Hadronic Matter at Extreme Energy Density

Hadronic Matter
at Extreme
Energy Density
ETTORE MAJORANA INTERNATIONAL SCIENCE SERIES
Series Editor:
Antonino Zichichi
European Physical Society
Geneva, Switzerland
(PHYSICAL SCIENCES)
Volume 1 INTERACTING BOSONS IN NUCLEAR PHYSICS

Edited by F. lachello
Volume 2 HADRONIC MATTER AT EXTREME ENERGY DENSITY

Edited by Nicola Cabibbo and Luigi Sertorio,
Volume 3 COMPUTER TECHNIQUES IN RADIATION TRANSPORT

AND DOSIMETRY
Edited by Walter R. Nelson and T. M. Jenkins
Volume 4 EXOTIC ATOMS '79: Fundamental Interactions and Structure

of Matter
Edited by Kenneth Crowe, Jean Duclos, Giovanni Fiorentini, and
Gabriele Torelli
Hadronic Matter
at Extreme
Energy Density
Edited by
Nicola Cabibbo
University of Rome, Italy
and
Luigi Sertorio
University of Turin, Italy
Plenum Press . New York and london

Library of Congress Cataloging in Publication Data
Workshop on Hadronic Matter at Extreme Energy Density, Erice, Italy, 1978

Hadronic matter at extreme energy density.
(Ettore Maiorana international science series: Physical sciences; v. 2)

"Proceedings of the Workshop on Hadronic Matter at Extreme Energy Density,
held at ... , Erice Italy, Dctober 13-21, 1978."
Includes index.
1. Hadros-Congresses. 2. Nuclear structure-Congresses. I. Cabibbo, N. II. Sertorio,
Luigi. III. Title. IV. Series: Ettore Maiorana international science series: Physical
science; v. 2.
OC793.5.H322W671978 539.7'216 79·18446
ISBN·13: 978·1-4684·3604·4 e·ISBN·13: 978·1·4684·3602·0
001: 10.10071978·1-4684·3602·0
Proceedings of the Workshop on Hadronic Matter at Extreme Energy Density,

held at the Ettore Maiorana Center, Erice, Italy, October 13-21, 1978.
© 1980 Plenum Press, New York

Softcover reprint of the hardcover 18t edition 1980
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
This book originated in the Workshop on "Hadronic Matter at

Extreme Energy Density," held at the Ettore Majorana Center in
Erice, October 13-21, 1978. The lectures have been expanded to
their present size, and the contributions of seven seminars have
been represented by abstracts which should stimulate the reader's
interest and guide him to the original literature.
The title of the book perhaps does not fully represent its
content but still is a good indication of the conceptual motiva-
tion of our Workshop.
The development of physics in recent years has filled in the

first details of the grand design which was initiated with the
theory of general relativity and aspires to a synthesis of all the
different interactions. However, this development has not been a
linear one but .has followed a divided pattern: general relativity
had its phenomenological domain in cosmology and had little to do
with high-energy elementary particle physics. It was progress in
the knowledge of symmetries in particle physics that fueled the
advance toward the present formulation of supergravity, thus help-
ing to heal this historical separation. The great program would
not have advanced so far if our attention had all the time stayed
focused at infinity, where the great issues are.
Other conjunctions between fundamental truths and researches

developing along their own paths may be found in statistical
mechanics. We may, for instance, remember the following histori-
cal steps: the theory of ideal gases and the first symptoms of
quantization; the understanding of black-body equilibrium; the
Fermi and Landau efforts to describe strong-interaction production
scattering by an ideal-gas approximation, and the huge effort to
overcome the limitations which this imposes, which is still going
on. Into this context we may also place the nonperturbative
statistical mechanics of Hagedorn and the first intuitive notions
of a new domain of phase transitions that followed from it, and
finally the new ideas of Hagedorn and Frautschi in the cosmology
of the hadron era.
v
PREFACE
The cosmology of the early moments after the big bang is the
milieu in which the conditions relevant to general relativity,
high-energy elementary particle physics, and high-energy
statistical mechanics naturally occur. In practice, the apprecia-
tion of this concept lay at the base of the Workshop and of its
triangular architecture: supersymmetry and supergravity -- chromo-
dynamics thermodynamics.
Sometimes the techniques, the immediate aims, and even the

long-range motivations of the participants seemed disparate, and
the unifying idea was hidden by the detail. But in the end it
became evident that a confrontation between scientists with
different interests but united by a nonrigorous idea was not
premature and certainly was worth the effort.
This is the year of Hagedorn's sixtieth birthday, and it is

a pleasure to have the chance to dedicate this book to him.
We wish to thank Antonino Zichichi for his leadership and

hospitality in Erice. We are also grateful to all the authors for
their friendly help in the process of assembling this book.
Nicola Cabibbo
Luigi Sertorio
CONTENTS
THEORETICAL LECTURES
Ericson Fluctuations and the New Argonne

Data on TIN Scattering . . . . . 1
S. Frautschi
Asymptotic Freedom and Color Superconduc-

tivity in Dense Quark Matter 18
S. Frautschi
Duality Transformation and Confinement in

Gauge Theories ..... . 29
F. Gliozzi
Thermodynamics of Nuclear Matter from the

Statistical Bootstrap Model . . . 49
R. Hagedorn
The Bag Model and Quantum Chromodynamics . . . • . . . . 149

J. Kuti
Quarks and Fermionic Geometry 187

J. Lukierski
Nuclear Fireballs in Heavy Ion Collisions 201

1. Montvay
A Critique of Astrophysical Applications

of Hagedorn's Bootstrap . . . . 219
W. Nahm
Particle Production in Soft and Hard Had-

ronic Collisions. Is There Evi-
dence for Hadronic Constituents? 241
J. Ranft
vii
viii CONTENTS
From Hadron to Quark Matter • • • • • • • • • • • • • • 269

H. Satz
Supersymmetry Approach to the Unification

of Interactions • • • • • • • • • 291
J. Schwarz
The Infra-red Behaviour of the Running

Coupling Constant in Yang-Mills
Theories •• • • • • • • • • • 313
F. Zachariasen
SEMINARS ON SPECIALIZED TOPICS
A New Viewpoint on De Sitter Gauge Invar-

iance in Gravity • • • • • • • • 335
P. Fre
Limiting Temperature, Lifetime, Size of the

Centrally Produced Hadronic Matter
and its Astrophysical Implications 337
L. Gutay
Evidence for Two Body Break-up at a Unique

Temperature in High Energy P-Xe and
P-Kr Collisions ••••••••• 339
L. Gutay
On the Application of Fields with Continu-

ous Mass Parameter 343
J. Lukierski
A Classical Theory of Strong Interactions

(with "Confinement") • . • • • • 345
E. Recami
Energy-Momentum Distribution in e+e- Anni-

hilation • • • • • • • • • • • • • 347
G. Pancheri-Srivastava and Y. Srivastava
Exponential Mass Spectrum and Pion Con-

densation • • •• • •••• 349
A. Tounsi
A Field Theoretic Description of Clusters

as an Approximation to the Statis-
tical Bootstrap Model • • • • • • • 351
L. ':urko
CONTENTS ix
"Small", "Large", and "Very Large" Trans-

verse Momenta in a Unified Hydro-
dynamical Description . . . • • • 353
R. Weiner
Contributors 355
Index 357
ERICSON FLUCTUATIONS AND THE NEW ARGONNE DATA ON TIN SCATTERING
Steven C. Frautschi
California Institute of Technology

Pasadena, California 91125
I. The Search for Highly Excited Hadron States: Introduction
}lost people nowadays think of hadrons as made of quark con~

stituents trapped in a potential well. The well mayor may not be
infinitely high; it must rise at least above the present energy range
to make free quarks so rare. As in any potential well, we expect to
find excited states right up to the top of the well. We also expect
the density of levels to rise rapidly with mass since there are so
many ways to form a highly excited level: anyone of the valence
quarks can be raised to an excited level, or more than one can be
raised at the same time, or quark pairs or gluons can be added.
The subject of this talk is an attempt to verify that hadron

excited states - specifically in the TIN channel - continue to occur
and become more numerous above the mass range covered in the Particle
Data Tables. The physical idea we attempt to use is that of Ericson
Fluctuations 1 • I will review these theoretical ideas, and then I
will describe some recent, as yet only partially published, experi-
mental results on this subject.
Specific theoretical estimates. Several specific models (statistical

bootstrap2, dual resonance mode1 3 , MIT bag mode1 4 ) suggest that the
number of hadron levels keeps increasing at high masses at the very
rapid rate
miT
dN _ p(m) ~ e 0 (1)
dm
miT
For concreteness I shall use p(m) ~ e 0 for all quantitative
2 STEVEN C. FRAUTSCHI
estimates in this talk. However, even if the level density grows at

a slower rate, most of the qualitative features should remain the
same.
Evidence from Particle Data Tables. In general levels can be divided

into two classes, separable and overlapping, for purposes of obser-
vation. Separable resonances are low enough so that no other reson-
ance with the same B,S,I,JP , ••• lies within one decay width of them.
Thus they are certainly separable in a partial wave analysis, and may
stand out as isolated peaks in certain reactions. Levels of this
type have individual names and are listed in the Particle Data Tables.
When we count them we find the results shown in Fig. 1 (here the
"experimental curve" includes only states with no charmed component.
However the charmed and cc states exhibit the same behavior - narrow
resonances near threshold, and then a smooth continuum from which
individual resonances cannot be separated).
Clearly one cannot use the/separable resonances to establish

mT
any distribution such as p ~ e 0,However, it is remarkable that
the7 observed curve is what one expects if p has the specific form
m T
e o. That is, for any given set of quantum numbers B,S,I,JP , ••• ,
one may have a stable ground state, perhaps a narrow resonance, and
then increasingly broad, ~losely-spaced resonances. Rough estimates S
-mIT
comparing the spacing e 0 between resonances to the width r ~ m
indicate that resonances begin to overlap at about 2 GeV for low 1T
J, and somewhat later for high J where the threshold for levels is
shifted upward in mass (Fig. 2).
R/n p (m)
miT
o
Ptheory = ce
PexperiIi1.ent
1 2 m
Fig. 1 Comparison of theoretical spectrum of hadron levels with

experiment.
ERICSON FLUCTUATIONS AND ARGONNE DATA ON 1TN SCATTERING 3
overlapping
resonances
m
(2 GeV) 2
Fig. 2 Schematic plot indicating where rp(mpJ) is large enough to

make individual levels (of a given J ,B,S,I, ... ) overlap.
We claim that the higher mass resonances are not individually de-
tectable because they overlap. Since they do not produce observable
peaks, even in individual partial waves, their existence must be
inferred more indirectly. This can be done by finding reactions
whose cross section is controlled by statistical factors associated
with the large number of overlapping resonances. Precedents for
this approach exist in nuclear physics where a similar problem of
overlapping resonances occurs.
The most familiar application of statistical ideas is to find

reaction rates given by phase space factors, so that the ratios of
different reaction rates are controlled by statistical competition.
This approach was introduced into nuclear physics by Bohr and others,
and has been pursued in hadron physics especially by Fermi 6 and
Hagedorn and Ranft 7 . However, the information this approach yields
on excited states is indirect, and in practice there is no general
agreement among particle physicists as to which reactions are
"statistical" and what has been learned about the particle spectrum
thereby. (The problems are: i) the short reaction time makes the
attainment of statistical equilibrium marginal at best, ii) even in
cases where equilibrium seems to be obtained and T is known, the
volume is not well known, i.e., the energy density is poorly deter-
mined, which frustrates attempts to deduce the level density.) Thus
more direct evidence for high mass resonances is needed. For this
we turn to a second kind of statistical argument.
II. Ericson Fluctuations in Nuclear Physics
Once more the appropriate concept originated in nuclear physics

(Ericson, 1960)1. The general idea is that where statistics apply,
there must be fluctuations.
STEVEN C. FRAUTSCHI
To pursue this point in some detail, let us consider the partial

wave amplitude for a reaction at intermediate energies as a sum over
resonances each represented by a Breit-Wigner term:
YabiYcdi
(2)
-E +E
i
- i .!:.2
We normalize IAJI to 1 at the unitarity bound. The number of over-

lapping resonances is
n ~ p(E)r (3)
where peE) and r now refer specifically to the density and width of
resonances with the quantum numbers B,JP, •.. appropriate to our
reaction (the main variation to keep in mind in the hadronic case is
EIT
peE) ~ eO). The typical partial width is
lyabi l 2 1
r
RJ - (4)
n
since in the elastic channel they must sum to about 1. In an
inelastic channel Ya b'Y d' may have either sign so the sum is smaller.
1 c 1
In general AJ may contain two terms:
AJ = AJC + OAJ (5)
AJ is the coherent contribution that occurs if the resonances tend to

b~ in phase. SAJ represents the fluctuations about the coherent
amplitude, resulting from variations in phase, partial width, and
spacing from one resonance to the next. SAJ is expected to fluctuate
in sign and have the characteristic magnitude
100JI ~.!.x rn~J:.- (6)

n rn
where lin is the strength of a typical resonance term and is the rn
result of summing over all resonances with random phases. It is
expected that SA is always present whereas AC depends on the reaction;
examples of large AC are forward elastic scattering, charge exchange
peaks, etc.
In a purely statistical reaction,
o (7)
and
1 -E/T
'---r.Je (8)
n
This ordinary Boltzmann factor is the Bohr expression which one sees
in poor resolution (here the averaging is provided by experimental
smearing over energy). But in good resolution one should see strong
fluctuations about the Boltzmann factor.
Even in a coherent reaction, the full rate IAc+ oAI 2 should show
fluctuations of relative height 2IoAI/IACI. Thus these "Ericson
fluctuations" are expected to occur in a broader variety of reactions
than do other statistical phenomena.
In all cases the fluctuations are characterized by an energy

width
LlE '" r (9)
and an angular width
1
(10)
L
max
The energy width is evident from the argument that when the energy
is increased by r, one gets a new set of overlapping resonances with
a new set of phase relationships. The angular width follows from the
usual arguments about how fast a Legendre polynomial can oscillate.
The fluctuations in E are expected to provide the most distinctive
indication for resonances, since coherent phenomena such as diffrac-
tive peaks also exhibit rapid variations in 8.
Ericson fluctuations are well established and a useful tool in

nuclear physics. Let us review a beautiful nuclear case as an
example of what we are looking for. The reaction
p + Fe 56 + p + Fe 56 (11)
has been studied 8 over the range E = 9.3 to 9.6 HeV at intervals of
2 to 5 KeV at each of several angles between 63 0 and 171 0 (Fig. 3).
At fixed energy, the dependence on angle is characteristic of a
diffraction peak with minima. It is at fixed angle that the cross
section exhibits rapid Ericson fluctuations as the energy is varied.
The fluctuations are visible at all angles, but oA/A is biggest at
the larger angles where A (i.e., the coherent diffraction term) is
small. From the observed fluctuations one learns that:
i) resonances are present in this diffractive reaction;

ii) the resonances have r '" 3 KeV;
Fig. 3 d~/dn for p + Fe 56 + p + Fe 56 measured in steps of 2 to 5 KeV

around 9.4 UeV (taken from Ref. 8).
iii) the resonances have roughly the level density expected from
nuclear theory (the level density is obtained from the data
by a correlation function method. In essence the method esti-
mates IAcI by smoothing over the data, and loAI by using our
result that the relative height of fluctuations is approximately
2IoAI/IAcI when Ac is present. Comparison with the expression
loAI ~ l/Ill then yields the number of overlapping resonances
n = rp from which p is readily deduced).
III. Some Theoretical Questions
In interpreting the peaks in a reaction such as p + Fe 56 -+ p +

Fe 56 , an ambiguity arises. The peaks can be interpreted either as
i) individual resonances, in which case
N N (12)
res peaks
or as
ii) overlapping resonances, in which case
N » N (13)
res peaks
To distinguish these possibilities, one needs further information.
For example, suppose that in addition to the elastic reaction it were
possible to study
p + Fe 56 ~ p + Fe 56
~p + Fe 56 *
~D + X55
-- - - (14)
If interpretation i) is correct, each new reaction could be inter-

preted in terms of the original set of resonances. If interpretation
ii) is correct, each new reaction would present an entirely new
pattern of peaks requiring the introduction of many new resonances.
If it were possible to study all open channels, the sum over all
resonances would turn out to be far greater than the number of peaks
in any single channel.
In nuclear physics, while the ideal of complete information is

not reached, enough is known to convince people that interpretation
ii) is right. The same question will arise in particle physics, but
here we lack the information to decide which interpretation is correct.
Another problem regarding the theory of Ericson fluctuations has

been raised by Hamer 9 and a number of other authors 10 • They point
out that when the levels get too close together, unitarity no longer
allows us to write the amplitude as a sum of resonance terms, as we
did in eq. (2).
For example, in a single elastic channel, S can be written
2iO
S e (15)
with the phase 0 real to satisfy unitarity. For a single resonance,
r )
o (E) = tan-1 ( 2 nl-E) (16)
gives the Breit-\Vigner form of S. For N resonances,
o(E) .2N tan- l( 2(M~-E)

1=1
r. )
1
(17)
is a generalization which is automatically unitary. We note that

for this choice of o(E),
N
S II S. (18)
1
i=l
is a product of Breit-Wigner terms. It approximately reduces to a

sum only when the spacing between resonances is greater than r i ,
i.e.,
r p < 1 (19)
so that all arguments of tan- 1 are small except the nearest (Mi- E)-l.
In the general many-channel case, the correct statement is that

unitarity allows the amplitudes to be written as sums of Breit-Wigner
resonance terms only when
nchannel
r p s 21f
(20)
This is most easily seen by transforming to the eigenchannels (linear

combinations of the physical channels chosen such that each resonance
couples to a single eigenchanne1). In the eigenchannels, condition
(20) reduces to the single channel relation r p < 1 applied to each
of the (nchannel) eigenchanne1s. Note that this condition does
allow overlapping resonances in each physical channel •
The relevance of condition (20) is that, in deriving Ericson

fluctuations, we assumed a sum over independent Breit-Wigner terms.
When the resonances are too close together (in violation of (20»),
unitarity does not allow them to be independent; in this case their
*Some of the papers listed in Ref. 10 analyze only the single or two-
channel case, where condition (19) is not satisfied by overlapping
resonances. As a result these papers reach unduly negative conclu-
sions concerning Ericson fluctuations, which are not justified in
the general multichannel case.
couplings cannot vary randomly. Quite different behaviors can re-

sult in this case, as Hamer illustrates with several examples.
It is interesting to note that in both nuclear physics (in the

usual Weisskopf model) and in the statistical bootstrap model for
hadron physics, condition (20) is satisfied with the equal sign. In
other words, one has the borderline case. This suggests that the
relations among flu~tuations, level density, and r will not have the
precise quantitative values discussed earlier. On the other hand
the relations do work approximately in nuclear physics, and we can
hDpe that they again work approximately in hadron physics.
IV. Ericson Fluctuations in Hadron Physics
Following the observation of fluctuations in nuclear physics,

Ericson (1964) suggested a search in hadron reactions1,11. Experi-
ments on pp-elastic scattering, the reaction which can be measured
with greatest precision, were carried out in 1966-67 and revealed a
very smooth cross section in both s12 and t 13 The most critical ex-
periment indicating the absence of fluctuations was that of Akerlof
et al. 12 , who fixed Sc.m. = 90° and varied the energy in small steps.
These experiments killed interest in the subject for some years.
In 1972 I returned to the subject 14 • By then it had become

accepted that the pp channel is "exotic" - i.e., no direct channel
resonances are found experimentally, and none is expected in the
standard quark model. I noted that one should look for Ericson
fluctuations in non-exotic channels which do have resonances, such
as nN, -KN, or -pp * .
I focused attention on
(21)
where precise data with the necessary closely-spaced energy intervals
above the separable resonance region is easiest to obtain. Data of
this type already existed for 0° (Otot) and 180° scattering up to Plab
of somewhat more than 5 GeV/c. It was shown that the familiar peaks
and dips in this data, though usually interpreted as single resonan-
ces, can also be interpreted as overlapping resonances (Ericson fluc-
tuations)14,15. The peaks are, of course, much more prominent at 180°
where AC is smaller. However, the largest fluctuations (largest oA/AC)
were expected in the region around 90°, where AC is smallest (Fig. 4).
*Even in non-exoti~ reactions, two-component duality complicates the

theoretical estimate of IOAI14. loAI is now reduced from (unitarity
bound/In) to (I~/In) where AR is the Regge exchange contribution
to the forward elastic amplitude. The vanishing of oA in pp is a
special case of this relation.
+1 cosS
Fig. 4 Schematic comparison of experimental dcr/dt for elastic

~p scattering at lab momenta of order 5 to 10 GeV/c with
the purely statistical rate <loAI2) which would apply if
the reaction proceeded solely via randomly coupled, ran-
domly spaced direct channel resonances with a spectrum
p(m) ~ exp (m/To). (Specific numbers were obtained in
Ref. 14 by use of the statistical bootstrap model.)
A preliminary indication that the region around 90 0 might exhibit

fluctuations was supplied by the Lundby group, which had a nearly
complete angular distribution for ~±p elastic scattering at a lab
momentum of approximately 5 GeV/c. By a re-analysis of their data,
they obtained dcr/dt at each of two energies separated by only AE c . TII • =
30 MeV I6 • They found suggestive though incomplete evidence for
fluctuations: the ~+p ~ ~+p data exhibited a striking shift from one
energy to the next*.
In 1974-75 an Argonne-Columbia-Uinnesota collaboration under-

took a systematic survey of
(21)
pp ~ pp (22)
around 90 0 • Specifically, they studied the kinematic region
* This shift was only partially reproduced in the subsequent Argonne

experiment l7 , which found dcr/dt varying less abruptly with energy
at most sand t.
ERICSON FLUCTUATIONS AND ARGONNE DATA ON 'll'N SCATTERING 11
-0.3 < cosS < 0.4

GeV {9.5 GeV/c 'If -p,pp
2--<p <
c lab 6 GeV/c +P
'If
at intervals of order
I1E
c.m. '" 35 UeV
2
I1t 0.1 GeV
Altogether this gave a measurement grid (Fig. 5).with some

25 different t measured at a typical s
50 different s measured at a typical t.
Results. A brief summary of the data has been published in Physical
Review Letters 17 • The authors have kindly supplied me with many
further details* in advance of full publication (these preliminary
findings are indicated by a t below).
i) The experiment originally hoped to check the djmensional counting

rule predictions 18 as well as fluctuations. As you can see from
Fig. 6, this hope was achieved in pp scattering, where no significant
fluctuations were observed and
def
dt Ifixed cosS
s-n (23)
cose = +1
Previous
Experiment ---+ s
cos e = -1
Fig. 5 Kinematic regions where systematic studies of elastic 'lfN

scattering at closely-spaced intervals in sexist.
*I am especially indebted to Larry Price for many discussions of

the data.
I>.)
PLAB (G eV/c) PLAB (G eV Ic) P LAB (GeV/c)
5 23456789 23456789 234567895
, I I • ,
{. • ~ + i • • i 10
10 .\ ....
'.. do- 71"-P-71"- '. 71" P-71" P (b) pp-pp eCl
'_.. di xIOO P (0) ". Qg-x 100
104 ··.dl 104
• 0
.... 8c .,,=100· .... 8c .,,=100 8c ,l90·
3
\t.:.· 103
10, ....,. ······1\
.'. ........... , ...
fB ........ S-8
NU
102 "!.
. ~
'.
do-
: 0
'\
'\\ .......
d o-x 0.01
dl ..
102
- '"t>
;;;- £P"xO.OI \
dl dt.:'0.OI ... t
.,
.0 >10 "-""
10.0 I;;;->
-I
::i..QJ
::i..'" l!)
"'-.",
l!) ".
..\·...6 CM
. . .-.. ••.•.••'ot~·
~ •• 8 =90. \ ....
.'....,
o .~
-i;1.; -I •••• • + 8c,,=80 '. -, -i;1.;
10 10
··.l8 CM=80
-2 -2
10 '. 10
t".."'.
• R e f.3 a Ref.3
103 • R e fA • R e f.5
163
• Ref, 5 • Ref. 7
• R ef.6 6 R ef.8
104 IC/
4 5 6 7 8 910 12 4 5 6 7 8 910 12 14 16 18 4 5 6 7 8 9 10 12 14 16 18 ~
s (GeV2) ( GeV2) s (GeV 2) m
<
m
Z
Fig. 6. Differential cross sections for elastic ~-p, ~+p, and pp
r>
scattering measured in the Argonne experiment 17 for some ."
::tt
representative angles near 90°. The lines are the s de- »
pendence expected on the basis of dimensional counting C
-I
rUles18 . Data from earlier experiments (Refs. 3-8 of the Ul
(")
Argonne paper) on ~-p and ~+p above 3 GeV/c are also plotted. :z::
works well at the larger s with n ~ 10 as predicted by the rule.

However in np scattering the cross section was too bumpy to extract
a meaningful s-n behavior. The bumpiness appears to have two sources:
the fixed dip at t = -2.8 GeV 2 (which is crossed by the fixed
Sc.m. = 90° curve in n-p + n-p, for instance, at around s = 8 GeV 2
causing the big drop in the figure at that point), and apparent
resonant structure at various values of s.
ii) Before studying fluctuations, one should separate out fixed

diffractive features. In the large t, large u region covered by this
experiment one finds no fixed -u structures, and only the well-known
t ~ -2.8 GeV 2 dip at fixed t.
iii) To avoid the t = -2.8 GeV 2 dip, it is best to follow fixed t

as a function of s. Doing so, one finds localized structures -
peaks, dips, and shoulders. In Fig. 7 we exhibit, by way of illus-
tration, the n+p data at several values of t.
iv)t The total number of clearcut candidates for structures (n-p plus
n+p) is about 20. Only one or two of the n+p structures are dupli-
cated in n-p. The typical width of the structures is r ~ 100 to
200 HeV.
v)t Are the structures known resonances? Few of the structures are
likely to be known resonances from the Particle Data Tables. They
appear throughout the experimental range 2.3 GeV < Ec . m• < 4.3 GeV,
whereas previously established nN resonances are limited to Ec • m.
S 3.2 GeV. The structures with Ec .m. < 3.2 GeV are generally
narrower than known resonances and are not centered on them.
vit) Do the structures represent individual resonances or over-

lapping resonances? I assume, first of all, that the structures are
resonance-related. Of course we don't know this, but the fact that
the structures are seen in np but not pp, and the width r ~ 100 to
200 HeV, tend to indicate resonarLt behavior.
It is a remarkable fact that the typical structure has b.t ~ 1

GeV; i.e., the structures appear as spots on an s-t plot rather than
stripes running across the pLot at particular values of s. This
implies that at each energy, several J (I shall take, somewhat arbi-
trarily, 3) cooperate to make the structure.
As usual we can entertain either the individual resonance or

overlapping resonance interpretation. In the former case, taking
into account the implications of b. ~ 1 GeV 2 , we have on the order
of 3 X 20 ~ 60 new ,[ncUv.'[dua.l fLe60nanc.e6. In the latter case, with
the usual estimates, we have on the order of 10 5 ovef!.tapp,[ng fLe60n-
anc.e6 .
CENTER OF MASS ENERGY (GeV)

2 2.5 3.0 3.5 3.0 3.5 4.0 2.0 2.5 3.0 3.0 3.5
3
10
-103.1
)(10 2
10
2
"~.~<
",5
"'u 10
'""
.>
(!)
" I
\
.0
:t.
5.0
)( 2
'0
"b
'0
-I
10
-2
10
5 10 7 10 13
Fig. 7 Differential cross sections for elastic n-p and n+p measured
at constant t in the Argonne experiment 17 • The top number by
each set of data gives -t in GeV 2 /c 2 • The bottom number, if·
present, is a scale factor by which do/dt has been multiplied
for presentation. The lines are purely to guide the eye; they
are not fits.
vii)t Correlation function test for fluctuations. This has been

carried out and shows that the fluctuations are well above the level
one would expect just on the basis of statistical errors. The rela-
tive amplitude of the fluctuations does not vary much with energy.
This indicates a behavior like Fig. 8a whereas I expected a behavior
like Fig. 8b (fluctuations decreasing by an order of magnitude over
the energy range surveyed). Thus apparently p is increasing more
slowly than exp(IS/160 UeV). Of course this does not rule out
Ericson fluctuations; it does suggest that our original estimate of
the number of participating levels was deficient in some respect.
dol
dt
cose
s
(a)
dol
dt
cose
s
(b)
2
Fig. 8 Schematic comparison of IAcI (the smoothed-out experimental
do/dt) and the purely statistical contribution to do/dt for
nN elastic scattering at a fixed angle near 90 0 :
a) as indicated by fluctuations in Argonne data 17 ;
b) as predicted on basis of statistical bootstrap model 14 •
To summarize, this experiment suggests numerous new resonances.

Whether one favors the individual resonance or the overlapping
resonance interpretation, in either case one concludes that the
nucleon excited states continue to high mass with an increasing
density.
v. Future Possibilities
One obvious line of development for this subject would be exten-

sions of the Argonne experiment. For example, it would be desirable
to
i) Extend the detailed study of da(TIp + TIp)/dt over a broader range

of sand t. The Argonne experiment suggests the rule of thumb that
the number of fluctuation structures seen will be proportional to the
area ~IS x ~t surveyed (provided that at each s, the ordinary forward
and backward peaks, where fluctuations are less visible, are excluded).
ii) Hake a comparably detailed study of the polarization in TIN + TIN.

This would allow the two TIN helicity amplitudes to be disentangled,
thus doubling the number of independent fluctuation structures seen
if the structures are indeed of the Ericson type.
iii) Study other reactions in the same spirit. For example, K-p and
~p would be interesting to compare because the latter is an exotic
channel.
Another reaction with unique features is
(24)
It is especially simple to interpret because only one spin-parity

state, JP = 1-, contributes. That spin state (along with JP = 0-)
happens to be one for which the spectrum begins at especially low
masses; thus the density of states is higher than usual and the
resonances may begin to overlap at a lower mass than usual. Because
of these special features, the e+e- reaction is the one where chances
are best for comparing a whole array of final states and seeing
whether the peaks in all final states can be explained by a small
number of resonances, or whether each new final state studied will
yield new peaks. Hy expectation is that above about 2 GeV, the
resonances will be seen to overlap by direct count [r times the
(number of observed peaks in an interval ~E) will exceed ~E] and one
will be compelled to interpret at least the broader I = 1 peaks as
cooperative phenomena involving more than one resonance. Comparisons
of the experimental structure in da/dt with statistical estimates of
the "noise level" daF/dt for each final state (TITI, TIp, .... ), along the
lines we have conducted for TIN + TIN, will then give a clearer picture
of p(m) and the degree of randomness in the cross sections.
References
1. T. Ericson and T. Hayer-Kuckuk, Ann. Rev. Nuc1. Sci. ..:!i, 183

(1966).
2. R. Hagedorn, Nuovo Cimento Suppl. 3, 147 (1965).
3. "Dual Theory," H. Jacob editor, North-Holland Pub. Co. (1974).
4. A. Chodos, R. L. Jaffe, K. Johnson, C. B. Thorn, and
V. F. Weisskopf, Phys. Rev. D9, 3471 (1974).
5. S. Frautschi, Nucl. Phys. B9~ 125 (1975).
6. E. Fermi, Progr. Theor. Phys. (Kyoto) 2, 570 (1950).
7. R. Hagedorn and J. Ranft, Nuovo Cimento Suppl. ~, 169 (1968).
8. J. Ernst, P. von Brentano, and T. ~ffiyer-Kuckuk, Phys. Letters 19,

41 (1965).
9. C. Hamer, Nuc1. Phys. BIOS, 153 (1976).
10. P. A. Ho1dauer, Phys. Rev. 157,907 (1967), and 171, 1164 (1968);
v. L. Lyuboshitz and M. I. Podgoretsky, Yad. Fiz:-24, 214 (1976)
[Sov. Journ. Nuc1. Phys. 24, 110 (1976)]. --
11. T. Ericson, "Fluctuating Phenomena and Statistical Reactions,"
CERN TH-406 (1964) (unpublished).
12. C. W. Aker1of, R. H. Hieber, A. D. Krisch, K. W. Edwards,
L. G. Ratner, and K. Ruddick, Phys. Rev. 159, 1138 (1967).
13. J. V. A11aby, G. Be11ettini, G. Cocconi, A. N. Diddens,
M. L. Good, G. Matthiae, E. J. Sacharidis, A. Silverman, and
A. M. Wetherell, Phys. Letters 12, 389 (1966).
14. S. Frautschi, Nuovo Cimento 12A, 133 (1972).
15. P. J. Carlson, Phys. Letters B45, 161 (1973).
16. F. H. Schmidt, C. Bag1in, P. J. Carlson, A. Eide, V. Gracco,
E. Johansson, and A. Lundby, Phys. Letters B45, 157 (1973).
17. K. A. Jenkins, L. E. Price, R. K1em, R. J. Miller, P. Schreiner,
H. Courant, Y. 1. Uakdisi, ~1. L. Marshak, E. A. Peterson, and
K. Ruddick, Phys. Rev. Lett. 40, 425 (1978), and 40, 429 (1978).
18. S. J. Brodsky and G. R. Farra~ Phys. Rev. Lett. 31 1153 (1973);
V. A. Uatveev, R. U, Muradyan, and A. N. Tavkhe1idze, Lett.
Nuovo Cimento 1, 719 (1973).
ASYMPTOTIC FREE DOll AND COLOR SUPERCONDUCTIVITY IN
DENSE QUARK MATTER
Steven C. Frautschi
California Institute of Technology

Pasadena, California 91125
In the present talk I shall deal with densities much higher

than nuclear - densities which, as far as we know, were achieved
only in the very early stages of the big bang. I shall review two
issues:
i) The current view that QCD predicts an approximately free quark-

gluon gas phase at high densities and temperatures, which is incomr
patible with the kind of non-perturbative confinement effects that
would be needed to yield a limiting Hagedorn temperature at high
densities.
ii) The recent suggestion of my student, Bertrand Barrois 1, that QCD

leads to a different kind of non-perturbative phase, analogous to
superconductivity, at high densities and low temperatures.
I. QCD at High Density and High Temperature
Are quarks and gluons compatible with a limiting temperature?
Of course it has always been clear that free quarks and gluons,
at energies so high that their masses could be ignored, would give
(E) ~ T4
just like a photon gas. Thus the temperature would rise as (E)1/4
and there would be no limiting temperature.
However until recently it seemed not excluded that the attrac-

tive interactions among quarks and gluons might create so many bound
19
states that the limiting temperature could emerge from quark-gluon

dynamics.
This hope was greatly diminished in 1975 when Collins and

Perry2, and Kislinger and t.10rley 3, provided strong arguments from
QCD that quarks and gluons approach a free gas state at high den-
sities and temperatures. Thus it seems that indeed (for zero or
relatively small barron excess, such as seems to be the case cos-
mologically) T ~ (E) /4 and rises without limit.
The argument of these authors can be stated roughly as follows.

Consider an assembly of N quarks and gluons at an average energy
such that rest mass can be neglected. If, for simplicity, we con-
sider only two-body reactions, the energy is
N
E .-_Ll (Ekin) i +
1
L Vij
ij
(1)
The first term is of order N whereas the second term is potentially

as large as O(N 2 ).
To argue that the kinetic energy nevertheless dominates,

Collins and Perry and Kislinger and l10rley introduced two physical
ideas:
i) Screening. Electric charge in a plasma is screened by neigh-

boring charges. Its potential is reduced from
V = e 2 /r (2)
ei
to
V (3)
ei r
where
mc -1
.-if ~ (rscreen) (4)
i.e., the Coulomb photon acquires an effective mass in the medium*

Now the potential energy ~ V.. of the i th charge is effectively
reduced from a sum over J 1J all N particles to a sum over the
neighbors inside the screening radius. As the density increases,
the screening radius falls so the number of neighbors interacted
with does not change greatly.
*\Vhat is calculated in lowest order is the ordinary vacuum polari-
zation diagram ,~. There is a mass term because the
intermediate states are modified by the presence of the fermi sea
of electrons.
DENSE QUARK MATTER 21
In QCD Collins and Perry (working at large baryon number NB

with T = 0, N'B = 0) and Kislinger and Horley (working initially
at large T with NB - Us = 0) found the same behavior for the
"electric" components of the gluons. The gluon mass grows as
where R is the interparticle spacing. In other words, the screening

radius grows as
r
screen
'" R·
V§gC
g22
(6)
Comparing
'" p.1 c "'1ic/R (7)
to
2
(V e",n).ln '" g /R (8)
where n is the nearest neighbor, we obtain
2
. ).1 '" g /hc
(V e",n).ln /(E k ln (9)
2
ii) Asymptotic Freedom. As R becomes small we need to know g only
in the small r reglon where it is reliably calculated:
:l... 1 (10)
-tic '" RIn. (r /R)
1f
Thus
o (11)
To complete the argument for V 1/Ek. 7 0 we must, of course,

sum over the neighbo~ ly~ng within e ln the screening radius.
Since r . '" R c/g, the number of neighbors which are
screenlng
strongly interacted which grows slowly from the value of about 1
it had at R ~ r1f' If all neighbors interacted with the same sign,
we would have crudely
(Ve~)i ~ g l
2 Rkc/g 2 3
d r p/r 4'1rg
2 Rltrc/g 2
J
o
dr rp
2'1rg 2 PR2 (1'ic/ l) (12)
where p is the number density. Putting in p 3/4'1rR3 we have
(V.).
e", ~
,... 31rc/2R (13)
so
(Ve",.)./(E
~
kin ).~ ,... 0(1) (14)
However this is assuming all neighbors interact with the same sign
(same charge or color state). This requires a highly ordered state.
Such a degree of ordering cannot be achieved at high T, where in
fact there are cancellations among Vij' which increase as the
effective number of neighbors rises. Thus Vet/Ekin indeed fails as
the density rises at high T. We shall come back to the possibility
of order at low T in the second part of our talk.
iii) Magnetic Effects. Since the medium provides a preferred frame,

the time-time, time-space, and space-space components of the gluon
propagator are on a different basis and need not exhibit the same
mass. In fact one finds in perturbation theory that
222
(me) gluon ,... ~ (p ) (15)
as above for electric terms but not magnetic terms. (A familiar
example: in electrodynamics, a magnetic mass implies the Meissner
effect. This is not obtained until one considers non-perturbative
effects. On the other hand, charge shielding is obtained in per-
turbation theory.) Thus we must consider the unshielded magnetic
potential separately. Fortunately the leading term at long range,
the dipole term, has the form
2
-+ -+ L
(Vmag) ij ,... g C\ . Bj"'" 3 (16)
r
which falls much faster with r than Ve~ did. Here, even i f all
pairs interacted with the same sign, we would obtain only a loga-
rithmically divergent contribution from the sum over distant
magnetic sources. In practice at high T many particles have the
opposite spin (or charge, or color) ana this logarithmic divergence
is cut off leaving a small though non-negligible term.
Discussion
i) The foregoing argcment contains some loopholes. For example,

it depends on asymptotic freedom. If there are more than l6~ quark
flavors, asymptotic freedom will be lost. Similarly a sufficient
number of fundamental scalars would destroy asymptotic freedom.
ii) Nothing we have said argues against the possibility that the
hadron mass spectrum has the form
p(m) ~ exp(m/T O) (17)
In fact the HIT bag model, which is hoped to be compatible with

QCD, yields this spectrum. In this model the more massive states
have larger volumes:
Volume ex: m (18)
Thus they do not have higher energy density than a low-mass hadron.
lihat happens in dense matter, then, is presumably that as the average
density increases past the density characteristic of an MIT hadron,
the MIT hadrons cease to exist as separate states and a phase
transition from confined quark bags to free quarks occurs, as dis-
cussed by Cabibbo and Parisi 4 .
The type of hadron that appears to be contradicted by the

asymptotic freedom arguments is the original Hagedorn hadron with
radius fixed at about 10- 13 cm. This type of hadron has a very high
internal energy density which, if the asymptotic freedom argument is
right, would not support the non-perturbative binding needed to make
hadrons in the first place.
II. QCD at High Density and Low Temperature
Recently Barrois has examined the consequences of QCD from a

somewhat different point of view. Horking at a given net baryon
number density excess
(19)
he has the temperature as a variable. At low temperature he finds

that qq pairs form, analogous to the electron pairs in superconduc-
tivity. This is a non-perturbative binding effect with an energy
gap between the low-lying ordered states and the continuum of dis-
ordered quark-gluon states. It did not occur in the previous argu-
ment of Collins and Perry because their argument was perturbative
(Collins and Perry did remark as a topic for further study that
superconductivity might occur).
At higher temperature Barrois finds a phase transition to the

disordered state, as in ordinary superconductivity. The disordered
state is the nearly-free quark-gluon soup of Collins and Perry, and
Kislinger and Morley. The phase diagram is shown in Fig. 1.
With regard to this diagram, we first make some comments on the

theoretical side. The reliability of both the perturbative and non-
perturbative calculations is greatest when the effective g2 is small-
i.e., at very close particle spacing, which is achieved at large ~NB
and/or large T3. Thus the phase assignments are hoped to be accurate
far to the right or upwards on the plot, but they become less reliable
as one approaches the origin, where we are presently incapable of pre-
dicting the confined-quark phase and must take it from experiment.
In view of the theoretical uncertainties in this region, there could
be still more phases, and of course the detailed location of the
confined-quark phase boundary is purely conjectural.
On the observational side, I have plotted on the phase diagram

the trajectory the Big Bang is believed to have followed. At each
baryon density the Big Bang temperature is far above the critical
temperature Tc (~NB) where su~erconductivity would hold. To put it
another way, the present cosm1C blackbody temperature would have to
be much lower than 3°K to have placed the temperature of the Big Bang
in the superconducting region. This would imply a universe dominated
by baryon excess from very early times.
T
/
c§'~
~
nJtee q';.,
''''I~
.Q;?
qq pcUJrA
,
\
NeUtron
Star\!
Fig. 1. Phase diagram for quark-gluon matter.

It is possible that the centers of neutron stars are in a

superconducting quark phase, but the effective gZ is still rather
large in this region and it is not possible to make ~eliable esti-
mates at present.
What would the qq condensate phase be like? As in ordinary

superconductivity, q's from near the Fermi surface would form Cooper
pairs of zero total momentum. These would have a spatial extent
related to the enerf~ gap, rather than the average quark separation
distance or the 10- cm. radius of ordinary hadrons. There would
also be collective excitations - a spectrum of translationally,
rotationally, and vibrationally excited states with similar spatial
extents.
The bound Cooper pairs would not be ordinary hadrons. This is

apparent from their spatial extent but, more importantly, from the
fact that a qq pair cannot be a color singlet. We are accustomed
to the idea that ordered low-temperature states break some symmetry
of the original Lagrangian, and this one is no exception. The qq
pairs would pick out a definite direction in color space, breaking
the original SU3 of color down to an SUZ subgroup (just as H' =msss
breaks flavor - SU 3 down to SUZ). For example it might be red-blue
quark pairs which participate in the ordering while the red-white
pairs remain disordered, unable to carry a lossless color current.
As far as the gluons are concerned, we recall that they had

already acquired a perturbative "electric" mass in a dense medium,
but that the perturbative estimates left the space-space ("magnetic")
components of the gluon propagator massless. The electric mass,
being a perturbative QCD effect, was of course a color singlet. In
the superconducting state, the gluon also acquires a non-perturbative
"magnetic" mass. This non-perturbative mass is not a color singlet-
five of the gluons become heavy while the three gluons associated
with the remaining SUZ color symmetry stay massless.
The techniques used by Barrois to reach these conclusions are

a relativistic generalization of those employed in ordinary super-
conductivity. One guesses at a symmetry-forbidden Green's function
(Fig. Za) and requires that (i) it be a consistent solution of the
Dyson equations (ii) it must lead to a lower ground state energy
than the ordinary symmetric solution. (e.g. Figs. Z b,c.) Any
consistent symmetry-forbidden solution will be non-perturbative
because perturbation theory cannot break the symmetry of the original
Lagrangian. Of course the original Lagrangian still plays an
essential role; in particular it must supply an attraction if we are
to obtain bound states such as Cooper pairs. In Barrois' calculation
it is the one-gluon exchange which supplies an attraction in certain
states*. To be confident that this attraction is not overridden by
(a)
e (b) (c)
Fig. 2. (a) Symbolic representation of a symmetry-forbidden Green's

function for Ivac) + Iqq); (b) vacuum bubble diagram for
lowest-order color-symmetric contribution to the ground
state energy (the propagators here and in Fig. 2c are modi-
fied by the Fermi sea of quarks); (c) vacuum bubble diagram
for contribution of symmetry-forbidden Green's functions to
the ground state energy.
two-gluon effects, etc., one must have small effective g2 - i.e.,

high energy densities.
The limitations of these techniques include not only the re-

striction to small effective g2, which prevents an attack on the
lower-density regime we inhabit, but also the non-systematic nature
of the initial guesses. While Barrois has indeed found that the qq
condensate has a lower free energy than unpaired quarks at low
temperatures, there may conceivably be some other condensate he has
not examined which has an even lower free energy** Another problem
is the lack of an "effective Lagrangian" for the interaction of the
qq pairs, which would permit estimates of color domain boundaries,
possible color vortices, and the like.
Even in its present preliminary stage, the study of ultrahigh

densities is useful for broadening our viewpoint on QCD. At
accelerator energies, QCD is seen as a theory which must somehow
* The simplest qq state with an attractive interaction is color-odd

(3), spatially even (S-wave), spin odd (S = 0), and flavor odd (1=0).
Thus the Cooper pairs are spinless and flavorless though they do
carry color. The collective excitations will have more complicated
quantum numbers. As pointed out earlier, a systematic organization
of pairs into attractive configurations is just sufficient to make
the potential energy competitive with the kinetic energy, and that
is what happens in the qq condensate.
**1n the high density limit he provides a dimensional argument 1 favor-
ing qq pairing over all more complicated correlations. The uncer-
tainty therefore arises mainly at intermediate and lower densities.
lead to colorless clumps, suppressing all long-range manifestations

of color. At ultra-high densities, while our wonder at never seeing
isolated quarks remains, the perspective enlarges: QCD seems to
predict several phases, colorless clumps are no longer useful degrees
of freedom, long-range color correlations are possible. Color con-
finement in hadrons is not the sole outcome of QCD (assuming it
indeed gives color confinement), but is rather a property of the
particular phase we happen to live in.
References
1. B. C. Barrois, "Non-Perturbative Effects in Dense Quark Hatter,"

to be published in Nucl. Phys. (1979). A more detailed
account is contained in Barrois' Caltech Ph.D. thesis
(unpublished; 1978).
2. J. C. Collins and M. J. Perry, Phys. Rev. Lett. 34, 1353 (1975).
3. M. Kislinger and P. D. Morley, Phys. Rev. D13, 2765 (1976).
4. N. Cabibbo and G. Parisi, Phys. Letters 59B, 67 (1975).
DUALITY TRANSFORMATION AND CONFINEMENT
IN GAUGE THEORIES
F. Gliozzi
Istituto Nazionale di Fisica Nucleare
Sezione di Torino
I. INTRODUCTION
Quantum Chromodynamics is widely considered as one of the

most promising candidates for a fundamental theory of the
hadronic world. Indeed the short distance behaviour of this
theory has been successfully extracted with the renormalized
perturbation techniques and even partly tested with experiment.
However the long distance structure of the theory poses a num-
ber of yet unsolved questions, owing to the non-perturbative
nature of the problem.
In particular, the seducing conjecture that the quarks are per-
manently trapped into colourless bound states (quark confinement)
has not yet been proved (or disproved).
Nevertheless in the past few years a considerable progress has
been made in the intuitive understanding of the long distance
features using the analogy between euclidean quantum field theory
and statistical mechanics in which the coupling constant acts
as the temperature parameter 1 ,2,3.
In many statistical systems, like ferromagnets, liquid
helium or superconductors, there is a phase in which confining
forces arise quite naturally.
For instance in the superconductors of type II the magnetic
field is concentrated in quantized flux tubes (Abrikosov vorti-
ces); then a pair of magnetic monopoles embedded in the
superconductor 4 ,5,6,7 is connected by an Abrikosov vortex which
29
30 F. GLiOZZI
conveys the whole magnetic flux of one monopole to its partner.

It follows that their potential is asymptotically linear, thus
they are permanently confined.
There are some important differences between these
statistical systems and gauge theories. First, the confined
objects are collective or topological excitations rather than
electric or colour charges.
Moreover the confining phase has a highly ordered structure,
while in gauge theories the high coupling regime of quark
confinement is associated with a maximal disorder of the vacuum
state.
It has been however realized that many dynamical systems, both
in field theory and in statistical mechanics, share a discrete
symmetry which exchanges one into the other such opposite
properties. A symmetry of this kind is called a duality trans-
formation: it is an involutory transformation which maps a
theory A given in a region with a small coupling constant (or
low temperature) into another theory B in a region with a large
coupling (or high temperature); similarly the local excitations
of A are mapped into the topological excitations of B and the
order parameter of A is transformed into a disorder parameter
of B.
If A=B the theory is said to be self-dual.
The first exemple of duality transformation and of a self-dual
theory in statistical mechanics was discovered in 1941 by
Kramers and 1;lannier 8 for the two-dimensional Ising model. They
showed that the partition function Z (~) of this model satisfies
the following remarkable identity
(1. 1) (t OI')~ B 'X~h B) - NZ (g )

where N is the number of points of the two-dimensional lattice
and
(1. 2) B
Thus if one knows the partition function in the low temperature
region ( f\ large) one can also describe the thermodynamic
properties at high temperature (B small).
In field theory the first example of duality transformation
was found even earlier by Dirac 9 ,10,11 in his theory of magnetic
DUALITY TRANSFORMATION AND CONFINEMENT IN GAUGE THEORIES 31
monopole (1931). Here the small and the large coupling constants
are the electric and the magnetic charges e and g, which are
related by the celebrated Dirac quantization condition
(1. 3) 'YI :: 0, :!: ", -

+ Z. )
Eq.(I.3) replaces eq. (1.2) of the Ising model.

Hore recently, a beautiful example of duality transformation
has been discovered by Coleman 12 and Handelstam 13 between the
two-dimensional Sine-Gordon model and the Thirring model.
Indeed it has been shown that the solitons of the Sine-
-Gordon equation, which are the topological excitations of such
theory, behave like the local field excitations of the Thirring
model and viceversa.
The coupling constants '" and g of these two models are related
by
(1.4 )
which again expresses, like eq.s (1.2) and (1.3), the duality
between large and small couplings.
If one could find a duality transformation in the models
where there is a confining phase for the topological excitations,
the dually transformed phase should describe the confinement
produced on local excitations (say, electric charges or quarks)
by a highly ordered structure of topological excitations, like
a Bose condensate.
In gauge field theories there are many types of topological
excitations, like the polyakov's2 instantons, ~he de Alfaro,
Fubini and Furlan's14 merons, the Handelstam's1) gas of monopoles
or the 't Hooft 16 fluxons. These authors argue that the
quark confinement is due to a sort of condensation of these
topological excitations even if it is not yet clear which one
of the topological excitations discovered up to now, if any,
actually contributes to confinement mechanism. Horeover an
explicit duality transformation is still laking.
The solution of these problems seems to be a formidable
task.
We limit ourselves in this lecture to describe some general
property of the duality transformations for a wide class of
32 F. GLiOZZI
dynamical systems.
17-20
We shall see that the general features of the duality
transformation, and in particular the complementarity between
condensation and confinement, do not depend very much on the
dynamical details of the systems but rather on the space time
structure of the topological and local excitations (part II
and III) \vhile the duality relation between coupling constants
depends on the underlying symmetry of the model. In particular
the Jirac quantization condition of electric and magnetic
charges is the general relationship between dual coupling con-
stants associated with U(1) symmetric models 20 • Similar relations
will be found in part IV for abelian gauge theories on a lattice.
II. HOMOLOGY AND TOPOLOGICAL EXCITATIONS
The natural language in which one can express the topological

properties of the Maxwell equations with magnetic monopoles and
their generalizations is that of the exterior forms 21 •
The algebra of p-forms is basically the one of antisymmetric
tensors of rank P.By definition a p-form is fp
(II. 1) f ': rpr·f" "'rf J.1C '" tA ~ ", .. d"'r
rft.}'of." r, P r.}"1. f
where , is an antisymmetric tensor with p indices and
"/'-;,c,,,..,,, ,L "~,,,. .• J,x.",,. may be identified with the oriented
volume element of a p-dimensional manifold.
For our purposes we need to define only two operations on the
forms, namely the dual • f of a form and the exterior
•. ..J f P
derlvatlve "" f •
One has, by definition
(11.2)
where d is the dimension of the ambient space and N is a

normalizing constant such that
(d..-p)r" rl-s
(II.3) :f: f- fp =(-,) T Jp
where s is the segnature of the space-time. Clearly ~ fp is
a (d-p)- form built up with a linear combination of the

components of the f, form.
The exterior derivative d. of fp f,
is given by
(11.4) --
Clearly cl f., is a (p+,). form. I t is immediate to verify the

identity:
(II.5) cL t Fp = 0
The O-form are simply the functions. For instance in d=3
case we can verify the following translation table
c:L fo grad I
r
~
cih ~ curl
d,
,f ........
t div r
r,t : ......-.
0
{ curl grad
div curl
= 0
In the language of forms the Gauss theorem can be written in a

particularly simple way
(11.6)
v Ptl ') V f+1
where Vf+1 is any (p+1)-dimensional manifold and Q

r:\ VP+J
indicates its boundary.
The Gauss theorem establishes a close relationship between
the language of closed forms (cohomology) and that of manifolds
without boundary. We have the following list of definitions in
both formalisms
Cohomology formalism Homology formalism

closed form f,: p-cycle=manifold without
tA. fp : 0 boundary: ~ ,f =
0
34 F. GLiOZZI
exact form fp: p-boundary B

f' .
: p-cycle wh~ch
is the border of a M ,~,
~p:' d, ~p_, manifold B' =? N p., .
~f"p'~O (because d,t.:o) dB P=0 (because ~l.:o)
{{!> f' 1 c { fp 1 ~ Bp 1 c {c'1
Cohomology space (group): Homology space (group):
Hp = ~Fpl/{for1 HP = f C PJ II BP}
In the Eucledean space In the Eucledean space
H is trivial: H = 11 HP is trivial: HP = U
p p
We are interested in theories of the form
(I1.7)
(n.s)
for p=2 and d=4 they are just the ordinary Maxwell equations,
but one has also other interesting examples. For instance d=3
p=1 describe,s the pseudo-particle gas of polyakov 2 , d=4 p=3
is the Kalb-Ramond theory of interacting strings 20 , etc.
If fp is defined in the whole (pseudo)-euclidean space,
Hp is trivial, i.e.
(I1.9)
where Q,'_I is the form associated to the potential.

We have the following generalized gauge invariance of the theory:
(I1.10)
If eq. (7) only holds in a submanifold of the space-time and
(n.11) f fp
cP
14p (~ijP) is no longer trivial: there are topological
excitations in the theory (e.g. monopoles). Indeed we derive
from Gauss theorem that cP cannot be a boundary
(rr.12) c p -:/= '0 V P+I

This is only possible if there is a region T 9 in which the
generalized homogeneous Haxwell equations (rr. 7) ar~ no longer
valid. Ff
is defined on the restricted manifold t1 = Tf etl._
In view of eq. (11.12) c must be linked to 7
P 1 . Two manifolds
of dimensions p and q might be linked only if
(rr.13)
q represents the space-time dimension of the manifold spanned

by a topological excitation coupled to the generalized Maxwell
theory. \ole call it the classical dimension of the topological
source. Likewise the classical dimension of the source J' J.
( :: minimal region in which . J
)
J
::t 0 is
",·ptl
-ptl
(11.14) Q -=. p-i

For instance in the ordinary Maxwell equations we have , : Q=i
indeed the electric and the magnetic sources describe a (world)
line in the space-time. In general one has, of course
(11.15) cl - 2
In the following table we give some examples of topological
excitations as well as the corresponding local or elementary
excitations
Theory Topological excitation Local excitation
f
Dirac- 't Hooft magnetic
q=1 electric charge 0=1
d.~4 Polyakov monopole monopole
Kalb-Ramond
string instanton q=O string 0=2
superconductors Abriko~ov vortex q=1 electric charge

{ or Cooper pair
magnet~c monopoleq=O electric current
superfluids Helium vortex q=1 Helium source 0=0
polyakov Euclidean instanton q=O electric
Elettrodynamics 0=1
current
36 F, GLiOZZI
The central point is now that the generalized Maxwell equations

are invariant under the dual transformation
(II.16)
provided that we exchange the topological excitations with the

local excitations.
Thus the duality transformation for this kind of theory is
rather trivial, at least at the classical level. We ,dsh to
show now that the quantization poses some strong restriction on
the consistency of this description.
In order to quantize we have to introduce the potential
"'p_. such that f: ti Q.po. which is impossible if J.I p -I II
A simple '-lay to trfvialize H w~tJ:0ut eliminating Tq is to
restrict f, to a submanifol~ H of 11 cl such that
(II.17)
where D
1iol is an arbitrary manifold whose boundary is the
classical support Tq of the topological source, i.e.
(II.18)
The manifold Dq+ 1 is the generalization of the Dirac string of

the ordinary Maxwell equation with magnetic monopoles. Note
that the cycles cP of Md that are linked to Tq have no., a
, "
non-zero ~ntersect~on , 'f 1
w~th ahe D~rac man~ 0 d D
q+1 ; then they
are no longer subsets of H • It follows that all the p-cycles
of Md are boundaries,then
(II.19) fP = c(, Cl. p_. ,....

