to a limit of 3.5%, within the virtually
complete acceptance for detecting at least one of the 2 muons from
the second J/~. The limits on associated production of charmed
particles are of course dependent on the branching ratio assumed
for their muonic decays, as well as the model of production.
Eventually, we shall calculate limits for the popular models, but
for now I simply state that the limit for a charmed particle accompanying a J/~ is < BR~ x 10- 3 ) , which for the usual assumpmeson + ~
tion of about 0.1 for the branching ratio, says that ~ 5 % of J/~'s
are produced in association with charm. Similar limits over a
somewhat more restricted acceptance have been reported for neutronnucleus collisions at Fermilab. 25

A. J. S. SM ITH

736

4.0
SINGLE MUON ACCEPTANCE

"-

IN THE X F - PT

(!)

><II

3.0

PLANE

= MUONS ASSOCIATED

WITH JlljI's

a.

rn

ID

oo
li?
<

.2 .3 .4 .5 .6

.7 .8 .9 1.0

XF

Fig. 23.

Sensitive kinematic region for the search for multimuons.

CONCLUSION AND ACKNOHLEDGEMENT

The experimental results on dimuon physics are becoming quite
precise. They are generally in agreement with the parton model, but
there is a disturbing feature in that the high average transverse
momenta are not easily explained. It is also puzzling that the differential cross sections of resonances and continuum are so similar
in their dependence on the kinematic variables. Finally, we all
wish to know why there is no evidence of muonic decays of charmed
particles, either as evidenced in copious yields of single W's cr in
extra muons accompanying J/W events.
I wish to thank Professor Zichichi for his kind invitation and
superb hospitality. It is also a pleasure to acknowledge the staff
of Fermilab and particularly the members of the Chicago-HarvardOxford Illinois ~-scattering group who have provided the spectrometer for our work.

737

PRODUCTION OF DIMUONS

REFERENCES

This research was performed at Fermilab and supported in part

by the NSF and ERDA.

The authors of this work are members of the Chicago-Princeton

(II) collaboration: K.J. Anderson, G.G. Henry, J.E. Pilcher,
E.I. Rosenberg, J.G. Branson, K.T. McDonald, G.H. Sanders,
A.J.S. Smith, and J.J. Thaler. In addition G.E. Hogan and C.
Newman participated in the analysis of the 225 GeV data.
Preliminary results were reported in K.J. Anderson et al.,
Phys. Rev. Lett. ~, 237 (1976).

J.J. Aubert et al., Phys. Rev. Lett.

Augustin et al., op cit, p. 1406.

S.D. Drell and T.M. Yan, Phys. Rev. Lett.

G.R. Farrar and S.C. Frautschi, Phys. Rev. Lett.

(1976).

V.V. Abramov et al., reported in the Proceedings of the 17th

Int. Conf. on High Energy Physics, London, 1974, p. V-53.

J.P. Boymond et al., Phys. Rev. Lett.33, 112 (1974); J.A.

Appel et al., Phys. Rev. Lett. 11, 722~1974) .

F.W. Busser et al., Phys. Letters 53B, 212 (1974), D. Bintinger

et al., Phys. Rev. Lett. ~, 72 (1975).

B. Rossi, High Energy Particles, Prentice Hall, 1956.

The p -width observed in p+~~ decay is narrower than in p+TITI

decay o because of the photon propagator in the former decay.

11,

1404 (1974); J.F.

~,

316 (1970).
~,

1017

..

10

C. Jarlskog and J. Pilkuhn, Nucl. Phys. Bl, 264 (1967);

C. Quigg and J.D. Jackson, LBL Report UCRL-18487; C.H. Lai
and C. Quigg, FNAL Preprint FN-296 (1976) .

11

F.W. Busser et al., Phys. Letters 55B, 232 (1975).

12

M. Binkley et al., Phys. Rev. Lett.

13

M. Bourquin and J.M. Gaillard, Phys. Letters, 59B, 191 (1975);

J.W. Cronin, in Proceedings of the International School of
Subnuclear Physics, Erice, 1975.

..

1l, 574 (1976).

A. J. S. SMITH

738

14

For example, the effects of the peculiar A-dependence of

high p~ processes reported by Cronin et a1., Phys. Rev. D11,
3105 (1975) are avoided.

15

K.J. Anderson et a1. , Phys. Rev. Lett.

K.J. Anderson et a1. , Phys. Rev. Lett.

16

T. Eichten et a1. , Nucl. Phys. B44, 333 (1972); W.F. Baker

et a1. , Phys. Letters 51B, 303 (1974) .

17

L.B. Leipuner et a1., Phys. Rev. Lett. 35, 1613 (1975);

H. Kasha et a1., Phys. Rev. Lett. ~, 1007 (1976).

18

D. Buchholz et a1., Phys. Rev. Lett.

19

This was emphasized by P.A. Piroue, Division of Particles and

Fields Meeting, Upton, L.I., October 1976.

20

H.D. Snyder et a1., Phys. Rev. Lett.

21

F.W. Busser et a1., Phys. Letters 56B, 482 (1975); Y.M.

Antipov et a1., Phys. Letters 60B, 309 (1976); also References 2 (Aubert et a1.) and 20-.--

22

G. A1tare11i et a1., Nuc1. Phys. B92, 413 (1975).

23

G.R. Farrar, Nuc1. Phys. B77, 426 (1974).

24

A. Okubo, Phys. Letters, 5, 165 (1963); G. Zweig, CERN Report TH-412 (1964); I. Iizuka, Prog. Theor. Phys., Supp1.
No. 37-38, 21 (1966); D. Sivers, Phys. Rev. D11, 3253 (1975);
R.M. Barnett and D. Silverman, Phys. Rev. D12, 2037 (1975).

25

M. Binkley et a1., Phys. Rev. Lett.

..

R, 803 (1976) ;
R, 799 (1976) .

~,

~,

932 (1976).

1415 (1976)

R, 578 (1976).

739

PRODUCTION OF DIMUONS

DIS C U S S ION
CHAIRNWN:

Scientific Secretaries:

Prof. A.J.S. Smith

Y. Afek and G.J. Tarnopolsky

DISCUSSION
ETIM:
1ihat is the parametrization of the continuum which you added to
the vector meson induced ~-pair production to finally compute the
inclusive single-~ spectrum?

SMITH:
VJe divide the continuum into mass regions, each parametrized
with similar functional dependence. The results of the fit appear
in the table shown.

ETIM:
How sensitive to this particular parametrization
inclusive muon yield?

~s

the computed

SMITH:
The main uncertainty comes from the extrapolation we made from
the lower measured values at x ~ 0.15 down to x ~ O. Fluctuations
arising from the fit parameters are smaller than the ones introduced
by the extrapolation.

PAULI:
From these experiments, can you make any statements about heavy
lepton production? Also, can any of the dimuon experiments be altered
to get more definite statements on the properties or non-existence of
heavy leptons in the 1-10 GeV range?

740

A. J. S. SM ITH

SMITH:
~10st dimuon experiments work because all particles but muons
are absorbed. If the produced heavy leptons both decay into muons,
their contribution would probably be buried under the dimuon yield
from other sources. One should really try the e~ channel.

ZICHICHI:

Heavy lepton production is also suppressed in hadronic reactions

since it is not easy to produce a time-like photon of at least 4 GeV.

YOON:
Can you explain why the ~' bump in the mass spectrum you
measured is so small?
SMITH:

There is a ~' signal, smaller than that observed by

Lederman et al. Our mass resolution is just sufficient to see the
bump. We then subtract a non-resonant background; the subtraction
could be off by a factor of two, but the signal is small. One
could ascribe this to the lower centre-of-mass energy of our experiment. Also, Lederman's sample is restricted to x ~ 0, while we get
a wide range of x values. The signal might be diluted there. Comparing the J and ~' yields, one should note that the branching ratio
B(~' + ~~) is only (0.9 0.2)%.
LITTENBERG:

Is your result inconsistent with the observation of a substantial

signal of direct photons at the Intersecting Storage Ring (ISR)? I
believe Farrar and collaborators have claimed they could explain all
the direct muons on the basis of the amount of direct photons actually
observed. Could you comment on this?
SMITH:

I have not studied the photon experiment carefully enough to

assess its implication. However, the ISR measurements are at a much
higher energy than our experiment. Also, we do see a substantial
continuum contribution throughout the resonance region. Perhaps this
corresponds to the Farrar contribution. Clearly we shall study this.
Anyway, I am sure our result is correct.

PHYSICS PITH THE SINGLE ARM f;PECTROMETEP AT FF.p.MILAB

P. Cutts
Brotffi University, Department of
Provi.dence, Rhode Island 02912

Phv~ic~
n~A

INTRODUCTION
I will descri.he ~ome of the uses of a ne~v spectrometer f aci.1ity now operational at the Fermi National Accelerator Laboratory.
Located in beam H6E of the Meson Lab, the Single Arm Spectrometer
(SAS) is fully instrumented for high energy counter/wire chamber
experiments. I will talk briefly about the spectrometer to ~ive
you a feel for the physics it makes accessible, but mainly I will
concentrate on experiments we've done, are doing, or hope fo do.
To build the Single Arm Spectrometer Facility, the large
group of people indicated below contributed their time, some apparatus, or both. This group formed the collaboration for the
initial experiment, a study of elastic scattering completed in
Fermi1ab Single Arm Spectrometer Group: D.S. Ayres, R. Diebold,
and G.J. Maclay, Argonne National Laboratory; D. Cutts, R.E. Lanou,
L.J. Levinson, and J.T. Massimo, Brown University; J. Litt, CE~
and Daresbury Laboratory; R.Meunier, CE~; B. Gittelman, E. Loh,
and M. Sogard, Cornell University; A.E. Brenner, J.E. Elias, and
G. Mikenberg, Fermi National Accelerator Laboratory; L. Guerriero,
P. Lavopa, G. Maggi, C. DeMarzo, F. Posa, G. Se1vaggi, P. Spinelli,
F. WaIner, E.N. Ane1li, Istituto di Fisica, INFN, Bari, Italy;
D.S. Barton, J. Butler, J. Fines, J.I. Friedman, H.W. Kendall, B.
Nelson, L. Rosenson, and R. Verdier, Massachusetts Institute of
Jechnology; B. Gottschalk, Northeastern University; R.L. Anderson,
D. Gustavson, K. Rich, D.R. Ritson, and G.A. Weitsch, Stanford
Linear Accelerator Center.

741

D.CUTTS

742

1975. Nuch of these results are already published [1], and I will
discuss them only briefly, dwelling on one particular application
of the data: an impact parameter analysis of elastic scattering.
In this same first experiment we obtained data on inelastic scattering in the diffractive and triple-Regge regions, and I will
show some results for the inclusive reactions:
a+p-+a+X
a+p-+b+X

and

where a = n , K-, p and b ~ a. This inclusive study is being

extended by the experiment now running in the Single Arm Spectrometer. A subgroup of the original collaborators, comprised of
people from Bari, CERN, FNAL, NIT, and those of us at Brown, have
added a detector around the target to measure multiplicity and
pseudorapidity distributions in association with a fonlard particle
detected by the spectrometer. I will discuss our plans and progress. Future experiments with SAS include a further inclusive
study, by those of us now running, which will use nuclear targets,
and an experiment to study the non-diffractive reaction n+p -+ K+r.+,
in the forward direction. (The collaboration for this latter experiment will include people from the original group not involved
in the current run.) Finally, I will speak about a proposal which
seems uniquely suited to some aspects of SAS, a search for heavy
long-lived particles.
THE SINGLE ARM SPECTRONETER FACILITY
The SAS Facility is based on a forward arm spectrometer
located at the end of the high momentum (to 200 GeV/c), high resolution secondary beam H6E at FNAL's Heson Lab. The spectrometer
sits at 0 0 to the beam and functions essentially as a fourth stage
to 116E, having focussing as well as bend magnets. A total of 8
Cerenkov counters are used in H6E and SAS for complete tagging of
incident and scattered particle identifications (n, K, or p).
Figure la shows the third stage of the beam line, with detectors
for measuring the momentum, direction, and type of particles incident on the H2 target. To choose the central scattering angle
we pitch the beam in the vertical plane, as indicated in Figure lb;
the maximum pitch angle at present is ~24 mrad, and about the pitch
angle SAS has an angular acceptance ~ 2 mrad (vertically). The
ability to run the beam at 0 0 into SAS proved to be very useful
experimentally. Figure lc illustrates the detectors in the spectrometer. Performance figures for SAS include the following:
[1] Fermilab Single Arm Spectrometer Group, Phys. Rev. Lett. }2,
1195 (1975); D.S. Ayres et al., to be published; R.L. Anderson et
al., to be published.

SINGLE ARM SPECTROMETER

nd

743

x-v

Focus

Hodolcop.1
H2 Target

Trigger
Counter

Oiffer.ntiol
eer.nkov

P Hodolcope

Figure la:

Diagram of the third stage of beam line M6E.

AVB2
(20 foot)

Incident
Beam

Spectrometer
Target

Figure lb:

Angle

Diagram of AVB system, determining scattering

angle setting.

Jaw

Multi Wire

",re.

Muon

B':~X~Y90xo~p"?rCh'm~")i'

Diff.~";"

Cerenkov

Figure lc:

T.; ....

Th ...

Counter

Cerenkovs

Counter

Diagram of the Single Arm Spectrometer, showing

detectors for particle identification, momentum,
and angle determination.

D.CUTTS

744

full acceptance

r.m.s. resolution

0.07%

~str.,

4.5%

~p/p

~p/p, ~O.l

mrad.

With SAS one measures the momentum, scattering angle, and

particle type of a single particle going forward in the lab. Its
primary physics application then is to study high energy scattering reactions, involving a forward secondary. The obvious first
experiment was elastic scattering: the good momentum resolution
allowed a clean separation of elastic peaks--but more practically,
the presence of easily observable peaks allowed the Facility to be
built and tuned!
Sample results of the elastic scattering experiment are shown
in the following pages. Our basic measurement was that of the
differential cross section do/dt for the six elastic reactions
(np, Kp, pp) in the kinematic region:
incident momenta:

50, 70, 100, 140, 175 GeV/c

(s from 94 to 328 GeV 2 )

momentum transfer: .02

~ It I ~ 0.8 (GeV/c)2

These data are plotted in Figures 2a and 2b. The following

figure (3) gives our results for the slopes [2] of the elastic
differential cross sections, as compared with measurements at
lower energies, and with pp results from the ISR. Note that over
an energy range the K+p and pp cross sections exhibit shrinkage
(b increases), np and K-p are essentially flat, and pp antishrinks.
We have integrated the differential cross sections to obtain the
total elastic cross sections, plotted in Figure 4 together with
the total cross sections from Carroll at al. [3], and the inelastic
differences. The np elastic cross sections are essentially flat,
K+p rises slightly, and pp, pp elastic cross sections are falling.
Both the K+p and pp inelastic cross sections are clearly rising,
as are their total cross sections in this energy range.
Elastic scattering is a fundamental hadron-hadron interaction,
and the data are important to a number of models and methods of
interpreting high energy strong interactions. In this talk I shall
describe just one such application, the derivation of a picture of
the scattering process in impact parameter space.
bt+ct 2
[2] The cross sections are well fit by the form e
we plot
the effective slope b eff at t = -0.2 (GeV/c)2.
[3]

A.S. Carroll et al., Phys. Rev. Lett. ]1, 928 and 932 (1974).

..

'tI

.....

' tI

10- 1

~ 10 2
E

(.!)

>4>

...

N lOS

Figure 2a.

'~

~)."
706-&v/c..

x/o.,

~SO~"I...

CK+,P)

0.00

10'

0.30

0.60

~,

' ...............

.........

'-.......

*. .'''-......

..... (,A."k.

....... 1'75"

~Vk.
~
..~ 1'10'

-T(GeV.. 2)

0.90

10. 2 ....
' ,.....~;;-;;-.L....J-:-!-:-L-L...LJ

'tI

*~IO'"

"lifO

....

=" . .

~.....

I:)~.

-::
~

> _ ~ 100~/O)""""""'"
~ 102 ~ \c
,tllk

N... '03 ~*\

E""

ro4~\,

105~
\.....

E(K+,P) -

106~

~~

'

\
\

'x (,a.Vk.

f'f

\,

-T(GeV .. 2)

0.60

"'-,

.....
0.90

\.x\..x
c,.'f, ' "

'x, '1(,0' ' "

-.....~,

'Xx '15" ~

0.30

",

'\.

If

~\..x,.x/,,'"' """

100

'x" lifo

'It

x/o

'X, )LIO)

'"z

'x,

'\

It

" X,

x_

\.

10"
0.00

;; 10

, I\.

'xx

f\ '. . X,~r:..\/((,"\'-~

10 x--

-N 103

K)st

(pi;P)

"'\c SO (,._V/c.

"'.~

~'\

cptp ) -

Elastic differential cross section data for TI+P, ~P, and pp.

0.90

(PI+,P)

0.60
-T (G~V 11112)

0.30

(PI+,P) -

'I
....
til

:0

-I

o
s:
m

:0

-I

(')

"0

en

s:

:0

m
l>

G')

en

"0

~b

.Q

...

>
~

':

'.

(Pr,P)-

(Pl~P)

'-

....

lft.'O'"

'~It'tI"'

.....,

"~

\,~,'''erGw"fc:.
',,-

"~''tO""Ie.

~
*)'.,0.

t,~

Goal(,

70

Figure 2b.

-T(GeV . . 21

(K;PI- (K~PI

' ......70

./0*

*'t,

J\

~~fD)

'''0
.6-e\(,

'a

0.00

10"

0.60

-T(GeV**21

0.30

I,

);1

~ev/,

'~I'l~

'-!.

0.90

~ lK..". I,."~Il
\.
'"
" 1~1011~

~ Irf.. '.

~ 101lE\:,,*.hll/e. ~

10 II! 'a.

41

t~6."1c.

10'

Elastic differential cross section data for

I~~!~--~~~~--~~~~~
0.00
0.30
0.60
0.90

10

I..

101~\'..

I ...

~.

Io'~E .... , ~

~ '~

I04~

l ... .~oc;.."k

P'.

101~

'.

10~ ' \

41

10S_~. " \

(P,P) -

J
I e

~.

"V/~"+,

freVk,

1T

gAg

p, K+p and pp.

-T(GeV .. 21

0.30

0.90

~l~~~~_ _~~~~~~~_ _~

0.00

10~

li)

\~:

"

~l'''t'05:0(0)
t

'I"

~ E\\ \J~ll

70 '\. Xlo"

(is,PI

~ lo3f \ \ ) \ ,
*

.f.
\

loe

.....

"'"

CIl

=l

("')

0.

Q)

CD

>

),

10

4'
10

20

20

,,:'(0,1,0
6f- y
Y

,".1
I

I
II

'

200

~_

200

500

20

1000

(GeV )

iii i ,

10

~--

2000

200

500

This
Experiment

100

K+p

s (Ge'P)

50

.".+ p

Slopes of the elastic differential cross section.

100

pp

.....

Figure 3.

50

100

iii

iii

_ , ... _.

50

-O~

12r-r---r-..,....,-T""T"""'-

......

..,.......

:0

--I
m

o
s:

:0

--I

("")

"'tJ

en

s:

:0

r
m

SQ
Z
Gl

100

ELASTIC
C

200

n+p

INELASTIC

300

,,

'

400

K+p

200

31

31

,' ,

300 400 100

200

pp

300

400

,
,
J

ELASTIC

I t

"

INELASTIC, ,

{I

ELASTIC
t

INELASTIC,

100

?1

15

18

TOTAL

Figure 4a: Integrated cross sections in millibarnes for n+p, K+p, and pp.

Figure

18

i :1

~ 19

TOTAL

.I>-

'l

(')

co

21t

Q)

a::

\I)
\I)

\I)

UJ

-I-0

---:z

.Q

ra

I::

II

lq

- 20[

en

t
TOTAL

Figure 4b:

300

400"~bo

,.

ELASTIC

200

..

100

~-p,

pp.

300 400
K-p, and

200

ELASTIC

pp

J JI I

I NELASTI C

, , J,
,

.1.

33

34

300 400

I f

K-p

INELASTIC

TOTAL

35

36

,
,,

Integrated cross sections in millibarnes for

200

1T-P

ELASTIC

100

INELASTIC

21'

'I
....
'0

:lI

-I

m
m

0
3:

:lI

-I

(")

en
"ll

3:

:lI

Cl

D.CUTTS

750

IMPACT ANALYSIS
From our elastic differential cross section data we make the
numerical transform [4]

to find the elastic scattering amplitude he1 as a function of

impact parameter b. The procedure has two minor problems: 1)
the data is in the form do/dt while what we really want is 1m A(t)
and 2) we have no data for It I > t 1 , where tl is generally 1.2
2

(GeV/c) We assume 1m A = A, and we apply a correction for the

lack of high It I data by studying one energy where our data extends
out to 1.4 (GeV/c)2.
Estimates of the ratio Re A(t)/Im A(t)
indicate that our assumption of a purely imaginary amplitude introduces an error less than 0.5% at all b. Separate analyses
using models of the high t behavior give the same results as those
I will present, except for a small scale shift of the order of
3% at b = O.
Having obtained h e1 (s,b) [= 1m hel(s,b)] as above we use the
unitarity equation
1m h e l(s,b)

-4llh e l(s,b)1 2 + G.1ne l(s,b)

to find the differential cross sections in impact parameter space:

dO tot
--2ndb

= 1m

he1(s,b)

dO ine1
---=::'::':::2:'=' = Gi l(s,b)
ndb
ne
The inelastic "overlap function" Ginel we find simply by subtracting the elastic from the total cross sections, both of \.,hich come
directly from the transform of our data. Thus from our measured
elastic do/dt we have a full description of the b-space scattering.
[4]

H.I. Miettinen, CERN preprint TH.1864 (1974).

SINGLE ARM SPECTROMETER

751

The first results one obtains in this analysis are the amplitudes 1m hel(s,b) for the six elastic reactions. All are nearly
gaussian in b, as expected for transforms of nearly exponential
t-space data. Over our energy range (s from 94 to 328 GeV 2 ) we
find 1m h 1 ~ independent of energy for all b, for the reactions

1/p and K~P. For K+p we observe a slight increase l-7ith energy,
at all b; and for pp w'e see a slight decrease with energy, at
all b. Finally, we find the amplitudes 1m hel for pp decreasing
at small b and increasing at larger b, as s increases. These
results are expected, as the amplitude integrated over b gives
the total cross section:

The absorption probability for a head-on collision ("opacity")

is given by the function Ginel(s,b = 0); our results are shown in
Figure 5. The striking feature is that the mesons are more transparent than the protons, with K slightly less opaque than n;
nip, K-p opacities are approximately constant while the K+p
opacity is increasing by ~2% over our energy range. Note the pp
and pp opacities are decreasing to the apparently constant values
of 0.94 found for pp at the lSR. As mentioned before, the absence
of large t data in our transform could cause a systematic shift
in these results, such that the opacities shown are ~ 2.5% low;
this effect, however, wouldn't change the general comparison between reactions. The following Figure 6 shows for the six reactions the calculated r.m.s. interaction distances defined by

the size of the K+p and pp interaction is increasing with energy

decreasing i~ pp, and essentially constant for nip and K-p. Thus
the change with energy in the total cross section is reflected
directly in the growth in the interaction size rather than, for
pp, in an increase in central opacity.
To examine more closely the energy dependence of the K+p and
pp cross sections, we have plotted the differential total cross
section differences, defined by

..

(.!)

(I)

'-"

en

..0

II

0.92

0.94

0.96

0.98

..

'----....

(!)

(I)

(I)

..0

II

500

pp

= 0).

1000

t.

2000

3000

{t

*pp this experiment

points in ref. 3

Inelastic overlap function Ginel(s,b

GeV

300

K+p

Figure 5.

S In

200

~ K-p

7T+P

100

'f + i #1 t

},. t 144

} ~ 1 tt *
} 7T-P

50

0.74

0.76

Qao

-- 0.a2

=len

()

til

...,

'-I

753

SINGLE ARM SPECTROMETER

.82

.80
.78

.76
.74

.82

.80

.78

"-

c:

.76
.74

K-p

I}

I}

pp

9
~

~ K+p

.88
.86
.84

.82
.80

*
I

100

200

pp
~ ~
~

300

00

S in GeV2
Figure 6.

Calculated r.m.s. interaction distances, in Fermi,

as a function of energy.

D.CUTTS

754

lI tot (b)

dO tot
ndb 2

do

175

- (~)
ndb2

70

A function of b, this quantity gives the difference between the

differential cross sections measured at 175 GeV/c and at 70 GeV/c.
We have also calculated the same expression subtracting 50 GeV/c
from 140 GeV/c data. Since the range in energy covered with each
pair of data (175-70 and 140-50) was similar, a comparison of the
two results is sensitive to the presence of systematic errors.
The values for 1I
(b) we obtained are shown in Figure 7, for pp
tot
+
and K p. Errors plotted are only statistical, but there seems to
be good agreement between the two independent pairings of data.
We show also the differential cross section differences calculated
for the "1/l"P reaction, where to isolate Pomeron exchange we define
the amplitude [5]

Note that 1I

t is negative at small b, for the pp reaction; in

to
other words, the pp differential cross section decreases at
small b and increases at larger b, while both K+p and "I/l"p increase at small b. [In looking at these graphs remember the
total cross section is gotten by integrating with ndb 2 ; the
pp total cross section rise with energy is in full agreement with
such an integration of the data shown.] The following page
(Figure 8) shows the same relations for the inelastic cross
section differences rather than the total. The K+p and "I/l"p are
more or less the same, but there is a dramatic difference in the
small b results for pp. Thus the decrease in the total cross
section in pp at small b is related to the decrease in the elastic
cross section, while the rise at large b (and overall rise) comes
from the inelastic cross section increase.
We have studied the crossover effects by examining do/dt
elastic scattering data for Kp and pip [1]. The idea is that
differences in the particle-antiparticle cross sections isolate
the imaginary non-flip w -exchange contribution. He have examined
the quantity
lI(xp) -

[5]

C. Quigg and E. Rabinovici, Phys. Rev. 13, 2525 (1976).

755

SINGLE ARM SPECTROMETER

0.02
0.0

pp

-0.02

o
Cl

1'15-70
I "to -So

-0.04

'".c
"'0

"'-

....0

"'0

004

<J

K+-

.~hr

0.02

Pa

('15-70
ILfo-50

x ''15"-'70

0.0

b in Fermi

- O. 02 L-...L--_--J..._~:::__--::":--~0.2
0.6
1.0
1.4
1.8
Figure 7.

Differential total cross section differences,

as a function of impact parameter.

D.CUTTS

756

pp

o \'15'-'70
o I 'to -50

0.02

0.0
-Q02

./

i/

-0.04

--

-0.06

,I

f/-vr::. of

pf>

.0
"'0

.b
(j)
C

0.04

o 175-70

-a

<l

140-50

0.02
0.0

b in Fermi

- 0.02 L-...I.-_---1-_ _.L...-----'------"----'

0.2
0.6
1.0 .
1.4
1.8
Figure 8.

Differential inelastic cross section differences, as a

function of b.

