Professional Documents
Culture Documents
Institute of Theoretical and Experimental Physics, B. Cheremushkinskaya, 25 117259, Moscow, USSR
Institute of Theoretical and Experimental Physics, B. Cheremushkinskaya, 25 117259, Moscow, USSR
North-Holland, Amsterdam
V A C U U M B A C K G R O U N D F I E L D S IN Q C D AS A S O U R C E
OF CONFINEMENT
Yu.A. SIMONOV
Institute of Theoretical and Experimental Physics, B. Cheremushkinskaya, 25;
117259, Moscow, USSR
Quarks and gluons in the vacuum background field with a finite correlation length d are
shown to be linearly confined inside white states. The string tension for an arbitrary representa-
tion of SU(N) is obtained in terms of d and the gluon condensate and agrees with numerical data
and large N behaviour. The QCD string picture is shown to emerge asymptotically at large
euclidean distances.
1. Introduction
A l , = ~ + a ~. (1)
m
(2)
follows:
when C is the straight line from y to x we shall omit the subscript C in (4). In the
present paper the exact principle of separation in (1) is unimportant and another
equivalent way can be used instead of (2). It is important to stress that the
wave-length spectrum of .~, is bounded from below and therefore A~ does not
influence the small distance behaviour, apart from corrections, like those for
constant background field [2]. Therefore the asymptotic freedom in the vacuum
filled with the .,~ field is not spoiled.
(ii) We assume that A~, F,~ are large in the QCD scale, ( F-2~ ) >> A~cD" In this
case the renormalized coupling constant a s - ln-l(ff2/A4ocD) will be small even at
large distances, and we can define asymptotic behaviour of quark and antiquark
Green functions beyond the confinement scale.
Since a s is small one uses the usual perturbation scheme and the calculation of all
physical amplitudes should be done in several stages. First one computes Green
functions in the background field exactly (since g-A---~- O(g°)). Then one takes into
account interaction via gluon exchanges perturbatively and finally the influence of
sources on the background and their motion in the modified background. The
scheme is assumed to be infrared finite and particular examples of the background
like a constant field or the instanton gas support this assumption.
For quasiclassical solutions A~, F ~ - O(1/g) and they should be densely packed
to form a strong background. As one can see, we do not confine ourselves to
quasiclassical solutions, any type of vacuum will do if a strong condensate is
formed.
In reality [2], (~/z2j,) = 0.5 (GeV) 2 and (F~2) is indeed much larger than A4QCD
for AQC D = 0.15 GeV, but the coupling constant at large distances is not small and
may be around 0.5-1. Requiring P~ to be strong, we have in mind a theoretical
limit, where our approach can rigorously be applied. In a full-fledged theory (F,~)-2
must be found self-consistently, and will probably settle down to moderate realistic
values with a S of order of one. We believe however that qualitatively and to some
extent quantitatively our approach is useful for the discussion of long-range physics,
as will be shown below.
Assumptions (i) and (ii) have immediate qualitative consequences which can be
compared with experiment and Monte Carlo data. We have already mentioned that
a similar picture has been used in [3] with reasonable predictions for hadronic mass
Yu.A. Simonov / Vacuum backgroundfields 515
scales in different channels, quark-gluon mixing etc. The OZI rule violation in S, Ps
channels can be understood [18] as due to the interaction of light quarks with zero
modes of background (anti-)self-dual regions in the QCD vacuum. The intermittent
self-dual and anti-self-dual regions of background field can explain the chiral
symmetry breaking [7].
Another consequence of the strong background field is that all condensates
(qq), ((as/vr)F~,F,,), ~t = (1/V)(((as/8"n')fFff" d4x) 2) are O(g °) (up to anoma-
lous dimension of q q ) - and should be numerically of the same order and large as
compared to A~QCDwhich is indeed true (resp. (270 MeV) 3, (330 MeV) 4, ( - 200
MeV)4). More than that, one can easily persuade oneself that both gluonic con-
densate and topological susceptibility can be reproduced if the density n = IF] of
one topological charge per fm 4 and a statistical law for the difference of the
instanton and anti-instanton density (n - ~) is used.
(iii) All physical amplitudes and Green functions are obtained by averaging over
background field configurations. To maintain the 0(4) and Poincar4 invariance one
must require that the background fields form what is called a homogeneous
stochastic ensemble. In this way the Green functions fY(x, y, z .... ) will depend on
differences x - y , y - z etc. and the momentum is conserved. The average of
irreducibly gauge-noninvariant amplitudes vanishes. These properties are familiar in
background field studies.