Clearly a has a singularity on Dq+ 1 (otherwise H would be
1 d
trivial iK- M ) which has no physical meaning, thusPdoes not
have measurable consequences.
This requirement poses some constraint in the quantum version
of the theory, since here one measures something more than the
field strength f 23. Indeed in this case the measured quantity
is the phase
p
( ,-.
&e.J Q.
(II. 20) e V'-·
which is a typical factor of the functional integral, where

Vp...
is a possible history of the classical source Jd.-
P+ I and
the "electric" charge is defined by
(II.21)
Because of the linking condition (11.13) the difference between

tyro possible histories V p'" - V'r.' ::. C p•• might be a cycle
linked to the Dirac manifold Dq.
Thus the condition of no physical effect of the Dirac singularity
is simply
(II.22)
which is the generalization of the condition of no measurable

effect produced in the surrounding space by a line of magnetic
flux in the Bohm-Aharonov experiment of Electrodynamics 11 •
From the Gauss theorem we deduce at once
(II.23)
p-1
where V f is a manifold bounded by a small cycle Clinked
to Dq.
Eq. (11.23) is the generalization of the Dirac quantization
condition for the theories defined by Eq. (11.7,8).
Eq. (II.23) could be considered as a consistency condition
on the coupling constant e d~e to the presence of a topological
excitation. Another important effect is produced when there are
many topological excitations in the system. He shall see that
if their number is sufficiently large they produce the
confinement of the elementary excitations (electric charges).
The best parameter to study the confinement of the "electric"
charges is the Hilson factor given in eq. (II.20). The signal
of confinement is that the expectation value of the Wilson
exponent is proportional to the volume 'tYP of a manifold which
has V,-I as boundary. This is the so called area law for the
Wilson integral.
38 F. GLIOZZI
III. CONFINEMENT IN DILUTE GAS APPROXIMATION
We present in the following a simple argument showing

directly the "area law" for the Wilson integral associated with
a p-dimensional source immerged in a gas of q-dimensional dual
or magnetic charges
First, note that the Dirac quantization condition allows
us to use the Gauss theorem even in presence of dual charges:
i eo J 0. p i t J tA ,," • ~V ,.,
(III.1) e "'='iV'·':. e"'·' ::: e
where ~~,., is the generalized "flux" through any manifold
Vr·'
wi th boundary 'If' . Notice that cPt'"
is an addi ti ve
functional of the magnetic charges:
N
(III. 2) ~ - ~ It-.
Vrt,- • ~C,
where C; label the whole set of coordinates specifying the
position and the form of the ~.L q-dimensional charge. Then
the expectation value of the Wilson factor can be written in
the following form when e is very large 20
(III.3)
where P
w( '. I •• 'AI) is the probability of the system of N
charges in the configuration specified by the coordinates
C1 , C2 ···CN•
The flux cp varies from 0 to 2-D' dep'ending on the mutual
pos1t1on 0 t e ~·tL charge and the V 11"" man1fold.
. . f h • It i s
approximately zero when their distance is large with respect to
the size of VptJ • According to Dirac's quantization condition
cf>. would reach its maximum value 211' if V'" were closed
c'l'a
V'·'; 0 ) and l i nked to C ~ • Since ? V"'* 0 this
value can be reached only for a very peculiar configuration of

the ~. ,." charge: ~t
. must b e I ~nked
. an d very near to V P to.
but rather far from the p-dimensional source, and its size must
be negligible in comparison with V P"'I in order that 4>".
do not appreciably differ from the flux through a closed
manifold obtained by adding to V ptl another suitable (P+1)-
-dimensional manifold.
We leave out for the moment the somewhat exceptional case q=O
(gas of monopoles) and take an elementary, illustrative example
out of a gas of electric currents (q=1) in a 3-dimensional space
(superconductivity). In this case the contribution to ~ due to
a circular current loop C of radius t orthogonal to a
segment V t passing through its center and joining two magnetic
monopoles (external sources) is
(III.4) cp: rr( '" + ~

)
v;a.tl"'" W~;'"
where x and y are the distances of the plane of c from the two
monopoles. We clearly see that cp= 'l,n when \"'~ 0 and ~/~:l0
as expected.
We now turn to the central point of our argument. We select
among all the configurations of the gas those where f=
r;r{A1)~ "'1
Such configurations have a well - defined structure: the ~
distribution of charges far from V pt-I is arbi trarYi then
there is a thick layer of vacuum wrapping the manifold V pH .
inside V,U lie charges of very small size linked to it. The'
configurations that appreciably differ from the above are
strongly suppressed by the damping factor 1 .
liTe can then
approximate the sum over all the configurations of the gas in
the eq. (IIL3) by those in which 1~1 :
itt,o., Z 1.. ~
(IIL5) (e >'Y ~ N!
The configurations wi th ~, i differ substantially from

those associated with the ground state of the gas ",here the
40 F. GLiOZZI
charges are evenly distributed in the space. Rather they

correspond to collective excitations of the gaSjtheir structure·
clearly suggests that their energy be approximately proportional
to the volume '\J' p+1 of V p.l~ and to the number N of
charges. Thus the probability L PN(1:,) is simply given by
the Boltzmann factor e- ~ N""fIt-.
. showing directly the "area
NI
law" for not too small values of 'li''''
(III.6) -Vl1 pte

e.
The case of a gas of magnetic monopoles (q=o) deserves a
special comment. To be definite let us choose d=3, and take for
Vt the minimal surface whose boundary is a current loop. In
this case, according to Coulomb law, the maximal value of cP
reached when the monopole lies on V"', is TT instead of tv
Configurations of this kind have ~ =(_I)N and are suppressed
in the general case (f i= 0) because there is no way to enhance
those with an even number of charges.
In the present case there is a natural way to order the monopoles
on the surface in pairs by setting the positive monopoles on
one side and the negative ones on the other side of V t • Such
configurations have ~ = 1 then we can derive the area law for
the Wilson integral like in the general case.
Note that the expectation value of the Wilson factor may
also be interpreted as the partition function of the gas
modified by an external dual source. We can then summarize the
previous analysis by saying that the interactions due to such
a source lean to promote only those configurations with i
The mechanism we have described coincides, in the particular
r::
case ,=0
with the one pointed out by Polyakov 2 in his
argument on the confinement of the electric charge in the compact
QED in the three-dimensional space-time.
IV. DUALITY TRANSFORHATION IN THE LATTICE VEl?SION OF GAUGE THEORY
One of the greatest advantages of expressing the topological

features of duality transformation in the Homology language is
that it is possible to extend in a straightforward way our
considerations to the lattice formulation of field theory, which
is one of the more powerful techniques to study phase transition

phenomena and high coupling expansion in field theory.
The relation between the Homology properties of a field
configuration and its lattice version is quite obvious.
Indeed it is sufficient to know the cycles which are not
boundaries. To this end we introduce a triangulation of the
ambient d-dimensional space of the theory. To be more specific,
we choose as a possible triangulation a cubic lattice in
d-dimensions.
The p-cycles of the triangulation are those linear combinations
of the p-cells of the cubic lattice which are without boundary
the O-cells are the points of the lattic~; the 1-cells are the
links; the 2-cells are the plaquettes (i.e. the elementary,
oriented squares of the lattice) and so on.
The Homology properties of a field configuration of a generalized
p-gauge theory (i.e. a theory built up with potential forms a )
are uniquely defined if we know the value of the integral-!,~~
over each p-cell C f since we can obviously evaluate with' it
the integral over any p-cycle of the lattice.
Thus in the lattice version of field theory the fundamental
objects are no longer the forms, but rather their integrals
over the cells.
We have already seen that the Gauss theorem provides a
means to translate the exterior forms calculus (Cohomology)
into that of their integrals (Homology). The last thing we
need to apply the Homology formalism to the lattice theory is
the lattice version of the dual .. 4.p of Q". Since *
tJ.p is
a (d-p)-form built with a , we have to define the quantity *4 pJ
where" Cp is a (d-p)-cel~ associated with C p . ."
In order to associate to each p-cell one and only one (d-p)-
-cell, one has to consider also the reciprocal or dual lattice.
The dual * L of a cubic lattice L is simply obtained by
shifting L of one half of the lattice spacing a in all the d
possible directions. Then it is easy to see that there is a
natural one-to-one correspondence between the p-cells of Land
the (d-p) -cells of it L .
We are now prepared to apply our considerations to the
gauge theories on a lattice.
We start with ordinary abelian gauge fields (p=1). Then
wi th each link I of Lit is associated the phase ~ = a. or,J
e e
42 F. GLiOZZI
equivalently, the phase factor , e _ e

_ i~t
which obviously
is an element o£ an U(1) gauge group.
The action S o£ the theory is chosen in such a way that it re-
produces the usual continuous action when one takes the zero
limi t o£ the lattice spacing Q, ... 0 •
The most popular (but not unique) way is constructed as follows.
Take a plaquette P, multiply the phase factors ~e o£ the four
links bordering P and sum over all the plaquettes and their
possible orientations in the lattice L. The result is
(IV.1)
where ~ is the coupling constant. Note that the sum over the
orientations makes S real.
The object of interest is the partition function. It reads
S
't I
(IV.2) ~ e
{'t 1
i
where
the variables 'e.
indicates the sum over all the possible values o£
associated with the links of L.
One can also restrict the set {~t J to any discrete
subgroup o£ U(1).
A particularly interesting choic~ is ZeN), i.e. the discrete
, '"11"
group formed by the N elements e w (n=1,2, ••• N), because
it can be considered as the center o£ a fully broken SU(N)
gauge theory (say by Higgs mechanism making all gluons massive)
which seems to be 16 one o£ the most important ingredients to
understand quark confinement.
The generalization o£ eq. (IV.1) to a gauge theory of type
p is straightforward: S is simply the sum over all the (p+1)-cells
of the products
. r;r ,., ,,/
(0". o£ the elements 'cl associated
,.,
. c.0.,(
Wl. th the p-cells borderl.ng G • Gauss theorem shows that the
operator T;1 ,., has the role o£ the exterior derivative d
actl.ng on where
~ I 0.,
(IV.3) -= e c.'
It is an useful and amusing exercise to verify the
following translation table from the exterior forms language
to the lattice manipulations.
Translation table
a)
<
b)
~ ,. , ~ ,. =1
(', f: 'J
n'It' () ,pIt)) n
G
c)
A~a,*~~
J( ~, cf ,. ,/+1)
in b) the C/ are all the p-cells bordering c/·'
while
in c) the product r-T is the one over all the (p+1)-cells
. . , J
hav~ng a s~de, ~n common.
In the particular case p=1 with the action (IV.1) the

property a) expresses in terms of lattice quantities the fact
that the 2-form associated with the plaquettes is derived by
the potential 1-forms defined on the links, and b) says simply
that the "magnetic" flux through each cube c 3 of L is zero.
The duality transformation in the exterior forms language
,... associated with the force
consists in expressing the (p+1)-form
in terms of the dual potential "*
it p (clearly the topological
excitations of L become the local excitations of *L).
In the lattice version we may proceed in an analogous way. Let
us take for simplicity p=1 d=4 and write the action (IV.1) as
(IV.4)
s ":: f'l.!. ~ ~f
or~ented
plaquesses
where ~ 1 are unconstrained elements of the gauge group.

Thus we have to add to S the constraint
(IV. 5) ~Tf Otf=O

1 E (" d
through a Lagrange multiplier k. e which is associated with
*~ dual to the cube c 3 • I,le have L
't nr',,) ~"e
the link
(IV.G) Z ((I,) = Z 2. T1 e -1
{'t1 .( "lIfel f pI
44 F. GLiOZZI
•
The idea is now to integrate first over the ~e ~. To this end
we have to transform the Boltzmann term into the product of
factors, each of which refers to only one plaquette.
It is then easy to verify that the factor with the Lagrange
multiplier can be rewritten in the form
(IV.7)
--
where the product II Q.f refers to the four links *e
of It L
bordering the plaquette .tr l dual to P.
multiplier ~*'
becomes the character
.
ide have not yet specified the range of values of the Lagrange
'ile may normalize
Ul#l
Of~!
h.*e
such that eq.(IV.7)
i.e.
(IV.S)
~J..~~~l) : ~tJ..¥l
This choice has many advantages. First of all inserting
(IV. 8) in (IV. C) vie see that the duality transformation is
directly related to the harmonic or Fourier analysis of the
Poltzmann factor, ,.,hich is just the historical way to do the
duality transformation in lattice theories. Moreover the set of
the indices k*"! of the characters of an abelian group G
forms itself a group ~ G called the dual of G. Thus under a
duality transformation a theory with a symmetry group G becomes
one ,.,i th another symmetry JIf G.
Ue have the follovling table,
which exhausts all the possible abelian groups
Table
G
U(~) compact
U(t) non-compact~ U(t) non-compact
Z (w) ~ 2 (N)
*" * G- G
can now answer the follovling question: what are the necessary
\:Ie
and sufficient criteria satisfied by a p-gauge theory on a
lattice with symmetry group G in order to be self-dual?

He have the following three conditions
i) The classical dimensions of the topological excitations (i.e.
the excitations of the dual lattice) must coincide vii th those
of the elementary excitations. Eq. (II.15) tells us that
2 P = tL-2
ii) G=" G
iii) The functional form of the action is invariant under the
characters expansion.
Condition i) tells us that for d=4 only the ordinary p=1
gauge theory might be self-dual. From condition ii) we see that
the simplest self-dual theories are those associated to a non-
-compact U(1) or to Z(N), although there are other possible
choices made with a suitable combination of the groups of the
above table such that one has again G ~ ~ (;. .
For a non-compact U(1) it is well known that the only
functional form of the Boltzmann factor which is invariant
under the Fourier analysis is the gaussian, where the action
may be considered as the quadratic approximation of the action
-
(IV.4).
Ide have
S 'L
- P
(IV.9) e J dp e e
= -Ob
where the coupling constants ~ and B are related by
( IV. 10 ) f B = (G IT )_1
'L
and "" and p are the U( 1) fields defined on the plaquettes of

the direct and the dual lattice.
It is easy to see that (J and B are related to the electric
and magnetic charges e and g by
(IV. 11)
f .-. eJ. "
Hence we find, as we expected, that the duality transformation
on the lattice version of a U(1) gauge theory yields again the
46 F. GLIOZZI
Dirac quantization condition.

It is also interesting to see what is the form of eq.s
(IV.9, 10) for ZeN) gauge theories. It is not yet known in
this case what is the general self-dual functional form of the
Boltzmann factor. However it has been shown 24 ,25 that the
standard action defined in eq. (IV.1) gives rise to a self-dual
ZeN) gauge theory for N~ 4.
It is straightforward to verify by direct computation the
following relationship for the harmonic analysis of the
i: ..f 89~(~/
Boltzmann factor.
(IV.12) e l~~e· ...· = ~

~ w
where ~e e 2(~) ~t::~&.:1 ~ I.. = O,~, 2, •• N·I
tJ:2. B :_1r.. ~ ~o.... l~ ~ = (e ~~ e._ 18J!

8 __ te. . e. (b-e.'
-! ~ (l
~ nt t
-' f
(IV.13 ) tJ::1 ~.
- 3 tPott'-.. ~" '1 e. 8
N-=4 B -= Q.1tt l Ill?

~ & F. - ~~_e-f
4-
~
with these notations the self-duality condition is simply
(IV.14) 0< Z (B)

It is worth noting that the duality relationship between coupling
constants in the N=2 case is just the same of the one obtained
by Kramers and Wannier (eq. 1.2) for the two-dimensional Ising
model.
Indeed our analysis shows that such a relationship depends only
on the symmetry group and not on the dimensions of the space.
The most important feature of a self-dual theory is that
this property enable one to locate exactly the critical points
under the assumption that the system undergoes only one phase
transition. Indeed from eq. (IV.14) we conclude at once that

the critical value 1\ is the solution of the equation
{- Co
(IV. 15)
B-=f
It is then easy to verify the following table
group critical coupling references
2('2.) foe. -:. i ~ ( ,+h ) 8
Z (3) /1, : !'3 t.. (,.. (3) 24, 25
Z (4) (\ _ ~(It\ll.)
c-
24, 25
15 16
The works of Handelstam and It Hooft have emphasized the
relevance of ZeN) gauge fields in understanding the quark
confinement in SU(N~ gauge theories. Indeed one can construct
classical solutions 6 of SU(N) gauge theory which represent
non-Abelian magnetic flux tubes (fluxons).
In the lattice version these solutions become just the
topological excitations of the ZeN) gauge theory25.
The phase where quarks are permanently confined is the one
in which the fluxons condense. Such a condensation is just the
phase transition described by the duality transformation. Thus
it is possible to have a crude extimate 2 5 of the critical
coupling constants in the SU(N) models caused by this
condensation.
REFERENCES
1. K.G. 11'1'ilson, Phys. Rev. D10, 2445 (1974).

2. A.M. Polyakov, Phys. Lett. B59, 79, 82 (1975).
3. A.A. Migdal, JETP 42, 413, 743 (1975).
4. H.B. Nielsen and P. Olesen, Nucl. Phys. B61, 45 (1973).
5. L.J. Tassie, Phys. Lett. B46, 397 (1973).
6. Y. Nambu, Phys. Rev. D10, 4262 (1974).
7. G. Parisi, Phys. Rev. D11, 970 (1975).
8. H.A. Kramers and G.H. r,"annier, Phys. Rev. 60,252 (1949).
9. P.A.M. Dirac, Proc. Roy. Soc. A133, 60 (1931).
10. E. Amaldi and N. Cabibbo in Aspects of Quantum Theory
(Cambridge University Press. 1972).
48 F. GLiOZZI
11. P. Goddard and D.I. Olive, New Developments in the theory

of Magnetic Monopoles CERN prep. TH 2445 (1977).
12. S. Coleman, Phys. Rev. D11, 2088 (1975).
13. S. Mandelstam, Phys. Rev. D11, 3026 (1975).
14. V. de Alfaro, S. Fubini and G. Furlan, Phys. Lett. 65B,
163 (1976),
C. Callan, R. Dashen, and D. Gross Phys. Lett. 66B, 375 (1977)
15. S. Mandelstam, Talk presented at the American Physical
Society, Washington D.C. (1977).
16. G. 't Hooft, Utrecht preprint (1977); D. Foerster, Cornell
preprint CLN 5394 (1978).
17. R. Savit, Phys. Rev. Letters 39, 55 (1977).
18. M.B. Einhorn and R. Savit Fermilab preprint 77/97 THY (1977)
19. M.E. Peskin, preprint HUTP-77/AO 83 (1977)
20. F. Gliozzi, T. Regge and M.A. Virasoro, Torino preprint
IFTT 312 (1978) and in preparation
21. H.Flanders, Differental forms (Academic Press, New York,
1963).
22. M. Kalb and P. Ramond Phys. Rev. D9, 2273 (1974).
23. T.T. i-lu and C.N. Yang Phys. Rev. D12, 3845 (1975)
24. C.P. Korthals Altes, preprint CNRS Marseille 78/P.1003 (1979)
25. T. Yoneya, Nucl. Phys. B144, 195 (1978).
THERMODYNAMICS OF NUCLEAR MATTER FROM THE
STATISTICAL BOOTSTRAP MODEL
R. Hagedorn
CERN, Geneva, Switzerland
1. Montvay
CERN and Universitat Bielefeld, Fakultat fur Physik
J. Rafelski
CERN, Geneva, Switzerland
ABSTRACT
We study the properties of nuclear matter within the

framework of a modified and generalized statistical bootstrap
model in which the volume of a fireball grows with its mass.
We find that the such described nuclear matter can exist in
two phases. In particular we consider in a numerical example
the high temperature (T < To ~ 150 MeV) regime of the
gaseous phase with a density of less than ~ 0.75 of normal
nuclear density.
49
SO R. HAGEDORN, I. MONTVAY AND J. RAFELSKI
1. INTRODUCTION
In order to understand high-energy heavy ion collisions or

even perhaps high-energy hadron-nucleus scattering, we must study
the equations of state of nuclear matter. From the point of view
of a theoretical physicist, the inverse statement is even more na-
tural: our ideas about the properties of nuclear matter at high
and low densities and temperatures can be tested in high-energy
nucleus-nucleus and hadron-nucleus collisions. Moreover, this
knowledge is certainly essential in order to understand the proper-
ties of hadronic many-body objects in astrophysics, such as neutron
stars, the Universe at early time, stellar collapse and perhaps
even quasars.
While we are aware of the possible richness of the nuclear

matter properties, in our approach to these problems we will con-
centrate on the gross features of nuclear matter that follow when
we incorporate into the description the following basic properties:
1) conservation of baryon number and clustering of nucleons (i.e.,

attractive forces leading to many-body clusters with well-defined
baryon number);
2) nucleon (isobar) excitations and internal cluster excitations
(i.e., internal degrees of freedom that can absorb part of the en-
ergy of the system at finite temperature, thus transforming kinetic
energy into mass);
3) approximate extensivity of nuclear matter (volume roughly
proportional to baryon number, i.e., effectively a short-range re-
pUlsion);
4) co-existence of a pion gas when the temperature is not equal to
zero (and behaving properly even in the absence of nuclear matter);
5) baryon-antibaryon pair creation;
6) "chemical" equilibrium between all constituents of the system
(nucleons, isobars, clusters, pions ••• ).
NUCLEAR MATTER AND THE STATISTICAL BOOTSTRAP MODEL 51
Our present work should be most trustworthy in the domain of high

temperatures and moderately high density, where details of the
interaction, Fermi and Bose statistics, as well as the quark struc-
ture of nucleons, are most likely negligible. Also not considered
explicitly is the isospin of the nuclei.
In order to derive the physical properties of such a system

which are independent of a particular choice of the two-body and
multibody interaction we employ a technique ("bootstrap") developed
for similar problems in elementary particle physics -- here, how-
ever, sufficiently modified to suit the different physical environ-
ment. An additional motivation in this direction is the recent
recognition that the understanding of nuclear matter at the satura-
tion point depends very sensitively on the character of the two-
body potential at short distances which is not well defined by two-
body reactions. It is possible to view the bootstrap technique only
as a convenient way to introduce some physical properties which can-
not be so easily defined by the choice of a specific potential but
which globally might even be more important than details of the
two-body force.
Let us now explain the general idea of the bootstrap descrip-

tion of the nuclear matter. Consider an assembly of b nucleons.
We can view it as an assembly of (b/2) two-particle clusters •••
or also as two (b/2) - nucleon clusters all possible divisions
will contribute to the number of states of the b nucleon system.
In turn, each cluster containing b. nucleons can be viewed in a
~
similar fashion. It is simple to write an equation characteristic

of such a system; neglecting for the moment all further degrees
of freedom but the baryon number b, the number of states will be
described by a function p(b) obeying the equation
52 R. HAGEDORN, I. MONTVAY AND J. RAFELSKI
where the sum runs over all the sets with j elements (clusters)
wi th b. e {l, ... , b - l}, such that the conservation of baryon num-
~
ber is secured by the Kronecker delta function 0K()' The exist-

ence of the quantity p as defined by the above equation is obvi-
ous by recursion, which, however, is not practical for actual cal-
culations. The above equation is already a true "integral"
equation in the bootstrap sense. To see this more clearly we con-
sider in the same context a continuous variable m, say "mass",
for illustrative purposes
(1. 2)
here Po ~s some given function expressing a limiting case in which

clusters do not contribute. As the lowest hadronic mass is > 0,
the sum over j runs in practice only over a finite number of con-
tributions with m. different from zero. This integral equation
~
expresses most clearly the fact that the cluster described by p(m)
consists of an arbitrary number of clusters of smaller mass, each
of which is in turn made of an arbitrary number of smaller clusters,
. .
etc., as can b e seen ~terat~ng h
tea b ove .
equat~on
*) •
The quantity P (derived from a more realistic equation) may

be taken as the level density of nuclear matter, provided that some
further statistical factors are considered and that Po expresses
the level density properly when other terms do not contribute.
Since the knowledge of the level density is sufficient for a com-
plete description of the physical properties of the system from
the point of view of statistical physics, the system is completely
defined once the proper bootstrap equation reflecting our physical
requirements is written. Clearly, one can write many different
bootstrap equations under given general constraints, each with some
parameters allowing an adjustment to experimental data. Here we
follow the simplest possible line of approach choosing the simplest
possible case in a numerical example.
Plan of the paper
Section 2 We discuss the bootstrap hypothesis first in the con-

text of a strongly interacting pion gas. The boot-
strap equation of the pion gas is solved and discussed.
Section 3 The mass spectrum and its Laplace transform are used
to obtain a thermodynamic description of the system.
Section 4 We write down, discuss and solve the bootstrap equa-

tion for nuclear matter. It is much richer than that
of the pion gas, which it contains as a special case.
Section 5 The general solution of the bootstrap equation for

nuclear matter leads to a corresponding statistical
thermodynamics. Our postulates result in two main
*) At this point the reader may have the uneasy feeling that we
are doubly (in fact more than doubly) counting states. This
problem has been discussed in detail in references 1) and 2), to
which we refer the reader. Here we must ask him to simply be-
lieve that our counting is correct, if clusters exist (see also
Subsection 3.2).
properties of the thus described system:
i) there exists a maximum temperature, which is of

the order of that of the pion gas (To",m);
7T
ii) there exists at all temperatures o ~ T ~ To a
critical baryon number density separating a low
density "gas phase" from a state where a conden-
sate and its vapour exist in equilibrium.
A numerical study is presented in which the simplest

non-trivial input spectrum is assumed; the corres-
ponding model is solved explicitly and the results
are displayed graphically. This case shows all es-
sential features but it is still too far from reality
to be taken as more than a qualitative prediction.
Section 6 The "liquid phase" requires different mathematical

techniques than the "gas phase". The general ap-
proach is outlined.
Section 7 Summary and outlook.
Our notation and units:
n = c = k (Boltzmann constant) = I
the only dimensional unit is I GeV
Metric : a • b =a ~
b
~
= aobo -+-+
- a • b.
Remark: Throughout this paper we use only Boltzmann statistics.

As the bootstrap approach leads to an extremely rich mass
spectrum, it is practically irrelevant whether a particular
cluster or particle is a boson or a fermion or a
Boltzmannion: it (almost) never happens that two equal
clusters occupy the same state.
2. THE STATISTICAL BOOTSTRAP METHOD
2.1 The statistical bootstrap model in particle physics (SBM)
The motivation for a statistical bootstrap model in particle

physics comes from two sources:
i) the abundant production of particles in high-energy p-p col-

lisions and a momentum distribution of these particles which
suggests that there might be some analogy to black body radi-
ation emitted from moving sources;
ii) the apparent existence of intermediate states in which lumps
of highly excited hadronic matter ("fireballs") are staying
together before decaying.
Thus it was tempting to describe the particle production process

as pion black-body radiation emitted from one or several fireballs
with a volume Va ~ 4TI~3m~). This idea was first proposed by
H. Koppe 3 ) and for this reason it is called the Fermi Statistical
Mode1 4). As for a statistical-thermodynamical description the
density of states o{E) is necessary and sufficient, we may ex-
press the Koppe-Fermi approach as follows:
~o6;(E) L00
X
(.21)3 :=: ~~r O\.!
)C. Sb(E-L V1-M ('3 til -+

2. . ." ) 0 (~Dj) II
i- ~ d,3
9 Pi
(2.l)
i=1 ' 'It i=I' i:1 l2J:)3

This is nothing else than the phase space density of a pion gas
with free particle creation. If one puts m= 0 and multiplies by
2 for the two helicity states of a light quantum, one obtains from
(2.l) all the usual formulae of the electromagnetic black-body ra-
diation (Planck's law) in the Boltzmann limit.
Before proceeding we shall write this formula, which is meant

to be valid in the rest frame of the emitting volume va, in a
56 R. HAGEDORN, I. MONTVAY AND J. RAFELSKI.
manifestly Lorentz-invariant formS). We introduce the four volume =

(rest volume) x (four velocity)
.
)
(2.2)
Then with
(2.3)
The right-hand side is the number of states in {p, d 3p} of a

scalar neutral particle of mass mn enclosed in vo, expressed
covariantly. Now we modify the definition of the density of states
= number of states ~n {p, d 4 p} (2.5)
and rewrite (2.1) as
(2.6)
*) We shall refer to 2V·p/(2n)3 d 4 p as the "Touschek measure".

where we write the one-particle term separately for convenience.

This is a relativistically invariant equation for the density of
states of a gas in which the interaction manifests itself only via
creation and absorption of Boltzmann pions.
The next important idea (see also Subsection 3.2) was to ad-
mit other particles than just pions, and in particular resonant
states of pions, just as if they were stable particles 6 ). Not
knowing which ones should be admitted and how many there are, we
might put them in a mass spectrum of admissible input particles
p. (m); the pion contributes to p. (m) a 0 function 0 (m-m1f ) ,
1n 1n
resonances contribute smeared-out 0 functions. For the moment,
p. (m) is a function which represents our (incomplete) knowledge
1n
of the true mass spectrum p(m).
Now (2.6) becomes
o(r~ ,D· V);: fLY 3 do (p~ AU.~ ) +

l,.2tt)
(2.7)
DO tk I)t
+,1;2.~ rrcr-t;, pjl:)~\""(NA.;)d. (r~-4I<~) d.·pjd.w.,

Note that we have restricted the one-particle state to have the
pion mass. Higher mass "one-particle states" are already contained
in the sum, namely when in any of its terms all Pi + mi. Our new
equation for 0(p2, P'V) describes the density of states of a many-
component gas: each species of particle contained in p. (m) is
1n
present in the gas. All these components are in "chemical" equili-
brium, since all kinds of reactions (e.g., 1f + 1f + W + 1f + 1f + 1f,
etc.) are now possible and neither the total particle number nor
that of any of the various components is fixed.
Such a system should already exhibit a good deal of the pro-

perties of a strongly interacting gas; we argue that our descrip-
tion will indeed contain the essence of strong interactions, once
the incomplete p. (m) is replaced by the true mass spectrum
1.n
p(m) -- which unfortunately is not known to us.
The key idea leading to the hadronic bootstrap is the obser-

vation that a(p2, p·V) can be related to the mass spectrum p(m).
Suppose we could insert the true mass spectrum p(m) into Eq. (2.7),
then a(p2, p·V) would be the density of states of a "fireball" of
hadronic dimension built up from all strongly interacting particles
in statistical equilibrium. Such a fireball is itself a highly ex-
cited hadron with mass m = /P2. For reasons of consistency it
should then be admitted as a constituent particle in fireballs of
larger mass. Hence it should already be present in the true p(m).
As both, a(p2, p·V) and p(m), are densities of states, it fol-
lows that if p(m) is the true mass spectrum, a(p2, p·V) is it-
self, apart from some minor kinematical differences, the true mass
spectrum at m =;pI. This statement establishes a new relation
between p and a, leading to an integral equation, the "bootstrap
equation" (BE). Physically it is equivalent to the postulate that
resonances and fireballs are one and the same and that fireballs con-
sist of fireballs.
In order to find the precise relation between p(m) and

a(p2, p·V) we consider the conceptual differences between them as
exhibited by Eq. (2.7): while p(m) counts all hadrons (as given
for example in the Tables of the Particle Data Group7» as being at
rest in their own confining volume, a(p2, p·V) is the density of
states of an object with mass m = IP2 allowed to freely move in its
confining volume instead of being at rest. This fact is also re-
flected by the dependence of a on the scalar product of p·V.
Thus a counts more states than p (and contains more information).
In order to relate a to p, we restrict this freedom by
NUCLEAR MATTER AND THE STATISTICAL BOOTSTRAP MODEL S9
requiring that p~ and V~ be parallel four vectors, i.e., have a

common rest frame. Then the first term of Eq. (2.7) becomes
which, apart from the factor vO/(2TI)3, is just the beginning of

p(m) •
Hence we multiply Eq. (2.7) by (2TI)3/vO and go to the com-

mon rest frame of p~ and v~. We obtain
(2.9)
ir=, p.1 "fc,=, ~.c""i)

t)\
)( 03C tl pi dill! 3 l
f .""
-1
Now the left-hand side has dimension m and represents the irtter-
na1 density of states of a system of mass m at rest in its own
confining volume vo; this density begins with o(m - m) and has
TI
a continuum for m > 2m • It therefore might be considered as an
TI
averaged mass spectrum (the true one is not yet continuous at
m~ 2mTI ), which asymptotically becomes physically equivalent to
p(m). In the first paper on statistical bootstrap8), the postulate
was accordingly that the logarithms of the densities of states (i.e.,
60 R. HAGEDORN, I. MONTVAY AND J. RAFElSKI
essentially the entropies of the corresponding systems) should be-

come asymptotically equal: define
(2.10)
Then the two equations
- 1 (2.na)
and
(2.l1b)
constitute the bootstrap postulate in its weakest form. In the

following deve1opment 9)-16) the formulation underwent several
changes: Frautschi 9 ) required p = p. = p (= some approximate
av l.n 13)
average over the true complete mass spectrum). Yellin wrote the
equation in invariant momentum space [see Eq. (2.15b)]
(2.12)
which is almost the same as Frautschi's version, but not quite,

since
J
d"Pi ~ (pf) = d3Pi L{""~)J.lp:-.: )dfu~ dpo, i
= d"Pi S
1:'(-1) /IIIi d ...
V~2+ _,2.
(2.13)
Comparing this integration measure with the one in (2.llb) we see

that the square root in the denominator of (2.13) makes equation
(2.12) dynamically different from Frautschi's. If all interaction
is believed to be contained in p(m) then (2.12) appears not to
be correct. As we do not really know, we might accept (2.12) as
well (in fact, it has been widely considered as "the bootstrap
equation"). Many applications have been based on both of them.
The results in particle physics are, to a large extent, independent
of which particular version one chooses. Namely, all equations of
this type (there exist various other forms of it which we cannot
discuss here) lead to one common characteristic result: asymptoti-
cally for m+ 00 the mass spectrum grows exponentially:
(2.14)
I
)
where To is a "limiting temperature" and where the values of

'a' and To depend on the version chosen. We will solve a BE
and prove Eq. (2.l4) further below.
When we decided to apply the above line of approach to in-

finite nuclear matter, we had to take extensivity (volume ~ propor-
tional to baryon number) into account. By trying to do that rela-
tivistically and consistently we became aware of several obscure
points in the statistical bootstrap model (SBM) which had to be
cleared up; the result was yet another version of SBM, which then
could easily be generalized to include the nuclear matter case.
2.2 SBM reconsidered
From the above discussion, it follows that two questions have

to be answered:
i) SBM has much to do with phase space; which is the correct

integration measure?
ii) a mass spectrum deals by definition with particles carrying

their confining volume. The co-moving volume should be in-
troduced covariantly and more consistently than by (2.8), (2.9);
furthermore, in that equation all fireballs have the same
volume vo, independently of their mass. While this mayor
may not be true for pionic fireballs (linear Regge trajectories
J = J o + am2 require that some relevant length of the par-
ticles should grow proportional to m), it cannot be assumed
for nuclear clusters. How can we incorporate covariantly ad-
ditive co-moving volumes?
We now discuss these two questions in detail.
2.2.1 The integration measure

For n free (Boltzmann) particles of mass m two sorts of
integrals are commonly used without much reflection: Invariant
Phase Space (IPS)*) and Invariant Momentum Space (IMS):
*) which has Touschek measure
(2.l5a)
(2.l5b)
For the light quantum gas (m = 0) we obtain the Stefan-Boltzmann

law by summing over n and going over to the thermodynamic descrip-
jl k
tion (see, e.g., Ref. 2)). The result is, with V = (VjlV )2
IPS (E) - l~t) Tit (2.l6a)
IMS <1£) :: (2ttB ~(3)) T '3 (2.l6b)
For a dilute gas we know that the first expression is correct, the
second wrong. Furthermore, (2.l5a) can be derived in a straight-
forward way from the elementary rule that the available number of
states in {p, d 3 p} of one particle enclosed in a volume V is
given by
(2.17)
while (2.l5b) cannot be justified that way. For more detail see
Ref. 17). Since in the absence of interaction p(m) + oem - m ),
1T
the density of states a(p2, P'V) should reduce to that of a free
pion gas. Therefore, we conclude that SBM should be formulated in
IPS (Touschek measure) as was done in the first papers and not in
IMS as has become customary since.
2.2.2 The co-moving volume
In deriving Eqs. (2.9)-(2.11) we have treated the volume un-

symmetrically by attributing it to the main fireball (on the left-
hand side of the BE), while its constituents described by the
p(mi ) (on the right-hand side of the BE) behave as point-like
particles all enclosed in the same common volume; as they, in
turn, are described by the BE and then acquire the same volume,
they must penetrate each other perfectly.
We now shall treat all particles on the same footing.
We start by defining a new density of states of a system en-

closed in a four volume V:
= (2.18)
number of states in {p,d 4 p; V, d4 V}
According to the conclusions reached in the foregoing subsec-

tion, we have used the Touschek measure to express the kinematic
part of the level density (cf. Eq. (2.4»; o(p,V) represents the
part due to internal degrees of freedom.
In a world consisting only of strongly interacting neutral

scalar Boltzmann pions the mass spectrum must start with a C func-
tion representing the pion:
(2.19)
NUCLEAR MATTER AND THE STATISTICAL BOOTSTRAP MODEL 6S
We generalize to arbitrary p by postulating a new BE of the type

of Eq. (2.7), in which we give additive volumes to all constituents.
This procedure enables us later to generalize straightforwardly to
the case of nuclear matter. Whether in the pion case it is better
than the old BE's remains to be seen.
(2.20)
This equation is relativistically invariant and is written in

Touschek measure. Furthermore we have postulated additive four
volumes such as to correspond to "dense packing". Strict addi-
tivity is not necessary but it is assumed here on physical grounds.
Mathematically, it is possible to introduce in the bootstrap equa-
tion some prescribed way of volume addition replacing
t(V -L: v· ACe]2 )

1M
>
""
O"(V-r. Vi )
;:1'1 'A lp~)
1=1
(2.21)
with arbitrary given A(p2); we will not pursue this possibility.
Next we require that all fireballs carry their four volumes

(i.e., and V.IIP.) because only then can we interpret 0 as
1. 1.
a mass spectrum. We therefore demand for all V, V.:
1.
(2.22)
Here V(m) = '\fV2 is a Lorentz scalar which may depend on m.

These requirements are imposed on the BE (2.20) by the ansatz
.2V·~cr(PIV)::II f(V-VCa)£) (~V.~ &-(p, V) (2.23)

(2.Jr) 2/(')
The o~ function implies that its factor is a function of

p2 = m2 alone:
(2.24)
Here B(m 2) represents the kinematic factor 2V·p/(2n)3, while

T(p2) counts the internal degrees of freedom. Our BE (2.20) now
becomes
The rrd~V. integration has already been done.

1
All volume 04 functions must have the same argument; fur-

thermore, we have Hence
(2.25)
for all n ~ 2 and arbitrary p .• Therefore,

1.
V(~)= A Mi-
(2.26)
S(~) = 2Arm,'/(1:CC)3
where A is some fundamental parameter of our mode1*) • We can esti-
mate its order of magnitude in various ways.
i) We may require that B(m 2 ) = Bo = 1/(4mm2 ), which yielded a

. 7T. ~ 14)
reasonable value for To 1.n ear11.er ca1cu1at1.ons ; we then ob-
tain
(2.27)
ii) We might fix A by requiring that for some mass ~
$ ~. $ 10 m7T ) V(~) = Vo = (47T/3)m;3. This gives
Af"-
'-
(2.28)
iii) We might choose A to correspond to the MIT bag model 18 ):
*) A-I is the constant (!) energy density of fireballs.

- A . mt. (2.29)
q.. (120 MeV)" ~
~. 2. . 10 3 [4eV -If]
(2.30)
loS.1f [MeV. ~-3J
which corresponds to 11 = 9 m7f in our previous estimate. We
shall consider A as a free parameter limited to 10 3 to 10~
-~
GeV
With Eq. (2.26) satisfied, the o~ functions for the volumes
factor out and we obtain the reduced BE
'" to"- "- ~ r ~ 2

Blp )1;lp') = Blp )00 (p-tturr: ) +
(2.31)
E~! SO'(p-£Pir[B(p~} ~cp~)d'fi

eO tIC II(
Now we have to identify the physical mass spectrum

This identification is not at all obvious. Remember that T(p2) ori-
ginates from o(p,V), which we called above "the part of the level
density due to internal degrees of freedom" while B(p2) (see Eq.
(2.24» comes from the "kinematical factor" 2V·p/(27f)3. One could
therefore argue that T(m 2 ) is the mass spectrum, while B(m2 ) is
some irrelevant reminiscence of external degrees of freedom. And
indeed: the mass spectrum should start with oo(m2 - m2 ) and not
7f
with B(m~)oo(m2 - m~). One can, however, also argue that B(p2)
is of dynamical origin and should be absorbed into the mass spec-
trum. Namely, by imposing co-moving additive volumes on our dynami-
cal equation (the BE) we have made the volume an inseparable,
dynamical part of the fireballs. Furthermore, if the volumes move

with the fireballs, one can no longer claim that the factor
2V'p/(2n)3 represents the kinematical degrees of freedom of a par-
ticle confined to some externally given volume V, since now the
particle is always at rest: 2V'p/(2n)3 + 2Am 2 /(2n)3. This factor
has thus become a dynamical factor.
We now take this latter point of view*); the spectrum can be

defined as to contain all p2 dependence and the factor can be
made constant such that the spectrum T(p2) starts with
oO(p2 - m2 ):
n
tv
B(p2)
T(p2) - '"
-C (f2 )
g{Ittf~)
(2.32)
g - B(mt~)
With the identity BT(p2) B(p2)T(p2) our new BE is
(2.33)
*) By adopting this interpretation we m~n~m~ze the difference

with the older BE's. The case of nuclear matter discussed
in Section 4 will lead us nearer to the first argument where
T(p2) is the mass spectrum.
It is identical to Yellin's BE (2.12). Although we started with

an equation written in Touschek measure, the co-moving volumes have
led to IMS measure (Eqs. (2.22)-(2.24». We keep in mind that B
might be taken to be proportional to p2 instead of being constant;
we are aware of the important consequences this will have, but we
cannot discuss them at length in the present publication.
We stress that we do not claim that our BE (2.33) be "the BE";

any of the other versions may be considered as well; the situation
is similar to the one in Lagrangian field theories: which, if any,
is "the Lagrangian"? Only experiment will give us the answer.
2.3 Solution of the bootstrap equation
The standard method of solving Eq. (2.33) is by Laplace trans-

formation. We define a four vector
. (2.34)
I J
and the two Lorentz invariant functions:
(2.35a)
(2.35b)
Applying the operator
(2.36)
to the whole BE (2.33), we obtain
(2.37)
£f(s) is a well-known function; Eq. (2.35b) gives
(2.38)
Define G('IJ by
(2.39)
then Eq. (2.37) can be written
(2.40)
and the problem is to invert this equation, that is: to find

G{f) = ~(S). The easiest way to do this is a graphical solution by
first plotting ~(G) and then consider the curve as G(,) • By
expanding exp (G) one sees that ~(G) = G + ••• ; with growing
G the exponential function takes the lead and Cf(G) goes expo-
nentially to -00. The maximum lies at Go = In 2 and has value
Cfo = In 4 -1; Cf"{Go):I 0 (Fig.2.la). The graphical solution
is presented in Fig. 2. lb. From the figure and CP" (G 0) of 0, it
follows that GC,) has a square root branch point at ~o and is
complex for ~ > ~o 10). We note for later use that to
~o = In 4 - 1 corresponds a value So for which
G (4')
1.2 ----..--_...
----
--- .............
08
"
,---------------
,
G(%)=ln 2 =0693
0.4 G (4') J physical
G
(a) (b)
Fig. 2.1 Graphical solution of the bootstrap Eq. (2.40)

(a) CI = <f'(G) ; (b) G = G(tf).
= (2.41)
As ~ (B) increases monotonically with liB, it follows that So

is the lowest value for which a real solution ~(B) of Eq. (2.37)
exists. With the knowledge of ~(B) our problem is, in principle,
solved, since the wanted mass spectrum follows by inverse Laplace
transformation
(2.42)
We can obtain some information about T(p2) even without explicitly

inverting Eq. (2.35a); we rewrite this equation by inserting
1 = fdm 2 OO(p2 - m2 ):
(2.43)
As we have just seen, G(<f) has a square root branch point at

Cf 0, and so has cp(S) at
So, since Cf is monotonous in S.
Since Kl (mS) behaves like exp (-Sm) for m -+ 00, Eq. (2.43) can
yield a singularity of cp (S) only if T(m 2) grows asymptotically
like exp (Som); a square root branch point requires
(2.44)
A more familiar expression for the mass spectrum is p(m) (cf. the
text and the equations from (2.7) to (2.11)), related to T(m 2 ) by
T(m 2 ) dm 2 = p(m) dm:
(2.45)
p(m) starts with 8(m - m) and, according to Eq. (2.44) behaves

'IT
asymptotically as
rv (2.46)
For later we note that, had we not absorbed the p2 factor into
T(p2), but defined T(p2) to be the mass spectrum, we would have
obtained
(2.47)
In Eqs. (2.44), (2.46) and (2.47) the value of So is independent

of the definition of the mass spectrum, while the constants and
higher order corrections are not. One can calculate the constant
as well as higher corrections to T(m 2) by expanding G(,) at the
branch point; one obtains
(2.48)
t/) -1
Rewriting this expansion in terms of S and applying ~ to it
is straightforward but tedious. The result is found in Ref. 14).
While this recipe yields still only an asymptotic expression for
T(m 2 ) -- which is, however, rather good down to almost m -- an
1T
exact expression for small p2 = m2 has been given by Yellin l3 ) in
terms of a series expansion of G(,) at ~ = 0:
(2.49)
Therefore GC,) is the generating function for the Yellin coeffi-

cients gn which are very interesting numbers by themselves: n!gn
is the total number of different ways to cluster n elements. The
gn as well as our Eq. (2.40) were already discovered in 1870 by
E. Sch ro··der 19),•
Using Eq. (2.49) we can invert the Laplace transformation

directly, because we know the original function ~ -l,<S) from
Eq. (2. 35b). Write
WI
C(>(~) -
(2.50)
The expression in brackets is the n-pion 1MS integral
(2.51)
U -1
Applying N to Eq. (2.49) yields therefore
(2.52)
The 1MS integrals are well-known functions, for which powerful com-
puter programs exist. Therefore, Eq. (2.52) is very useful at not
too large p2, since the sum has actually only a finite number
of terms -- it is cut off at n ~ Ip2/mn by the momentum o~ func-

tion and by the condition P o -> m7T •
Had we used IMS measure in Eq. (2.6), the density of states of

the pion gas would have read
(2.53)
while now
(2.54)
One sees that the rapidly decreasing lIn! has been replaced by the
(exponentially increasing!) g. Thus the Q in Eq. (2.53) have
n n
been multiplied by n!g, which is the total number of possible
n
ways to cluster n pions; this factor represents the whole boot-
strap dynamics.
With the help of Laplace transforms one can easily prove a

recurrence relation for Q (p2,B):
n
(2.55)
for any choice of nl, n2 ••• n~ such that L~=l ~ = n. Insert-

ing the expression (2.54) into the BE (2.33) and using the recursion
relation (2.55), one obtains a similar recursion relation for the
20)
gn
(2.56)
This equation determines the g uniquely. For practical ca1cu1a-

. n 21 ) .
tions another recursion re1atLon LS more useful:
(J -
(flk -
(2.57)
Given Eqs. (2.48), (2.49) and (2.57) the bootstrap equation (2.33)
can be considered as solved.
3. THERMODYNAMICS
We have solved in Section 2 the BE with the help of the Laplace

transformation. The same mathematical procedure is used in statis-
tical thermodynamics to obtain the partition function from the den-
sity of states. This coincidence has had the effect that the Lap-
lace transformation ~(S) of the mass spectrum ~(p2) and the
Laplace transform Z(S,V)
of the density of states of a thermody-
namical system containing particles with the mass spectrum T(m 2 ),
have sometimes been confounded. We expect a relation between ~(S)
and Z(S,V) -- and we will exploit it below -- but conceptually

these two quantities are different.
3.1 The partition functions of the one-component ideal gas
Consider an ideal relativistic Boltzmann gas with one sort of

particle of mass m enclosed in an arbitrary, macroscopic external
volume Vex. From Eq. (2.4) the number of states in {p, d"p} .of
].l
one particle in the four volume Vex is
].l
(3.1)
From this, the one-particle partition function (the super-

script denotes "non-interacting") is defined by
(3.2)
Here the four volume Vex is an arbitrary external parameter (a box

II
of arbitrary volume having an arbitrary four velocity) while before,
in Section 2, we took the volume to be the dynamically determined
proper co-moving volume of the particle.
Thus, in Eq. (3.2) Vex is a constant with respect to the

p integration. Furthermore, a has from now on the physical
II
meaning of the inverse temperature four vector:
r
-
(3.3)
where the Lorentz invariant T is the usual temperature (remember:

the Boltzmann constant k = 1) in the rest frame of the thermo-
meter.
z~O) is by construction a function of the invariants a 2 , v!x

and allV ex • As it seems not very useful to consider a description
II
where the thermometer moves (fast) with respect to the container of
a gas, we take here all parallel to Vll :
.
}
(3.4)
where u~ ~s the common four-velocity of the thermometer and of

the container (we drop from now on the subscript "ex" of the
volume V of the container).
We then obtain in the common rest frame of S~ and V~
(0)
ZA (~,V) (3.5)
(Compare this to Eq. (2.38) which followed from taking V~ paral-

lel to p~, while S~ was arbitrary.)
From the one-particle partition function the N particle

partition function is found
(3.6)
The grand canonical partition function*) is
with A being the fugacity. From z(O)(S,V,A) nearly all rele-

vant quantities can be found by logarithmic differentiation, in
particular
*) We keep the bad habit of physicists to use the same symbol for
mathematically different functions; the information is con-
tained in the arguments.
-~ ~ A.,Z(9)(A VA)
V 'b~ \-, , energy density
pressure (3.8)
particle number density
and so on. We introduce the relativistic chemical potential (equal

to ~non-rel + m) by
'A- (3.9)
~ = 0 (A = 1) corresponds to black-body radiation of quanta with

rest mass m.
3.2 The strongly interacting pion gas
A gas of strongly interacting particles (of one kind*» en-

closed in an arbitrary volume at arbitrary temperature and chemical
potential may be described
either as a gas of one single sort of particle with a complica-

ted interaction
or as a non-interacting gas consisting of an infinity of

different species with a mass spectrum appropriate to
the interaction in question.
This statement has been discussed in great detail in Refs. 1)

and 2).
*) The generalization to several different species is straightfor-

ward.
What it claims is that if the mass spectrum of the interac-

tion is known, replacing the interacting one-component gas by an
ideal infinite-component gas and weighing the different components
according to the mass spectrum, generates the same distortion of
phase space as the interaction would do. An example is, for in-
stance, a dilute He gas; usually it is not described as a gas
of protons, neutrons, electrons with a Hamiltonian containing QED
and strong interactions; instead one uses the mass spectrum (here
essentially one state with mass, spin etc. of He 4 ) and calculates
the properties of an ideal Bose gas of He atoms, considering the
latter as elementary.
Accordingly we make here the hypothesis: should the mass

spectrum of strong interactions be known, it could be used to arrive
at a statistical-thermodynamical description of strongly interacting
particles in terms of an ideal gas of infinitely many components.
The above hypothesis that the mass spectrum adequately represents
the otherwise unmanageable interaction is used, not only in dealing
with a strong interaction gas, but the SBM which finally yields the
mass spectrum is built itself on this same hypothesis. Taking now
the attitude that SBM has provided us with the correct spectrum,
the corresponding statistical thermodynamics of the strongly inter-
acting pion gas follows by simply generalizing the formulae of the
ideal gas given in Subsection 3.1.
The one-particle phase space measure (3.1) becomes now the

"one-fireball" phase space measure
(3.10)
Accordingly we find the "one-fireball" partition function
(3.11)
We do not need to find 1(p2) by an explicit inverse Laplace trans-

form from ~(S) as given by Eqs. (2.42) and (2.43), because Zl
can be computed directly from ~(S). Recall Eq. (2.35a):
(3.12)
Obviously
(3.13)
Our postulate that V~ and S~ should be parallel allows to go to

their common rest frame and rewrite Eq. (3.13) as
(3.14)
We know from Section 2 that no real ~(S) exists for

-1
S < So ~ ~ (see Eq. (2.41» and therefore no real Zl(S,V)
exists for S < So. Thus
(3.15)
is a critical temperature.
We do not specify now what happens at To, because that will de-
pend on the power of m in front of the exponential of ,(m 2 )
(see Eqs (2.44)-(2.47) and Table 3.1). We shall come back to this
in Subsection 3.3.
In Section 2 we also mentioned another possible definition of

the mass spectrum (text following Eq. (2.31», where the quantity
(3.16)
obeys, as before, the BE, but where T(m 2 ) is taken to be the mass
spectrum. Because of the identity (3.16), ¢(B) is the same func-
tion as before:
(3.17)
but now the partition function
(3.18)
obeys a different relation with ¢(B). We find it by inserting
1 == ) Oo(p~M4't)dltU.t.
in the integrals of Eqs (3.17) and (3.18) and integrating over p:
(3.19)
(3.20)
Inserting the integral of the formula 22 )
(3.21)
into Eq. (3.20), we find by comparing with (3.19)
(3.22)
Comparison of Eqs. (3.22) and (3.14) shows how important the proper
definition of T(m 2 ) is (cf. discussion after Eq. (2.31». Indeed,
Eqs. (3.22) and (3.14) lead to different physical behaviour near the
critical temperature To. We recall that GCY) = $(S) has a square
root singularity: $(S) = $0 -const·(S - So)%. Therefore, Zl(S) ~
~ const·(S - So)-%, Zl ~ const - const·(S- So)3/2. The correspond-
ing energy diverges in the first case: E(S) ~ const·(B - Bo)-3/2
I
and remains finite in the second: E(B) = Eo - const·(B _Bo)1.
We now return to the discussion of the thermodynamics follow-
ing from the choice of T(m) made in Section 2, Eq. (2.32): from
Zl(S,V) we find the N fireball partition function
-I
N'
(3.23)
and the grand canonical partition function

(3.24)
The relations (3.8) now become
£(~/V, ~ ) - -& ~ ~Z (~.V.;\)

P(~,V/~) - IV ~2 (~, V, r. ) (3.25)
tn(~,VrA) -~ 1.'R,u Z l~ V A)
V ra" 'I
Here, however, n(B,V,A) is the average number of fireballs present.
For this n(B,V,A) we have the ideal gas equation (due to the
linearity of In Z in A)
P = m·T (3.26)
while the corresponding equation in terms of the average number of

pions (contained in all these fireballs together) would look horribly
complicated. This result (3.26), which in the framework of this
model is exact, shows once more how simple things become once the in-
teraction is hidden in the mass spectrum.
At this point we can ~eneralize from the pure pion Boltzmann

gas to a hadron gas with correct statistics: we have to replace in
Eq. (3.11) T(p2) by the full hadronic mass spectrum and build in
Bose and Fermi statistics. We first rewrite Eq. (3.11) in the

S,V rest frame inserting 1 = f OO(p2 - m2) dm 2
(3.27)
Assume now we had solved a BE for Bosons and Fermions, then we

would have obtained two changes:
i) the mass spectrum is split into a Bose part and a Fermi part
(3.28)
ii) the logarithm of the grand canonical partition function

Eq. (3.24) will then become 8)
""Z( ~,V,,,) ~
cIT r r ~ r (",,)-~~ (tJt)lt1ktl(2.(M~*,)tLtt

2.'lt't ) tIt:, th,1
p
L\s )F U
(3.29)
for details see Refs. 8), 10) and 17).
This equation can serve to illustrate the influence of

Bose-Einstein and Fermi-Dirac statistics on SBM: in the above
version it was shown8 ) that
(3.30)
Furthermore, with the asymptotic behaviour of K2(X) it fol-

lows that 8 )
i) only the first term of (3.29) has a singularity at So.