757

SINGLE ARM SPECTROMETER

as a function of impact parameter b. The results are shown in

Figure 9: data on the particle-antiparticle differences are plotted
for two energies, as well as the corresponding expressions for pp.
For both reactions we find the difference to be peripheral centered roughly around 1 Fermi, and decreasing rapidly with energy fully
consistent with a simple Regge-pole interpretation.
The impact analysis gives us a measure of the overlap of two
hadronic density distributions; from the pp data we can extract
the proton distribution and then apply the result to 1Tp and Kp
to find the hadronic density distributions for 1T and K. Figure 10
shows the comparison of our results for the Fourier transform of
the hadronic densities with the Fourier transform of the hadronic
charge distribution, as suggested by Chao and Yang [61. Our
results are the lines, and the data points are measurements of the
form factors Fl and GE , for pp [71, and the pion form factor [81,
as a function of t. Our agreement with the model is good, for pp
and 1Tp, and we make a prediction for the Kp form factor.
We have tested in impact parameter space quark model predictions for the amplitudes as a function of b. The comparisons
are generally ~20% off, as is the amplitude relation proposed by
Lipkin [9]:
A(1Tp)

= 21 A(Kp) + 31 A(pp)

To test the relation, one picks particular combinations of the

charge states to satisfy Regge exchanges. Two such combinations
give the following relations:

Rule 1:

Rule 2:

32

a(1T p) +

31

a(1T p)

7l----+----~1---_---

a(K p) +

a(K-p) -

21

t
-

a(pp)

a(K+P) +

a(1T p) +

21

- 1

a(pp)

- 1

a(1T p)

Our results are shown in Figure 11. The relations are good in the
region of 1 Fermi and are worst at large b.
[61
[7]
[8]
[9 ]

T.T. Chao and C.N. Yang, Phys. Rev. 170, 1591 (1968).
L.E. Price et al. , Phys. Rev. D8, 2063 (1973).
c.s. Bebek et a1., Phys. Rev. Dl3, 25 (1976), and G.T. Adylov
et al., Phys. Letters SIB, 402 (1974).
H.J. Lipkin, Phys. Letters B56, 76 (1975).

0.75

1.25

1.50

50 GeV/c
175 GeV/c

1.75

Particle-antiparticle differences as a function of impact parameter.

b in Fermi

1.00

0.50

0.0
0.25

(K-p)-(K+p)

-.061-

Figure 9.

<l

xt) -.02

0..-

. .0
....-

.02

.06

(0)

=len

o
o

til
00

-xb

<J

0.25

0.50

1.00

b in Fermi

0.75

,f

1.25

1.50

50 GeV/c
175 GeV/c
1.75

Particle-antiparticle differences as a function of impact parameter.

0.0

(pp)_(pp)

~o6tlftfl

-.02

.02

Figure 9.

0 ..-

..0

,-

.06

(b)

C.1I
'0

.....

:0

-t

m
m

o
s::

:0

(")

-a
m
-t

C/l

s::

:0

l>

~
Z
Gl

760

D. CUTTS

(a)

0.8

+ FIP

...--

-+..........
a..w

2 G~

{>

0.6

pp at 175GeV

0.4

(!)

...--

---LL-

-+-

a..

0.2

(b)

-+-

..........

----K+ p at 175 GeV

--7r+ p at 175 GeV

0.6

t::LL

0.4

9
0.1

0.3

0.5

0.7

0.9

1tl in GeV2
Figure 10. Comparison of proton form factor (a) and pion form
factor (b) measurements (data points) with calculated transforms
of proton, pion, and kaon hadronic density distributions (lines).

a:

z:

...

..J

:)

CI

Figure 11.

0.0

",--~.....

til

III

CjI

CjI

0."

...........~......

O.U

Q
I

1.1

1.2

I.'

'"
ci

;
I;

;
N

ci

-IN

2.0

I.' I.'
!

...
I

. .J

-r

; 0.0

Ii

III

w
-:
Q D

...

0.0

0.11

D."

1.1

1.2

&fERMI)

0.1

D.'
C.O

CjI

..

ci

III

ci

'"

l~

I:' 1.0

I.'

LIPKIN RULl 2 lEST

Quark model amplitude relations in impact parameter space.

8 (fE""I)

0.1

0.'

LIPKIN RULE 1 TEST

"
0.

:ll

--I
m

o
s:

:ll

--I

(')

."

(/)

s:

:ll

z
G')
r
m

762

D.CUTTS
INELASTIC RESULTS

As part of the initial data taking with the Single Arm

Spectrometer, we recorded events for which the momentum of the
particle detected in SAS was below the elastic peak. Some of
these inclusive events came "for free" in the elastic data taking;
and at most of the elastic settings (a given scattering angle,
determining t) we also took additional inelastic data by lowering
the momentum of the spectrometer in steps up to 25% below the beam
momentum. This data, then, covered the kinematic range

1.0>x>0.75
.03

<

It I

<

0.7 (GeV/c)2

for the five incident momenta

(or up to 1.2
in some cases)

50, 70, 100, 140, and 175 GeV/c.

Note the Feynman variable x is simply determined (to an excellent
approximation, at our energies) by the momenta measured in the lab:

x '" 1

Mx 2 '" Pspectrometer
s
Pbeam

Since we tagged for each event the incident particle type (n, K,
or p) and the particle type in the spectrometer, we took data
simultaneously on 9 reactions:
a+p-+b+X
where a,b = n, K, or p. Together with two beam polarities (in
this first experiment, sign of b = sign of a) we have data for
18 reactions at 5 energies and a number of kinematic points in
the ranges indicated above -- generally referred to as the "projectile fragmentation" region.
How do we study this data? We have used the "triple Regge"
picture (see reviews by Boggild and Ferbe1 [10] and Fox [11]) as
a guide.
In this model one describes the single particle inclusive cross section by a sum of terms involving different combinations of the possible exchanges i, j, k, each with residue
[10]
[11]

H. Boggild and T. Ferbe1, Ann. Rev. Nuc. ScL, 24, 451 (1974).
G.C. Fox, p. 180, in High Energy Collisions, AlP-Conference
Proceedings #15 (1973).

763

SINGLE ARM SPECTROMETER

p
Triple Regge picture.

functions and having t, M2, s dependences given by the appropriate

a(t) 'so With R = Reggeon and P = Pomeron, one 'vrites the terms
expected to be dominant as: PPP, PPR, RRP; their contributions
to the cross section are:
1
M2

(PPP)

1
M3

(PPR)

'V -

do
dtdM 2

'V

'V-

(RRP)

or, using the relation M2 ~ (1 - x)s:

do
dtdM 2

'V

(1 - x)-l

'V

(1 - x)-3/2 s-1/2 (PPR term)

'V

const.

(PPp term)

(RRP term)

We have not as yet made a proper Triple Regge Analysis, using all
terms and the full t, M2, s dependences; what we have done so far
is to use the above picture to help parametrize the data.
We have divided our inelastic events into three categories:
diffractive, inclusive (without quantum number exchange), and

764

D.CUTTS

exchange (of strangeness or baryon number).

"diffractive" are of the type:

The data grouped as

a+p-+a+X
with kinematics close to the elastic, specifically:
0.975

<

<

0.995

2.4 GeV 2 < M2 < 9 GeV 2

I will say only that this data has the characteristics that:

M2 ~
dtdM2

is independent of M2, for fixed t

in all six channels; and, further that for all channels the data
"factorizes" -- namely, the expression above is the same for all
reactions, when normalized by the appropriate elastic cross section.
These results are just those expected if the processes are dominated
by a single term, PPP, common to all channels.
Data of a similar type (particle a -+ particle a) but with a
larger missing mass (M2 > 4 GeV2) we studied separately, including
events over the full range of x (x > 0.75) and .03 < It I < 0.7
{GeV/c)2. There is an enormous amount of data (over 500 graphs)
of the type shown in Figure 12, and at present our analysis has
only fit events to a simple relation suggested by the triple Regge
picture and partly by the data itself. We suppose the invariant
cross section depends only on two terms, one that varies as l/l-x
(like PPP) and one that varies as (l-x). The latter term may be
some combination of PPR or RRP terms, perhaps, but describes the
data well. For each term we allow sand t dependence, writing

+ D (E- 1 / 2 - 0.1{1 - x)
1

A2 e 2t

C t2

{1 + D2 (E- 1 / 2 - 0.1) (~)

1

where E is the incident laboratory beam energy. We have separated

parameters depending on sand t, writing a form suggested by our
2
elastic data (e Bt + Ct ) and typical energy dependence expected
in the Regge picture.

765

SINGLE ARM SPECTROMETER

p+p~p'+X
I Lf 0 G-<t.:vk
j(::'

Cjl(.

I~~~~-L-L~~~~~~~-L~~~~~~

1f+'1-p -+ rr+ '1- X

5"0 G-e v'/e.

I-tJ <: O. '2.

(),~

O. <3
Figure 12.

0.9

Sample inelastic data.

D.CUTTS

766

TABLE I

Channel
1T
1T

p-

Al

Bl

C1

Dl

(mbs/GeV 2)

GeV- 2

GeV- 4

GeV~

113

-8.00 0.30

6.21 0.50

-0.68 0.60

120

-7.76 0.32

5.33 0.50

1.15 1.00

612 30

-8.32 0.16

5.24 0.24

-0.28 0.40

609 50

-8.42 0.50

6.25 0.80

-2.15 1.50

K+

69

-7.66 0.80

6.85 1.00

-1.29 L73

K-

65 14

-7.75 1.30

5.90 1.50

-3.81 4.00
~

A2

B2

C2

D2

2
(mbs/GeV )

GeV- 2

GeV- 4

GeV~

2.30 0.12

-5.21 0.15

1.72 0.20

3.64 0.40

1..2.50 0.14

-5.04 0.16

1.20 0.22

3.39 0.46

6.08 0.36

-7.24 0.14

2.58 0.22

2.69 0.28

6.09 0.70

-6.53 0.60

-0.66 1.50

8.16 2.00

1.30 0.10

-3.64 0.32

0.83 0.44

5.51 2.00

1.56 0.14

-3.85 0.40

0.64 0.60

7.31 2.50

Channel
+

1T
1T

p
p

K+

------

SINGLE ARM SPECTROMETER

767

Our results for this fit are shown in the table above, from
which we can make the following observations:
1.

The parameter A2 ; 0;

all channels exhibit a diffractive

peak at x '" 1.
-1

2.

We find C2 '" 0 or small: the (1 - x)

a simple exponential t dependence.

3.

In contrast, Cl is large: the (1 - x) term has a strong

change in slope with t.

4.

At high x, there is large energy dependence (the parameter D2 is large), but

5.

we see very little energy dependence away from x = 1

contribution has

(D l '" 0).

-1

From the energy dependence in the (1 - x)

term, we conclude the
contribution is not completely from PPP but has some other part,
possibly PPR (which goes as s-1/2). These terms must have roughly
the same relative size for the different reactions to explain the
observed factorization at high x, as shown in Figure 13. Here
we plot the ratios of the invariant cross sections as normalized
by the corresponding elastic cross sections, as:
el

do 'If_I __
do 'lf_
__

R IR
'If

dtdM2 dt
do
d el

----2-1
dtdM2

Tt0

Note in the region near x = 1 all reactions factorize, while down

to x = .75 the 'If and K ratios remain together but increasingly
fall below proton ratio. The implication in a simple Regge picture is that the same exchange dominates both 'If and K, but that
an additional contribution to p + p + p + X enters away from the
diffractive limit.
A third grouping of our inelastic data was made by selecting
events for which the particle identified in the spectrometer indicated a change of strangeness or baryon number relative to the
incident beam particle. Because of the low statistics on these
inelastic "exchange" reactions, we concentrated our analysis on
the six channels with high incident beam flux:

768

D.CUTTS

;.0

{<v+-j

~
OJ

'<S

.,'.,"
...
V

Rp

5"0 (i..eVit.-

oL
0

orIS

(),~

LO

I=:

R1<+/Rr

t:.. ."2-

o,q

o.~s

X.

(tl

o.qS

t
I
706R.'v'/c.

~
~
V

o~'t.::

0,%

o.~

o ..~

o,q5

0,9

o.~s

RK/cRrr-

f
o,CjJ

Figure 13.

ILl-O ~v'1c
o,z.~

tt 1< o. if

Ratios of invariant cross sections normalized by

elastics, as a function of x.

(),QS

769

SINGLE ARM SPECTROMETER

reaction
'If

expected exchange

+ P -+ K+ + X

'If

+p-+ K- + X

'If

+p-+ P + X

'If+ + p -+ p

+X

* '"

0.3

<IN '" -0.5

aK

P + P -+ 'If+ + X
P +p-+ K+ + X

We parametrized the cross section explicitly according to the

Triple Regge Model outlined above; writing the invariant cross
sections for these reactions (a + p -+ b + X) as:

Our results are given in the table below. We find the effective
trajectory <leff '" -.6 for reactions probably dominated by nucleon
exchange, <leff small and positive for those involving K exchange,
and <leff large and negative for A exchange reactions.

These values

are in good agreement with those expected by the exchange picture.

l-2<l
To show the data and this (1 - x)
eff dependence explicitly, we
plot in Figure 14 the logarithm of a normalized cross section

vs. ln(l - x). The slopes of channels expected to have similar

exchanges are clearly in good agreement. The relative normalization (the residue function S) agrees less well, except for the
'If -+ K reactions.

+
+p-+K

1T

+p-+1T+

Ii

+ P -+

+ P -+ P

1T

P -+ K

+
+p-+K +

-+

Jf

1T

Channel

0.2 - 0.20

0.2 - 0.20

0.2 - 0.65

0.2 - 0.65

0.1 - 0.65

0.1 - 0.65

Effective
-t Range
in GeV2

50 - 175

SO .;. 175

140 - 175

50 - 175

SO - 140

50 - 140

Laboratory
Energy Range
in GeV
(GeV-2)

laI 2/161Ts 0 2

0.108 0.015

0.270 0.025

0.240 0.022

0.128 0.012

0.135 0.014

0.155 0.015

TABLE II

-0.65 0.20

-0.65 0.15

-0.50 0.10

-0.55 0.15

0.25 0.06

0.20 0.06

<leff

3.0 1.0

4.4 0.5

3.1 0.5

2.0 0.3

3.7 0.3

3.7 0.3

b eff
(GeV-2)

~
(J)

(')

"i
"i

771

SINGLE ARM SPECTROMETER

(ij

<-t> = 0.3 GeV 2

.".++p - K++

9 ""-+P - K-+

0.1

0.01 L--...L-L.-.L..L.UL..LL_~--L-L-..L-J-~
0.1
1.0
I-x
Figure 14.

Comparison of inelastic exchange reactions.

D.CUTTS

772

PHYSICS IN PROGRESS
The Single Arm Spectrometer as a detector provides excellent
particle identification with good momentum and angle resolution,
for forward scattered particles. Although its aperture is relatively small it can be easily set at a particular value of (x,PT)
by selecting the spectrometer momentum and angle; with successive
runs at different settings it can span a very wide region in the
forward single particle inclusive kinematics, as illustrated below.

0.4

0.2

0.6

0.8

1.0

Vl

--

.1

/'

~(l/

-I
~I

",

-t' /""
6>'
~,

/""
..0,
~,

DATA

~\
('

~\

0.5

~~\
/'
/,...~,

~',
/'
~\

). - - - - - -

PT
GeV/c

/
1.0

- -

/'

- - -- - --

(STATISTICS LIMIT)

A subset (Bari, Brown, CERN, Fermilab, MIT) of the original SAS

collaboration is taking advantage of this feature. He are engaged
in a broad study of inclusive scattering, extending the single
particle coverage begun in the earlier experiment to as wide a
kinematic region as possible and including data with opposite
as well as the same sign particles detected in the beam and spectrometer. Additionally, we have added detectors around the target
(see Figure 16) capable of measuring the charged multiplicity and
pseudorapidity distributions on an event-by-event basis, in
association (or not) with a forward particle detected by the spectrometer.
This inclusive study is currently in progress. We have data
already at 100 GeV/c and 175 GeV/c with positive beam (n+, K+, p
data taken simultaneously) filling most of the kinematic area

773

SINGLE ARM SPECTROMETER

~
q pla.nes wire ~hambers
II--<'~-- ,.., 1.8 In -----,.,~
Figure 16.

Multiplicity detector.

shown, for SAS set on positives, and some data with the spectrometer set on negative particles. The run is resuming shortly;
we will finish data taking at 175 GeV/c positive beam, take data
with negative beam at 175 GeV/c and 100 GeV/c, and possibly explore 50 GeV/c for the maximum extent in s. The analysis is progressing well. The software to compute single particle cross
sections from SAS detector measurements was developed for the
earlier experiment, so the main problem has been to handle information from the multiplicity detector. Our analysis is now
able to unravel the wire chamber hits and Cerenkov pulse heights
into reasonable multiplicity distributions, and we are proceeding
to examine the data already on tape.
An interesting feature of the apparatus is that, by setting
the spectrometer at 0 0 (pitching magnets off), we can use SAS in
veto to make a total cross section trigger. Thus SAS can be used
to measure the integrated cross section as well as map out the
differential cross sections, and -- with the detector around the
target -- measure multiplicity distributions associated with
either. "SAS-veto" data has already proved useful in allowing
us to compare our results with well known pp or np charged multiplicity and prong distributions. Another check on our data has
come from on-line "scaler" cross sections for the single particle
inclusive reactions; where there is overlap there seems to be good
agreement with earlier measurements.

774

D. CUTTS

CONCLUSION:

FUTURE EXPERIMENTS HITH SAS

I will finish my discussion of physics with the Single Arm

Spectrometer by mentioning several experiments due to be run in
the next year. The suhgroup now taking data has heen approved to
study "A-dependence of Inclusive Scattering and Associated Multiplicity." He will explore similar kinematic regions as indicated
in Figure 15 using Be, AI, Cu, Sn, Pb targets, recording as hefore
the multiplicity and pseudorapidity distributions for each event,
and Hith added information from a neH detector for nO's at large
PT. A separate subgroup (ANL, Cornell, Fermilab, SLAC) of the
original collaboration will make a "Study of n+p -+ K+Z::+ and
n+p -+ K+Y*+."

This experiment will study these non diffractive

reactions at It I < 0.6.

As a final topic I will mention a proposed "Search for Heavy

Long-Lived Particles." Those of us currently doing the inclusive
studies have noted the unique suitability of the N6-SAS (beam,
spectrometer) combination for a mass search. Set at 0 the spectrometer is essentially a 4th stage of the beam; together \-lith
M6 there are 8 Cerenkov counters (3 differential) available, as
well as long flight paths. He have proposed to use combined
Cerenkov and time-of-flight techni~ues to search for heavy stable
(T ~ 10-8 sec.) particles.
Three C counters can be used as a light
particle veto, for ~10-10 rejection; and others can be spread
throughout the mass interval up to 5 GeV (mass), above ",hich
time-of-flight measurements become accurate. With this scheme
the entire mass region can be sampled simultaneously with one beam
setting. We will measure pulse heights in the Cerenkov counters,
for better mass determination, and an existing muon detector (at
the end of SAS) would tag muon-like events. The experiment as a
whole would improve sensitivity over previous searches by a factor
of ~200, and could set an upper limit for heavy particle production of ~4 x 10- 5 ~b/GeV/c str., per nucleus. This corresponds
to sampling ~1010 incident beam particles, of both signs, and
"ould take in all about 2 \-7eeks. This mass search seems an exciting possibility to us and a good example of the varied and
interesting physics accessible with the Single Arm Spectrometer.

775

SINGLE ARM SPECTROMETER

DIS C U S S ION
CHAIRMAN:

Prof. U. Cutts

Scientific Secretary:
DISCUSSION
PAULI:
Is there some special reason why the predictions of the simple
quark model and the Lipkin quark model should diverge at large b,
impact parameter?

CUTTS:
Actually, I think that we would expect the data to blow up
compared to quark model predictions at large t
small b.

LIPKIN:
It is rather surprising that the quark model gives you results
at all for this. We have ignored the fact that the pion radius is
different from the proton radius. We should put in form factors
here but if we do, we wind up with a theory with more parameters than
data.

AZIMUTHAL CORRELATIONS IN PARTICLE PRODUCTION

AT LOW

p~

Gisela Ranft
Sektion Physik, Karl-Marx-Universitat
DDR - 701 Leipzig

Summary
The inclusion of Bose-Einstein statistics in an independent cluster model explains the experimentally observed differences in the production of like or unlike
mesons at small separation in rapidity and small azimuthal angle. This effect accounts for an important
part of the observed short range correlation. It also
shows up in the azimuthal asymmetry parameter B as a
function of the mean transverse momentum of the two
produced particles under study. This azimuthal asymmetry parameter as function of rapidity B (AY) basically
allows to discriminate between local or non-local conservation of transverse momenta. Together with single
particle spectra and rapidity correlations the azimuthal dependence explains the observed band structure
in two-particle inclusive production in a conventional
way.

This lecture is based on several papers done together with J. Ranft.

777

778

G. RANFT

1. Introduction

When the study ot" inclusive reactions became :fashionable it was soon realized that a wide variety of models is able to explain single particle spectra s~~
= E ~~F at least in their general trend, but that
correlation data - in particular in differential formallow to discriminate between various proposed models
for the production mechanism. The investigation of
two-particle correlations in rapidi ties ~2.%~1 d':Jz.
revealed a short range nature in the non-diffractive
component which could be explained by intermediate
cluster production /1/. From the y~, y~ correlation
data if followed that in the non-dif1"ractive component
of particle production the number of clusters increases
wi th energy, IWl.":> IV ~s I the average cluster mass is
independent of energy, around
.(1'1 > ~ 1 .5 to 2 GeV,
and the average decay multiplicity of a cluster into
charged particles is
<: &1c.h.~ ~ 3 to 5. These numbers
rely strongly on the value 01 the correlation function R

~ _

c;("N

~r"cl3rL

c;f3N

cl'r" .

c;('!.N

(1 )

d~t=>z..

in the centre of the rapidity plot; i.e. at y~ ~ 0,

y~ ~ O. There also the diffractive component contributes. Moreover, at small rapidity dift"erence 6y = y" - Y2of the two observed particles the Bose effect of identical particles simulates a short range effect. Studied also as function of the transverse momenta p~"
and Pl. and their relative angle <p, this eft"ect enhancesZ.the two-particle correlation at ~ ~ 0 0 and
A P.L = 1P.L.t - pJ.z.1 ~ O. This eff"ect reduces the amount
of short range attraction which has to be explained by
clusters; it makes cluster masses and decay multiplicities smaller such that only
<!VI) 'V 1.2
to 1.5 GeV
and
(Llc:.J..'> "" 2 to 2.3 per cluster.
This suggests even more that clusters are a superposition of actually identified resonances.

PARTICLE PRODUCTION AT LOW PI

779

In Section 2 we shortly describe how the Bose effect

is incorperated in the cluster model in the volume
element of creation of the considered two like particles. In Section 3 we stress that the Bose effect
acting at the birth of the two like particles and the
second order interference effect (a la Hanbury BrownTwiss) are not identical but rather act in the same
direction. Examples are quoted how to discriminate
between these two effects. Sections 4 and 5 deal with
applications of azimuthal correlations, such as the
use of the azimuthal asymmetry parameter B (A y) to
distinguish between local or non-local conservation
of' transverse momenta and the explanation 01' the observed band structure in invariant two-particle masses
as a particular form of two-particle correlations in
y, PL and azimuthal angle

2. The Bose effect for like particles and the cluster

model

2.1. The data

The difference between like and unlike two-particle

correlations - attributed to the Bose effect - is seen
when both particles 1 and 2 are in the same momentum
state, p" = P&. It is however so strong as to show
up also in partially integrated distributions such as
in
see
- Gl~cfI for small ~y (integrated over P~i
:-'>
.....
A. =_ "n~
-II
rJ.~' P.I..'&..
Fig. 2. 1 ) It is
Y'
vv~
I Pl.., II

rl.'l.-l

- the azimuthal asymmetry as function of

13 (l~~)

ell ~i4

( > :[)

d%q, CC\J> I)

~ (P ~ I)
+ ct~~ ccp ~ I)
-

Ay
(2.1 )

for small ~y tintegrated over P~i ; see Fig. 2.2)

- the correlation i'unction C ( AY, t:P ) for small Ay
and cP ~o
(integrated over P.J.i ; see Fig. 2.3)
- the azimuthal asymmetry B l 6p~) for small ~y and
small APl. = IP.l. . 1 - 11.1.2.1
tsee Fig. 2.4)
- the azimuthal asymmetry B (q~) for small Ay and ~p~
with qL = (IP:,,\ + IP1. 2 / )/2.. (see l!'ig. 2.~).

780

G. RANFT

ol}J

d<P

f like
t ""(i ke
6~ ~.4

1.(,

-1.1

;!f

Jf

Fig. 2.1
a)
b)
The experimental distribution dN/J~ for like and unlike
pion pairs at ~y ~ 0.4 in the reaction l-r~~~X at 40
GeV/c. It is 2 ~ n ch ~ 20. a) Experiment. b) Theory.
From ref. /2/.

+eike
t ""tik.e

'b(ll':!)

- Ol

+.~

1.1. 2

Fig. 2.2 a)

2.g

AJ

Fig. 2.2
The asymmetry parameter B as function of AY for like and unlike
pion pairs without restriction in
Y1 and Y2
a) in the reaction Jj-P"'JTJT)( at 40
GeV/c for 6 ~ n ch ~ 20. From ref./2,
b) in the reaction pp -'JrJr X at
205 and 102 GeV/c from the Rochester
- Michigan - ANL - FNAL - SUNY colla
boration as given by T. Ferbel, Proc.
V-th Internat. Symposium on Many Par
ticle Production, Leipzig 1974, p.38 1

781

PARTICLE PRODUCTION AT LOW PI

o 0

Jr+Jf-

,t+.J+

.2.

"

2.

lOS, "102. qe.V/e

or

3F-.r

'I

Fig. 2.2 b)

2.2. The cluster model and Bose effect

We describe the observed short range rapidity dependence within a model with neutral clusters; in this
model the two observed particles can originate from
one cluster (one-cluster term, denoted by 1F) or from
two clusters (two-cluster term, denoted by 2F).
The short range rapidity correlation of two particles
of the one-cluster term is contained in the function

Q Z. ( ~'"

'jz..,

rs)
S

~ ex.nI (- (~~~&2.)~)

(2.2)

with the correlation length L. The long range rapidity

correlation of two particles is due to both particles
emerging from difrerent clusters, thus

(2.3)
where Q~(y, (5) describes the rapidity dependence in
the decay of one particle. Both functions Q., and Q2.
are calculated from the mass and the rapidity distribution of the clusters as well as their decay distribution in rapidity and multiplicity /6/. They are normalized to 1. For small y, Q~(y, (8) is constant representing the rapidity plateau.
Further rapidity dependence is contained in the Bose
effect, see eq. (2.7) below.
The clusters have the average charged decay multiplicity <v:>, its second moment is <v(v-1. Thecharged multiplicity in the collision is~, + or denote the observed kind of secondaries in the cluster
decay (v+ , v_) or the collision ( n+ n_). We assume
I

782

G,RANFT

.,

that their average values

<'V>., , <. v(v-1
are
independent of 11 ; the average values <::''l1..) and
<. h (n -1)'> are functions 01' s.
The inclusive rapidity distribution is /2/ for like
particles
J
like
VI. <J;I'I

Qz(I;j.,/,;!-z"S) <"",">
+

Q'" (,:/" ,S) Q

-1 (

1,-

':111 S) L.::::' 1'11.. ( 'l1. -

-1 )

~V(V-1)'>

~v>

~ v(v-")~ ]

> - c:::1111. '>

~ v>

( 2,

t..;)

lMode{
--"

2.00 qwk.