A crucial question about the strong background field is whether it can ensure
confinement. Some progress in this direction has been described in [19-24]. It has
been shown there that the linear confinement between quark and antiquark indeed
occurs if the background field is uncorrelated for distances larger than some finite
value, called the correlation length, d [22, 21, 23].
To obtain this result an intuitive, but not rigorous method of hamiltonian
averaging, has been used in [19-21] and a very efficient method of cluster (or
cumulant) decomposition has been used first in [22] and later in [23, 24]. To make
the method applicable to the nonabelian case one should introduce so-called
path-ordered cumulants [23]. Using those the area law has been proved in [24] for
the fundamental charges, provided that the correlator (3) falls off with the correla-
tion length d. In this way confinement has been proved for infinitely massive quarks
and the question emerges whether confinement occurs for other colored objects,
includin~ ~luons.
It is the purpose of the present paper to answer this question. In doing so we
introduce the cluster expansion in the framework of the path-integral representation
for the partition function and Green functions. In sect. 2 we write the grand
partition function and integrating over background, fields obtain an effective
lagrangian for quarks. We do not obtain the most general effective lagrangian for
gluons, quarks and ghosts, and instead we disregard for simplicity gluons and ghosts
(owing to the smallness of g ( F ) ) and retain only quarks interacting with the
background fields. As an application we verify that for external quark currents
516 Yu.A. Simonov / Vacuum background fields
(heavy quarks) there appears a linear confining potential as in [22, 24] but now for
any quark representation of SU(Nc). At this point we can seriously check our
model. The adjoint quarks are confined in the model, in O(g°), whereas they can be
screened by gluons in higher orders in g. If gluons (i.e. the fields in (1)) are
operative in confinement, adjoint charges should not be confined, whereas in the
suggested model they are screened only due to interaction with gluons via small
g(ff) and therefore up to large distances one should see an adjoint string tension.
The numerical experiment seems to support the adjoint quark confinement up to a
distance of 5 lattice units [25]. It is found in sect. 2 for string tension in different
representations j that
2. Basic equations
We are considering euclidean gauge-field configurations in euclidean space-time
and decompose the gauge field in a usual way into a background and a quantum
part as in (1), A. = A. + a., exploiting the familiar background field formalism [36].
A.(x) can be described by collective coordinates, as in the instanton vacuum, and
then a . is constrained by the background gauge condition D.(A)a. = 0 and by the
orthogonality conditions with respect to zero collective modes [6]. A similar proce-
dure can be used for another way of separation different from (2). Using the
background gauge for a.(x) and introducing ghost fields q0 and ep+, we can write
the partition function as [36]
z= f~L~a.~p+~q~/exp[-(So+Sl+S2+S3+S~)], (6)
where
while S 3 includes higher order terms in a., ghost terms and a gauge-fixing term,
which we shall not use here.
The usual strategy in dealing with (6) is to assume some background field
characterized by a number of parameters, to integrate over quantum fluctuations a.,
keeping quadratic terms $2, and in this way to obtain an effective lagrangian in
terms of background field parameters. Finally minimizing the free energy F = - In Z
we get an optimization procedure for the latter and in this way we obtain an
optimized vacuum. To the full extent this procedure has been used for the instanton
vacuum [6].
518 Yu.A. Simonov / Vacuum background fields
= 1
f Xe-S A b(X). (11)
To integrate the A-dependent exponents in (6) one can use the cluster expansion
[31] (sometimes also called the expansion in cumulants) valid for any functional
depending on a stochastic function, e.g. for f ( A ) linear in A~ we have
(exp(-f(A)) =exp ~
m=l
((f"(A-)))(--1)m m!
(12)
((f)) = (f),
( ( f 2 ) ) = ( f 2 ) __ ( f ) 2 ,
and so on.
In our case the exponent in (6) has the A-dependent parts in S 1, S 2, S 3 and S¢.
As was discussed in the previous section, if one is interested in the behaviour of a
system of quarks in the background field A-,, one should take into account the
interaction of quarks with A-, exactly, i.e. to all orders in g, since ,4, is supposed to
be large, A--~- O(1/g). Still we consider g small, even at large distances, since g ( F )
is renormalized in the strong background field, therefore one can neglect interaction
between quarks via gluon exchanges and gluon emission in the first approximation.