This is the first singularity encountered if S decreases
from +00;
R. HAGEDORN, I. MONTVAY AND J. RAFELSKI 87
ii) the integral over the rest of the sum (n =2 ••• 00) is
finite even if the first term becomes infinite at So.
Now it depends entirely on the power of m in front of the

exponential in Eq. (3.30) , how the first term of the integral
(3.29) behaves for S -+ So: it is finite at S~ if the power
"a" of m in the spectrum (3.30) is larger than 5/2.
Thus it follows that for a ~ 5/2 the first term alone

need be considered when T -+
-
To. As this first term repre-
sents Boltzmann statistics, this means that for a ~ 5/2 sta-
tistics can be ignored at T -+ To, while for a > 5/2 it may
not. In contrast to the Boltzmann partition function (3.24),
for which In Z(S,V,A) is linear in A, the Bose + Fermi
partition function (3.29) has a more complicated A depend-
ence. Therefore in this case the ideal gas equation (3.26) is
no longer valid (except for a ~ 5/2 and T -+ To). If P is
taken as the mass spectrum (Eq. (2.47)). we obtain a = 4 and
statistics becomes relevant at To (though perhaps not import-
ant numerically).
3.3 Physics near To
We have seen how the BE provides us with the function ~(S)
from which Zl(S,V,A) = In Z(S,V,A) can be calculated; Zl(S,V,A)

then serves as the generating function for physical quantities. In
all versions of SBM we find an exponential mass spectrum
.(3.31)
with To of order m. While the small variation of To ~ m from

TI TI
version to version is of no physical importance. the nature of the
system, when T -+ To, depends critically on the power of m in
Eq. (3.31). We study this now by using Eq. (3.27) to determine
how the type of singularity at To depends on the power "a" of
the mass.
We replace in (3.27) T(m 2 ) dm 2 by p(m) dm:
(3.32)
As we are interested in the behaviour at T + To (S + So) we

denote all quantitities which are constant in this limit by the
symbol C (at each place where it occurs, C may have a differ-
ent value and/or dimension). Using Eq. (3.31) and the asymptotic
formula
(3.33)
we obtain
(3.34)
M is a mass large enough to justify the use of the asymptotic

formulae (3.31) and (3.33), the +C stands for the non-singular
integral from m to M. With
1T
~-~.= ¥ o
~ C tr;-T) = CAT (3.35)
Eq. (3.34) can be written as
+C
(3.36)
As we know that the integral converges at 00, we expand the exponen-

tial and obtain
(3.37)
Thus
(3.38)
The case a = 5/2 yields (see Eq. (3.36»
(3.39 )
I
In the following table we list the most interesting quantities for

a = 1, 2/2 ••• 8/2, namely P = presure, n = fireball number den-
sity, = energy density, OE/E mean relative fluctuations of E,
E
Cv = dE/dT = specific heat and v 2 = square of velocity of sound.
Our new hypothetical interpretat~on of BE's has added the

cases a = 2 and a = 4 to the previously studied interval
5/2 ~ a ~ 7/2. In fact, the discussion following Eq. (2.31) sug-
gests that it is possible to split the quantity B(m 2 )T(m 2 ) in any
way we like into two functions of one of which we call "mass
spectrum". T(m) and T(m) corresponding to a = 2 and a =4
respectively, are extreme choices; hence 2 ~ a < 4 seems now a
reasonable range of physically possible a values. With the
exception of the velocity of sound, which is independent of
"a" when T -+ To, and the specific heat, for which we ahTays
find Cv -+ 00 when T -+ To, all other quantities cover the whole
range from non-singular to singular when "a" changes in the in-
terval {2,4}. The lesson is that much depends on the seemingly un-
important power of m in front of the exponential mass spectrum.
Table 3.1 8
a P n e: oe:/e: Cv = de:/dT v2 = dP/de:

.r
C C C C
1/2 L\T2 L\T2 L\T3 C + CL\T !:f.T'+ CL\T
C C C C + CL\T3/4 C CL\T
1
L\T3!2 L\T3!2 L\T 5 !2 L\T7!2
3/2 C C C C + CL\Tl/2 C CL\T

L\T L\T L\T2 L\T3
::l :tJ
C C C C + CL\Tl/4 C
~ 2 CL\T :J:
L\Tl!2 L\Tl!2 L\T 3!2 L\T 5 / 2 »
C>
m
"0
t1 To To C C g
CD 5/2 C In L\T C In L\T C CL\T :tJ
....<: L\T L\T2 ,2
0 :-
~
CIl C C C s::
..... 3 Po - CL\Tl/2 no - CL\Tl/2 CL\T o
'< L\Tl!4 L\T3/2 2
III
L\T U2 -I
CIl <
CIl »
C C -<
~
CD
7/2 Po - CL\T no - CL\T e:o CL\T »
p. L\Tl!2 L\T 2
c
f-
::l :tJ
e:o - CL\Tl/2 C C »
~ 4 Po - CL\T 3 / 2 no - CL\T 3 / 2 CL\T "TI
L\T3!4 6.Tl/ 2 m
r
en
"
An important new feature for SBM is our Eq. (2.26) which tells
us that the energy density in fireballs,
:: ..!. (3.40)
A
is finite, constant and of the order of the rest-energy density of
a proton. Therefore it occurs to us (in agreement with the contri-
bution of W. Nahm to this volume) that it is not reasonable to ap-
ply the thermodynamics derived from the BE beyond the point where
the energy density E{T) becomes larger than l/A (~ 10- 3 GeV~).
Furthermore, one may argue that the choice of T{m 2 ) as the mass
spectrum (cf. Eqs (3.16) - (3.22» is physically more consistent
than our conservative choice L{m 2 ), because T{m2 ) leads to
a =4 and therefore to a finite energy density at T ~ To (see
Table 3.1) in accordance with the above Eq. (3.40).
In the next section, where we consider nuclear matter, we

come much nearer to the choice T{m2 ). The growing volume will
be an essential ingredient of the nuclear matter BE and the whole
picture will, therefore, be more internally consistent.
4. NUCLEAR MATTER BOOTSTRAP
4.1 The nuclear bootstrap equation
According to the aims described in the Introduction, we now

generalize the BE (2.20) to the case of nuclear matter. We postu-
late the following BE for the level density of "nuclear clusters"
with baryon number b (-00 < b < 00):
(4.1)
Equation (4.1) is not a single BE, but a member (with baryon number
b) of an infinite set of coupled integral equations, each having
its own input term. The (k!)-l is necessary for correct count-
ing. The non-vanishing pion and nucleon mass ensure that for any
finite p2 the set (4.1) has only a finite number of equations:
Ibmax I -< Ip2/m.
p Therefore the solutions for any finite p2 can
(in principle) be built up iteratively by starting with
4m~ ~ p2 ~ m~ and by increasing stepwise this interval to include
higher and higher Ibl. This, incidentally, also allows to prove
that for any p,V,b, Equations (4.1) have a physical solution.
This equatiOl. fulfils the requirements set up in the Introduc-

tion:
1) Conservation of baryon number b and clustering of nucleons:

the baryon number (number of baryons minus number of antibaryons)
is conserved with the help of the Kronecker 0 (b - Lb.) function.
K 1.
The infinite Qet of density functions a(p,V,b) corresponds to the
admission of nucleon clusters with any baryon number b, four mo-
mentum p and four volume V.
2) Nucleon (isobar) excitation and internal cluster excitation:

internal cluster excitation is contained in the p2 = m2 depend-
ence of cr(p,V,b) and single nucleon (isobar) excitation is con-
tained in the same way in cr(p,V,l).
3) Extensitivity of nuclear matter is ensured by the volume

elf-function.
4) Co-existence of a pion gas is contained in the equation with

b = 0 and in all others by the presence of factors
cr(p. ,V. ,b. = 0) on the right-hand side.
~ ~ ~
5) Baryon-antibaryon pair creation (and annihilation) is built

in by allowing -00 < b., b < 00. Then on the right-hand side an
~
arbitrary number of clusters (Ib i ) and anticlusters (-Ib i ) may

occur.
6) "Chemical equilibrium" between all constituents: this is

expressed by the infinite set of coupled integral equations (4.1)
which allows all multibody reactions between clusters Q.:
~
Q., + III .,. ... + Q,.. ~ "" .,. Q'I. .,. ...
U'., of Ll'
Cor. .
compatible with band p conservatiou.
The input terms, except that for b= 0 (pion) and for

b = 1 (nucleon) specify particular features of the model, namely,
i) details of nuclear interaction may be represented by giving

clusters (e.g., alpha particles) a special weight.
ii) The equations (4.1) deal with Boltzmann particles without

charge and spin. Introducing spin, isospin and statistics would
be possible but complicated. We can obtain a similar physical
effect by assigning to an input nucleus of baryon number band
volume Vb a mass ~ which is greater than (b • m ).
p
4.2 Separation of the volume dependence
We follow closely subsection 2.2.2. We introduce
(4.2)
where the function B describes the p2,b dependence of the V·p

~
term, while 1: that of (J (cf. Eqs. (2.22) to (2.24». Now we
rewrite Eq. (4.1), integrate over the IId"V; and require that all
volume 0" functions have the same argument. We find in the pre-
sent case the condition
(4.3)
from which it follows as before that, for all i,
V(~·. ,I..)
, = A 1M., (4.4)
The constant A is independent of i and therefore equal to that

defined earlier (Eqs (2.26) to (2.30». It therefore follows that
(4.5)
is, in fact, independent of b and the same as the B(m2) de-

fined in Eq. (2.26).
The volume 0" function can now be factored out on both

sides of Eq. (4.1) and what remains is a new bootstrap equation
for the function T(p2,b):
(4.6)
The essential step now consists in the proper extraction of the

mass spectrum T(p2,b) from the function T. We follow here con-
sequently the recipe developed in Section 2, Eq. (2.32). ,Thus
for each b we now write
(4.7a)
Introducing for convenience the coefficient Bb through
(4.7b)
we can write the bootstrap equation for the mass spectrum as

The bootstrap equation (4.8) is much richer than that for the
pion gas; we have allowed the presence of arbitrarily complicated
clusters characterized by the baryonic number b.. For b = 0 we
1
have a description of meson - fireballs - but in order to under-
stand these fireballs properly, especially when baryon-antibaryon
clusters are among their constituents we have to obtain a solution
for the function -r for all values of b.
We remark that our definition (4.7b) makes Bb proportional

and in this way leads us nearer to the case B m2 ,
to ~ 'V dis-
cussed in Section 2 after Eq. (2.31), although not exactly to it,
since Mb is the mass of the b cluster in the ground state.
4.3.1 Laplace and L transforms of the mass spectrum
In order to solve the nuclear bootstrap equation we apply the

same methods and techniques as in Section 2. The new task is the
treatment of the b dependence. This is done by defining the
"L transform":
.t)
l [{('-)1: = 2:
J.-= -00
-
'A 4 ~ (' 1 - (4.9)
Hence f~(A) = L[f(b)] is the generating function of f(b)*). We

mUltiply the entire bootstrap equation with Ab and sum over b.
Defining the L transforms of -r(p2,b) and of the input term,
respectively:
(4.l0a)
*) We use the expression "L transform" to stress the formal ana-

logy with the Laplace transform: L is the discrete counter-
part of !t .
(4.l0b)
where
we find that the bootstrap equation takes the form of the pion
bootstrap equation, however, with a much more involved input func-
tion LOA:
(4.11)
En passant we note that this confirms the general bootstrap philo-

sophy that the input function characterizes the "raw material" while
the integral equation imposes the dynamics onto it. The dynamics
should be more or less independent of what the raw material is (but
it will depend on kinematics, statistics, etc.)
Once we solve the bootstrap equation for the function LA it

is, in principle, possible to invert the L transform by integrat-
ing along a circle C around the origin ~n the A plane:
B~?;(f~t. ) '" 2;' ~ -til 8wr}. (~~ A)

(4.12)
c
We need to consider the Laplace transform in p of the function LA
in order to proceed further (compare Eqs. (2.35a), (2.35b)):
(4.13)
(4.14)
These two equations define two basic quantities we will often refer
to further below. In complete analogy to the case of pionic boot-
strap we find now the usual bootstrap equation (see Eq. (2.37»:
(4.15)
For later convenience we introduce the Laplace transform of

T(p2,b)
(4.l6a)
(4.l6b)
The quantity BbTo in Eq. (4.l6b) is the input term of Eq. (4.6).
The relation between ~ and ~b is given by the inverse L trans-
form:
~~(~,{, ) = ~ ~ ~ <P l~, '). ~ (4.17a)
and similarly C
(4.17b)
Here 'f b (8, b) is the term in the Laplace-transformed input func-

tion that is associated with the b nucleon cluster, cf. Eq.
(2.38). Thus we have explicitly
cp(~,~) = L:'~'Cf>",(~,.(,) (4.18a)
<p. (A '-) = C 8 k ~'l. K. ((~~.) (4.18b)

~ rr :tJ. ~ "~M.f,
In Fig. 4.1 we give a short summary of the relations between the
four functions introduced above: L, LA' ~b' ~.
Fig. 4.1 Logical connection between the mass

spectrum L(p2,b) and its Laplace
~) and L transforms.
4.3.2 Properties of the transforms
The bootstrap equation (4.15) for the doubly transformed func-

tion ~(S,A) has a real solution wherever in the (S,A) plane the
input function cP satisfies the condition (cf. Eq. (2.41) and
Figs. 2.la and b)
(4.19)
Thus along a curve S ; f(A) in the (B,A) plane defined as

c c
the boundary of this domain
q>(A
\.- t )
r. )
t
= &c.tt-1 (4.20)
a qualitative change in the behaviour of the properties of nuclear

matter may occur. Quite aside from the physical questions, we have
to ask for a mathematical solution of the bootstrap equation beyond
this boundary line. As we have previously argued by a recursive
argument, a physical solution for T(p2,b) exists for any p2. Our
~(S,A) is the Laplace-L- transform of BbT(p2,b):
which does not exist in this form everywhere in (0 ~ S < 00) E)

~ (1 ~ A < 00). However, once defined in a domain where it does
exist, it fulfils Eq. (4.15) which then permits analytical con-
tinuation of ~(S,A) beyond the limit (4.19) into the whole
(complex S) €> (complex A) domain (see Section 6).
We remark here that the analytical continuation beyond

cpo has never been considered in the case of pionic bootstrap,
since there this limit on ~ led to a limiting temperature;
the energy of fireballs diverged at this point and made a transition

from our world to the new domain impossible. Now the presence of
baryons changes this -- the introduction of A leads to the exist-
ence of a new region with T < To but <f > In 4 - 1 as we will
see in the next section. We will find in our present model again
a boundary T To, at which the energy density diverges -- but
this limit is not at <f = In 4 - 1, except when A = 1.
As long as ~ satisfies the condition of Eq. (4.19) we can

use the Yellin expansion (see Eqs. (2.49), (2.52) and (2.54» to
obtain ¢ explicitly as a function of ~. With
(4.21)
we can find the explicit form of ¢b(S,b) by computing the coef-

ficient of Ab in (4.21) with the help of Eq. 4.l8a)23)
(4.22)
Here the sum over all sets of n. is restricted by the conditions

J
fVL· (4.23)
J
This expansion considers in each term "£" objects consisting of
groups of n. "elementary" clusters with baryon number j, such
J
that the total baryon number is conserved. This power series ex-
pansion in "non-interacting elementary clusters" <f b (B,j) has the
same physical meaning as the expansion of an interacting field in
free-particles -- it is a series of products of free objects
and only the expansion coefficients contain the deeper physical in-
formation. The behaviour of this power expansion illustrates the
various physical regimes contained in the bootstrap equation, which
describes a loosely interacting system at relatively low tempera~
tures and densities where the expansion (4.21) works well, further
a strongly interacting system with dominant clustering, where the
series (4.21) converges badly and finally a qualitatively different
regime, where the Yellin expansion (4.21) does not exist.
With a particular choice of the input function in Eq. (4.6),

namely,
(4.24)
which neglects all detailed nucleon structure and statistics ef-

fects we obtain for the transformed input function ~(S'A) the
expression
Cpo (~ ) = Sl( 2t""! I(~ (~tNl( ) I ~-T(

(4.25)
q>1 (~) =- BNil""! Kllp",.)/~IIf"

Neglecting for the moment the possibLe presence of antinuc1eons
(A» 1f)..), we can make Eq. (4.22) more explicit, since only
j = 0,+1 will contribute
(4.26)
This expression describes the expansion of a b-nucleon cluster

into free nucleons and an arbitrary number of pions.
Upon a further inverse Laplace transform we can obtain from

Eqs (4.22) and (4.26) the mass spectrum (cf. Eq. (2.54) and text
below it). As will be seen in the next section, this will not
be necessary -- we will be able to determine the thermodynamic
properties of nuclear matter alone from the study of the function
~ and its functional dependence on <f as described by Eq. (4.15).
5. THERMODYNAMICS OF CLUSTERED MATTER
5.1 The partition function
Let us consider a cluster with baryonic number "b" enclosed

in an "external" four volume Vex. Then the one-cluster partition
1.1
function is given by Eq. (3.11):
(5.1)
the only change being the dependence of the mass spectrum on the
.
baryon~c number b *) •
When n such clusters are present, but each with the same
b, we find for the n cluster function the usual expression"
cf. Eq. (3.23):,
M.
&~ 2 141,- (~(v (~ )

(5.2)
l(M,~ (~tV,"') ==
*) As in Section 3 we drop henceforth the superscript "ex" on the

four volume.
When clusters with different "b" are present, then we have to

compute the product of the different contributions in Eq. (5.2).
Let us consider the case in which "i" clusters are present. The
sum over all possible partitions of b nucleons into R. clusters
gives us the partition function of b baryons assembled into
i clusters:
(5.3)
The sum is over all partitions of b baryons into i clusters

with n. being the number of clusters having baryon number j:
J
(5.4)
In order to obtain the partition function of an arbitrary num-

ber of clusters having together b baryons, we have to compute in
Eq. (5.3) the sum over all possible numbers of clusters "t",
since each such configuration is possible. This has the net ef-
fect in Eqs. (5.3) and (5.4) that the restriction In. = i is
J
removed:
(5.5)
We have made the constraint on baryonic number explicit.

The grand canonical partition function Z is the L trans-

form of Eq. (5.5):
(5.6)
It is straightforward to carry out the sum over b when Eq. (5.5)

is inserted into Eq. (5.6) and we obtain
All values of n. are allowed and the set {n j ~ a} depends on

J
j only through the fact that there are j members of the set.
Since all j are permitted, the order in which the infinite sum
and product are evaluated is irrelevant, provided that the sum
converges. Under this assumption we obtain
r
.0 110. ot
Z(~(V/))= J= -~ L
0.:0
~~ [AaZ1f,.) (~\Vla)]
el<r! to/) '>.;i Z. ~ (~,

(5.8)
= 41 V ,S ) }
or
~Z(~,V,,.) - Z-1 (~( V, A ) (5.9)
(5.10)
106 R. HAGEDORN,I. MONTVAY AND J. RAFELSKI
Note that the existence of Zl(S,V,A), the one-cluster grand

canonical partition function, is not assured. In fact, often only
the canonical partition function Zb exists, Eq. (5.5). When an
analytical expression for Z can be found then we can recover the
physically relevant quantity Zb by the inverse L transform
(5.11)
We will return to this point in Section 6.
5.2 Partition function of nuclear matter
Thus we see that we need only to compute the one-cluster

grand canonical partition function Zl to determine the grand
canonical partition function Z, Eq. (5.9). This is an easy
task -- we recall the definition of the function ~b in Eq.
(4.16) and find from Eq. (5.1)
in the common rest frame of the volume and the "thermometer". We

recall that S is related to the physical temperature by Eqs
(3.3) and 0.4)
(5.13)
Inserting Eq. (5.12) in (5.10) we find

(5.14)
Would it not be for the b dependence of the function Bb , Eq.

(4.7b):
(5.15)
we would have already the analogue of Eq. (3.14), since according

to Eqs. (4.10), (4.13) and (4.16)
(5.16)
In Order to proceed further we have to make an assumption about

the b dependence of the cluster mass Mb. For the present we
choose to consider the case
(5.17)
where
,
m,
7T
m..
1'1
are the pion and nucleon masses, respectively.
Through Eq. (5.17) we have implemented explicitly the assumption
that the mass of a ground state cluster is proportional to the
baryonic number (this assumption might be given up in more refined
models). We now find for the grand canonical partition function
(5.18)
In order to sum the expression (5.18) using Eq. (5.16) we generate

b- 2 in the sum by a double integral over A:
where the particle-antiparticle symmetry
(5.20)
was used. For the sum in Eq. (5.18) we obtain
The second sum can be written by observing that a contour integral

of Eq. (5.19) yields
In the same way we find an expression for cj>b (S,O):
(5.22)
We insert Eqs. (5.22), (5.21) (5.20) into Eq. (5.18) to obtain the
final result
(5.23)
On first sight it might seem that the pion and nuclear contribu-
tions to In Z are additive -- however, when we recall Eq. (4.22)
we realize that cj>b (S,O) contains already a great deal of nuclear
contributions; the curly brackets describe that part of In Z
which is not generated via baryon-antibaryon creation but enforced
onto the system by a given fugacity A ~ 1. We note that the
multiple integrals in Eq. (5.23) may be reduced using:
(5.24)
5.3 Physical Properties of Nuclear Matter
Given the grand canonical partition function Z(S,V A) we

want to obtain the quantities of physical interest for nuclear mat-
ter. The energy density, pressure and baryon number density are
respectively
t(~IV,) ) = - ~.t~ . . Z (~,",? ) (5.25)
'P(~IVIA ') = ~ ~Z(~IV,~) (5.26)
0/ =: ,,(~,\I,).) = ~~ iuZ (~IV, ~ ) (5.27)
Of further physical interest are the energy per baryon
~(~,V/').)
(5.28)
"(~IV, ~)
and the relative fluctuations of the energy density and the baryon
number density
(5.29)
!
~ (& V \):
~ "t Y2=
7
'I: ( - )
'\) (I' ~'2. (5.30)
All these quantities are directly related to ~(S,A) through

Eq. (5.23). As we know that ~(S,A) is singular along the curve
Sc = f(A)
c given implicitly by Eq. (4.20), we expect that this
curve separates two regions of different behaviour of nuclear
matter. In fact, ~(S,A) and In Z(S,V,A) are complex beyond
this curve and Eqs. (5.25-30) make no sense there. On the other
hand, it will turn out that in approaching this curve (except at
A = 1) the energy density remains finite so that it is physically
NUCLEAR MATTER AND THE STATISTICAL BOOTSTRAP MODEL III
possible to cross this border line (in contrast to the border line
e = Bo whose crossing would require infinite energy). We there-
fore expect that the line ec = f(A)c is a critical curve separa-
ting two different phases of nuclear matter.
5.4 A Speculation about the Early Universe

Of particular interest is the case in which no baryons are
brought in (A = 1); then the total baryon number (number of
baryons minus number of antibaryons) is zero.
Only in the canonical description (5.5) can this be achieved

rigorously by putting there b = O. In the grand canonical de-
scription =0 is obtained by setting A=1 (i.e., ~ = 0).
Indeed, from Eq. (5.18) it follows that
(5.31)
since corresponding terms with band -b cancel each other. In

this case one cannot use Eq. (5.30), since = O. However, we
may evaluate hI - b 2 = hI; from (5.18) we have
(5.32)
For S + So the derivative of the sum diverges, as can be seen

from Eq. (4.22), because for all b, -a~b(s,b)/as is positive for
S + So and the derivative -a/as of each single contribution
to it in the Yellin expansion is positive (see Eq. (4.18b».
Therefore, if we can show that the derivative of some contribution
to ~b(S,l) diverges, we have proved that b 2 diverges, all con-
tributions having the same sign. We pick out the contribution to
~b(B,l) coming from the subset {n.} = {no = £ - 1; nl = 1; all

other nj = O} and we call this co~tribution ~~l)(B,l). Hence
lO
= ~.(,. (~11) + contributions of remaining subsets
(5.33)
One easily sees that the above subset of the sum (4.22) yields
(5.34)
Here G(~) is the bootstrap function (2.40) which has, for

B + Bo, infinite derivative (see Fig. 2.lb). Hence -d~~l)(B,l)/aB
and therefore the whole sum (5.32) diverges. Thus
- (5.35)
(5.36)
We now apply our model to describe a charge symmetric early Uni-

verse. We then have = 0 at all times and temperatures, but
just after the first ~ 10- 23 sec the temperature will be near To
and then, in spite of = 0, we will have + 00 in any
finite volume. Therefore there will exist regions with large 
z ±1 which will never completely disappear upon expansion and
cooling down. The mechanisms for preserving such fluctuations,
once they are there, have been discussed in the beautiful and im-
mense work of Omnes and co-workers 24 ); we present here a non-
perturbative mechanism for generating such large initial fluctua-
tions in a charge symmetric big bang.
5.5 Nuclear Matter Without Baryon-Antibaryon Pairs
While the general formula (5.23) describes all possible cases

of cold and hot matter, low and high baryon number density, we will
be interested fro~ here on in properties of bulk nuclear matter:
that is, the case when a certain number of nucleons is already
present in a given volume. Unless T ~ To, we expect only moder-
ate contributions from baryon-antibaryon pair production, since
~ »T o • Therefore we further simplify our result and neglect
antibaryon production. We can implement this by restricting b
to be positive in all previous formulae. We note that in do~ng so
we allow uncompensated baryon production which is, for T ~ To, a
small effect, since ~/To ~ 7.
The bootstrap equation is then as before
1 (5.37)
but the input term that describes only "raw" pions and nucleons
takes the form
(5.38)
The relation between the grand canonical partition function and ~
is given by Eq. (5.18)
(5.39)
The sum in Eq. (5.39) can now be obtained by integrating from

zero to A:
(5.40)
In the next sub-section we illustrate our model by some numerical

results obtained studying this equation.
5.6 The Different Phases
In the remainder of this section we will study the physical

properties of the model defined above. We begin by considering in
more detail the point ~o = tn 4 - 1, where the function G(~)
has a square root singularity. This point corresponds to a curve
A = f(S) in the (A,S) plane, defined implicitly by the equa-
c c
tion
Thus (with i = TI,N)
(5.41)
it "1
As in Section 3, Eq. (3.9), we introduce the chemical potential II
by
'). = e ~r (5.42)
~ [GeV]
1.0
0.8
0.6
III
0.4
0.2
o 50 100 150
T[MeV]
Fig. 5.1 The critical curve ~c = ~c(Tc) in the ~T plane se-
parating the gaseous phase (I) from the "liquid" phase
(II). The dot-dashed line would be the critical curve
if pions were excluded. Region (III) is inaccessible
(T > To): infinite energy density. For T = 0 the
critical chemical potential equals the nucleon mass;
note that this is not its maximum value.
and consider the function

c
= f(T),
~
c
where T = S-l that fol-
c C
10w8 from Eq. (5.41). As shown in Fig. 5.1 this line divides the
(~, T) plane into two parts. For ~ < ~ C (T)
c
we have Cf < Cfo
and we know that the grand canonical description is valid there.
At ct = Cfo we are on the critical curve corresponding to a
singularity of £n Z. We record the interesting behaviour of
~c(Tc) for small Tc (large Sc)' This can be found analytically
employing the asymptotic expansion for the Hankel functions
.1L e-i.
K'Yl.(~) = \JJf [
1 + -q. ",,2:., + -.. ]
(5.43)
2~ g i.
We find, using Eqs. (5.41)-(5.43), the relation for large Sc :
(5.44)
We see that as long as ~c - ~ «mn , the last term in Eq. (5.44)

is negligible compared to the first term. This is just the case
for large Sc (small T ), since when neglecting it we obtain
c
solving Eq. (5.44):
rlA,.. . -- ~ t.I - ~2. Tc. ~(~ IT')

c
(5.45)
Depending on the choice of A as discussed in Section 2 we

find T' ~'15-40 MeV. From Eq. (5.45) we see that ~c increases
c
as a function of T initially, until Tmax = eT'. Beyond this
c c c
point it drops continuously until ~
c
= 0 at Tc To '" 149 MeV.
We note that the behaviour of the chemical potential for T ~ To

is similar even when the pion term is switched off entirely
(dashed line in Fig. 5.1). This is visible in Eq. (5.44), where
the last term is almost everywhere small, except when 13
c
'V 11m1T •
The limiting temperature To is now a solution of the equa-
tion (5.41) with AC 1. However, since the nuclear term (~)
is exponentially small at 13
c
~ 11m1T ~ lITo we expect that the
limiting temperature is little changed from that of pionic boot-
strap. The change of To induced by the possible baryon produc-
tion is obtained by expanding Eq. (5.41) around So. We find
-4 -I
As ~S is positive, ~ 4.2 x 10 MeV , the change of To IS
negative: the limiting temperature is slightly lmvered (by about

11 MeV).
There are three domains shown in Fig. 5.1. In domain I

enclosed by the function ~
c
(T)
c
the grand canonical description
is valid; in domain II, above the critical curve, we have
q > 1n 4- 1, but T < To. Thus this is a new bootstrap region
opened up by the introduction of nuclear matter -- the description
of the physical quantities should be canonical, since the grand
canonical partition function does not exist for If > <.fa. It is
possible, however, to consider the analytical continuation of the
grand canonical function into this domain -- inverse L transform
can then be used to find the canonical quantities. Henceforth, we
will call region I the gaseous phase (because it contains the
region of small density) and region II "liquid phase" [because

it is approached if at fixed temperature the baryon density (that
is A or ~) increases]*). Region III characterized by T > To
is a domain that cannot be reached from the physical phases in
those bootstrap models that give divergent energy density at
T = To. As we have demonstrated in Section 3 an alternative
treatment of the bootstrap model can allow a transition even to
this region -- we will, however, not discuss further this pos-
sibility here.
We cannot exclude that in models with more general input

functions (see Eqs. (4.10) and (4.18» a further phase develops
for large baryon densities. However, this is not so within our
simple model of pions and nucleons where we neglect most of the
details of nuclear structure. In particular, for T + 0 and for
~ corresponding to V ~ 1 we might need more detailed input than
we have considered in the present simplified study.
5.7 Numerical Methods
We need to compute the different derivatives with respect

to S and A of Eq. (5.40). Since ~(S,A) = G('(S,A», we only
need the function G(tf) and its derivatives with sufficient
precision in order to calculate the quantities of physical inte-
rest. This is done by considering the expansion (2.48) of G(~)
at ct, :
(5.47)
*) In Section 6 we shall see that in region II in fact two phases

co-exist: liquid and vapour, in equilibrium. We write "liquid"
for short, but put it within quotation marks to emphasize that
it is not the pure liquid phase.
Equation (5.47) defines the remainder 8G which can be taken to

have the polynomial form
AQ(Cf) = ~ tl$Cpt\\ (5.48)

(\\:'0
Since we know the inverse function Cf = <f (G), we can easily

fit the coefficients a. We find that even for quite small
n
N (= 3) already a very satisfactory result is obtained. This is
partially due to the fact that Eq. (5.47) with 8G neglected is,
in itself, a very good approximation of G since the maximum error
occurs at Cf = 0 and is
Also, at <I = cto the proper analytic behaviour is obtained

from Eq. (5.47) for G(~) and its first and second derivatives.
Thus to one per mille accuracy the expansion (5.47) is already
quite adequate, however we have included in numerical calculations
the polynomial terms, Eq. (5.48), in order to achieve relative ac-
-8
curacy of 10 • Another merit of the expansion (5.47) (5.48) is
its analytical integrability in Eq. (5.40). Thus we have succeeded
in obtaining In Z in terms of known functions. The computation
of the different physical quantities, though tedious, is an elemen-
tary exercise now. The results were obtained and graphically pro-
cessed by the CERN Interactive Computing System SIGMA 25). An in-
dependent check of our calculations has been done with the Yellin
expansion which follows easily from Eqs. (2.49) and (2.57). These
two completely different methods yield the same numerical results
for the critical curves separating gas and "liquid".
5.8 Baryon Density in the Gaseous Phase
We begin the discussion of our results by considering the

baryon nwnber density v, Eq. (5.27) along the gas-"liquid phase"
limi t. As a unit of V we will choose the "elementary" volume of
one baryon, VN =~, as introduced in Eqs. (4.4), (2.26). In
the nuclear bootstrap we fix the value of A by requiring VN to
be the volwne available to a nucleon in cold nuclear matter at the
saturation point. We find
V., :: ~ (ti.2 f_)'3 = q. 1O!t [ae.v - 3 ]

A = . q, . 10 3 [ CeV - 4 ] (5.49)
which is in the range of ~ we have estimated in Section 2 (in

*)
Eq. (2.28) it gives ~ ~ 11.3 m~) . The baryon nwnber contained in
the elementary vo1wne VN follows now from Eqs. (5.27) and (5.40)
(5.50)
We find upon differentiation
(5.51)
At the critical line we just have <f~ + AC CfN = CPo ' so
*) All figures have been computed using the A value of Eq. (2.27).
The value given in Eq. (5.49) would yield To around 200 HeV.
0.6 ~ _ _ _---'-_ _ _ _.l......-_ _ _-..1-_~

o 50 100 150
T[Me'Zl
Fig. 5.2 Critical baryon number per nucleon volume VN
as a function of the temperature. The dot-
dashed line results if pions are excluded. The
unexpected shape of the critical curve is seen
to be due to the co-existence of pions. Region
(I) is gaseous, (II) fluid. For T > To the
baryon density diverges. Note that ordinary
nuclei li~ in the liq~id phase but that at T ~
~ To the gaseous phase is reached at any den-
sity.
(5.52)
The first term is the only one remaining in the absence of pions
and is shown as a dashed-dotted curve in Fig. 5.2. Since for
T ~ To we have ~/T» 1, the asymptotic form (5.43) for the
Bessel function in Gr N can be used to determine v. Therefore
we find
VNV I = - ( I + ~ T Ifit" ). 4l- (5.53)
c.;,. ",;n~
Even including pions this expression is correct for low temperatures
7n,crit «A crit GPN,crit· The complete expression

since as before LD
(5.52) is shown in Fig. 5.2. We see that the onset of the pion compo-
nent lowers the phase transition density, but at high temperatures the
density increases sharply again. We remark that we have computed
the results of Fig. 5.2 once as a limit in the gaseous phase and
separately as a limit in the "liquid phase". The exact agreement
of both computations serves as a check of the numerical procedures
involved.
We notice that for T < To - 0 (with 0 a few MeV) the tran-

sition from gaseous to "liquid phases" occurs always below one (one
baryon per unit volume is by definition the normal nuclear density).
This justifies a posteriori our choice for the names of the different
phases. The sudden rise of the transition density as T approaches
To, if confirmed in more elaborate models, could help to obtain in-
formation about the magnitude of To from nuclear physics experiments.
z
c
nr
m
VNV
»
:IJ
t vNvf
s::
II
»
II ~
J m
0.8 I-t\ O.S :IJ
, .,.,.... ....... ._. __ . j1 »
z
0
-f
:I:
I
~l' -----l~-l--r-rl m
0.6 ~ 0.6 ,!:"el en
-f
--r-~",(J9J 1
»
-f
en
-f
n
»
~,"'- I III r
0.4 W- I 1::>'"'- .'1 I.~ I 0.4 ~1 J.._~ tIl
r'rl!,
f-
" I 0
0
-f
I en
-f
:IJ
I »
."
0.21- / / / I I I J.o~ 0.21-1 I I I L _-.1 s::
0
0
m
r
( 0) ( b)
Fig. 5.3 (figure caption on following page).
~
-
Co)
....,
.".
VNV VNV
./ II
,T=20 MeV II 1.1.= 0.94 GeV /././
O.B T=50 MeV ./ O_B
T=BO MeV ./ I.I.=O.B-GeV .,.",./
./ ,I
0.6 I 0.6
/
0.4 0.4
T= 145 7 T=149 MeV

:::0
0.3 0.2 :::c
»
G'l
m
o
o
:::0
Z
o o 20 30 ;-
P ~0-5 P~ :s:
o
Z
(c) (d) --I
»<
Fig. 5.3 Baryon number per nucleon volume VN in various representations up to the critical curve. -<
(a) against chemical potential with isotherms; (b) against the temperature with V as »
z
parameter; (c) against the pressure with isotherms; (d) against the pressure with V o
~
as parameter. Po = A-I ~ proton rest energy density ("internal proton pressure"). The :::0
dash-dotted line is the critical curve, region (II) the liquid phase. The white lower »"T1
right corner in (c) is due to the impossibility of having no baryons at high temperature m
r
(unsymmetry of our input term). en
A
NUCLEAR MAnER AND THE STATISTICAL BOOTSTRAP MODEL 125
In Fig. 5.3 we show the baryon density in the gaseous phase.

In (a) as a function of chemical potential with temperature being
the parameter (isotherms), in (b) as a function of temperature,
with the chemical potential as a parameter. In (c) and (d) we have
eliminated the chemical potential from (a) and replaced it by the
pressure, Eq. (5.26), in units of Po = ~/VN = A-1 , Eq. (5.49).
Since in Fig. 5.3a ~ = 0 implies a finite baryon density in
particular noticeable for T ~ 120 MeV we cannot find for
T > 120 MeV the corresponding pressure in Fig. 5.3c below a cer-
tain density. This is a consequence of taking a single baryon
(without antibaryon) in the input term (5.38). By this we have
allowed free baryon production without compensating antibaryon pro-
duction. The simple input (5.38) was chosen to reduce numerical
computation: we did not anticipate such a visible effect, otherwise
we would have included antibaryons. We stress that this minor ~n
consistency does not change the main results of the model.
5. 9 Baryon Energy in the Gaseous Phase
The energy contained in the unit volume VN can be obtained

easily from Eqs (5.25) and (5.40):
Both (5.54) and (5.50) are functions of ~ and T and we can

eliminate numerically either one of these physical parameters in
Eq. (5.54) and replace it by v, Eq. (5.50). Since T has
a better intuitive meaning, we eliminate the chemical potential
from (5.54) and consider
(5.55)
dropping henceforth the lower index \I. The results are shown in
Fig. 5.4. Here the isotherms T = constant are shown for VNE
as a function of VNV. We record the nearly linear behaviour (in the
gas phase) of the energy density: E ~ C1 + C2 v with temperature
dependent constants Cl. C2 • We reca1l that for very sma1l V(T) our
neglect of antibaryons is not justified. But above VNV = 0.1 and
T $ 120 MeV our results should be in~ependent of this approximation.
Even better insight can be obtained inspecting the energy per

baryon, excluding the rest mass:
(5.56)
shown in Fig. 5.5. For small temperatures (T < 30 MeV) and den-
sities this should be just the usual 3/2 T, which we actually
find for T = 20 MeV. For higher temperatures, as we can see in
Fig. 5.5b, this is the lower limit of the thermic and interaction
energy E~r. For T = 50 MeV and higher, we have a large pion
component; thus the energy per baryon (total energy divided by to-
tal baryon number), which also includes the energy of the pions,
stays high above the lower limit 3/2 T. We note that our interac-
tion energy is, by definition, positive -- our nuclear mass ~
for the input nucleon should, in principle, include all the. binding
effects at saturation, thus be really - EB• Therefore, at den- ~
sities lower than the saturation density in the gaseous phase, the
thermal energy 3/2 T is the lower limit on the energy per baryon.
Furthermore, we note that within our model the thermic energy domi-
nates the picture between ~ 20 and ~ 60 MeV, at which point the
onset of pion and resonances excitation becomes important.
1.6 II
T=120 MeV
1.2
T= 80 MeV
0.8 T=5O MeV
=20MeV
0.4
a 0.2 0.6 0.8
Fig. 5.4 Energy per nucleon volume VN as a function of

baryon number per nucleon volume. Isotherms up
to the critical curve separating gas (I) from
liquid (II). As the rest mass is included in
the energy per nucleon volume, the lower part of
the diagram remains empty.
to.)
00
e: ~'Kie~ e: ~r [Ge'll
10 1.0
\
\ ,
\
,I,
\
T t/n ..... _~ 0.8
8 \
,,
\
\ (b)
,, '1 \
,, \
6 (a) 0.6
T=140Me~
,,
\ I \ , II
,,
4 \ '-J II 0.4
:JJ
:r::
»Gl
T=50 MeV m
2 0.2 o
o
:JJ
Z
T=20MeV :-
I ~ -- , s:
I I I) I ..
o 0.2 0.4 0.6 0.8 VNv 0' 0~2 0'4 0'.6' 0~8 VNv .. o
Z
-l
Fig. 5.5 Energy per baryon (minus rest mass) as a function of baryon number per nucleon volume. <
»
(a) from 0 to 10 GeV - (b) from 0 to 1 GeV with isotherms up to the critical curve sepa- -<
rating gas (I) from fluid (II). Note that energy per baryon is not the energy per carried
»
z
by a baryon but : total energy of the system divided by total baryon number minus rest o
<-
mass : E/b - mN. At very low temperatures (T = 20 MeV) it is ~ 3/2 T. :JJ
»"T1
m
r
en
A
It is straightforward to isolate the thermic term from

Eq. (5.54). In fact recalling the rules of chain differentiation
we obtain from Eqs. (5.25), (5.27) and (5.40)
£= - -\? (5.57)
The first term expresses the pion-nucleon interacting component

and, as discussed in a previous subsection, is small at tempera-
tures below 60 MeV. The second term is just the "free" nucleon
term at density v, which in the non-relativistic limit gives
us the usual 3/2 Tv.
5.10 Baryon Pressure
We turn now to a brief discussion of the pressure of the

nuclear matter in our model. In Fig. 5.6 we show the same results
as in Fig. 5.3c, but now P as a function of the inverse baryon
density for fixed temperatures T 20, 50, 80, 100, 120, 140,
145 MeV. For the interpretation of this figure we note that from
Eqs. (5.18), (5.26) and (5.27) it follows that P = f(\,T) and
V/(VN<b» = g(\,T) so that in this figure we plot f(\,T) against
g(\,T). We may view this as a P-V diagram with fixed {,T}
or as a p_-l diagram with fixed {V,T}. Adopting the first
interpretation we see that P falls with increasing V (as ex-
pected), but not like l/V (as for the ideal gas), since that
would require that g(\,T) f(\,T) be a function of T alone in
contradiction to Eq. (5.18). The absolute magnitude of pressure
rises very rapidly as we approach the critical temperature, in ac-
cordance with the conclusions that can be drawn from Table 3.1.
On first sight, more surprising are the results presented in

Fig. 5.7, where the energy per baryon (excluding the rest mass) is
plotted as a function of the pressure at constant temperature.
PUO-5 pJ
T=145 MeV
20
T=140 MeV
15
II
10
T=120 MeV
5
T=100 MeV
T=80 MeV
T=5O Me
T=20MeV
0 2 3 4 5 6
llV N V
Fig. 5.6 Pressure against inverse of baryon number per nucleon

volume (i.e., V/«b>· VN): volume per baryon in
units of the nucleon volume). Isotherms up to the
critical curve separating gas (I) from fluid (II).
Po .. rest energy density of proton ("internal proton
pressure") = ~/VN = I/A.
NUCLEAR MAnER AND THE STATISTICAL BOOTSTRAP MODEL 131
E gr [GeV]
T=120 MeV
T=100 MeV
II
2 8
Fig. 5.7 Energy per baryon (minus rest mass) as a function of

pressure. Isotherms up to the critical curve sepa-
rating gas (I) from fluid (II). The energy per bary-
on plotted here is not the energy carried by the
baryon (cf. caption to Fig. 5.5).
These functions decrease as the pressure increases, because we ~n-

. 2bnr t h e energy 0 f t h e p~ons,
1 d ~n
cue . resonances, etc., and express
it per baryon. Consider the isotherm T = 100 MeV: along it, the
pion component remains the same, no matter how many baryons we force
into the system. If we increase the baryon number (by increasing
~), more baryons must share the same amount of pionic energy; hence
the energy per baryon decreases. At the same time the pressure in-
creases, since there are now more particles. Therefore the "energy
per baryon" is a misleading concept at temperatures where pions con-
tribute significantly to the energy density; it should be used only
to characterize the initial state of nucleus-nucleus collisions.
6. THE "LIQUID PHASE"
6.1 The Canonical Partition Function
The grand canonical partition function Z of nuclear matter ~n
the case of the input term (pion and nucleon) defined by Eq. (5.38)
can be written,
(6.1)
Here the functions ~ (S) are defined as in Eq. (5.41)

ex
(6.2)
while the modified functions Gik)(CP 1, cp 2) are defined as
aLt\Cf),-= d,t a(If) (6.3)

~'f"
and
at)
(~)
(ql}<f.,..1:=
(IL) (6.4)
c; ". . (If.) <ft ):=
o
From the definition in Eq. (6.4) it follows immediately that
'Oatlr.} ('"et,)
I(
'bf,
. - Ci, q
, (Il)
Ie.. , (CP,) eft )
a~ltl (~, <f 1

(6.5)
Ck)(
~G k "1.l fl"J '= ) to
~Cft
The pion bootstrap partition function Z (S,V) in Eq. (6.1) does
'!T
not affect any of the following considerations. The corresponding
contributions for the thermodynamical quantities can be simply added
at the end to the ones obtained from Z'!TN. Therefore, in what follows,
we shall consider only Z'!TN containing the nucleons and the '!TN inter-
action.
In order to study the thermodynamical properties of the "liquid-

phase" we have to abandon the grand canonical description because
of the singularity corresponding to the gas-"liquid phase" transition.
We can go either to the grand microcanonical or to the canonical en-
semble with fixed baryon number. (The procedure is rather similar to
the one in Ref. 26), where the Bose-Einstein condensation was con-
sidered in the grand-microcanonical ensemble.) We choose the canon-
ical description which is simpler here (but the grand-microcanonical
description is also possible and equivalent).
The connection between the grand canonical and canonical par-

tition function is (L transform and its inverse)
Z1f~ (~,V, A) :- to A Z1fN (~,V,.(,. )

oG "
(6.6)
We introduce as integration variable AGPN instead of A and write

the second relation in the form
tU~a;h'/'I! )+,/.,ifrr a~h,lf,..m

where
r (11f, ~N ,rioN ,'1.1( i ),): '" tkf

(6.8)
ol l<"(~-N) rJ. 1Mt~ 1(1.( ~ (t(1l)
1
0- 0 I_
11- K,(~I\U.N) J 1('0-

~N K,l~~t(
The integration in the A plane goes, in principle, along an
arbitrary curve encircling the origin. In practice it can be easily
evaluated if, e.g., there is a saddle point of the integrand on the
real aXiS. In the gas phase the saddle point exists and we can pro-
ceed according to the well-known Darwin-Fowler method which can be
found in textbooks (see, e.g., Ref. 27). Therefore, we only state
the final result without going into details. In the thermodynamical
limit V + 00, b/V = V fixed we have
(6.9)
where A A is the saddle point determined by the equation

s
(6.10)
This equation has a solution only if the baryon number density V

is smaller than the critical value V (6). Otherwise (i.e., in the
* and the relation in
"liquid phase") there is no saddle point
Eq. (6.9) is no longer valid.
The thermodynamical quantities can be obtained from the canon-

ical partition function as
( = - ~~~ .G...Z (~,V, ~)

Jv lM. Z (~,
(6.11)
p~ =- VI -6- )
Using Eq. (6.9) in the gas phase and adding, of course, the pion
gas contribution coming from Z , the same equations of state can be
1f
derived for the gas phase as in the grand canonical ensemble (cf.
Subsections 5.5 to 5.10).
6.2 The Canonical Partition Function Z1fN in the "liquid phase"
Let us turn to the "liquid phase". As can be seen from Eq.

(6.10), if the baryon number density V approaches the phase transi-
tion value v*(6), the saddle point at A = AS goes to the singularity
in the A plane at
(6.12)
For V > V (13) (i.e., in the "liquid phase") the saddle point dis-
appears. *As Eq. (2.48) shows, G(<p) has a square root cut in the
complex A.. plane extending from 'lo to 00 This cut generates a
cut in the A-plane extending from AO(13) to 00
In this case we can integrate along the curve C shown in

Fig. 6.1. The integral splits into the contribution from the cut
and from the circle.
IrnA
Re A
Fig. 6.1 Path of integration for Eq. (6.7) when the

saddle point has disappeared at Ao(13).
Along the cut, the relation G(~*) = G(GP)*, which follows

from Eq. (2.40), yields the discontinuity relation
~
£.--+0
a (Jr.)
k (AiLE.) = (6.13)
The functions H~k)(A' ~ 2) and h~k)(A' GP2) defined by Eq. (6.13)

can be expanded in a power series in the variables [A - AO( GP 2)J and
[A - AO( q>2)]1/2, respectively. Using Eqs. (2.48), (6.4), (6.5) one
finds
We shall see below that this is all we need in the thermodynamic

limit (V 7 00, b 7 00, V/b = v).
. (k) tD (k) tD
In terms of the functlOns HK (A, 'f 2) and 11K (A, '1 2) we now
obtain from Eqs. (6.7) and (6.8) by integrating along the curve C
shown in Fig. 6.1 the following expression:
(6.16)
>c elcrt ~ [o(N J./~h,% ).f ".,lf -H~ I~, !PI!' >JJ )( lf
l< Ju. { *' ["II

N
{:ll'Alfw ) +0111 Cflf tli I).,'f. )] + l ~
fArfb..,
Below, we shall show that the contribution of the circle can

be neglected in the thermodynamic limit (V + 00, V = b/V fixed),
therefore we consider only the cut where we write
(6.17)
therefore
AI
Z'lfN(~IVI~ )= "N(~1~,r
rD d.~ .&.,.'(')., "VN )
i ~ ): e... "
~o(~) (6.18)
X IM.{k [olN()(l,lf'. )+o<.CPrlO;(~,!f1 ~ }