?:,OOQeN(L

,0'+
,Ot

,o~

-h

Fig. 2.3
Comparison of (6. y, CP) correlations in the cluster model
wi th data on pp Jr":Jr-X and pp ...... Jf-J[- X at 200 and
300 GeV/c by Oh et al., ref. /4/. In the model the peak
in the like pion combination is due to Bose statistics.
From ref. /5/.

783

PARTICLE PRODUCTION AT LOW Pi

)t -

t1k.e.
",,,,Gke

Fig. 2.4
The azimutha.l asymmetry parameter B( AY ~ .5, AP.l..)
as function of 4r.L~lrl.I-lrr~\for Jf-P -Jf!Jf:!:X and Jf'-p""":rr+J[-X
at 40 GeV/c. The theoretlcal curves are from the cluster
model described in the text. From ref. /6/

like

)( ---I.mlilct..
.~

If

.3
/

.2-

...,
0

-,,,

4:/

i/

.;-

''I

t
~

,~

,g

qJ.,

ClINic...

Fig. 2.5
The azimuthal asymmetry parameter J3(D.~ ~.5)dP,~.2c0as function of q.l.",(lp7.1-t1r7D/2.for Jlp ... Jf!Jr!X and JI-p .... i-ti-x'at 40 GaV/c.
The theoretical curves are from the cluster model described in the text. From ref. /6/.

784

G. RANFT

for unlike particles

lAo.,l; kf..

Xl. (~+I~_IS)

=.

(2.5)

Here the first term is the one-cluster term, the second

term is the two-cluster term.
There are three sources of transverse momentum and azimuthal angle dependence
i)
phase space for not too large transverse momenta
P~i
behaves as /8/

~
~ ( \f)

ii)

P.1. 1 )

pJ.~

'V

-1 -

J/

2.

73 PJ. r~~
4

2.

~P~)

r+.

cos't'

(2.6)

with the azimuthal asymmetry parameter B ~ ~_~

depending on the number 01' produced particles.
This ~ dependence suggests the use of a semiinclusive description. In general B is also a
function of the cluster mass M and the cluster
transverse momenta
<: 1cJ..'> /3/.
the p~ and ~ dependence of the Bose effect for
like particles, for Which the parametrisation /2,3/

f(-p" r~-. -P"

J

I 6.

Ll

r.L ) = 1

i-

(1.)(

- r"J"l. - (/77" - f5:.z/]

r L-=- (PII"
2. ( 4i:,)
2. (Ar)~
1.

(2."1)

= ,

can be adopted. For two particles having equal

momenta ~
~~
f reproduces the factor 2 demanded by statistics /2/; it gives a Gaussianlike smearing around ~ = ~~ and for very different 'momenta P1 =i= p1..1 gives the factor 1. Via
the uncertainty principle, the parameters
dPu ~~
are related to the transverse
6)(.1. ~ %PJ.
J
~
c;nd long~tudina~ extension AX" ~ %17
of the
1nteract10n reg10n.
,-u
iii) the transverse momentum cut-off. Adopting a thermodynamic law for the cluster decay this is contained in the term
Wt.J.

(-E/T)

(rt ~ m" r
IV

(.)(p

eJ'r (- h-1.LCosl".j/r)

(2.8)
is the particle I s transverse mass.
=

785

PARTICLE PRODUCTION AT LOW PI

The 6Y, PJ.i and <V dependence due to the short range
term and the Bose effect for the one-cluster contribution to the two-particle distribution is given by

02. (A~/S) T-1F (D.~(<P)

so

N1fc:i~-1e(~2.dl).1."c(P.Lz J{A~ -(~1-'jJ)~,

(- ~'.L"co~k{~1-")) exr(- V\1.L2.COS;(~2.-')1

P.L. P.L2 e~r

{1 + '~r [- 2('p)' ( mi,'i"t..~, - m <i..L.~L)

i ,

r~, p~ - 2. pi,ri, eos

n
L

(2.9)

and for the two-cluster contribution to the two-particle distribution is given by

- ';41>.1)'-

f ",: + P.l: - LP-,< r... <vOl

<'Ds

(2.10)

We obtain the inclusive two-particle distribution for

unlike particles (+,~)
I\r Io\hlike
J',
(
l.

N'1

':f +, 'j - I \f' J

rr
J.+

(.11+11_)

.c:: v+ >

11\

J.-

L..

r.L +

PJ. - - J..0:
) _

e/C.h
r [- h1.l.+ TCost.. 'j.j.

'1Il.

'>

(-1 -

Jr
Z.

-' (J,

G.I.

11\

-J-'---,J-J-J,:---.J-e;{-~+ CA~_ 1/\'1' ~ PJ.+

Je~r[- "".L_ cos

k ~-1
T

1S -t - E.l+ P.l_ cos ~) +

'1 F

<. r v~

r.L-

786

G. RANFT

for like particles (1,2)

'1

0\...

N,

[1

f- .. ,

cJ~ .. d.~2. e<tP olPJ.1 dpJ.2.

p". F"L e-,I" [ - rn.... CCS~( ~,- , ) j "-'r[-"'". <;sk (~,-,~

P{- ;(Ar t (""-. Si"l..~. - ...~. si. L..~.

- 2.p.L .. r.L1CO~~]

+-

d..o-;.",

.('1-1.)

r p,~
+

Pi:

,(\f(V--1\~ (-1 - Jf 1S~~ r.l~r.l,. c.osrh\

Lv>

2.

L..

1P'

rl.Y

"t)

N2 t=>J.~rJ.2. .)(p (- WtJ.~W\J.Z.)[1 + ey:r (-2.~&Ply-t ht4S;i~k~"

- ""<, $iL..~,r - 2~"fi), {

r> ri~ -2 p. . FL, Coscp 1)]

.Q"('j1,S)O,('j2./s)rL.(,\\C,,,_"

- <,,,,,> ~V(V-1

1.

LV>

(2.12)
Here the transverse momentum dependence is made explicit. 1 is the rapidity of the cluster. The normalisation constants N~ and N, are given by the normalisations of Q (y,s) and
Q,.,(y~ 'Yz.,s) to 1.
The data can be described with the parameters of the
cluster decay
Lv("-")~

L...v>

= 0.4

and

Lv.')

1.3 ;

whose numerical values are suggested by data on pp

annihilation and the statistical bootstrap model, the
long range asymmetry parameter
B2F ( e:.y> 2)

={

0.08
0.06
0.04

at
at
at

40 GeV/c
100 GeV/c
200 GeV/c

787

PARTICLE PRODUCTION AT LOW Pl

and the asymmetry parameter B4F due to one-cluster

decay, which is related to ~v > , the cluster mass M
and the cluster transverse momentum <kL~

13

iF

(2.13)

_"-(. v- "

:>

Good numerical values are

0.12

= 0.08 .

and

The comparison ot' the theory wi th the data is shown in

Figs. 1.1 to 1.5. Here we have defined in Fig. 1.4
A(Ll~,P.L112

1S (A~, q.1.1 tlb.L

.c. a) = A( d )
,.
.A'j, PoL",

- C(A~/r..1.,6)
~

(A~, r~."

6)

and in Fig. 1.5

(2.14)
J)

<,A~)

J) (Ll~)

with

E (.d~)

+ E

(2.15)

L~'j)

\="..L +4/t.

A(A~/P.L"/o)=~G(4> J" dr.J.~Jfz..(ct>,A~1 r.1.., P.l.z)

11(1-

and

J)

(A~) -

r.l.. . ~/2..

iclq, Jdp<, [} df.L~Jf..

T

17.,

(2.16)

'"

r..

<;'/
-Orz.

l.l"

C) +- JdpLf.,c... )
P.l. +0/2-

(2,17)

and
C( AY,p.L ,0)
(E( 6Y
being defined similarly to A( AY,P.J..:' 0-)
(D( ~y
for in the range
o to 11/2... In eq. (2.14) (corresponding to Fig. 1.4)
we integrate over events inside a rIng of width d
around the collision axis for fixed \ P':., \ = P.l. .
In eq. (2.15) (corresponding to Fig. 1.5) we in~egrate
over all events besides of those inside the ring of
width 0 This ring shifts with PJ. = \ P:4\ .
The parameters ot- cluster extension \n eq. (2.10) have
the values
~p~ = 0.3 t 0.05 GeV/c
6P"
suggested by the data.

0.15

0.05

GeV/c

788

G,RANFT

3. Bose effect and second order interference

The difference between like and unlike particle correlation also bas been explained by tbe analogy to the
Hanbury-Brown-Twiss ef1'ect /8/. An analysis was proposed in terms 01' the so called A A plot which studies the two-particle distribution over the (qo , 6.q~ )
plane with
q" = \ E .. - Ez.1 (~, E" and ~ ,E",
are the momenta and energies of tbe two particles) and
~ qL
is given in tbe following way;
If

n=
than
and

(P:

P:.)

I;"

P:I

WI.

- -

b q.L

=W

n
-

- = ...

and

w,,

PI! - Pz. ,

This analysis has been carried through in a number of

experiments; it leads to a peak for small qo and small
Aq~ , from which according to the probability to observe like meson pairs
(3.2)

(I is the Bessel function) the interaction radius R

and the interaction time 't' has been determined by
fit /9/. This effect is due to the path ambiguity of
two identical bosons being emitted at two dif:1:'erent
space time pOints, which demands a symmetrisation in
the co-ordinates of the two particles. This e:1:':1:'ect
amounts to intensity correlations which are constructive for bosons; it will be present even with randomly emitted particles (similarly to the astronomical
case, where this second order inter1'erence serves to
determine the radius of' stars, uses randomly emitted
photons). The effect described in Section 2 however is
related to a dynamical e1'fect (i.e. a correlated emission) which can be understood via statistics.
A clear discussion of the distinction of these two
e1'1ects was given by Kripfganz /10/ who also proposes
examples in which these ef1'ects can be studied separately. In particular a quantum statistical treatment
including full isospin conservation predicts correlations or unlike bosons, such as ~+S- These correlations are not present if only charge conservation is

789

PARTICLE PRODUCTION AT LOW PI

taken into account. Here no second order interference

can build up. Although these correlations are in the
order of a few per cent only in most cases (and therefore undetectable), they are big and negative in the
decay

r,"" lI=o) q .. -t1)

--

~+ J(-

(t..(Jro) .

It should be possible
to search for this effect.
Another example is the Bose peak in ~.~. ; in the
decay Jfo _ 2} second order interference is superimposed to the eff"ects present in the lTD Jr" distribution, it makes the peak in )1 correlations more narrow.

4. Azimuthal correlations and local conservation of

transverse momentum

The hypothesis of local conservation of transverse momentum in non-dif"f'ractive multi-particle production

processes was proposed by Krzywicki and Weingarten /11/.
It was used to derive a bound on the Pomeron slope.
Due to dif"t"icul ties in detecting all neutral particles
a direct experimental test of this hypothesis is rather
diff'icul t.
However, it seems possible to test the concept of local p~ conservation in experiments detecting only charged particles if the azimuthal asymmetry parameter (2.1)
is used. In order to exclude diffractive processes,
only large multiplicity events should be considered.
In ref. /12/ perf'ormed two Monte Carlo calculations
producing particles in an UJM like fashion; intermediate cluster production is not important for the argument.
The two calculations are
i) transverse momentum of each particle is compensated
by all other particles regardless of their distance in rapidity, as in the usual UJM.
ii) transverse momentum of each particle is compensated
locally in rapidity space. In events with n particles with rapidi ties y" (i = 1, . , n) we define
weights
~
)

~
/'0

"'.'1

(4.1 )

and balance the transverse momentum of' particle j

by all other particles i = j according to the
weights W'j

790

G.RANFT

." -,,

toeCiL
- - - UJM ikt
~l. OO~ ~uv,tIiOIl

d-'"

-5

'

If

L_ ....

-,L-J- s-l_..,:'
I

.1

L...J

- , 1

.4

L-..,L-__~__-+__~~__~__~__~

.~

.~

.!>

-,r

,,

'2..

~.

Lt.

s.

L_ ~

.2.~

__-L.......:L~

L __- - - ,

L"",-_r-1_ .,-- __ -

, 1
1--1

IIL-___--~--~~~~~~~
,
1.
z.
~.
If.
5.
"A~

.1

L_ .,

~1-----L..,~
I

'---...,'-------------

'I.

2.

~.

If.

C\.U

r.L

s.

Fig. 4.1
The azimuthal asymmetry parameter B( ~y,p~) for charged particles plotted as function of AY for three different Pol cut-offs in the two models
(i) with UJM like P.L conservation, - - (ii) with local P.L conservation, - - - - - -

PARTICLE PRODUCTION AT LOW PI

791

In both calculations the transverse momentum compensation is done for all particles produced in the event, but
we calculate the asymmetry parameter defined in eq. (2.1)
only by considering the charged particles.
In Figure 4.1 we plot B(~y) as function of AY
a) without p~ cut-off .
b) only for particle pairs with P-'-i and ~:i>O.3 GeV/c
c) as case b) but with p. and P.l.j "7 O.b f5:eV/c.
In all three cases we find:!l
i) in the uncorrelated jet-like model, BC4y) is independent of 6 y and
ii) in the model with local p
conservation strong
short range behaviour is present.
No data is available on B(~y) of charged particles for
large multiplicities only, excluding diffractive events
savely. In the data on B(AY) shown in Figure 1.2 b)
leading particles have been subtracted and in this way
also some diffractive events have been eliminated. These
data show very strong long range behaviour rather similar
to our calculation i) (i.e. non-local p~ conservation).
It is difficult to believe that all of this is due to the
diffractive component, however, no firm conclusion can be
drawn. Weingarten /12/ also has presented analytical arguments that it should be possible to test local compensation of transverse momentum in experiments detecting only
charged particles. His conclusions are similar to ours but
we here are more specific ~n proposing to use the quantity
B(AY). These conclusions are also reached by GraBberger
et a1. /14/.
It also was studied in the Monte Carlo calculation /15/
whether by rotation the secondaries by an angle g (arotUld
an axis perpendicular to the collision plane) the observes
long range correlations in B(AY) (see Figure 1.2 b))
could be manufactured from Monte Carlo events produced
with local p~ conservation. This would have demonstrated
that in particle production even at low P.1. there is an
axis of preference (different from the collision axis),
which would minimize the correlation length. The non-local p~ conservation - suggested by the data - then
would De a result of the rotation of the particle jet relative to the collision axis. This way the peak of B~~)
at small AY in the case of local p~ conservation is
reduced, but even rotations by rather large angles do not
produce the long range effect present in the data (e.g.
for rs = 56 GeV, the peak is reduced from B(4Y ~ 0)
~ 0.53
to B(AY ~ 0) ~ 0.38 for e ~ 10 wi thout a significant change for 4 y ~ 2.

G. RANFT

792

5. Two-particle correlations and band structure

The band structure observed in the invariant mass spectrum of two hadrons i and j (i and j being p, p, Jr", Jr- or
K-) in the reaction p + Be -. h. + h + X /16/ can be
considered as a special kind of ~ two~particle correlation, since the invariant mass is Mij ~ (Pi + Pj)L.
The measurements are done at xi = Xj = 0. It is /17/

dN

aM ..
IJ

-=
II

(11 +

'Q ij ( 0

t:

0))

(0 J

r~;) t"

j (0

F~ )

; "lIj :0

. '2i-b ( '1
MU

- lr 'B t:>J.i

P~i

J.')z.

cos

cp\
)

5. 1 )

i
2
i'
where f 1 (x,p.L) and R J(Yi=O'Yj=O), rl.i' B and ~ are
the single particle spectrum of particle i and the rapidity correlation, the transverse mass, the azimuthal asymmetry and the azimuthal angle of the two particles i fond j.
In the experiment the main contribution is due to cos ~
= -1. With thermodynamic single particle spectra f 1 (x,pJ.)
for p~ <1.5 GeV/c and a fit of the constituent interchange model to spectra at higher p~, as well as with
Rij (0,0) ~ RJicl,.Jf,"",-(O,O) '!:!.. 0.6 used for all combinations
of secondaries, the observed band structure at 30 GeV/c
can be understood (Figure 5.1). However, due to the rise
with s of the single particle spectra at x = 0, which is
different for the various pairs of secondaries, the band
structure is predicted to change drastically with energy.
As an example, Figure 5.2 shows the change of the band
structure at Mij = 2 GeV from 30 GeV/c to 300 GeV/c.

PARTICLE PRODUCTION AT LOW Pl

793

I~r
~
If

--rp

Jrp

rr

-..-.11-

+- J(r

-- Icr

pF

z.

1.S

If.

Ifs

-.- Jl'T+
.... Jf - Jr+ "R. Sc.

() le;rt

".5

r~

--- v+\c

I
fle-\{-

-- -

1f+ f

'CT-

-ICK-

r.

\(-r

M..

:J

(4tN )

Fig. 5.1
Comparison of the observed (ref./16/) two-particle mass
distributions at )0 GeV/c with eq. (5.1), see text. From
ref. /17/.

G. RANFT

794

Fig. 5.2
Change of the band structure of the two-particle
invariant mass at Mij =
2GeV from 30 to 300 GeV/c
as predicted by eq. (5.1)
with thermodynamic single
particle spectra. From ref
/17/.

.z.g

10

ao Cit-VIc.

M....
U 2. y('tV

~OO qe-V(c.

PARTICLE PRODUCTION AT LOW Pl

795

References

/ 1/

G. Ranft and J. Ranft, Nucl. Phys. B 53 (1973) 217.

G. Ranft, J. Ranft, M. Sabau, K.G. Fadeev, and A.N.
Solomin, Nucl. Phys. B 86 (1975) 63.
/ 3/ G. Ranft and J. Ranft, Nucl. Phys-:13 92 (1975) 207.
(1975)
/ 4/ B.Y. Oh et al., Phys. Letters
/ 5/ G. Ranft and J. Ranft, Phys. Letters bt"8 (1975)313.
/ 6/ G. Ranft, J. Ranft, R. Kirschner, N.G. Fadeev, S.
Lutov, and R.M. Nazargulov, Nucl. Phys. B, to appear.
/ 7/ G. Ranft and J. Ranft, Nucl. Phys. B 83 (1974) 285.
/ 8/ G.I. Kopylov and M.I. Podgoretzki, Jad. Fiz. 18 (1973)
656.
E.V. Shurjak, Phys. Letters 44B (1973) 387.
G. Cocconi, Phys. Letters 49B (1974) 459.
/ 9/ G.1. Kopylov, " A () plot", Dubna preprint E2-8549
(1975).
A. Para. Talk at the IV Internat. Seminar on High
Energy ~hysics Problems, Dubna, June 1975.
/10/ J. Kripfganz, "Quantum statistical correlations in
the frame work of an uncorrelated jet model", KarlMarx-University pre print KMU-HEP 7607 (1976).
/11/ A. Krzywicki, Nucl. Phys. B 86 (1975) 296.
D. Weingarten, Phys. Rev D 11 (1975) 1924.
/12/ J. Ranft and G. Ranft, "Azimuthal asymmetry and local p
conservation", Karl-Marx-University preprint ID~U'HEP 7514 (1975).
/13/ D. Weingarten, Rochester pre print 1975.
/14/ P. Grassberger and J.L. Meunier, private communication (1976).
/15/ J. Ranft, unpublished.
/16/ J.J. Aubert et al., Phys. Rev. Lett. 35 (1975) 639.
/17/ J. Ranft and G. Ranft, Phys. Rev. Lett. 36 (1976)
988.

/ 2/

G. RANFT

796

DIS C U S S ION
CHAIRMAN:

Prof. G. Ranft

Scientific Secretary:

F. Bopp

DISCUSSION
FERBEL:
I just wish to stress that the latest measurements on zone-zone
correlations from the 30 inch bubble chamber for 200 GeV/c and
300 GeV/c pp collisions indicate that data obey the sort of features
expected from local compensation of transverse momentum -- as pointed
out by Heingarten et al. Using these results, Heingarten has extracted
a lower limit of 0.2 GeV- 2 for the slope of the Pomeranchuk trajectory.
The value of a(Pomeranchuk)is ~ 0.25 GeV- 2 , which implies that local
compensation of transverse momentum in multiparticle production is
responsible for a large fraction of the shrinkage of the elastic
scattering distribution with increasing energy.

RANFT:
So far, most analysis leads to different conclusions.
transverse momentum conservation could not be established.

Pure local

WEILL:
What is the connection between the correlation described here and
the "Goldhaber" effect studies in pp pionic annihilation? In this one,
if one restricts 6p -- the difference of the momenta of the pions
one observes no effect for like-pion pairs; but one observes an
increase in the negative correlations for unlike pions.

RANFT:
The considered effect should be of the same nature. The reported
observations could arise from the use of wrong variables or insufficient
narrow intervals.

797

PARTICLE PRODUCTION AT LOW Pi

BASILE:

To what extent do diffractive processes contribute to the

correlations?

TITI

RANFT:
Diffractive processes are negligible, as events with high
multiplicity (n c > 6) have been selected.
BASILE:

If particles were not identified, how can you obtain the Bose
effect?

RANFT:
The particles are predominantly pions.
particles is, in this respect, negligible.

The influence of other

BERLAD:

Is one to understand the rise of low mass pp pairs from 30 GeV

to 300 GeV as growth of exotic cluster production with energy?

RANFT:
No, they are completely explained in terms of single particle
distributions, which rise at x = 0, and two-particle rapidity correlations.
BERLAD:

Do the correlations at small 6y and 6~ appear in the same form

in low and high multiplicity inclusive data?

RANFT:
Yes, the effect is consistent at different multiplicities.

MONOPOLES

Patrizio Vinciare11i*
CERN
Geneva, Switzerland
1.

INTRODUCTION
In these lectures I would like to discuss some recent work

done in collaboration with Walter Troost at CERN. 1 )

general conditions, we are led to predict the
magnetic monopoles.

Under very

existence of

To put our work in perspective, I will

review briefly what I

regard as key developments in the history

of monopoles.
The symmetry of the free Maxwell equations under transformations of electric into magnetic fields led Dirac to suggest
..

t h e~r

ex~stence.

2)

In his approach, monopoles are introduced in

the theory in analogy with electrons, via independent field

degrees of freedom.

Though motivated by symmetry requirements.

Dirac's extension of electrodynamics was a matter of choice, as

for the introduction of a muon field.

The inclusion of monopoles

had however immediate implications of considerable theoretical

799

P. VINCIARELLI

800

interest, since it poses very stringent constraints for the

consistency of the theory.
a)

I have in mind:

Dirac's famous "quantization condition," which could

account for the universality of the electric charge;

b)

gauge invariance of the second kind, which could

account for the masslessness of the photon;

c)

"duality" of the electromagnetic field, which could

account for our living in 4-dimensional space-time.

On the other hand, Dirac's extension was faced by what then

appeared to be a technical nuisance:

the description of the

electromagnetic field in the presence of a magnetic monopole in

terms of the e.m. potential.

Dirac adopted a

potential with a

singularity line, the (in-)famous "Dirac string", and modified the

connection of the e.m. field tensor to the potential:
F

]JV

(a x A)

~V

fictitious string field.

(1)

To many people this "trick" is not esthetically pleasing.

Objectively, it led to the introduction of dynamical coordinates
and momenta to describe the motion of the string, a fictitious
notion.

It took a real "tour de force" of Dirac to derive the

theory from an action principle. 3 )

Also, it has been a source of

ambiguities, as witnessed by controversies about the precise

character of the "quantization condition.,,4)
In spite of these and other difficulties, in light of the
simplicity of the idea and, undoubtedly, the reputation of his
author, much effort has been put over the years in (unsuccessful)

MONOPOLES

801

experimental searches, yielding ever increasing lower bounds on

the monopole mass and decreasing upper bounds on its production
cross section.

(Un-)fortunately the mass

of the monopole is a

free parameter in Dirac's theory, which can be adjusted to

accomodate any lower bound.
The disillusionment caused by continuing experimental failures
was enhanced in recent years by theoretical developments.

The

advent of unified gauge theories of weak and electromagnetic

interactions modified our visions of symmetries, while prodiving
an alternative explanation for the universality of the electric
charge.

It is amusing to note that these developments

weakened the case for Dirac's theory, but eventually suggested

a conceptually deeper and more beautiful realization of Dirac's
original suggestion.
In unified gauge theories, the non-abelian gauge group is
spontaneously broken and electromagnetic gauge invariance remains
as a left-over unbroken U(l) subgroup.

Then, as

demonstrated

by Polyakov and 't Hooft,S) particles carrying a magnetic charge

with respect to this subgroup appear as soliton solutions to the
classical field equations.

These solutions do not require the

introduction of external sources.

bound states of the theory with

They correspond to bona fide

a non-vacuum quantum number, a

magnetic charge, which may be identified with a topological

charge. 6)

The energy of the solutions.is finite.

The mass of

the bound states, which for small values of the gauge coupling

802

P. VINCIARELLI

constant is given by the energy of the classical solutions, is

therefore calculable and was estimated to be E ~ M fa ~ 10 4 GeV.
w
This large value would account for the failure of past searches
in accelerator experiments.

It would also predict the failure

of similar searches in the foreseeable future.

The existence of magnetic monopoles is therefore compelling
within the theoretical framework of unified gauge theories.
However the framework itself is special, e.g. for its reliance
of the Higgs-Kibble mechanism for

spontaneous symmetry breakdown.

It is then of interest to abstract from previously known examples

and inquire about

the most general conditions leading to the

existence of magnetic monopoles as solitons.

2. - THE MODEL OF IT HOOFT AND POLYAKOV
The model of It Hooft and Po1yakov5 ) is based on the SU(2)
gauge Lagrangian:*)
(2)

(3)

(4)

Here

~a and V(~) denote a triplet of scalar (Higgs) fields and

*)Our conventions are gWV

= diag.

(1,-1,-1,-1); sabc is

completely antisymmetric with s123= 1.

803

MONOPOLES

its self-interacting potential leading to a non-vanishing (vacuum

expectation) value of

~a in the ground state solution:

o
with the usual" choice of gauge.

(S)

The 1,2 components of the

vector field will then acquire a mass (eF) whereas the third
component will describe the surviving Abelian electromagnetic
interaction.
The field equations admit also a soliton solution of the
formS)
~a

= .!....
r

~(r),

A(r) ,

(6)

with
~(r) ~
~

F,

A(r) ~ _1_
~
er

(7)

To interpret it, 't Hooft proposed to use the following gauge

invariant definition of the e.m. field:
(8)

which reduced to the familiar one in the usual (vacuum) gauge:

F

(9)

~v

Expression (8), evaluated for the solution (6), yields

r

k
(10)

representing the radial magnetic field generated by a magnetic

804

P. VINCIARELLI

pole of strenght lie located at the origin:

1
471'

1"
i

2'

ijk F
E

1
- 8(x)
e
-

jk

(11)

Hence, the interpretation of the solution (6) as

describing a

magnetic monopole carrying two Dirac units of magnetic charge g,

corresponding to n=2 in the quantization condition:
eg =

2'1

(12)

It is tempting to wonder about the "origin"

Arafune, Freund and Goebe1 6 ) noted

that

F~v,

of this charge.

as defined in

Eq. (8), may be rewritten in the form

~v

+H

(13)

~v

where

~v

WV

(14)

1
e

d ~b d ~c

(15)

They observe that the magnetic current

(16)
for the solution (6) reduces to
(17)

leading to an expression for the magnetic charge

Q which

MONOPOLES

805

involves only the asymptotic behavior as R

of the triplet

00

of Higgs fields:
lim

= R~

Q = 4TI

8TI

f (d 2a).iE ijkE b

e S2
R

a c

"a

"b

"c

d. dk
J

(18)

They then proceed to identify the magnetic charge of the solution

(6) with (equivalent) topological characters of the Higgs fields
configurations (Kronecker index, Brouwer degree, homotopy class,
Poincar~-Hopf index).