Following this logic we concentrate on S+ (10) and disregard in this paper other
terms in (6). With the help of (12) we have
(14)
Yu.A. Simonov / Vacuum background fields 519
where ( D r ) = 0r - ig<Ar), and the effective quark action contains the terms of the
cumulant expansion (12), starting with the second order in "4r:
(15)
w h e r e j ~•a( x ) = - ~Trr a~b(x). One should keep in mind, that ('4r) can be nonzero due
to the inhomogeneous term in the gauge transformation:
( 7a --b
Ar(x)A.(y)) ) = (x __ Xo)p(y-xo)oJ{pr,o.(x,y),
ab (is)
where
pg,o~,l,X,
..)¢{'ab l y)= f01ctdctf01fldfl</-a
F;r(x0+ O~(X__XO))ffobv(xO+
(19)
The bilocal correlator of fields entering (19) can be written in a form, which is a
520 Yu.A. Simonov / Vacuum background fields
slightly improved version of the form used in [24]:
<<F a,~(z)Ft~(z'))) =
~8~b<trF~(O)>[(N:__ 1 ) ( D ( O ) D+ ~ ( O ) ) l- 1
<tr F2,(O)>
- 6(N~2 - I ) ( D ( 0 ) + Ox(0)) gPtt'oP(Z' Z'), (20)
Nc2 - ]
~-~t~t"= - - . ], (21)
2No
one can define through (20) the gauge-invariant correlator (which is equivalent to
(20) in the chosen gauge, if x 0 lies on the straight line connecting z and z')
(trF~(0))
<<tr~(z"z)F°"(z)q~(z'z')F~(z'))) = 1 2 ( D ( 0 ) + Dl(0)) g°" . . . . (22)
tion of the SU(Nc) group. To this end we choose the current of the infinitely heavy
quark and antiquark to be
j}~ = 0, i = 1,2,3,
g2qaq°ltrF~(O))
so.= 1 2 ( ~ - ~ ;71(o))
f foTa~d,'ff~dzdz'D(z-z',~-~'). (25)
where the string tension o ( j ) in the representation j of the quark charge qa = Ta(j)
is expressed through the quadratic Casimir operator
o(j) = g2d2(trF+(0))C2(j)
12(N? - 1) (29)
522 Yu.A. Simonov / Vacuum backgroundfields
g2d2
o ( f ) = 2 - - ~ (tr F:a(O)) • (30)
This result coincides with the one obtained in [22,24]. Here ( t r F ~ ( 0 ) ) is in the
fundamental representation and is the standard gluonic condensate [2] (for N c = 3)
o(j) 4j(j + 1)
(32)
3
which is 2 at large N c and coincides with the prediction [28] based on the
factorization at large iV,..
For small R, R << d, one obtains from (25) a quadratic potential, which for SU(3)
fundamental charges reads
3. Green functions
this point, that physically measurable Green functions and amplitudes are gauge
invariant, and one should average over vacuum fields gauge-invariant correlators
(Green functions), while an average of a gauge-noninvariant object is zero (if an
object is irreducibly gauge-noninvariant, i.e. it does not contain linearly a gauge-
invariant part [42]).
Therefore, of physical interest are Green functions describing evolution of a white
state of some number of quarks, antiquarks and gluons, defined on some space-like
surface in the 4-dimensional space, where the final state is again a gauge-invariant
(white) state defined on another space-like surface at a later time-like coordinate.
Confinement in this language follows from the fact, which will be proved in this
section, that colored constituents in a white state cannot go far from each other,
since they appear to be connected by a kind of string. For an abelian theory without
monopoles the strings do not appear and colored constituents may go astray. This
enables one to consider in this case also separate charges as one usually does in
QED.
We start with the Green function of a white system consisting of a quark at the
4-point ( y ) and an antiquark at (3) connected by the space-like contour integral
~P%(Y, Y) (4). The final state is characterized by x, ~ and ~%(x, £) respectively. The
Green function is
In the spirit of our approach we first neglect the quark-antiquark interaction via
gluon exchanges and quark-pair creation taking into account only interaction with
the background field, then the quark and antiquark propagators factorize in (35):
The quark (antiquark) propagator in the external field satisfies the equation
[ (O
i x0-~ -gA~(x))+m] (37)
portant at large distances, we can write the integrand (38) as a path integral [44]
where the ordering is in the parameter o, 0 ~<o ~ s and [d4z] = lq N=1 d4z(n)/(2 rre)2,
z,(O) =y, z~(s)----A. The same representation is used for S(Y, ~) and we also note,
that the path-ordered exponent actually does not depend on s, therefore in (36) one
can separate out the Wilson loop factors [45]
=
ss: dsdse -'~2(s+s)</[d4z][d4z]exp (s: s:.))