From the study of the behaviour of the functions Hik ) and hik ) near
the tip A = Ao(S) of the cut we can infer that f(A, VVN) is a de-
creasing function of A near Ao(S); hence if b + 00, only the tip of
the cut is important (if the circle is put not very far from the tip).
As a consequence, we can use the approximations in Eqs. (6.14) and
(6.15) to evaluate the integral. After a somewhat long but straight-
forward calculation we get the simple result
~~~N(~IV,~l ~ "t..-l ctii~ )+~~~J~/V/~. ;:~~~)

(6.19)
(Note that in the thermodynamic limit finite powers of b can be and,
in fact, were neglected in ZnN') Comparing Eqs. (6.19) and (6.9) we
can see that the canonical partition function in the "liquid phase"
is the same as in the gas phase; only A has to be replaced by the
s
V independent value Ao(S).
6.2.1 The Contribution from the Circle
In order to really establish Eq. (6.19) we have to prove yet

that the circle contribution in Eq. (6.16) can indeed be neglected.
For convenience let us rescale the variable A in the function r in
Eq. (6.8) and introduce
(6.20)
with 0(8) > 0 arbitrary for the moment. We have, of course,
The singularity of the function r in the A plane is at
(6.22)
We shall show that in the "liquid phase" the circle can be

omitted in the thermodynamic limit assuming that the saddle point
contribution is dominating the corresponding integral in the gas
phase. (This is sufficient as the Darwin-Fowler method is well
established in the gas phase.) Let us choose a baryon number den-
sity va in the gas phase such that
(6.23)
where a(E) is such that for Va the saddle point is at
(6.24)
Later on we let € go to zero [hence a(€) + OJ. For the density va

we choose the integration path Cva + Sva going across the saddle
point, whereas for the density V + V (S) we use C + S (see
* V V
Fig. 6.2).
1m }.,
-----4------4------±~--;_~--~~--------~Re}.,
Fig. 6.2 Paths of integration used in the proof that the

contribution of the circle Cv can be neglected
in the thermodynamic limit.
We know from the dominance of the saddle point in the gas phase
that in the thermodynamic limit
S(... )
CUI
/ ~
.s "«1
( ... ) --7> 0 6.25)
Here and below ( ... ) stands for the integrand ln Eq. (6.21) belonging
to the appropriate density (va or v). On the other hand, the above
explicit calculations show (see Eqs. (6.9) and (6.19) that in the
/~
case cr(S) = AO(S)-l we have
( ... ) 1 (6.26)
S~
At the same time
-- (6.27)
The last strongly oscillating factor cannot be cancelled as V + 00,
therefore Eqs. (6.25)-(6.27) imply
~( ... ) / ~( ... )-? o (6.28)
Cv Sv
This proves the dominance of the tip of the cut in the "liquid
phase".
Equation (2.40) implies G(A + 4rri) = G(A) + 2rri. Hence in

addition to the cut shown in Fig. 6.1 there are further cuts from
AO + k·4rri to IJO + k·4rri; k = ±l, ±2, For b + 00 these cuts
cannot contribute, since they are far away from the paths shown in
Figs. 6.1 and 6.2 (note: 4rr/Ao ~ 30).
6.3 The Equations of State for the "liquid phase"
Using the expression obtained in Eq. (6.19) for the canonical

partition function of the "liquid phase" it is easy to derive the
equations of state from Eq. (6.11). Let us introduce the notation
v*(S) for the critical baryon number density, E*(S) for the critical
(relativistic) energy density and P (S) for the critical pressure.
*
In terms of the functions G(k) we have:
K
"- (A)'V =I('L(~"N) r CPl () (Il) ~ (1\)+ VI - : I(L(~-r) CifJ)(~C~)cr. (rt)'

• ,- N K, (~tMN1'"". •r , r ,- ImN 21'~_1I" I 0 ,. , I I· ~
1(1.c~"N) i-[k'1l~"'.)J2 (I) r; ~,

E.~ (~) ="", (~),., N K. (~fM ..) +V1('M1f 2.&1 ~"I G LC,w (~}J +
Mt'4 k1.l~411II) ~(R.tN )C1(f)fcp' (11)' +
+ 11 2.1t 'I. ~ 1M1(r I L I( ,. ~
A'l('~ (R) cp'1~)'!!!! [1('1(~"II) )_((('tl~tu')J2]

+ '1. 0 \- ) I.,· 'J V" K. (p...)
R(B""
r tI 1(1 C~ .... ) t
"lll(~ (~) rr. (A);'\V",,: kt l ~"1) Kt.(@ ..... )

t \..( I • ,.. , It ,. 'J VN 21t' ~"" 11 K.~'*N) +
+Q~(~(~)cpR(~) V.4Il: 1(.&£(3"....) RCp-l!) +

I V."~ -21t~ ~1M(1I"
t(l'&)().C~)(P,(B)' V;~! IIC'l~tM.))2. .

'Z.l O\" I « r'} VN ~N l2Jt1 ~~M J
1?lR) RV = K2.l ~ 'M.) r(P'(A (Il) CP (A)) of

.at r \'" N I< f (~IWM N) Y 1 • \- I 1( ,.
+ V, ..: K~( f.-rr) tfl(').!~l If, (1\)'

1m" 21\'l~lMr) 1 0 ,I,""J+
trl1t v.N2Itt~')IM1t
I(~(~".)C:) &.II!.~
Ir'!j
(6.29)
Here the function
'Q(>c) : (6.30)
was also used which actually gives the average energy per particle
in a relativistic ideal (Bo1tzman~ gas by mR(mS).
From Eqs (6.19) and (6.11) it follows that in the "liquid

phase"
r1~)" £ (M+~-'J (I!,)I flit! Ic'l(~"".\ VI"': kl(~"") 1

• '\-~ L AI K. (~CMN) 21l1~o(t) ~-I
(6.31)
For the derivation of these expressions the differentiation rules

in Eq. (6.5) have to be used. From Eq. (6.31) we see that in
the "liquid phase" the critical gas characterized by ",*(S), £*(S)
and P*(S) co-exists with arbitrarily large nucleon clusters res-
ponsible for the phase transition. The average energy per baryon
number in the large cluster's component is
(6.32)
The large cluster's phase ddes not contribute to the pressure, hence
P is constant along the isotherms (and equal to the pressure of
the critical gas). Hence, P*(S) is the vapour pressure when the
nuclear gas and its condensate co-exist. What we have called
"liquid" turns out to be really a co-existence of two phases. We
understand fully the pure gaseous phase and the state of two co-
existing phases, but not yet the pure "liquid". For this the volume
of the clusters must be introduced also on the thermodynamical level.
The property P = P* and the coincidence of the isotherms with the
A = constant curves in the "liquid" phase (A = AO(S» is also
• • d •
true f or t h e Bose-E1nste1n con ensat10n 26) • This shows a structural
similarity between our "nuclear liquid" and the Bose-Einstein con-

densate. The difference lies in the internal excitability of the
large clusters manifested, e.g., in Eq. (6.32), while the large Bose-
Einstein clusters are in the ground state (w 11q
" = m = the mass of
the boson).
7. SUMMARY AND OUTLOOK
We have proposed a bootstrap model for the description of

nuclear matter at finite temperature. We bootstrapped the level
density (related to the mass spectrum) of nuclear matter. We con-
sider the latter to consist of arbitrary clusters of: nucleon
clusters and their excitations, of pions, of mesonic and baryonic
resonances, and of the corresponding antic1usters. In a simplified,
approximate version of our model which contains only Boltzmann pions
and nucleons as elementary building blocks, we compute n~erica11y
the physical properties of nuclear matter. In this model, there is

only one free parameter A taken to be equal to the inverse rest
energy density of the nucleon.
In studying this model we obtain a non-perturbative, exact

solution (describing actual qualitative physical properties of nu-
clear matter) despite the high and ambitious aims outlined in the
Introduction. We consider our present calculations nevertheless
more as an exploratory study than as a final result. However, some
of the general features we find are rather model-independent and
should survive further elaborations:
1) Considering the grand canonical partition function, depend-

ing on the chemical potential and temperature, we find three dif-
ferent situations:
a) a gaseous state (containing for ].l + 0, T = 0, the empty vacuum),

and characterized by the presence of easily moveable but strongly
interacting nuclei and pions, all in arbitrary states of excita-
tion;
b) a "liquid" phase at larger baryon densities, and

c) a supercritical (unphysical) region above T = To 150
MeV, where the energy density becomes infinite.
2) The transition to the "liquid" phase occurs at about 0.65-

0.75 of the normal nuclear number density and at finite energy
density, except when T approaches To, where the pure gaseous
phase persists through high density and where the energy density
becomes very large. We stress once more that what we have called
throughout this paper the "liquid phase" is really the co....,existence
of two phases, vapour and liquid, in equilibrium. This is obvious
from Eq. (6.31) which says that the pressure depends only on the
temperature but not on the density, just as it is usual for the
vapour pressure. We are presently unable to describe the pure liquid.
3) In our actual description we find a limiting temperature

To ~ 150 MeV. At this temperature the energy density diverges.
We have noted, however, that this is a subtle point which touches
on the limits of the validity of our present interpretation of the
mass spectrum. In this respect we recall that the volume of fire-
balls grows now with the fireball mass thus the average density
should be finite for T + To (see end of sub-section 3.3). In
further elaborating the alternative description mentioned in
Section 3, we would have found a model with a finite energy den-
sity at To so that the presently forbidden region beyond To
would have become accessible, though only in a microcanonical, non-
thermodynamical description.
4) Above T = 20 MeV but below 60 MeV we find that the

energy per baryon obeys roughly the simple relation ~ 3/2 kT.
Below 20 MeV our model includes too little nuclear structure to
have enough predictive power; above 60 MeV pion degrees of free-
dom absorb an increasing amount of the total energy so that the
"energy per baryon" (= total energy/number of baryons) exceeds more
and more the energy which the baryons carry themselves.
Perhaps the most important aspect or our work at the present

level is the better understanding of several conceptual aspects
new to the bootstrap model, namely:
a) the definition of the mass spectrum, as the level density

o(p,b,V) with the co-moving volume vllp;
b) the connection between the bootstrap function ~ (obeying
the bootstrap equation) and 1n Z (describing arbitrary
macroscopic systems);
c) the now possible existence of a physical world for finite
energy densities beyond E(To).
Looking ahead, we hope to enlarge our model by making the in-
put more elaborate, by maintaining the particle-antiparticle sym-
metry and by considering the particular position of alpha clusters.
It seems that a more profound study of the "liquid phase" will be
rewarding since much of the structure of the liquid (maybe even
the existence of a new "solid" phase) depends on the amount of
nucleon structure we include in the input terms. An obvious first
step in this direction is the possible introduction of effective
masses « free masses) of the bound nucleons, a feature very
likely relevant to the understanding of the saturation of nuclear
matter in the bootstrap description. We must also incorporate
Fermi and Bose statistics and investigate the version leading to
a finite energy density at To.
REFERENCES
1) R. Hagedorn, Cargese lectures in Physics, Vol. 6, Gordon and

Breach, N.Y. (1973).
2) R. Hagedorn, "Thermodynamics of Strong Interactions", Lectures

given in the Academic Training Programme of CERN 1970/71.
CERN Yellow Report 71-12 (1971).
3) H. Koppe, Phys. Rev. 76 (1949) 688.
4) E. Fermi, Progr. Theor. Phys. 5 (1950) 570.
5) B. Touschek, Nuovo Cimento B58 (1968) 295.
6) S.Z. Be1enkij, Nuclear Phys. 2 (1956) 259.
7) Particle Data Group, Phys. Letters 75B (1978) No. lo
8) R. Hagedorn, Supp1. Nuovo Cimento 3 (1965) 147.
9) S. Frautschi, Phys. Rev. D3 (1971) 2821.
10) W. Nahm, Nuclear Phys. B45 (1972) 525.
11) C.J. Hamer and S. Frautschi, Nuovo Cimento 13A (1973) 645.
12) S. Frautschi, Nuclear Phys. B91 (1975) 125.
13) J. Yellin, Nuclear Phys. B52 (1973) 583.
14) R. Hagedorn and I. Montvay, Nuclear Phys. B59 (1973) 45.
15) E.M. Ilgenfritz and J. Kripfganz, Nuclear Phys. B56 (1973) 241.
16) P. Fre and L. Sertorio, Nuovo Cimento 28A (1975) 538.
17) M. Chaichian, R. Hagedorn and M. Hayashi, Nuclear Phys. B92

(1975) 445.
18) A. Chodos, R.L. Jaffe, K. Johnson, C.B. Thorn and V.F. Weisskopf,
Phys. Rev. D9 (1974) 3471.
19) E. Schroder, Zs. f. Math. u. Physik, 15 (1870) 361. See also,

L. Comtet "Advanced Combinatorics", D. Reidel Pub1. Co.,
Dordrecht/Boston (1974).
20) E. Etim and R. Hagedorn, Nuclear Phys. B131 (1977) 429.

21) J. Engels, K. Schilling and H. Satz, Nuovo Cimento l7A (1973)

535.
K. Fabricius and U. Wambach, Nuclear Phys. B62 (1973) 212.
22) I.S. Gradstein and I.M. Ryzhik, Table of Integrals, Series and
Products, Acad. Press, N.Y. (1965).
23) M. Abramowitz and Irene A. Stegun, Handbook of Math. Fcts,

Dover, N.Y. (1970).
24) R. Omnes, Phys. Reports (Phys. Letters C), 3C (1972) 1, and

subsequent work.
25) R. Hagedorn, J. Reinfelds, C. Vandoni and L. Van Hove,

"Sigma, A New Language for Interactive Array-Oriented
Computing", Revised edition by H.E. Rafelski, CERN Yellow
Report 78-12 (1978).
26) I. Montvay and H. Satz, Nuovo Cimento 39A (1977) 425.
27) A. Munster, "Statistical Thermodynamics, Berlin (1969).

THE BAG MODEL AND QUANTUM CHROMODYNAMICS
Julius Kuti*
Laboratoire de Physique Theorique et

Particules ~lementaires, Universite de Paris-Sud
Batiment 211, F-9l405, Orsay, France
1. INTRODUCTION
The purpose of this paper is to discuss some recent develop-

ments in the bag model .from the closing date of a previous article
(1) in April, 1977.
We have witnessed important progress, I believe, regarding the

field theoretic foundation of the bag. In particular, there is now
a serious attempt (2-7) to derive the bag picture from quantum
chromodynamics (QCD) which is the leading candidate for the field
theory of strong interactions. The hope is to fill the gap theor-
etically between the QCD Lagrangian and the rather successful bag
phenomenology.
The key assumption in constructing the phenomenological bag

model of hadrons within framework of quantum chromodynamics is to
visualize the QCD vacuum as a perfect (or nearly perfect) dielectric
substance (1,8). It is suggested, therefore, that a long-range
order, which may be expressed perhaps in terms of the chromoelectric
susceptibility of the vacuum, is responsible for this phenomenon
(2-6) •
In a relativistic theory as QCD, the dielectric constant E

is the inverse of the magnetic permeability, ~, of the substance.
* Permanent Address: Central Research Institute for Physics,

H-1525 Budapest 114, Post Office Box 49.
149
150 JULIUS KUTJ
Consequently, vanishing dielectric constant implies infinite mag-

netic permeability. It follows then from our assumption £~ that
the QCD vacuum is a perfect (or nearly perfect) dielectric and
paramagnetic medium in response to color gauge fields.
Now, by concentration of energy, a small domain (bubble) in

normal vacuum phase (£ = 1) may be created in the medium of the
QCD vacuum where £~. Inside the bubble, £ is nearly one, quark
and gluon fields behave according to standard perturbative QCD
quantized in the small volume, with boundary conditions for the
gauge fields.
Accordingly, the surface of the bubbl~ is impermeable against

the gluon chromoelectric induction fields D (a=1,2, ••• 8) and
against the chromomagnetic fields H. Theyacannot penetrate into
the perfect dielectric and paramagn~tic medium which behaves as an
"electric" superconductor with chromoelectric Meissner effect.
We picture the hadrons as small bubbles of the normal vacuum

phase immersed in the dielectric medium of the QCD vacuum. The
interior of the bubbles is occupied by constituent quark and gluon
orbits. This is the bag.
To create a "normal" bubble in the medium requires an amount

of energy p per unit volume, and an amount of energy S per unit
surface. Here, p may be associated with the QCD vacuum pressure
exerted on the bubble by the medium. The pressure of the constit-
uent quarks and 9luon fields from inside is balanced by the QCD
vacuum pressure p and surface tension S.
The quark bag model is the invention of the ingenious M.I.T.

group who introduced the vacuum pressure p to stabilize hadrons (9).
The surface energy S per unit surface area was introduced by a
group of us in Budapest (10). Any decent field theoretical form-
ulation of the bag generates such a term (2-5).
It is important to note that a closed hadron domain of finite

energy immersed in the perfect dielectric medium is always colorless
in the terminology of QCD. In the limit £~, the energy of a color-
non-singlet domain becomes infinite. The bag binds quarks and
gluons into observable color-singlet hadrons.
When we closed our review article (1) for Physics Reports in

April, 1977, we said: "We do not attempt to derive S or p from
some microscopic structure of the physical vacuum in gauge theories,
though it did not escape our attention that the above-discussed
physical picture may be related to instantons in QCD or to other
vacuum phenomena." The present report is concerned with some of
the new developments along this line.
THE BAG MODEL AND QUANTUM CHROMODYNAMICS 151
The notes are based on my Erice lectures in October, 1978.

After I gave these lectures at Erice in a single long run, I re-
peated them in November, 1978 at the Laboratoire de Physique
Theorique et Particules Elementaires in Orsay, and at the Istituto
di Fisica Teorica dell'Universita di Torino in January, 1979.
During these reruns, I made some improvements and changes.
The organization of the paper is as follows: In Section 2, I

describe a recent attempt of Callan, Dashen and Gross (3) to picture
the QCD vacuum as a paramagnetic and dielectric medium. Section 3
is a further elaboration of this suggestion, describing bag for-
mation in the presence of static and pointlike quark sources.
Section 4 is devoted to some problems related to the bag picture
for light quarks. Section 5 comprises my conclusion and outlook.
The Appendix is an elementary introduction to instantons (11).
2. THE QCD VACUUM IN SEMICLASSICAL APPROXIMATION
2.1. pseudoparticles Constitute a Polar Medium
Callan, Dashen and Gross observe (3) that the response of the
pseudoparticle vacuum against color fields can be treated as the
four-dimensional version of the magnetostatics of a polar medium
provided the density of the pseudoparticles is small and they don't
overlap. Following their reasoning, I will often refer to the re-
sults of Appendix B. The reader is advised to go through Appendix
B first.
The vacuum-to-vacuum amplitude, <vacle-HTlvac>, in four-

dimensional Euclidean space for a Yang-Mills theory with SU(3)
symmetry is proportional to the classical partition function Z of
static chromomagnetism, and Z is given by
F~v)
z= J[dA ] exp[- __1__
~ 4l f Tr(F
~v
d 4x] (2.1)
The temperature is in correspondence with the squared coupling con-

stant, 1/g2 ~ l/kT and the energy of a given field configuration is
'41 J Tr (F ~vF ~v ) d 4x •
As we indicated in Appendix B, the saddlepoints of Z are

localized pseudoparticle solutions which are randomly distributed
in Euclidean space according to the grand canonical partition
function of a gas. With a cut-off on the scale size of pseudo-
particles, the density of the gas is less than one and they don't
overlap.
152 JULIUS KUTI
The effective coupling, g(Ap) also is reasonably small with

the same cut-off, so that perturbation theory can be applied around
the saddlepoints in the functional integral (2.1). Z then becomes
the thermodynamic partition function of a four-dimensional linear
static gauge theory in the medium of pseudoparticles (non-Abelian
self-coupling is small). In the statistical physics of the system
we have to include the pseudoparticles, their pairwise interactions
and the linearized pseudoparticle-gauge field interaction.
The Polar Medium. Dipoles in four-dimensional space are

characterized by anti-synunetric tensors. The long-range field of
an instanton of size p is the same as that of a colored, permanent,
magnetic dipole whose dipole moment is
Da (2.2)
).1\1
There are independent dipole moments for each component of color.

The interaction energy of a dipole D~\I with a weak external field
Ba is
).1\1
£ (P , R) (2.3)
The interaction energy, £(P,R), depends upon the size of the in-
stanton and its orientation R in the color space.
The interaction energy of two dipoles is easy to calculate

from Equation (2.3) if B~\I is replaced by the long-range dipole
field of the instanton. The chromomagnetic field of an instanton
(anti-instanton) is self-dual (anti-self-dual). The corresponding
dipole moments D~\I have the opposite self-duality properties.
Therefore, two self-dual (or anti-self-dual) dipoles do not inter-
act. This is why instantons do not interact with instantons, only
with anti-instantons.
The above picture developed by Callan, Dashen and Gross is

equivalent to a gas of permanent magnetic dipoles in four-dimensional
Euclidean space. The particle number is not fixed, and we have a
grand canonical partition function Z in Equation (B.17). If the
gas is not dense (f < 1), the interaction between an instanton and
anti-instanton is of dipole-dipole type with a long range l/R~
where R is the separation of the pseudoparticles.
2.2 Large Permeability
A gas of permanent magnetic dipoles has a positive suscep-

tibility, so that the permeability is greater than one. The di-
polar medium in an external field may have a net dipole moment
density with a color magnetization tensor ~\I' The chromomagnetic
induction field
1 a
'2 e: llvaS F as
is the microscopic magnetic field whose sources are the external
sources and the dipoles.
Only the external sources will create the chromomagnetic field
(2.4)
When external sources are not present, the chromomagnetic field H

and the color magnetization M vanish. For weak fields we write
Ba II Ha
llV llV
where X is the chromomagnetic susceptibility and II is the chromo-

magnetic permeability of the medium.
Let us calculate the permeability ll. In a weak external field,

a
the energy of the instanton is given by Equation (2.3). The
B llV '
Bo1tzman factor for the instanton which depends on the orientation
is
exp[e:(p,R)]
P (p ,R) (2.5)
f [dR] exp [e: (p ,R)]
where e:(p,R) is taken from Equation (2.4). The mean dipole moment
is
< D~v > = J[dRab ] P(p,R) D~v (2.6)
with an integration over the orientations of the dipole in color

space.
From an evaluation of Equations (2.5) and (2.6) in a weak ex-

ternal field we can calculate the net magnetization of the ins tan-
ton gas
(2.7)
where n(p) is given by Equation (B.21) for a dilute gas. In this

approximation
154 JULIUS KUTI
llD.G. 1 + 47T2X
D.G.
1+47T2 Jp5dp D(p) 87T 2
g2p,p)
£.4
8
, (2.8)
or
I
00
7 -x
llD. G. 1 + const. dx x e (2.9)
x
c
where the critical lower limit x of the integration is about x ~ 14
at which point the instanton den~ity becomes large. The corres~on
ding value of llD.G. is about 11, indicating a strongly polar medium.
However, in such a strongly polar medium, the energy of a di-

pole is determined by the local microscopic field (B~v)local which
is different from the weak external field (B~V)ext. Onsager's
treatment (16) of polar dielectrics can be adapted to four-dimen-
sional magnetostatics (3). The method treats the dipolar feedback
interaction correctly. The net result is a permeability II which
is given by
II llD.G. - 1 + !(llD.G. -1) 2 + 1 (2.10)
where llD.G. is the result of the naive calculation in Equation (2.9).
The new value of the permeability is about II = 20, a very

strongly polar medium indeed. There is a strong belief, however,
that a more sophisticated calculation which treats the breakup of
instantons into merons (2) at larger densities, and perhaps sig-
nificant quantum effects (6) and non-classical meron field con-
figurations (7) actually will make II diverge leading to perfect
paramagnetism. How it happens is not yet well understood. Never-
theless, the pseudoparticles, when the gas is not dense (f < 1)
already produce a large permeability and perhaps the precise
mechanism which makes II actually divergent is not of utmost impor-
tance for phenomenological applications.
2.3 Second Phase by Color Magnetostriction
In the above estimate of the permeability, we have taken for

n(p) the perfect gas density (B.2l). The pairwise interaction of
pseudoparticles and the presence of external fields will change
the perfect gas density. The refined treatment, together with an
interesting signal for a first-order phase transition in the
chromomagnetic susceptibility as a function of the strength of the
applied external field, will be sketched at the end of this sub-
section.
Here we only note that in the undisturbed vacuum state, the

density of pseudoparticles and the permeability (large or infinite)

are uniform. External sources (quarks) will change the density and
permeability from point to point. Though the spatial variation of
the pseudoparticle density would increase the free energy, the gain
in the electrostatic energy of the chromoelectric field of the
quark sources from the corresponding variation in space of the per-
meability is larger.
This observation is very important for the ded vation of the

bag picture from QeD. Around quarks, a finite region of space
(domain) is created (Figure 1) where the pseudoparticle density is
nearly zero and the vacuum permeability is approximately one. There
is a sharp transition to dense vacuum phase at the boundary of the
domain.
Figure 1. The quark-antiquark pair through its chromoelectric

field creates a normal vacuum domain in the nearly
perfect paramagnetic and dielectric medium.
It is easy to see the consequences of baglike spacetime var-

iation of n(p) and~. For simplicity, we choose a static bag with
a time-independent boundary. The fluctuating linearized field-
strength tensor F~v in Minkowski space is related to the chromo-
electric field E~, and the chromomagnetic induction field Ba by
E~ = Fa. and B~ ~ ~£ •• k Fa].k. E1 and BaJ.' are the microscopic
-1- oJ. J. J.] a
fJ.elds. The intensity tensor G~v is determined by the external
sources only and related to F~v by
a (2.11)
F~v
156 JULIUS KUTI
The chromoelectric induction field Da and chromomagnetic field Ha

are given by D~ = G~i and H~ = ~Etjk Gjk The color dielectric
constant E and chromomagnetlc perm. ]1 are related by E = 1/]1 as a
consequence of Lorentz invariance.
Let us take a static bag with a sharp boundary. Inside, we

take n( p) - 0 and E = 1/]1 ~ 1. Outside we have the (nearly) per-
fect dielectric and paramagnetic medium. It follows, then, for the
normal component E~, and tangential component B~ that
(2.12 )
on the boundary. The covariant form of Equation (2.12) is

F~V nv = 0 for a = 1, ... 8, where nv is the space-like normal to
the bag surface. This is the M.I.T. bag boundary condition (9)
for the gauge fields.
Magnetostriction and Phase Transition. We consider instantons

now in the presence of an external constant color field. For a
static configuration we shall take a chromoelectric field
For not very strong fields, the instanton medium can be treated as
linear.
The corresponding thermodynamic potential (17) depends upon

the instanton density n(p) and external field E:
F (n,E) = F
o + F.ln t - !2 ED (2.13)
where F is the free energy of the perfect gas and F. t contains

the dipgle-dipole interaction energy of the pseudopa}~lcles.
One now minimizes the free energy F(n,E) in Equation (2.13)

with respect to n for a given external field E. The resulting
density of instantons is
n(p) = no(p) exp [6 x(p) ~:i

(%)4_ TI2E2 p4 x (p) ] (2.14)
4(1+]12)
where n (p) is the density of instantons of scale size p in the
perfectOgas approximation. R is a hard-core cut-off on the dipolar
potential (Figure 2) and the optimal value is determined to be (3):
R _ 2.2p
Our medium is very similar to a van der Waals gas with a repulsive
core and an interaction energy proportional to -1/r4.
.,- ,
I
,.
~,
,,"
--
,.., ,."
Figure 2. The interaction energy between an instanton and
anti-instanton. The hard core at R serves to
leave out overlapping pseudoparticles from the
partition function.
The hard-core cutoff prevents a catastrophic collapse of the

molecules where the significance and well-defined meaning of ins-
tantons would be lost. Particles can be created in the gas; hence,
the grand canonical partition function Z in Equation (B.17).
Equation (2.14) should be solved, together with the formula

for the permeability ~ = lie in the medium:
(2.15 )
~2f d~ x(p) n(p)p4 (2.16 )

2 p
where x{p) = 8TI21 g2(Ap) is the effective coupling.
The dependence of n(p) on E explains the strong reduction of

pseudoparticle density in high field regions. Large-scale size
instantons suffer large reduction because the dipole moment is
proportional to p2. The external field acts as an infrared cut-
off. For large E, the van der Waals gas is dilute and calculable.
Callan, Dashen and Gross solve Equations (2.14) -(2.16) with the
choice R = 2.2p and calculate n(p), ~ and D as a function of E3.
Their result for D(E) is plotted in Figure 3.
The following features of the solution are important. The

density of instantons along the curve in Figure 3 is small up to
point 3 where f = 0.15. The external field is not too big to
cause saturation in the induced magnetization. Therefore, the
approximations used in the calculation of the curve certainly are
acceptable up to point 3.
158 JULIUS KUTI
4
tr'flfi~I." ~•
.Ie"". ~ •••
1 , ,
Figure 3. First-order phase transition in external
chromoelectric field.
The following features of the solution are important. The

density of instantons along the curve in Figure 3 is small, up to
point 3 where f = 0.15. The external field is not too big to cause
saturation in the induced magnetization. Therefore, the approx-
imations used in the calculation of the curve certainly are accep-
table up to point 3.
There is an instability starting at point 2, where the instan-

ton gas becomes unstable. This instability signals a first-order
phase transition to a condensed (liquid?) phase. One notes the
close analogy with the equation of state of a van der Waals gas.
At Point 3 the approximations break down and merons take over.

The guess is that ~ will rapidly increase .(and probably diverge)
as indicated by the dotted line. The Maxwell construction tells
us the critical field strength E at which the phase transition
c
sets in:
E '" 61-. 2
c
There are two phases in equilibrium with this value of the field.
The difference in the free energies per unit volume of the

dilute and dense phase is approximately
p (2.17)
where ~1 and ~2 are the permeabilities of the dilute and dense

phases, respectively. The difference in free energy is identified
with the vacuum pressure (or volume energy) p as it was introduced
in Section 1. Since ~1 ~ 1 and ~2 ~ ~, we find
p = 18 A4 (2.18)
which is the the energy density required to create the abnormal

(dilute) phase.
The estimate of p is, of course, only qualitative. What seems

to be likely is the existence of two phases: a dilute phase with
small coupling (hadron phase) and a dense vacuum phase with ~ ~ ~
(~ 20) •
3. HADRONS WITH HEAVY QUARKS
3.1 Bag Formation in Color Fields
Let us consider a color singlet of a very heavy quark-anti-

quark pair separated by a distance, R. The color fields of the
point-like heavy quarks will expel the instantons by magneto-
striction frOm some domain. A bag will be created in dilute ins-
tanton phase.
The problem simplifies when R becomes very large. In this

limit, the flux tube in dilute phase becomes a cylinder with a
quark and antiquark at the two ends (Figure 4). In our linear
approximation, the quarks will be treated as static Abelian sources
of charge square 4/3 • g2 where the factor 4/3 comes from the color
singlet nature of the state. As we shall see in Section 3.3, this
is just the Coulomb part of the problem.
Figure 4. Elongated bag with static quarks.
The chromoelectric field is constant inside the cylindrical

flux tube with a constant magnitude E ,
c
160 JULIUS KUTI
(3.1)
The diameter d of the flux tube is determined by identifying the

flux 1(4/3)g with lTd 2E /4)1. v.lith)11 ~ I we find
c 1
I (3.2)
d '" 0.4 I
The chromoelectric flux gives a linearly growing interaction

energy between quarks. The energy per unit length along the flux
tube is (1)
E2
7T d 2
E: = 2 2~1 -4- '" 34 A2 (3.3)
if we neglect surface effects. From the charmonium spectroscopy

we know that E: must be equal to 1 GeV/fermi, approximately. This
determines the value of A to be somewhere around 100 MeV.
It is important to note that the bag shields the quarks inside

it from large color field fluctuations in the vacuum. This is like
a chromoelectric Meissner effect. The QeD bag is like a cylinder
of color superconductor which expels the color field.
3.2. Phenomenological Lagrangian
We may adopt a phenomenological description (4,5,8) of the

long-range order of the vacuum through the use of a phenomeno-
logical scalar field o. The dielectric constant E: will be repre-
sented by this scalar field in the simplest form:
__
0_ = 1 _ E:
(3.4)
o
vac
Inside the hadron we want 0 '" 0, and therefore, E: '" 1. Outside the
hadron, 0 = 0 , and E: = O.
vac
The phenomenological Lagrangian density is assumed to be in
Minkowski space (1):
1 a ajJv 1 - jJ+;;+ - 1 jJ a
~ = - '4 € F jJV F + '2 \jJ (y d)1 + im) \jJ - g\jJ '2 Aa y \jJ AjJ
- ~ (djJo)2 - U(o) (3.5)

where E is given in terms of 0 through Equation (3.4). ~ is the

quark field and m is the quark mass matrix. The function U(a) has
an absolute minimum at 0 = a and a local minimum at a = 0 with
vac
U (0
vac
) o (3.6)
and
U(O)=p. (3.7)
The detailed form of U(a) (Figure 5) is not important. It can be

chosen to be a quartic function of a.
Figure 5. The field potential U(o) •
Since a is only a phenomenological field describing the long-

range electric order of the vacuum, its short-wavelength components
do not exist in reality. All a-loops will be ignored, and a will
be approximated by a classical field. It describes the instanton
medium in some continuum limit.
Our computational strategy is best described as follows:

For a given classical a configuration in space, we calculate the
Hamiltonian from Equation (3.5). The Yang-Mills field is treated
in a spatially varying dielectric E. After quantization, we can
calculate any stationary quantum state of quarks and gluons in the
spatially varying dielectric medium. At the end, the energy ex-
pression must be minimized with respect to the classical 0 field
whose quantum fluctuations are regarded as the artifact of the
phenomenological approximation.
The parameters of the function U(o) can be chosen such that

the calculation simplifies. The minimum energy states are almost
sharp bags in the vacuum with E = O. Inside the bag, E ~ 1 and
there is a sudden change across the surface of the bag. To create
the inside phase costs energy p per unit volume according to
Equations (3.6) and (3.7) (also, see Figure 5).
From the gradient term of the 0 field in Equation (3.5), some

surface energy S per unit surface is generated. It describes the
energy of the phase boundary in the phenomenological continuum limit.
162 JULIUS KUTI
The phenomenological bag Hamiltonian becomes a relatively

simple expression,
H = HQCD(E) + pV + SA (3.8)
where V is the volume inside which E = 1. A is the surface area

of this domain. H C (E) is the Yang-Mills Hamiltonian in a spat-
ially varying diel~c~ric where E suddenly jumps from its value of
1 in a closed domain to the value E = 0 outside this domain. The
surface of the bag is regarded as static in this approximation.
The chromoelectric field is tangential and the chromomagnetic field
is normal at the surface of this domain.
3.3 Coulomb Gauge
For applications in charmonium spectroscopy and the T-family,

it is useful to work out the QCD Hamiltonian of the spatially
varying dielectric in Coulomb gauge (5,18).
Let us consider the Lagrangian density
:t' = - 1:. E (x) Fa Fa]lV +:t (3.9)

4 ]lV F
where the Yang-Mills part is given in a static dielectric charac-
terized by E (~). The fermion part:t'F describes the quarks and their
coupling to the Yang-Mills field,
(3.10)
The volume energy and surface energy will be added to the Hamil-
tonian at the end of the calculation.
From Equation (3.9)
so that our canonical variables in the Hamiltonian formulation are
and
a
D plays the role of the canonical momentum of the Yang-Mills fields
cgnjugate to Aa •
]l
We have well-known constraints in the theory. In Coulomb

a
gauge A, can be decomposed into a transverse and longitudinal part.
Introdu6ing the transverse projection operator p, "
~J
P
ij = ~ ", A-I "
Yij - o~u OJ (3.11)
h
were t"-1,~s t h e ~nverse
' 0f th I
e Lap ace tor, we can d e f'~ne the
opera
transverse part of A~ and D~ by
AT, p" A ,
a~ ~J aJ
and
T
D , = p, , D ,
a~ ~J aJ
respectively. The longitudinal part of A~ is set to zero in Coul-
omb gauge. The vafiables Aa, DO, and di t,,-loj Daj can be elimin-
ated in terms o·f D , , AT" 1/J0 anda~ which become our independent
dynamic variables.a~ a~
In terms of the canonical Dirac brackets we have, symbolically,
T
{ AT,
a~, ~j }* 0,
{ DT , T }* 0,
ai Dbj
(3.12)
{ T
Da~" ~j }* Poo,
~J
{ 1jJ ~ }i' =i yO,
and the rest of the brackets are zero. On the right-hand side of
the third and fourth brackets in Equation (3.12), a 0 function, is
understood but not written explicitly.
We now shall give the Hamiltonian in C~ulomb gauge. The

transverse chromomagnetic induction field B ,
a~
T T l T T
Bai = - £ijk OJ Aak + 2 g £ijk fabc Abj ACk '
and the Coulomb Green's function Gab (x) will be used in the Hamil-
tonian. Gab (~, y.) is defined as the inverse of the operator
0i (0ab 0i + g fabc A~i)

164 JULIUS KUTI
With the above notations, the Hamiltonian is
H= f 3{ 1
~
TTl
al. - al. - -
T
al. -
T
d x -2e:() D ,(x) D ,(x) + -2 e:(x) B ,(x) B ,(x) +
al.-
".+
'f'
(~) (-iaV
++ + 8m) ljI(~) + gljl + (~) 2
1 Aa -+
a ljI (~) -+T
Aa (~) +
al.,e: (x)
(3.13)
T -
9 D ,(x)
al. -
e: 2 (~)
where
p a = ljI + 1 Aa
(x) -2 ljI (x) + f b A.T , (x) DT , (x)
- - a c --bl. - Cl.-
is the color charge density. The volume energy pV and surface

energy SA of the bubble is added to the Hamiltonian.
At the end of a calculation for a hadron we have to minimize

the energy of the bubble in terms of all possible shapes for the
surface. Naturally, with an educated guess, we usually start the
calculation not far from the minimum shape.
In applications of Equations (3.13) weThave to impose boundary

c~nditions at the surface of the bubble. D (x) is tangential and
B (x) is normal at the surface, if e: sharp1~ jumps from one to
a -
zero on the boundary.
3.4. Static Bag with Point-like Quark Sources
As an application of the Hamiltonian in Equation (3.13), let

us consider a pair of point-like colored quarks which are coupled
to the non-Abelian gluon fields. With very heavy quarks, their
kinetic energy in Equation (3.13) can be omitted, and they are
considered as nailed down sources to fixed points in space.
It can be shown (1,18) in Coulomb guage that in the lowest

order in the small coupling constant g, the transverse fields can
be neglected and we get an effective Abelian problem for the long-
itudinal field inside the bag:
-+ -+
curl E = 0 , curl B = 0 ,
-+
div B = 0 ,
div -+E = / 4
'3 (3.14)
together with the boundary conditions

+ + + +
n • E = 0 n x B = 0 (3.15)
+
The quark positions are denoted by £1' and £2' and n is the normal
unit vector to the surface of the bag.
The minimum energy shape of the bag can be determined from the
equation
1 +2 1 1
-2 E + S (- + - )
= P (3.16)
Rl R2
at the boundary. Here, l/R, and 1/R2 are the principal curvatures
in two orthogonal direction§ at a given point of the static surface.
In n numberal solution to the problem (1), the quark-gluon

coupling was set at the value
2
L = 0.2 (3.17)
3 'IT
and the strength of surface energy was
220 MeV
s (3.18)
fermi 2
In the numerical example, p = 0 is chosen for simplicity.
The shape of one quadrant of the bag in longitudinal section

for different q-q separations is shown in Figure 6 from the avail-
able numerical solution.
Figure 6. One quadrant of the bag in longitudinal section.

166 JULIUS KUTI
The computer calculation confirms our intuitive picture about

the static confinement force for colored quarks separated at large
distances: a chromoelectric vortex tube develops between the color
charges for large separation. The radius d/2 of the cylindrical
flux tube for infinite s~paration can be calculated with the prev-
iously fixed values of g /3n and S,
d =1 fermi.
The value of d/2 is shown in Figure 6, indicating the rapid con-

vergence of the bag's width to that of the ideal cylindrical vortex
tube with increasing values of r.
This rapid convergence is even more dramatic in the energy as

a function of the q-q separation. It is shown in Figure 7.
The slope of the potential VCr) in the vortex region is 1

GeV/fermi in our numerical example. The ideal chromoelectric vor-
tex solution with linearly rising potential energy, and with the
correct slope, sets in at rather small separation, around r = 0.5
fermi.
VC1')
.2 (GeV]
Figure 7. The static bag energy as a function of the q-q separa-

tion. The solid line is the sum of the chromoelectric
field energy and surface energy.
The potential energy is of coulomb form at short distances

according to our expectations. The form of VCr) in Figure 7 was
used in the calculation of the charmonium spectrum (10).
The qualitative features of the numerical results were con-

firmed by analytic calculations in a two-dimensional bag model (19).
4. HADRONS WITH LIGHT QUARKS
4.1. Goldstone Vacuum in QCD
The applications of the pseudoparticle vacuum of QCD in the

description of light hadrons is more problematic. One has to study
quark propagation in the instanton gas and the related problem of
chiral symmetry breaking. We shall recall here the arguments of
Callan, Dashen and Gross (2,3).
The light quarks have a very small mass parameter in the QCD
Lagrangian (the value of m in Equation (3.10». However, ins tan-
tons generate an effective four-fermion interaction which breaks
the chiral symmetry of the vacuum. This picture as calculated in
Hartree-Pock approximation (2,20) in the instanton vacuum suggests
an effective quark mass m(p) which depends on the momentum.
If the density of instantons is large, m(p) becomes very large

at small p. Inside the bag, the density of instantons is low; we
have approximate chiral symmetry there, and the quark masses are
close to the value of the parameter in the Lagrangian. Outside
the bag, the instanton density is high and a very large quark mass
will be generated. We have a mass bag which confines the quarks
to a region where they are light. This hypothesis has not yet been
proven in QeD.
The boundary condition for the quark field is determined by

the discontinuity of the quark mass at the phase boundary (surface)
of the bag. If we assume mq ~ 00 outside the bag, the M.LT.
boundary condition
(4.1)
on the surface arises with the explicit introduction of chiral

symmetry breaking (21).
If the QCD vacuum phase outside the bag is a Goldstone vac-

uum with zero mass Goldstone bosons (pions), these will be coupled
to the quarks at the surface of the bag, restoring chiral symmetry.
Therefore, we expect modifications in the phenomenological bag
equations, although the details are not yet clear.
4.2 Phenomenological Lagrangian
To be less ambitious, we forget about the problem of chiral

symmetry and try to incorporate mass bag effects (4,5,22) into
our phenomenological Lagrangian (3.9 - 3.10). This can be achieved
by coupling the phenomenological field to the quark fields:
!f = i £ (x) p~\) p a ].!\) +!fp + ib-,

168 JULIUS KUTI
a
E: = 1 -
a
vac
i
2
1jJ (yfl a+
].l
+ i (m+fa)) l/J
- 21 (d].l 0)
2
- u(a) ,
where f is the strength of the quark-a coupling.
Outside hadrons, quarks pick up a large mass fa vac from the

a vacuum. Inside hadrons, where a ~ 0, quarks have their "bare"
masses.
It can be shown that the phenomenological Lagrangian of

Equation (4.2) in semiclassical approximation reproduces the re-
sults of the original bag model to a good approximation.
4.3. Do Confined Gauge Fields Confine Quarks?
The boundary in Equation (4.1) may be interpreted as follows:

Outside the bag, the vacuum acts as a scalar confinement potential
with infinite walls. What happens in the presence of quark-g1uon
coupling?
Let us consider slowly moving quarks in adiabatic approximation

with simplified geometry (1). We imagine the phase boundary to be
an infinite plane at z = 0 (Figure 8), with its normal vector along
the z-axis. We have a semi-infinite dielectric slab with E: = 1 for
z > 0, and E: = 0 for z < O.
''.C,: 0
$.: .... .
: .. .
~f-+---->
.. 1
Figure 8. The semi-infinite slab.
The plane z = 0 is a reflecting dielectric mirror against an

nbe1ian gauge field A (x) with the well-known boundary conditions.
Let the gauge field bM coupled to a quark spinor field that is not
restricted by boundary conditions on the surface. The plane at---
z = 0 is transparent against free quarks in the absence of quark-
gluon coupling. Nevertheless, as we shall see, quarks cannot get
through the dielectric mirror because of the gauge field A dragged
along by the quark color charge. ].l
The motion of a charged-point quark will be studied in the

region z > O. There is an instantaneous coulomb interaction between
the quark and the dielectric. The instantaneous color electric
field which is tangential on the plane at z = Q is calculated by
introducing the image charge of the quark at ~lm (x,y,-z) as shown
in Figure 8. The position of the quark is denoted by ~ = (x,y,z).
The potential ¢(~') of the color electric field is given by
(4.3)
where r' is an arbitrary point in the region Z > O.
The first term in Equation (4.3) gives the divergent coulomb

self-energy of the point-like quark when evaluated at r' = r. It
is independent of ~ and can be absorbed in the mass of-the quark.
The second term, however, gives a z-dependent finite mass term in
the one-particle Dirac equation. This mass term is identical with
the potential energy between the dielectric and the quark. The
force between the dielectric mirror and the charged particle is
repulsive for both the point quark and its antiparticle.
From the solution of the Dirac equation for the point quark
with the scalar self-mass term a/4z, it follows (1) that the quark
is repelled from the E = 0 phase.
In this simple example, the quark becomes confined because the

gauge field is confined.
4.4 Deformation Energy of a Six-Quark Bag
There is a recent attractive application of the bag idea to

the calculation of the nuclear force (23). More precisely, the
deformation energy of a six-quark bag was estimated.
The starting point of the calculation is a six-quark bag with

deformations under the constraint that there are three-quark clus-
ters with the quantum numbers of the nucleons inside the bag. The
results are interesting and encouraging.
When the two nucleons are very close to each other, there is
a repulsive soft core in the interaction. It is generated by the
chromomagnetic gluon force between quarks. At large cluster sep-
aration, we have a deformed bag with increasing separation energy.
The individual nucleons as represented by three-quark clusters can
be easily recognized in this configuration.
DeTar's calculation is the first important attempt to calculate

the nuclear force from the first principles of QeD.
170 JULIUS KUTI
5. CONCLUSION AND OUTLOOK
We have seen that instantons and merons may form a condensate

in the QCD vacuum which eventually develops into a bag-like picture
for slowly varying or static color gluon fields. It remains an open
question concerning what is the influence of light quarks on the
confinement picture. They bring in two problems. Virtual light
quark-antiquark pairs may change the structure of the vacuum quite
significantly; also, it is not clear whether the confinement of
light quarks follows from the confinement mechanism for static gauge
fields in the gluon sector of the theory. The nature of dynamic
cluon excitations remains equally puzzling.
Recently .. another confinement mechanism was suggested in the

gluon sector in terms of a vortex condensation mechanism (24). It
is of great importance to understand whether instanton and meron
condensation on one hand, and vortex condensation on the other hand,
may operate in the QCD vacuum as two independent confinement mech-
anisms (25), or if they have a drastic influence on each other. To
clarify this situation seems to me the most challenging problem for
the near future.
ACKNOWLEDGMENTS
I thank Professor Luigi Sertorio for the kind hospitality ex-

tended to me at the Erice Workshop. Part of this work was prepared
while visiting the Laboratoire de Physique Theorique et Particules
Elementaires in Orsay. I thank Professor J. Tran Thanh Van for his
hospitality and the Orsay theory group for discussions.
APPENDIX A. INSTANTON IN QUANTUM MECHANICS
The Double-Well Potential
Let us consider a potential with two minima and a symmetry

Vex) V(-x) as in Figure 9. To be explicit, we take Vex) =
=
~A2(x2_a2)2. There are now two classical zero-energy configurations,
x = fa, to the Hamiltonian
2
H = E.:.
2m
+ Vex)
For an approximate quantum ground state, we may construct two
Gaussian wave functions 1/JG(x-a), peaked at x = a, and 1/JG (x + a)
peaked at x = -a. The width of the Gaussians is determined by the
potential and corresponds to an oscillator frequency, "_ ~ A~2.
Figure 9. The double-well potential.
The Gaussian ,,,ave functions are not parity eigenstates and

we have to take a superposition:
~(x) = C+ ~G(x-a) + C_ ~G(x + a) •
Since ~(x) must be a parity eigen state, c+/c_ = ±l follows. How-

ever, the ground state is not degenerate in this approximation be-
cause of tunneling. The antisymmetric state is higher in energy
by an exponentially small amount.
To see the role of instantons (12) in removing the classical

degeneracy of the ground state and restoring the x ~ -x symmetry,
we turn to the path integral formulation of quantum mechanics. The
Euclidean (imaginary time) sum over histories is given by (13)
N f [dx] e~ S
1
--
en HT \ x.>
~
(A.l)
where \xi> and \xf> are position eigenstates. H is the Hamil-

tonian and T is a positive number (time difference), N is a norm-
alization factor, S is the Euclidean action
~
S f 2
dt [1.m(dX)2+V1 (A.2)
-T 2 dt
2
and [dx] designates integration over all functions x(t) with the
boundary conditions x(-T/2) = Xi and x(T/2) = x f . For large T
asymptotically
N f [dx] (A.3)
172 JULIUS KUTI
where 10> is the ground state and E

o
is its energy.
The left-hand side of Equation (A.l) can be evaluated in the

semiclassical (small n) limit when the functional integral is
dominated by the stationary orbits X,
~~ = _ m d 2 x(t) + V' (x) o. (A.4)

dt 2
Introducing the eigenfunctions, x n ' of the second variational

derivative of S at X,
(A.5)
the functional integral becomes a product of Gaussian integrals in

semiclassical approximation:
[I<!xl ok s ok = S (xl IT >~, - 1:.

en
S (x) 2
[det(-ot+V"(X»]-
~
n
(A.6)
If there is more than one stationary orbit as a solution to Equa-

tion (A.4), each contributes a term like Equation (A.6) to the
functional integral.
Equation (A.4) describes a particle of mass m moving in a

potential minus V. Therefore,
E = ~m (~~) 2 _ V(x)
(A.7)
is a constant of motion. Ebr the double well of Figure 9, we have

the equivalent of Figure 10 for the motion of the particle at a
given value of E.
-v
Figure 10. The equivalent problem for Equation (A.4).

We want to compute
-HT
<al e- HT la> <-a I e I-a> (A.8)
and
<a I e -HT I- a> = < - a I e-HT I a> (A.9)
in the semiclassical limit. Two solutions to the Euclidean equation

of motion, (A.4), with E = 0, are x = ±a. There is, however,
another solution in the T+oo limit,
A
x(t) = a tanh a rm t (A.10)
where the particle starts at the top of the left hill at time -T/2
and moves to the top of the right hill at time T/2.
The solution in Equation (A.10) is called an instanton, or

pseudoparticle (12). It is centered at t = 0, and can be shifted
by any fixed amount of time. There also is an anti-instanton re-
placing t by -t in Equation (A.IO). The energy E of the instanton
L
solution vanishes in the large T limit. The action of an instanton,
+a
So = I dt [~ (:)2 +
m V(i)] = dx/2V. (A. H)
is a finite and well-defined expression. The solution in Equation

(A.IO) is well-localized in Euclidean time, having a size on the
order of l/w.
For large T, there are other approximate solutions which

correspond to a widely separated sequence of instantons and anti-
instantons. Evaluating the functional integral in the semiclass-
ical approximation, we have to sum over all such configurations
with n pseudoparticles centered at tl ••• tn (Figure 11).
Fbr n well-separated pseudoparticles, the Euclidean action is

n so. The contribution of the determinant to the semiclassical
approximation is
1
;::;2 wT
(A.12)
in the T+oo limit. The term Kn in Equation (A.12) comes from the
location of the n widely separated pseudoparticles. The exponential
term with its multiplier in front comes from the fact that V" = w2
everywhere except at the locations of instantons and anti-instantons.
174 JULIUS KUTI
X.
'l
_T
.t -1.4 r ~
I.~ t
) it\S~"",i.,,, QI1 l i i u-l....tort
7
-q,
\.. .z.
Figure 11. Pseudoparticles distributed along the t-axis.
The integration over the location of the pseudoparticle cen-

ters brings a factor (l/nl)T n Finally, we get
1
<a I ~fi HTI a> (~) ~e -~wT \'
1ft'! L
(A.13)
n even
since n must be even if we want to get back at a for the orbits of
the functional integral. For <al exp(-(l/~)HT) I-a>, we get the
same expression as Equation (A.13) except that the sum runs over
odd values of n.
From Equation (A.13) and the odd matrix element, it follows

that we have the two lowest energy eigenstates
1
1 -r S
2~w± hKe l l 0 (A.14)
This is the result expected from the Schrodinger equation. The

energy eigenstates are approximately the spatially even and odd
combinations of harmonic oscillator states centered at the bottoms
of the two wells. The degeneracy is removed by tunneling as des-
cribed by the instanton solution in Euclidean time.
Periodic Potential
Before examining the Yang-Mills theory, let us consider a

periodic potential (Figure 12) in quantum mechanics.
V{~)
Figure 12. Periodic potential.
We now have an infinite number of classical zero-energy con-

figurations x = an, n = 0, ± 1, . . . . Accordingly, there is an
infini te number of G aussians 1/IG (x-an) , each peaked at the zero-
energy configuration x = an. There is a symmetry in the problem:
shifting x by a does not change V(x).
We find a family of states parametrized by S,
inS
I
n
£ 1/I G (x-an) (A. IS )
which possess the symmetry. Their degeneracy is removed by tun-

neling and we get a continuous energy band E(S) parametrized by S.
This is the description of a Bloch wave in a periodic crystal.
Let us investigate how this well-known result of quantum mech-

anics can be obtained by pseudoparticle description in Euclidean
time. As one notes from Figure 13, the instantons are similar to
the previous ones in the case of the double well.
-I/(X)
Figure 13. The equivalent potential.

176 JULIUS KUTI
Instantons can start at any initial position x=na and go to the

next one at x = (n+l)a. Anti-instantons can go from x=na to
x=(n-l)a.
In contrast to the double well, there is no constraint that

instantons and anti-instantons must alternate. However, the total
number of instantons minus the total number of anti-instantons must
be equal to the change in x between the initial and final position
eigenstates in the functional integral.
The result of the summation over the dilute gas of instantons

and anti-instantons is
00 00
1 1
<n+a
-- HT
en n a> (;)2 2 WTL
e-!
\'
I.
n=O n=O
1
n!n!
1
(KefiSOT)n+n
-
. on-n-n +n (A.l6)
+ -
where n is the number of instantons and n the number of anti-ins tan-
tons. It is easy to find the continuum of energy eigenstates
labeled by the parameter S, and with the energy band of a Bloch
wave:
E (S) = -21 n w + 2 n K cos S (A.17)
This is the correct result of quantum mechanics in semiclassical

approximation.
APPENDIX B. PSEUDOPARTICLES IN QCD
Degenerate vacuums and the Schrodinger Equation
Let us consider the pure Yang-Mills theory with Su(3) sym-

metry. FiJ;'st, we shall discuss the model in Minkowski space and
use a Schrodinger representation for field theory. It is like our
first discussion of the double well in real time using the Schro-
dinger picture and coordinate representation.
The Lagrange density is
.!t' = - .!.4 F)l'J F

a a)l'J
(B.l)
where a summation over the color index a = 1,2, ••• , 8 is under-

stood. The field-strength tensor F is given in terms of the
a)lV
field potentials A by
au
F
a~'V
a~ Aa\) - a
\)
A
a~
+ g fabc ~~ A
. c\)
(B.2)
The Hamiltonian density is

1 (~2
71'= 2
+ :B2)
a a
1 ijk
2 E: Fajk
+ +
where E is the chromoelectric field of the theory. B is the
chromom~gnetic field. We shall adopt the guage A~ = O~ temporarily.
+
The wave functional ~[A] in field-diagonal represent~tion de-
pends upon the dynamic variables A of the Yang-Mills theory.
a
The Schrodinger equation is
+
E ~ [A] (B.3)
and scalar products of wave functionals as well as matrix ~lements

of observables are defined by functional integration over A .
a
To study the ground-state wave functional in semiclassical
approximation, we have to find the zero-energy, c-number field
configurations. Zero-energy classically requires 71'= 0; hence,
F~\) = O. Consequently, the gauge potential A ~ must be a pure
g~uge. A~ = 0 is such a configuration, and t~e ordinary pertur-
bation theory above the Fbck space starts from a ground state
(Fbck vacuum) which is a Gaussian around A~ = O.
However, there are other pure gauge configurations with!Jt' = 0

that are separated by a potential barrier from A~ = 0 and also from
each other. They are similar to the local minima of the periodic
potential in Figure 12. To find them, it is convenient to define
a matrix-valued vector field, A~(X), by
A = -ig Aa Aa
~ 2 ~
where g is the gauge coupling constant and the A's are Gell~ann's
matrices. The matrix-valued field-strength tensor F is defined
~\)
by
F
~\)
=a ~
A -
\)
a\)
A + [A , A ] •
~ ~ \)
(B.4)
In matrix notation, a pure gauge is

178 JULIUS KUTI
(B.5)
where g is ~ unitary, position-dependent matrix. The choice

g = I gives A = 0 and~= 0 classically. However, there are
infinitely many gauge copies of this configuration with zero-energy