One is thus led to believe that the following conditions

are necessary for the existence of monopoles as solitons:
a)

the imbedding of U(l) in larger gauge groups;

b)

the spontaneous breakdown of such groups to U(l)

via the Higgs mechanism, in the presence of multiplets
of spinless fields.
Is this really necessary?

3. - MODEL NO. 1 IN THE RADIAL GAUGE

Let us consider the SU(2) gauge Lagrangian

where n

denotes a real spinless isovector field subject to the

+2

constraint n

M2 7)

This constraint may be implemented with

the addition of a Lagrange mUltiplier term

'i!w l

(20)

806

P. VINCIARELLI

to the Lagrangian (19).

The constraint eliminates one of the

+

three field degrees of freedom of n.

The model which is obtained

may be understood as a suitable strong coupling limit of models

defined by the Lagrangian (2).
The equations of motion are

(21)

(22)

In analogy with (6), we look for static soliton solutions of the

form

(23)

In contrast to (6). this ansatz involves only one reduced

function A(r), which controls the amplitude of the gauge fields.
The amplitude for the isovector field is constrained to

M.

The

orientation of n in isospin space is not defined at the origin,

which is therefore an exceptional point.
we shall carefully proceed to:

To avoit ambiguities

a) solve the Lagrange equations

with the ansatz (23) in any domain which

b) extend the solution to the whole of

excludes the origin;

space in the sense of

distributions, checking for the absence of external sources.

807

MONOPOLES

Substitutions of the ansatz (23) in Eqs. (21-22) for any

domain exciuding the origin leads to the reduced field equations:

A>O)

(24)

(25)

where

K (A.) = 1 -

e.A...

A (..I\..) .

(26)

Equation (25) determines the Lagrange multiplier function A(r).

The dynamical equation (24) corresponds to the reduced
Hamiltonian:

(27)

It is immediately seen that a solution of Eq. (24) will have a

finite energy only if the following boundary conditions are
satisfied:

I+0
Equation (24) actually implies:

l.A."'6:).
I

(28)

P. VINCIARELLI

808

With the help of standard mechanical analog techniques, one may

convince oneself that indeed there exists a solution of Eq. (24)
in the domain

r > 0 with these asymptotic behaviors.

In Fig. 1

the solution, constructed numerically, is displayed (for later

reference, this is denoted as the

1t

= 1 curve).

We must now extend this solution to the whole of space in the

sense of distributions.

We shall proceed carefully since the

singularity in the n-fie1d at the origin, though very mild,

might induce some people to suspect the presence of an external
source in the n-field equation.

The problem is well posed if

and ~ [where it. is the Lagrangian

ona
Odin
including the multiplier term (20)] are uniquely defined as
the variations

dis tributions, which is the case if they are locally summab1e.

As implied by Eq. (28), corresponding to the presumed solution,

(30)

and they are indeed locally summab1e.

We must then check that,

for any test function T(x), the functional

(31)

vanishes.

To this end, we break up the volume integral into two

parts extending to the interior and exterior of a sphere of

809

MONOPOLES

radius

R, J R and J R respectively.

Clearly

J R = O.

By

partial integration.

"SR=

S a3", [T<''t)~~ +(~T)

;11.