- ~ ~ 2 d a - ¼ z-2d6 .W (40)
where W(x, Y, y, f~) contains a Wilson loop W(C) passing through the points
x, Y, y, y and relativistic corrections in the form of Wilson loops with insertions of
~,,A, (insertions of this kind have been considered in [46])
where 'Pc(x, x') is a path-ordered integral along the contour C as in (4) but with A_~
in the adjoint representation. Here f,~ is made of au and to lowest order in g(F)
the correlator (f(x)f(y)) can be expressed through the gluon propagator in the
Yu.A. Simonov / Vacuum backgroundfields 525
external field
D~,~(x, y) = ( x l ( - b 2 8 ~ + 2 gF~T
. a ) X[y), (44)
where
For D~ one can use the proper-time representation (38) and path-integral form of
(39). As a result one obtains an expression similar to (40) with m 2 = 0 and the
Wilson loop operator W(x, x', y, y') in the adjoint representation. In this way the
problem reduces to the calculation of the Wilson operator in the background field
A~, which has been done in [24] with the help of cluster expansion (12). The
contribution of bilocal correlators can be recast in the form
1 {-
(W(C)) = --trexp g2E fdo..(utfdoo.(u'l((F .(ulFo (U')l), (45)
Uc p<#
o<Tp
One should keep in mind that C and S(C) depend on the paths over which the path
integral is taken in (40), (42). For large time T and separation R between the quark
and antiquark the dominant contribution to the path integral is from classical
trajectories with fixed initial and final points, while quantum corrections contribute
to the pre-exponential factor. Therefore one obtains an effective action of the QCD
string, proportional to the surface, but only at large distances, Ax, a r >> d. When
both Ax, Ar are small ( << d), the interaction in (45) is nonlocal and not equivalent
to a simple potential interaction [47,48]. When the time T in (42) is large as
compared to the correlation length d, but the separation R is small, R << d, we are
in the situation considered in [41,49], and the correlator D ( z - z', r - r ' ) enters
effectively at 5 = U = 0, and integrated over ( r - r ' ) . This contributes a static
correction (34) to the heavy quark-antiquark potential. From this point of view the
potential picture of heavy quarks and even more so the string picture of light quarks
can be considered as a large-scale approximation of the genuine picture, which is
strongly nonlocal at small scale, and is given by the effective action in the exponent
of (45).
526 Yu.A. Simonov / Vacuum background fields
where subscripts are suppressed for simplicity. If the multilocal correlator in (47)
falls off with a correlation length d', and the average field strength is B, then the
expansion in the effective action in terms of multilocal correlators is in the
parameter Bd '2, and one can keep only the bilocal term (45) if Bd '2 << 1. We
assume that this condition is fulfilled; from the qualitative point of view higher
cumulants (correlators) also yield linear confinement and therefore hardly bring
drastic changes of our results even if the series in cumulants converges slowly, e.g. at
Bd '2 -- 1 [50].
We now turn to the one-particle Green functions, which can be written in the
gauge invariant way as in (3) or, for a quark as
4. Summary
As one can see from sects. 2 and 3 there are four main properties of the described
confinement mechanism.
(i) The confining force appears between colored constituents in a white system.
This happens already in the first approximation, where particles do not interact with
each other and only interact with the background field. How then the confining
force, i.e. a strong interaction, can emerge? The answer is that by requiring the total
system to be white, we impose strong correlation on the phases of charged particles.
We actually introduce what can be called a "prestring" or an "Urstring" when we
multiply a physical state ~(x)~(y') with a phase factor (3) ~c(x, y) to make it
gauge invariant. It is exactly from this element that the actual string appears after
averaging in the background field.