classically. Some of the gauge copies, which can be joined by a
continuous gauge transformation to the identity, are trivial and
physically uninteresting.
It is important to realize that there are gauge potentials A~

that can be obtained from each other by gauge transformations g(r)
which are not continuously related to the identity transformation.
An example is
(B.6)
where r designates the first three Gell-Mann's matrices. gl (~) in

Equation (B.6) gives a vector potential:
(B.7)
-+
Al(y) is pure gauge, the corresponding field-strengths and class-
ical energy are zero, though we shall see that this configuration
is separated by a barrier from A~ = O. It is another local minimum
of the field potential energy which must be kept in the construction
of the true ground state.
It can be shown that the physically relevant zero field poten-

tial energy configurations comprise a denumerable set classtfiied
by an integer n = 0,1,2, •••• The representative of the n class
is
with g. (r) given by Equation (B.6) The corresponding gauge po-

tentials-are:
In the above classification, we required the boundary condition
(B.9)
The relaxation of (B.9) might bring in only configurations that are

separated by an infinite-energy barrier from the ones in Equation
(B.a) and therefore can be omitted.
+
We now can form functional Gaussians '¥ [Al peaked around each
A. preserving the gauge invariance of the theory, we can construct
anfamily of vacuum wave functionals parametrized by an angle 8,
+ \' in8 +
'¥ e [A]. = L e '¥ [A] • (B.10)
n n
These states are not degenerate in energy, since there is tunneling

through a finite barrier between the local minima A (~). The Yang-
Mills vacuum is a Bloch wave. To demonstrate this,nthe pseudo-
particle solutions in Euclidean space-time become relevant.
The Pseudoparticle and Tunneling
We turn now to the path integral formulation of field theory

in ~uclidean space-time. The vacuum-to-vacuum amplitude, <vacl
e- H Ivac>, in four-dimensional Euclidean space for our Yang-Mills
field is proportional to the path integral
z f [d AlJl exp (B.ll)
in matrix notation. We ignore here some gauge-fixing and normal-

ization problems. Aa satisfies some boundary condition in Equation
(B.ll) when the Eucl~dean time x4 goes to ± ~.
The expression for Z in Equation (B.ll) can be interpreted as

the classical partition function of static chromomagnetism in four-
dimensional Euclidean space. A (x) is then regarded as the spatial
components of a gauge potentiallJin a five-dimensional, non-Abelian
gauge theory (four space + one time). The static chromomagnetic
field in the four-dimensional Euclidean space is defined as
(B .12)
in close analogy with the three-dimensional case.
Ear a more explicit analogy with a thermodynamic temperature

T, the inverse of the coupling constant,
1 1
g2 kT
180 JULIUS KUTI
may be factored out in the Euclidean action S,
S = 2:...2 1:.4
g
I Tr (F
Jl\l
FJl\l) d 4 x (B.13)
The integral may be interpreted as the static chromomagnetic energy

of the system with energy density E(X) given by:
E(X) = -14 (B.14)
We shall say that A specifies a configuration of the class-

ical continuous medium, fn thermodynamic equilibrium at temperature
g2 = kT and described by a partition function Z in a canonical en-
semble.
The statistical weight of a given field configuration A (x)

in our thermodynamic system is determined by: Jl
d P - [dA ll ]
z
In the semiclassical approximation, we may evaluate the func-
tional integral in Equation (B.ll) by saddlepoint integration. The
saddlepoints are solutions to the Euclidean Yang-Mills field equa-
tions:
oS = 0 (B. 15)
oAll
in close analogy with Equation (A.4).
The important solutions to Equation ~B.15) were given by

Belaving et al. (14) in Landau gauge (d A = 0) as
Jl
4n p2
all\)
(B. 16)
where the numerical tensor is defined by
1
nall \I = E
oall\l
+
2
E
abc
E
bCll\l
The solution in Equation (B.16) is called an lnstanton, or pseudo-
particle localized around x = 0 in Euclidean space. There is also
an anti-instanton by the replacement
- 1
n = £ - - £ E
all\l oall \I 2 abc bCll\l
Arising from the scale invariance of the classical theory,

there is an arbitrary scale parameter p in the solution (B.16).
Independent of p, the Euclidean action of the solution is 8n 2 jg2.
It is not difficult to see in the AO = 0 gauge that the in-

stant~n describes vacuum tunneling between A
= 0 and the local min-
imum Al in Equation (B.7).
The Dilute G as of Pseudoparticles
There are approximate saddlepoints to the functional integral

in Equation (B.ll) with the superposition of n widely separated
instantons and n widely separated anti-instan!ons of scale sizes
Pi (i=l ••• "+) ;r Pi (i=l ••• n).
Each instanton (anti-instanton) has 12 collective variables:

four for the position in four-dimensional Euclidean space, one for
the scale size P. (P.) and seven for the orientation within SU(3)
determined by a Unit~ry matrix. There is an integration over each
collective coordinate in the partition function Z.
We speak about a dilute distribution of instantons and anti-

instantons with an extension p. (p.) in four-dimensional Euclidean
space, if the distances betwee~ th~ pairs of pseudoparticles is
much larger than their sizes. In Figure 14, the black spots are
instantons, the white ones are anti-instantons. Vacuum tunnelings
are happening in localized regions of the three-dimensional space
wherever instantons (or anti-instantons) are found.
O. o. ·
• 0 • • •O. 00
0.0 O.
• 0
Figure 14. The pseudoparticle vacuum in four-dimensional
Euclidean space-time.
Fbr a dilute gas of pseudoparticles, the partition function

can be evaluated by saddlepoint integration with the result
+
dP~ +
Z ~
D(p~) , (B.17)
(p~) 5
~
where, using Pauli-Villoirs regulators, we find

8 2 g
D(p)=l.S'lO-3 ( _ n 2 ) 6 e- 8n j 2(Ap) (B.18)
g2 (Ap)
182 JULIUS KUTI
In Equation (B.17), VT is the large volume of the system in four-

dimensional Euclidean space, and g(Ap) is the renormalized value
of the coupling constant.
The functional integral in the saddlepoint approximation car-

ries the standard ultraviolet divergences of ordinary perturbation
theory (one-loop corrections) and requires renormalization. The
end result (2,15) is to replace everywhere the coupling constant g
with the effective coupling constant g(Ap) of the renormalization
group
11 2 (B.19)
1 + 8"ii7 g R-n].lp
Here, ].I is the renormalization mass, and the constant g2 is the

temperature in the partition function Z. The effective coupling
constant g2(AP) equals g2 at an instanton size P = 1/].1. Fbr a
pseudoparticle
1
11 R-n IP (B.20)
The expression (B.17) for Z is identical to the grand canon-

ical partition function of an ideal gas of instantons and anti-
instantons with equal average densities
n(p) = D(p) (B.2l)

p5
for the size of the pseudoparticles between p and p + dp.
We shall estimate now the energy E(8) of a 8 vacuum whose

construction was indicated in Equation (B.10). In the dilute gas
approximation, we have to sum over all widely separated instanton
and anti-instanton configurations. On the basis of Equation (B.17),
the contribution of instantons and anti-instantons of size p to the
energy is:
HT _8TI2/g2
<8 I e - I8> -+ E (8) = EO - 2VD cos 8 e (B. 22)
where EO is the zero-point energy of the Pock vacuum around A = O.

One observes the similarities with a periodic potential in qu~ntum
mechanics.
Fbr hadron physics, the vacuum with 8 = 0 is chosen. The

second term on the right-hand side of Equation (B.22) must be in-
tegrated over the scale size p of the pseudoparticles. However,
the integration cannot be extended to infinity, since the dilute

gas approximation ~ perturbation theory around instantons would
break down at the same time.
The gas is dilute, for consistency, if the pseudoparticles

occupy only a small fraction of the available four-dimensional
volume VT. Let us calculate this fraction f(p) for pseudoparticles
of scale size less than p. We take the volume of a pseudoparticle
of scale size p to be that of a four-dimensional sphere of radius
p (~n2p4) which leads to
It
P
f(p) = n 2 p4 2 n(p) dp
o
Introducing the notation x = 8n 2 jg2 and using Equation (B.20)
in the asymptotically free regime (small p), we can write:
00
'V
f (x) D (x) (B. 23)
x
The function f(x) is plotted in Figure 15.
0.0f. If#.)~
o.o't
0.02,
'----'-_....i...._-'---+x
10 11. ,+ ,(,
Figure 15. The fraction f of space-time occupied by pseudopart-

icles smaller than a given scale size p in an SV(3)
Yang-Mills theory.
We observe that the pseudoparticle gas is dilute for x ~ 14.
REFERENCES
1. Hasenfratz, P. and Kuti, J., Physics Reports, 40C, Page 75,

1978.
2. Callan, C. 3., Jr., Dashen, R. and Gross, D. J., Physics

Letters, 66B, Page 375,1977, and Physics Review D17, Page
2717 I 197a:-
184 JULIUS KUTI
3. Callan, C. G., Jr., Dashen, R. and Gross, O. J., Physics

Letters, 78B, Page 307, 1978, and Princeton University Preprint.
4. Friedberg, R. and Lee, T. D., Physics Review, D16, Page 1096,

1977, and CU-TP-118 Preprint, 1978.
5. Lee, T. D., CD-TP-127, Preprint, 1978.
6. Pagels, H. and Tomboulis, E., Nuclear Physics, B143, Page 485,

1978.
7. Glimrn, J. and Jaffe, A., Physics Review, ~, Page 463, 1978.
8. The first suggestion along this line can be found in:

't Hooft, G., TH-1902 CERN Preprint, 1974, and Kogut, J. and
Susskind, L., Physics Review 09, Page 3501, 1974. For more
details, see Reference 1. --
9. Chodos, A., Jaffe, R. L., Johnson, K., Thorn, C. B. and

Weisskopf, V. F., Physics Review, 09, Page 3471, 1974.
10. See, for example, Hasenfratz, P., Kuti, J. and Szalay, A. S.,
Proceedings X th Recontre de Moriond, Meribel, Vol. 2, Page
209, March, 1975; Gnadig, P., Hasenfratz, P., Kuti, J. and
Szalay, A. S., Proceedings of the Neutrino '75 IUPAP Conference,
Vol. 2, Page 251, 1975; Physics Letters, 64B, Page 62, 1976.
In References 1 and 9, the notation B was-USed for the volume
energy which I changed for p as a reminder of vacuum pressure
(see, also, References 4 and 5). I also changed the notation
for surface tension from cr to S.
11. fbr a more detailed and thorough discussion of instantons,

see the following reviews: Jackiw, R., Review Modern Physics,
Vol. 49, Page 681, 1977; Coleman, S., Harvard lhiversity
Preprint, 1978; Crewther, R. J., TIL 2522-CERN Preprint.
12. Polyakov, A. M., Physics Letters, 59B, 1975; Nuclear Physics,

B120, Page 429, 1977.
13. Abers, E. S. and Lee, B. W., Physics Reports, 9C, Page I,

1973.
14. Belavin, A. A., et al., Physics Letters, 59B, Page 85, 1975.
15. 't Hooft, G., Physics Review, D14, Page 3432, 1976.
16. Onaqger, L., J9.urnal American Chemical Society, 58, Page

1486, 1936; Frolich, H., Theory of Dielectrics, Clarendon
Press, Oxford, 1958.
17. Landau, L. and Lifshitz, E., Electrodynamics of Continuous

Media, Addison-Wesley, 1960.
18. Kuti, J. and Szlachanyf, K., unpublished.
19. Giles, R., Physics Review, 18D, Page 513, 1978.
20. Caldy, D. G., Physics Review Letters ~, Page 121, 1977.
21. Chodos, A. and Thorn, C. B., Physics Review, D12, Page 2733,
1975.
22. Creutz, .'1., Physics Review, D10, Page 1749, 1974; D12,
Page 443, 1975.
23. DeTar, C., Physics Review, D17, PP 302, 323, 1978.
24. For a few, see, for example: 't Hooft, G., Nuclear Physics,
B138, Page 1, 1978; Yoneya, T., Nuclear Physics, B144,
Page 195, 1978; FOerster, D., Physics Letters, 77B, Page 211,
1978; Nielsen, N. K. and Olesen, P., NBI-HE-78-24; Preprint;
Nielsen, B. H. and Ninomiya, M., NBI-HE-78-39 Preprint.
25. Glimm, J. and Jaffe, A., Nuclear Physics, B149, Page 49, 1979.
QU_~ AND FERMIONIC GEOMETRY
Jerzy Lukierski
Institute of Theoretical Physics

University of I~oc~aw
50-205 Wroc~aw, Cybulskiego 36, Poland
INTRODUCTION
The supersymmetric generalization of Einsteins
gravity was firstly obtained in the framework of four-
dimensional Minkowski Q]'T (1,2] .Recently it appears
however rather evident that the geometric interpreta-
tion of supertransformations requires the introduction
of additional anticommuting variables, which describe
together with four Minkowski coordinates the suoer-
space as a basic geometric manifold
Xp --+ X.-. I ~ -+ X;u., 9-,.;. /1/

space-time supers pace extended
(r : OI1.t;~) (. . =1, t.,~,~) superspace
C~ = .. ,1.··.N)
'ilie conventional space-time points acquire an additio-

nal internal structure, described by a finite-dimensional
Grassmann algebra. Such an internal space is purely
fermionic and contains the degrees of freedom descri-
bing spin as well as internal charges lin the case of
extended superspace formalism/.
Our main idea here is to introduce the geo;retri-
zation of' the internal degrees of freedom of hadronic
matter on the quark level via the supersymmetry scheme.
If we assume that the quark masses are due to the sy-
mmetry breaking we should get in the bosonic sector the
conformal transformations in its spinorial SU(!,t.) form.
We should assume also that the "maximal" geometric
187
188 JERZY LUKIERSKI
conformal group and internal symmetries do factorize

/i.e. conformal and internal symmetry generators co-
mmute/. In such a way we are led to the uni1ue choice
of graded conformal groups S;U(2.,'l. ;"'l [3,4, with the
following bosonic sector
(conformal group) • U(Ml

(internal symmetries)
/2/
unified on the level of spinorial supercharges, It is

interesting here to recall that the group SV(~,L\M)
is recently acknowledged on purely algebraic ground as
the best candidate for the unification of geometric and
internal symmetries.
If N = 1 , we obtain the supergroup which i.s gene-
rated by the Wess-Zumino superconformal algebra 5 with
the internal abelian V(" group. This the simplest case
we shall discuss in Sect 2. The fermionic representation
space is spanned by 4 complex anticommuting variables
t4l (4.: 4,1.I,It) which transform under SU(t,l) as a
conformal spinor named also a twistor [6] • We shall
call further these variables the quark-twistor (4T)
coordinates. Our first aim in Sect. 2 is to show how
acts on the ~T variables the Wess-Zumino superconfor-
mal group; we shall introduce there also the superconfor-
mal-invariant metric two-form '!
In such a scheme the U(1) internal symmetry group
/called 11 1s -transformations" in conventional discu-
ssion of SU(2.,t.'1) /generates the baryonic charge.
In our model there is also a place for the colour group-
the choice of cothlex fermionic QT variables leads to
the Uc., colour. Sect. 3 we shal! show that
a7 the replacement SU(t..,'I.\1,) ... sutt.,'I.; N) leads
only for ". ~ and N • 6 to the baryonic charge ~; the
internal symmetry group U(M) decomposes into baryonic
U(1) If flavour S U(") ,
b/ the replacement of complex CiT variables by the
guaternionic ones leads to the nonabelian SUtl.) colour
group; the choice of SU(\) colour leads to the QT fer-
mionic space with octonion-like coordinates.
*We shall use in this lecture the differential geometry

on Grassmann manifolds /see e.g.[7-91/. Because ourQT
fermionic variables are complex, we shall use here the
generalization of the differential geometry of Klhler
manifolds to fermionic spaces with complex Grassmann
coordinates.
QUARKS AND FERMIONIC GEOMETRY 189
Further, in Sect. 4 we shall consider for "=1 quark-

twistor (&Tl string obtained from c~mplex QT variables.
The passage from QT geometry with a given metric two-
form to the reparametrization-invariant action de-
scribing one-dimensional spinning QT string is a straight
forward procedure. It appears that the dynamics of GtT
string is described by a nonlinear two-dimensional super-
conformal-invariant tr -model [10], which can be written
in two equivalent forms:
- using independent QT string variables /see [10] /
20£. --.. i III (~,tl /3/
_ using two-dimensional supertwistor field (till CS,-.:-) , u.(~"t'))
wi th ~ fermionic and \A. bosonic constrained by addi-
tional subsidiary condition with SUt2,'l.·,t) group acting
linearly.
The second form of QT string model discussed here
can be described as fermionic C. PCIt) G"-model., and the
variables tal"'- play the role of preguark striW coordi-
nate§. Such a model requires the introduction of Abelian
colour gluon field which however due to field-current
identity can be reexpressed in terms of GTvariables.
An interesting result is the interpretation of the cur-
vature R of Grassmann aT space as the inverse of di-
mensionless colour coupling constant. If R ~ 00 i.e.
colour coupling is switched off, the Wess-Zumino super-
algebra is contracted as follows
..
where Tit denote anticommuting flat translations of Q,
/4/
variables. We obtain therefore the semidirect product

of a bosonic and fermionic sectors and without colour
coupling the supersymmetry is destroyed.
~order to describe hadronic string coordinates
we should introduce the dynamics of composite It and
22t string variables; such a procedure one can call
the hadronization of QT string coordinates. To look
into the hadronization problem is our task in the near
future.
SUPERCONFORMAL FERMIONIC GEOMETRY
Graded Lie groups can be represented by graded
• So-called C pc...) G" -models have been introduced by

Golo and Perelomov [111 and Eichenherr [121 • For re-
cent discussion of CPC,,) «r-model see [13] •
190 JERZY LUKIERSKI
matrices /see for example [141 . We can define the ele-

msnts of superconformal group as linear transformations
of the supertwistor space (t~,~)which preserve the fo-
llowing supertwistor norm
Five-dimensional supertwistor is a fundamental re-

presentation of superconformal group provided that the
statistics of t~ and ~ are opposite. One can choose
i/ bosonic supertwistors, obtained by adding to four
commuting twistor variables (t",ta"t",t .. ) fifth complex
Grassmann coordinate
u.t. = {\A"u,.l = u.*1: 0 /6/
In such a way one gets the Ferber's supertwistors [15].
ii/ fermionic supertwistcrrs, obtained by the assumption
that four complex twistor coordinates are Grassmann
variables·
/7/
The fifth complex coordinate ~ is chosen bosonic.
Supertwistor norm is invariant under the following
25-parameter group of linear transformationsl
a/ Conformal transformations / 15 generators A~p\S.p.
=
where A.-.p 1r -AI'-' , S.-." slW 1 S-.II(.. 0 I
-to' = s -t \4' =~ S E SU(t,2.) /8/
b/ Four com~lex Grassmannn rotations /8 anticommuting
parameters!
~ t~ = ~ EtA. ""
/9/
~ 'A. • -1. ii:
R
~ =t»"
'TWistor space with anticommuting components were in-
troduced in 16 ; such variables also parametrize
the coset space SU(,.,1.\4>/SU(2.1).V(1)COnsidered in (171.
wh~re Etl=~~+iE:iS an infinitesimal complex Grassmann

tWl.stor
c/ A-phase transformations /generator A. /
t' = e ~~~ u.' = e"""'",- /10/
d/ B-phase transformations /generator 8 /
-t:'=e"~t. ,-,-'=e"('"" /11/
which do not belong to the superconforma1 group.
The generator of the transformations 79/ can be
written as follows
Gal = E: A'" +~! 5'" /12/
where A'" and S'" denote eight supercharges. The super-
conformal algebra takes the following form [10]
{A ... , A"1- a ~~ (S"'t'i'lA,"P) .

.
-it. (~" +2A "ltlf')
\ StC., Cit"1 =
it.
{ A", C3p.' = Atl~ /13/
We shall assurne that the fundamental quark coordi-

nates are invariant under the 8-transformations. In or-
der to eliminate this UCt)degree of freedom we consider
supertwistor components Ct~.~) as the homogeneous coor-
dinates on projective Grassmann manifold describing QT
variables:
/14/
one can call the supertwistor coordinates the "pre-
quark"variab1es and one can introduce the following in-
verse formulae:
/15/
where 'f "'f+~ under 8 -transformations and describes

the '~idden" degree of freedom which is lost during the
projection /14/. Using /15/ we get the following expre-
ssion for the superconforma1-invariant metric on the
complex Grassmann projective p1ane*
'We obtain here the so-called Fubini-Study metric on pro-
jective plane with complex fermionic coordinates.
192 JERZY LUKIERSKI
<l s t. = (~! ltet + ~4~ ) c:lz" clz =

~2" ()z. ")ZA ~'ZB 8
. d2! rl14 _ ~ (2: d.2tt)(2r d.t"> /16/

=- 1. 1 + ~t. Rt. ( ~ + !!!~)t
~L ~~
=
where Z" (<t", 1:)'
We see that
a/ In the limi t R~ Co we obtain flat fermionic Cit,.. space,
with the group of motions described by inhomogeneous
U(',~) group with anticommuting flat supertranslations
. q" -. !~~E~(derived from the Grassmann rotations /9/
J.f ""It. -. ..)
*
b/ For R 0 the supertranslations are "curved", and
the nonlinear generators A.,.. S", in the parametrization
/14/ are given by the formulae
A...=t(~..-!~I'~..)+H~+~j1tl''') /17/
Set = (~.l + 2"/:1 \r:>-( ~-- 2tt\P)

The parameter ~ in the formulae /13/ and /16/
describes the sectional KAhler curvature of the complex
Grassmann manifold with QT coordi.nates. It will be ar-
6ued in Sect. 4 on the ground of 2-dimensional Q1'" string
model that the dimensionless parameter R can be identi-
fied with the inverse of the colour, coupling constant t.
DESCRIPTION OF FLAVOUR, BARYONIC CHARGE AND COLOUR
In order to find the physical interpretation of
A -transformations we should generalize our <itT geo-
metry by considering the superconformal group SU(2.,1.,M).
/5/ is now replaced by the following manifold
\t.,1.+ \t2,\ 1._ \-t~ \1_ ,t..\" +1\\,,\"+.·· .. +("" ..\,1. = R 2. /18/
where u... ... """
are com'plex bosonic coordinates. The
A -transformations /10/ are rewritten for the sUEer-
conformal group SU(1.,1.;If) in the following way (41:
. .
\.\.1: e" N.l u'i.
~
t \ : e."~ t /19/
The variables (-ttL ....... ) describe the prequark coordi-
nates. In order to introduce the QT coordinates, the
formula /14/ can be generalized as follows [18]:

r (\ \'1.)
2"-;,, :
'T ~'L -\;ol
u..
T 2.. \'1.
'''''\ =, ""
+ ., +lu." /20/ ,1
It is easy to check that this is the most general rela-
tion between the sU'Jertwistor and QT coordinates lea-
ding to desir~d transformation properties of QT va-
riables under the flavour group S\)(N). Substi tuting
/ 19/ in /20 / one gets
o' . - D·.. (1-!)o(o . /21/
~ "';" - ~ .L~j"
'.'ie see that only :'or tl =~ and 1'/=' one obtains that
'.
f"',,,~ e
~ ~.( .
2.c.;" 1"1= ~ /22/
We conclude that the A -transformations fit well for
the description of the haryonic charge' , because one
gets for physically reasonable SV(~l and SCl(~)flavour
groups the physical value Ibl: Js for single quark varia-
bles.
In order to relate the B -transformations with
the colour group let us observe firstly that these
transformations parametrize the U(t) fiber bundle atta-
ched to every point of the Grassmann complex ~T plane.
If we formulate the dynamics of Q,T variables in terms
of pre quark supertwistor coordinates one has to intro-
duce an Abelian gauge field which eliminates the depen-
dence of quark dynamics on the parameter ~ /see /15//.
One can say therefore that colour group describes the-
se degrees of freedom which are "lost" if we pass from
pre quark coordinates to QT variables.
It can be easily checked from the formula /20/
that also in the presence of S\)(N) flavour group the
transformations S stay Abelian. If we wish to obtain
nonabelian SU(~)fibre bundle over the pOints of Grass-
mann plane we should introduce the guaternionic suuer-
twistor coordinates. In such a case the projection by
means of the formula /15/ would mean an introduction
of SVU.) gauge freedom over the points of quaternionic
projective Grassmann Q, plane. The introduction of 5U(l)
colour group requires further generalization: one should
consider octonion-like coordinates. Colour obtained in
this way leads to SUn) as the largest possible colour
group.
FROM GEO~mTRY TO FIELD THEORETIC MODELS
One can cpnsider the replacement t." -. t"'('l") and
wri te the action for the fermionic Ci\T 'point" uarticle
194 JERZY LUKIERSKI
determined by the metric form /16/. Here we assume that

the hadronic quark matter forms a linear manifold - Qr
string.
One can write the action for aT string by conside-
ring the two-dimensional !!prequark" fields t4. C\,d and
'-\.(\,1:') which are restricted to the moat degenerate
orbi t /5/ of the superconformal 5u(~,'1;4) group. The
action has a form (j =S.,'t".~o)
S = i ~d.2.t {(V.. t ..\IlV'-4:"-+ (9,1.\)"."'4, + l (Itl:l+lw.l"-Rt.) \
.2. i. .0,4 /23/
where the local Abelian gluon field 8, (~) introduced
by the Bubsti tution
~ .. .... Vi. .. )~ - .. 't 8~ /24/
eliminates the colour degree of freedom described by the
6 -transformations / 11/. The fields ~ and 8~ play
the role of Lagrange multipliers. We obtain
Itl2.+ 1'4.'2. :: +! tIC.+ ""."" = ~1. /25a/
t"*V"t4l.-lV'--l,,fitC. 't 1.\·9~"'- - (V·Lc.c.)*'-\. =0 /25b/

The equation /25b/says that the total colour cu-
rrent is eEual to zero. Substituting /25&1 Into /25b/
one gets t e fIeld-current identity
B~ = -'" (~: ~ t"'+ tA-it; 2: "'-) :II k~ /26/
provided that
R'&. = 1. /27/
2.t
where ~~ describes the colour current for a system
without the gauge fields.
The composite gluon field 8~ characterize the
mapping of the supertwistor fields onto the projective
Q.T Grassmann plane C~'~)"'2. (arbitrary So,S .. , and
the scalar field ..,C\0,1..) /see /11// define the U(,\)
fibre over every point !.(SO.j4) • In the Euclidean
version of the model one can study the nontrivial to-
pological configurations of the field B. , determined
by the following formula for the topological charge
/28/
=ii SfA. t E~j (d,-t::~j.t~-t ~~ 'A,.)j t.c.) =

2
= t.n S4~~ ~~j [(V"t:)VA-tel ... (Vi.""-)Vi",-l

where the last equality is valid due to the constraints
/25a-b/.
Now we shall calculate the admissible values of't.
It is easy to observe that due to /15/ and /25a-b/ one
gets (-r=(S:+t:)%. ~.~S ... "So)
if the fields 1~(~4,t~ are sufficiently well behaving

at .,. ... 110. Introducing polar coordinates S"tl -. ~)tC.
one gets from/28-29/ that /see e.g. (193/
If we assume that the dependence of the fields t~,~

on a( is continuous, we obtain 1(-,2.1\):: 11'1k ... ~(-'D)
/ k integer/ and ~Q : k • We see therefore that one can
divide the two-dimensional supertwistor fields ~~(~),~(t)
into the homotopy classes "4(54) characterized by
the integer values of the integral /28/. In particular
one can introduce the instanton and antiinstanton so-
lutions, satisfying the selfduality equations
+. ...L±
V" ti = ± .. ~"j Vi,;.c, /31/
v~ '-t~ =± i, ~':j Vi ,-,-1
For these solutions the action is determined by the
topological charge
5 = tTiQ.
Introducing 2-dimensional Euclidean Dirac matrices ~,~&
one can write the equations /31/ in a Dirac form
196 JERZY LUKtERSKI
where
/33/
If we introduce by means of the formulae /14/ GlT

string variables, one obtains from /32/
/34/
or in Dirac form
...
5"~ l~ 'X.c. = 0 Xci.. == (~~) /35/
w~ere _{lltl ,'XI'!'\= 0 _ Using the complex variables ~. f1+
+"j,..\ =~.. - i. SJ. one can write the general solution
of /351 as follows:
~!~!t(\) t;=2~(i) /36/
We see therefore that the fermionic QT instanton and
antiinstanton solutions satisfy the free two-dimensio-
nal Dirac equation, and are described by the formulae
/36/-
Abelian gluons and W-" _
In this Section we described only the model with
The quaternionic 6lT'string
model as well as the generalization to arbitrary
can be constructed in analogous way.
HADRONIZATION OF QT FERMIONIC COORDINATES
Let us summarize briefly our approach to the de-

scription of quarks and extended space-time structure:
a/ The points of space-time occupied by the hadronic
matter are characterized by ~~ fermionic variables
with R'" :(1ty4l:faO. The one-dimensional continuous
quark system is described by QT string.
b/ All the degrees of freedom of hadronic matter at a
given point are composed out of the elementary fermio-
nic coordinates represented by the QTvariables.
In order to introduce the coordinates for hadrons,
describing e.g. spin and internal charges, one should
answer the following two questions:
i/ How to define from Qr variables the comoosi te
hadronic string coordinates,
ii/ How e.g. the dynamics of 6lT string discussed
in Sect. 4 can be translated into the dynamics of com-

posite hadronic coordinates.
The answer to these questions is made easier by
the group-theoretic descriptioll of the quark model of
hadrons: one should look for 21 /mesonization of~T
string/and .ftt /baryonization of ~T string/ composi-
tes. In orner to obtain only these composites one sho-
uld introduce in suitably complicated way the colour
group: for example in the case of octonion-like string
coordinates one would obtain that exotic products of
string variables/different from it... andfLt/ are not
allowed on purely algebraic ground-[20, 21J. Other pro-
perties of hadrons require the introduction of non-
trivial flavour grou~ and suitable syw~etry breaking
terms; these generalizations will be considered in
our future work.
FINAL REMARKS
One can say that the space-time manifold is intro-
duced in order to provide a framework for the geometri-
zation of energy-momentum distribution; the geometry of
space-time becomes Riemanian and the scalar curvature
is non-zero if the masses are present. Here we discu-
ssed another manifold which is described by purely fer-
mionic QTvariables. One can say that the presence of
these variables in the extended geometric framework
provides the place for the geometric introduction of
strongly interacting hadronic matter. In similar way
like mass determines the space-time geometry we suppo-
se that the coloured sources imply the curved super-
conformal geometry of QTvariables. The colour coupling
constant has a geometric interpretation as a Kibler
curvature of complex GlT Grassmann space. The geometry
wi th flat shifts of fermionic variables 9... .... 9~'tE4. dis-
cussed e.g. in the super Poincare- invariant theory
and further gauged in supergravity corresponds therefo-
re to the case of quark degrees of freedom with swit-
ched-off colour interaction.
In this lecture we did simpl:1.:fy to the extreme
the misterious relation between the space-time and
fermionic geometry - we introduced QT string variables
which do not couple these two geometries at all. It
is the simplest possibility, due to the two-dimensio-
nal nature of string dynamics, in which the coupling
between the external and internal /gauee/ symmetries
i8 introduced by choosing the gauge-fixing condition.
In order to study however the connection between the
supergravity and our approach the relation between
fourdimensional space-time manifold and fermionic
198 JERZY LUKIERSKI
curved Q,T geometry should be studied and clarified,

with the possibility of incorporating the Penrose theo-
ry of composite space-time coordinates 1ft and leptons
as quarks without the colour charge.
REFERENCES
1. D.Z. Freedman, P. van Nieuvenhuizen and S. Ferrara,
Phys. Rev. D~), 3214 /1976/
2. J. Wess and • Zumino, Phys. Lett. 62B, )35 /1976/
). R. Haag, J. Lopuszanski and M. SohnIUS, Nucl. Phys.
B88, 257 /19751
4. S:-Ferrara, M. Kaku, P. van Nieuvenhuizen and P. K.
Townsend, Nucl. Phys. Bj29, 125 /1977/
5. J. Wess and B. Zumino, ucl. Phys. f70, )9 /1974/
6. R. Penrose, Journ. Math. Phys. 8, ) 5 /1967/
7. F.A. Berezin and G.J. Kac, Math7 Sbornik, 82, )4)
/1970/
8. C. Fronsdal, Lett. Math. Phys. 1, 165 /1976/
9. F. Mans o1,l.ri , Journ. Phys,. ,18, 52 /1977/
10. J. Lukierski, ICTP, preprim IC/78/82, July 1978
submitted to Journ. Math. Phys.
11. V. L~ Golo and A.M. Perelomov, ITEP-62 Moscow Uni-
versity preprint
12. H. Eichenherr, Freiburg University preprint, 1978
1). D'Adda, P. diVecchia and M. Lfischer, Nordita pre-
print, 1978
14. P.G.O. Freund and I. Kaplansky, Journ. Math. Phys.
17, 228 /1976/
15. A7 Ferber, Nucl. Phys. Bl)2, 55 /1978/ .
16. J. Lukierski, Stony Brook preprint ITP SB 78/8,
Nuovo Cimento Letters, in press
17. F. Gfirsey and L. Marchildon, Journ. 1~th. Phys. 19,
942 /1978/; Phys. Rev. ]11, 20)8 /1978/ --
18. J. Lukierski, Lecture at Symposium on Math. Methods
in the Theory of Elementary Particles, Lib1ice
/Czechos1ovakia/, June 1978; Wroc~aw University pre-
print No 436, July 1978, Czech. Journ. Phys., in press
19. M.M. Ansourian, Phys. Lett. 70B, )01 /1977/
20. M. Gdnaydin and F. Gfirsey, Phys. Rev. ~, ))87 /1974/
21. G. Domokos and S. K6vesi-Domokos, Journ. Math. Phys.
12, 1477 /1978/
See [22] where the composite space-time coordinates
are expressed by two twistors with commuting compo-
nents. See also Rzewuski et all [2),24] who introdu-
ced in analogous way the spinor space as the funda-
mental geometric manifold.
22. R. Penrose and M.A.H. Mac Callum, Phys. Rep_ 6C

242 /1972/
23. J. Rzewuski, Nuovo Cimento 2., 942 /1958/
24. R. Ab~aroowicz, J. Mozrzymas, Z. Oziewicz and J. Rze-
wuski, Rep_ in Math. Phys., in press.
NUCLEAR FIREBALLS IN HEAVY ION COLLISIONS
I. Montvay
Centre for Interdisciplinary Research
University of Bielefeld, F.R. Germany
ABSTRACT
The properties of nuclear fireballs produced in energetic

heavy ion collisions are considered. Questions concerning hadro-
chemical equilibrium and possible phase transitions are dealt with
in more detail.
I. INTRODUCTION
Energetic collisions of heavy nuclei offer a quite unique

possibility to study in the laboratory the properties of very
dense and hot nuclear matter. The relativistic energy range is
particularly interesting because of the particle production
processes occuring in the collision of individual nucleons.
The study of relativistic heavy ion collisions is a vast and

rapidly developing field incorporating interesting aspects of
particle and nuclear physics as well as statistical physics. In
the present review I shall not ~ry to cover all the recent de-
velopments. Instead, I shall mainly concentrate on a few aspects
such as the question of hadro-chemical reactions and if possible
phase transitions. (For recent compilations of the topic see
201
202 I. MONTVAY
e.g., Refs. [1-3]).
To have an idea about the conditions occurring in relativistic

collisions of large nuclei let us briefly consider a few characteri-
stic features of the kinematics. Ek' If
is the kinetic energy
1.n
of the individual projectile nucleons in the laboratory system and
~ is the nucleon mass, then the c.m. velocity of the nucleons is
v = I Ekin
2~+Ekin
(1.1)
In the c.m. system the shape of the nuclei is Lorentz-contracted

in the direction of motion by a factor
( 1.2)
The relativistic regime, where v is not negligible compared to

1, begins somewhere about Ek'1.n ';;: 500 MeV. At Ek'1.n = 2000 MeV
the motion is already quite relativistic.
If we consider the central collision of two equal nuclei with

radius R, then the time difference tl between the first moment
of touch and the complete overlap of the two Lorentz-contracted
spheres is
= R!2~'.
Ek'1.n
(1.3)
A nucleus with A nucleons has a radius roughly equal to
(1.4)
Entering into the relativistic regime means a drastic decrease of

t 1 • For uranium nuclei (A = 238) at Ekin 2000 MeV (1.3) gives
-2. 3
a value about 2 • 10 s. (For comparison, the lifetime of a
typical hadronic resonance, like p-meson for instance, is about
NUCLEAR FIREBALLS IN HEAVY ION COLLISIONS 203
If the tw.o nuclei w.ould just g.o en t.op .of each .other filling
the cemmon .overlap v.olume unif.ormly (a situati.on, .of c.ourse, c.om-
pletely unrealistic but still instructive) the density w.ould be
(1. 5)
where
(1. 6)
is the density .of the undisturbed nuclei f.oll.owing fr.om Eq. (1.4).
There is a c.onsiderable rise in the relativistic regime f.oll.owing
fr.om the ever increasing degree .of L.orentz-c.ontracti.on.
Due to the interacti.on .of the tw.o nuclei the kinetic energy
.of the nucleens in the c.m. system is certainly rand.omized t.o seme
extent. The degree .of randemizatien ("thermalizatien") is, .of
c.ourse, an interesting and impertant questi.on. Assuming cemplete
thermalizatien (i.e., all the energy transfermed t.o thermal energy)
the temperature .of the preduced het and dense piece .of hadrenic
matter ("fireball") can be calculated .once the thermedynamic
equatiens .of state (in particular, the specific heat) .of the
nuclear matter are kn.own. Taking again the simplest case .of
symmetric c.ollisiens (i.e., central cellisiens .of twe equal nuclei)
and assuming, fer a first guess, ideal relativistic gas equatien
.of state fer the nucleens, gives fer the temperature T the
fell.owing equatien:
(1.7)
Here the functi.on R(x) is defined as

204 I. MONTVAY
3 Kl (x)
R(x) = - + - - - - - ; > + 1-
2x
+ (1.8)
x K~(x) ex + 00)
and mNR(~) is the average thermal energy of nucleons in the

relativistic ideal Boltzmann gas [4]. (In the non-relativistic
limit ~/T + 00 one can recover from Eq. (1.8) the well-known
thermal kinetic energy term 3kT/2 , as we are using kB It =1).
o. zmann
Using the ideal gas equation of state for such a dense system
is, of course, unrealistic. The interaction of the nucleons leads
to strong cluster formation in the nuclear matter, therefore it is
completely different from an ideal gas. The effects of clusteri-
zation can be calculated from the statistical bootstrap equations
[5]. Let us take, for simplicity, the bootstrap equation with
nucleon and pion input terms only. This gives presumably a good
approximation in the relativistic regime we are interested in. It
leads, instead of Eq. (1.7), to the equation valid for the nucleon
fluid [5]:
/1 + Ekin' =
2~
I(~
T' T
m'IT) (1. 9)
where m'IT is the pion mass and ~I (T'

~ Tm'IT) is the average
thermal energy per nucleon given by
+ _3_ +
2~
a (1.10)
o v m 3
'IT 'IT
Here is the bootstrap parameter equal to a typical hadronic
volume:
...... 4'IT -3
=-m (1. 11)
3 'IT
Note that in the non-relativistic limit (~ + 00) Eq. (1.10) gives

again th.e ideal gas result. Non-relativistically, however, the
nuclear binding effects are non-negligible, therefore stable nuclei
must also be included in the input term modifying the above result.
From Eqs. (1.9 - 10) it follows that the temperature can never
exceed the Hagedorn temperature To where
ao = ~ K (mTI) (1.12)
m
TI 1 T 0
Numerically, it follows from Eq. (I. II) that To ~ 217 MeV.
A very important factor in the collision of such (from the

hadronic point of view) "big" objects like, say, an U nucleus,
is the geometry of the collision. For large impact parameters
only the periphery of the nuclei get hurt and most of the energy
is carried away by the remaining, relatively undisturbed pieces
of the two nuclei. The picture is completely different for small
impact parameters when most part of the matter of nuclei intercept
each other. It is to be expected intuitively that the general
features of the collision events are dominated to a large extent
by geometry. From the point of view of producing hot, dense
hadronic matter central collisions are presumably much more
interesting.
Another factor is the relative size of colliding nuclei. For

very asymmetric collisions, when (mostly) the projectile nucleus
is much smaller than the target, the typical picture is that the
projectile cuts out a tube from the target (with a length depending
on impact parameter) leaving most of the target untouched. Hence,
the two nuclei play quite different roles. For equal nuclei, how-
ever, the situation is nicely symmetric.
The generally accepted picture in fireball models of heavy

ion collisions [6-8] (see also Ref. [1-3] for a more detailed list
of references) is that the parts of the two colliding nuclei
206 I. MONTVAY
intersecting each other's trajectories form a "nuclear fireball"

in some strongly interacting "chaotic" state described thermody-
namically (and sometimes hydrodynamically). The remaining parts
(sometimes two, sometimes one depending on the geometry) run away
only a little (or not at all) excited. It is usually assumed that
the fireball is at least approximately described by equilibrium
states, therefore notions like temperature pressure, collective
flow velocity etc. have some meaning in it. The temperature T
determined by Eq. (1.9) refers, in fact, to the fireball in the
case of symmetric collisions. For asymmetric collisions, when the
fireball is produced from a number of A nucleons from the
p
projectile and nucleons from the target the corresponding
equation is
2Ap A t Ek'1.n
(I. 13)
(A +A )2m...
p t N
We can see from here that the symmetric case A At is the mos t
p
favourable one for producing high temperatures.
In the symmetric, central collisions the time available for

the formation of the fireball is just the time tl from the first
touch of the nuclei to the complete overlap. The very highly
compressed matter explodes then into the surrounding vacuum. As
the system is quite big, some sort of hydrodynamic collective
motion must be there during the expansion as the outer layers are
experiencing a rather big pressure form inside and none from out-
side. (Note in this respect the difference from high energy
single hadron collisions where colour confinement may provide some
sort of external pressure to the quark system). During the ex-
plosion period the thermal energy is transformed into the kinetic
energy of the collective motion, therefore the gas is cooling down
and is getting more dilute all the time until the nucleons "lose
contact" with each other and the system is "breaking up" or
"freezing in" conserving its actual state later on. This "frozen
in" state is available for experimental observation. The hard
problem is, of course, that we have to figure out the whole story
from this "frozen in" picture.
The breaking up point of the fireball can be determined in

different ways. In a pure hydrodynamical picture a plausible
condition is that break-up happens at the time instant, when the
velocity of divergence (perpendicular to the radial expansion
velocity) is reaching the average velocity of the thermal motion
[9]. Another criterion might be the stopping of hadrochemical
reactions producing some sort of hadrons, say pions [10]. Still
another, rather good possibility is that the break-up is at the
phase transition point from nuclear fluid to nuclear gas state
[5], which is roughly at a nucleon number density
'" In2
\)
=V (1.14)
* 'IT
In the fluid phase the nucleons have the strong tendency to form
big clusters consisting of many nucleons, whereas no such strong
tendency is present in the gas phase equivalent to a free gas of
clusters. Numerically, using the value in Eq. (1.11) for the
characteristic hadron volume v , the density in Eq. (1.14) is
'IT
about 1/3 of the standard nuclear density Po in Eq. (1.6).
It is clear that the main problem of the thermodynamic,

hydrodynamic or any statistical models is whether or not the short
time and the relatively small number of particles (about 500
nucleons and hence about 1500 quarks in the U + U case, however!)
justify the use of our macroscopic equilibrium concepts. Surely,
it is much better to consider the process as successive efforts
of the system to reach some sort of nearly equilibrium situation
rather than a smooth adiabatic succession of equilibrium states.
In other words, transport phenomena are always very important even
208 I. MONTVAY
if equilibrium considerations can give a first insight into the

problem.
II. HADRO-CHEMICAL REACTIONS
Once the temperature T of a gas of hadrons (nucleons,

pions etc.) is high enough, particle production becomes essential
in the collisions among hadrons. This means reactions transforming
different hadronic states into each other. The situation is simi-
lar to ordinary chemical reactions, therefore we can say that
''hadro-chemical reactions" are going on. Actually, the threshold
for resonance production in a nucleon gas is near T ~ 50 MeV.
Above this temperature hadron chemistry is important.
The effects of hadro-chemical reactions in heavy ion collisions

were considered in detail in a recent paper [10]. The main
assumptions of this paper are that:
i) thermal equilibrium sets in faster than chemical equilibrium,

therefore one can speak about hadro-chemical reactions going
on at some given temperature;
ii) the hadron gas is considered as a multicomponent, relativistic

ideal Boltzmann-gas.
The motivation for the first assumption is that for reasonable

temperature values (100-200 MeV) the average collision energy is
relatively low, therefore the elastic collisions (collisions
leading to no hadro-chemical reactions) dominate. The elastic
nucleon-nucleon cross-section is very large at small (Eko~n <100 MeV)
energies and the low lying resonances (A-resonance in the Nn system
and p, w-resonance in the nn-system) are all elastic. The second
assumption is more questionable as the gas is far from being dilute.
Nevertheless, once the use of the S-matrix for individual colli-
sions is allowed the situation is not so had. Namely, resonances
are taken into account, therefore an essential part of the
interaction among hadrons is included [11, 12].
The equations governing the rate of hadro-chemical reactions

were derived in Ref. [10]. For the simple case of a spatially
homogenous gas consisting of a single sort of ground state hadrons
(called "'IT-meson") and a single sort of resonances ("p-meson") the
equations for the time derivative of densities v (t) and v (t)
'IT p
are [10]:
n
dv (t) Kl (mpB) [ v'IT(t) Qp(B)]
'IT 00
nr
L V (t) -
dt n=2 n K2 (mpB) p Q (B)n
'IT
(2. I)
dVp(t) 00 Kl (mpB) n Qp (B) ]
L r V (t)
dt n K 2 (mp B) [-Vp(t) + 'IT Q (B)n
n=2
'IT
Here m and mp are the masses, B = B(t) is the inverse

'IT
temperature, K (x) are Bessel functions and
1 ,2
(2.2)
The reactions taken into account by Eq. (2. I) are the formation
and decay of the resonance:
p ~ n'IT (n 2,3, ... ) (2.3)
(For the true p-resonance only n = 2 matters, but here we can

leave this more general). The total width of the resonance is
r , which is a sum over the partial widths rn
00
r L r (2.4)
n=2 n
Note that Eq. (2. I) involves a narrow resonance approximation. To

take into account the resonance width one has to express dv /dt
'IT
210 I. MONTVAY
by the scattering cross-section (see Ref. [10]) and put a Breit-

Wigner form there.
The chemical equilibrium is reached, according to Eq. (2.1),

when the absolute activities
(2.5)
are equal to 1.
In the numerical calculations of Ref. [10] nucleons, ~-re
sonances, pions and p-mesons were taken into account. It turned

out that the time necessary to reach chemical equilibrium is of
the same order of magnitude as the total reaction time in heavy
ion collisions. Therefore although the ratios of particle numbers
of different sorts of hadrons usually do not reach the equilibrium
values, they are also not very far from them.
At temperatures near the Hagedorn temperature To (i.e.

laboratory energy per nucleon near 2 GeV) the inclusion of only
the lowest lying resonances is not sufficient. The fast (exponen-
tial) increase of the resonance mass spectrum implies that near
To the excitation of higher resonances becomes essential.
The generalization of Eq. (2.1) for an exponentially in-

creasing spectrum is, in fact, not difficult. Let us consider the
nuclear bootstrap with pions and nucleons [5] when there are
resonances ("clusters") with baryon number b = 0,1,2, •.. The
mass spectrum TdM2 of mass M and baryon number b = ° clusters
is [5]:
00 1-1 2
L gl Bn Po 1 (M ) (2.6)
1= I '
whereas for b > we have

(2.7)
Here are the Yellin-coefficients (see e.g. [13]), B denote

1T,p
the bootstrap parameters
m.v.
B. = _1._1._ (2.8)
1.
and the phase-space integrals are defined as
(2.9)
For numerical calculations it is better to go over from the

continuous spectrum of clusters to a discrete one putting all the
states nearby in mass to some discrete mass value. Let us denote
the degeneracy of the nlth discrete level with baryon number b
by d(n, b). We forget, for simplicity, about charges, isospins
etc. and put for the stable states with n=b
d(b,b) I . (2. 10)
Choosing the discrete mass values as
m for n =0
m
n { 1T
n~ for n > I
(2. II)
we put because of ~2~ (2n-l) ~ for the unstable states (n>b):
d(n,b) = (2n-l) ~ (2. 12)
In accordance with "reciprocity" [13] see also [14]) we assume

that the average resonance width for the transition nb+(n1b1 .•• nkb k )
212 I. MONTVAY
is
(2. 13)
. 8(b, k
L b. ) Pk (2.
1
m
n
-+ m2. ,
nl
... , m2. )
nk
i=1
Here 8(b,c) denotes a Cronecker-delta, Pk is the k-body phase

space integral with masses m , ••• m defined in analogy with
nl ~
Eq. (2.9), r 00
is the (constant) asymptotic resonance width [14],
and according to Ref. [5] we have to put
for b = 0
(2. 14)
for b > 1
In he nucleon gas consisting of (n,b) clusters the transitions

nb # nlb l , ... , ~bk can be considered as hadro-chemical ones
changing the chemical composition of the gas. Assuming thermal
equilibrium with a time dependent inverse temperature Set) and
spatially homogenous densities Vnb(t) of cluster types (n,b)
one can easily write down the hadro-chemical reaction rate equations
corresponding to Eq. (2. 1) :
dVnb (t) co
snb(t) + L
dt
n'=n
o(n,n') O(b,b')] • Qn'(S) d(n',b') [i\'b,(t) - Anlb~t) •.•
A~bk (t) ] (2.15)

Here snb(t) is an eventual external source term and the absolute

activities ("fugacities") of the components (n,b) are introduced
as
v
A = nb (2.16)
nb d(n,b)Q (S)
n
The hadro-chemical equilibrium is at
for b = 0
(2. 17)
{(I A )b/b'
for b, b' of 0
n'b'
The time dependence of the temperature can be determined from

energy conservation. The energy density in the gas is
00 00
E(t) L L V b(t) m R[m S(t)] (2. 18)
n=o b=o n n n
where R(x) is defined by Eq. (1.8). If the change of the energy

density is given (by external sources etc.) then the change of the
inverse temperature can easily be calculated:
dS(t) 00 ddt) 00 dVnb (t)

-dt L m2. V bet) R' (m S) = - d t - L m R(m B) dt
n,b=o n n n n , b=o n n
(2.19)
Here R'(x) denotes the derivative of R(x).
Although it seems less advantageous from the point of view of

the numerical study, the continuous analogue of Eqs. (2.15) can
also be easily written down. In the continuous case we have instead
of Vnb(t) the density distribution V(M2.,b; t)dM2. and the
corresponding absolute activity
(2.20)
214 I. MONTVAY
The widths corresponding to reciprocity are given by
(2.21)
k
• IS (b, L
i=1
b.)
~
... , ~)
The continuous analogues of Eqs. (2.15), (2.19) are:
aV(M 2 ,b; t)
at
00
L
I Jkn dM~
-,
00
L
k=2 k. i=1 ~ b.=o
~
(2.22)
and
aV(M ,b;t) MR(MS) (2.23)

at
Coming back to the discrete equations, I studied numerically the

set of equations truncated at some upper limit for the cluster
mass. In order to keep the number of equations at a reasonable
level (actually 21) I took m < 5~ .• For the initial conditions
n - l.'l
and for the source terms I have chosen typical values in the heavy
ion collision model of Ref. [10]. In this model it was assumed
that for symmetric central collisions in the "ignition period"
(from the first moment of touch of the two nuclei at t = 0 to the

moment of complete overlap at t = t 1) the nuclear fireball fills
the overlap region, whereas in the subsequent "explosion period"
the volume is growing according to the hydrodynamical model [9]
and the corresponding amount of collective flow energy is sub-
tracted from the thermal energy. It is assumed all the time that
the fireball fills the volume uniformly and I took for the volumes
also Lorentz-contraction into account (unlike it was done in Ref.
[10]).
The numerical results show that the chemical composition does

not change appreciably in the "explosion period" hence, the abundance
of different hadronic states reflects mainly the previous "ignition
period".
The need for more time to establish chemical equilibrium at such

high temperatures is, in fact, to be expected on general grounds as
near to critical points relaxation times always get longer. Such
critical points in our case are the nucleon gas - nucleon fluid phase
transition and the maximum temperature point at T=T.
o
The former
one implies large nucleus formation and the second one very heavy
hadron formation. Both have to happen through a rather complex coup-
ling structure of a large number of states requiring a considerable
amount of time.
Note, that in the case of low p~ hadron-hadron collisions a

similar situation seems to occur, namely thermal equilibrium shown
by the universality of the transverse energy cut-off [15] but no
chemical equilibrium, as the inclusive cross-sections are dependent
on the quantum numbers like strangeness, charm, baryon number etc.
III. PION CONDENSATION
The history of the nuclear fireball begins in the "ignition

period" by a very hot and compressed phase. During this period
216 I. MONTVAY
hadro-chemical reactions are fast and there are a considerable number

of pions and resonances produced (even if time is not enough to reach
perfect chemical equilibrium). In the subsequent "explosion period"
the hot and dense matter is cooling down and the hadro-chemical reac-
tions become slow. The previously produced pions (increased by the
decay products of the resonances) are much more numerous than re-
quired by the low temperature. This means that the absolute activity
of pions A is increasing to large values. As it was pointed out
n
In Ref. [10 ], once the chemical potential ~ = TInA reaches the
n TT
value of the pion mass m a Bose-Einstein condensation of pions
n
begins. This happens at the end of the fireball's life, not much be-
fore the break-up, therefore it is easily observable. (Events that
occur earlier in the fireballs history are "washed out" to a large
extent from its "memory" by the later thermalization).
In the non-relativistic situation the hadro-chemical reactions

are extremely slow, as the average energy of collisions is far below
the resonance production threshold. Therefore, the number of pions lS
constant and one can consider the pion condensation at fixed pion
number density. By lowering the temperature T the condensation be-
gins at T=T*. As it was shown in detail In Ref. [16] by the use of
the grand-microcanonical ensemble, below the temperature T* the
critical pion gas coexists with a condensed phase consisting of in-
finitely large pion clusters.
In the relativistic regime particle production is important, the

number of pions is not constant, the hadro-chemical reactions become
fast and in the true equilibrium we have always A I. Therefore,
n
there is no pion condensation. As we saw, however, in the previous
Section, for rapidly changing processes occuring e.g. in heavy ion
collisions, the time lS not enough to reach true chemical equilibrium.
The reason is the large net of resonant states with a complex coup-
ling structure, making the transfer from one sort of states to the
other relatively slow. Therefore, as it was advocated in Ref. [16],
there is a possibility for pion condensation even in the relativistic
regime. In general, it is clear that hadro-chemical type processes

are important for the possible phase transitions in high temperature
hadronic matter.
PEFEPENCES
I. M. Gyulassy 1n Proc. Int. Symp. on Nuclear Collisions and their
Microscopic Description, Bled, Yugoslavia, 1977, Physica.
2. J.R. Nix, Theory of High Energy Reavy Ion Collisions, preprint

LA-UR-77-2952.
3. H. Feshbach, Relativistic Heavy Ions, Lectures at les Houches

Summer School, Aug. 1977.
4. I. Montvay, Equations of state for relativistic quantum ideal

gases of massive particles, Budapest preprint KFKI-1978-48.
5. R. Hagedorn, I. Montvay, J. Rafelski, this Proceedings.
6. G.F. Chapline, M.H. Johnson, E. Teller, M.S. Weis, Phys. Rev.

D8, 4302 (1973).
7. G.D. Westfall et al., Phys. Rev. Lett. lZ, 1202 (1976).
8. A. Mekjian, Phys. Rev. Lett. 38, 640 (1977); Phys. Rev. C17,
1051 (1978).
9. J.P. Bondorf, S.I.A. Garpman,J. Zimanyi,Nucl.Phys. A296,320(1978).
10. I. Montvay, J. Zimanyi, Hadron chemistry in heavy ion collisions,

Budapest preprint, KFKI-1978-45.
II. E. Beth, G.E. Uhlenbeck, Physica, ~, 915 (1973).
12. S.Z. Belenkiy, Nucl. Phys. l, 259 (1956).
13. R. Hagedorn, I. Montvay, Nucl. Phys. B59, 45 (1973).
14. S. Frautschi, Nucl. Phys. B91, 125 (1975).
IS. E.H. de Groot, H.Satz, D.Schildknecht, Phys. Lett. 77B,418(1978).
16. I. Montvay, H. Satz, Nuovo Cimento 39A, 425 (1977).

A CRITIQUE OF ASTROPHYSICAL APPLICATIONS OF HAGEDORN'S BOOTSTRAP
W. NAHM
CERN - Geneva
Talk given at the Workshop of Hadronic Matter at Extreme

Energy Density, Erice, October 1978
Abstract
Hagedorn's bootstrap should not be applied to hadronic

matter at densities large against nuclear densities. The
correct predictions of the thermodynamical model do not use
any relation between the mass of the fireballs and their size,
whereas the astrophysical applications depend on the unreason-
able assumption that the size is independent of the mass.
Moreover, the most spectacular prediction of the bootstrap,
namely violent black hole explosions yielding 10 15 g in the
last millisecond, is completely unfounded, even if such an
assumption is made.
219
220 W.NAHM
1. - INTRODUCTION
Astrophysicists discuss many different equations of state

for hadronic matter at energy densities large against
1 GeV!fm 3 1) However, for the calculation of black hole decay
two models are the most popular : conventional field theory and
Hagedorn's bootstrap. The models tend to be treated on the same
footing. Expensive experiments have even been justified with the
assumption that the bootstrap has a reasonable chance of being
correct. Here I want to show that astrophysical applications
of the bootstrap should not be believed, though Hagedorn's mo-
del is very successful in describing hadron collisions at acce-
2)
lerator energies In part this is a self-criticism.
In the asymptotically free non-Abelian gauge theories fa-

voured today for the non-gravitational interactions, matter at
high energy densities behaves like radiation 3). This means in
particular that the pressure is about one third of the energy
density, and that the energy density increases like the fourth
power of the temperature. For other renormalizable theories, it
might well increase with a different power of the temperature,
but there is no indication that a maximal temperature exists.
In theories with an infinite number of fundamental fields one
might have a maximal temperature or no homogeneous thermal equi-
librium at all, but the divergences of such theories probably
cannot be made sense of anyhow. Gravity also behaves differently,
as the classical energy is not bounded from below if black holes
are present. Gravity of course is weak and does not pose much of
a problem below the Planck temperature of 10 20 run,
but theories
with strongly coupled tensor fields might well have an unexpected
behaviour. Nevertheless, our present understanding of high
energy physics indicates that below the Planck temperature the
free gas behaviour is qualitatively a reasonable approximation.