VR

- SS=-' (J'ao-)i.

~~~Q..

T(){)

,1- ] +

S tlp'tlQ..

(32)

R.

in obvious notations.

Since

and

a.T
are test functions and
1.

therefore bounded, Eqs. (30) imply that the volume integrals

be made as small as desired by a proper choice of R.
test is with the surface integral.

The

can

real

Now since,

(33)

this integral is bounded by

(34)

and can therefore be made infinitesimally small as well.

JR

=0

and J

= O.

Thus

proving the absence of external sources.

The e.m. field tensor for the present model may be defined,
in analogy with Eq. (8). to be:

P. VINCIARELLI

810

For our solution it takes the same value as in Eq. (10) and
implies a magnetic charge:

..L

(36)

e..

Thus we claim to have found bona fide monopole-solitons in

theories defined by the Lagrangian (19).
This result suggests that the answer to the questions
raised at the end of the previous section is:

no.

In fact,

making a conventional choice of gauge in the Lagrangian (19),

so that

(37)

we obtain

(38)

where

(39)

We recognize this as a theory of two massive charged vector

mesons interacting with a U(l) (massless) gauge field.
excitations (Higgsons) are not present.

Spinless

Since the Lagrangian (38)

811

MONOPOLES

is expected to be equivalent to the Lagrangian (19), this theory

should exhibit the same monopole-solitons.

We will show that this

is indeed the case and proceed from there to demonstrate the

existence of monopole-solitons in much wider a class of theories.
4. - THE DESCRIPTION OF MONOPOLES IN ABELIAN GAUGES
The transformation connecting "abelian gauges", as defined
by Eq. (37), to "radial gauges", as defined by Eq. (23), is
singular.

This singularity, of topological origin, will reflect

itself in drastic differences of language in the description

of the same physical systems, monopoles in particular, in the
two classes of gauges.

We prefer the description in terms of

abelian gauges because, as we shall see, it will enable us to

present a simple physical interpretation of the binding forces
which are responsible for
solutions.

the existence of monopole-solitons

Thus we will understand why such solutions exist,

carry a magnetic charge. have a finite energy. etc

From another point of view, the description in terms of abelian
gauges is also more fundamental, since it is in this class of
gauges that the
formulated.

vacuum state and perturbation theory are

Finally, only the description in terms of abelian

gauges will be available to us in extending our analysis to

theories where

U(l) is not imbedded in larger gauge groups.

Such imbeddings are in fact inessential for the generation of

monopole-solitons.

They playa cosmetical role, via the use of

radial gauges, in allowing a singularity-free description of

P. VINCIARELLI

812

monopole solutions in terms of ordinary functions with a domain

which extends to the whole of space.

This is not possible in

abelian gauges, where the description will involve the use of

"sections" instead of functions, a
mathematics of fiber bundles and

notion borrowed from the

deeply connected to the gauge

principle, which was recently suggested by Wu and Yang. 8 )

This

notion is not yet familiar among physicists and some were tempted
to conclude that

"sectional monopoles" are not the same as

"radial monopoles".

In fact there is no intrinsic difference

between them.
The apparent difference is similar to that experienced in
the description of the surface of a sphere in terms of spherical
(e,~)

and cartesian (x,y,z) coordinates.

The surface of the

sphere (the analogue of the monopole) does not change with the
coordinate system, but in the first case its description
involves points of singularity (the analogue of the Dirac string)
or a split into "sections", *) whereas in

the second coordinate

systems, which used more variables and is redundant, the

description is automatically singularity-free.
of U(l) into larger gauge groups, such

The imbedding

as SU(2), SU(3), etc.

serves a similar purpose, that of providing a special

singularity-free "coordinate system" for the monopole, at the

*Overlapping regions with associated singularity-free coordinate

systems and coordinate transformations between different systems
in their overlap.

MONOPOLES

813

expense of some redundancy.

The difference is only one of

language.
We will adopt the language of sections.

While referring the

reader to the original articles of Wu and Yang 8 ) for a

com-

prehensive discussion, we recall here the basic ideas which

necessary to follow our subsequent developments.

are

Space in the

presence of a monopole is divided into two regions, Ra and

~.

With the monopole located at the origin and in terms of

spherical coordinates, these may be chosen to be:

(40)

(41)

where

0<0::

in each region.

1T
2 , The e.m. field is described by a potential

For example:

in obvious notations.

Both Eqs. (42) and (43) define Dirac

potentials with a string singularity which is, however, outside

their respective domains.

In the overlap region

P. VINCIARELLI

814

the two potentials are related by a gauge transformation

(45)

where

(46)

The requirement that the gauge transformation be univa1ued

yields Dirac's quantization condition for the monopole charge,
Eq. (12).

The pair of potentials then defines a "section"

which is analytic everywhere, and describes the true field of

a monopole without the need to subtract off any fictitious term:

(49)

via Gauss' theorem.

Charged (gauge dependent) matter fields are

also defined as sections:

815

MONOPOLES

:("1'40) 'fI b ).1

~4.(X):: SAJ, lfb()f.).1

XE. R~b .

(50)

Equations of motion for e.m. potentials and matter fields are

written down for each element of a section in the corresponding
domain.

They are compatible in virtue of Eqs. (45) and (50).

When not explicitly carried out, this splitting of equations is

to be understood in the remainder of this paper.
5. - A PHYSICAL PICTURE OF MONOPOLE-SOLITONS
So far little has been said in the way of understanding why
monopole-solitons exist.

Arguments based on topological con-

siderations in the radial gauge can hardly be regarded as a

satisfactory explanation.

Furthermore, such arguments neces-

sitate the imbedding of U(l) in larger gauge groups,

would like to repudiate as inessential.

~ich

we

In this section we

hope to fill the gap by advancing a simple physical interpretation of the origin of monopole-solitons.

In the next

section this simple physical picture will serve as a

guide

in constructing the ansatz which will exhibit monopole-solitons

in the model defined by the Lagrangian (38).

Then our intuition

will be confirmed.
Let us perform the following conceptual experiment.

Given

a point monopole charge, represented for the moment by an

external source, we place it in a medium described by a matter

816

P. VINCIARELLI

field carrying a spin and a magnetic moment, which we may think

of as an infinite set of magnetic dipoles.
with the e.m. field, the medium will

Via the interaction

polarize:

/
-_._-

Because of the singularity in the e.m. field at the location of

the external point source, we expect the matter field to become
singular at the same point.

If this singularity has the right

strength, then:
a) the infinite accumulation of oriented magnetic dipoles
at one point will build up the monopole charge selfconsistently, in absence of external sources, via a
phenomenon which is converse to charge screening;*)
b) the dipoles

will reinteract with their coherently

generated field and be attracted to the center of the

monopole; the ensuing negative interaction energy will
compensate for the infinite part of the energy required
to create the e.m. field of the monopole.

MONOPOLES

817

Thus particles carrying a magnetic charge could bootstrap themselves into the theory in the form of solitons.
We shall see that this indeed occurs in a wide class of
models characterized by the following general conditions:
1)

U(l) gauge invariance;

2)

the presence of charged matter fields with an intrinsic

spin and a positive anomalous magnetic moment.

The second requirement may be understood by exploring in more

detail proposition
theories.

b) within the framework on non-linear field

In fact the interaction with the normal component of

the magnetic moment arises in such theories as part of the gauge

invariant kinetic term of the matter field.

This term contributes

*)The fact that a (monopole) charge can be constructed by superimposing dipoles can be understood in ordinary linear electromagnetic theory as follows.
2-+

-+

The equation for the e.m. potential

-+

is V A(r) = e(r), where e(r) is the source term.

-+

-+

charge located at a we have e(r)

-+

moment d,

e. (r)

-+:t

-+-+

d YO (r-a).

For a point

-+ -+

o(r-a) and, for a dipole

Consider now a

distribution of

dipoles, centered around the origin and radially oriented with

2
density l/a.
:t

-+-+
yo(r-a)
=

:t-+

Ya/a

-+
The source term then becomes e(r)

f d 3 a(Va/a
-+
3
-+-+
)o(r-a),
-+

f3-+3
d a(a/a ).

by partial integration.

Now

-+-+

o(a), so that e(r)

o(r), i.e., the source corresponding

to a point charge at the origin.

Note the essential role played

by the 1/r2 singularity in the dipole distribution at the origin.

Thus, by accumulating an infinite number of dipoles, a charge
may be obtained.

P. VINCIARELLI

818

a positive definite quantity to the Hamiltonian and cannot

therefore compensate for infinities in other terms, such as the
self-energy of the e.m. field.

This cancelling

role can be

taken up by the anomalous component of the magnetic moment.

6. - MODEL NO. 1 IN AN ABELIAN GAUGE
Armed with the intuition just developed and the section
formalism, we can now go back to the model Lagrangian (38) and
construct an ansatz for monopole-soliton solutions.

Our physical

picture suggests that the polarization vectors of the charged

-+ ) be aligned along the radial direction with

vector mesons (W
respect to the location of the point monopole charge.

This

requirement essentially determines the structure of the presumed

solution (g

= lie):

(51)

(52)

(53)

(54)

MONOPOLES

819

where the regions

Ra and

are as specified by Eqs. (40-41).

Notice that the consistency conditions (45), (46) and (50) are
satisfied.

-+3

-+-

The sections A , W , and Ware everywhere analytic.

The fields, in vector notations, are given by

= eA:1.
\

.It

(55)

(56)

(57)

(58)

(59)

The equations of motion which derive from the Lagrangian

(38) are:

(60)

P. VINCIARELLI

820

Our ansatz satisfied Eqs. (60) trivially, while reducing Eqs. (61)
to the single non-linear equation:

(62)

Not surprisingly, this is the same as the corresponding

dynamical equation obtained from our analysis in the radial
gauge, Eq. (24).

Needless to say, the effective Hamiltonian

also coincides with the corresponding expression obtained in

the radial gauge, Eq. (27).
section 3), the g

Thus, following the steps of

lIe monopole-soliton implied by the

Lagrangian (38) may be easily constructed:

it is physically

undistinghishable from its radial gauge ancestor.

At this point, before proceeding to obtain monopolesolitons in wider and more interesting classes of models, we
should demonstrate explicitly the claims made on the basis of
our intuition in the previous section, in particular concerning
the role of the anomalous magnetic moment of the Wand the
origin of the magnetic charge.

821

MONOPOLES

The Lagrangian (38) describes charged massive vector

mesons carrying an anomalous magnetic moment:

(65)

in addition to the normal component implied by minimal coupling

of the e.m. field to the vector mesons.
(55-56) and (29), it is the anomalous

As

implied by Eqs.

component (65) which

+3
reduces in F , Eq. (56), the Coulomb singularity of the magnetic

field

f: , Eq. (55), at the origin to:

(66)

This logarithmic singularity is sufficiently mild and does not

cause any problem.
in integrable:

In particular the energy density in Eq. (64)

as we predicted in the previous section, this

may be interpreted by saying that the attraction exerted by the

magnetic charge on the anomalous magnetic moments of the W's
offsets the infinity in the

Coulomb

possible for a monopole to exist as a

self-energy, making it
finite energy soliton

solution.
It is with the anomalous moment (65) that we must find also
the mechanism for the origin of the magnetic charge carried by
the soliton.

For the purpose of this identification, let us

consider the Bianchi identities:

822

P. VINCIARELLI

o.

In particular for a

(67)

3 we have, in vector notations:

This equation enables us to express the magnetic charge in terms

of vector meson fields:

(69)

where

SR denotes the surface of the sphere of radius R and VR

the exterior of this surface.

Since the second term vanishes

identically, we obtain:

(72)

MONOPOLES

823

Now, for our soliton solution, we find

(73)

leading to

(74)

Thus we may concisely say that "monopole-solitons are made from

dipoles" in the sense that the latter, in addition to playing a
crucial role in the binding of the soliton, are responsible for
the generation of its magnetic charge.

7. - MODEL NO.2:

ELECTRODYNAMICS OF VECTOR MESONS

In the previous section we demonstrated that it is possible

to dispense of Higgs fields and still retain monopole-soliton
solutions.

This was done in the context of a model with an

underlying SU(2) gauge invariance of the

Lagrangian, explicitly

broken down to an exact U(l) gauge invariance by mass terms.

The latter invariance is obviously indispensable for the
existence of monopoles, but there is no obvious need for the
imbedding of U(l) in larger gauge groups or the invariance of
the interaction Lagrangian with respect to any such group.
physical picture presented in section 5 suggests that such

The

P. VINCIARELLI

824

imbeddings are inessential to the formation of the soliton and

of its magnetic charge.

Now we will demonstrate that this is

indeed the case, a conclusion which will appear as a byproduct

of the extension of our analysis to a wider class of models.
Let us consider the Lagrangian

'I!.:. -

[~~ + (,)( e. (W;' w;:, - W; W;)] +

(75)

where

~ A" -

This

d)) ~ I

(76)

Lagrangian describes two massive charged vectors mesons,

with an unspecified anomalous magnetic moment proportional to)(, ,

in interaction with the photon field.

The quartic se1f-

interaction of the charged mesons incorporated into (75) is

dictated by very general physical requirements on the classical
theory, such as causality of propagation.

From another view-

point the presence of this term ensures the positivity of the

Hamiltonian.

In summary, the

Lagrangian (75) is the most

general one describing three vector mesons with an exact U(l)

MONOPOLES

825

gauge invariance, parity

conserving interactions and involving

only dimensionless coupling constants. 9 )

For l( = 1 the Lagrangian (75) reduces to the Lagrangian
(38) with the vertices of an SU(2) Yang-Mills gauge model and
for which we already exhibited monopole-soliton solutions.
However, in the light of our physical picture, there appears to
be nothing that requires the anomalous magnetic moment of the Wls
to take on its gauge theory value )c= 1:

we expect monopole-

soliton solutions to persist, though with different characters,

for all values of

~>

O.

A generalization of the ansatz which

we have applied successfully to the model with l(

=1

immediately suggests itself:

(78)

(79)

A(:I.) ::

c0-\'\e ~
e.~
a 1lf
--I

(80)

~E: R.J,

w\'I.) = ~ '" tJ\..)

e....JI...

where the regions

(40-41).

Ra and

e.. i. f( ~

(i'tp .. i, fe ) .I

are again as specified by Eqs.

The fields, in vector notations, are given by:

(81)

P. VINCIARELLI

826

(82)

(83)

(84)

(85)

(86)

The equations of motion which derive from the Lagrangian

(75) are:

-0 ~V

t,

e.

1.0

-l)

~'\) -

W+ c- )
r~~

(87)

Our ansatz satisfies Eqs. (87) trivially and reduces Eqs. (88)
to the single non-linear equation:
(89)

MONOPOLES

827

This is similar, though not equivalent, to Eq. (62).

The

effective Hamiltonian is:

The boundary conditions which must be met by a solution of Eq.

(89) for finiteness of the energy are the same as in Eq. (28).
However, Eq. (89) now leads to asymptotic behaviors near the
origin which differ for

')( ~

1 :

>t > 1

(93)

The asymptotic behavior at infinity remains as in Eq. (29).

Mechanical analogue techniques indicate that a solution of
Eq. (89) with these asymptotic behaviors indeed exists for any
~

> O.

Numerical solutions corresponding to )(

are displayed in Fig. 1.

0.5 and)(

The e.m. field for our solutions,

given by Eq. (82), is of course independent of l( and corresponds

to a point magnetic charge

g = lie.

P. VINCIARELLI

828

er

Iw:tl

1.0

"K

=1 (gauge theory value)

0.5

4
Mr~

Figure 1

829

MONOPOLES

It is interesting to compare the monopoles arising in

models with different values of )(.
masses

Figure 2 shows their

E (classical energies) and radii R:

(94)

(95)

R..fl

::.

~ ~~ ~Jl. ~ ~ (./L)
'"

(96)

where we made use of an equipartition theorem implied by a

trivial condition of monimum of the functional (91) under the
scaling operation K(r)+K(Ar).

Notice that, although for)(

1,

the gauge theory value, monopoles are relatively heavy, as

)l. +

as

00

X+

their mass becomes arbitrarily small.

0 the same diverges:

On the other hand

in fact we would then be approaching

electrodynamics of minimally coupled vector mesons in the presence

of a Dirac monopole*).

*)Notice however that as the limit

X + 0 is taken our solution,

while acquiring an infinite energy, remains sourceless, which

is not true for a Dirac monopole solution.

830

P. VINCIARELLI

4
gauge theory value

I
I

I
I

I
I

I
I
RM

e2 E
4n: M

4
Figure 2

831

MONOPOLES

8. - HODEL NO.3:

ELECTRODYNAMICS OF NON RELATIVISTIC SPIN

1/2 PARTICLES

So far we have discussed models where the "constituents" of

monopole solitons are spin 1 particles.

However our physical

picture of "monopoles made from dipoles" does not require that

the latter be carried by spin 1/2 fields.

Unfortunately at the

classical level relativistic c-number spinor field theories are

ill-defined (e.g., the Hamiltonian is not bounded below).

To

circumvent these difficulties and enable ourselves to expose in

a simple and unambiguous way interesting aspects of the problem,
we will adopt for the moment Ga1i1eian c-number spin 1/2 fields.
Of course, in general this will amount to more than a nonrelativistic (low velocity) approximation, as the Pauli principle
will also be ignored.
With this reservation, we introduce the following
Hamiltonian density:

)(=.;; t,+[~.(-<.1'-"A)t~+ }:+[C:.(-1.1 +,,-"A>)""x.1

(97)

... 1r)

(cp+++ 1-+1--) +t[~- ~~ (q,tot 4> - '1-+if!""t)]~)

involving two oppositely charged spin 1/2 (two-component) fields

with the same mass,
field

-+

~.

and

X,

in interaction with the magnetic

Notice that in addition to the "normal" magnetic

moment coupling, as generated by a non-relativistic expansion

832

P. VINCIARELLI

of the Dirac equation, an "anomalous" magnetic moment interaction

is also included by a term of the form

(magnetic field - spin

density)2, implying also a spin-spin self-interaction.

This

is done in analogy with our models 1 and 2 for vector fields.

In the vector case such a form is needed to meet general physical
requirements 9 ) on the classical theory, whereas in the present
case we do not know of similar restrictions.

Note however that

the suppression of the spin-spin term in Eq. (97) would destroy

positivity.
Motivated by our physical picture of monopole-solitons
which suggests

(98)

we

make the following ansatz:

(99)

(100)

(101)

833

MONOPOLES

(102)

(103)

(104)

Notice that

(105)

corresponding to the Dirac unit of magnetic charge, g

= 1/2e.*)

The Euler equations implied by the Hamiltonian (97) are reduced

by this ansatz to

.J

*)This is to be contrasted with the value g

(106)

l/e for which we

exhibited solutions in our previous models with vector matter

fields.

The justification for this difference is the existence

of a connection between the possible values of the magnetic

charge for monopole-solitons and the spins of their constituent
matter fields:

g e

= s.

P. VINCIARELLI

834

which, apart from a redefinition of parameters, is the same as

Eq. (89).

At this point monopole-soliton solutions may be

constructed as in our model No.2.

9. - MISCELLANEOUS CONSIDERATIONS
Thus monopole-solitons exist in theories which contain:
a)

the (massless) photon field;

b)

charged matter fields carrying a positive anomalous magnetic

momen t.

Note that:
1)

The constituent matter fields may carry spin s

= 1/2,1, ..

and correspondingly there will exist monopole-solitons with

magnetic charges g
2)

1/2e, l/e, ... ;

As it is evident in the section formalism, Eq. (46), each

value of g defines a homotopy class of U(l);

3)

The imbedding of U(l) in larger groups is inessential;

4)

The mechanism (Higgs or otherwise) by which such

larger

gauge groups are broken down to U(l) is a fortiori also

inessential;
5)

The monopole-solitons that we have constructed have zero

electric charge.

Solutions exist for arbitrary values of

the electric charge and may be constructed by a simple

extension of our ansatz;
6)

The monopole-solitons which we have constructed have zero

angular momentum.

MONOPOLES

835

Let us now come to the mass of our monopoles, which for

small values of the coupling constant

may be calculated in

good approximation ignoring quantum effects.

If we identify the

W fields of our model No. 1 or 2 with the intermediate vector

boson of the weak interactions having )( = 1, the gauge theory
value, the monopole mass would be predicted to be very large
(E ~ 10 4 GeV), as in the case of It Hooft and Polyakov.
easily accessible value (E

A more

~ 10 2 GeV) would be obtained if the

same models could be applied to describe the interactions of

+

particles such as the (-(750 MeV) mesons.

It is unlikely that

lower mass monopoles with spin 1 matter fields, as constituents,

exist because of the apparent lack of light constituents of
this type.
We know however of light spin 1/2 charged particles:
electron and the positron.

the

They do not carry an anomalous

magnetic moment at the Lagrangian level.

However, they acquire

one by 2nd order radiative corrections and it is positive:

l( = a./27T.

Let us then assume that we can apply the Hamiltonian

(97) to provide an effective description of the electron-positronphoton system.

Since)( is very small, a solution of Eq. (106)

is approximately given by

(107)

836

where

P. VINCIARELLI

m is the electron mass.

Our speculative application

of model No. 3 then leads to a monopole of mass:

(108)

For technical reasons we have not yet assessed the accuracy

of this estimate.

837

MONOPOLES

REF ERE NeE S

1)

W. Troost and P. Vinciarelli, "Monopoles from dipoles",

CERN preprint TH.2195 (1976); and other unpublished work.

2)

P.A.M. Dirac, Proc. Roy. Soc. London A133, 60 (1931).

3)

P.A.M. Dirac, Phys. Rev.

See also:

li,

817 (1948).

J. Schwinger, Particles, Sources and Fields,

Vol. 1 and 2 (Addison-Wesley, 1970 and 1973).

4)

B. Zumino, Strong and Weak Interactions, 1966 Int. School

of Physics, Erice, edited by A. Zichichi (Acad. Press,
New York and London), p. 711.

5)

A. M. Polyakov, JEPT Letters 20, 194 (1974);

G. 't Hooft, Nuclear Phys. B79, 276 (1974).
See also:

T. T. Wu and C. N. Yang, in Properties of Matter

under Unusual Conditions, edited by H. Mark and S. Fernback

(Interscience, New York, 1969), pp. 349-354.
6)

See, e.g., J. Arafune, P. G. O. Freund and C. J. Goebel,

J. Math. Phys. 16, 433 (1975).

7)

The Lagrangian (19) has been proposed as the basis for a

model of weak and e.m. interactions by L. D. Faddeev, Sov.
Phys. Dokl. 18, 382 (1973).

8)

T. T. Wu and C. N. Yang, Nuclear Phys. Bl07, 365 (1976);

Stony Brook preprint ITP SB 76-11 (1976).

9)

G. Velo, Nuclear Phys. B65, 427 (1973).

838

P. VINCIARELLI

DIS C U S S ION
CHAIRMAN:

Prof. P. Vinciarelli

Scientifiec Secretaries:

A. Audrito and A. de la Torre

DISCUSSION
KLEINERT:
In what sense is your monopole charge built from magnetic
dipole moments?

VINCIARELLI:
Not in the sense that, starting from dipoles, you or I can make
a monopole. We cannot make a monopole because there is an infinite
potential barrier associated with the generation of the charge. The
sense of "monopoles from dipoles" was made precise by the Bianchi
identity, which I have discussed and for which I have given a physical interpretation: the monopole charge is built from an infinite
accumulation of magnetic dipoles at one point.

KLEINERT:
In your third model, you
in order to have a monopole.
will have to invoke radiative
not afraid the same radiative
"puff up" into many mesons?

need an anomalous magnetic moment K

Thus, within a fundamental theory, you
corrections to create K = O. Are you
corrections will make your monopole

VINCIARELLI:
The danger is not the possibility of decay into ordinary particles,
which is excluded because of the charge of the monopole, but rather
that the finiteness of the energy in the one monopole sector, which
is implied by our effective Lagrangian, will not be retained by the
full-blown quantum field theory.

839

MONOPOLES

KLEINERT:

Let me suggest using the technique that is standard in superconductivity: first, make a meson Lagrangian out of your fermion
one, analogous to the Ginzburg-Landau equation, as I have explained
in my lecture -- then you have certain radiative corrections included.
After this, solve for a monopole in the meson theory.
VINCIARELLI:

We are presently investigating a similar approach.

ROSSI:

Do you think you could obtain a derivation of your effective

Lagrangian by summing up the quantum contributions to the effective
action, at least to first order in h -- loop expansion?
VINCIARELLI:

We are now making an

atte~pt

along similar lines.

YANG:

In your second example, the fact that you have a finite total
energy is very interesting; but I do not believe you can claim that,
for your solution, VH = 0 is valid everywhere. If you accept my
statement, then your variational equations are, for your solutions,
not valid at the origin. The solution is still interesting because
it has finite energy. It does not, however, satisfy the field equation everywhere. I do not believe you can cure this problem with
the idea of sections.
VINCIARELLI:

I never said that the magnetic field of our solutions is divergenceless everywhere in space. In fact
_ I

VH -

ijk

di Fjk =

o(~)

if the point-like magnetic charge is located at the origin.

Our equations are valid and satisfied everywhere by our solutions
if these are formulated in terms of sections. We need sections, or
alternative formalisms which also exploit U(l) gauge invariance, only
for the purpose of describing unambiguously the electromagnetic field
in the presence of a monopole charge. This charge is, in our case,
self-consistently generated from within the -- conventional -- field
theories we study, without the need to introduce external sources
or new degrees of freedom carrying magnetic charge.

QUARKS,

COLOR AND OCTONIONS

F.BUCCELLA
Istituto di Pisica "G.r-.1arconi" - Rm,1A.

The enigmatic properties of quarks have been described by introducing for them a new SU(3) degree of freedom, which is an exact
symmetry+, with the additional constraint that only states scalar under this new group, named SUc (3) to distin~i$h it from the
SU(3) of isospin and strangeness, can be observed(ZJ. This assumption implies that quarks, which transform under SUc (3) as the fund~ental representation, cannot be observed alone but only in pairs
qq (mesons) or in triplets (baryons). This new degree of freedom
accounts for the symmetry in the others quantum numbers of the baryon wave function and succesfully explains (n-72:t) or reduces
(R value) previous discrepancies. The main purpose of this talk is
to show that the octonion algebra supplies a natural framework both
for the SU(3) character of the new degree of freedom and for the
non observability of non singlet states(3).
There have been already attempts to connect isospin to the qu~
temion algebra(4); in fact its three immaginary units e1 have
the product rule:
ei ej

= -

Sij

+ (ijk ek

which is evidently invariant under rotations in the three-dimensio

nal space of the ei' s : one has then an SU(2). However one would
like to identify one of the ei's , for instance e3 with the
ordinary uni t of quantum mechanics i = [q, pJ
Therefore one should consider only the transformations which leave
e3 invariant, namely the rotations around the third axis: this

it

+ This is equivalent to assume that quarks are parafeTffilons of order

three LlJ .
841

F.BUCCELLA

842

reduces SUeZ) to U(l). In conclusion the quaternion algebra is co~

nected to SUeZ), but if it is considered an extension of the complex numbers of quantum mechanics one is left with U(l).
Let us consider now octonions with their seven imaginary units
multiplying as (5)
e1eJ = - 51J
.. + h1J.k ek
1)
where h is an antisymmetric tensor equal to one for (lZ3) and the
others six triplets obtained from it with a cyclic permutation on
the cycle (1Z43657) (for the others triplets hJ.jk =0)
It is useful to introduce the so called split octonions(S):
l( e -1e
. )
uxl = 2
l
4

l( e +1e
. )
ul = 2
l
4

u Z= ~(ez+ie5)
u3= ~(e3+ie6)
1

u o = 2(1 +ie 7)
In the basis of the

u' s

eq.l) reads:

:It

uiu/ijk~
u. u~=-S .. u o
1 J
1J
uou.=u.
1

uou.=O

Z)

u.uo=O
1

:It

U.Uo=U.
1

:It

uouo=O
Eqs.Z) show the intimate connection between octonions and SU(3); in
deed if the ui's and the ut's are classified resnectively in a
3 and 3:t representations of ~U(3) and Uo and u~ in aI, the nroducts just written show invariance under SU(3) (in fact two 3's
combine antisymmetrically to give a 3~ the product of a 3 times

QUARKS, COLOR AND OCTONIONS

843

a ! gives a singlet and the multiplication by U o and u! does not

change the transformation properties under SU(3)). All together
the seven irnmaginary units of the octonions build up a 7 represen
tation of GZ' which is the symmetry algebra of the octonions,(S)while the SU(3) considered is the sub algebra of GZ which leaves e 7
invariant. Beyond being SU(3) invariant eqs. Z) enjoy the property that no 6 is present in the product of two 3' sand n.o octet in
the product of a 3 times a 3~:therefore three 3's combine to give
a singlet as well as a pair 3 and 3 X;this is just what is expected
for the SUc (3) degree of freedom. In fact the split octonions alg~
braic structure arises if one defines in a six-dimensional vector
space, which is a 3 t!l 3 ~representation of SU(3) , an SU(3) invariant product whic? closes in a space which contains only 1, 3 and
3~epresentations. 6)
The property, which makes octonions a particularly appealing
framework for SUc (3), is the fact that they supply a rather natural
explanation for the non-observability of non-singlet states. In
fact the quarks, which are assumed to transform as a 3 of SUc (3) ,
are described by the ui's; but these numbers do not associate
( (uluZ)u3 f ul (uZu3) ), which is an undesired property for an
observed quantity; however the subspace of the SUc (3) singlets~
spanned by U o and u~, is just the ordinary complex algebra.( )
The identification of SUc (3)as the subalgebra of GZ' which
does not act on one (e 7) of the immaginary octonion units brings
to a non trivial connection between the color and flavor groups.
In fact, given the flavor group F , the color-flavor group SUc(3)~F,
has to be included in a larger one, which contains GZ' where F does
not commute with the six generators which complete SUc (3) to GZ'
To this extent one employs e Z' which is unaffected by SUc (3), in
representing F; the action of the additional generators of GZ
enlarges e7 to 0 0 (the space of the purely immaginary octonions).
Let us consider the simple case where F is the SU(3) of isospin
and strangeness. The SU(3) algebra may be represented by the Gell-Mann rna trices )... ~ times the immaginary unit i (the commutator of
two i Aa is a combination of the others). The antihermitian trace
less 3x3 complex matrices may be decomposed into the real antisym~
metric (called A) and the purely immaginary symmetric ones (called
i So). In order to establish a connection with SUc (3) one identifies the irnmaginary unit which multiplies So with e 7 ; the SU(3)
of flavor is spanned by e 7Sot!lA.
The six generators needed to complete SU (3) to GZ rotate e 7
into the others six immaginary units; so one chas GZt!lOo~Sot!lA which

844

F. BUCCELLA

shows symmetry between the seven octonion immaginary units.

The products 0 0 ~ So transfonn as a 7 ~ 5 under G2 ~ SO(3)
(SO(3) is the Lie algebra generated by A) and, with an appropri~