Yu.A. Simonov / Vacuum background fields 527
and for large distances I x - 2[ and for a gauge-invariant (white) system one
generalizes (49) to the form
The form (50) is for the gauge-invariant system qq and can be obtained from the
Wilson loop in (42) noting that f0xA.~c~a(T)d'r= fcA~dz~ = f s F 1 4 d 0 1 4 . For a gauge-
noninvariant system, e.g. for the qq system one obtains instead of (50) the sum of
two independent terms qlA4(xl)+ q2A4(x2). This illustrates point (i). Now the
528 Yu.A. Simonov / Vacuum background fields
average of (50) over field distributions is zero and one must consider the quadratic
contribution to the action, which due to the cluster expansion reads
ASerf=q2ford1"ford'c'fXdzffdz'((F14(z,'r)Ft4(z','r')>). (51)
Eqs. (51) and (52) illustrate point (iv). Indeed ASeff is an averaged square of the
sum (50) of statistically independent deviations and it is proportional to the number
of steps, ] x -Y[/d. Therefore the linear confinement is linear because the mecha-
nism is that of the random walk problem, and distance measures the total number
of steps.
Finally, why magnetic monopoles? Here, in the 2-dimensional problem they do
not appear, because we can arrange a statistically independent, and homogeneous in
the plane distribution of the field, El4 without monopoles. Now let us add two more
dimensions (y, z). If F14 does not depend on y, z, the Bianchi identities still are
satisfied, but the field distribution in the 4-space is not 0(4) symmetric, and the
general form of the correlator is more complicated than (20).
And now if we insist on having a completely symmetric distribution of fields, then
F14 should depend on all four coordinates (and other components F,~ appear). At
this point the Bianchi identities are violated and magnetic monopole currents
appear at the boundaries of the domains to match a change of F14 as a function of y
and z.
This simple example enables one to understand the crucial role of magnetic
monopole currents in creating a homogeneous stochastic distribution.
That "disorder through monopoles" is part of a general phenomenon, common to
both statistical physics and field theory, where disorder transitions are caused by
condensation of topological objects [53].
5. Conclusion
and bilocal correlators have the form assumed by us. The formalism proposed
should have a connection to the dual long-distance lagrangian [54] since both
fundamentally seem to be based on the same physics. On a more phenomenological
level there are many directions where the approach can be developed further: heavy
and light quarkonia, hadron structure etc. These topics will be discussed elsewhere.
References
[1] M. Gell-Mann, R.J. Oakes and B. Renner, Phys. Rev. 175 (1968) 2195
[2] M. Shifman, A. Vainshtein and V. Zakharov, Nucl. Phys. B147 (1979) 385, 448
[3] V.A. Novikov, M.A. Shifman, A.I. Vainshtein and V.I. Zakharov, Nucl. Phys. B191 (1981) 301
[4] C.G. Callan, R. Dashen and D.J. Gross, Phys. Rev. D17 (1978) 2717, ibid. D19 (1978) 1826
[5] E.V. Shuryak, Nuel. Phys. B203 (1982) 93,116, 160, ibid. B214 (1983) 237
[6] D.I. Dyakonov and V.Yu. Petrov, Nuel. Phys. B245 (1984) 259
[7] D.I. Dyakonov and V.Yu. Petrov, Nucl. Phys. B272 (1986) 457
[8] C. Callan, R. Dashen and D. Gross, Phys. Lett. B66 (1977) 375
[9] M.I. Polikarpov and A.I. Veselov, JETP Lett. 45 (1987) 113
[10] S.G. Matinyan and G.K. Savvidy, Nuel. Phys. B134 (1978) 539
[11] H. Leutwyler, Nuel. Phys. B179 (1981) 129