ASTROPHYSICAL APPLICATIONS OF HAGEDORN'S BOOTSTRAP 221
Hagedorn's bootstrap yields very different results. There

is a maximal temperature To N ~, such that one can reach arbi-
trary high energy densities at finite temperature, which would
be very important for astrophysics. In the original version of
the model the equilibrium pressure increases only logarithmically
with the energy, in another version it even becomes constant.
This is possible because the fluctuations of the energy density
near To become very big and the energy is concentrated in a
few fireballs. This also has a further interesting consequence
the establishment of thermal equilibrium takes a long time. In
particular, the expansion of the universe should easily disrupt
the equilibrium, and the later decay of the fireballs down to low
energy densities might create lots of entropy. This has been
used by Carlitz, Frautschy and myself to explain the high value
of the entropy per baryon 4). I shall say something more about
our model and its deficiencies in Section 3. What is more im-
portant, the prediction of a maximal temperature has much increa-
sed the hope of astronomers to observe the final explosions of
black holes.
2. - BLACK HOLE EXPLOSIONS ACCORDING TO THE BOOTSTRAP

According to Hawking 5), the temperature of a Schwarzschild
black hole is inversely proportional to its radius
T-
and for its mass one has
(2 )
(we put 1'1= c = k= 1). This yields an entropy
S " f~ ~ ~ - r7fGM 2
.
222 w. NAHM
If To really is the highest temperatQre, the black hole should

explode after it has reached a radius R rv O.1 fm, i.e., a mass
Mrv 10 14 g. To see what is meant by explosion consider the en-
tropy of a hadron of high mass, which according to Hagedorn is
S ~ ~ (4)
To'
When the radius of a black hole has decreased to (2rrT )-1, the
o
entropy of a hadron of the same mass starts to get bigger than
that of the black hole. According to the original version of
Hagedorn's bootstrap, the volume should be comparable, too.
Thus it becomes very probable that the black hole will change
into a hadronic fireball.
In Ref. 4) it was argued that the decay time of a fireball

of this mass should lie between 10- 4 sec and 10 5 y. Astro-
physicists usually use the lower limit or even shorter decay
times, yielding very powerful explosions, I insisted that the
decay time is 10 5 y, if one really sticks to the model 6)
However, this calculation neglected gravity, and Hawking and
Page correctly pOinted out that this is inadmissible, as for
fireballs heavier than Planck's mass of 10- 4 g the gravita-
tional energy cannot be neglected against the kinetic energy 7),
though of course below 10 14 g the gravity contribution to the
fireball mass is small. They argued that gravity should increase
the decay rate, as in thermodynamical equilibrium the collisions
between the fireballs would become more numerous, and therefore
also the decays. Unfortunately this argument is completely wrong,
as it is based on a given equilibrium density of fireballs, in-
dependent of the gravitational interaction. But if one starts
from an equilibrium with G=O and then turns on gravity, most
of the collisions would yield fi~eballs of higher mass, until a
new equilibrium is reached, with a much lower density of fire-
balls and drastically reduced rates of collisions and decays.
Indeed, Hawking and Pagels argument starts at a wrong point.

Intuitively, fireball fragments should have greater difficulty
to get apart, if they attract each other gravitationally.
Taking gravity into account, one also expects a smooth

transition between hadrons and black holes. For masses less
than 10 14 g, black holes should spontaneously change into
hadrons, and for masses larger than 10 15 g, hadrons will be
inside their Schwarzschild radius. For hadron masses between
10- 5 g and 10 14 g the spectrum should interpolate between
Eqs. (3) and (4). Thus let us repeat the bootstrap calculations
taking into account gravity.
Let a(E) be the density of states per energy, i.e.,
o (E) == J£ ~(INt)d(€-'f"&+~')c/l,j~1 (5)

(271-)' JJ
where p(m) is the hadronic mass spectrum. Neglecting unessen-
tial complications the bootstrap equation should look like
where r is of the order of 1 fm. With

o
this yields
It£) ~ ,...,QX I(1;.f(E;)I7;€;T~

~, "~J
,,;,.Ej =E]. 0
(8)
Below 10 14 g the gravitational contribution can be treated as

a perturbation
224 W.NAHM
7: ({€) r:: E + G/"(E) -+ O(GV, (9)
yielding
~ Li h(E,)) (10)
that is
E~
h(€) ::
-~ID (11 )
Thus
E GE~
SeE) - t ~ rs (€ ) -::: -=;=: -I- i r. T:.
o • D
which interpolates between Eqs. (3) and (4).
To evaluate the decay probability into fireballs with ener-

gies E. we have to take into account both the phase space and
l
the final state interaction. In particular, the kinetic energy
of the fireballs should be sufficient to get them at least a
distance of 2r apart.
o
Now the preceding equations show immediately that for all

E. the value of ~f(E.) is approximately the same, such that
l l
phase space does not favour particular decay configurations, as
far as exponential factors are concerned. On the other hand, a
decay product of energy E1 needs at least a kinetic energy of
GE 1 (E-E 1 )/(2r o ) to get away, such that the decay probability
is suppressed by a factor exp(-GE 1 (E-E 1 )/(2r o To )). Thus typical
decay products have at most an energy
(13 )
Because of limitation of space a fireball cannot decay into

more than, say, 12 parts at once. The time per decay is cer-
tainly larger than ro' as the velocities are small. Thus
-dE
dt
~
2'17;
-GE ( 14)
This is rather slow, and in the last second much less than 10 9 g
will be freed. In the conventional model the explosion energy
will be higher, as already the electromagnetic radiation yields
a contribution
- ( 15 )
where 0= rr2 /60 is the Stefan-Boltzmann constant. Thus due to

the electromagnetic radiation alone, the black hole will evapo-
rate 10 9 g in its last second, which is more than for the
bootstrap, but still not very high. A few fields more yield a
factor of 10, but that's it. Thus for reasonable densities of
black holes, one has not much of a chance to see their "death
cries". Astronomers have tried hard to see the y's from the
explosions, but in vain. Still the searches have some use as
they put limits on the black hole producing fluctuations in the
early universe 8). But in my opinion one should decide about
such searches without taking the bootstrap into account, and
this for two reasons. Firstly and contrary to the present be-
lief of astrophysicists, the bootstrap would yield less violent
explosions than field theory, as I have tried to show. Secondly,
I see no experimental support whatsoever for the bootstrap type
226 W.NAHM
behaviour of hadronic matter at high energy densities, in spite

of the successes of the bootstrap at accelerator energies.
Before discussing this point, let me say something about

the application of the bootstrap to the cosmological entropy
production, which does not fare much better than the applica-
tion to black hole explosions.
3. - THE BOOTSTRAP AND COSMOLOGICAL ENTROPY PRODUCTION

Using the bootstrap, Carlitz, Frautschi and myself tried
to derive the density distribution in the universe and the
large entropy per baryon 4) Simplifying the arguments, we
reasoned like this. The causally connected pocket of our uni-
verse has about the correct mass for a black hole. From
Einstein's equations we know that this feature is approximately
time invariant. Thus let us go back to 10- 23 sec after the
big bang, where bootstrap hadrons just should come into being.
The mass of a black hole of this size is 10 15 g rv 10 39 Illrr' and
this will also be the approximate mass of our hadrons. Within
somewhat less than 10 39 x10- 23 sec they should decay into 10 39
particles, among them a sizeable fraction of photons. Then
thermal equilibrium should be restored. Before this restoration
the numbers of photons and baryons are approximately equal, but
afterwards the number of photons will have increased by a large
factor, which is to be compared with the observed number of about
108 or 10 9 photons per baryon.
Let us evaluate this factor. According to Einstein's equa-

tions, the volume increases like the square of the time. Thus
before restoration of thermal equilibrium the number density of
photons is 10- 39 fm- 3 , whereas the energy density is about
10- 39 m fm- 3 (both figures will be larger if the fireball
TT
decay time is shorter). Now the equilibrium number density of
photons increases like the third power of the temperature, which
is proportional to the fourth root of the energy density. Thus

during the restoration of thermodynamical equilibrium the pho-
ton number has to increase by a factor of somewhat less than
(10 39 )1/4, which is certainly in the right ballpark.
However, this scenario is in contradiction with many other

observations. Almost no primordial helium is produced, as the
radiation era is too short. We tried to transform this vice
into a virtue, noting that the gravitational growth of density
fluctuations are unimpeded, if there is no radiation to damp it.
However, we greatly overestimated the growth rate by an unjusti-
fied application of linear perturbation theory, as has been
pointed out by Peebles 9)
As a consequence, the observed large
inhomogeneities, which extend over 10- 2 of the radius of the
causally connected pocket of our universe, cannot have arisen
from a homogeneous initial state. After all this may be a good
thing, because initial inhomogeneities tell us a bit more about
what happened before the big bang.
As far as the photon number per baryon is concerned, there

is now hope for a better theory using initial conditions with
baryon number zero, and C, P and T violations arising natu-
rally in unified gauge theories 10),11). But even without the
problems concerning helium production, etc., we should have been
reticent to apply the bootstrap to cosmology. This may sound
strange, because Hagedorn's model was the first one to give rea-
sonable and quite often accurate prediction for hadronic cross-
sections, and is still widely and successfully used by experi-
menters. But I shall try to dissect the model into two pieces
with no deep connection between each other. One half alone yields
the experimental success, the other half yields the controver-
sial astrophysical predictions, and is responsible for the most
important wrong predictions at accelerator energies.
228 W.NAHM
My presentation of the model will be as elementary as pos-

sible. Moreover, I shall describe it not as it really came into
being, but in the dry, idealized, and inaccurate way such things
are usually described in textbooks.
4. - THE THERMODYNAMIC MODEL AD USUM DELPHINI

The success of the model depends on the correct description
of the longitudinal and transverse dynamics of hadron collisions.
The longitudinal dynamics has simply been fitted in a way yielding
Ilmiting fragmentation. The transverse dynamics depends on the
exponential mass spectrum of the resonances. These are the good
features.
On the contrary, the astrophysical applications depend on

the assumption of a universal volume for all hadronic resonances
and fireballs, which is (a) unreasonable, (b) unnecessary for the
empirical success of the model, and (c) not in agreement with the
observed hadron spectrum. On the contrary, strings and bag models
share the good features of the bootstrap, but avoid the bad ones.
Hopefully these models can be derived from QCD.
Let me consider these points separately. Early attempts to

predict cross-sections for hadron collisions in the 30 GeV range
on the basis of phase space alone failed completely. In particu-
lar the obvious peaking in forward and backward directions could
not be reproduced, but had to be explained from the internal
structure of the hadrons. Phenomenologically, the effect was
described by intermediate states with a number of fireballs
moving along the collision axis. For the moment we regard fire-
balls just as highly excited pieces of hadronic matter with no
important externally imposed constraints, such that for their
decays a distribution according to phase space is reasonable.
For heavy fireballs the recoil can be neglected, and the

decay products have a thermal spectrum, with temperature
T :=
where p(m) is the density of states per mass of the fireballs.
Now let us consider the one-particle spectrum coming from

a collision with total energy 2E. Apart from quantum number
effects, which are unimportant for our purpose, we have to take
a thermal spectrum in the rest frame of each fireball and to
transform it by a Lorentz transformation into the laboratory
frame. The fireballs have to have small transverse momenta, as
otherwise we could not describe the peaking along the collision
axis. For simplicity let us assume that their transverse momenta
are zero, but later we shall show that more reasonable values do
not change the results significantly.
Let us denote the thermal spectrum at temperature T with

G(p,T), where p is the momentum of the emitted particle. Then
wi th x= piE, where P is the fireball momentum,
(17)
Here f(x,m,E) is up to a normalization the mean density of

fireballs with given m and x. With L
we denote the Lorentz
x
transformation to the rest frame of the fireball. We still have
to determine one function T(m) of one variable and one function
f(x,m,E) of three variables.
I shall explain below how Hagedorn found that

230 W. NAHM
for sufficiently high masses, such that according to Eq. (16)
T(M) ~ To . (19)
Note by the way that a kinetic energYlof order To restricts

the momenta to values of order (MTo)2, such that Eq. (5)
yields
(20)
Equation (19) helps a lot, because we can integrate over M,

obtaining
where only one function of two variables remains to be determined,

namely
F (X, £") • ji {lei M, E} cIM. (22)
Now Hagedorn compared with the experiments and found that F(x,E)
is practically independent of E, if the collision energy is
high enough. Later this phenomenon was rediscovered by Benecke,
Chou, Yang and Yen and called limiting fragmentation.
Well, I have been cheating a little bit, because Hagedorn

did not use
x .. (23)
but
(24)
where v is the fireball velocity, y the corresponding

Lorentz factor
(25)
and Yo the Lorentz factor of the incoming particles. For

large y this makes no difference, as the variables x and
A are approximately equal. For small velocities, A is of
course a poor choice, as it depends non-analytically on v.
However, the central region is anyhow better described in terms
of the rapidity, which for vanishing transverse momenta is just
(26)
Using this variable, one obtains experimentally a fiat distri-

bution near y= O. This means local invariance of the hadronic
matter with respect to longitudinal Lorentz transformations.
Therefore such a distribution stays flat, when the fireballs
decay.
In terms of A such a flat distribution in y yields
At 30 GeV this function should be reasonable up to A = 1/3,

one unit of rapidity away from the maximal rapidity. Indeed
between A=0.02 and A=0.3 it agrees well with Hagedorn's
F(A), of course up to a normalization. However, Hagedorn kept
his F(A) finite at A = 0, which introduces at y =0 a spu-
rious hole into the y distribution. Moreover, below A = 0.1
an energy dependence should be taken into account.
These unrealistic features of the longitudinal dynamics

can easily be remedied by working with y instead of A 12)~14),
232 W. NAHM
which indeed is done at ISR energies. The unrealistic assump-

tion of zero transverse momentum of the fireballs also causes
no problems. From the decay recoils the fireballs will anyhow
acquire some transverse momentum which approaches a thermal
distribution. Thus one may as well assume a thermal distribu-
tion from the start, which is by far enough to fulfil the un-
certainty relation. Even then the corresponding transverse
velocities will be so small that one can forget about them.
In such a model it should not make much of a difference

at which generation of fireball decay one starts. In particu-
lar the individual contributions of high mass resonances are
washed out, though it is useful to take the very last generation
of resonances explicitly into account. The first generation of
fireballs are not even well-defined entities. Regarding the
hadronic matter created by energetic collisions one just con-
siders regions which are small enough such that no longitudlnal
peaking of the relative velocities is marked. But the hadronic
matter need not split into such parts before particles are
emitted.
Such a behaviour of the longitudinal dynamics had to be

assumed in Hagedorn's model, whereas nowadays limiting frag-
mentation and a flat y distribution can be derived in bag and
string models 15)-17). The other basic ingredient for the suc-
ces of the model is the exponentially increasing mass spectrum
of the fireballs. On the contrary, nothing need be assumed about
their size, except that it should be not too large, such that the
states can be produced in typical collisions. This is not an im-
portant restriction, as the mean observed size of fireballs
("clusters") is anyhow only 1-2 GeV, independent of the collision
energy.
One might even argue that at these low masses the exponen-
tial increase is just an accident, and that it would be suffi-
cient to use just the resonances of the naive quark model. How-
ever, the fireball excitation function f(x,M,E) might well have
a tail reaching to much higher masses. In particular one should
note that the thermodynamical model allowed a successful calcu-
lation of the production rates of anti-deuterons and anti-He 3 18)
In QeD one may also expect the production of larger fireballs
in collisions where multigluon exchange yields flux tubes with
a colour flux in a higher SU(3) multiplet.
To gain a better understanding of the significance of a

fixed fireball volume in Hagedorn's bootstrap, let us look at
his derivation of the exponential mass spectrum, and let us
compare it with the collective excitations of strings and bags,
which also yield an exponential spectrum. Hagedorn made one
basic assumption which was hotly contested in the sixties, but
seems to be universally accepted today: there is no qualitative
difference between resonances and fireballs.
Thus we may build up the fireballs inversely to their

decay by letting the decay products resonate with each other.
Both Hagedorn's model and the bag and string models assume that
the interaction is rather soft : for a given location in phase
space of the constituents the resonance is essentially uniquely
determined and the binding energy is not large. This means in
particular that P(M) cannot increase faster than exponentially,
because otherwise
(28)
for high masses.

234 W. NAHM
That hadronic interactions are soft, in other words that

there are no strong forces involved, is up to a point well born
out by the experiments, as large transverse momenta are rare.
Note also that for a spectrum increasing faster than exponen-
tially the time for every decay step of a fireball would increase
with its mass, as the phase space for the decay products would
be smaller than that of the fireball itself. Pion emission
would dominate, thus the decay time would behave like
t ",...., ,
whereas evidently no hang-up in fireball decay is observed. On
the contrary, for black holes just such a phenomenon is to be
expected, as according to Eq. (3) log p(M) is proportional
to M2.
In general one might expect that soft interactions yield
as states of mass M1 +M2 usually can be obtained in very many

ways by putting states of masses NI1 and M2 together. For
example in a pion gas one may distribute the pions in many dif-
ferent ways into two packages. But if the inequality (30) is
true, p(M) increases slower than exponentially. Only if the
fireballs somehow remember the way they have been put together,
one may have an approximate equality in (30) and correspondingly
an exponential spectrum.
So far the agreement between the conventional theories and

Hagedorn's bootstrap, but now comes the big difference. If one
considers collective excitations in string or bag models, the
size of a fireball increases proportionally to its mass. Thus
if one puts two fireballs together, they still occupy different

volumes and the parts still can be identified to some extent.
Instead of the inequality (30) one therefore obtains an appro-
ximate equality. In other words, the entropy is an extensive
quantity, just as for ordinary matter
(31 )
Therefore the exponential increase of p(M) is a rather trivial

geometrical effect, which does not say much about the equation
of state for high densities. Consider for example ordinary water
in a state dominated by droplets. Volume and mass are propor-
tional, and their density of states per mass increases exponen-
tially. The entropy of the total system can be obtained in the
usual way from the entropy of the droplet. But now increase
the density until the droplets fuse, and still further: one
obviously has to switch to a completely different description
of the system.
Of course one may use slightly different versions of this

geometrical effect, having either a constant energy per volume,
a constant surface tension, or a constant tension along a string.
The difference matters only if one calculates the power of M
in front of the exponent in p(M). Configurations of open
strings are the easiest to count. Adding one unit length to
the string, the number of configurations increases by some
constant factor
c (AE) o-(E) I
such that the denSity of states with energy E is a pure

exponential
236 W. NAHM
For open surfaces one can add unit pieces of surface anywhere
along the boundary, which yields some multiple counting. If
~(E) is the length of the boundary, one finds
.1.
For smooth surfaces, should increase like E2, such
that
(35)
Similarly for volumes
(36)
p(M) is always given by Eq. (20).
A geometrical effect of just this kind is responsible

for the exponential increase of the mass spectrum in the old
Pomeranchuk model of fireball decay, as he also considered
fireballs with a volume proportional to their energy 19),20).
However, one again obtains a different power of M in front
of the exponential, as he admitted arbitrary distributions of t
the energy within the volume, but fixed the geometry of the
volume itself.
In contrast, Hagedorn does something completely different.

In the original version of his model, he put all the degrees
of freedom into one volume of fixed size, with a radius given
by the pion mass. In this model one cannot simply explain

why the hadrons remember how they have been put together, one
has to postulate it. In other words, as, e.g., a p(rrrr) re-
sonance must be different from the corresponding (prr)rr reso-
nance, even if the momenta involved are the same, one must
postulate that the formation of resonances is not associative.
Describing strings in this language, one has associati-

vity but not commutativity
Putting the pieces together in the order ACB yields obviously

a string different from ABC, but (AB)C and A(BC) are the
same.
In my opinion, the assumption of a constant volume of all

resonances cannot be maintained. On the one hand one cannot
identify a dynamic mechanism which confines them to one fermi
cube, as large forces are incompatible with the assumptions of
the model. But the assumption of a constant volume is also
impossible to reconcile with the observed linear Regge trajec-
tories.
Consider the expectation value of the orbital angular

momentum
)(f~ '"
..l ...J ..
.( &:
~ r
'"
-->n th
where p is the momentum on the n constituent. Obviously,
238 W.NAHM
(I) (38)
where R is the radius of the system. Thus for a constant

volume the maximal angular momentum of hadrons of mass M
should be
;v,M (39)
J
if the constituents are enclosed in a potential well of infi-

nite height. For a confining potential V(r)rvexp(r!ro) the
maximal angluar momentum is somewhat larger, but only by a
factor log(Mr o ), as not much phase space outside r=r o is
available. On the contrary for strings and bags
(40)
and
(41 )
as it is observed. Thus for hadrons on the leading trajecto-

ries one sees directly that their energy density does not vary
strongly with M.
20)
Gorenstein, Miransky, Shelest and Zinoviev ,and
Hagedorn, Montvay and Rafelski at this workshop introduced
versions of the bootstrap in which both non-associativity and
a volume proportional to the fireball mass contribute to the
exponential rise of the mass spectrum. Such models are threa-
tened by Occam's razor, but there are more serious objections,
too. The volume effect alone yields one non-degenerate leading
Regge trajectory. If one now adds non-associativity, the mul-

tiplicity along this trajectory should increase rapidly, in
contradiction to the observations. One can also argue, of
course, that deep inelastic scattering, large transverse mo-
menta, W spectroscopy, etc., more and more reveal a rather
conventional internal dynamic of the hadrons, which is diffi-
cult to reconcile with non-associativity at distances of the
order of one fermi.
5. - CONCLUSIONS
When people tried to apply the bootstrap to hadronic
matter at large energy densities the results have been wrong
(as for black hole decay) or inconsistent with the observations
(as for entropy production in the early universe). There is no
point in looking for improved applications, as an analysis of
the foundations of the bootstrap shows that its fireballs should
be objects with an energy density of order 1 GeV/fm 3 • We have
to accept that our laboratory experiments yield no direct obser-
vations of the states of high energy density relevant for cos-
mOlogy. Of course, states of high energy density are initially
created by colliding Lorentz contracted protons, but they are
far from thermodynamic equilibrium.
Nevertheless, the bootstrap might yield interesting phe-

nomenological models for phase transitions near energy densities
of 1 GeV/fm 3 • Below this energy density one should have a gas
of hadronic resonances and bound states. Near the transition
point they should cluster to larger and larger fireballs, and
finally fuse. Above the transition point one has to describe
the state in terms of the hadronic constituents. This direction
has been explored by Cabibbo and Parisi 21), and by Hagedorn,
Montvay, Rafelsky and Satz at this workshop. This is the density
regime where the bootstrap may have useful astrophysical applica-
tions.
240 W. NAHM
REFERENCES
1) V. Canuto - Ann.Rev.Astron.Astrophys. 12 (1974) 167

13 (1975) 335.
2) H. Grote, J. Ranft and R. Hagedorn - Atlas of Particle

Production Spectra, CERN (1970).
3) M.B. Kislinger and P.D. Morley - Phys.Rev. D13 (1976) 2765.
4) R. Carlitz, S. Frautschi and W. Nahm - Astronomy and

Astrophysics 26 (1973) 171.
5) S.W. Hawking, Commun.Math.Phys. 43 (1975) 199.
6) W. Nahm - Nuclear Phys. B68 (1974) 111.
7) S.W. Hawking and D.N. Page - Astrphys.J. 206 (1976) 1.

8) B.J. Carr - Astrophys.J. 201 (1975) 1.
9) P.J.E. Peebles - "The effect of a lumpy matter distribu-

tion on growth of irregularities in the expanding universe"
(October 1973), unpublished.
10) M. Yoshimura - Phys.Rev.Letters 41 (1978) 281.
11 ) J. Ellis, M.K. Gaillard and D.V. Nanopoulos - CERN Preprint

TH. 2596 (1978).
12) C.E. DeTar - Phys.Rev. D3 (1971) 128.
13) L. Van Hove - Physics Reports 1 (1971) 347.
14) J. Ranft - Phys.Letters 41B (1972) 613.
15 ) F.E. Low - Phys.Rev. D12 (1975) 163.
16) J. Kogut and L. Susskind - Phys.Rev. D9 (1974) 3501.
17) X. Artru and G. Mennessier - Nuclear Phys. B70 (1974) 93.
18) R. Hagedorn - Nuclear Phys. B24 (1970) 93.
19) l.Ya. Pomeranchuk - Dokl.Akad.Nauk SSSR 78 (1951) 889.
20) M.l. Gorestein, V.A. Miransky, V.P. Shelest and

G.M. Zinoviev - Phys.Letters 45B (1973) 475.
N. Cabibbo and G. Parisi - Phys.Letters 59B (1975) 67.

PARTICLE PRODUCTION IN SOFT AND HARD HADRONIC COLLISIONS.
IS THERE EVIDENCE FOR HADRONIC CONSTITUENTS?
Johannes Ranft
Sektion Physik, Karl-Marx-Universitat,
Leipzig, G.D.R.
ABSTRACT
After a short review of some properties of hadronio
multi-particle systems which show the presence of a
statistical mechanism, we describe particle production
in soft hadronic collisions using the quark recombi-
nation model. Finally particle production in hard
collisions, in particular large transverse momentum
processes and the occurence of gluon jets in deep
inelastic processes is discussed within the framework
of QCD perturbation theory.
1. THE STATISTICAL DESCRJPTION OF HIGH ENERGY MULTI-
HAVRON SYSTr;MS, SUCCESSES AND FA ILURES
I will discuss multihadron production in hadron-
hadron scattering in the framework of the Statistical
Bootstrap Model /1/, the Thermodynamic Model /2, 3/
and the Independent Cluster Emission Model /4/ with
clusters decaying according to the SBM. Only a few
characteristic sucoesses and problems of these models
are discussed.
Suocesses:
(i) Limited transverse momenta of produced particles
following from the finite temperature T ~'30 -
180 MeV belong to the firmest predictions of
these models and have been observed at least in
241
242 JOHANNES RANFT
the range P"...c. 1 GeV up to the highest primary

hadron energies accessible to experiment so far.
(ii) Predictions for particle production ratios in
the Thermodynamic Model are due to the term
i
exp( - p~ + m2 iT).
It was shown in I~/ t~at the
average multiplicities ot produced 1( - , K-, P and
p agree very well w~th the data in the range
10 ~ s ~ 2000 (GeV) •
(iii) The Thermodynamic Model predicts in the central
region deviations from Feynman scaling at finite
energies. These deviations are essentially due
to the rise of the temperature T with riSing
energz density. Especially for heavy particles
like p, where these deviations are predicted to
be largest, they agree rather wei~ with experiment
13/.
(iv) The dependence of average transverse momenta on
the mass of the produced particle was predicted
in /2/ to behave like
< r(om.) >
L
~_&)O
.. ,1 ~ """T •
,
( 1)
_I 2 2 2'
due to the term exp(--Yp + Pit + m /T). For Pions,
Kaons and Antiprotons, the only particles where
data were available, the agreement with experi-
ment was good. Also the production of the J,~
(3.1 GeV) particles was consistent with this
rule /3/. Recently it was found by Satz /5/
that even uncorrelated multihadron systems and
~epton pairs produced in hadron collisions
behave like (1).
(v) It was first found from the analysis ot inclusive
and semiinclusive two particle correlations:
that particle production in the central rapidity
region is dominated by central clusters or fire-
balls with masses <m>",1 - 2 GeV /4/. In the
recent years more and more experimental evidence
has become available, that particle production
in the central region is largely via the pro-
duction of well known resonances.
(vi) The effeots ot. Bose statistics on multihadron
systems produoed in hadronio ool!isions has
been predioted since long /6/. Two particle
IS THERE EVIDENCE FOR HADRONIC CONSTITUENTS? 243
correlations R(Y1' Y2) = '-P2(Y1' Y2)/ <f>1 (Y 1)·

4J 1 (Y2) - 1 for identical particles like Jr-Jr-
were predicted to differ in shape from 'JT+Jf-
correlations (which are dominated by cluster
production) and increase with decreasing
azimuthal angle ~~ between the two ~-/7/.
Beautiful experimental data are now available,
which show the presence of this effect /8/.
Problems for statistical models:
(i) The leading particle effect and in particular
the quantum number dependence of particle pro-
duction in the fragmentation region are diffi-
cult to be described in the above mentioned
models. When a description of these data is
possible as in the Thermodynamic Model, then
only by using phenomenological fireball dis-
tributions.
(ii) The production of particles with large transverse
momenta was not predicted and is difficult to
describe. It is very doubtful that the event
structure of large PJ.. react ions can be described.
(iii) The jet structure observed in the reaction e+e-
hadrons cannot be understood.
Conclusion
Multihadron production at p~<1 - 1.5 GeV/c shows
many characteristic features which suggest the presence
of a statistical mechanism. However there are many
features, especially in hard processes which point to
an underlying dynamics of constituent interactions.
~herefore a better understanding is to be gained in the
future of the relation of this underlying dynamics to
the statistical mechanisms and of the regions where
the constituent dynamics on one side and statistical
processes on the other side dominate.
2. IS THERE EVIDEUCE FOR HADRONIC CONSTITUENTS FROM
SOP!' HADRONIC PARTICLE PRODUCTION PROCESSES?
THE QUARK RECOMBINATION MODEL
No predictions for QeD are available for particle
production in soft hadronic collisions. The quark-
gluon constituent picture can therefore in this region
only be used on a phenomenological basis. Here we want
to study particle production in the hadron fragmenta-
tion region using the quark recombination model. As
244 JOHANNES RANFT
phenomenological input, only quark and gluon dis-

tributions are to be used as determined from hard pro-
cesses.
The longitudinal momentum distribution of leading
protons in low PoL proton proton collisions were first
interpreted by Pokorski and Van Hove /9/ as due to the
recombination of three valence quarks of the original
proton. The relevance of quark.. partons for particle
production in the fragmentation region was also dis-
cussed by Ochs /10/. Das and Hwa /11/ explained the
production of mesons in the fragmentation region of pp
collisions as due to the recombination of quarks. This
model was subsequently used and refined /12, 13/ and
applied to Baryon production by Ranft /14/. The model
as presented here /15/ has not yet been fully des-
cribed. Other models for low PoL particle production via
quark-partons were discussed repeatedly /16/; no
attempt to review all the different approaches will be
made here.
2.1. The Formulation of the Quark Recombination Model
Das and Hwa /11/ considered the recombination of a
meson out of' an u valence quark and a d sea quark.
The inclusive 11- production cross section in the frag-
mentation region is obtained as
a()
-d
I =
r d.l<,.
fX-
cix.:z.
-x- 2J
F(X-f/ X '"R(x",Xz./X) .
( 2)
)( f'1" ~:1f+ 4 Z.
Here x is the Feynman x of the pion. x:J and x 2 are the
momentum fractions of the uand d quarks, F(X" X2) is
the two-quark distribution runction and R(x 1 , x 2 ' x)
is a recombination function.
The quark-antiquark recombination fUnction is
assumed to scale in li = Xi/X and it is assumed that
the recombination out of just one quark-antiquark
pair dominates. Das and Hwa /11/ use the simple
symmetric form
(3)
'R (~ ~J ~")2.)
')
= tXT :E -r ~(~"::>1\ +~":>:z.. _/1)
,)" "::>2.
where the parameter <X is detennined from the sum rule
" If
Jd~. ~d~1. 1«(~,>~.) ~"1

o
(4)
and becomes r/..r= 6. It is easy to check that the powers

of ~1 and ~2 in (3) dO not influence the resulting
pion distribution strongly.
Das and Hwa use a factorized two-quark distri-
bution function
( 5)
I="z. (X., J x~ -::: ~ J:"t.c., va.! (x,,) ~ (Xz) (A - X.t - x~
where the Fj(x)/x are the single quark distributions,
B is a cons~ant and (1-x1' -x 2 ) is a phase space factor,
Inserting (5) and (3) into (2) we obtain
d6" x (6)
d. x
pp-+Jr+ ~
~
= o(lr " : x (d. x., F\.\. v"l (x-t') l="_ ()( - X',,)
I ... d. .
Das and Hwa /11/ and Duke+and Ta~lor /13/ find that
the production of ~~~, K and K can be well des-
cribed by this method using reasonable quark dis-
tribut ions as input.
This method was generalized by Ranft /14/ to
baryon production
a.cr I
cJ..~ f r =
~ d..v..1 d...~2... clx:3
x-::-
X" 3 -x F3 (x,,/)(2. Xa) ~3
(7)
(><,.,)<2. X""x.\
-+ 10 -, I I .. 'J
where H3 is a recombination function parametrized as

(8)
with 0<. = 120 obtained from the normalization condition.

Also h~re the resulting baryon distributions do/dx
do not change noticeably if higher Ii powers are used
in (8). Assuming a factorizing three quark distri-
bution one obtains finally
do-I
d'l< =o(b ~"
(oi-X)
2.
t
(J
J(
~
jrAX-"
x' X1
('Ie.,) d)(2. ~("2) ~()( - x" -)(2.) .

(9)
PI'''' to X 0
o
The production of protons, neutrons and A hyperons
could well be described /14/ by this method.
246 JOHANNES RANFT
These applications of the quark recombination model

indicated that the method is rather promising. One
finds however easily two unphysical features of this
scheme.
(i) The factorized mul tiquark distributions used do
not fulfil sum-rules like
5F2.(l<.1.; XzJ dKz. = ~('j(A) .

( 10)
~ear to the kinematic boundary the model becomes

therefore unreliable.
( ii) According to (6) all meson distributions behave
in 'the l1mit x ~1 wrongly as (1 - x). This factor
comes from the phase apace term in (5) and is
conneoted to point (i) above.
These two problems were the reasons to introduce
well defined multi-quark distributions into the quark
recombination model /12, 15/ as described in the next
Section.
2.2. Multi-Quark Distribution Funotions Calculated
from a Generalized Kuti-Weisskop:t' Model
The original Kuti-Weisskopf model /16/ gives too
flat sea quark distributions; therefore we have to
generalize it. In order that all momentum sum rules be
fulfilled properly we start from expressions for the
exclusive n parton density, Where the momentum sum
rULe can be imposed by a I-function.
f
a ~~.
d'P""(lI.1J ".))(,,,,)=.2CPo))f r .(x.)1\.L , r .(", ).(11)

1=" t V,I I .
~=1
""'J' ' ""'&1
t£ IJ ""
n d. ....
,~ ~ 5(11- ~ ><j) ,
l"''' "j J:.-1
We have ns kinds of sea part one (quarks or gluons).
The 'Cotal number of' partons is
111\. =. 3 + 1IIl... .j.. 11It..2,. 1- .•• of M..."'s (12)
The f' ,1(X) are valence quark matrix elements, the
fs i(~) are sea quark matrix elements. Different f'rom
Kut~ and Weieskopf /16/ we ohoose
rs . (x) =- Q~ S. (l<.) . (13)

\' " -a" )
where g 2 is a cou~ling constant and s~(x) a function.

The dis!ributions (11) define the "excIusive" n parton
densities in an uncorrelated jet model applied to
partons instead to produced particles. In order to
obtain "inclusive" n parton distributions we have to
integrate over all x variables of sea partons and sum
over all n, each n has to be summed from 0 to~. The
calculation of sucH phase space integrals is well
known and we do not give all details of the calculation.
Furthermore we restrict ourselves to only one kind of
sea parton. The distribution of n sea partons is
('P.) -""" ... ( 14)
II
~
d'P"'(><iI, ... jX""x)=: ," tC"j)dl'j cS('-t- LXj -x)
""'. j=.o Xj j"''' I
where Z(P ) is a normalization factor introduced to

cancel thg primary momentum dependence of the integral,
x is the sum of all xi of valence partons and sea par-
tons lett out of the Integral. Integrating over all xi
and summing over n we write
,2._ A
¢C-t- x) = (-t-x)
IV
d>C-1-)() =: (15)
= ~
1>11 -+ ~
2 ("Po) f
"=0
(~"Lt· I ex
"""" h I
"Po)
with
(V
t\- It
$:- 1(;) .
~-)(
I/W. ()()~) = ( ... SC)(.;) &"'i d"(A-)C - (16)

J)
o Q.
'="1 XI .... "
The lower integration limit a ~O( 1/P ) gives a week

momentum cut-off. After some calcula~ion we obtain
~(H) !CDS[V - fS(~-. ~-1]e'~rkc(.v-. 4-X)] d.V

c
-00
J J
(17)
with the definitions

(iI-X)~
S(~
'))
.f - }() ... - i) 0
S( ~)
~
.. i-. \A. d~
(18)
and
1-X
2'(~}-1-X):'~· S~)c) (t>'r(-i~)()d)C _ h.L-1~)(] (19)
Cl.
We insure the proper normalization of' the parton dis-
tributions
~ ~cLfM.lX.,) ... ) X/~J == -1 (20)
248 JOHANNES RANFT
by the normalization factor
~A ~ Jrf'{J
3
Jf~1
I
d.;,i
I
r ,(Xi) \f~(I\_£ Xi)
til',. I'=-'\
(21 )
We obtain all possible single and multiple parton dis-

tributions as nroducts of' matrix elements and phase
space integrals.
We give a few examples:
1 valence quark
r
l.t(x) .: tlA.
I
V CX) 1\ )(elx
~
, . "':..,
2. -
Xct.
r V(><Iot,)
"" d..t.d, t",
2., l:
t (KiA.)
1..,11
~ ('\ - x -)(\4. -
.a..
(22)
xJ
~
1 valence and 1 sea quark
F(X k ,.) X;;:,,) = f ..,v (){ ...) fJ/".r,s) t\ Sd: . . ~.( ""2.. t(
(23)
j
~ (1\-'1( v..~v -X'",-2.. -}(J)
,tt.c.;,.., v (l(\.t;L) TcrdV(X.J V) '1' r.{, 11\.1'
3 valence quarks
( 24)
1 sea quark
S (x) ~ ts (x.) 1\ (d.K"'1\ lAx: ""a.. cil<'tA,. r (x·) ( eX) (25)
J X',.
-"
l(
t.\~
)(~ t'\ V
CII.'
14.,\ t\.t, v ,
IA.~.
· +c1 lf(Xd,)t cp (t\ - X - xlA.", - x""~ - X~) ,

Computer programmes are available which calculate all
the phase space integrals efficiently. We are now in
a position to choose all matrix elements in such a
way that the quark distributions calculated agree with
the ones measured in hard collisions.
The original choice of Kuti and Weisskopf was
f,,(x) =X
of - U(. to)
j
r
ts Cx) =~1.,
(26)
With this simple choiiCe all integrals can be evaluated
:in closed form
(27)
but the sea quark distributions ~ecome too flat, this

is easily visible by inserting g = 3 and ~(o) = 1/2.
We described the difference between the u and d
quark distributions and the steep sea distributions
following from experiments in lepton pair production
with the following choice.
f"" (x) == X" - oax) (28)
fd. <-x) = )(1-«(X) (~-x)[" +-A VX (O\-K)1
t sex) - ~~ (" - x) It;

By a suitable choice of the parameters g2=2g~ + 2g~ +
2g; + gglue' Kg1Ue ' Ksea and A it is possible to fit
the deep inelastic structure function as measured at
SLAC. This fit is insensitive to the shape of the
gluon distribution, therefore we fix Kglue such that we
obtain Fglue(X) ",(1 - x)5; The fit is also rather in-
sensitive to the shape of the sea quark distributions,
theret"ore Kses. could be fitted trom data in lepton pair
production. Following Duke and Taylor /13/ we determine
Ksea from the hadron distributions in the next Section.
2.3. Inclusive Single Particle Distributions in the

Quark Recombination Model
Here we can only describe one or two problems in
detail and leave most details for future publications
/15/. We consider the ratio ot' ~/Jr- mesons as function
of x in the quark recombination model
(29)
~;_lpp ... 7-
= ----------------------
¢~"'-(I\-)() ~j(tc.t(J()+~(x.-X")c!)(,,
where the phase space integrals th 0j (1 - x) are de-
fined as follows ~~
250 JOHANNES RANFT
Ph (t:A.'lCj tJ,.'lCj r .r ~
41ij{oi- lC ) =j lei 'lCj ti(~i) tj(~j) yr(I\-)C - X, - Xj) (30)
The J(" +/Jr- ratio (29) depends exp1icite1y on the sea quark
matrix elements. Therefore we can fit the experimental
data to fix Ksea/13/. In this way we obtain a good des-
~~il~!~nb~~a!~gd~l~~ :(~)~if1_!)a§~9s~~n~~~:n~i:l~!-
data on lepton pair productio~. After this fit also the
shape of the Jrfo';r:']r distribut ion is very well a.escribed,
see Fig. 2 and no free parameter is left in the model.
A second problem is the normalization of the sea
quark a.istributions to De used in the quark recombi-
nation model /13/. In hadronic collisions, a long time
is available to form the resulting hadrons. During
this time the gluons fluctuate continuously into quark
antiquark pairs and can contribute as such to the re-
combination into hadrons. Therefore we choose in the
model only to keep two parton recombinations into mesons
and three parton recombinations into baryons but with
enhanced sea quark distributions, enhanced so much that
they carry the momentum of the gluons as well.
pp-+l[tx PT .0.8
!.i=45
10
/
j
8
.. :/
~
!)~t
II
....... V'
.- ~~
t#
o
o 0.5
x
~ig. 1. Experimental data for the ~+/~- ratio measured

at the CERN - ISR and at Fermilab are compared
with the 1('t/Jr- ratio in the quark recombi-
nation model.
pp ... It+X ••• PT' O.~5

/.i.45 ••• 0.85
10' .'--
.:---r--.
1'--r-... ~
:-... r-.. ............ ~
~ ~~
~ "
~1\
IAi'
i\ .\
~
O.S
x
Fig. 2. Comparison of the Feynman :x dependence of :.-+

production at the CERN-ISR with the prediction
of the quark recombination model.
Resonance production is known to play a large role

experimentally. The quark recombination model can be
formulated in such a way that a quark-antiquark first
forms a hadron state, say ror mesons out of the two
lowest nonetts, the pseudoscalar and vector nonett,
and that the resonances decay subsequently into stable
hadrons. It can be argued that the good description of
resonance production is more important f"or the model,
since these are the originally produced particles and
their distributions are not influenced by decays.
To describe hadron production in meson hadron
collisions we need the quark distributions in mesons.
These distributions are more difficult to measure in
hard collisions, only some data are available from
lepton pair production. However we might use the quark
recombination model and experimental data on hadron
production in the meson fragmentation region to learn
more details about quark distributions in mesons.
2.4. The Quark Recombination Model and Correlations
Between Hard and Soft Hadronic Processes
Correlations between particles produced in the
hadron fragmentation region and hard processes play an
unique role in testing the physical picture of the
quark recombination process. The quark recombination
252 JOHANNES RANFT
model makes definite predictions for these correlations.

Relevant experiments include
(i) correlations between deep inelastic lepton-hadron
scattering and hadron production in the target
fra~aentation region
(ii) correlations between lepton pair production and

low p~ particle production in both fragmentation
regions
(iii) correlations between large p~ particle production
and low p~ particle production in the hadron
fragmentation regions (spectator jets).
A few examples are the following :
Antineutrino scattering on protons proceeds according to
( 31)
At small XBj the scattered u quark is likely to be a sea

quark and tfle particles produced in the proton fragmen-
tation region are expected to have a similar composition
as in low p~ hadronic reactions. At large values of xBj
the scattered u quark is one of the valence quarks. <
Il1he re fore t the system fragmenting contains only u and
ona d valence quark, and fragments into equal numbers of
Jf+ and ;rr-mesons. In J!iig. 3 the JT+/Jr- ratio predicted is
plotted as function of the momentum fraction z and for
different values of XBj •
Correlations between massive lepton pairs and
particles produced in the fragmentation region were al-
ready predicted in /12/ and /14/. Here I diseuss results
obtained with the generalized Kuti-Weisskopf quark dis-
tributions. According to the Drel1 Yan model
( 32)
1"<-
1"- .
•
If
____ iir
==~---===~------vp x-o..,
Fig. 3. 'I'he 1t+ IJf- ratio in the proton fragmentation

region in deep inelastic antineutrino proton
scattering. vp_ fU+JT~X as predicted by the
quark recombinat~on model :for different
values of' X:sj.
lepton pairs are created by quark antiquark annihilation.

'l'ne u quark nas tne ~argest coupling to the photon. At
large values oiL= Q Is valence u quarks dominate the
lepton pair production.
r
J[-
/ klif;.h,,,,,J;
10 PP-Jr!·x It..r~"
/""'" tr.~~'r
t..rf-o" rair r -M'
s-
..,.. ,"£..sO..,,,
/
G,
4 1"'''''0.0-1
2-
~-o.o.
--... J:.O.11
J .. o.1b
0
0 ·l .~ ·6 ·8 1.
X
:Fig. 4. The Jf+ I Jr- ratio in the pro_ton rragmentation

regions in reactions pp_le J[tx triggered by
a mass ive lepton pair as preo.icted by the
quark recomoination model f02 dirrerent
values of the parameter r= Q Is.
254 JOHANNES RANFT
Consequently in the parton fragmentation regions

the X+/JI'- ratio will decrease with rising t;.
In Fig. 4 the 7+;3(- ratio is plotted as function of the
Feynman x for different ~ values.
Correlations between large p~ jets and leading
partons were studied in detail in /4/. The comparison
with existing experimental data was encouraging.
2.5. Summary. Quark Recombination
There is good evidence that hadron production in
the fragmentation region is related to deep inelastic
structure functions. The following tests and pre-
dictions of this approach are significant :
(i) The old independent cluster production model is.
recovered, resonance production is predicted in
detail. The progress against older formulations
of such models is in the detailed prediction of
the quantum number dependence of particle pro-
duct ion.
(ii) The model can be tested through correlations
between hard and soft processes.
(iii) The production of resonant and nonresonant two,
three, ••• particle systems can be predicted and
seems to agree well with recent ISR measurements.
(iv) 'l'he model offers a method to detennine quark dis-
tributions in mesons from low P1 data.
Problems which are not well understood presently
include the following :
(i) How can such a model be justified from QCD ?
(ii) What is the origin of scaling violations at
low P,l. •
(iii) What is the interplay between quark recombination,
which seems to domlnate in soft processes and
quark fragmentation which dominates in hard pro-
cesses.
3. EVIDENCE FOR QUARKS AND GLUONS IN HARD PROCESSES
In this Section we discuss the fOllowing problems:
How to calculate hard scattering cross sections from
QeD perturbation theory. QUD perturbation theory is
applied to particle production at large p~. It is found,

that present data are consistent with the ~CD predictions
and predictions for further checks are made. Finally
we discuss the QCD prediction that gluon jets shOuld
be observed in deep inelastic lepton hadron COLlisions.
3.1. How to Calculate Hard Cross Sect ions from QCD Per-
turbation Theory. Factorization of Mass Singu-
larities
Hard scattering cross sections, like the production
of large p~ jets or lepton pairs in hadron collisions
r
are calculated in the quark-parton model from the e%-
pression
dO-A" = L~,r..
d.. a. dx" F"'A ix..) F.,.. t~.) d......... (3)
where Fa/A(%a) is the probability to find parton a with

momentum fraction %a in hadron A. These distributions
cannot be calculatea. f'rom QeD presently. dO"ab is the
hard parton-parton scattering cross section, which is
to be calculated perturbatively.
If we cons ider as an e%ample the "production of
lepton pairs in hadron hadron C011is~ons in low~st 2
order in the QeD coupling constant g 1/10g (Q /1\ ),
AJ
the f'ollowing contribut ions should be considered
Bom term qq....., {[ (34)
Order g 2 : qq
- .....,fl+
-
(35)
--"""---r
256 JOHANNES RANFT
From configurations in the graphs (35) where colinear

quanta are2em~tted, mass singularities arise behaving
like 10g(Q /m ). ~he pe~u2bation series becomes a
power series in glog Q /m which cannot be expected to
converge.
Using the leading log approximation it was round
during the last year /17/ that these mass singularities
can be absorbed into the parton distribution functions
and parton fragmentation functions. This leads to a
redif'inition of the parton distributio~s and parton
fragmentation functions which become Q aependent with
the same dependence as originally found from the opera-
tor product expansion and renormalization group for
aeep inelast ic lepton hadron scattering.
From these areuments the following changed pre-
scription for the calculation of hard. scattering cross
sections from QCD perturbation theory emerges
d"A ... ~ ~ ~ cI. x.. G( Xb 1'.;.,,,(. ~ p') F.,,, ex., Q') a.cr-" b
(36)
I
only the Born term cross section is to be used for the

rard scattering cross sect ions dO"ab.
Calculations using besides the leading log terms
also the next to lead,ing log terms /18/ have recently
been reported. These calculations indicate that in
some processes corrections to (36) might become im-
portant.
3.2. Hard Scattering of Quarks and Gluons According to
QCD, Large p../. Single Particle Distributions,
Effective .l:'ower Behaviour and Inf'luence of' Parton
Transverse Momenta.
The QCD predictions for single particle distri-
butions at large P.L. were first calculated by Cambridge,
Kripfganz and Ranft /19/ and Cutler and Sivers /20/.
'11he quark-quark, quark-gluon and gluon-gluon scattering
cross sections which dominate the large p~ particle
production are the following /19/
a .•
,
QJ' _ Q j Q
I I I
J. cJ.ts- =-"",l..irs'L+v.:2.. +~ .. SZ.TtL_d.. ~
c;(t sz.. <j L tz. u 'J 3
tA,:L.
52.1
v...t
(37)
d.Js- ::. TrO<l..[_ t u.l..tS'2.. i- Lt,l..4-S2..] (38)

olt: s 1. 9 l4..5 e.
The running coupling constant

odQ2.) -= 12.r (40)
2.5" h (Q2./I\2.)
is used and quark and gluon distributions F(x,Q2) were
used with deviations from scaling as predicted by QCD
/21/.
Due to the gluon contributions, the single par-
ticle spectra are larger and steeper in P.L than accor-
ding to the quark-quark scattering term only.
The Q2 dependence of the running coupling constant,
the quark distributions and the quark fragmentation
functions change the effective power behaviour of the
single particle distributions
(41)
from the canonical n = 4 to higher values n = neff-

The scale violations are only logarithmic, therefore
the effective power neff is not independent of rs and x.l.
In Fig. 5 /22/ we give for three different energies in
the range of present accelerators and as function of P.l.
this effective power. neff falls steeply with increaSing
collision energy fa and is not independent of XL or P.L
in the range of present data.
5
5 10 15 20 p.. GlNlc
l<'ig. 5. The effective power neff' (Xl.' t'S) for different

energies. Solid curve: QCD prediction without
parton k.L0 Dashed curve: <k.1.. 2 >1f2:1 GeV/c.
258 JOHANNES RANFT
As discussed first by Coutogouris et al./23/ and

Field /24/ using Monte Carlo calculations, the effective
power neff is increased further by taking the parton
transverse momenta into account. To study the effect of
the parton k~, we use an approximate analytical method
proposed by Ranft and Ranft /22, 25/. With this method,
using factorized structure functions
J='()(,G:1- 1 k,l.) = F(~.G22.) ~(k~) (42)
with F (x,Q2) as given by asymptotic freedom and

Gaussian k,l. distributions
(43)
'itk,L)
we obtain the single jet distribution with parton k~
in terms of the single jet distribution without
parton kJ,
H(p.L.I fS, (. k:'> &0)
ci: cr (.
E d =: H(r~)rst~k~'»= " (44)
P )2.J
"
= 4 r21'
Ja.t-
Sds'" I'hr~~ f3: /s/<::k.J.>=O)~[
~ \ f (p~- iih
2
S
<k2.>
p~ ~ 1c.a.2.). .£,.
In Fig. 5 we plotted also the effective power neff

( re, <kt» for the total jet distribution with
<ki> =1(GeV/c)2. In the P.... region arround pJ,= 5 GeV/c
we find indeed an effective power near to n = 8 as
found experimentally. In the case with parton k~,
the power depends much stronger on x~ or p~ , de-
creasing rapidly with rising X.L.. In a recent ISR ex-
periment, Clark et ale /26/ find in a range O.2~:X ~~
0.45 the power n = 6.6 + 0.8, in qualitative agreement
with our result. -
The agreement of the QCD calculations with data at
ISR energies rs = 50-60 GeV is rather good with or
without parton k,l., see /19, 22/. At Fermilab energies,
the QCD predictions without parton k,Lwere well below
the data /19/. In ~ig. 6 we compare the prealction for
~ing~e jet distributions using average k~ values of
< kJ.> = 0, O.?, and 1 GeV /0 with the single jet dis-
tributions of Bromberg et a1./27/. For the highest koL
value, good agreement is obtained also at Fermilab
energies.
pp-JET' X
data: Bromberg et al.
• ~ 3GeVJc
-28
• >4 '
• untJjasE'd
::- -29 '.'.
~ - - - (k:~ -1 GrNlc
Q7 •
"e
&
-30 --' a
a. "31
~
.g
w -32
.§' •
-33
-3"0 1 2 3 4 5 6 7 8 R.GeVJc
Fig. 6. Comparison of' the QCD prediction :for the large

p~ single jet cross sectionoin pp collisions
at fS= 19.4 GeV and e = 90 • 'Ithe data is due
to Bromberg et a1./29/. ~he three curves are
the QCD predictions for~kL2>V2 = 0,0.7 and
1 GeV/c.
3.3. QCD Predictions for Large p Particle Production

at Higher ~nergies and by Polarized Hadrons
At presently available energies, the QCD pre-
dictions are consistent with the data, neTerthe1ess
some doubt remains. Therefore, experiments at the next
generation of accelerators, the pp colliders at the
CERN-SPS, and Fermilab and ISABELLE, will provide
crucial tests for the applicability of QCD pertur-
bation theory to large P.l. hadronic react ions. At these
energies the QCD predictions are by a factor 10-50 .s
higher than the extrapolations of present data with a l'
law, see /19/. L
The QCD predictions for large p~ particle pro-

duction by polarized hadrons were discussed by Ranft
and Ranft /28/.
According to QCD the quark quark scattering cross
sections t'or definite helicities of the two quarks are
r- s2-
1- o~·
IJ ~ - J;. z.
'J 3
260 JOHANNES RANFT
olo:;- (45)
de =
and for definite transverse polarizations of the two
quarks
cto-1'+ no<;
= s2.
~t
d~~ 1r 0<.$
l...
=
Cltt- &2.
From (45) we obtain at e = 90 0 the asymmetry
c{fY+- i- d...~_
i 1-j
'" =
cU:. d..t
=
{ 0.' (47)
A£..J... c10+i-
+- d.-<11--
o.LtS I=j
~ oU:
The corresponding asymmetry for transverse polarization

can be obtained from (46) and is smaller than (47).
Using furthermore the distribut ions of quarks with
definite helicities within protons with definite heli-
cities for instance according to a broken SU(6) model
/29/ we obtain the asymmetries for large P.l.. part icle
production by two protons with definite helicities
E d~6++ E (.{~a+- (48)
c;f&f' tJ.3p
In Fig. 7 we plot Ahel (p J.. ' fa) as funct ion of x..L

for two values of fS. The measurement of such
asymmetries would be an important verification of the
QeD predic'tions.
J b
b
J
b
b
b
b
.1 ~L_-; - - -':-/_-: - - - r.-~-

/
/
/
A
A
A - - - 50.3 •
7
27. GeV
O. .2 .4 f> .8 1.
x.
Fig. ? The asymmetry Ahe1 :for large PoL reactions with

polarized protons as function of x~ •
3.4. QCD Predictions for the Production of Three Large

Pl.. Jets /30/
CCD predicts large P.J- events with additional gluon
jets. There are many possible three jet processes. At
large x~ tr.e dominant contribution is expected to be
=L
the process quark + quark~quark + quark + gluon
according to
~ ~~ =t(49)
The azimuthal angle ~ dependence of the production of
two large p~ particles for a fixed value of the total
transverse energy EL~ p~ + PL2-+ PoL (p~ is the
transverse momentum balaJcing ~1 + 3p;) ~as measured
by Cobb et a1. /31/. If the jets are produced at
9 = 90 0 the variable E~ plays for the large p~jets
a similar role as rs in e+e- -.hadrons or the hadronic
energy W in deep inelastic collisions. The large 'cross
section for large E..L. around e~ 1800 is evidence for
the jet structure of large P..L. events. In this experi-
ment the usual trigger bias effects, which occur if
one or two particles are selected according to their
transverse momentum, are not present.
The cross section measured by Cobb et a1. /31/
was howevgr also found to be rather large at e values
around 90 where only rather small contributions are
262 JOHANNES RANFT