te definition of their Lie product, build up, together with G2
and A, the exceptional algebra F4 (7); F4 in fact can be obtained
from SU(3) by generalizing the complex numbers to octonions in
the 3x3 traceless antihennitian matrices and by adding the symmetry group of octonions G2(8) (the complex numbers have no continuous symmetry).
The generators needed to complete SUc (3) ~ SU(3) to F4 transform as a (3,6~)ffi(3~,6) under the two SU(3)'s (the SUc (3) content
is easily understood since both the transverse octonions eir7 and
the six generators which act on e7 transform as a 3ffi3* under SUc (3)).
The special role plaid by the left-handed currents in weak interactions makes ra~er appealing to consider, in the framework of
unified theories(9 , the chiral groups. The main point is to repr~
sent the operator which interchanges the corresponding vector and
axial vector charges in tenns of e 7 ; since this operator has squ~
re +1, while e~ = -1, the simplest thing to do is to write it as ie7
where i is an immaginary unit; so SUL(3)~SUR(3) is spanned by:
e7SoffiAffiiSoffiie7A
The extension of SUc (3) to G2 brings to:
G2ffiOo~(So+iA)ffi(A+iSo).

The generators just written build up, again with an appropriate definition of their Lie product, the exceptional algebra E6 ,(7)
which consists of the generators of the three SU(3)'s and of a
(3,3~,3*)ffi(3,3,3) representation of SUc (3) ~ SUL(3) ~ SUR(3).
In connection with the conjectured existence of two others quarks of
charge -1/3(10), SU(6) may be proposed as a possible flavor group.
An extension of SUc(3)~SU(6) can be obtained by identifying the
generators, previously employed to build chiral SU(3) ~ SU(3) , as
a basis of the collinear subalgebra A CAd-.(1.cr;.)/t,1; if one adds to
i, corresponding to the x direction, the immaginary units j and k,
corresponding to the y and z directions, one~ gets the complete set
Qo of the immaginary quaternionic units; A(e) /2) may be identified
with the symmetry of quaternions SUO(2). In conclusion a basis for
the SU(6) algebra is given by: e7So~AEBQo~SoffiQo~e7AffiSUQ(2). With
the previous extension method one is lead to:
G2ffiOo~(So+Qo~)ffi(A+Qo~So)ffiSVQ(2)

which give rise to E~7), consisting of the generators of SUc(3)~

~U(6) and of a (3,15 *)ffi(3*,15) representation of this group.
The partitions performed for the generators of the exceptional

QUARKS, COLOR AND OCTONIONS

845

algebras considered are rather similar; in fact the extension of

the real 3x3 matrices to the complex and to the quaternionic ones
brings from F4 to E6 and E7 respectively: in E7 there is also the
symmetry of quaternions SUQ(2); the extension of Qo to 0 0 and of
SUo(2) to the symmetry of octonions G2 brings from E7 to Es (7).
The groups E6(11) and E7 (12) have been avocated in connection with
the attempts of building up a unified theory of strong and weak
and electromagnetic interactions.
In this framework the fundamental fermions, quarks and leptons,
are classified in the fundamental representation of these groups.
Since the exceptional groups may be characterized as the symmetries
of the Jordan algebras on the real (F 4) complex (E 6), quaternionic
(E7) 3x3 hermitian dctonionic matrices (7), it is not surprising that
these mathematical objects playa role in describing the fundamental representations of these groups.
In particular the smallest non-trivial representation of F4
is described by a traceless hermitian octonionic matrix; the basis
of this representation can be written as: So ffi 0 0 ~ A.
Under SUc(3)~SU(3) Soand e7A transform as the (l,S) while
ei,,1 as ~ (3,3) ffi (~3*)representation. The reducible represent~
tion 27ffi2~of E6 is described by a complex 3x3 octonionic hermitian
matrix: under F4 this object transforms as a 26ffilffi26ffil, which makes easy to understand that the SUc(3)~SUL(3)~SUR(3) content of
the 27 is (3,3,1)ffi(3~1,3ryffi(1,3~3). Finally the 56 representation
of E7 consists of a complex 3x3 octonionic matrix and of a complex
number: from its SUc(3)~SU(3) content one can extrapolate, grouping together the SU(3) multiplets with the same behaviour under
SUc (3)J the SUc(3)~SU(6) content of the 56 to be (1,20)ffi(3,6)ffi
ffi (3;6) (12) .
Quarks (antiquarks) are identified with the triplets (antitri
plets) of color and leptons with the singlets.
A remarkable property of this approach is the proper electric cha~
ge universality between quarks and leptons: this is easily seen at
the F4 level, where they belong respectively to the 3 and S representations of SU(3), and is a consequence of the fact that the trag
sverse (associated to quarks) and longitudinal (associated to leptons) octonions anti commute or commute respectively with the immaginaIY unit e7 involved in the definition of the generators of
SU(3) (S) .

846

F.BUCCELLA

REFERENCES
1) O.W.Greenberg; Phys.Rev.Lett. 13,598 (1964).
2)-H.Fritzsch and M.Ge11-Mann; Proc. XVIth International Conference on High Energy Physics, NAL, vo1.2, p.135.
-M.Gell-Mann; Acta Phys.Austriaca, Supp1.IX, 7.33 (1972).
3)-M.GUnaydin and F.GUrsey; Phys.Rev. D9, 3387 (1974).
-F.Gtirsey; The John Hopkins Univ., Workshop on Current Problems
in High Energy Particle Theory, p.15 (1974).
4)-D.Finke1stein, J.M.Jauch, S.Schiminovitch and D.Speiser; Journ.
Math.Phys. 1,207 (1962).
-D. Finkelstein, J.M.Jauch and D.Speiser; Journ.Math.Phys. ~,136
(1963).
-G.G.Emch; He1v.Phys.Acta 36,739,770 (1963).
5) M.GUnaydin and F.GUrsey; Lettere a1 Nuovo Cimento
and Journ.Math.Phys. 14,1651 (1973).

~,401

(1973)

6) R.Casa1buoni, G.Domokos and S.Ktlvesi-Domokos; Nuovo Cimento 31A,

423 (1976).
7)-See for instance R.D.Shafer; Introduction to Non Associative
Algebras (Acad. P ress,1966).
-N.Jacobson, Exceptional Lie Algebras (M.Dekker 1971).
8) F.GUrsey; Kyoto International Symposium on Mathematical Prob1e
ms in Theoretical Physics, ed H.Araki, p.189 (Sprin~er 1975).
9)-J.Pati and A.Sa1am; Phys.Rev. D8,1240 (1973).
-H.Georgi and S.L.G1ashow; Phys.Rev.Letters ~,433 (1974).
-H.Fritzsch and P.Minkowski; Ann. of Phys. 93,193 (1975).
-H.Fritzsch, M.Ge11-Mann and P.Minkowski; Phys.Lett. 59B,256
(1975) .
10) R.M.Barnett; Phys.Rev.Lett. 34,41 (1975).
11) F.Gursey, P.Ramond and P.Sikivie; Phys.Lett. 60B,177 (1976).

QUARKS, COLOR AND OCTONIONS

847

12)-F.GUrsey, P.Ramond and P.Sikivie; Phys. Rev. D12,2166 (1975).

-F.GUrsey and P.Sikivie; Phys.Rev.Lett. 36,775 (1976).
-P.Ramond; Nuc1.Phys. B110,214 (1976).

FIELD THEORY APPROACH TO THE

STATISTICAL BOOTSTRAP
Etim-Etim
CERN. Geneva+

1. INTRODUCTION
There is currently a new trend in particle physics
as is evidenced by the rapidly growing interest in the
dynamics of extended structures and the mass snectrum
they sustain.
It is assumed that these structures can be described by quantum field theory in which an infinite
summation over some suitable collective modes takes into account the essential non-perturbativ~ features of
the manv-body problem. The new approach therefore goes
beyond canonical perturbation theory.
Statistical Bootstrao ~odel was devised to
describe precisely the kind of extended objects (fireballs = hadrons) for which the above field theory description is assumed to apolv. From this point of view
it is natural to enQuire if a consistent field theory
T~e

+ On leave of absence from Laboratori Nazionali Frascati.

Italy; Present Address: Z.i.F University of Bielefeld.
Germanv
849

E. ETIM

850

underlies the bootstrap idea. This is indeed the case. The

bootstrap constraint can infact be respected by a non-lo~al field theory. We therefore propose to generalize
the statistical bootstrao model to a bootstrap field
theory of which we shall construct here the simplest non-trivial element - the two-ooint function.

2. HAGEDORN-FRAUTSCHI AND YELLIN REPRESENTATIONS

It was first hinted by Yellin that there is a formal analogy between the statistical bootstrap model and
a non-polynomial Lagrangian field theory. Unfortunately
for the simplest and most studied non-polynomial Lagrangian
( 1)

where
is the major and A the minor coupling
constant and
(Po(X') a free scalar field. the spectral
function of the "superpropagator"
A.2.6 (p2)

(2 )

grows less than exoonentiallv for large rn=~

Im~( ~2.)J
l!

r~
7.

REG

) 00

exp(a m 2/3)

( 3)

There is no doubt that one can find a non-polynomial Lagrangian with a spectral function which grows exponentially as required or at worst one that over-shoots it.
However quite apart from the inherent complexity of
working with non-polynomial Lagrangians. the main difficultv is in the approach itself - the specification of
an interaction Lagrangian and consequently an equation

STATISTICAL BOOTSTRAP

851

of motion. In the bootstrap approach, the bootstrap constraint is the "equation of motion".
Its most familiar formulation is in the feed-back
A HADRON
r---")

is a composite of an undetermined
number of all kinds of HADRONS, each
of which in turn------,

from which it is clear that our business is with the

way hadrons are assumed to be constituted. An infinite
hierarchy belies this structure. In order to appreciate
this point of view it is instructive to examine briefly
how one would construct a hadron otherwise.
It is enough for our purposes to compare the phase
space density of the composite system with the density
of mass states of the constituents.
The density of states of N independent particles in
a box of volume
and total energy E is given by

vot

NJ
u (E V ) ==r.~ l b(E-~EL)<b (~Pi)TTd p~
(?

L(.;l.l!)3

:!

N-;

I -

N 34

';--1

(4)

If the N particles are identical a factor of liN! is

also required in the RHS.
For simplicity we ignore quantum numbers, charge,
spin, isospin, etc.

is necessary for "confinement"

852

(a)

E.ETIM

For a meson considered as a quark-antiquark pair,

q q , N = 2 and
(Sa)

while the mass spectrum is just a delta function.

(Sb)

(b)

For a baryon as a quark triplet N = 3 and

--:--E~~)-7)

E5

(6a)

The mass spectrum on the other hand is a sum of delta

functions
3

: : 4:
I. -:.

(c)

b(YYl--rr1i)

(6b)

For the slightly more interesting case of a meson

formed of a gas of an elementary input boson b.
N 00 ie meson - bb, bbb. bbbb Nb,
(N....,. 00 ) 0 n e fin ds

0-0 0 (E, I/o)

.r. ~! ON
00

N= ~

---t)
E -? co

P(yn)

(E, Vo )

exh(a.E
r

3/4

= b(m- YYlb)

(7a)

(7b)

From these simple examples it is evident that infinite

compositeness is necessary if the density of all states
of motion CJ(E, V" ) must grow fast enough to approach
an exponential. This condition is however far from sufficient to ensure the bootstrap equality

853

STATISTICAL BOOTSTRAP

(8 )

as the last example clearly shows. The special strong

interaction dynamics responsible for (8) does not therefore rely only on the wealth of elementary constituents
available. Secondly eq. (8) cannot be valid for arbitrary volume Vo . In fact if the volume Vo is very small
the contribution of the N-particle cluster to the LHS of
(8) ;s suppressed by the factor VoN as against the
linear Vo dependence of the states contributing to
.f> (m) (cL eq.(4. If on the other hand Vo is very
large the N-particle clusters dominate overwhelmingly
in (j (E, Yo) and the equality in (8) becomes impossible.
There is thus a critical volume Va at which the density
of all states of motion of all possible non-interacting
N-clusters CT{E,Vo) becomes equal to the dynamical
density of mass states p em). Vo is a crucial parameter which sets the mass scale. At such a volume eq. (8)
is an extremely revealing and interesting equality between the many intrinsic degrees of freedom of a massive
resonance as counted by the mass spectrum and the equally
large number of dynamical states of motion which can be
generated just by clustering.
Such equality presupposes that no hadron is elementary. Consequently if the input boson in example (c)
above is composite one can write symolically
b

==

bb.

bbb,

bbbb Nb.....

(N~ 00

Hence if for fixed N ( N = 2, 3, 4 ) one iterates eq.

(7a) for each factor b the density of states of motion
( j (E, VO ) becomes

854

CT(E,Vo )

Nj
N
= L: ~[_~)3l b(E-~E)O (~~)x
N

00

1\1_

p-"'-

"...

~J

t(3)

.i: 1

1T (f(m~,VO) dmi d
i. -:.1

E. ETIM

.3~

L -1

(9)

F~

which on making use of eq.(8) can also be rewritten as

Eq.(lo) is the mathematical formulation of the constitutional assumption of the feed-back. It was first translated into this form by Frautschi. It has been observed
that eq.(lo) is a reformulation of Hagedorn's hadronic
thermodynamics in the language of the microcanical ensemble. This is essentially correct if one is satisfied
with taking Laplace transforms. But from the point of
view of Hagedorn's original arguments for the equilibrium
of hadronic matter it goes very much beyond thermodynamics.
According to Hagedorn equilibrium of hadronic matter.
taking place in the short time of about 10- 23 sec is not
brought about by a large number of collisions between particles, and therefore requires no relaxation time. It is
an equilibrium between the enormous number of competing
decay channels of an excited hadron. The probability
weights of these decay channels are given directly by
the S-matrix. There is thus no implication for the application of classical statistical thermodynamics nor the
assumption that S-matrix elements should have no symmetry
properties and no momentum dependence. Consequently in the

STATISTICAL BOOTSTRAP

855

approxiMation in which it is valid eq.(lo) represents not

so Much the Laplace transform of a partition function but
an important information on the constitutton of a hadron
which the S-matrix must contain. We shall call it the
cluster or Hagedorn-Frautschi representation.
Eq.(lo) can be solved exactly for,P(m). It was
first done by Yel11n. To exhibit the solution explicitly
in covariant form let us make use of the fact that the
critical volullle Vo in eqs. (8) and (10) sets a mass
scale in the theory to introduce a lowest Mass d1screte
state in the mass spectrum at m .0 where for later convenience we write

.Vo

~(41t"B)3IZ
3

= 1/4-ltfYl o
2

(11)

bo(~~ m~) = 9 (Po) b(p:l_ m!)

The presence of the term

o (~2_m;) in the RHS of
eq.(lo) is not strictly faithful to the idea that no
hadron is elementary. However from the way it was introduced the -elementar;tv of the particle with mass .0
reflects the length scale to which one probes. It is not
overt.ly inconsistent w;th the bootstrap philosophy, for
what is uelementarv" ~t one length scale lIIay be composite U

856

E.ETIM

when one orobes to smaller distances.

Yellinls solution of eq. (10 1 ) is
00

Bp(p2) =~ <3N QN(~;l) ; 'J1 =1

N=-1

( 11)

where

is the N-body phase s"ace and the gN are numerical coefficients which can be computed iteratively by substituting
(11) into (10 1 ) or more straightforwardly by using their
Laplace transforms. We shall refer to eq. (ll) as the
granular or Yellin representation.

3. BOOTSTRAP FIELD THEORY

In the statistical bootstrap model (SB"') eq.(l1) is
nothing more than an interesting analytic solution of eq.
(10 1 ) . The point of view of bootstrap field theory (BFT)
is radically different. According to it eos.(lol} and (II)
orovide two different but nhYsically equivalent ways of
interpreting the comnlex structure of a hadron. For this
reason it is argued that the two representations should
be unitarily equivalent. In other words. if in the granular picture a hadron is described by the field (P(x) and
in the cluster picture by
q?(X} then there should exist
a uni tarv transformation U such that

<p (x.)

U <P(X) U

.+

( 13)

857

STATISTICAL BOOTSTRAP

Actuallv much less will he renuired; as U will be implemented, the fact th~t it is an isometry (U+U = 1. (UU+)2
= UU+) will suffice. Armed with U the SBM follows as ~
very s~eci~l version of bootstrap field theorv, obtained
bv making use of an interesting decomoosition of ohase
soace (totally ordered manifold)

where dN is the number of

o~rtitions

of N.

Eo.(14) is a bootstrap equation, indeed the bootstrap

equation. The Frautschi bootstrao exoresses the fOrm invariance of (14) under the (order-oreserving) defornations

Df'J (p2)
.o.n~{9t)

---+)

91'" 1 (p2)

).Qn~'9:)

(15)

858

E. ETIM

Qn~ (9t)

~11.;' QYl.~ {9:)

The problem reoresented bv eQs.(14)-(16) belongs to Algebraic Tonology. The dynamics in BFT is not in eqs.(lo')
and (15). These equations are nothing more than an interesting way of exnressing a particular normalization condition (quantization). The dYnamics of the mass soectrum
is completely snecified by an abstract operation, called
the cluster oroduct, which is defined so as to simulate
strong interactions.

3.1

CLUSTER PRODUCT

The three main features of strong interactions we

wish to simulate are: (1) the state of two free "ions is different from that of
a 2 -meson and that of three free oions from that of
W. Symbolically

=f

7t Tt
T[ 1C rr

:4=

J'
OJ

*9

7C ::f A-1

( 17 a)

etc.
Introducing formally a bracket symbol e.g.

f
GO

A1
etc.

(TL TC )
( 7C 7r: rr:; )

( fIT)

( 17b)

STATISTICAL BOOTSTRAP

to indicate that jJ
from three, Al from
eo.(17a) as

TCTC" =*TI:n; rr

859

is formed from two pions, OU

p ,lL and so on, we can rewrite

(TCT[)

4- ("rc n: 7C::) =1=- ~ TC ) TC

* ((TC

7C )7[ )

etc.

( 17 c)

The bracket operation is thus non-associative. Mathematically bracketing is a combinatorial problem, and
as such has been known for a long time. Given a sample
of size N e.g. the product
(17d)
the number of different ways of putting brackets between various factors in the product is called the
bracketing coefficient bN The difference between one
bracketing problem and another arises from the type of
addtional constraints imposed on the formation of
brackets e.g, bracketing only two numbers at a time,
and that difference is reflected in bN,
Bracketing is our mechanism for resonance formation starting with a given input, It consists in associating with each monomial ~N a Hilbert space HN of
degenerate states of dimension bN, Each of the N factors in (17d) carries a basic unit of an additive
"~uantum number".
(2) The square of hadron masses are quantized ieee
( 18)

860

E.ETIM

(3) The mass snectrum of hadrons grows eXDonentiallv; in

other words the degeneracy at the mass level m2=m 2 (N)
grows exnonentiallv with m(N). An important property
of the bracketing coefficient is th~t it too grows exponentially with some algebraic function of N. Hence
if the mass soectrum is arranged to be equal to bN
the function (18) is known, at least asymototically.
Consider then a set(+)

( 19 )

of creation (+) and annihilation (-) operators in momentum

soace. Their (generalized) Wick oroduct is

( 20)

a: ,

Introduce an abstract operation

which transforms
Wick nroducts into the creation and annihilation operators
of particles.

+)

rr; +
-

do not stand for the corresponding

particles in the Rosenfeld tables.

.P

861

STATISTICAL BOOTSTRAP

(: W(rr~p~ W~' ; p)--~) (TCt).p~W~ .. oj p)

(n:-, f~

W-, ..

~;

p)

(21)

and satisfies the following conditions:-

(: 1 .

For all single-narticle onerators

+
P-Cp)
ur+ (p)etc.

P) - IT. (P)
p)
P + (p)

(n; ;

(f .;

(22)

, e.,tc

Hence on the vacuum of the ooerators in (19)

(1 0 )

10)

+
+
Because (TC- P-1
-oarticle operator
l

W, .... , r
-+

( 23)

) is bv definition a single-

imDlving

([
C h as

(24)

lC

the refore no i nvers e

-I, The decay of a res on ance
is therefore not the inverse of
It is much more comolicated and will not be considered here.

d: .

E. ETIM

862

<C. 2. For all local ooerators.7t-+ p-+ . W--+ .......

irresnective of their number. Cn:~f>~W, . ,.; p
is local and the commutator

[ (n: -1 f

-,

W -, , ,;

p) Cn: of, f ..., w ~ ... ; 9)] =

!

c - ~e;z. ( 25)

*0

is finite for all

The combination (tW= W"* will be called a bracket

or cluster or Wick star product in the set (19). The
particles of which (7[1:,p, W~# ... ;
are
the creation and annihilation operators will be
called clusters or simply resonances.

P )

*"
eXDected.

A few oroperties of Ware that

(i)

it is. as
(IT(.r

=*=- ( 7Tf) W)

non-associative

=F ({w;n;) p) =F- (TC f W)

(ii) it is commutative

(iii) it is distributive over addition

Cre [f> + OJ]) (7t" P) +(rc w)

3.2

SBM OR PHASE SPACE BOOTSTRAP AS SPECIAL CASE OF BFT

(26)

( 27)

STATISTICAL BOOTSTRAP

863

be a free scalar field of mass mo. Define the new operators

0.0 q;)) =

t{iiC2
-1)10

b (J~ ~ ~: ) a o( p)
0

a.;(P)= ~ bc (p.2- m ;) a~(p)

( 28)

with commutation relation

According to conditions ([:.1 and (:2 the operators in

(30) are local and have commutators

[Ct(P~N);

Q+C'1;

N'l] = ~N~" ~(4-)(P-9)N 1Q",(p2)

( 31)

E. ETIM

864

is a rewriting of the physics contained in eq.(17b). Eq.

(31) is then the most naive generalisation of eq.(29) and
gives rise to the following mapping

HN={a+(p{YlI

1 '\'"

.Y)~) i 1\= ~ 11, i i):2]---4 Q N(rJ( 32a)

i-=I

tV

kr-Ja(~,N)r ~b1N ao(P)~f-:1 C2;urL "

vf'l;

..f.

,.,

SS+Cr-.r 9Jlf V~ ..

1,::"

'L:.~

t ..

-1

~_2

N-Jlt l 1" +1t.

- ':l. ~
I

0.('1; /11; ) &4-4 i (32b)

between the Hilbert soace HN of dimension bN and phase

SD&Ce.b N = (d N - 1)+ is the number of different bracketings of first ( ) and second (( ) ( ) (
order for a sample of size Nand J is an isometry (J+J=l.
(JJ+)2 = JJ+)_
The 1n.ca1 soace-time fields corresDonding to the operators
in (30) are defined by
,
.' .

(Ptl{X)

=Ji"-Sd+rvV(a(~ ,N) e,j>X + a+(p,,,JQ,-'P")

( 33)
{X)
with a(p,l) = ao(p). and simi lar1v for 'Pn.,Yl;.z,'" Yl..
in terms of a(o,n,n 2 n 1 ). Consider next the operators
defined by

It is dN-1 and not dN because a(p.N) by definition.

a(p.1,1.
J1)
'-_
",'

865

STATISTICAL BOOTSTRAP

(34)

( 35)

DO

C(\") = Clo (PJ+t2. ~ ACp,-t)

(36)

with space-time fields

,~

rt'1 ~ X)

't'

=~

N::=1

O(N
Ar.7I
I
'V N
I'l !

cp. ( X)
N

o(~ == 1
( 34 I

866

E. ETIM

( 36 I
With

n~

restrictions on the coefficients

these fields are not local. From

eos.(31). (34) - (36) the vacuum expectation value of
their commutators ~re

rs

where ~ ( x-y,s) is the free field commutator for a particle of mass

and

867

STATISTICAL BOOTSTRAP

and imolies

ex (0)

(40)

Eo.(39) determines (40) only in modulus. In fact from

( 41)
and

Z 0 I LC(p), C-4- ( CO ] I 0)

wh i ch. wi th

C}N

= 10< N 1.2

l-t, (lI"f\.2.,' ,.">1,1.)

is exactly

what one would obtain by substituting (15 1 ) in (16). From

eo.(41) and the commutation relations (31) it is easy to
see that (42) follows from the operator equation

Comparing (32b) and (43) we see that if phase soace bootstrap (the Frautschi e~uation) is all one needs U can very

E. ETIM

868

well be iust an isometry. EQ.(43) is then consistent with

(32b) if
( 44)

In turn (44) is consistent with (42) if

where

P(N) is a oartition of N

and

o('n..

ex n, "',0( Y\t, ~(N -~,n,L)

,,-

l.="

(46)
From eas.(15). (16). (32b). (39), (42), (43) and (44) we
see that many algebraic systems. e.g. functions, real and
complex numbers, operators, the subspaces of a totally
ordered topological soace, obey bootstrap equations of the
same general form. Thus one has only to verify the existence of a bootstrap rel~tionshio in a simple system (in
our case the category of N-bodv (N=1.2 ) phase space)
and then reoroduce it in other algebraic systems by suitable homomorohisms. Given this fact the Frautschi boots-

STATISTICAL BOOTSTRAP

869

trap equation is clearly nothing more than an interesting

way of expressing a particular normalization condition (eq.
(31)) for the states in HN Actually one need not go beyond
eq.(34) to derive the Frautschi eouation: for fixed N subA

Cl(~/'i

stitute

= 0( Y\"

-n,)

Q.(1i, -n:)

1/(d N-l) to r~ then

into eq.(32b) and change

Phase space

I_~_c_o_n_s
l_
each s_t_s_O_f_p_h_a_s_e_s_p_a_c_e_s_--,
of which in turn - - - .
_i

yields the bootstrao equation upon taking commutators and

summing over N.
3.3

CORRELATION BETWEEN

~ASS

AND SIZE OF CLUSTER

By definition, the cluster operators a(p,n,n 2 n l )

( n,.~/.q1) are local and have finite commutators for all
p2. This definition actually stipulates a definite correlation between the mass mN = ~
and the size
N= /n"I"t"'/}l,;l....;."'-t 'n.L
of a cluster. Consider,
for simplicity, the case 1=1 and take the limitN-tCX>
in eq.(31). Condition Q:2 then implies the inequality
(47)

0<'
and hence the said correlation. If for small mo and
mN
mo ' nt4(m~) is aooroximated by

(m2)~
N

TCB.2

&2.(N-1)J(ti-2>!

(1tB'7Y'-2.)
T

f'4

N-.2

(48)

870

E. ETIM

then (47) is satisfied only if

(49)

where e ~
2.78 If the mass-size correlation is the
same for all clusters then the square of cluster masses are
Quantized
:2
1'YlN

.2
mo2 + b (N-1) ,-6 = 8 m o/
e

( 50)

4. BFT IN GENERAL

BFT is concerned not with the structure of phase space

but how to formulate dynamics in the Hilbert soace

The definition of the cluster products in eo.(30) with

only nhase soace correlation was done for the sake of
simplicity and to get directly to the results of SBM. In
general we have
.
l

.." ,_

~nl"\ "n ~rJ-1

j

1.

t.

'Vn,~ fll.! '" T\~ ~

(S
b (P-iq;)\JJ..lP/t'~"leJJTQlLi:n,)J 1;
,l4)

~=i

,.

ArL;(I.l"'l.'1

(52)

871

STATISTICAL BOOTSTRAP

fn terms of wave functions lV "

i

1\

,<VI'"

'~j ~

"':L' ,,7

1 (~,1f1.)which

.(

may depend on additional quantum numbers represented

collectively by ~. With (52) the normalization of the
states in HN is no longer given by (31). At this point
we can incorporate the information that mass and size of
clusters are correlated directly in the normalization of
states and quantfze as follows

~~Y\,f ;12" .. 'Yl,R.(p 1, Q,; '-n; n;,,1\ t

X

Note th at; f

bo ( p2._ m ~((ll

I (

9)

fl;t' ..

l;=: ~A)/ ~"e i' x'

niL) ~ 4(p- 9)

~A'" =r ~,.\ t!;r:J,

N

and HN in eq.(51) has dimension dN

The ooerators

! I L.
.-.,0

A,,(I') =')JI'>+

Q">,, n.

'l.ttl')

(55)

N==:t t;~ YI ..ot-t\],-t' ~Tle= N

are not local. Their mass spectrum is given by the vacuum
expectatfon value of the commutator

872

E.ETIM

wfth

fOo,2) == ~o(r~m;)-t

2:
00

d,~ ~o (p~n1:)

(57)

N:-::1

m~

Making use of

olN

-N~

= M~

b (N-l) and the recursion relation

N-i

o-(N-.t) ell.
2:
t=o

( 58)

one gets
C>O

6 f(~2) -

d N Wt.J (p2)
L:
N=o

( 59)

where

and cr ('n) is the divisor function, that is the sum of

all divisors of n. Eq.(59) is the IIYellin expansion" in
this case and (58) the IIbootst rap ll equation analogous to

873

STATISTICAL BOOTSTRAP

(42). Eq.(58) is a group theory (linear) decomposition

with resoect to the quantum number(N-l). that is the mass
sauared operator.
We are unfortunately not yet fully equipped to go bevond the mass sDectrum and define the S-matrix. We still
have to define an operation corresponding to decay and
consi der in some more detai 1 the field theory properties
of this non-local model. However what has been done is
clear and can be concisely stated: Given the C~ -algebra
generated by the creation and annihilation operators of a
"oartic1e" carrying a basic unit of an additive "quantum
number" a multi-nary non-associative bracket oroduct can
be defined which maps the monomials

a1 a 1 a 1
<

( 17d)

N-times
into N-degenerate vector spaces. The degeneracy of the subspaces HN consisting of products with at most two overall
bracketings is exponenti a1 in .{t:f as N ~ 00 The
"quantum number" N can be anything provided it is additive.
In the statistical bootstrap model it is the number of
particles in phase space. In the dual resonance model it
is spin.
As far as the mass spectrum is concerned the difference between these two models is in the different normalization of the states.
In fact ea.(46)
Q(

IIi(N)

= lX n

0( 11,
I

i
,. "~nn ~(N-~ 1'L,);
.to

'L:-t

Q.

.~ 1

(46)

874

E. ETIM

can also be used to generate HN In the dual model the

O<'"(V are commuting operators and the stateS
-+ "
(/(j?<N) I 0),

O(~/(N)lo> are orthogonal for

The need for bracketing does not

a r i s e be c au set he s tat e s 0( ~ I 0). 1'11 >/,/ are ass ume d g i ve n
from the beginning. If these states are normalized with
respect to phase space and eq.(44) is imposed the statistical bootstrap model is obtained. The relationship between the two models ends here.
?(t-))::fiP'(/Il)

ACKNOWLEDGEMENTS
The work reported here was done in collaboration with
Prof. Hagedorn. He is however not responsible for any inaccuracies in this report. I would like to thank G.C. Rossi
for valuable discussions on combinatorics and the SBM, to
F. G1iozzi for very helpful information on the point of
view of the dual model and to W. Nahm for explaining his
own point of view of SBM and the dual model.

STATISTICAL BOOTSTRAP

875

REFERENCES
1.

R. Hagedorn; Suppl. Nuovo Cimento 3 (1965) 147

CERN lecture Notes, CERN 71-12 (1971)

2.

R. Hagedorn and I. Montvay; Nucl. Phys. 859 (1973) 45

3.

S. Frautschi; Phys. Rev. 03 (1971) 2821

4.

S. Frautschi and C.J. Hamer; Phys.Rev. 04 (1971) 2125

5.

,J. Yellin; Nucl. Phys. 852 (1973) 583

6.

W. Nahm; Nucl. Phys. 845(1972) 525

7.

l. Comtet; Advanced combinatorics, D. Reidel Publishing Company, Dordrect (Holland, 1974)

8.

R.D. Schafer; An Introduction to Non-Associative Algebras, Academic Press, New York (1966)

9.

E.H. Spanier; Algebraic Tonology, McGraw-Hill (1966)

Chs. 1 and 7

10.

J. Riordan, An Introduction to Combinatorial Analysis,

John Wiley, New York (1967)

E.ETIM

876

DIS C U S S ION
CHAIRMAN:

Prof. E. Etim

Scientific Secretary:

M.A. Ichola

DISCUSSION
LITTENBERG:
How do you verify the form of p(m) at high energies when you
cannot identify individual resonances?

ETIM:
It is not necessary to resolve individual resonances at very
high energies in order to test the form of p(m) there. In fact,
the theoretical mass spectrum is a continuous function in the energy
region where the resonances merge into a continuum.

WILKIE:
The model is similar to the generalized Veneziano model and gives
a spectrum of bosons of higher and higher spins. Is it possible to
put in internal quantum numbers?

ETIM:
It is not possible to give a proper treatment of angular momentum
in this model, and this should be done before any attempt is made to
include isospin.

FERBEL:
I have often heard the statement you have just made -- namely
that the mass spectrum grows exponentially. Could you explain what
that means?

STATISTICAL BOOTSTRAP

877

ETIM:
That the spectrum grows exponentially! It means that if the mass
is allowed to go to infinity the mass spectrum p(m) behaves as an exponential exp (bm) where b is a constant.

FERBEL:
I know that, but what does it mean and what does it imply?

ETIM:
It is based on the bootstrap assumptions I drew attention to
in the lecture. It implies that at very high energies, the number
of open channels in a given hadronic reaction increases with energy
and does so exponentially. Implications for cosmology have been
discussed in various papers by Hagedorn, Frautschi, Huang, and
Weinberg.

FIFTY YEARS OF

S~ffiTRY

OPERATORS

E.P. Wigner
Princeton University

A BIT OF HISTORY
The title of my address, so kindly provided by the leadership
of this conference, clearly indicates that I should be principally
concerned with the role which the symmetry and invariance principles
play in quantum mechanics, with their applications and effectiveness it was precisely 50 years ago that these were recognized. Nevertheless,
I like to say a few words about the role of symmetry principles in prequantum theory because the co~parison of this role with the role played
by the same principles at present seems to me very interesting.
Well, the first application of symmetry principles in physics
is almost 150 years old. In 1830, J.F.C. Hessel determined 32 crystal
classes l ) , those finite subgroups of the three-dimensional rotationreflection group which have only elements of the order 1, 2, 3, 4, or
6. This was soon followed by the determination, simultaneously by
Schonflies 2 ) and by Fedorov 3 ) , of the 230 space groups, that is the
230 discrete subgroups of the euclidean group which contain three noncoplanar displacements. Let me admit at this point that it always
fills me with admiration that neither Schonflies nor Fedorov have
missed a single one of the 230 space groups - neither did Hessel miss
any of the 32 crystal classes. They must have been very careful
workers. Of course, their interest in the problems they solved was
motivated by the early ideas on crystal structure, dating back many
years to Sten0 4 ). Indeed, according to classical mechanics, if the
very few atoms the positions of which are not determined by crystal
symmetry, occupy equilibrium positions, the same will be true also
for those atoms - infinite in number in an infinite crystal - the
positions of which are determined by the crystal symmetry, rotational
879

880

E. P. WIGNER

or displacement. In this sense then, the concept of crystal symmetry

is rigorous in classical physics, as is also the wealth of consequences
for the properties of the crystal which follow from the crystal's symmetry. As will be mentioned later, none of this is true if classical
mechanics is replaced by quantum mechanics - the classical crystal
symmetry then becomes an approximate concept and can be best motivated