[12] A.L. Mil'shtein and Yu.F. Pinelis, Phys. Lett. B137 (1984) 235
[13] Yu.A. Simonov, Yad. Fiz. 46 (1987) 317
[14] E.V. Shuryak, Phys. Lett. B153 (1985) 162; B193 (1987) 319
[15] G. 't Hooft, in High energy physics, ed. A. Zichichi (Editrice Compositori, Bologna, 1976);
S. Mandelstam, Phys. Reports 23C (1976) 245
[16] Yu.A. Simonov, Yad. Fiz. 42 (1985) 557
[17] V.A. Novikov, M.A. Shifman, A.I. Vainshtein and V.I. Zakharov, Nucl. Phys. B174 (1980) 378;
V.A. Novikov, M.A. Shifman, A.I. Vainshtein and V.I. Zakharov, Fortschr. Phys. 32 (1984) 585
[18] B.V. Geshkenbein and B.U Ioffe, Nucl. Phys. B166 (1980) 340
[19] Yu.A. Simonov, ITEP preprint N68 (1987), unpublished
[20] Yu.A. Simonov, ITEP preprint N99 (1987), Yad. Fiz., in press
[21] Yu.A. Simonov, ITEP preprint N188 (1987), Yad. Fiz., in press
[22] H.G. Dosch, Phys. Lett. B190 (1987) 177
[23] Yu.A. Simonov, ITEP preprint N211 (1987), Yad. Fiz., in press;
N.G. van Kampen, Physica 74 (1974) 239
[24] H.G. Dosch and Yu.A. Simonov, Phys. Lett. B, submitted
[25] C. Bernard, Phys. Lett. B108 (1982) 431; Nucl. Phys. B219 (1983) 341
[26] J. Ambjorn, P. Olesen and V. Peterson, Nucl. Phys. B240 (1984) 189
[27] W. Bardeen and V. Zakharov, Phys. Lett. B91 (1980) 111
[28] J. Greensite and M.B. Halpern, Phys. Rev. D27 (1983) 2545
[29] Y. Banks, R. Meyerson and J. Kogut, Nucl. Phys. B129 (1977) 493
[30] J.L. Cardy, Nucl. Phys. B170 (1980) 369
[31] T. De Grand and D. Toussaint, Phys. Rev. D22 (1980) 2478, ibid. D24 (1981) 466
[32] G. Bhanot, Phys. Rev. D24 (1981) 461
530 Yu.A. Simonov / Vacuum background fields
[33] J.S. Barber et al., Phys. Lett. B147 (1984) 330; B152 (1985) 221;
J.S. Barber et al., Nuel. Phys. B257 (1985) 515
[34] G. 't Hooft, Nucl. Phys. B190 (1981) 455; Phys. Scripta 25 (1982) 133;
G. Mack, Fortschr, Phys. 29 (1981) 135
[35] A.S. Kronfeld, G. Schierholz and U.-J. Wiese, Nucl. Phys. B193 (1987) 461;
A.S. Kronfeld, M.L. Laursen, G. Schierholz and U.-J. Wiese, DESY preprint 87-073 (1987)
[36] B.S. De Witt, Phys. Rev. 162 (1967) 1195, 1239;
J. Honerkamp, Nucl. Phys. B36 (1971) 130, B48 (1972) 269;
G. 't Hooft, Nucl. Phys. B62 (1973) 444
[37] N.G. Van Kampen, Phys. Rep. C24 (1976) 171
[38] P.A.M. Dirac, Phys. Rev. 74 (1948) 817
[39] N.A. Campbell, J.H. Jorycz and C. Michael, Phys. Lett. B167 (1986) 91
[40] P.H. Damgaard, Phys. Lett. B194 (1987) 107; B183 (1987) 81
[41] I.I. Balitsky, Nucl. Phys. B254 (1985) 166
[42] S. Mandelstam, Phys. Rev. D19 (1979) 2391
[43] J. Sehwinger, Phys. Rev. 82 (1951) 664
[44] M.B. Halpern and P. Senjanovic, Phys. Rev. D15 (1977) 1655
[45] S. Samuel, Nucl. Phys. B149 (1979) 517;
J. Ishida and A. Hosoya, Progr. Theor. Phys. 62 (1979) 544;
G.S. Iroshnikov, Zh. Eksp. Teor. Fiz. 90 (1987) 1922
[46] E. Eichten and F. Feinberg, Phys. Rev. D23 (1981) 2724
[47] M.B. Voloshin, Nucl. Phys. B154 (1979) 365
[48] H. Leutwyler, Phys. Lett. B98 (1981) 447
[49] U. Marquard and H.G. Dosch, Phys. Rev. D35 (1987) 2238
[50] H.G. Dosch and Yu.A. Simonov, to be published
[51] C. De Tar, J.E. King, S.P. Li and L. McLerran, Nucl. Phys. B249 (85) 621,644;
M. Campostrini, A. Di Giacomo and G. Mussardo, Z. Phys. C25 (1984) 173
[52] J.E. Mandula and M. Ogilvie, Phys. Lett. 185B (1987) 127
[53] A.M. Polyakov, Nucl. Phys. B120 (1977) 429; Phys. Lett. B59 (1975) 80, 82;
Yu.A. Simonov, Lectures at the 22 Leningrad Winter School, ed. LINP, 1985
[54] M. Baker, J.S. Ball and F. Zachariasen, CALT-68-1416 (1987) and refs. therein