expected from the dominant two jet terms. Kripfganz

and Schiller /30/ found that the processes (49) explain
at large E~ the orger of magn1tu~e of the experimental
data around e = 90 , see Fig. 8. Therefore the further
study of three-jet events in large PL reactions an~
their comparison with the predictions t"rom QeD per-
turbatjon theory should be very interesting.
3.5. Gluon Jet Effects in Deep Inelastic Le~ton Hadron
Reactions
'rhe occurence o! gluon jets is easier to study
theoretically in e+e annihilation and deep inelastic
lepton hadron scattering, where the number of lowest
order Feynman diagrams which ha~e to be considered is
smaller. Gluon jet etfects in e e- annihilation were
considered in detail by DeRujula, Ellis, Floratos and
Gaillard /32/.
In deep inelastic lepton hadron collisions the
dominant two-jet term is due to the diagram
( 50)
o
III
X - all final state jets
~ 40 -- qucuk final state jets
VI
~
~20
;
~ O.j....-L.~...;u.=.,....~
0- 90-
Fig. 8.
duction of two *'
The az imuthal aneile <l> de-pendence of' the pro-
at large p. t"or fiJeed E~
= P~1 + p~? + PL1· The data \s due to Cobb et
ale The carves are calculated by Kripfganz
and Schiller according to QCD three~jet
contribut ions qq - qqg.
To lowest order in «s the following processes contri-

bute to three - jet production e ('
t (,' ~ .I' ( ___~
(51)
-"""-'1
if
p------ p
The two processes with emlssion or a gluon dominate at
presently available energies. The rise of hed.ron trans-
verse momenta relative to the direction of q due to the
processes (51) was studied by Altarell~ and Martinelli
133/.
Azimuthal asymmetries due to these processes were
predicted by Georgi and Politzer /34/, Cleymans /35/
and Mend~z /36/. Georgi and Sheiman /37/ studied the
cross section do/dT~, where ~iS defined as
( 52)
They found that the H.L distribution, which at large
is less dependent on nonperturbative effects, can be
u:.
predicted more safely than average values like <P,L'> •
Ranft and Ranft /38/ studied the 3 jet cross sections
(51) in terms of the variables spherocity S and thrust
T. In the following we will discuss these cross
sections. The variables S /39/ and T /40/ are defined
as follows .
S=(~t~i'" (z.lrlIV- (53)
tr L.1p..I)
(54)
T==
For an ideal jet at ra~1l:I one expects S .. 0 and T .. 1.
For an ideal spherical symmetric t'irebaLl S .... 1 is ex-
pected. At finite energies nonperturbative effects,
the fragmentation of quark jets into hadrons with
limited PL ' leads to Sand T distributions of forms
like
ctN
d.S
<'V S
(AS)~
e..xrJ(-.L).
\ AS ) (55)
264 JOHANNES RANFT
dN
dT
N (" - T)
(AT)t. e.X)' -
. - T)) j
(-1AT (55)
where W is the total hadronic energy in the processes

(50) and (51).
Average values of t S > and .( 1 - T> as function of
the hadronic energy W were measured and found to de-
crease with W as expected /39/, thus giving evidence
for the jet structure of the hadronic system. Due to
the nonperturbative effects the average values are
however less well suited to show the existence of the
third jet as predicted according to (51)0 The thrust
or spherocity distributions are better suited for this.
In Fig. 9 we give the perturbative (according to (51))
and nonperturbative cross sections da/dT /38/. For
hadronic energies W>12-15 GeV the nonperturbative two-
jet background no longer dominates and the perturbative
three-jet contribution should be detectable. It would
be even more direct evidence for gluon jets, if in-
spection of the events out of this nonperturbative tail
would give direct evidence for the third gluon jet.
However gluon jets are expected to fragment into a
much wider, less jet-like hadronic final state so that
this second possibility seems to be more remote.
- - 2 jet non-perturbative 1.1&.

/' adT
- 3 jet perturbative
(without smear)
<;~. II
W-6GfN-" ..-::
0.' :. 2 GeN' I!.T-Q04)../ I
~ / I
,/ / /
/ / / 0.1
W-6GtN I
10 _ t ,/
/
/
/ I
20- / / / 0.01
,/,/ / /
,/ / I
ll'/ ~24 /
6T-(Xl3 ~ 20Gl\/, toT -0017
.6 7 .8 .9 T _ 1.
Fig. 9. Perturbative (3 jet) and nonperturbative (2

jet) contributions to the cross section drr/dT
as function of thrust for different values
of the total hadronic energy w.
3.6. Conclusion. Is There Evidence for QCD ?

Besides of its theoretical attraotiveness,
the present evidence for QCD is due to
( i) the nonscaling of struoture funct ions found in
deep inelastic collisions
(ii) the scaling in ~ ot the lepton pair production
in hadronic oollisions
(iii) the existence of jets in e+e annihilation and the

behaviour of R
(iv) the existence of jets in lepto production
(v) the consistency of large Pk data with QCD pre-
dictions.
Presently no convincing evidence is yet available for
predicted effects like
(i) gluon jets in e+e- annihilation, deep inelastic
lepton hadron interaction, lepton pair pro-
duction or large P.1.. reactions
(ii) the Q2 dependence of the quark fragmentation
functions.
References
/1/ R. Hagedorn, Supple Nuovo Cim. 3, 147 (1965).
/2/ R. Hagedorn and J. Ranft, Supple Nuovo Cim. 6,
169 (1968);
Nucl. yhys. B48, 157 (1972)
/3/ J. Ranft, Nucl. Phys. B105, 139 (1976); this
paper describes the last version of the T. M.,
see also J. Ranft, contribution to the Work-
ShOp on Theoretical £hysics: Hadron Thermo-
dynamics, Erice 1975.
/4/ reviewed in J. Hanft, Fortsch. Phys. 23, 467 (1975).
/5/ H. Satz, preprint CERN-TH 2532 (1978).
/6/ u. uo~dhaber et al., ~hyS. Rev. 120, 300 (1960)
/7/ J. Ranrt and G. Ranft, Phys.Lett. 57B, 373 (1975);
Nucl.Phys. B92, 207 (1975).
/8/ Data by V. ~. Kenney et ale pre~ented by R. E.
Diebold at the 19. Int. Conf'. on High Energy
Physics, Tokyo 1978.
266 JOHANNES RANFT
/9/ L. Van Hove and S. Pokorski, Nucl. Phys. B86, 243

(1975)
/10/ W. Oche, Nuel. Phys. B118, 397 (1977).
/11/ K. P. Das, R. C. Hwa, Phys. Lett. 688, 549 (1977).
/12/ T. A. DeGrand, H. I. Miett1nen, SLA.C-PUB-2036( 1977).
/13/ D. W. Duke, F. E. Taylor, Fermilab-Pub-77/96-THY
(1977).
/14/ J. Ranft, Phys. Rev. D18 (1978) 1491•
/15/ T. A. DeGrand, D. W. Duke, H. I. Miettinen,
I.Inami, J. Ranft and H. Thacker, to be
published
/16/ H. Goldberg, Nuel. Phys. B44, 149 (1972);
R.BIankenbeeIer, S.J. Brodsky, Phys. Rev. D10,
2973 (1974);
S.J. Brodsky, J. F. Gunion, SLAC~UB-1939
(1977).
V. V. Kniazev, A. K. Likhoded, V. A. Petrov,
and A. N. Tolstenkov, Serpuehov prepr1nt
OTF 77-106 (1977).
/17/ F. C. Ern~, J.C. Sens, CERN-prepr1nt 1978;
V. Cerny, P. Liehard, J. Pisut, Phys. Rev. D
( 1977) ;
V. V. Anisovieh~ V. M. Shekter, Nuel. Phys.
B 55, 455 (1973).
/18/ J. Kuti, V. F. Weisskopf, Phys. Rev. D4, 3418
(1971).
/19/ H. D. Politzer, Nuel. Phys. B129, 301 (1977);
C.T.Saehrajda, Phys. I,ett. 73B, 185 (1978);
K. H. Craig, C. H. Llewellyn Smith, Phys. Lett.
72B, 349 (1978);
w. Furmanski, Cracow prepr1nts TPJU-10/78,
1278;
D. Amati, R. Petronzio, and G. Veneziano,
Nuel. Phys. B 140 (1978) 54.
J. Kripfganz, l,eipzig, preprint KMU-HEP-78-00
( 1978);
Yu. L. Dokshitsera D. I. D'Yakonov, and S. I.
Troyan, I'roe. 13 Winter School of LNPI,
Leningrad 1978.
/20/ J. Kripfganz, JJeipzig preprint KMU_HEP-78-12 (1978)
G. Al tareIIi, R. K. ELl is, and G. Martinelli,
MIT preprint CTP 723 (1978),
J. Abad, and B. Humpert, Wisconsin preprints
1978;
H. Georgi, Havard preprint HUTP - 78/A003 (1978).
/21/ B. R. Combridge, J. Kripfganz, and J. Ranft,
Phys. Lett. 70B, 234 (1977).
/22/ R. Cutler and D. Sivers, Phys. Rev. D17, 196 (1978).
/23/ A. J. Buras, and K. J. F. Gaemers, Nucl. Phys.
B 132, 249 (1978).
/24/ J. Ranft, and G. Ranft, Leipzig preprint KMU-HEP-

78-06, submitted to Aota Physica Polonica (1978).
/25/ A. P. Contogouris, R. Gaskell and S. Papadopaulos,
Phys. Rev. D in press (1978).
/26/ R. D. Field, preprint CALT-68-633 (1977)
/27/ G. Hanft, and J. Ranft, Nuovo Cim. Lett. 20, 669
(1977).
/28/A. G. Clarketal., CERN prep rint (1978).
/29/ C. Bromberg et al., Proc. 8. Int. Symp. on Multi-
particle Production, Kayserberg 1977, p.13-89.
/30/ J. Ranft and G. Ranft, Phys. Lett. 778, 309 (1978).
/31/ F. E. Close, Nucl. Phys. B80, 269 (1974);
R. Carlitz, J. Kaur, Phys. Rev. Lett. 38, 673
(1977).
132/ J. Kripf'ganz, A. Schiller, Leipzig pre-print KMU-
HEP-78-10 (1978) P~8. Lett. 79 B (1978)
317. .
/33/ J. M. Cobb et a1. t Phys. Hev. Lett. 40, 1420
(1978).
/34/ A. DeRujula. J. Ellis, E. G. Floratos and M. K.
Gaillard, CERN-TH-2455 (1978).
/351 G. Altarelli and G. Martinelli, Roma preprint
1978.
/36/ H. lieorgi and H. D.Politzer, Phys. Rev. TJett. 40,
3 (1978).
/37/ J. Cleymans, Bielefeld preprint OUTP 29/78 (1978)
/38/ A. Mend~z, Oxford preprint OUTP 29/78 (1978).
/39/ H. ueorgi and J. Sheiman, Havard preprint H~~-
78/ A034 (1978).
/40/ J. Hanft and G. Hanft, Le ipz ig preprint KMU-HEP-
78-15 (1978), Phys. Letters B (1979) in-press.
/41/ A. Vayaki, Aachen-Bonn-CERN-London-Oxford-Qaclay
Collaboration, CERN preprint C~RN/EP/Phys 78-28
(1978).
FROM HADRON TO QUARK MATTER
H. Satz
CERN - Geneva
and Department of Theoretical Physics
University of Bielefeld*)
ABSTRACT
We discuss the transition of strongly interacting systems

from a hadronic resonance gas to a gas of quasi-free quarks. By
limiting the hadron density, we obtain an equation of state which
yields a quark phase at small and a hadron phase at large values
of the resonance coupling constant. Separating these regimes,
there is a transition region with a behaviour of the Van der Waals
type, providing a negative specific heat and, at the critical
point, infinite temperature fluctuations.
*)
Permanent address
269
270 H. SATZ
I. INTRODUCTION
If we consider a system of many strongly interacting

particles at high density, we encounter two competing descriptions.
On the one hand, our picture of hadronic forces suggests hadronic
matter as an assembly of many different types of resonances. On
the other hand, the quark basis of resonance structure accounts
for each resonance as an orbital excitation of a bound quark-
antiquark or of a bound three-quark system; very dense strongly
interacting matter is therefore a multiquark system. Thus it is
tempting to think of two different phases of strongly interacting
matter 1), a hadron phase and a quark phase (cf., Fig. 1), and to
study the transition between these phases.
(a) (b)
Fig. 1: Strongly interacting matter in

(a) hadron and (b) quark phase
The possible two-phase nature of strongly interacting systems

has in recent years attracted more and more attention - triggered
both by quark and by hadron interaction studies. One may start
with a theory of constituent interactions, such as quantum chromo-
dynamics (QCD), and look for transitions between bound state and
"free" quarks 2 ), or one can check if hadronic interaction models,
perhaps because of their implicit quark basis, lead to singular-
ities interpretable as critical points 3). Bridging the two sides,
FROM HADRON TO QUARK MATTER 271
one can try to construct phenomenological models [bag model 4 ), Van

der Waals modelS)] containing by assumption both phases character-
ized by their most important features, and then study how phase
transitions may take place.
The construction of statistical mechanics for QCD 2 ), i.e.,

for a gas of coloured quarks with gluon interaction, is a task
which will certainly receive much attention in the future, if QCD
does persist as our theory of constituent interactions. Since the
perturbative methods of the weak coupling limit cannot yield the
strongly coupled hadronic phase, additional, perhaps more pheno-
menological, inputs may be necessary to obtain a connection
between the two regimes. For a further discussion of these prob-
lems, see Ref. 2). We shall here restrict ourselves to the other
two approaches: to critical features of hadronic resonance gases,
and to the phase transition between hadron and quark matter in a
Van der Waals type of picture.
Before going into details, let us remain for a moment on a

qualitative level. With increasing density, a system of hydrogen
molecules (Hz) experiences a phase transition from gas to liquid;
however, at sufficiently high temperature, the difference between
gas and liquid phase disappears. Quark confinement attributes to
hadronic states an infinite binding energy, so that a system of
hadrons in a given volume could exist at all energies; therefore,
we expect a hadron-quark transition, if it exists, to be of higher
order 6 ) .
The statistical bootstrap model 7 ) provides a one-phase

description of multi-hadron systems at all energy densities: a
gas of hadronic resonances contained in a fixed volume. The
spatial extent and the compositeness of hadrons, however, provide
serious difficulties for such a picture. If the size of hadrons
does not decrease with mass, then we must require them to become
mutually transparent to each other in order to accommodate an
272 H. SATZ
unbounded number within a given volume*). Compositeness appears

even more serious: when thirty protons are in one hadronic volume,
it is hard to conceive how the constituent quarks could remember
"their" hadron. It then seems more appropriate to consider the
system as a gas of quarks, and if quarks are point-like, this new
phase could exist up to arbitrarily high densities.
Let us consider an atomic analogue here as well. A gas of

metal atoms represents in our language a gas with confined valence
electrons as constituents (Fig. 2a), the confinement being
achieved by a bag set up through the ionic Coulomb field. In-
creasing the density sufficiently leads to solid metal, with the
valence electrons "free" in the lattice and no longer associated
to a particular ion (Fig. 2b). For strongly interacting particles,
the confinement forces (gluons) must provide for the quarks (and
perhaps also for some phonon-like gluons) the analogue of the
lattice.
(0 ) (b)
Fig. 2: Metal in (a) gaseous and (b) solid phase
*)
This problem can be avoided by formulating the bootstrap model
at constant particle density [see Ref. 8)]
II. LIMITING THE HADRON DENSITY
Our aim in this section will be to determine a partition

function for strongly interacting systems; from this, the statis-
tical description of such systems can then be obtained by the
conventional formalism of statistical mechanics. We shall arrive
at the desired partition function by modifying the statistical
bootstrap equation of the resonance gas approach such as to res-
trict the possible density of hadrons, requiring above a certain
values a quark gas description instead.
The level density of an ideal gas of hadrons and hadronic

resonances is given by
a(P2.) = 8('+)(P) + ~ B~ J ~ {d'+P1oP(P~)} 8('+)(~1 P1o-p), (1)

N=l N. i=l
where P denotes the over-all four-momentum of the system, and

B is a parameter characterizing its co-ordinate space extent:
with B = 2~V, we obtain in the non-relativistic limit the grand
microcanonical level density for a system of (Boltzmann) pions of
mass ~,contained in a volume V. The resonance spectrum is
described by p(P2.); the choice
(2)
would restrict us to pions only, so that p(P2.) - 8o(P2.-~2.) con-

tains the higher excitations. The statistical bootstrap approach 7)
determines this excitation spectrum through the resonance gas level
density, i.e., statistically:
(3)
with a denoting the over-all strength of the excitation.

274 H. SATZ
Equality of spectrum and level density [strong bootstrap condi-

tion 7b )] requires a = 1. Equations (1) and (3) completely
determine the system. Inserting (3) in (1) gives
where the summation times a comes from a single massive resonance

composed of multi-resonance subsystems, while the remaining
summation comes directly from multiple resonance contributions.
From the grand microcanonical formulation we now pass to the

grand canonical. The partition function of the system is given by
(5)
with
(6)
denoting the generating function of the complete resonance spectrum,

while
(7)
is that of a pion alone. The Laplace transformation of Eq. (4)

then yields
Z(S) • 1 + B~
o (8) + (l+a)
(8)
as the determining equation for the partition function of a

hadronic resonance gas in the statistical bootstrap approach.
Carrying out the summation leads to the more familiar form
~o(B) - 2 ln Z(B) + 1 - Z(B), (9)
where we have set a. '" 1.
Equation (8) clearly shows that a given volume B may con-

tain arbitrarily many hadronic states. To modify this 5 ), we permit
only N ~ NH hadrons at fixed B; any contributions from higher
N must be in the form of quasi-free quark-antiquark pairs, rather
than in the form of hadrons or hadronic resonances. Hence in
Eq. (8) we must replace
(10)
where <PQ (B) is the generating function for the qq pair. We

shall return shortly to the form of <pQ(B); we note here only
that relation (10) implies no direct interaction between the
different qq systems, just as we had assumed none between
different hadtonic resonances.
For simplicity, we now fix NH-2; the formal extension to

higher ~ is discussed in Ref. 5b), where it is found to provide
qualitatively the same behaviour. Our modified partition function
is then determined by the equation
Z(B) = 1 + B<p (B) + (I+a.) ln 2 Z(B) ~ (11)

o N=2
For a. = 0, we have as unique solution the ideal gas
Z(B) II: (12)
provided that then
( 13)
276 H. SATZ
Hence a may now be interpreted as a constant measuring the

strength with which qq pairs are coupled to different hadronic
states. For a = 0, we have only one such state, accountable
either as a hadron or as a qq pair. If qq pairs can yield
different hadronic states, a must be greater than zero.
Let us now consider the generating function ~Q(S). The

level density of. N free identical quarks together with their
antiparticles is given by
where we have given the quarks an effective mass ].12 = q2 and

q
neglected for the moment their Fermi statistics. The corresponding
grand canonical partition function is
(15)
with
(16)
as single-quark generating function. Since.for small S (high

temperature)
( 17)
we take <P in Eq. (11).

Q= 2~
q
Finally, we introduce Fermi statistics for the quarks. The
generating function (16) of a quark must then be replaced by9)
(18)
the Boltzmann form (16) is the first term of the sum.

Relations (11), (17) and (18) thus determine our modified

partition function. For simplicity we shall take the zero coup-
ling form (13) to hold at all values of a; then we have
eB~o(S)_l_B~ (S)
2(S) " 1 + &0,(8) + (1+") 1n'Z(8) [ (
B~o (S)
),' 1 (19)
as the hadron-quark bootstrap equation.
Before going on to the solution of Eq. (19), we note that it

would also be possible to "truncate" the bootstrap equation (8)
additively, requiring the summation up to N = N to contain only
H
hadrons, above N = NH only quarks. The resulting form
2(S)
(20)
10)
is, however, found to yield the same qualitative behaviour as
Eqs. (11)-(19), so that we here concentrate on the latter.
III. SINGULARITIES, INSTABILITIES, PHASE TRANSITIONS
In this section, we shall show that the partition function

determined by the hadron-quark bootstrap condition (19) provides
three distinct types of behaviour. For sufficiently large
coupling a, we essentially recover the conventional bootstrap
picture, while for sufficiently small a we have a quasi-free
quark gas at all energy densities. Separating these extremes is
a "coexis tence regime". bounded on both sides by critical values
a and a of the coupling parameter a; it is in this region
c c
that energy density variations take the system from a hadronic
resonance gas phase to a quasi-free quark gas phase.
278 H. SATZ
Before studying the behaviour of the modified partition

function (19), we recall briefly that of the conventional bootstrap
model ll ). Differentiating Eq. (9), one finds that a~o/az
vanishes and hence aZ/a~o or az/a8 diverge when
~0(8) 2 In 2 - I (21 )
so that Eq. (21) defines a critical temperature 8 = 8o • since

a2 "'0
,n laz = - 1 there
2
2'
one has near 80
(22)
and thus
(23)
The functional behaviour of Z(8) is shown in Fig. 3a. Since the

energy density
~(8) = (-I/B)(a In Z/a8) (24)
diverges at 8 = 80 because of Eq. (23), the temperature To=1/8 0

is a limiting temperature of the system, attained only at infinite
energy density (cf., Fig. 3b). For comparison, we show in Fig. 3
both times the corresponding behaviour for an ideal gas.
Let us now consider Eq. (19). We see that
a(B~o)/az = [1-2(I+a)Q In Z/Z] I [1+(I+a)Q' In 2 Z] (25)
N-2
Q ::
ex>
L (B~o) IN! Q' :: aQ/az (26)
2
vanishes when
FROM HADRON TO QUARK MAnER 279
Z - 2( 1+a.)Q In Z. (27)
Z(Jj) E(Jj)
FG
To To
(a) (b)
Fig. 3: Temperature dependence of partftion function (a),

and energy density (b) for statistical bootstrap
system (resonance gas) and free gas.
However, if in addition we have
In Z = I, (28)
then also the second derivative a2B~o/az2 vanishes. Hence
determines a critical value ac of the temperature, and
a critical value of the coupling. Further study5) shows that for

280 H. SATZ
a > a c ' aBcpo/az vanishes at some point Bo' where is

still negative; at a = a
both derivatives vanish at
c'
while for a < a , the first derivative remains finite for all
c
The resulting functional form is shown in Fig. 4.
z (~)
~-I
Fig. 4: Temperature dependence of the hadron-quark partition

function for resonance coupling a > a (a), a=a (b),
c c
and a < a (c); curve (d) shows free gas behaviour.
c
In the case a > a , we thus have effectively bootstrap

c
behaviour: Z has a square-root branch point at Bo and the
energy density diverges as B + Bo. For a = ac' Z has a cube-
root branch point instead; this also yields a diverging energy
density at B = Bc • For a square-root branch point in Z, one
obtains the level density
(31 )
with a = -3; correspondingly, the cube-root yields a = -17/6.

In general, form (31) provides diverging energy densities at

B = So for all a > -1/2.
For our partition function (19), this regime ends with a=a •
c'
for a < ac' d~O/dZ never vanishes and hence we always have a
finite energy density. In Fig. 5, we show s(S), as obtained
from Eq. (19) and (24), for various a < a
- c
E(/3)
/3-1
Fig. 5: The hadron-quark energy density for different
resonance couplings ac > a 3 > a2 > a l •
The most striking feature is the existence of a temperature range

TA < T < TB, in which the specific heat CB (dS/dT)B = is negative;
TA and TB are determined by the requirement
(dS/dT)B = O. (32)
We shall return shortly to the implications of such a situation;

here we note that the range of negative CB decreases with de-
dreasing a, until for a < a, CB is always positive. The
282 H.SATZ
conditions
(33)
determine the critical value a

c
We thus have three distinct regimes of coupling in the hadron-
quark equation (19):
i) for a > a c ' we have always a positive specific heat CB;

the temperature is bounded, T < To' and the energy density
E increases monotonically with T, diverging as T ~ To;
ii) for a < a < a , the temperature becomes unbounded, but
c c
there is a temperature region of negative specific heat, in
which the energy density decreases with increasing tempera-
ture;
iii) for a < ac , the specific heat is once more always positive,
the temperature is unbounded, and the energy density in-
creases again monotonically with T.
While the first regime, as already noted, displays characteristic

bootstrap behaviour, the third behaves as a conventional homogenous
gas. The second regime, as we shall now show, contains aspects of
both.
For a just below a, we have

c
(34)
at T - T , so that the energy density almost diverges there; hence

c
in the temperature range T < T , we observe practically bootstrap
a
behaviour. On the other hand, Eq. (19) for a < a has the high
c
temperature solution
Z(e) = (l+a) eB~o , (35)

so that for T » Tb , the energy density £(13) approaches that of the

free quark gas. In the region between Ta and Tb , where the
specific heat is negative, the system thus undergoes the transition
from hadronic bootstrap behaviour to quark gas form.
A negative specific heat, just as a negative compressibility,

violates the basic statistical requirement of maximal entropy7).
In this sense our Eq. (19) is quite similar to the Van der Waals
equation, where one encounters a region of negative isothermal
compressibility
KT = (-I/V)(av/ap)T (36)
between the gas and the liquid states of the system (cf., Fig. 6).
The problem is resolved by introducing two phases in coexistence
and consequently replacing P(V,T) by the Maxwell construction
peT) for V' > V > V' with VA' and V' determined by the equal
A - - B' B
area rule. For V > VA' we then have a homogenous gas, from VA
to V~ gas and liquid in coexistence, and for V< V~ a homogenous
liquid. The coexistence region becomes smaller with increasing T,
until it vanishes at the critical point T • T , where the Van der
c
Waals equation provides a diverging compressibility and hence in-
finite volume or density fluctuations: the system at the transition
point oscillates between the dense liquid and the less dense gas phase.
Fig. 6: Van der Waals P-V diagram,

for T< T and T = T
c c
284 H. SATZ
The hadron-quark system of Eq. (19) can be treated quite

analogously (cf., Fig. 7). Defining TA and T~ as above by an
equal area rule, we have a bootstrap phase for T
TA, a quark
~
gas phase for T ~ TB, and a coexistence regime in between. In

this region, the temperature increases at constant over-all energy
density, since slow massive hadrons change to fast light quarks -
rest mass is converted into kinetic energy. The coexistence region
decreases with decreasing coupling strength a, until it vanishes
at a =·a o • At that point, Eq. (19) provides a vanishing specific
heat and hence infinite temperature fluctuations: the system is
at the transition point between a hot quark gas and a (relatively)
cold hadronic resonance system.
€(Jl)
Fig. 7: Energy density of hadron-quark systems with

resonance couplings a c ' a c > a > a c ' and a c
We can thus note the correspondence
P <--> E: T <--> a
V <--> T
between a Van der Waals system and our Eq. (19). The phase
transition in the former is associated to volume or density
fluctuations, in the latter case to temperature fluctuations. In
Fig. 8 finally, we show the comparison between the P-T diagram
of the Van der Waals system and the corresponding E - a diagram
in our case.
p E
RG QG
G
a-I a-I a-I

c c
(0 ) (b)
Fig. 8: P-T diagram for Van der Waals systems (a),
and E-a diagram for hadron-quark systems (b).
IV. CONCLUSIONS AND OUTLOOK
With Eq. (19), we have given a description of strongly

interacting matter such that for increasing energy density we
first attain a region of resonance gas behaviour, associated with
very slow temperature increase (cf., Fig. 9). For large resonance
coupling, a > a the system never leaves this regime. For
c'
a c < a < a c' however, an increase of energy brings us to a region
of ins tabi li ty, where the system fluctuates between a resonance
gas and a quark gas phase. A further energy increase here leads
286 H.SATZ
to a stable quark gas regime. For very weak resonance coupling,

a < c
-a,the system is always in this phase _. it never exhibits
clear resonance gas behaviour.
__------------a c
a >a c
€
Fig. 9: Temperature variation in the hadron-quark system.
Together with the formalism of statistical mechanics,

Eq. (19) provides an equation of state for strongly interacting
matter - one which interpolates between hadron gas and quark gas.
Where can this equation of state be of relevance? Astrophysical
applications are evident, as both hadron stars and quark stars
have been extensively discussed 1),12). Furthermore, phase
transitions may also playa role in the description of multihadron
production processes, in the connection of hard and soft inter-
actions as well as in accounting for the different behaviour of
longitudinal and transverse spectra. Looking at the interaction
region of an energetic proton-proton collision, we encounter in
the longitudinal direction due to Lorentz contraction an energy
density suggesting a quark state - and hence are led to expect
distributions similar to those from e+e- annihilation, where the

"smallness" of the virtual photon leads to high energy density.
In the transverse direction, however, the system can rather
easily expand into a hadronic state. Thus, a two-phase equation
of state for hadronic matter may well provide a new impetus to
the hydrodynami cal mode I of Landau 13) •
We close with a comment on the relation between the bag

model 4 ) and our picture. The bag model prescribes the existence
of an upper bound on energy density for the hadronic phase. This
implies a growth of hadron size with mass, so that the hadronic
particle number density becomes also bounded - similar to the
behaviour found in the Pomeranchuk model 14 ) or the Kiev-style 8 )
bootstrap model. Such a bound suggests a first order phase
transition - in contrast to our higher order transition, which is
directly connected to the input of mass independent hadron size.
********
Stimulating discussions with J. Baacke, T. ~elik, R. Hagedorn

and especially I. Montvay are gratefully acknowledged.
288 H. SATZ
REFERENCES
1) N. Itoh, Progr. Theor. Phys. 44, 291 (1970;

F. Iachello, W.D. Langer and A. Lande, Nuclear Phys. A219,
612 (1974);
N. Cabibbo and G. Parisi, Phys. Letters 59B, 67 (1975);
J.C. Collins and M.J. Perry, Phys. Rev. Letters 34, 1353 (1975);
G. Baym and S.A. Chin, Phys. Letters 62B, 241 (1976);
G. Chapline and M. Nauenberg, Nature 264, 235 (1976);
B.A. Freedman and L. McLerran, MIT Report 541 (1976) - un-
published;
G. Chapline and M. Nauenberg, Phys. Rev. D16, 450 (1977);
M.B. Kislinger and P.D. Morley, Phys. Letters 67B, 371 (1977).
2) D.A. Kirzhnits and A.D. Linde, Annals of Physics 101, 195

(1976) ;
M.A. Polyakov, "Thermal properties of gauge fields and quark
liberation", ICTP preprint IC/77/135 (1977;
T. Banks, R. Myerson and J. Kogut, Nuclear Phys. B129, 493
(1977);
B.A. Freedman and L.D. McLerran, Phys. Rev. D16, 1130, 1147,
1169 (1977);
G. 't Hooft, Nuclear Phys. B138, 1 (1978);
V. Baluni, Phys. Rev. D17, 2092 (1978);
C.G. Callan, R.F. Dashen and D.J. Gross, Phys. Letters 77B,
270 and 78B, 307 (1978);
L. Susskind, "Hot quark soup", SLAC preprint SLAC-PUB-2070
(1978) •
3) R.D. Carlitz, Phys. Rev. D5, 3231 (1972);

N. Cabibbo and G. Parisi, Phys. Letters 59B, 67 (1975);
D. Miller, Phys. Letters 64B, 193 (1976);
s. Eliezer and R. Weiner, Phys. Rev. D13, 87 (1976) ;
A.1. Bugrij and A.A. Trushevski, "High temperature phase

transition in hadron matter", Kiev preprint lTP-76-35E (1976);
I. Montvay and H. Satz, Nuovo Cimento 39A, 425 (1977);
J. Letessier, A. Tounsi and H. Satz, Nuovo Cimento 45A, 580
(1978) ;
H. Satz, "Critical behaviour of hadronic matter", CERN pre-
print TH. 2493 (1978), Phys. Rev. D, in press.
4) A. Chodos, R.L. Jaffe, K. Johnson, C.B. Thorn and V.F. Weiss-

kopf, Phys. Rev. D9, 3471 (1974);
F. Cooper, G. Frye and E. Schonberg, Phys. Rev. D11, 192 (1975);
G. Baym and S.A. Chin, Phys. Letters 62B, 241 (1976);
J. Baacke, Acta Phys. Polon. B8, 625 (1977).
5) a) T. ~elik and H. Satz, "Phase transitions l.n hadronic

matter", CERN preprint TH. 2518 (1978); Z.Phys.C., in press;
b) T. ~elik, "The structure of phase transitions in hadronic

matter characterized by finite number of fireballs",
Bielefeld preprint Bl-TP 78/17 (1978).
6) See, e.g., L.D. Landau and E.M. Lifshitz, Statistical Physics,

Pergamon Press (1958).
7) a) R. Hagedorn, Nuovo Cimento Suppl. 1, 147 (1965);

b) S. Frautschi, Phys. Rev. D3, 2821 (1971);
c) P. Fre and L. Sertorio, Nuovo Cimento 28A, 538 (1975).
8) M.l. Gorenstein, V.A. Miransky, V.P. Shelest, G.M. Zinovjev

and H. Satz, Nuclear Phys. B76, 453 (1974).
9) M. Chaichian, R. Hagedorn and M. Hayashi, Nucl. Phys. B92,

445 (1975).
10) I. Montvay, private communication.

290 H. SATZ
11) W. Nahm, Nuclear Phys. B45, 525 (1972).
12) See, e.g., B. Freedman and L. McLerran, Phys. Rev. D17,

1105 (1978), and further references therein.
13) L.D. Landau, Izvest. Akad. Nauk , Ser. Fiz. 12, 51 (1953).
14) I.Ya. Pomeranchuk, Dokl. Akad. Nauk 2!, 889 (1951).

SUPERSYMMETRY APPROACH TO THE UNIFICATION OF INTERACTIONS
John H. Schwarz%
Laboratoire de Physique Theorique de l'Ecole Normale
Superieure, 24 rue Lhomond 75231 Paris Cedex OS, France
ABSTRACT
Various trends and problems in supersymmetry theory are discussed.
The requirement that gravitation be consistently unified with the
other interactions in a finite theory is very restrictive with far-
reaching consequences for the other interactions. A specific example
based on the spinning-string theory is described in some detail. In
this model the elementary particles lie on linear Regge trajectories
with a common slope determined by Newton's constant. As a result the
spectrum suggests a limiting temperature corresponding to the Planck
mass, rather than to the pion mass as in hadronic string theories. The
superspace approach to supersymmetry is also presented. While this
formalism has so far only been applied to the reformulation of known
theories (in a more geometrical way) it is hoped that it will provide
a framework appropriate for obtaining insights and extensions that would
otherwise be elusive.
%John Simon Guggenheim Fellow on leave of absence from the California

Institute of Technology.
291
292 JOHN H. SCHWARZ
I. EXTENDED SUPERGRAVITY THEORIES
Ultimately a completely unified theory of interactions must in-

clude gravitation. There is an understandable tendency among part-
icle physicists to neglect it as an unimportant force, a fact which is
undeniably true for elementary particles at energies low compared to
the Planck mass (~IO I9 GeV). Nevertheless a very severe restriction
is placed on theories when one requires that they allow for the in-
corporation of gravitation in a way that allows for finite calculation
of quantum corrections. It is unreasonable to suppose that just any
otherwise acceptable theory of electroweak and strong interactions
will have this property. Therefore by focussing on those theories
which incorporate gravitation in a satisfactory fashion one may hope
to single out a small class of theories for the other interactions
as especially promising candidates. In the next section we will argue
that N =4 supersymmetric Yang-Mills theory (obtained as a limiting
case of the spinning string theory) [1,2] is especially deserving of
study on these grounds (as well as on others). Before describing the
logic that singles out that theory. let me briefly review the status
of extended supergravity theories [3,4] . They allow a closely related
but somewhat different approach to the problem of finding a consistent
unification of gravitation and the other interactions.
The central question in the whole subject is whether supersymmetry
--dynamically broken, of course-- is relevant to the real world. It
is a very attractive possibility not only because it is mathematically
pretty, but being apparently the only allowed nontrivial extension
of Poincare symmetry, it is required to unify particles of different
spin in a single irreducible representation. Therefore a unification
of the graviton (spin 2) with the gauge bosons of electroweak and
strong interactions (spin I) in a single irreducible representation
does not seem to be possible without supersymmetry. As yet there is
no experimental evidence for (or against) supersymmetry. The uncovering
of any such evidence (e.g. the discovery of a gravitino or gluino or
whatever) would, of course, be a major event. [t is conceivable,
SUPERSYMMETRY APPROACH TO UNIFICATION OF INTERACTIONS 293
however, that the mass scale characterizing the symmetry breaking is

too large to allow for the djrect observation of supermultiplet part-
ners. In this case lall need not be los t, since it could turn out that
a strongly broken supersymmetry theory would single out a small class
of phenomenological gauge theories with more relations among parameters
than is usually the case.
The most thoroughly studied class of supergravity theories are
the ones based on the super Lie algebra OSp(N,4). This algebra contains
generators for SO(N) x Sp(4), where the SO(N) describes an internal
symmetry, while Sp(4) gives the Poincare algebra by contraction. In
addition there are N 4-component Majorana-spinor generators S!
(~ = 1,2,3,4 ; i = 1.2, ...• N) with anticommutation relations of the
form [5,6] (aside from possible central charge~)
. .
I
t
S' SloP I
Do.)
= J (I )
where P r and Jr~ are the translation and Lorentz generators respect-
ively. e is a dimensionless constant. characterizing a de Sitter
radius, that vanishes when one contracts to obtain Poincare invariance.
The only other super Lie algebras that have been seriously studied
are the superconformal ones SU(2,2;N) which contain the conformal
group SU(2.2) and SU(N) as subgroups. While algebraically elegant
theories have been constructed [7J based on this algebra, they
involve higher-derivative interaction, so that their interpretation
as particle theories appears problematical.
The SO(N) theories with e = 0 have irreducible representations
whose particle content is easily inferred from (I). For the massless
case (which is the one of greatest interest) one finds that the I.R.
contains 2N physical helicity states with the multiplicities of the
various helicities given by
N!
(2)
where Jmax denotes the maximum helicity in the I.R. Evidently the
z
294 JOHN H. SCHWARZ
minimum helicity is given by
Jmax _ N
(3)
z 2
When Jmax F N/4, the TCP conjugate multiplet containing the opposite
z
helicities must be added in. Supergravity multiplets are, by defini-
tion, ones for which Jmax = 2. The multiplet is evidently self
z
conjugate (with respect to TCP symmetry) when N = S and contains
spins exceeding 2 for N ~ S.
Interactions can be introduced among the fields of a supergravity
multiplet in such a way that the spin-two (vierbein) field interacts
according to Einstein's theory and supersymmetry is preserved. When
such theories are constructed, they turn out to have surprisingly
good ultraviolet behavior. Namely, on the mass shell the divergences
appear to cancel through the two-loop order in perturbation theory
(S] . This is in striking contrast with what happens when gravity lS
coupled to "external" matter. In this case nonrenormalizable

infinities always arise at the one-loop order. Thus the higher level
of symmetry appears to pay substantial dividends. To be sure, there
is considerable concern that problems will arise at the three-loop
order. This may indeed be the case, at least for certain choices of
N, but as far as I know it is by no means established to be the case.
Wouldn't it be bizarre for a property that holds through two loops
to begin failing at the three-loop level ? (This rhetorical question
will reoccur in another context in the next section.)
When N does not exceed S/supergravity theories have the following
pleasant features : no fundamental fields with spin ~ 2, one graviton,
N spin-3/2 gravitinos that gauge the N local supersymmetries, and
N(N - 1) /2 massless vector fields that can be arranged to gauge the
internal SO(N) group. So long as one insists on disallowing additional
external matter, which would destroy the ultraviolet behavior of the
theory, all the elementary gauge fields of nature must be embedded
in this list. Thus the largest possible Yang-Mills group would appear
to be SO(S). Unfortunately this group is not quite large enough to
contain all of known physics --namely SU(2) x U(I) for the electro-
weak interactions and SU(3) for QCD. One can embed SU(3) x U(I) x U(I),
but then the ~ bosons are missing [9] . This is a very serious
problem in this approach. One might imagine that other states arise
dynamically and perhaps even fill out SU(8) multiplets, for example.
This is wild speculation, however, and I know of no reason to
seriously expect such good fortune.
Even if the problem of the group is overcome for extended super-
gravity, there are other difficult problems that must be confronted.
In particular, when Yang-Mills self interaction is introduced so as
to gauge the SO(N) group with coupling constant e (10] one obtains
a cosmological constant at the same time that is some 60 orders of
magnitude larger than the experimental limit (de Sitter radius of
10- 30 em instead of 10 30 em) for reasonable choices of e.One may
hope, following a suggestion of Hawking (II] that in a quantum theory
such a term would only play an important role on the scale of a Planck
length and would not imply a large effective de Sitter radius on a
macroscopic scale. Alternatively one may hope that some cancellation
can be rigged up, preferably in a "natural" way. Another serious
problem concerns the breaking of supersymmetry. For theories with
N > 2 the usual Higgs type mechanisms (12) appear inapplicable, at
least based on fields occurring explicitly in the Lagrangian. The
theory must find some clever way of breaking its own symmetry. There
are simply no knobs to turn from the outside. In summary this is a
very grandiose program that is running up against some equally
imposing obstacles.
II. ALTERNATIVE APPROACH BASED ON THE SPINNING STRING THEORY

One other alternative is known for a theory in which the graviton
interacts with other particles in a renormalizable fashion. This is
the string theory or dual resonance theory~ [13J. Only three schemes
~The renormalizability, achieved by absorbing infinities in the dual

slope and coupling constant parameters, is not fully established, but
appears likely to be true.
296 JOHN H. SCHWARZ
of this type are known to have a spectrum free from ghosts. The first
is the Veneziano model which exists in 26 dimensions (I time and 25
space), but contains a tachyon and no fermions. The second is the
spinning-string theory (14], which requires 10 dimensions (9 space
and I time). This is the theory we will discuss in greater detail
below. The third theory is a two-dimensional string theory [IS] with
extended (N = 2) supersymmetry in its gauge algebra. It turns out to
be equivalent to a supersymmetric nonlinear (J"" model [16] in two
dimensions x .
When first invented, the spinning-string theory was proposed as
a phenomenological theory of the hadrons. In the Born approximation
the state·s all lie on linear parallel Regge trajectories. The
universal slope~~ was naturally taken to be about I GeV- 2 ~n the
context of a hadronic model. Since string theories have an exponent-
ially increasing particle number density as a function of mass, this
implied the existence of a limiting temperature close to the value
suggested by Hagedorn and Fral1tschi based on statistical bootstrap
ideas [17] . What Scherk and I have suggested [18]in previous works,
and which I wish to review here, is an alternative interpretation in
which the states of the string theory are taken to be elementary part-
icles. In particular, a massless spin-two state in the closed string
sector is identified as the graviton. Then the requirement that its
interaction be of the usual Newtonian strength implies a universal
, -34 -2
slope controlled by the Planck mass, 0( ~ 10 GeV. When this
assignment is made one is left with a well-defined finite spectrum
of massless states and a plethora of string excitations having masses
of the order of 10 17 GeV, and hence quite unimportant for most
purposes. Of course, in this scheme the limiting temperature is also
taken to be of this magnitude. This suggests that the Hagedorn
XThe field theory version of this model was discovered some time
after the string theory. The identity of the two theories was not
noted despite some overlap in the authorship.
temperature of 150 MeV which describes the hadron distribution (and

can perhaps be thought of as string-like excitations of tubes of
gluonic flux) may only be an approximate notion valid for a limited
energy regime, and that there is no true limiting temperature below
the Planck energy, where all sorts ot confusing issues arise.
It is amusing to remark that this radical shift in viewpoint
for the string theory away from a model of hadrons and toward a
theory of fundamental fields has a close historical parallel in the
evolution of thinking about Yang-Mills theories. The latter started
out as a model of pions, rhos, and nucleons and has now culminated
in the highly successful QeD and SU(2) x U(I) theories. I believe
that if the string ideas to be described are to have a chance of
being right they must reduce to these Yang-Mills theories in the low-
energy limit and only represent an extension or modification at
energies higher than those of present experiments. Fortunately, a
construction satisfying this "correspondence principle" appears to
be possible, except for one nasty detail that will be discussed.
In the original formulation of the spinning-string theory [14J
the meson sector contained two types of particles which were referred
to as even and odd "G parity". The states of lowest mass were an odd
G-parity pion with o<')n.~ = - 1/2 and an even G-parity pion with
0(' .m.~ O. The obvious problems for a hadronic interpretation were
that these particles were on Regge trajectories one-half unit too
high, there was no place for strange particles, and the dimension of
space-time needed to be 10. It was obvious that a theory without
tachyons could be obtained by considering only the even-G subspace.
This seemed pointless at the time not only because it did even further
violence to a hadronic interpretation, but because when the fermions
were added into the theory the exchange of even G mesons in fermion-
fermion scattering would give rise to the odd-G ones in the crossed
channel by duality. This problem was overcome in a very nice work of
Gliozzi, Scherk, and Olive [I] . They noted that by restricting the
fermions to be simultaneously Majorana (real in the Majorana represent-
ation) and Weyl (right-handed) at the same time as restricting the
298 JOHN H. SCHWARZ
boson sector to "even-G" states only, that one had a truncation of

the theory consistent with duality. Furthermore they gave compelling
evidence that in this case the theory has 10-dimensional supersymmetry.
This was shown explicitly in the low-energy limit for the zero-mass
states. Also it was proved that at each of the higher mass levels
there are exactly the same number of bosons and fermions. They also
made the amusing remark that the Weyl and Majorana conditions are
compatible only when the space-time dimension D = 2(mod 8), D = 10
being the first nontrivial case. This was, of course, the dimension
that the string theory was known to require anyway.
A string theory contains open strings and closed strings in
interaction with one another. One must choose an integer n to describe
a U(n) symmetry of the theory, which is introduced by the method of
Chan and Paton D9J ' which will not be reviewed here. The important
point to note is that all open string states belong to the adjoint
representation of U(n), whereas all closed string states are singlets.
In the supersymmetric interpretation of the theory the massless
open strings interact at low energies (compared to the Planck mass)
as a supersymmetric Yang-Mills field theory with gauge group U(n),
while the closed strings describe a supergravity multiplet in inter-
action with it. If we neglect all interactions of gravitational
strength we are left at low energy simply with the Yang-Mills theory.
It is described by the 10-dimensional Lagrangian [1,2]
(4)
The action formed from this Lagrangian is invariant under the global
infinitesimal supersymmetry transformations
i € r; "AQ. (5a)
L += f""o. E
r ll
(5b)
These formulas look the same as in the D 4 case, however we must

note the following
1) 1J4 ~ r,.. and 0;11 ~ ~ r" are 32 x 32 Dirac matrices
satisfying
2) Lorentz indices take 10 values 0,1,2, ... ,9 with

.. = -I, i = 1,2, ... ,9.
~OO = 1, YI\ ~~
3) Aand e are Majorana-Weyl 32-component spinors (which we
recall is possible in 10 dimensions).
4) A~ describes 8 physical helicity states, one for each
transverse dimension of space-time.
5) ~ also describes 8 physical helicity states. Starting with
32 complex components, the Majorana condition makes them real and the
Weyl condition reduces their number to 16. Then the Dirac equation
implies that there are 8 physical modes.
Obviously a Yang-Mills theory in 10 dimensions has little to
do with the real world and has disastrous ultraviolet behavior (al-
though it comes from truncating a theory with very good ultraviolet
behavior). The way to get a sensible 4-dimensional interpretation is
to take 6 of the spatial dimensions to be compact with a spatial
extent corresponding to a very high momentum [20J . Obviously there
are a number of topological possibilities for this compactification
all of which are probably consistent with the basic requirements of
the string theory. The most symmetrical choice is a 6-dimensional
sphere, but another choice that keeps the mathematics simpler is a
generalized torus corresponding to a direct product of 6 circles.
The only important requirement for the consistency of the string
theory is that none of the six radii be exactly zero since this
would change the number of modes and lead to certain contradictions.
300 JOHN H. SCHWARZ
To illustrate the dimensional reduction consider a scalar field ¢'

~n ten dimensions. Letting x be the coordinate of 4-dimensional
spacetime and y the coordinate of the compact 6-dimensional space,
one may perform a Fourier series expansion
• (6)
If the 10-dimensional mass is m~o then the 4-dimensional mass of the

field ¢ \,n\ (x) is obviously
mJ.. (7)
10 •
In the limit L. ~ 0 the masses go to 00 for all but the n = 0 level,

~
and in this limit we can pass from a 10-dimensional Lagrangian to a

4-dimensional one by ignoring all y dependence of ~ (x, y) and
simp ly taking
( ll.
t)b 0
ddt:. £.10
(Note that I take ~ 4 and 0<...
.p to have the same dimens ions
10
--namely the usual one of 4 dimensions).
At the same time as y dependence is dropped, the fields that
transform by a nontrivial representation of 0(9,1) have to be
decomposed into 0(3,1) representations. Thus the 10 vector A~ is
decomposed by the rules
t-"= 0,1,2,3
} m 1,2,3
displaying a 4-vector and 6 spin-O fields. To decompose the 32-

component spinor It we first need a representation of the Dirac
matrices. Taking
rr = "ItA- ® 18 ,r-:::o"l/ l )3 (8a)
r (0oc. O{W\)
}
m+ 3 : (8b)
'IS' ~
TYI 0 m: '.I J..) 3
r ll\+b =
"S- ® (~~ O~) (8c)
o -~
where the ()( I sand r'S are 4x4 real antisymmetric matrices
satisfying
0<. '"' 0( 'n, == _El'nf\.P o<P 8 M

"-
(9a)
pmp'n. ==
_ E"'lIr
r P 8"'''' (9b)
t..o(~.1 rtl.] = 0 ~
(9c)
One satisfies the Majorana-Weyl condi tion by taking
I (10)
~
where °l:i are four 4-component Majorana spinors} and we have taken for
the charge conjugation matrix
(11 )
and for the Weyl condition ~I A - A)

(12)
•
Substituting these formulas in eq. (4) one obtains the 4-dimensional

Yang-Mills theory
( 13)
302 JOHN H. SCHWARZ
The theory in eq. (13) has N 4 type supersymmetry. To see this

one simply sets
i 1,2,3,4 (14)
in analogy with eq. (10) and substitutes ~n the transformation

formulas (5a) and (5b). This results in
( 15a)
(15b)
( 15c)
This theory is special in a number of respects: I) First of all it

is derived from the string theory in which gravity is incorporated
in a renormalizable way. 2) It has the largest space-time symmetry
(N 4 supersymmetry) that a Yang-Mills theory can have. 3) It
apparently has vanishing charge renormalization (this has been checked
through the two-loop order) (21] , and it is the unique 4-dimensional
theory with this property. It is safe to assume that many other
striking features of this theory rema~n to be uncovered.
For these reasons I feel that this theory deserves special
attention as a candidate for a unifying gauge theory of electroweak,
strong, and associated interactions. As a gauge theory in four
dimensions any semi-simple group is allowed, but the string origins
demand that a unitary group be chosen. Since all the particles must
be assigned to the adjoint representation, SU(6) would appear to be
the minimal choice that can contain the standard SU(3)xSU(2)xU(I)
theory with the usual multiplet assignments. To now make physical
identifications a lot of symmetry breaking is required. The breaking
of the SU(n) group is probably not a problem since the classical
potential in eq. (13) has valleys of zeros including points with

almost any breaking of SU(n) one desires. A much more serious problem
is the breaking of supersynnnetry --obviously Bose-Fermi degeneracy
must be removed. Triplet quarks and octet gluons should be light,
while triplet gluons and octet quarks should be heavy, etc. This
problem has not yet been solved (it is the nasty detail alluded to
earlier). Unfortunately the Higgs-type mechanisms that Fayet has
discussed for the breaking of N = 1 and N = 2 supersynnnetry (12) are
not applicable in the N = 4 case. It is a problem that the theory
must solve by itself. Our task is to figure out how it does it.
Unfortunately one can't do any serious phenomenology untiJ this
problem is solved. One amusing implication is suggested, however.
Since the fermions come in quartets (of the SO(4) subgroup of the
superalgebra) which will be split if and when the supersynnnetry is
broken, it is natural to suppose that the quarks and leptons which are
falling into doublets of the weak SU(2), will form four such doublets.
The theory will contain many other fermions as well, but it is
tempting to speculate that the others will all be much heavier. If
the mechanism of supersymmetry breaking were understood, one might
even be in a position to say something about the systematics of
quark and lepton masses.
In conclusion, by considering a theory in which gravity occurs
with the infinities of quantum corrections under control, we have
been led to focus attention on a very specific gauge theory of the
usual interactions of particle physics. This theory deserves further
study to determine whether it could possibly be right. If it should
fail, then perhaps an other scheme involving an infinite spectrum of
Planck-mass states that cuts off the divergences of gravitation should
be sought. No other examples are presently known.
III. SUPERSPACE FORMULATION OF SUPERGRAVITY
Supergravity theories have traditionally been formulated as

304 JOHN H. SCHWARZ
ordinary field theories in four-dimensional space-time. Thus, for

example, pure N = 1 supergravity in the first order formalism may be
written as [3]
(16)
where
v = d-et V;
I5r 4J1' = ~t li'r •
The field Vr is the vierbein and ~~s is the connection field, which
is varied independently in the first-order formalism. R V'" Vl> Rrr~
= is
r s
the curvature scalar. The action formed from this Lagrangian is
invariant under the local infinitesimal supersymmetry transformations%
sV'"r = - ~ K (X r
'tho \V ( 17a)
S~ = ~l}o(.1 (17b)
where «(x) is a space-time dependent infinitesimal anticommuting

Majorana spinor parameter. This type of formalism has been extended
in a large number of elegant works to describe coupling to external
matter multiplets, generalizations to extended supergravity with
N = 2,3,4,8, and gauge couplings corresponding to a finite de Sitter
radius (4]
It is generally recognized that this formalism constitutes a
rather special choice of gauge. In this choice the fundamental
invariance group OSp(N,4) has its different pieces handled in very
different ways. The natural suggestion is that just as ordinary
%There is also a formula for S tU~rs, which is not needed when ~~s
is expressed in terms of the other fields using its equations of'
motion ("one-and-a-half-order formalism").
gravity theory (N = 0) can be described as a geometrical theory of

space-time, so the supergravity theories can be described as purely
geometrical theories of superspace --an extension of ordinary space-
time to a space with additional anticornrnuting coordinates. One way
of making this natural is to note that the representation of the
Poincare algebra given by
+ )
(18)
where €~ ~ are infinitesimal parameters for Lorentz transformations

and ~~ describe infinitesimal translations, can be extended to a
representation of the supersyrnrnetry algebra by
(19a)
-I
e·~ et + 0(
Y I
·r ~ (
rv )""
~ "''''tS'': r e~ ) ( 19b)
where e"i are the four components of a Majorana spinor satisfying
In the case of extended supersyrnrnetry a representation ~s expressible

using N such e 'so
A geometrical description ~n superspace of supergravity theory
(eq. 16 slightly generalized to include a cosmological term and
vector-spinor matter) was carried out in collaboration with M. Gell-
Mann, P. Ramond, and L. Brink [22,23,24] Earlier pioneering efforts
on geometrical theories in superspace were made by Arnowitt and Nath
[25J as well as Wess and Zumino [26] , in particular. Many of our
results appear to be closely related to ones of these authors [27J
(and others [28J ), although the precise connections are not always
easily seen. With this clearly understood let me now describe our
efforts without attempting to trace the historical origins of each
concept.
306 JOHN H. SCHWARZ
We use a generalization of the vierbein formalism of ordinary

spacetime --the "vielbein" formalism of superspace. In this approach
the vielbein v:. (x, e) and the spin connection h~ (x, £7) are all
treated as independent fields. In particular we will find an action
[241 from which the equations of motion are obtained by varying all
components independently. This is to be contrasted with some other
formalisms in which external constraints are assumed. The meaning of
indices in the case N = 1 is
spinor vector Lorentz trans.
Tangent abc mn p rs, tu, ...

space
Base
space 0( rt f v f
We also use collective 8-valued indices
A (a,m), B = (b,n), ...
and 14-valued group indices
X (A,rs), Y = (B,tu), ...
The generalization to N >1 is obvious~ but we will only consider

N = in this paper~ We also set the gravitational coupling constant
K= {41YG = 1. The "charge" e is related to the de Sitter radius
R by e = ~/R. An inverse vielbein is defined by
and a Latin superconnection by
The superconnections are gauge fields for local (in superspace)

Lorentz transformations. For such a transformation with infinitesimal
rs
parameter € (x, e) one has
+ • (20)
Accordingly one can define a derivative covariant with respect to

local Lorentz transformations (but not general coordinate transform-
ations) by
J.. I "5
+ .;L n"
+ X"-5 ) (21)
where X are the generators of Lorentz transformations. Denoting

rs
13 A and Xrs collectively by Gy , "flat" superspace without super-