by the Born-Oppenheimer approximation. It was rigorous, though, in
the world picture prevalent at the time its founders established it a fact worth remembering.
In classical physics, the theory of crystal symmetry was the most
colourful consequence of the symmetry properties of space-time but it
was not the only remarkable consequence. The other remarkable consequence was the validity of the various conservation laws for energy,
linear and angular momentum, motion of the centre of mass. Actually,
these were known before their connection with the invariance properties
of space-time (Galilei invariance originally) was discovered by Hamel,
Klein, and Nother 5 ) . Again, these are rigorous consequences of the
invariance properties of space-time, valid for isolated systems,
i.e. closed physical systems, not subject to forces due to objects
not included in the system. The derivation of the conservation laws
from the invariance principles required great skill - its discoverers
also deserve our admiration.
The last and most important pre-quantum mechanical application
of symmetry principles is, as we all know, due to Einstein. He was
also the one who first emphasized the fundamental importance of these
principles - the special theory of relativity is a direct consequence
of them. The general theory of relativity also seems to be founded
on a symmetry principle but, as will be discussed further below and
as was already recognized 60 years ago by Kretschman 6 ) , the way the
symmetry principle is applied in this connection is rather different
from its application in the cases mentioned before. As I said, I
will return to this point later.
Before sketching the role the symmetry principles played in the
last 50 years, in particular in quantum mechanics, it may be good to
insert a few remarks on the essence of these principles, on their
role in the structure of physics. It will be my endeavour to be
brief in this because I have already had the opportunity and pleasure to discuss this question at a meeting similar to the present
one 7 ).
EVENTS, LAWS OF NATURE, INVARIANCE PRINCIPLES,
KINEHATIC AND THEN DYNANIC
The physicist endeavours to find the laws of nature, that is the
correlations, between events. We know that if a heavy object is at
the time to at a distance Zo from the ground and is at rest, it will
be at time t at the distance z = Zo - g(t-to)2a from the ground. The

FIFTY YEARS OF SYMMETRY OPERATORS

881

two events between which this law established a correlation are the
positions and velocities at the times to and t. As we all know, the
law just given can be greatly generalized - it is given in the simple
form only as an illustration of what "law of nature" and "correlation
between events" mean. We also know that the events between which
physics establishes correlations has undergone drastic changes in the
course of our discipline's history: it was, originally, the occupation
of definite positions by the objects, it became the magnitude of various
kinds of field strengths at all points of space later - in quantum
mechanics it seems to be the outcome of an observation. Naturally,
in order to make a prediction of a future event on the basis of the
knowledge of past events, the system the events of which we are considering must be free of unknown outside influences, it must be
"isolated" at least to such an extent that the unknown part of the
outside influences is negligible. This can create serious problems,
both practical and theoretical. However, if we disregard these problems, the laws of nature become verifiable and they describe truly
amazing correlations between events, in amazingly simple and attractive mathematical language. It is well to emphasize, nevertheless,
that the laws of nature only give correlations between events, they
do not describe or predict all the events we experience - they would
not even if they were perfect. This fact appears in the usual mathematical formulation of the laws of nature as the needed input of the
initial conditions, the position and velocity of the falling object
in the example mentioned.
The preceding characterization of the laws of nature, and of the
role of initial conditions, may be short and incomplete but we are all
cognizant of their essence. The same applies to the concept of the
symmetry or invariance principles, at least the kinematic ones - in
contrast to the laws of nature these describe correlations between
the laws of nature, that is correlations between the correlations
between events. They postulate the equivalence of several ways to
describe the events - the correlations are postulated to be the same
in all these descriptions. If we can translate from one description
of the events to the other equivalent descriptions of these events,
the postulate of the identity of the correlations provides a great
deal of information about these correlations.
Let me illustrate this point on a very simple, in fact trivial,
example. It is well known that Newton's first law as formulated
already by Galileo "any velocity once imparted to a body will be
rigidly naintained as long as there are no causes of acceleration
or retardation" is invariant under Galilei transformations (and
also under Lorentz transformations). It is not commonly observed
that the converse is also true if it is also assumed that the motion
is fully determined by the initial position and velocity, i.e. if
the position at time t, to be denoted by X(t;x,v) is a uniquely
defined function of t and the initial position x and velocity v

E. P. WIGNER

882

X(O;x,v)

x, (O;x,v)

=x

(1)

X, denoting the derivative of X with respect to the first variable.

This, of course, is not true for bodies of asyometric nature - in fact
for such bodies the definition of the "position" may be quite conplex.
In the following, however, the existence of a uniquely defined
X(t;x,v) will be assumed and proved, on the basis of Galilei or
Poincare invariance that
X(t;x,v) = x + vt

(2)

In order to prove (2), we first consider X(t;O,O), i.e. a body

at rest and at the origin at t = O. If this were different from 0,
rotational invariance would be violated. Renee,
X(t;O,O)

(3)

0 .

If we now describe this body's motion from

coordinate system which is at the position
a uniform velocity -v, the body will have,
at time t the position
X(t;x, v) = X(t;O,O) + x

the point of view of a

-x at time 0, and which has
in this coordinate system,
+ vt

which, together with (3), indeed gives (2) - if we assume that X

uniquely determined by its variables.

(3a)
~s

It may be worth observing that the preceding argument cannot be

carried out if the space is one-dimensional. In this case, in fact,
X(t;x,v)

i gt

+ x + vt

(4)

is consistent with Galilei or Poincare invariance - though not if

reflection invariance is also present - g being a number characteristic of the body, i.e. independent of x,v,t but possibly different
for different bodies. A similar situation holds in quantum mechanics
also: the derivation of the representations of the Poincare group
(to be considered later) yields less far-reaching results if there
is only one space-like dimension. The limitation to symmetric
bodies can, though, be eliminated.
What the preceding argument shows is that Newton's first law is
not only compatible with Galilei invariance, it is a consequence
thereof - or of Poincare invariance.
This example was worked out in some detail because it renders it
less surprising, at least in the case of elementary systems, that the
equations of motion can be derived from invariance principles also in
quantum mechanics. Altogether it is perhaps the most important function of invariance considerations to eliminate assumed laws of nature
which are incompatible with them. Another important function, to be

FIFTY YEARS OF SYMMETRY OPERATORS

883

discussed later, is the derivation of some of the consequences of

these laws - many more in quantum mechanics than in classical theory
where this is restricted to the consequences of the conservation laws.
Important as these are, they are quite restricted.
The preceding considerations refer to kinematic invariance
principles and what I would like to call active ones. They postulate
correlations between different sets of events, sets which can be
transformed into each other by an invariance transformation. They
actually increase our knowledge of correlations between events. If
we find that hydrogen peroxide reacts with a permanganate and discolours it now and here in Erice, we can infer that this will be
the case also at another time, at another place, and will be true also
on a rapidly moving airplane. In contrast, there are invariance laws
which postulate a variety of descriptions between the same events.
The oldest example for this is the description of the same electromagnetic field by a variety of potentials, obtainable from each other
by the addition of an arbitrary gradient. All these postulate the
same correlations between "events" which are, in this case, the
magnitudes and directions of the electric and magnetic fields associated with the points of space-time. Unimportant as this may appear
to be, it does unquestionably have a certain mathematical attraction
to have a variety of descriptions of the same events. The most important non-active invariance is the basis of the general theory of
relativity - the same field, that is the same set of events, can be
described by all the coordinate systems which can be transformed into
each other. All the true consequences of the theory, all the correlations, can be obtained by means of coordinate systems chosen in particular ways, as characterized most concretely by Fock S ). Surely,
ordinary mechanics can also be formulated not only in terms of
Cartesian coordinates but in terms of any type of coordinates - in
fact Lagrange's equations do that. But the physical content of
the equations is not increased by the more general formulation only the mathematical attraction of the theory is. As was mentioned
before, this was first recognized by Kretschman 6 ) . Nevertheless, the
mathematical attraction of the more general formulation is so great
that the condition of the invariance of the gravitational equations
under arbitrary coordinate transformations has not been seriously
questioned. Similar invariances under "gauge transformations" have
been introduced also in modern quantum mechanics and they play an
important role in particular in particle theory.
Let me now come to the last kind of invariances I want to mention - the dynamic invariances. These are also of increasing importance. They refer to particular types of interactions which show
definite and often surprising regularities. The oldest such regularity
is the very closely equal interaction between a pair of protons, a
pair of neutrons, and between a proton and a neutron pair. This
equality of the interactions is approximate
it has nevertheless
many interesting consequences which will be a bit further discussed

E. P. WIGNER

884

below. Even more interesting is the SU(3) symmetry between particles,

the "eight-fold way", discovered by Gell-Hann and Ne'eman 9 ) though it
is even less accurate. I am sure all of us here are familiar with
it.
Let me herewith conclude the general discussion of the various
types of symmetries and go over, first to a discussion of the fundamental differences in their applicatiofr in classical and quantum
physics, then to a more specific discussion of the conclusions they
led to in the various areas of our discipline.
WHY DO THE SYMMETRY PRINCIPLES CONTRIBUTE MORE
TO QUANTUH THAN TO CLASSICAL MECHANICS?
Two of the reasons for the increased importance of symmetry
principles are rather obvious. The description of a state of the
system is more involved mathematically in quantum than in classical
theory. The latter describes the state of n particles by 6n numbers
(positions and velocities), the first one by functions of 3n variables a much more complex concept. If the symmetry principles make in both
theories the same relative contribution to the solution of the equations, their actual contribution is much greater in quantum than in
classical theory. The second obvious reason is that quantum mechanics
has been applied in much more widely extended areas than were the
domains of classical physics. In particular, the latter took the properties of its objects, size of the atoms, density of the substances,
as given from the outside, whereas quantum theory tries to derive
them from its basic principles.
There are, however, in addition to the foregoing, other reasons
for the increased effectiveness of the symmetry postulates. The first
and most general one of these is that the states of a quantum mechanical
system form a linear manifold. In particular~ the superposition of two
or more states with the same definite energy*) (or the same momentum
or the same value of any other quantity) is still a state with the
same definite energy (or the same momentum, etc.). It further follows
from the rotational invariance of the energy that if we subject a
state with a definite energy to a rotation, it still will have the
same definite energy. This is, of course, true also in classical
theory. However, in quantum mechanics it is also possible to make a
superposition of the initial and all kinds of rotated states and each
such superposition will still have the same definite energy. If we
superpose with equal weights states subjected to all possible rotations, we obtain either 0 - that is no state - or a spherically symmetric state. Hence, the existence of spherically symmetric states except for single particles non-existent in classical theory - is, in
*)

By this we mean that its state vector is a characteristic vector

of the energy operator.

FIFTY YEARS OF SYMMETRY OPERATORS

885

quantum mechanics, a consequence of the rotational invariance of the

energy. Even if the procedure sketched above gives 0, i.e. no
spherically symmetric state, it is possible to construct finite sets
of states, each member of which, 1f subjected to a rotation, becomes
a linear combination of the members of the finite set to which it
belongs. This is a mathematical theorem, known by now for a long
time. Hence, the linear character of the state vectors, the possibility to form superpositions of them, guarantees that the states
with a definite energy have simple properties with respect to rotations. This plays an important role in the theory of simple systems
and leads to the theory of angular momentum J. This has an important
role in the theory of small systems - atoms and molecules - but is
not very relevant in the macroscopic domain. After all, no one wants
to construct a spherically symmetric state of a crystal - even if it
does follow from the theory that such states exist.
The linear character of the states in quantum mechanics has other
interesting consequences. If we subject the states not only to rotations but to all possible Poincare transformations (Lorentz transformations and displacements) we can form sets of states - infinite sets
though - with remarkable properties. Each set remains "closed" under
any Poincare transformation, i.e. if we subject any member of any
set to such a transformation, it becomes a linear superposition of
the original members of the same set - a rather simple set characterized
by two numbers, the rest energy of the states of the set (which is the
same for all members of the set) and the angular momentum of the states
at rest (called the spin, also the same for all members of the set).
Since the Poincare transformations include time displacement, the time
development of each member is also determined. Uhat is sketched here
is the theory of the irreducible representations of the Poincare groupIO)
- surely not a new subject. It follows from what I said, conversely,
that every state of any system can be considered as a linear combination
of states each of which belongs to an irreducible representation of the
Poincare group so that its time development is well known. This seems
wonderful, but the trouble is that we do not know, in general, how to
decompose the state vector of an arbitrary system into vectors which transform by such a representation-the state vectors are usually given in
terms of other quantities. Only in the case of elementary particles,
all the states of which transform to superpositions of each other
by an irreducible representation of the Poincare group, does the
existence of this symmetry give a complete description of the behaviour of all states. Even this is, however, remarkable - it also
reminds us of the similar situation in classical mechanics where the
admittedly much simpler character of the motion of the particle also
follows from the Ga1i1ei or Lo'rentz invariance.
The preceding discussion emphasizes the importance of the linear
character of the quantum mechanical states as facilitating the application of the invariance principles. In fact, this character has a
decisive role in the application of the symmetry relations in quantum

886

E. P. WIGNER

mechanics. Another circumstance vlhich increases the role of the

symmetry relations is the importance of the response of systems to
small perturbations of the total Galilei or Poincare symmetry. Such
a perturbation is caused, for instance, by a weak homogeneous magnetic
or electric field or some other outside interaction which destroys
part of the Galilei or Poincare symmetry, leaves some of it valid
and has a simple behaviour even under those symmetry operations which
it destroys. It is a consequence of the effect of a weak, homogeneous
magnetic field that it splits an energy level with angular momentuhl J
into 2J+l equidistant levels, the splitting being proportional to the
intensity of the magnetic field. There are many other kinds of perturbations about the consequences of which the symmetry principles
furnish equally far-reaching information ll ).
The last reason for the increased effectiveness of the symmetry
principles in quantum mechanics, perhaps the most important one, is
the existence of approximate symmetries in virtually every area of
application of this theory. The oldest example herefore is the RussellSaunders coupling in atomic spectra. It is based on the fact that the
velocities of the electrons in the atoms, particularly of those on the
outside, are small as compared with light velocity. As a result, the
magnetic fields they create are small and have little effect on the
energy levels, either by influencing the orbits, or by interacting
with the electronic spins. The consequences of the assumption that
the effect of these interactions is small can fill books, they include
Lande's interval rule, his g formula, and many other conclusions.
Naturally, all of them are approximate.
It may be good, in conclusion, to recall the case in which the
symmetry relation is less effective in quantum than in classical
mechanics. Crystal symmetry is, at least at low temperature, a
rigorous concept in classical theory - in quantum mechanics it is an
approximate one, just as are those mentioned in the last paragraph.
It would be valid if the nuclear masses were infinitely larger than
the electronic mass - since this is not truly the case, it is approximate. Born and Oppenheimer have developed an approximation schemel 2)
which brings this out clearly, but the effect of the departure from
the approximation has not been analysed for crystals, even though a
number of cases are known where it breaks down badly. This presents
a very interesting problem, but a fully logical solution, pointing to
all kinds of deviations from the classical theory, does not seem to be
easy.
This concludes what I wanted to say about the reasons of the
greater effectiveness of the symmetry principles in quantum rather
than in classical theory. A very short review of the application
of these principles will follow a few critical remarks of a general
nature on the invariance principles.

FIFTY YEARS OF SYMMETRY OPERATORS

887

SOME ADVERSE REMARKS

It is natural to ask at this point: how sure are we that the
"rigorous symmetries", those of the Poincare group, are really valid?
Do we have any reason to doubt them?
There is no question that they are valid to a very high degree
of accuracy. They may not be absolutely accurate because every actual wave function extends to infinity and, of course, the true physical space is not flat, it has a gravitational curvature interfering
with the Poincare symmetry. If one makes an estimate of the inaccuracy
of the symmetry on this basis, one arrives at a fantastically low
number, perhaps unobservably low. But perhaps not quite. Anyway,
there is at present no experimental reason to attribute weight to
this observation.
It should be mentioned though, that what we now believe to be
the true rigorous symmetry is smaller than what we thought of before
1956. Before that, it was believed that space reflection is also an
accurate symmetry, i.e. that if a succession of events is possible,
i.e. compatible with the laws of nature, the set of events which, if
reflected in a mirror looks like these, is also compatible with the
laws of nature. This symmetry was first suspected to be absent in
the process of S-decay by Lee and Yang 13 ) and their surmise was confirmed by Wu and collaboratorsl~). Thus space reflection, although a
highly accurate symmetry in processes not affected by the interaction
responsible for S-decay, is not a truly accurate symmetry. The same
seems to be true of "time reflection" or reversal of the direction of
motion.
Neither of these observations affect the remaining part of the
Poincare group, the so-called proper group. Also, for most phenomena,
the effect of these "symmetry breaks" is negligibly smalL Nevertheless, the fact that some symmetries in which we had full confidence
are broken - violently under appropriate conditions - makes one less
certain about the absolute validity of the remaining symmetries.
Are there any other reservations about the present concepts of
symmetry and invariance? The connection between the invariance principles and the conservation laws was mentioned before. Can all conservation laws be derived from symmetry principles? No - the conservation laws for electric charge, for baryon and lepton numbers, do not
follow from any accurate symmetry known at present. Are these going
to undergo modifications? We do not know.
Let me now conclude by saying a few words on the application of
the symmetry principles in the various principal areas of quantum
mechanics - the role of the different symmetries which are recognized
to be approximate providing the most interesting conclusions.

E. P. WIGNER

888

THE ROLE orr THE SYHHETP3 PRINCIPLES IN

SPECIAL AREAS OF PHYSICS
The symmetry principles' first application in quantum mechanics
was to the theory of atomic spectra. Their effectiveness in this
area was most eloquently stated by Max von Laue as quoted in the
Preface of the book published on this subject 11 ) by the present
writer: "almost all rules of spectroscopy follow from the symmetry
of the problem".
Some of these symmetries are essentially rigorous, such as the
aforementioned equality of the energy of 2J+I states, or the splitting
of these energy levels into 2J+I equidistant levels by a weak homogeneous magnetic field. The same applies to the formulae of Honl
and Kronig giving the relative strengths of the optical transitions
originating from the members of such a split level. Some other
rules, such as those of Lande also mentioned before, are approximate,
the approximation being based on the assumption that the average
velocities of the electrons are small as compared with light velocity.
It is not possible to go into detail on these questions at this
occasion - books can be, and have been, written on them.
The role of symmetry principles is less prominent In molecular
Surely, the rules given before which follow rigorously
from Poincare invariance are valid here also but they are less important because the density of energy levels is, as a rule, so great
that one does not closely investigate individual energy levels. Of
the approximate symmetries, the most important one is valid in the
Born-Oppenheimer approximation 12 ) - in this approximation the motion
of the nuclei can be treated essentially classically, the potentials
acting between them being given by the electronic energy levels which
correspond to the specified positions of the nuclei. It follows that
in this approximation the symmetry theory of classical mechanics is
valid, and indeed, many molecules show a classical symmetry - the H
atoms of CH 4 are practically always close to the corners of a regular
tetrahedron. In classical theory, at very low temperature, they
would always be exactly located at the corners of such a tetrahedron.

physics.

There is one exception to the very close accuracy of the classical

theory: if a molecule contains several identical atoms, as does CH 4
or even H2 , the symmetry or antisymmetry of the wave function in the
variables of these atoms influences the rotational structure drasticallylS) and renders also certain transformations very unlikely: we
have orthohydrogen and parahydrogen and they transform into each
other very slowly16).
Most of the remarks made above on the theory of molecular systems
apply also in solid state physics. The energy levels of a solid are,
in fact, practically infinitely dense on the energy scale. There is,
however, an approximation which goes beyond that used for molecules.

FIFTY YEARS OF SYMMETRY OPERATORS

889

Once the positions of the nuclei are fixed - and since the EornOppenheimer approximation is used and since this gives results close
to those of classical physics, they do form a regular crystal lattice the potential field acting on the electrons is a field with a definite
crystal symmetry. The further approximation which is made is to
attribute an individual wave function to each electron 17 ). In atoms,
this is called the Hartree-Fock approximation. The electrons then
move in a field with the crystal symmetry and this has many interesting
consequences 18 ). In particular, the energy levels form zones, called
Brillouin zones, and the electrons can jump from one level of such a
zone to one close by. The electric conductivity and many other properties can be calculated from this picture and the agreement with
the experimental findings is truly surprising. According to the
picture used, if all Brillouin zones are either empty or fully occupied, an electric field of normal intensity cannot cause any electron
to change its position in any of the zones and the electric field
thus cannot induce a current. This is indeed confirmed by the fact
that materials, such as diamond, in which, according to the theory,
all Brillouin zones are filled or empty, are insulators - their electric conductivity is billions times smaller than that of metals. This
agreement between the very approximate theory and experiment is truly
surprising - no real explanation seems to be known therefor.
In nuclear physics 19j , at least the basic rules of the coincidence
of the energies of 2J+l states, where
is the angular momentum in the
rest frame, is verified. The magnetic splitting of the lines has not
been observed because its magnitude is inversely proportional to the
mass of the constitutents, and hence about 2000 times smaller than in
atomic spectra - unobservable for realizable magnetic fields. However,
the other consequences of the,Poincare invariance appear to hold.

In

The very close equality of the proton-proton, neutron-proton,

neutron-neutron interactions leads to a new approximate symmetry. If
it were complete, isobars such as 7Li and 7Be would have identical
energy spectra, and the two spectra are indeed very ~imilar in this
and in all similar cases. In the case of isobars such as He, 14 N,
140 , the spectra would be identical only for 14e and 14 0 - many of
the levels of 14N cannot be realized in 14e and 14 0 because of the
exclusion principle. Indeed, the spectra of 14e and 14 0 are very
similar and there is an energy level of 14N which corresponds to each
of the energy levels of 14e and 14 0 - the three levels form what is
called an isospin triplet. A similar situation prevails throughout
the whole nuclear table and several rules that can be derived for the
properties of the members of isospin multiplets are well satisfied.
Their energies differ, but the dependence of the difference is, as
expected, to a very good approximation only a quadratic function of
the neutron-proton difference. The B transitions between them satisfy
the symmetry rule also very closely. The isospin concept turned out
to be a very useful and rather accurate one.

890

E. P. WIGNER

If one further assumes that the nuclear forces are independent

of the spin - this corresponds to the Russell-Saunders model for
atoms discussed above - one obtains further regularities. These are
a good deal less accurate than the isospin regularities because the
nuclear forces are spin-dependent. In spite of this, as has been
shown particularly by Radicati, the resulting "supermultiplet theory"
has several interesting confirmations.
Let me now come to the last area I will mention, particZe physics Z (
Since our whole meeting was devoted to it, I will not discuss it in
detail., The approximate symmetry, the SU(3) symmetry, leading to
the "eightfold way" is not very closely realized. Nevertheless, it
has led to many very interesting conclusions which were confirmed
experimentally. In particular, it led to predicting the discovery
of new particles to complete the multiplet postulated by the SU(3)
symmetry. It also led to a rather simple expression, the Gell MannOkubo formula, for the mass differences between the members of a
multiplet. The basis of the symmetry, on the other hand, is, in my
opinion, not as clear as in the case of the other approximate symmetriel
Several attempts were made to derive them; very interesting ones, but
not fully convincing.
Well, it was a pleasure to review the problems and accomplishments
of the symmetry theories, even though the review had to be, naturally,
very superficial. I often wonder how long our division of information
into the three groups of initial conditions, laws of nature, and
i~variances will persist. At present it is interesting and useful
to use these concepts.

FIFTY YEARS OF SYMMETRY OPERATORS

891

REFERENCES
1)

J.F.C. Hessel, Ostwald's Klassiker der exakten Naturwissenschaften, No. 89 (Leipzig, 1897), p. 91.

2)

A. Schonflies, Krista11systeme and Krista11struktur (Leipzig,

1891).

3)

E.S. Fedorov, Zap. Hin. Obsk. 38, 1 (1891) (Trans.:

4)

N. Steno, De solido intra solidem naturaliter contento dissertationis prodromus (Florence, 1669).

5)

G. Hamel, Z. Hath. Phys. 50, 1 (1904).

F. Engel, Nach. Ges .Wiss-.-Gettingen, 1918, p. 375;
ibid. p. 235.

6)

A. Kretschman, Ann.

7)

The role of invariance principles in natural philosophy,

Rendiconti della Scuola Internazionale di Fisica Enrico
Fermi, Corso XXIX (Academic Press, New York, 1964).

8)

V. Fock, The theory of space, time, and gravitation (Pergamon

Press, London, 1957).

9)

M. Gel I-Mann, Phys. Rev. 125, 1067 (1962).

Y. Ne'eman, Nuclear Phys. 26, 222 (1961).

10)

E.P. Wigner, Ann. of Hath. 40, 149 (1939).

11)

E.P. Higner, Gruppentheorie und ihre Am7endung auf die Quantenmechanik der Atomspectren (Vieweg, Braunschweig, 1931). Somewhat updated English translation: Academic Press, New York,
1959.
M. lIamermesh, Group theory and its application to physical
problems (Addison Wesley Publ. Co., Reading, Mass., 1962).

12)

M. Born and J.R. Oppenheimer, Ann. der Physik 84, 457 (1927).

13)

T.D. Lee and C.N. Yang, Phys. Rev. 104, 254 (1956).

14)

C.S. Wu, E. Ambler, R.W. Hayward, D.D. Hoppes and R.P. Hudson,
Phys. Rev. 105 1413 (1957).

15)

G. Herzberg's books, in particular his Spectra of Diatomic Holecules (Van Nostrand, Nevl York, 1939 and 1950) describe very
vividly both the applicability of the classical picture and
the deviations from it in the case of identical atoms.
I. Kovacs' Rotational Structure in the Spectra of Diatomic
110lecules (Hilger, Londori, 1969) gives even further details.

d~r

Hin. Soc.).

E. Nether,

Physik 22,575 (1917).

892

E. P. WIGNER

16)

K.F. Bonhoffer and P. Harteck, Z. fur Physikalische Chemie B4,

113 (1929).

17)

J. Callaway, Electron band theory (Academic Press, New York, 1964).

18)

~1.J.O.

19)

A very interesting discussion on this subject was presented

at the 14th Solvay Conference: Symmetry properties of
nuclei (Gordon and Breach, 1974).

20)

M. Ge11-Man and Y. Ne'eman, The eightfold way (W.A. Benjamin,

New York, 1964).
Highlights in particle physics (ed. A. Zichichi) (Editrice
Compositori, Bologna, 1973).

Strutt, Ann. der Physik 85, 129 (1928).

F. Bloch, Z. fur Physik 52, 555~1928).
R. Peierls, Ann. der Physik 4, 121 (1930).
L.P. Bouckaert, R. Smolucho~ki and E. Wigner, Phys. Rev. 50,
58 (1936).

CLOSING CEREMONY

The Closing Ceremony took place on Saturday, 7 August 1976.