gravitational disturbances is described by the algebra
}
(22)
Z
where CXy are the structure constants of OSp(I,4) and the [ •• ~ 1
symbol means connnutation or anticonnnutation "as appropriate". In the
general case of curved supers pace one has
(23)
Z
where RXY are the superspace curvatures and torsions. One other
convention we have found convenient is to raise and lower spinorial
. ab -.I ab
tangent-space indices with the charge conjugation matrlx C (= 0 0
in the Hajorana representation).
In previous work [22,23J we showed that equations of motion and
transformation laws implied by eqs. (16) and (17) would follow from
making a special gauge choice in superspace provided the covariant
superspace equations (including a de Sitter charge e and axial-vector
external source J~ describing vector-spinor matter) were the following
308 JOHN H. SCHWARZ
-2e( ()rs)
ab
o
(24)
2 t JU
E rsu 5
Now these 420 equations describe the dynamics of the 112 fields ~
and hAs . Therefore they are highly redundant. It has been shown by
MacDowell [29] as well as by the Caltech group [30] that a subset of
112 of these equations can be selected from which the others follow
as a consequence of the Bianchi identities :
Gz 1 + cyc. perm. o (25)
The problem whose solution I want to discuss now is the construction

of an action which when varied with respect of ~ and h~ gives a
set of 112 equations from which the equations in (24) follow. This
action will be expressed in terms of a density in superspace
(26)
since we want to obtain equations that are local in superspace.

The method of solution for JC is a very general one. (We have
verified in particular that it can be used to obtain the usual
Einstein theory with cosmological term). We begin by considering the
most general expression linear in R~
(27)
where
v graded det (VA) (28)

A
is a scalar density. Thus the action so constructed will be invariant

under general coordinate transformations in superspace provided only
. a constant an d BCD
that A 1S . covar1ant.n
X 1S . I pract1ce
. we tak e BCD
X
to be the most general linear function of J~ using covariant constants.
The next step is to form the 112 equations of motion obtained from
eq. (27) by varying v~and h~s. Then we require that these equations
be satisfied in the special case of no supergravity (i.e. when R~B =
C~B)' This almost, but not completely, determines B~D. (It does give
a complete determination for ordinary gravity theory.) To do better we
include the generalization R~ = f~B where the f's are the extension
of the structure constants to include the dependence on the external
vector-spinor source J~. This is information that can be obtained in
principle without knowing supergravity theory in advance. We now have
a unique (up to an overall scale) determination of ~
.p • &.)C\. ..
r;I.../V OC - 31. 0"" b Rh.4...
-.1.4e 3'
v a.1>
+ ~ elo)&) R'"Q.!.
'"'. 1"(,.
+ "" L \ 5
rs:" ) o..b
J&.S
,S
.J5
R"'It.. . (29)
If one is only interested in pure supergravity theory he can set

e = J~ 0 and consider just the two terms in the first line. Either
of these two terms separately gives a subset of the equations of
motion. A specific ratio between their coefficients is only determined
1n the presence of the external source.
Having obtained eq. (29) the remaining task is to verify that
it gives correct supergravity equations in the general case.
310 JOHN H. SCHWARZ
We have explicitly formed the equations of motion in ref. (24) and

found this to be the case. It also gives a minimal set of 112
equations, somewhat different from the selections of refs. [29,30J
We believe that the procedure by which this result has been obtained
can be generalized in a straightforward way to N> I, although this
still remains to be done.
The formal developments described above have so far only been
used to rewrite a known theory in another way. In doing so we were
able to expose a larger symmetry than was apparent previously --14
local invariances in superspace (6 local Lorentz invariances and 8
general coordinate invariances) rather than 14 local invariances in
space-time (6 local Lorentz invariances, 4 general coordinate
invariances, and 4 supersymmetries). As a consequence all the fields
of the theory can be viewed as geometrical. This formalism may prove
useful for studying various technical issues in supergravity theory.
It also suggests a number of possible new theories to investigate.
What happens when one chooses different values for the coefficients
BXCD ?. One doesn I t destroy local Lorentz • •
~nvar~ance or general co-
ordinate invariance by doing so, but the theory is changed. It should
be interesting to investigate whether such theories are well behaved
and what their particle content is. If they contain more states, this
would suggest that the action we constructed contains additional symme-
tries beyond the ones we identified that single out this particular
choice of coefficients. That would also be worth understanding. Finally,
let me remark that it should be possible to construct a description
of the spinning-string theory in a superspace with 10 Bose and 16
Fermi dimensions.
I am indebted to M. Gel I-Mann, P. Ramond, and L. Brink who were

my collaborators for the work reported in the last section. I am also
grateful to P. van Nieuwenhuizen for a careful reading of the manu-
script.
REFERENCES
F. Gliozzi, J. Scherk, and D. Olive, Nucl. Phys. BI22 (1977) 253.
2 L. Brink, J.H. Schwarz, and J. Scherk, Nucl. Phys. BI21 (1977)77.
3 N = I supergravity was formulated by
D.Z. Freedman, P. van Nieuwenhuizen, and S. Ferrara, Phys. Rev.
DI3 (1976) 3214
S. Deser and B. Zumino, Phys. Lett. 62B (1976) 335.
4 For a review of extended supergravity theories and a complete
list of references to them see :
J. Scherk, LPTENS 78/21, Invited talk at the July 1978 Cargese
Institute.
5 A. Salam and J. Strathdee, Nucl. Phys. B80 (1974) 499 ;
P.H. Dondi and M. Sohnius, Nucl. Phys. B81 (1974) 317.
6 R. Haag, J.T. iopuszanski, and M. Sohnius, Nucl. Phys. B88 (1975)
257.
7 M. Kaku, P.K. Townsend, and P. van Nieuwenhuizen, Phys. Lett.
69B (1977) 303 ; Phys. Rev. Lett. 39 (1977) 1109.
8 For a review see P. van Nieuwenhuizen, CERN preprint TH.2473,
Orbis Scientiae (Coral Gables) 1978.
9 M. Gell-Mann, to be published.
10 D.Z. Freedman and A. Das, Nucl. Phys. BI20 (1977) 221.
II S. Hawking, lecture given at San Francisco EST meeting, Jan. 1978.
12 P. Fayet, Nucl. Phys. BI13 (1976) 135.
13 "Dual Theory", edited by M. Jacob, North-Holland Publ. Co. (1974);
J. Scherk, Rev. Mod. Phys. 47 (1975) 123.
14 A. Neveu and J.H. Schwarz, Nucl. Phys. B31 (1971) 86.
15 M. Ademollo et al., Nucl. Phys. BIll (1976) 77.
16 P. Di Vecchia and S. Ferrara, Nucl. Phys. BI30 (1977) 93
E. Witten, Phys. Rev. DI6 (1977) 2991.
17 R. Hagedorn, Nuovo Cimento Suppl. 3 (1965) 147
S. Frautschi, Phys. Rev. D3 (1971) 2821.
18 J. Scherk and J.H. Schwarz, Nucl. Phys. B81 (1974) 118
Phys. Lett. 57B (1975) 463.
19 Chan Hong-Mo and J.E. Paton, Nucl. Phys. BIO (1969) 519.
312 JOHN H. SCHWARZ
20 Many of these ideas were discussed previously in ref. 18 and

J. Schwarz, Caltech preprint CALT-68-637, Orbis Scientiae (Coral
Gables) 1978.
21 E.C. Poggio and H.N. Pendleton, Phys. Lett. 72B (1977) 200
D.R.T. Jones, Phys. Lett. 72B (1977) 199.
22 L. Brink, M. Gell-Mann, P. Ramond, and J.H. Schwarz, Phys. Lett.
74B (1978) 336 (Erratum 76B (1978) 664) ; Phys. Lett. 76B (1978)
417.
23 L. Brink, M. Gel l-Mann , P. Ramond, and J.H. Schwarz, Caltech
preprint 68-656, to be published in Nucl. Phys. B.
24 M. Gell-Mann, P. Ramond, and J.H. Schwarz, Caltech preprint
68-677, to be published in the proceedings of the XIX Inter-
national Conference on High Energy Physics, Tokyo 1978.
25 R. Arnowitt and P. Nath, Phys. Lett. 56B (1975) 177 and 65B
(1976) 73.
26 J. Wess and B. Zumino, Phys. Lett. 66B (1977) 361 and 74B (1978)
51.
27 See, for example, R. Arnowitt and P. Nath, Northeastern Univ.
Phys. Lett. 78B (1978) 581.
28 W. Siegel, Harvard preprints HUTP-77/A068,A077,A080,A089 and
HUTP-78/AOI4,A023 ;
W. Siegel and S.J. Gates Jr, Harvard preprint HUTP-78/AOI9
S.J. Gates Jr. and J.A. Shapiro, MIT preprint CTP # 709.
29 S. MacDowell, Yale preprint 1978.
30 M. Gel I-Mann, P. Ramond, and J.H. Schwarz in Proceedings of the
Dirac Symposium, 1978 (to be published).
THE INFRA-RED BEHAVIOUR OF THE RUNNING
COUPLING CONSTANT IN YANG-HILLS THEORIES
F. Zachariasen
CERN
Geneva, Switzerland
1. INTRODUCTION
In Quantum Electrodynamics (QED) the infra-red (IR) structure

of the theory is well understood. In Quantum Chromodynamics (QCD)
it is not, though considerable effort has been spent in recent
years in studying it. One reason for this interest is the hope that
(because of the apparently stronger IR singularities in QCD over
those in QED) confinement can be understood through the IR proper-
ties of QCD. Indeed, it has been argued l ) that an "effective po-
tential" in momentum space between quarks is proportional to
g2(q2)/q2 where g(q2) is singular in the IR (that is, as q:2 -+- 0)
then the "effective potential" produces confinement. In particular,
if g2(q2) ~ 1/q2 as q2 -+- 0, this can be crudely translated into
a potential growing linearly with distance for large distances.
There are clearly difficulties and uncertainties with these ideas.
First the identification of g2(q2)/q2 as an "effective potential"
has been made only in a leading log study of the IR properties of
QCD, and an investigation of non-leading logs casts doubt on it 2 ).
Second, g2(q2) is not really gauge invariant, and it is clearly
uncomfortable to artribute a physical result, such as confinement,
to a non-gauge invariant quantity.
313
314 F. ZACHAR IASEN
We do not, here, propose to involve ourselves in the discus-

sion of whether or not the above speculations are correct, but we
do wish to study what can be said about g2(q2) in the IR limit,
and the remarks in the above paragraph can be taken as our motiva-
tion.
A convenient way to identify g2(q2) is to study the gluon

propagator in axial gauge, for in this gauge the gluon wave func-
tion renormalization constant determines g2(q2) directly:
(1.1)
Within the class of axial gauges, Z, and hence g, 1S gauge in-

variant.
In QED we recall that a convenient method of obtaining the IR

behaviour of the electron propagator is to use the Dyson equation
combined with the Ward identity to generate an integral equation
whose solution defines a function having the same IR behaviour as
the true propagator 3 ). We may hope that an analogous procedure
can be used in QeD to obtain the IR behaviour of the glue propaga-
tor.
For the sake of completeness, we give a brief outline of the

QED case. The Dyson equation reads (for convenience we take a
spinless electron)
:: f 2. 1'-1 6
L
~ e 61. J" ~/( I) (It) s ( ,- l! )
. (z.,-l~ ( f..., - f¥.) 1; (" 1- ki I.). (1.2)
RUNNING COUPLING CONSTANT IN YANG·MILLSTHEORIES 315
Here S is the electron propagator,
f) (1) ;
/"
is the photon propagator in a gauge specified by the parameter ~ ,
and rv is the proper vertex function in which a photon of momen-
tum p-k connects to one of momentum k.
In general, the vertex can be written
r (f. f-~' 1)
/ ~ J
::. (zr-I.)
7
r,
z.
(2. 1- I.)
r (k -
7
z:.
(ZI-I.)·/c
It
) r z. (1.3)
where rl and rz are scalar functions of pZ, (p_k)z and k2•
Now the Ward identity fixes rl in terms of the electron pro-

pagator:
...... s-'1,) - S-'(,-I&)

I , ::
/"_ (,-le)'-
(1. 4)
r2 is, however, not determined. Thus in general we cannot use (1.2)

to evaluate S.
In the IR limi t, p2 -+ mZ (or small k), however, we know

1:
that D(k z ) -+ Z3/k2 (where the physical charge e = zl eo) and
the rz term in r does not matter. In this limit, therefore,
316 F. ZACHARIASEN
(1.2) does become a closed integral equation defining a function

having the same IR behaviour as the true propagator. This equa-
tion is
(1.5)
..
The equation 1S linear and may be solved by a Laplace transform.
In the IR limit, the solution is
t'r(l'! ,.,.) +
j
~(s)
S(,) ~ ds e
6
where
(,. /.) L)
b t.r J 'Co
4~1
I~ ~~
(i Z
f ~
I e
t'l ( ! ~ ,-/,1. J
It. J_ z.,' t.
I/. /==""
(1.6)
The IR limit (p2 + m2 ) is controlled by s + 00. As s + 00,

RUNNING COUPLING CONSTANT IN YANG-MILLS THEORIES 317
(1. 7)
1+
-)
) (1.8)
In QED this is, of course, a well-known result 4 ). At issue

here ~s whether or not we can use a similar technique to obtain the
glue propagator in QeD.
First of all we shall limit ourselves to a pure Yang-Mills

theory and ignore quarks. The reason for this is our belief that
the dominant IR singularities in QeD come from the pure glue sec-
tor of the theory: quarks feel the effect of confinement but play
no role in generating it. We are thus dealing with a massless
theory, in which the only mass scale is provided by the ultra-
violet (UV) cut-off A. (Upon renormalization, this of course is
translated into a renormalization point M or a renormalization
group invariant -- and thereby physical mass ~.) In a massless
theory, leading log UV and IR approximations coincide. Since QeD
is UV free, we know the leading log UV value of g:
(1. 9)
Formally, this is also the leading log IR result, but since in QeD
b < 0, there is a singularity as q2 + 0 and log q2/A2 + -00.
Hence the result is not valid in the IR: we must go beyond a lead-
ing log calculation. Fortunately, the Dyson equation/Ward identity
approach will permit us to do this.
318 F. ZACHAR IASEN
The Dyson equation for the glue propagator is shown in Fig. 1.

(We will write it down in detail in Section 2.) Our procedure will
be to eliminate the triple and quadruple glue vertex functions ap-
pearing in it using the Ward identities satisfied by these func-
tions, hoping that, as in QED, the undetermined parts of the verti-
ces are irrelevant in the IR regime.
What basis do we have for this hope? In QED the "moment form
factor" (essentially f2) is known to be free of IR singularities,
and, furthermore, the spino logical coefficient of it in the vertex
vanishes with the photon momentum. Hence it can be neglected. In
pure Yang-Mills theory we have not been able to find a proof that
the undetermined parts of the vertex are irrelevant in the IR. As
we shall argue in the following section, we can show that they are
irrelevant in the regime where anyone of the three momenta in the
three glue vertex is small relative to the other two. But, unfor-
tunately, we also need to know the vertex in the regime where all
three are comparable. So, at present, our use of the Ward identi-
ties to eliminate the vertex functions in the Dyson equation must
be looked on as an assumption.
This assumption can also be looked at as a conjecture that

the UV and IR singular parts of functions in massless pure Yang-Mills
theories are intimately connected. Suppose we look for simplicity
at a function of a single momentum q in a massless theory. The
unrenormalized function A(q/A, go) renormalizes according to
where the factor Z is present or absent according as A is UV

infinite or finite. If we choose M = q, (1.10) reads
RUNNING COUPLING CONSTANT IN YANG·MILLS THEORIES 319
..- ~ (1/A, JAN-II, dt,J) .

( 1.11)
Now let q + O. If A has a finite limit (which it automatica1-

ren
1y does if g(q) has, as in massless QED, for example) then the
only IR singularity comes from Z, so only their functions which
are uv singular are IR singular. We know from the UV renorma1iza-
tion program that all functions in the vertex which are UV singular
are fixed in terms of the propagator by the Ward identity. Hence if
all IR singularities are in Z(q/A), all IR singular parts of the
vertex are also fixed by the Ward identity. In QED this is true,
because A (1, g(q)) in (1.11) is indeed finite. In QCD (es-
ren
pecia11y since g(q) may be singular as q + 0) we do not know
if this is true: it must be taken as an assumption. (We remark
that a one loop calculation, is finite in QCD, so the assump-
A
ren
tion is trivially true at this level.)
Once the vertex functions are eliminated, we obtain a (non-

linear) integral equation for a function whose IR behaviour is then
supposed to coincide with that of the true gluon propagator. It is
this equation which we propose to obtain and to study, and from which
we hope to extract the IR behaviour of g(q2).
~--
Fig. 1
320 F. ZACHAR IASEN
2. THE PROPAGATOR AND VERTEX FUNCTION
In axial gauge the gluon propagator ~~~(q) describing the

propagation of a gluon with four-momentum q satisfies
:: ()
(2.1)
where n is the direction of the gauge. It can be written ~n the

form
(2.2)
- A
in terms of two scalar functions A and B. These depend on the

dimensionless variables q2/A2 *) and
-
... (2.3)
\<]e define
(2.4)
so that n'N O.
*) The UV cut-off A2 provides the only mass scale in the theory,

so q2 is referred to this. Upon renormalization, the mass
scale will be translated into a renormalization point ~12, or
equivalently, a renormalization group invariant mass ~2 which
has physical meaning.
ab
\fuen B= 1 and A = y, ~ becomes the free gluon propa-
( ~v
gator ~ O)ab. We shall henceforth suppress explicit colour in-
~v
dices, in the interest of simplicity.
We define an "inverse gluon propagator" II through

~v
~~ Tf).v - /'&11 ~ H~/,." (2.5)
*)
The extra term 1n (2.5) results from the constraint (2.1) •
Since qvII~v = II~vqv = 0 we can write
~v - ( ~I - 'J-f. ) ,
+ (I
/'''
- Y,. ~ f" - (/- y)
~,.~
,.-- + tf:J!/ } I. (2.6)
Evidently the scalar functions f and g satisfy
nil'
Yf- d - ~ ~" I
,.,' /- ~" (2.6a)
and
(2.6b)
*) We could avoid this term by adding ~q~qv/(q·n)2 to ~~v and

taking the limit ~ + 0 at the end of the calculations.
322 F. ZACHAR IASEN
and are related to A and B by
8 : A :. (2.7)
The inverse propagator satisfies the Dyson equation
+ 1 fa;.
(2.8)
. L\~'I1' (,') " 7;,~/1I (~~,,'-,.)
which is represented graphically in Fig. 1. In Eq. (2.8), k'

q-k and dk stands for
(2.9)
where CA is the Casimir eigenvalue of the adjoint representation

of the colour group. reO) is the bare vertex function: explicitly
"r
('J
)., >'a..As
(1,.I 9a7 '.1) :.
.,.. ~ye.l"c. ,e r- ....d -,. ~.~j
(2.10)
Equation (2.8) is plagued with overlapping divergences which

makes it inconvenient for determining the UV behaviour of IT
]l\!
Nevertheless we may expect it to be useful for studying the infra-
red behaviour of IT since in this case the relevant values of
]l\!
the magnitudes of the momenta k, k' and q should be much less
than the UV cut-off A.
The proper vertex r appearing in (2.8) satisfies the Ward

identity
(" ) - ~ ~ If.) (2.11)

I .. •
In the kinematic regime where all momenta are small (compared to A)

we have shown that (2.11) completely determines the vertex function
r. If the corresponding Ward identity for the four-gluon vertex
also determines it in the same kinematic regime, then the Dyson equa-
tion (2.8) becomes a closed set of non-linear integral equations de-
termining the infra-red behaviour of the gluon propagator,
We have not yet completed our analysis of the four-gluon ver-

tex. Nevertheless, we feel that the results obtained by ignoring
the four gluon term in (2.8) (even though this violates the expli-
cit gauge invariance of the theory) and using the results of our
study of the three gluon vertex are of sufficient interest to justi-
fy an interim report at this time. We do not anticipate that our
basic conclusion will be altered when we do include the four-gluon
term.
Even without the four-gluon term, (2.8), together with our

solution for r, yield a non-linear integral equation for the
infra-red behaviour of the propagator. We shall search for solu-
tions of this equation in which the propagator is an infra-red singu-
lar factor, independent of the gauge choice n]l' times the spino-
logy of the free propagator. We do this for several reasons: first,
324 F. ZACHARIASEN
because this is the structure of the IR behaviour of the electron

propagator in spin 0, spin ~, and spin 1 QED. Second, because it is
evidently true to lowest order in perturbation theory. Third,
because it yields a great simplification of the equation.
Thus we look for solutions to the propagator in which f 0

and
only; that is, solutions of the form
(2.12)
with ·1
Z
Such a solution is equivalent to a propagator in which
A ~ Y8
so that
L3
7"
I, J :
ll'~") (I _ ~ If.,
,'\. 7~
",.
1- ~~)
",'"
(2.13)
-- i'/'~" J ~~/,)
/ .
'"
With the above form for IT , we have shown that the most
]JV
general expression for r consistent with the Ward identity, Bose
symmetry, and free of kinematic singularities is
- I
(2.14)
'a: '-' I.) I

• J~
.JI F (11 ; ",1& J"" - (1,) "I. (9. }I", ]
[(9" '.I X~ J,. I

- (~.,~ )('1,,) ~I]
We have here temporarily re-instated the colour indices a,

b, c; thus f abc denotes the structure constants of the colour
group. The functions F and G are free of kinematic singulari-
ties; F is symmetric under an interchange a#- b, qI +r q~ while
326 F. ZACHARIASEN
G is antisymmetric. Aside from this, F and G remain Ullueter-

mined, since the tensors multiplying them satisfy the homogeneous
Ward identity.
We note that the first line in Eq. (2.14) reduces to the bare
vertex if Z is replaced by unity. The second line has a struc-
ture similar to that of the vertex function in QED. The undeter-
mined parts involving F and G are all proportional to a pro-
duct (ql)Al(q2)A2(q3)A3 and hence vanish when anyone of the
three momenta vanishes. If indeed this is the relevant regime for
the infra-red, then we are led to drop the F and G terms in
(2.14). Thus (2.8) and (2.14) together yield a closed non-linear
integral equation for the function Z.
Let us recapitulate the assumptions needed to obtain this equa-

tion:
i) The infra-red structure of IT is a gauge invariant scalar

1 ~v
function Z- (q2/A2) times the free rr~v.
ii) The contribution of the four-glue vertex in (2.8) can be ig-

nored.
iii) The F and G terms in the expression (2.14) of the vertex

do not contribute in the IR regime.
Of these, only (iii) is really essential, and its validity must be

studied further.
With regard to (i) and (ii) we make the following comments:

First, we have carried out an analysis of the Ward identity for r
without assumption (i); that is, including both the functions f
and g in TI. We have obtained a general expression for r, free
of kinematic singularities, analogous to and a generalization of
(2.14). The use of this form in (2.8) then yields a pair of coupled
integral equations for f and g. This more general situation will
be discussed in a separate paper.
Second, we are now engaged in a study of the Ward identity for

the four-gluon vertex which will fix it in terms of f and g.
Thus we will be able to include its effect in (2.8) as well. This
will also be reported elsewhere.
3. THE INTEGRAL EQUATION AND ITS "SOLUTION"
Given the ansatz (2.12), and the resulting vertex (2.14), the
Dyson equation (2.8) becomes an integral equation for Z(q2/A 2 ).
+
L
z
j f'l..'"
Lit. ~/I..) 1(1,')
(3.1)
where ir(O) is given in (2.10) and ir is given in (2.14) with

the F and G terms left out. (~, we recall, is defined in
(2.13).)
It is important to keep in mind that (3.1) defines a function

Z which is ~ the true gluon propagator, but only has the same
IR limit. Thus, whi Ie the Z(q) which in the solution of (3.1)
is defined for all q, only the q2 -+ a limit has physical signi-
ficance. Therefore, the mass scale, which for the true propagator
comes in because of the need for a UV cut-off, can appear quite dif-
ferently here. Here, in fact, we have two choices. We can insert
no cut-off and look for solutions Z which are well behaved as
328 F.ZACHAR IASEN
q2 + 00. In this case the solutions must scale, so that Z = z(q2fM 2)

where M is arbitrary. Alternatively, we can put in a cut-off
and allow badly behaved ZI S as q2 + 00; then A2 becomes the
mass scale and Z = z(q2fA 2 ). This, we emphasize, makes a different
equation from the first choice. In either case, however, we care
only about the q + ° behaviour.
First of all, on grounds of gauge invariance, we expect
rr~v(q)lq=o = 0. The right-hand side of (3.1) is readily evaluated
at q = 0, noting that
(3.2)
in this limit. We find after a tedious but straightforward calcula-

tion, that the triple gluon terms contribute
To guarantee II
~v
(0) = 0, then, we must require the integral to
vanish. This means we must have
(3.4)
Equation (3.4) should be viewed as a constraint on the func-

tion Z(k). The constraint is evidently satisfied if Z(k)
const/k 2 , but can of course also be satisfied in other ways.
We now wish to look for solutions Z(k) to the integral equa-

tion (3.1), subject to the constraint (3.4). To simplify the spino-
logy, let us mUltiply (3.1) by n n /n 2 • This removes the quadrup1e-
].1 V
glue term from the equation, because the bare quadruple-glue vertex
transfers the gauge direction n to one of the internal propagators
].1
to which it is orthogonal.
The quadruple-glue tadpole diagram (see Fig. 1) is also elimi-

nated, because fdk ~AA(k) = o.
Thus we obtain the integral equation*)
-
Z
I
OZ(J,) - ~('j.)
( 'if#' (9.,. "J r (3.5)

?t /.') h.1e. k &- ,. "
,
~O'I - "2 t,., 1.'1 v ~ r¥'

2{ I.) ,,·Ie '
-,
Ie'" ~
(1+ )
*) We can obtain another independent scalar equation from (3.1) by

taking the trace -- that is, by multiplying by 0].1V. If our
ansatz (2.12) is correct, this equation will lead to nothing
new. If it is not, then it, together with the equation obtained
from n].1nv/n2 • IT].1V provide two coupled equations for the two
functions A and B appearing in the propagator, as outlined
in Section 2.
330 F. ZACHARIASEN
".{/,.-J"j
I
h L
(3.5
cont'd)
Equation (3.5) is now a scalar non-linear integral equation for

the function Z(q). What we should like to do is to study this equa-
tion, to see whether or not it has solutions. If it does we should
like to study their uniqueness, and finally to identify the IR be-
haviour of those solutions. As a matter of practice, however, we
are unable to carry out this program. Instead we content ourselves
with the following.
We assume there exists a solution which has a power singularity

Z(q) + (M2/q2)a as q2 + O. Equation (3.5) then provides a con-
straint on what this power can be. To see how this constraint comes
about, we rewrite (3.5) in the form
v- J "c It ~/J..) K.
- I I +- ~-, Jc7f:'
J It"-
-:;..
(3.6)
~(1- )
I -f -Y
y-' Ie L j ,~
~/4 J ~(I.' I L
where the dimensionless kernels K and L can be read off from

(3.5). Now suppose Z(q) + (M2jq2)a. Then (3.6) will, as q2 + 0,
take the form
- (3.7)
where the functions F and G are evaluated by carrying out the

integrals in the numerator and denominator of (3.6). Then, pro-
vided a > 0, the two sides of (3.7) will coincide if there is a
value of the power a such that
F(I(~ y/
/ ~
~(~ y)
(3.8)
for all y.
We have not yet carried out this procedure for an arbitrary

power a, because we have not been able to, in general, evaluate
the functions F(a,y) and G(a,y). We can, however, note the
following. As a + 1, both F and G become singular like
(1 -a )-1. (This is easily seen from an inspection of the kernels
K and L defined through (3.6).) The coefficients of these singu-
larities are determined by the region k« q in the integrals in
(3.6). In this region, the integrals can be evaluated easily, so
that the coefficients of the singularities can be calculated. They
turn out to be equal: that is, as a + 1,
+ ...
332 F. ZACHARIASEN
and
C{y)
+
1- «.
with the same coefficient C('i). This again suggests that the
power a = I characterizes the infra-red singularity of Z, so
that as q2. -+ 0,
(3.9)
It is important to emphasize that the appearance of a =I

is not a coincidence, but has its roots in the form of the Ward iden-
tity, and in particular in the fact that the vertex involves the
structure
J-(Ic.}- r(J.'/
Ie. L_ 1:. "
As we already saw in (3.3), other powers than a I appear not

to work.
Conversely, it is even more important to bear in mind that

we have not shown that (3.9) is true; we have only shown that
if there exists a solution with a power singularity, then there is
a suggestion that the power is one. More work is needed to confirm
this suggestion.
Nevertheless, the l/q2. behaviour is very encouraging. A

naive interpretation of its meaning suggests a confining effective
potential V(r) ~ r for large r. Also note that the power we
obtain is independent both of the coupling constant (and therefore
our "solution" is a non-perturbative one) and of the colour gauge
group. Thus any non-Abelian gauge group produces the same IR be-
haviour, and hence the same confining effective potential.
RUNNING COUPLING CONSTANT IN YANG·MILLS THEORIES 333
He have, unfortunatley, no deep physical understanding of

why the precise power a = 1 should arise. Artificial models ana-
logous to the equation discussed here can give singularities with
different powers 5 ). Other arguments for confinement, also giving
rise to linear potentials, are often based on the existence of
classical solutions to the Yang-~-'!ills theory such as instantons
or merons. The existence of such solutions does not affect the
form of the Dyson equation or Hard identity used here. Thus any
physics which these solutions represent should be included in our
approach, and presumably r.J.anifests itself in the existence of unusual
non-perturbativ~ solutions. Perhaps it is one of these which we have
found, though any ~onnection of our result to instantons is, to say
the least, completely obscure at present.
In any event, we feel that the results obtained here are suf-
ficiently encouraging that the Dyson equation-Hard identity approach
is worth pursuing further. In particular, we need to go beyond the
small momentum expansion, to study the existence and uniqueness of
solutions. We also need to fully include the quadruple glue term,
to assure ourselves that it does not alter anything. And we need
to incorporate quarks and to clarify the connection between a
(1/q2)2 singularity and confinement.
The work described here is being carried out in collaboration

with James Ball and with Earshall Baker. As is evident from the
preceding, the work is not yet finished. We hope we will have
further results to report in the near future.
334 F. ZACHARIASEN
REFERENCES
1) J.N. Cornwall and G. Tiktopoulos, Phys. Rev. DID, 2937

(1977) •
2) J. Frenkel et al., Oxford preprint 67/76 (1976).
3) J.M. Cornwall and G. Tiktopoulos, see Ref. 1).
4) E.S. Fradkin, Nuclear Phys. 76,588 (1966).
5) J.S. Ball and F. Zachariasen, CALTECH preprint 68-647 (1978).

A NE\o] VIEWPOINT ON de SITTER GAUGE INVARIANCE IN GRAVITY
Pietro Fre
Istituto di Fisica Teorica - Universita di Torino
C.so M. D'Azeglio, 46 - 10125 Torino - Italy
In this contribution I report on a new method to implement

de Sitter Gauge Invariance in the theory of Gravitation
recentely developed by myself,,2. The purpose is that of ob-
taining Einstein Gravity as the form taken in a particular gauge
by an SO(2,3) covariant theory. Due to the isomorphism between
SO(2,3) and Sp(4,R) the theory can be subsequentely rewritten
in four-dimensional spinor notation as a symplectic theory. This
is the first step in the construction of an orthosympletic gauge
theory, namely a formulation of supergravi ty3 based on the su-
persymmetry algebra OSP(4/N)4.
The main idea of the method is the following. Besides RAB
the curvature 2-form of the de Sitter connection W AB • I intro-
duce a funfbein 1. -form E' A and an auxiliary a-form ~A
satisfying ;c,AX,A= • ~ EA is required to be orthogonal to
)(~. There are two theories. In the first the orthogonality
constrain is satisfied setting EA=~:D X" (D= covariant
differential, ~ = mass parameter). ~
f
The corresponding action is in explicit covariant notation that
of 5 :
= ~ ~B Reo Xe.
R/\ e
j6k.'I""
-.It
ASCDe
which yields the equation of motion:
335
336 PIETRO FRE
oc.P D
E"SC.DE I'.. A XE = 0
A
In the soldered gauge where )( = (0,0,0,0,1), equation (2)
contains both the Torsion and the Curvature equation of Einstein_
-Cartan-Trautman Gravi ty6, 7. In the second theory EAt=.1. J)X
A
and the action is 1 ,2: ~
~ AS ~ D E
(3) A.z= -d/~6ktJR I\EI\E J, ~ABcDE
In the soldered gauge this theory gives back usual gravitation
without cosmological term but with a modified energy-momentum A
tensor. Topological solutions of theory 1 can be found where/(
realizes a non trivial mapping of two S sPheres 2 • As pointed
out the starting point in the construction of manifestly cova-
riant Osp(4/N) supersymmetric actions is the symplectic spinor
transcription of theories (1) and (3). It is the following:
where 1(~ is the symmetric sp( 4) curvature spinor and EJ X

are antisymmetric spinors.
REFERENCES
1. P. Fre - CALT-68-661 (1978) unpublished.

2. P. Fre - On de Sitter Gauge Invariance and de Sitter Spinor
Calculus in Gravitation Theory. Torino preprint - submitted
to Nuovo Cimento.
3. D.Z. Freedman, P.V. Ni euwenhuizen , S. Ferrara - Phys. Rev.
D13, 3214 (1976).
S. Deser, B. Zumino - Phys. Lett. 62B, 335 (1976).
4. P.G. Freund, 1. Kaplansky - J. Hath. Phys. 11, 228 (1976).
5. S.W. Hac Dowell, F. Hausouri - Phys. Rev. Lett. 38, 739 (1977).
6. A. Trautman - Bull. Acad. Pol. Sci. Sere Sci. Hath. Astron.
Phys. 20,185,503,895 (1972).
7. Y. Neeman, T. Regge - Rivista N.C. 2, 1 (1978).
LIMITING TEMPERATURE, LIFETIME, SIZE OF THE CENTRALLY PRODUCED
HADRONIC MATTER AND ITS ASTROPHYSICAL IMPLICATIONS*
Laszlo Gutay
Purdue University
West Lafayette, Indiana 47907
ABSTRACT
Studying the reaction p + Al ~ P + charged + X we measured the

lifetime of the centrally produced hadronic matter1 ; To ~ 2.2 X
10-24 sec. From the reaction p + p ~ p + charged we determined2
the limiting temperature (T) and the average transverse momentum
at infinite energy « p »; To = 117 ± 4 MeV and = 188 +
5 MeV. Using the same ieaction and the intensity int~rferometer
technique we determined the size (r) and lifetime (T) of the
centrally produced hadronic matter; r = 0.73 ± 0.11 fermi
T = 1.97 ± 0.57 x 10-24 sec. From the velocity of sound
(v/c) at high densities was determinedtv/c ~ 1 or 0, which can
be used to construct the equation of state of neutron stars.
*Work partially supported by the U.S. Department of Energy

(1) L.J. Gutay et al., Phys. Rev. Letters, 37, 468 (1976).
(2) A.T. Laasanen et a1., Phys. Rev. Letter~ 38, 1 (1977).
(3) C. Ezell et al., Phys. Rev. Letters, 38, 873 (1977).
337
EVIDENCE FOR TWO BODY BREAK UP AT A UNIQUE TEMPERATURE
IN HIGH ENERGY P-Xe AND P-Kr COLLISIONS*
Presented l by Laszlo Gutay
This experiment was conducted at FNAL. Using the newly developed

warm gas jet facility, hydrogen-noble gas mixtures were injected
into the circulating proton beam. During injectio~ the beam was
accelerated between 20 to 400 GeV/c. The target mixtures reported
here were 90% H2 - 1010 Xe and 82% H2 - 18% Kr by partial pressures.
Fragments, emerging from the interaction region, were accepted if
they satisfied the ~E . E . VETO trigger. A typical mass spectrum is
shown in Fig. 1. In attempting to find a mass independent dis-
integration temperature,we plotted the kinetic energy distributions
for each fragment mass emerging from Xenon and Krypton. To our
surprise, the inverse logaritmic slope, (temperature) varied bptween
15 and 20 MeV, about a factor of two larger than the nuclear
binding energy shown in Fig. 2. There is a clear break in the
mass dependence of the temperature in the Carbon-Nitrogen mass
range. In trying to understand the origin of this mass dependence,
we were able to show that those fragments which are heavier than
Carbon had emerged as a decay product of a common parent
(progenitor). The mass of this progenitor is denoted by A* and it
is about twenty nucleon masses less than the target mass(A~ ~ A-20).
*Work partially supported by the U.S. Department of Energy

(1) The authors of this paper are: J.A. Gaidos, L.J. Gutay,
A.S. Hirsch, R. Mitchell, T.V. Ragland, R.P. Scharenberg,
F. Turkot, R.B. Willmann, and C.L. Wilson.
339
340 LASZLO GUTAY
8000r-------------------------------~
95 fLm thick 6E detector
B p+ Xc ..,. fragment
7000 EF >15 MeV
6000
II)
~ 5000 c
w
>
~ 4000
o N
ffiCD 3000
:E o
:J
Z 2000
1000
5 6 8 9
Fig. 1
In other words, we showed that

p+A-+ A* + X
P
Al + A2
f f
If all fragments heavier than carbon emerged as the result
of a two body decay of their progenitor A;, they can be lumped
together, after kinematic corrections, to determine a single
excitation function of A;. In a two body break up, the total break
up kinetic energy (Ep) in terms of the masses (A p ' Af) and the
kinetic energy of one of the fragments (Ef) can given as
Ep AE Ef
Ap - Af
Thus Ef Ef
dn Ap
exp (- Ep ) = exp (_
dE p kTp kTp Ap - Af ) =ex( kTf)
EVIDENCE FOR TWO BODY BREAK-UP 341
20 Q Xe TARGET
+ Kr TARGET
18 -
16
14
>(lJ
~
12
t-
10 -
o 10 20 30
Af FRAGMENT NUCLEON NUMBER
Fig. 2
Thus, the kinematic correction to the temperature is

AP - Af
Tf = Tp
Ap
From the overall fit to all fragments (Af > Ac rbon) we deter-
mined Ap and Tp for Krypton and Xenon. The vafues are; Tp(Kr)
14.5 ± 1 MeV, Tp (Xe) = 15.0 ± MeV, Ap(Kr) = 60 ± 5, ~ (Xe) =
110 ± 10. The resulting Tf is shown in Figure 2 as a straight
line. The good agreement between the straight line and the
individual values of Tf verify the validity of our hypothesis for
the existence of a common progenitor (Ap) and temperature (T p )'
ON THE APPLICATION OF FIELDS
WITH CONTINUOUS MASS PARAMETER
Jerzy Lukierski
Institute for Theoretical Physics,
Universi ty of Wroclaw,
Cybulskiego 36, Poland
We introduce the interacting fields with continu-

ous mass parameter by the substitution:
0-
A,2.(y.) -9 C(~):. ~ fl'K.~r('lC.I.) "f(lC ;"...)
in local interaction Lagrangean le.g. lw..t=toA2&1./.

We get in such a way the field-theoretic model, descri-
bing in perturbation theory the subset of Feynman dia-
grams of local exact theory [1].
Models with fields \of (.;"...) were recently used for
the description of clusters (2-4] . There are three
possible approaches to the description of objects with
continuous mass parameter:
al field-theoretic approach, with asymptotic fields
'f(lC\",a.) ... 'f~t(-- ;",. .) describing correlated mul tipar-
ticle states (5] ,
343
344 JERZY LUKIERSKI
b/ S-matrix approach [6], with all objects with con-

tinuous mass parameter converting in pre-asymptotic
region into stable decay products, in accordance with
physical unitary condition,
c/ hybrid approach, with fields \f ()( ;'3tio) describing
"central" region, and S-matrix description of forma-
tion and decay regions.
Recent applications[2-4] are made within the
scheme c/, with clusters produced in tree approxima-
tion. The spectral function ~(~~ is usually chosen
to have a Breit-Wigner enchancement factor, i.e. it
can be related with the original nonlocal interaction
between the" A -constituents". The next step is the
consideration of unitarized approximations /see [6] /.
References I
.l.J. 1ukierski, Acta Phys.Austr.Suppl.VI, 518 /1968/;

Fortschr.Phys. ~,85 /1975/
2.L. Turko,Nucl.Phys. Bll4, 535 /1976/
3.T. Grabinska and 1. Turko, Phys.1ett. 62B,219 /1976/
4.E. Etim and R.Hagedorn, Nucl.Phys. Bl31, 429 /1977/
5.J. 1ukierski, Nuovo Cimento 60A, 253 /1969/
6.J. 1ukierski, Nuovo Cimento 23A, 716 /1974/
A CLASSICAL THEORY OF STRONG INTERACTIONS (WITH "CONFINEMENT")
E. RECAMI
Istituto di Fisica Teorica,
Universita di Catania,
Catania, Italy.
1 = We refer here to recent work done by P.Caldirola, H.Pavsic,

=
P.Castorina and the present author. ,
1 "2
When physicists took due account of the electromagnetic phenome-

na (besides the mechanical ones), it was necessary to leave Gali-
lean Relativity in favour of Einstein's. Since we are now confronted
with sub-nuclear forces, it is perhaps adviceable to look for a new
Relativity. A first possible generalization of relativistic theories
has been attempted in refs. 3. Here we shall mention another general i-
zation, drawing its inspiration partly from the "Erlangen program" of
physics put forth in ref. ", and partly from the observation that the
symmetries of the most important classical equations 5 haven't been
fully exploited. As a first step, let us assume all physical laws to
be covarinat also under discrete dilatations: x' px ,( p discrete,
]l ]l
]l = 0,1,2,3). Then, let's recall that:
i) The dimensionless coupling-constant squares for gravitational
and strong interactions are Gm 2 /Jic '" 1.3 x lO-"o; Ng 2 /Jic '" 15, where
G and N are the gravitational and strong universal constants in
vacuum, respectively; m and g represent gravitational-charge (mass)
and strong-charge. 1 The values are calculated for m= and the pp1T
.
coupl~ng. Let us call: p -= Gm 2/2Ng '" O. 910-"1
x 1T ·d enta 11 y, ~·f
. I nc~
we call go the average magnitude of the quark 1 strong-charge, then
(N=G=1;n=2,3): go=g/3=m/n/p "'~; so that we expect that the
"small black-holes" predicted by some Authors could merely be identi-
fied with quarks, which appear to possess a strong-charge just equal
(in suitable units) to the Planck-mass: g' '" 10-5grams.
ii) if R"'1026 m is our cosmos "radius" and r is the hadron (pion)
radius, then: r/i", 10-"1 _ P
345
346 E. RECAMI
2 = The last eqs. suggest our cosmos and hadrons (typically,

pions) --both considered as finite objects l -- to be assumable as si-
milar systems governed by laws differing only for the scale-factor p
(which carries R into r and gravitational tensorial field into strong
tensorial field). We can therefore postulate: -Inside our cosmos
(="gravitational universe") the Einstein eqs. with cosmological term
(G=1): R _1g RP -Ag = -87TT /c"'· -Inside hadrons (="strong uni-
Jl\! 2 fl\! p. fl\! . Jl\! '
verses) the scaled El.nstel.n eqs. (N=G=l):
'" '"
R _1g "'p '"
R -Hg = -87TS /c '" (1)
Jl\! fl\! P Jl\!
2 U\!'
where dimensional considerations show l that S =p-lT ; H=p-lA; A~
"'lO-s6 cm; H-l"'O.l barn. It is even possible tg\!deriv~\!a whole "nume-
rology",l connecting our cosmos with the strong "micro-universes";
e.g.: M(cosmos mass)"'lOS"'kg; M=p-2m ; m "'10-28kg . Moreover, if (inter-
• 1 7T 7T . .
nal ) gravl.tons have zero rest-mass, then hadron constl.tuents Wl.ll
exchange (internal) "spin-2 gluons"l with zero strong-charge (and zero
mass; and v=c). One can geometrize the strong-field, since hadron-con-
stituen.ts must be attributed an inertia coinciding with their strong
(not gravitational) charges; so that an Equivalence Principle l holds
inside (and in the micro-neighbourhood l of) hadrons. Fr.om eq.(l), the
geodesic eq. in vacuum of a test-constituent g" in the strong field of
another constituent g' yields in the radial case (inside a hadron)
c[NJ=[GJ;N=l): d 2r/dt 2 = k 2 (l.-2g'/c 2r+Hr 2/3)(2g'/c 21+2Hr/3). In
the large-distance case (r;Sr '" 1 fm) one gets the radial confining
force: F '" -g"c 2Hr/3 a: -r, corresponding to a confining potential
Va:-r2. In the small-distance case (r«r), we get --when we assign
to g" the suitable "kinetic-energy term" (J/g"r)2-- the radial poten-
tial (N=p-lG): V '" -Ng'/r+ (J/g,,)2/ r 2; and for J"'l>i such eq. gives:
V~O for r"'O.Ol fm, and eventually (m in GeV /c 2): J/l>i '" m2 . For the
difficult problem of interactions between two hadrons, cf. the Refs.
REFERENCES
1. P.Caldirola,M.Pavsic & E.Recami: Nuovo Cimento B48:205(1978);
Phys.Letters A66:9(1978); Lett.N.Cim. (in press).
2. P.Caldirola: Lett.N.Cim.23:83(l.978); Nuovo Cim.A45:549(1978);
E.Recami: Rep.INFN/AE-78/5 & 6, to appear in.: "Gentenario
Einsteiniano" vol., M.Pantaleo,ed.,Giunti-Barbera,Florence,
1979); P.Caldirola: ibidem (to appear); D.D.lvanenko: ibi-
dem; P.Castorina & E.Recami: Lett.N.Cim.15:347tl976); 22:
195(1978); R.Mignani: Lett.N.Cim.l6:6(1976).
3. E.Recami, ed.: "Tachyons, monopoles, and Related Topics",
North-Holland, Amsterdam, 1978.
4. L.Fantappie: "Opere Scelte", UMI, Bologna, 1973, p.873.
5. A.Einstein: Preface in: "50 anni di Relativita",Firenze,1955.
ENERGY -MOMENTUM DISTRIBUTION IN e+e- ANNIHILATION
G. Pancheri-Srivastava and Y. Srivastava

INFN, Laboratori Nazionali di Frascati, Italy, and
Northeastern University, Boston, Mass., USA.
In this lecture we present a model for hadronic momentum

distribution in e + e - annihilation based on an approximation to the
gluon distribution in QCDl. Using an expression derived by Etim,
Pancheri and Touschek2 for the 4-momentum distribution d 4 P(K)
of the emitted radiation in any QED process, earlier we were ab
Ie to obtain (an exponential) transverse momentum damping 3 . Ii!"
the present work l we have extended that analysis to the energy as
well as the longitudinal momentum distribution of the radiation.
Similar formulae are also valid for the colorless states of
QCD with only minor changes. Omitting the details of derivation
(see ref. 1), we quote below the results for the longitudinal (K\I)
and transverse (K,l.) normalized momentum distributions:
T(l+8) ~- 1
1 2 2
<. KI\ ") ZL K fL:J:. (ZL) (1)
~
2 2 2
2 .''IT (M2)
dP(K.l) 1
'(iiT(l;~)
zP/2 K (ZT) , (2)
dK,!. <Kl.) .ft _1 T f!._l
2
22 r 2 (p/2)
+ Work supported in part by the NSF, USA.
347
348 G. PANCHERI·SRIVASTAVA AND Y. SRIVASTAVA
v3i r( 1f.!.)
zT = r(#/2) ( 3)
The above formulae are applied to obtain the K" and K di-
stributions of single hadrons under the assumption that the hadro-
nic 4-momentum distribution follows that of the radiation. Figu-
res 1 and 2 show the theoretical prediction and the experimental
result from SPEAR at 7.4 GeV 4
10,.....------------, 10.------------...,
I do"
ddi<. I do"
(GeV·') 4 dX"
0.1 0.1
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
K. (GoV) XII
Fig. 1 Fig. 2
The present approach may well be a viable alternative to the
quark fragmentation models. For gluon jets, we obtain that ~.Kl.)
from quarks and gluons is roughly the same, then (n > glue »
>} ~ n '>quark' a result in qualitative agreement with results from
DORIS on Y-resonance.
References.
1. G. Pancheri-Srivastava and Y. Srivastava, Frascati Preprint
LNF-78/46 (1978); (To be published).
2. E. Etim, G. Pancheri and B. Touschek, Nuovo Cimento 51,276
(1967); G. Pancheri, Nuovo Cimento 60, 321 (1969); G. Panche
ri-Srivastava, Phys. Letters 44B, 109 (1973). -
3. G. Pancheri-Srivastava and Y. Srivastava, Phys. Rev. D15,
2915 (1977).
4. G.Ranson, SLAC-PUB-2118 (1978).
EXPONENTIAL HASS SPECTRUM: AND PION CONDENSATION
A. Tounsi
Laboratoire de Physique Theorique et Hautes Energies,
Universite Paris VII
Th e POSS1. b 1. 1 1ty
. 0 f condensat10n
. ( 1 ) at h1g
. h energy 1n
. a
relativistic quantum gas of pions with fugacity A > 1 gives
the gas a structure similar to that obtained in the statisti-
cal bootstrap model. We consider(2) a more general system of
bosons with mass spectrum of the statistical bootstrap type:
( 1)
The partition function of such a system, taking into ac-

count Bose-Einstein statistics, can be written:
Z((', Y,A) =V<ff~~,;·:ft (i '.. tf:+> ~~~ -A:"") k J

-L
where f= TI and V is the volume of the system.

This partition function has two real singularities:
A detailed study of this partition function and of physical

349
350 A. TOUNSI
quantities (energy density, mean particle number, mean energy

per particle) shows that the behaviour of the system depends
crucially on:
i) the relative value of j8. and/~
ii) the value of the power of the mass in (1).
Different situations are then possible:
The condensation cannot occur. For any value
of a. For -7/2 ~ a < -5/2. The asymptotic
strong bootstrap condition reduces to:
2
2a + lla + 15 =0
whose solution is a= -3 (as found by Nahm).
For a < -7/2 the bootstrap condition is sa-
tisfied but To= J30 is phase transition tem-
perature reached at finite energy density.
It could correspond to the quark liberation
suggested by Cabbibo and Parisi(3).
Condensation sets in at T... =

-,
P...
and pre-
vents the bootstrap condition from being
satisfied.
pion condensation and bootstrap mechanisms

compete directly. For -7/2 ~ a :S -5/2 the
bootstrap dynamics is shown to dominate,
while for a < -7/2 pion condensation takes
over.
REFERENCES
(1) J. Montvay and H. Satz - Nuovo Cimento 39A, 425 (1977).

(2) J. Letessier, A. Tounsi and H. Satz - Nuovo Cimento 45A,
580 (1978).
(3) N. Cabbibo and G. Parisi: Phys. Lett. 59B, 67 (1975).
A FIELD THEORETIC DESCRIPTION OF CLUSTERS
AS AN APPROXIMATION TO THE STATISTICAL BOOTSTRAP MODEL
Ludwik Turko
Institute of Theoretical Physics,
University of Wroclaw.
Cybulskiego 36, 50-205 Wroclaw, Poland
We consider a field theoretic iescription of clusters

production in high energy proton-proton collision.
Clusters are described by means of a local quantum field
with a continuous mass spectrum. Asymptotic states of
this field describe correlated multiparticle states i.e.
clusters (1, 2J •Two models are considered. The first one
is a solvable eikonal model for cluster production [3].
The cluster field is given as
where g(~) is the probability distribution of observing

a cluster of mass squared s before switching-on the
interaction. Having cross sections for the production of
n clusters we calculate the observable mass spectrum.It
turns out that an interaction causes a change in the
351
352 LUDWIK TURKO
shape of the spectrum and shifts an averaged cluster's

mass [4].With increasing energy of the primary particles
this effect is decreasing and the observable mass
distribution tends to the "free" one. This result has an
interpretation in the framework of the statistical
bootstrap model [5] and corresponds to the existence of
the limiting temperature.The next considered model is
a field theoretic multiperipheral cluster model [6].
Cross sections for cluster production are calculated.We
obtain a behaviour of the cluster mass spectrum similar
to the previous one.We obtain also a violation of
Feynman scaling at ISR energies.In the next step cluster
decay is introduced in a phenomenological way and semi-
inclusive distributions of pions are calculated.We get
a good agreement with data.
REFERENCES
1.J.Lukierski, Acta Phys.Austr.Suppl. VI /1969/ 518

2.J.Lukierski, Nuovo Cim. 23A /1974/ 716
3.L.Turko, Nucl.Phys. B114 /1976/ 535
4.L.Turko, 1FT Wroclaw preprint 417/77
5.R.Hagedorn, CERN Report 71-12 and further literature
quoted there
6.T.Grabinska,L.Turko, 1FT Wroclaw preprint 438/78
II SMALL II , IILARGEII, AND liVERY LARGEII TRANSVERSE MOMENTA
IN A UNIFIED HYDl"l.ODYNAMICAL DESCRIPTION
Erwin M. Friedlander+ and Richard M. Weiner¥

Lawrence Berkeley Lab., University
o£ Cali£ornia/USA+
Dept. o£ Physics, University o£ Marburg,
Marburg/Germany¥
Two apparently unrelated e££ects, viz. the behavior of

transverse momentum spectra at IIlarge li PT in p_p collisions 1)( )
and the lIenhancement li o£ such spectra l.n
" p-nucleus co 11""
l.Sl.ons 2 ,
are shown to £ollow in a natural way £rom a hydro dynamical
model in which the space-time evolution o£ the system is taken
into account.
Up to PT = 5 GeV/c a single value (close to u 2 = 1/7) £or
the velocity o£ sound in hadronic matter gives a consistent
description o£ all experimental £acts. Recent observations(3)
at very large PT(5-15 Gev/c) require a jump to u2~1/4,
suggesting the possibility o£ a phase transition.
Re£erences:
1. B. Alper et al., Nucl. Phys. 1£, 395 (1972).
2. J.V'I. Cronin et al., Phys. Rev. D11, 3105 (1975).
3. A.G. Clark et al., Phys. Lett. 74B, 267 (1978);
A.L.S. Angelis et al., CERN-report 1978.
353
CONTRIBUTORS
Steven Frautschi California Institute of Technology

Pasadena California 91125
Pietro Fre Istituto di Fisica, Universita di Torino
Erwin Friedlander Lawrence Berkeley Laboratory

University of California
Ferdinando Gliozzi Istituto di Fisica, Universita di Torino
Laszlo Gutay Department of Physics Purdue University

West Lafayette Indiana 47907
Rolf Hagedorn CERN Ginevra
Julius Kuti Central Research Institute for Physics

H-1525 Budapest 114
Jerzy Lukierski Institute for Theoretical Physics

University of Wroclaw Cybulskiego 36
Istvan Hontvay Centre for Interdisciplinary Research

University of Bielefeld
Werner Nahm CERN Ginevra
Giulia Pancheri Department of Physics Northeastern Univ.

Boston Massachusetts
Johann Rafelski CERN Ginevra
Johannes Ranft Sektion Physik Karl Harx Universit~t

Leipzig
355
356 CONTRIBUTORS
Erasmo Recami Istituto di Fisica Universita di Catania
Helmut Satz Department of Theoretical Physics

University of Bielefeld
John Schwarz California Institute of Technology

Yogi Srivastava Department of Physics Northeastern Univ.

Boston Massachusetts
Ahmed Tounsi Laboratoire de Physique Theorique et

Hautes Energies. 75230 Parig
Ludwig Turko Institute for Theoretical Physics

University of Wroclaw Cybulskiego 36
Richard Weiner Department of Physics

University of Marburg
Fredrik Zachariasen California Institute of Technology

INDEX
Bag Gauge group U(n), 258

with heavy quarks, 164 Gauge theory
with light quarks, 167 on a lattice, 41, 42
Black hole self dual, 45
death cry, 225 Gluino, 292
entropy, 221, 222 Gluon jets, 262
gravitational interaction Gluon propagator
of fireballs, 222-224 in axial gauge, 320
Cluster Group
bar ion cluster equation, 52 superconformal SU(2,2,N), 192,
nuclear cluster equation, 92 196, 197
one cluster grand partition superconformal SU(2,2,1), 192,
function, 106 194-196
Correlations Hadro-chemical reactions, 208-215
color correlations, 27 Hadron resonances
Cosmological entropy, 226-227 separable,.overlapping, 2-4
Curvature of Grassmann space Hagedorn
R :; 00, 189, 192 dense pion packing, 65-67
R = 00, 189, 192 limiting temperature, 61, 82, 90
Duality transformations mass spectrum, 61, 74
in statistical mechanics, 30 nuclear bootstrap equation, 92
in field theory, 31 old bootstrap equation, 59, 237
Dyson equation Instantons
for electron propagator, 314 in QCD, 176
for gluon propagator, 322 dilute gas, 152
for Z(q2/A 2), 327 net magnetization, 153
Ericson fluctuations phase transition in presence
in nuclear physics, 5-8 of a chromoelectric field,
in hadron physics, 9 158
Excitations Mass spectrum
topological, local, 35 Hagedorn, 1, 23
Exterior forms MIT bag, 1, 23
cohomology, homology, 32, 33 dual, 235, 236
Fluctuations Meissner effect
in scattering amplitudes,S chromoelectric, 160
in barion number density, 110 Merons
and perfect paramagnetism, 154
357
358 INDEX
Quark hard scattering, 256 Quarks plus hadrons

Quark recombination model truncated bootstrap, 275
meson production, 245 its phase transitions, 277-285
baryon production, 245 Spinning string theory
generalized Kuti-Weisskopf with Majorana-Weyl
model, 246 fermions, 297, 299
hard and soft processes, 251 4-dimensional inter-
Nuclei collision pretation, 299
thermalization, 202-204 String density of states, 235
geometry, 205, 215 Super Lie algebra OSp(N,4), 293
Phase (hadronic) Higgs mechanism for N > 2, 295
of free quarks and gluons, N = 1 Lagrangian, 304
19, 20 Vertex functions
ordered phase, 23, 24 in QED, 315
diagram for quark-gluon in QCD, free of kinematic
matter, 24 singularities, 325
gaseous bootstrap phase, 117 4-glue vertex, 326
liquid bootstrap phase, 117 Vielbein, 306
Planck temperature (energy), Ward identity
220, 292, 295, 296 in QED, 315
Quark coordinates in QCD, 323
complex twistor, 188
octonions, 188, 193
quaternions, 188, 193

Hadronic Matter at Extreme Energy Density

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Hadronic Matter at Extreme Energy Density

Transféré par

Droits d'auteur :

Formats disponibles

Hadronic Matter

Volume 1 INTERACTING BOSONS IN NUCLEAR PHYSICS

Volume 2 HADRONIC MATTER AT EXTREME ENERGY DENSITY

Volume 3 COMPUTER TECHNIQUES IN RADIATION TRANSPORT

Volume 4 EXOTIC ATOMS '79: Fundamental Interactions and Structure

Plenum Press . New York and london

Workshop on Hadronic Matter at Extreme Energy Density, Erice, Italy, 1978

(Ettore Maiorana international science series: Physical sciences; v. 2)

Proceedings of the Workshop on Hadronic Matter at Extreme Energy Density,

© 1980 Plenum Press, New York

All rights reserved

No part of this book may be reproduced, stored in a retrieval system, or transmitted,

This book originated in the Workshop on "Hadronic Matter at

The development of physics in recent years has filled in the

Other conjunctions between fundamental truths and researches

Sometimes the techniques, the immediate aims, and even the

This is the year of Hagedorn's sixtieth birthday, and it is

We wish to thank Antonino Zichichi for his leadership and

Ericson Fluctuations and the New Argonne

Asymptotic Freedom and Color Superconduc-

Duality Transformation and Confinement in

Thermodynamics of Nuclear Matter from the

The Bag Model and Quantum Chromodynamics . . . • . . . . 149

Quarks and Fermionic Geometry 187

Nuclear Fireballs in Heavy Ion Collisions 201

A Critique of Astrophysical Applications

Particle Production in Soft and Hard Had-

From Hadron to Quark Matter • • • • • • • • • • • • • • 269

Supersymmetry Approach to the Unification

The Infra-red Behaviour of the Running

SEMINARS ON SPECIALIZED TOPICS

A New Viewpoint on De Sitter Gauge Invar-

Limiting Temperature, Lifetime, Size of the

Evidence for Two Body Break-up at a Unique

On the Application of Fields with Continu-

A Classical Theory of Strong Interactions

Energy-Momentum Distribution in e+e- Anni-

Exponential Mass Spectrum and Pion Con-

A Field Theoretic Description of Clusters

"Small", "Large", and "Very Large" Trans-

California Institute of Technology

I. The Search for Highly Excited Hadron States: Introduction

}lost people nowadays think of hadrons as made of quark con~

The subject of this talk is an attempt to verify that hadron

Specific theoretical estimates. Several specific models (statistical

estimates in this talk. However, even if the level density grows at

Evidence from Particle Data Tables. In general levels can be divided

Clearly one cannot use the/separable resonances to establish

Fig. 1 Comparison of theoretical spectrum of hadron levels with

Fig. 2 Schematic plot indicating where rp(mpJ) is large enough to

The most familiar application of statistical ideas is to find

II. Ericson Fluctuations in Nuclear Physics

Once more the appropriate concept originated in nuclear physics

To pursue this point in some detail, let us consider the partial

We normalize IAJI to 1 at the unitarity bound. The number of over-

AJ = AJC + OAJ (5)

AJ is the coherent contribution that occurs if the resonances tend to

100JI ~.!.x rn~J:.- (6)

In a purely statistical reaction,

In all cases the fluctuations are characterized by an energy

LlE '" r (9)

and an angular width

Ericson fluctuations are well established and a useful tool in