The Director of the School presented the prizes and scholarships as
specified below.
PRIZES AND SCHOLARSHIPS
Prize for Be6~ Student - awarded to William J. MARCIANO Rockefeller University - New York, NY, USA.
Thirteen scholarships were open for competition among the participants. They were awarded as fellow:
P~ck M.S. Blackett Scholarship - awarded to William J.
MARCIANO, Rockefeller University, New York, NY, USA.

Gunna4 Kallen Scholarship - awarded to Barbara YOON, MIT,

Cambridge, MA, USA.

Jame6 Chadwick Scholarship - awarded to Nigel H. PARSONS,

Glasgow University, Glasgow, UK.
Amo~-de

Shatit Scholarship - awarded to Gideon BERLAD, TECHNION,

Haifa, Israel.

Scholarship - awarded to Myron R. PAULI, MIT,

Cambridge, MA, USA.

An~e Lag~gue

G~o Racah Scholarship - awarded to Giora J. TARNOPOLSKY,

ETH, Zurich, Switzerland.
Gio~io Ghigo Scholarship - awarded to Arturo GARCIA, Oxford
University, Oxford, UK.
E~co P~ico Scholarship - awarded to Fred POSNER, Harvard
University, Cambridge, ~~, USA.
P~~ P~~w~k Scholarship .- awarded to Thomas D.B. WILKIE,
University of Durham, Durham, UK.

893

894

CLOSING CEREMONY

Scholarship - awarded to Junko SHIGEMITSU,

Cornell University, Ithaca, USA.

G~~ Quane~

Anto~o Stan9hetti~ Scholarship - awarded to Alberto C. DE LA

TORRE, Gesamthochschule, Wuppertal, Germany.

Atbento Tom~~ Scholarship - awarded to Hung PRAM QUANG, CERN,

Geneva, Switzerland.
Etto~e Majo4ana Scholarship - awarded to Paolo ROSSI, Scuola
Normale Superiore, Pisa, Italy.

Prize for Be6.t Scien:tt6-i-c.

MIT, Cambridge, MA, USA.

SeM~Y

- awarded to Myron R. PAULI -

Prize for Hono~y Be6.t Student - awarded to Eugene P. WIGNER,

Princeton University, Princeton, NJ, USA.
The following participants gave their collaboration in the
scientific secretarial work:
Yachin AFEK

William J. MARCIANO

Orlando ALVAREZ

Bruno MATHIS

Andrea AUDRITO

Nigel H. PARSONS

Dario BISELLO

Myron R. PAULI

Fritz BOPP

Hung PRAM QUANG

Franco CERVELLI

Fred POSNER

Andre DEGRE'

Gabriele PUGLIERIN

Alberto C. DE LA TORRE

Paolo ROSSI/Pisa

Massimo FALCIONI

Junko SHIGEMITSU

Barry A. FREEDMAN

Pasquale SODANO

Arturo GARCIA

Giora J. TARNOPOLSKY

Paolo GIUSTI

Karl-Ludvig WERNHARD

M. Alimi ICHOLA

Thomas D.B. WILKIE

Bernard JANCEWICZ

Barbara YOON

David R.T. JONES

PAR TIC I PAN T S

Yachin AFEK

TECHNION - Israel Institute of

Technology
Department of physics
HAIFA 32000, Israel

Orlando ALVAREZ

Harvard University
Department of Physics
CAMBRIDGE, HA 02138, USA

Ivan ANDRIC

Universitat Bielefeld
Abteilung Theoretische Physik
Universitat Strasse
48 BIELEFELD, D

Andrea AUDRITO

Istituto di Fisica dell'Universita

Corso }1assimo d 'Azeglio, 46
10125 TORINO, Italy

MilIa BALDO-CEOLIN

Istituto di Fisica dell'Universita

via }1arzolo, 8
35100 PADOVA, Italy

Haurizio BASILE

CERN
EP Division
1211 GENEVA 23, Switzerland

Rene BERGER

Husee des Beaux Arts

Palais de Rumine
1005 LAUSANNE, Switzerland

Gideon BERLAD

TECHNION - Israel Institute of

Technology
Department of Physics
HAIFA 32000, Israel

895

896

PARTICIPANTS

Dario BISELLO

Istituto di Fisica dell'Universita

via Harzolo, 8
35100 PADOVA, Italy

Volker BLOBEL

Universitat Hamburg
II. Institut fur Experimentalphysik
Notkestieg 1
HAlffiURG 52, D

Fritz BOPP

Gesamthochschule Siegen
Fachbereich HathematikNaturwissenschaften-Physik
Holderlinstrasse 3
59 SIEGEN 21, D

Franco BUCCELLA

Istituto di Fisica dell'Universita

piazzale delle Scienze, 5
00185 ROt-fA, Italy

Elio CALLIGARICH

INFN - Sezione di Pavia

via A. Bassi, 6
27100 PAVIA, Italy

Giovanni CARARm1EO

CERN
EP Division
1211 GENEVA 23, Switzerland

Roberto CASALI

INFN - Sezione di Pisa

via Livornese
56010 SAN PIERO A GPJillO (Pisa),
Italy

Franco CERVELLI

INFN - Sezione di Frascati

Casella Postale 70
00044 FRASCATI (Roma), Italy

Reinhold CHRISTIAN

Institut fur Theoretische Physik

der Universitat wi en
Boltzmanngasse 5
1090 WIEN, Austria

Luisa CIFARELLI

CERN
EP Division
1211 GENEVA 23, Switzerland

Jean CLEYl1ANS

Universitat Bielefeld
Abteilung Theoretische Physik
Universitat Strasse
48 BIELEFELD, D

897

PARTICIPANTS

James W. CRONIN

The University of Chicago

The Enrico Fermi Institute
5630 Ellis Avenue
CHICAGO, IL 60637, USA

David CUTTS

Brown University
Department of Physics
PROVIDENCE, RI 02912, USA

Luigi DADDA

Politecnico di l1ilano
piazza Leonardo da Vinci, 32
20133 tULANO, Italy

Richard H. DALITZ

The University of Oxford

Department of Theoretical Physics
12 Parks Road
OXFORD OXl 3PQ, UK

Andre DEGRE

Centre de Recherches Nucleaires

Labo PNPP-HE
B.P. 20/CRO
67037 STRASBOURG CEDEX, France

Alberto C. DE LA TORRE

Gesamthochschule Wuppertal
Fachbereich 8 - Physik
Hbfkamp 82-64
56 WUPPERTAL, D

Bernard D'ESPAGNAT

Universite Paris XI
LPTPE
Batiment 210
91 ORSAY, France

John C. ECCLES

"Ca' a la Gra'"
6611 CONTRA (Locarno) TI
Switzerland

Etim ETU1

CERN
Th Division
1211 GENEVA 23, Switzerland

Christian W. FABJAN

CERN
EP Division
1211 GENEVA 23, Switzerland

Massimo FALCIONI

Istituto di Fisica dell'Universita

piazzale delle Scienze, 5
00185 ROMA, Italy

898

PARTICIPANTS

Thomas FERBEL

The University of Rochester

River Campus Station
Department of Physics
ROCHESTER, NY 14627, USA

Harm FESEFELDT

Max-Planck-Institut fur Physik und

Astrophysik
Institut fur Physik
FohEinger Ring 6
8 UUNCHEN 40, D

Barry A. FREEDHAN

Massachusetts Institute of Technology

Room 6-416
Center for Theoretical Physics
CAMBRIDGE, HA 02139, USA

William F. FRY

University of Hisconsin-Hadison
Department of Physics
1150 University Avenue
MADISON, WI 53706, USA

Arturo GARCIA

The University of Oxford

Department of Theoretical Physics
12 Parks Road
OXFORD OXI 3PQ, UK

Paolo GIUSTI

CERN
EP Division
1211 GENEVA 23, Switzerland

Hichel GOURD IN

Universite Paris VI
LPTHE
Tour 16 - ler etage
4 place Jussieu
75230 PARIS CEDEX 05, France

U. Alimi ICHOLA

College de France
Laboratoire de Physique Corpusculaire
11 place Harcelin-Berthelot
75231 PARIS CEDEX 05, France

Bernard JANCEWICZ

Uni versi ty of vlroclaw

Institute for Theoretical Physics
ul. Cybulskiego 36
50205 WROCLAW, Poland

Peter D. JARVIS

Imperial College of Science and

Technology
The Blackett Laboratory
Prince Consort Road
LONDON SW7 2BZ, UK

PARTICIPANTS

899

Kjell JOHNSEN

CERN
ISR Division
1211 GENEVA 23, Switzerland

David R.T. JONES

The University of Oxford

Department of Theoretical Physics
12 Parks Road
OXFORD OXI 3PQ, UK

Hagen KLEINERT

Freie Universitat Berlin

FB 20, WE 4
Arnimallee 3
1000 BERLIN 33, D

Otto KOFOED-HANSEN

Danish Atomic Energy Commission

Research Establishment RIS0
4000 ROSKILDE, Denmark

Johann KUHN

Max-Planck-Institut fur Physik und

Astrophysik
Institut fur Physik
Fohringer Ring 5
8 MUNCHEN 40, D

Elliot LEADER

University of London
Westfield College
Kidderpore Avenue
LONDON NW3 7ST, UK

Endre LILLETHUN

University of Bergen
Department of Physics
Allegt. 55
5014 BERGEN, Norway

Harry LIPKIN

The Weizmann Institute of Science

Department of Nuclear Physics
REHOVOT, Israel

Laurence S. LITTENBERG

Brookhaven National Laboratory

Physics Department
UPTON, NY 11973, USA

l\Tilliam J. MARCIANO

Rockefeller University
Physics Department
NEW YORK, NY 10021, USA

Bruno 11ATHIS

Freie Universitat Berlin

Institut fur Theoretische Physik
Arnimallee 3
1000 BERLIN 33, D

900

PARTICIPANTS

Andrew l1cPHERSON

Rutherford Laboratory
CHILTON, Didcot
Oxon., OXll OXQ, UK

Ezio HENICHETTI

Istituto di Fisica dell'Universita

Corso Massimo d'Azeglio, 46
10125 TORINO, Italy

Adriana HINGUZZI RANZI

Istituto di Fisica dell'Universita

via Irnerio, 46
40126 BOLOGNA, Italy

Claudio ORZALESI

CERN
Th Division
1211 GENEVA 23, Switzerland

Oliver E. OVERSETH

The University of Hichigan

The H.H. Randall Laboratory of
Physics
ANN ARBOR, HI 48104, USA

Nigel H. PARSONS

Glasgow University
Department of Natural Philosophy
GLASGOW G12 8QQ, UK

Myron R. PAULI

Hassachusetts Institute of Technology

Room 6-4llA
Department of Physics
CAHBRIDGE, HA 02139, USA

Dubravko PEVEC

Ruder Boskovic Institute

P.O. Box 1016
41001 ZAGREB, Croatia, Yugoslavia

Hung PHAH QUANG

CERN
Th Division
1211 GENEVA 23, Switzerland

Fred POSNER

Harvard University
Department of Physics
CAl1BRIDGE, liA 02138, USA

Giuliano

CERN
Th Division
1211 GENEVA 23, Switzerland

PP~PARATA

Gabriele PUGLIERIN

Istituto di Fisica dell'Universita

via F. Marzolo, 8
35100 PAD OVA , Italy

PARTICIPANTS

901

Gisela RANFT

Karl-Harx-Universitat
Sektion Physik
Linnestrasse 5
701 LEIPZIG, DDR

Leonardo ROSSI

INFN - Sezione di Genova

viale Benedetto XV, 5
16132 GENOVA, Italy

Paolo ROSSI

Istituto di Fisica dell'Universita

via F. Uarzolo, 8
35100 PADOVA, Italy

Paolo ROSSI

Scuola Normale Superiore

piazza dei Cavalieri, 7
56100 PISA, Italy

Jun John SAKURAI

CERN
Th Division
1211 GENEVA 23, Switzerland

Dieter SCHLATTER

CERN
EP Division
1211 GENEVA 23, Switzerland

Stuart J.

University College of London

Department of Physics and Astronomy
LONDON WC 1, UK

SHP.u~OCK

Junko SHIGEIIITSU

Cornell University
Laboratory for Nuclear Studies
ITHACA, NY 14853, USA

A.J. Stewart SllITH

Princeton University
Department of Physics
Joseph Henry Laboratories
Jadwin Hall
P.O. Box 708
PRINCETON, NJ 08540, USA

Pasquale SODANO

University of Alberta
Department of Physics
EDt-1ONTON T6G 2J, Canada

Uartin SOHNIUS

Uax-Planck-Institut fur Physik und

As trophysik
Institut fur Physik
Fohringer Ring 6
8 UUNCHEN 40, D

902

PARTICIPANTS

Giora J. TARNOPOLSKY

L.H.E.
Eidgenossische Technische Hochschule
ZURICH, Switzerland

Val L. TELEGDI

The University of Chicago

The Enrico Fermi Institute
5630 Ellis Avenue
CHICAGO, IL 60637, USA

Jean-Uichel THENARD

CERN
EP Division
1211 GENEVA 23, Switzerland

Sam C.C. Ting

Uassachusetts Institute of Technology

Department of Physics
CAUBRIDGE, UA 02139, USA

Giovanni VALENTI

CERN
EP Division
1211 GENEVA 23, Switzerland

Patrizio VINCIARELLI

CERN
Th Division
1211 GENEVA 23, Switzerland

Guy von DARDEL

University of Lund
Solvegatan 14
223 62 LUND, Sweden

Gustav Adolf VOSS

DESY
Notkestieg 1
2000 HAl1BURG 52, D

Raymond \VEILL

Institut de Physique Nucleaire

B.S.P. - Dorigny
1015 LAUSANNE, Switzerland

Steven WEINBERG

Harvard University
Lyman Laboratory of Physics
CAl1BRIDGE, UA 02138, USA

Karl-Ludvig WERNHARD

CERN
EP Division
1211 GENEVA 23, Switzerland

Gian Carlo WICK

Columbia University
Physics Department
P.O. Box 132
NEW YORK, 10027, USA

903

PARTICIPANTS

Eugene P. WIGNER

Princeton University
Department of Physics
Joseph Henry Laboratories
Jadwin Hall
P.O. Box 708
PRINCETON, NJ 08540, USA

Thomas D.B. WILKIE

University of Durham
Department of Physics
Science Laboratories
South Road
DURRAH DHl 3LE, UK

Catherine WILQUET

Inter-University for High Energies

ULB-VUB
avenue de la Plaine, 2
1050 BRUXELLES, Belgium

Chen Ning YANG

SUNY
Institute for Theoretical Physics
STONY BROOK, NY 11794, USA

Barbara YOON

Hassachusetts Institute of Technology

Room 6-415
Department of Physics
CAliBRIDGE, IrA 02139, USA

Thomas YPSILANTIS

CCSEH
91016 ERICE, Italy

INDEX

Abelian gauges
monopoles in, 811
monopole solitons in, 818
Antiferromagnets, 27
Antihadron-nuclear total cross
sections, 558
Antineutrinos, 442
scattering with SU(4), 474
Antiquark distributions, 493
Asymptotic behaviour, 253, 265,
267, 274
background ~p scattering,
274, 280
Atoms, parity violation in, 431
Atomic spectra,
symmetry and, 888
Azimuthal correlation
local conservation of
transverse momentum and,
789
in particle production at
low p, 777
Backlund transformation, 106
Backward scattering, 274, 280,
398
Bags
charge form factor, 158
exchange, 137, 140
exchanged states, 130
four bag coupling, 129, 174
pomeron and, 140 '
oscillating, solition
solutions, 107
qq Greens functions, 123
quark motion in, 116
SLAC, 106

Bags (cont'd)
three bag coupling, 126, 174
Band structure, 792
Baryons, 181
construction of, 173
coupling, 200
exchanges, 276, 280, 285
magnetic moments, 389
number, 232
selection rules, 203
Bianchi commutability theorem,
87
Bilocal currents, 325
Bilocal hadron theory, 292
Bjorken canonical scaling,
446, 459
Bloch-Nordsieck problem, 9, 50
Bose effect
cluster model, 781
for like particles and cluster
model, 779
second order interference and,
788
Bosons, 184
critical phenomena in, 48
Green's functions for, 5
Higg's, 45, 46, 48
in nucleus, 182
second order phase transition,
5, 51
W-Boson production, 691
Brower's result, 96
Calculus of exterior forms, 112
Cartan's theorem, 108
CERN intersecting storage rings,
611
905

906

Chan-Hong Mo model, 248

Charm, 180, 232, 253
evidence for, 674
Charmed mesons, 674
semileptonic decay, 677
Charmed particle production,
OZI rule and, 228
Charmonium model, 671
J/$ family interpreted by,
673
Charm violation processes, 45
Cluster model, Bose effect,
779, 781
Coherent droplet model, 81
Collective tube model, 684
hadron nucleus collisions in,
683
new particle production and,
688
Colliding trajectories, 283
Colour, 189, 382, 384
in quark gluon theory, 328,
354
Confinement, 383, 387
Confusion theorem, 396, 441
Creation and annihilation
operators, 78
Critical phenomena, 1, 2
in boson systems, 48
calculation of eigenvalues,
31
external field problems, 20
fixed points, 26, 27
fixed points and scaling
laws, 16
floating cut-off, 8, 21, 33
Gaussian fixed points, 26
invariance of eigenvalues, 21
one-loop equations, 23
renormalization theory,S, 6,
11, 15, 21, 32, 36, 50, 51
three-dimensional field
theory and, 6, 8
Cross section data, 233, 234,
235, 236, 237, 252
Crystal symmetry, 880, 886
Current mixing terms, 349
Currents and normalization, 131
Deep inelastic coupling, 176

INDEX

Deep inelastic scattering,

146, 154, 244
pomeron contribution to, 149
scaling phenomena in, 149
Deuterium targets, inelastic
processes with, 697
Deuteron, disintegration of, 419
Dilepton continuum with high
invariant mass, 664
Dilepton production
cross section, 493, 494
importance of, 485
in proton-nucleus collisions,
485
from vector meson decay, 667
yield, 498, 501
Dimuons
cross sections, 728
non-resonant, 726
production, 486, 487, 495,
670, 739
experimental, 703
by pions and protons, 701
pion beam measurement, 505
results, 708, 739
results at 225 GeV/c, 716
scaling, 495, 504
single lepton puzzle, 714
transverse momentum, 495, 506
variables of, 702, 709
Dirac electron motion of, 63, 70
Dirac quantization, 63, 70, 72
Drell-Yan model, 493, 495, 665
Duder's paradox, 77
Dyons, 73
Eigenvalues, 51
calculation of, 31
invariance of, 21
Eigenvectors, infra-red
attraction or repulsions,
17, 51
Elastic scattering
antiparticle and particle, 561
impact parameter analysis, 742,
750
neutral currents and, 422, 423
Regge theory and, 259
Electron
detection efficiency, 539

INDEX

Electron (cont'd)
identification of, 539
production of, 646
wave functions, 53
around monopoles, 57, 69,
76
Electron-muon coincidences, 652
Electron-nucleon scattering,
412, 494
Electron-positron annihilations,
231, 426, 433
Electron-positron scattering,
jets from 510, 511
Electron-positron collisions,
511
distribution of events, 513
jets from, 511
Energy flux cascade model, 580
External field problems, 20
Factorization, 284
Fermilab
dimuon production at, 701
hadron physics at, 555
single arm spectrometer, 741
Fermions, 89, 106
Fermion-antifermion pairs, 188
Fermion-Bose-8alpeter equation,
310, 311, 355
Fermion fields, anti
commutability, 106
Fibre bundle theory, 64, 65
8U(2), 75
Fictitious Wtheory, 136
Finite energy configuration, 93
Firesausages, 121, 174, 177
production of, 141
Fixed points, 16
Gaussian, 26, 34, 36, 52
tricritical, 27
Wilson-Fisher, 27, 31, 35
Floating cut-off, 8, 33, 39
versus renormalization, 21
Fluctons, 509, 511, 516
Form factors, 78

907

Gargamelle experiments (cont'd)

432, 445, 446, 449, 464,
467
analysis of, 451
features and predictions, 452
Gauge invariance, 385
Gauge theory, 883
monopoles and, 805
neutral current interactions
and, 391, 393
Gaussian fixed points, 26, 34,
36, 52
Generalized color models, 189
Glauber theory, 697
Globe, parametrizing, 55
Gluons
algebraic derivations, 371
colour, 382, 384
confinement and, 383, 387
external, interaction with
hadrons, 324
in hadronization, 292
heavy, 389
mass limits, 329
Regge theory and, 327
role of, 290
theory of, 299
quantization, 304
without colour, 329
very heavy, 327
vertices for, 366
Goldhaber effect, 796
Gravitation as gauge field, 76
Green's functions, 49
renormalization group
equations, 15

Hadrons
bare, 291, 385
interactions, 292
building, 181
charmed, 677, 680
coupling, 175, 188
decay properties, 122
deep inelastic scattering, 146
density, 81, 83
G-parity, 246
direct current coupling, 145
Galilei, 881, 882
emission, 323
Gargamelle experiments, 391, 393, excitation into multiparticle
395, 402, 404, 418, 425,
systems, 577

908

Hadrons (cont'd)
e+-e- annihilation,
142, 187
heavy leptons and, 678
infinite component field,
312
interactions, 123, 125
jet-like production, 513
large PI , 150
mass, 388
neutrino-induced reaction,
394
new, 464
off shell effects, 130
perturbation theory and, 125
physical states, 117
production, 510
high transverse momenta
data, 685
from nuclear targets, 685
scaling phenomena, 142
scattering, 626
scattering amplitude, 125
irreducible kernal V6 ,
126
kernal Va' 129
semi-inclusive FS decay, 152
spectrum, 183
strange and non-strange, 254
totally inclusive FS decay
at large angle, 150
vector dominated contribution,
143, 145
Hadron - hadron interactions
dimuon production, 701
elastic scattering, 744
inclusive single lepton
production, 669
lepton production, 174, 175,
663
from vector meson decay,
667
Hadron - hadron scattering,
73, 81, 136, 160, 744
bag exchange, 137
high energy, 139
production of resonance, 136
Regge behaviour, 139
Hadronic currents, 407
Hadronic density distributions,
757

INDEX

Hadronic matter, 115-172

currents and normalization, 133
quark theory, 115
similarity of, 584
Hadron interactions, 555
elastic scattering cross
sections, 556, 559
hydrogen targets, 584
inclusive particle production,
578
inclusive production spectra
and limiting behaviour, 589
inelastic diffraction
production, 569
inclusive, 574
selected exclusive channels,
569
meson annihilation crosssection, 558
multiplicities, 584
neutral-proton charge exchange,
565
neutrino-induced
exclusive, 419, 443
inclusive, 410, 443
nuclear targets, 578
off shell inclusive scattering,
598
particle production
energy dependence, 590
energy flux cascade model,
580
space time development, 578
total cross-sections, 556
two-body charge exchange
reactions, 564
Hadronization
functional techniques, 293
gauge invariance and, 385
Regge theory and, 388
quark theories, 283
Hadron - nucleon collision, 581
in collective tube model, 683
Hadron nucleon cross section, 558
Hadron - proton collisions,
meson production, 595
Hadron scattering, 238
Hagedorn-Frautschi
representations, 850
Hard scattering, 508, 509
Heavy lepton production, 740

909

INDEX

, 230, 485, 555

production of, 494, 701, 717
735
cross sections, 724
J/$ family, 671
charm and, 674
intermediate states, 671
interpretation by charmonium
model, 673
J particle, 230
Ideal mixing in aZI rule, 212,
Jets, 507, 640
217, 220
charge multiplicity, 536
Impact parameter analysis of
collimated, 534
elastic scattering, 742,
coordinate systems, 532
750
coplanar, 626
Inclusive particle production,
detection of, 516
acceptance and normalization,
578
nuclear targets, 578
532
Inclusive scattering, 772
centre of mass motion, 535
Inelastic average charged
from electron-positron
particle multiplicities,
scattering, 510, 511
general features, 516
588
Inelastic diffraction production,
mass of, 535
properties of, 509
569
inclusive, 574
from proton - proton collision,
selected exclusive channels,
510, 514
569
sphericity concept, 536
Inelastic exchange reaction,
Kaons, 553
767
Inelastic lepton scattering,
density distribution, 757, 760
Kaon pair states, 205, 206,
473
Inelastic processes with
250, 251
deuterium targets, 697
Kaon wave functions, 232
Inelastic scattering, 762
Kinematic invariance, 881
Korteweg de Vries equation, 86
bilocal currents in, 325
categories, 763
data, 765, 766, 767
Laws, 881
Infinite potential energy, 92
Leptons
Infra-red divergencies, 8, 9, 11,
classification, 845
50
decay, 646
Instanton, 99, 112
heavy, 678, 740
Interactions
inclusive production, 669
in Regge framework, 255
kinematics, 542
theory of, 290
production, 174, 175, 646
Invariance, 879
in hadron collisions, 663
Newton's first law, 882
inclusive. 669
transformation, 883
prompt, 702
vality of, 887
single lepton puzzle, 713
Isobaric analog states, 179
Lepton - hadron scattering,
Isoscalar isovector interference
deep inelastic, 680
413, 415, 425
Lepton pairs, production of, 664

Helicity rule, 398, 400

Higg's boson, 45, 46
Higher order paradoxes of aZI
rule, 205, 207, 228
Hilbert space of sections, 57
Homotopy, 94, 97, 108, 112
definition of groups, 111
relation of groups with
connections, 108

J/~

910

INDEX

Lepton pairs production (cont'd) Monopoles (cont'd)

importance of, 485
evidence for, 71
in proton-nucleus collisions,
existence of, 802
gauge theory and, 800, 801, 805
485
rates, 498
harmonic Y ,1m, 58
by two photon process, 493
completeaess of, 61
from vector meson decay, 667
examples and analyticity, 61
yield, 501
explicit expressions, 59
Lumps, 110
history of, 799
magnetic charge, 820, 838
Massive quark model, 125, 174
mass, 800, 835
Mass mixing terms, 349
models
Mass spectrum, growth of, 876
No.1. in radial gauge, 805
Mesons, 153, 181
't Hooft and Polyakov, 97, 108,
charmed, 674
109, Ill, 802
semileptonic decay, 677
origin of magnetic charge, 820
coupling, 200, 208, 209, 340
regions, 54, 70
decay, 192, 250
Schrodinger equation, 62
dileptons from, 667
singularities, 54
decoupling high states, 120
soliton type, 74
degeneration, 350
space in presence of, 813
equations of motion, 117
spin of, 76
exchange, 285
SU(2). 72, 74
interactions, colour and, 384 Monopole solitons, 815, 834
momentum space wave function,
electrodynamics of
119
non-relativistic spin ~
production, OZI rule, 225
particle, 831
model No. 1 in abelian gauge,
Regge pole exchange, 258
scattering, 320
818
model No.2, 823
strange and non-strange, 235
Mueller-Regge phenomenology, 592
vector, 345, 346, 349, 352
Multimuon events, 735
electrodynamics of, 823
Muons, 440
wave function, 117
identifiers, 538
Meson - nucleon scattering, 269
origin of, 491
Meson selection rules, 203
production, 670
K Mesons, production, 537,
Muon - nucleon scattering, 426
545, 546, 593
W Mesons, production, 691
Muon pairs
Molecular physics, symmetry
non-resonant, 726
production of, 426, 486, 487,
and, 888
495, 701, 739
Monopoles, 53, 799
in abelian gauges, 811
experimental, 703
Dirac and, 800
pion beam measurement, 505
Dirac equation in field, 63,
results, 708, 739
results at 225 GeV/c, 716
70
single lepton puzzle, 713
division of space outside, 55
transverse momentum, 495,
electrodynamics of vector
506
mesons, 823
variables of, 702, 709
electron wave function
scaling, 495, 504
around, 57, 69, 76

INDEX

Nambu-Jona-Lasinio model, 352

Narrow continuum states, 230
Nature, laws of, 881
Neutral currents, 477
diagonal, 478
quark parton models, 470
left handed models, 465
Neutral current interactions,
391
astrophysical theory, 396,
420
confusion theorem, 396
cross section ratios, 415,
416, 417
forward-backward asymmetry,
430
gauge theory and, 391, 393
hadronic part, 407
isosca1ar-isovector
interference in, 413,
415, 425
1eptonic aspects, 430, 439,
440, 441
models, 409, 432
parity violation and, 430
SPT heresy, 396
strength of, 418
theories of, 394
universality, 439
without neutrinos, 426
Neutral kaon system, 197
Neutrino
massless, 394
nature of, 439
neutral current interaction
and, 393
scattering with SU(4), 474
Neutrino - antineutrino
inclusive reactions, 446
Bjorken canonical scaling,
446
energy distribution, 447, 449
Fermi1ab data, 453
Gargame11e data, 445, 446,
449, 451, 452
Neutrino-electron scattering,
397, 433, 441
Neutrino induced exclusive
hadronic reactions, 410,
419

911

Neutrino interactions
apparatus, 537
detection efficiency, 549
electron detection efficiency,
539
e+~- events, 537
characteristics, 540
energy dependence of
production rate, 544
loss due to simulated
Da1itz pairs, 540
properties of, 551
electron production, 548, 550
energy involved, 552
experimental details, 538
hadron involvement, 552
KO phenomena, 546, 552
loss of events at energies
Ee+ < .8GeV, 544
positron production, 548, 550
rate of ~-e+, 540
scanning, 538
strange particles, 545
Neutrino scattering, bi1oca1
currents, 325
Neutron, electrical dipole
moment, 47
Neutron beta decay, 421, 422
Neutron - proton charge exchange
scattering, 269, 565
New degrees of freedom, 185
Newton's first law, 881
Non-relativistic spin ~ particles,
831
Nuclear physics, symmetry in, 889
Nuclear targets, hadron
production from, 685
Nucleon cross sections, 233
Nucleon - deuterium inelastic
scattering, 286
Nucleon - nucleon scattering, 269
Nucleus
bosons in, 182
mu1tipartic1e production, 583
Nucleus - nucleus collisions, 691
jets from, 508
Oct onions , 841
Off shell behaviour, 117, 130,
139

912

Off shell inclusive scattering,

598
Okubo-Zweig-lizuka rule, 179,
189, 190, 322, 373
application to new particles,
227
basis of, 195
charmed particles production
and, 228
double forbiddenness, 193
dynamics, 197, 199
at SU(3) level, 199
evidence for, 195
examples, 190
exchange degeneracy and, 247,
248
experimental tests, 222
higher order paradox, 193,
194, 195, 196, 205, 207,
228
ideal mixing and symmetry
cancellations, 212, 217,
220
meson production and, 225
mixing K1-K 2 analogy, 196
multiparticle production,
250, 251
quark line selection ru'les,
200
SU(3) symmetry, 224
in SU(6)w symmetry, 211
symmetry and, 198
dynamics, 196
three point functions, 205,
245
unitarity paradox, 205, 207
validity of, 216
violation, 206, 215, 223,
224, 246
quantitative estimate,
217
One loop equations, 23
derivation of, 38
Opacity, 751
Oscillating bags, soliton
solutions, 107
Parity violation, 430
Particles
Bose effect, 779

INDEX

Particles (cont'd)
correlation in high PTfinal
states, 626
heavy long-lived, 774
leptonic decays, 646
multiplicity, 535, 536
production, 578
energy dependence, 590
energy flux cascade model,
580
inelastic multiplicities,
579
production at low p,
azimuthal correlation, 777
Particle - nucleus collisions,
691
Particle nucleus scattering, 683
Perturbation theory, 125
Photons, 740
coupling, 383, 396
dilepton production, 493
directly produced, 654, 662
exchange, 358, 382
Pion charge-exchange scattering,
564
Pion exchange, 214
Pion form factor, 82
Pion-induced lepton pairs, 664,
667
Pionium, 680
Pion - nucleon charge exchange
reaction, 256, 260
Pion poles, 286
Pions, 796, 797
density distribution, 757, 760
emission, 201
exchange, 287
production, 252, 424, 593, 594
semi-inclusive reactions, 425
spectra, 589
wave functions, 232
Plasmonization, 386
Poincare's fundamental group, 96
Poincare symmetry, 882, 885, 887
Pomerons, 139, 149, 154, 236,
257
exchange, 247, 283
factorization, 284
four-bag coupling and, 140
nature of, 236

INDEX

Pomerons (cont'd)
primeval, 124, 133, 135, 174
two component, 238
Preparata's lines, 383
Projectile fragmentation region,
762
Proton, density distribution of,
757, 760
Proton-induced lepton pairs,
664, 667
Proton - nucleus collisions,
dilepton production in,
485
Proton - proton interaction,
507
W boson production, 691
charged particles in the
hemisphere opposite to
640
charged particles produced
in n hemisphere, 636
cross sections, 621, 622,
623, 624, 625
differential cross sections,
616, 618
direct lepton production,
646
electron muon coincidences,
852
electron production, 646
jets from, 510, 514, 640
J production, 652
large scale elastic
scattering, 615
large transverse momentum
phenomena, 619
mass distributions in high
p events, 522
P outTdistribution, 640
particle correlation high
PT final states, 626
single particle inclusive
distributions, 620
total cross sections, 611
transverse momentum, 626,
631, 633, 634, 635
two-photon events, 626, 629
Psi, 230
Quantum field equations,
instanton, 99, 100, 102, 112

913

Quantum field equations (cont'd)

non-trivial solution, 100, 102
vacuum values, 93
Quantum field theory, 1
critical phenomena, 1-52
cut-offs, 51
extended objects, 110
fixed points and scaling
laws, 16
one-loop equation, 38
renormalization, 50, 51
group equations, 11
statistical bootstrap model
and, 849, 856
cluster product, 858, 870
general remarks, 870
Hagedorn-Frautschi and
Yellin representation, 850
phase space, 862
statistical mechanics and, 2
two, three and four dimensional
space, 6, 8, 95
at zero temperature, 6
Quantum mechanics
linear states, 885, 886
symmetry and, 884
symmetry and invariance and,
879
Quantum numbers
exchange of, 594
local compensation, 596
Quantum solitons, 109
Quarks, 173, 181, 187
additive model, 239
bare, 389
bilocal, 299
charmed, 180
classification, 845
color and octonions, 841
coupling, 345
to mean states, 120
currents coupling to, 133
diagrams for 3 point functions,
202, 203, 204, 205
elementary transitions, 474
fictitious ~ theory, 136
flow of lines, 318
forming isospin doublet,
131, 133
G parity, 246
gluon theory, 291, 299

914

Quarks (cont'd)
gluon theory (cont'd)
algebraic derivation, 371
colour in, 354, 382, 384
confinement and, 383, 387
extended, 324
Nambu-Jona-Lasimo model
and, 352
quantization, 304
Regge theory and, 327
without colour, 329
hadronization of theories,
283
ideal mixing and symmetry
cancellations, 212, 217,
220
line models, 320
cancellations and
degeneracies in, 215
rules, 200
mass, 47, 253, 336, 340, 347,
348
model, 209
prediction, 775
motion, 116
production, 459
scattering amplitude, 323
scattering in mesons and
baryons, 237
'sea', 154
selection rule in SU(6)w
symmetry, 211
strange and non-strange,
245, 252
SU symmetry, 211, 304
in three bag coupling, 127
valence model, 154, 411,
434, 460
Quark-antiquark pairs, 180,
181, 187, 222, 227, 231,
702
annihilation, 244, 702
scattering on proton, 234, 235
Quark confinement, 116
Quark parton model, 412
energy distribution, 455
left handed current models,
454, 465
modified, 461
right handed current 460, 470

INDEX

Quark parton model (cont'd)

SU(3), 450
two-compartment, 450
Regge behaviour, 139, 141
Reggeization, 137, 154
Regge states, 125
Regge theory, 256, 280, 282, 287
elastic scattering and, 259
hadron interactions, 561, 568
quark-gluon theory and, 327
hadronization, 388
scattering amplitudes and,
259, 264
Relativity, 99, 880
Renormalization theory, 30, 32,
36, 50, 51
cut-offs and, 21
group equations, 11
Green's functions for, 15
Renormalized mass,S, 6
Russell-Saunders coupling, 886,
890
S-matrix elements, 2, 49
Salam-Weinberg model, 392, 404,
408, 413, 415, 420, 430,
432, 443
Scalar fields in five dimensions,
34
Scaling laws, 16
Scattering, See Elastic
scattering, Inelastic
scattering, Inclusive
scattering, etc.
Scattering amplitudes, Regge
theory and, 259, 264
Second order interference, Bose
effect and, 788
Sections, Hilbert space of, 57
Semi-conductors, 354
Signature, 282
Sine-Gordon equations, 106, 107,
110
Sine-Gordon solitons, 107, 111
Single leptons, solution of
puzzle, 713
SLAC bag, 106
Solid state physics, symmetry
and, 889

INDEX

Solitons, 85, 803

energy, 91
magnetic charge carried by, 821
monopole, 74, 815, 834
electrodynamics of
non-relativistic spin ~
particle, 831
model No. 1 in abelian
gauge, 818
model No.2, 823
particle physics and, 88
quantum, 109
Sine-Gordon, 107, 111
stability, 91, 92
three dimensional, 89, 91
topological, 92, 101, 109
in Yang-Mills theory, 99
Soliton-antisoliton solution, 88
Spectrometers, single arm, 741
future developments, 774
Sphericity, 511
Spin effects, 79, 80
Split field magnet, 515
Stars, evolution of, 397
Statistical bootstrap model,
field theory approach to,
849, 856
cluster product, 858, 870
general remakrs, 870
Hagedorn-Frautschi and
Yellin representation, 850
phase space, 862
Statistical mechanics
field renormalization, 5, 6,
11, 15, 21, 30
field theory and, 2
fixed points, 16, 26, 27, 33,
34
functional techniques, 49
Strangeness, 232
interactions, 233
Super conductors, 352, 353,
382, 839
SUc(3), 841
SU symmetry, 304
SU(2) symmetry, breakdown of, 351
SU(3) symmetry, 194, 196, 207,
232, 890
breakdown of, 347
OZI rule and, 199, 224

915

SU(3) symmetry (cont'd)

quarks and, 841
simple model, 207
SU(4) symmetry,
neutrino-antineutrino
scattering, 474
SU(6) symmetry, 181, 197, 245
nuclear, 182, 185
sicilian, 181
SU(6)w symmetry, OZI rule in,
211
Symmetry
atomic spectra and, 888
in molecular physics, 888
in nuclear physics, 889
in quantum mechanics, 884
in solid state physics, 889
validity of, 887
Symmetry breaks, 887
Symmetry cancellations in OZI
rule, 212
Symmetry operators, 879
Tbilisi conference, 663
Temperature Green's functions,
3, 49
Transverse momentum, 789
Tricritical fixed points, 27
Two-particle correlation, 778,
792, 797
Unitarity paradox, 205, 207
Vacuum states, 102
Valence quark model, 411, 434
Vector mesons
decay, 667
dileptons from, 667
electrodynamics, 823
Veneziano model, 876
Wave equations, 85, 86, 89
Wilson-Fisher fixed point,
27, 31, 35
Wrapping number, 102
Yang-Mills theory, 99
Yellin representation, 850
Zone-zone correlations, 796