Asymptotic Limits and Sum Rules For Gauge Field Propagators

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

EFI 93-71

arXiv:hep-th/9406081v1 14 Jun 1994

ASYMPTOTIC LIMITS AND SUM RULES


FOR GAUGE FIELD PROPAGATORS 1

Reinhard Oehme and Wentao Xu


Enrico Fermi Institute and Department of Physics
University of Chicago
Chicago, Illinois 60637, USA

Abstract
For gauge field propagators, the asymptotic behavior is obtained in all
directions of the complex k 2 -plane, and for general, linear, covariant gauges.
Asymptotically free theories are considered. Except for coefficients, the functional form of the leading asymptotic terms is gauge-independent. Exponents
are determined exactly by one-loop expressions. Sum rules are derived, which
generalize the superconvergence relations obtained in the Landau gauge.

Work supported in part by the National Science Foundation, Grant PHY 91-23780

Interesting sum rules for the structure functions of propagators can be


derived on the basis of their analytic properties, together with the asymptotic
behavior for large momenta as obtained with the help of the renormalization
group. For systems with a limited number of matter fields, one obtains superconvergence relations for the gauge field propagator in the Landau gauge
[1, 2]. These relations are of interest in connection with the problem of confinement [4, 3]. Other results are dipole representations, and information
about the discontinuity of the gauge field structure functions. They indicate
the existence of an approximately linear quarkantiquark potential [6, 5],
and are important for understanding the structure of the theory in the state
space with indefinite metric [7].
It is the purpose of this note to present results for the gauge field propagator in general, covariant, linear gauges. We obtain the asymptotic terms
for large momenta, and for all directions in the complex k 2 plane. Sum rules
are derived, which generalize the superconvergence relations of the Landau
gauge. An important aspect of our results is the gaugeindependence of the
functional form of the essential asymptotic terms. Only the coefficients of
these terms depend upon the gauge parameter.
We consider a nonAbelian gauge theory like QCD, with the gaugefixing
part of the Lagrangian given by B ( A ) + 2 B B, where B is the usual
auxiliary field. For 6= 0, B can be eliminated by B = ( A ). In order
to define the structure function of the transverse gauge field propagator, we
write
Z

dxeikx h0|T A
a (x)Ab (0)|0i =

iab D(k 2 + i0)

(k k g k k g + k k g k k g )

(1)

with A A A . We assume the general postulates of covariant


gauge theories. Important are Lorentz covariance and simple spectral conditions, as formulated in references like [8, 9] for state spaces with indefinite
metric. Exact Greens functions should be connected with the formal perturbation series in the coupling parameter g for g 2 +0, at least as far as
the first few terms are concerned. Topological aspects of the gauge theory
are not expected to influence the asymptotic behavior we consider here.
As a consequence of Lorentz covariance, and the spectral conditions mentioned above, it follows that the function D(k 2 + i0) is the boundary value of
1

an analytic function, which is regular in the cut k 2 plane, with a branch line
along the positive real axis. In contrast to the situation for higher Greens
functions [10] , explicite use of local commutativity is not required for the
two-point functions [1]. Using renormalization group methods, together with
analyticity, we obtain the asymptotic behavior for k 2 in all directions
of the complex plane. We present first the essential leading terms, leaving
derivation and details for later.
For the analytic structure function D(k 2 ), we find for k 2 in all
directions:
!

00
0

k2
k D(k , , g, )
+ CR (g 2 , ) 0 ln 2
+
0

!1
k2

+
+ C1 0 ln 2
0

(2)

The corresponding asymptotic terms for the discontinuity along the positive,
real k 2 axis are then given by
!

/ 1

00
0
00
k2
2
2
2
2
k (k , , g, )
CR (g , ) 0 ln 2
+
0
| |
!2
k2

+
+ C1 0 ln 2
0
| |

(3)

The parameters and their limitations are as follows: The function D is normalized at the real point k 2 = 2 < 0, where
k 2 D(k 2 , 2 , g, )|k2=2 = 1.

(4)

The anomalous dimension of the gauge field is given by (g 2, ) = 0 ()g 2 +


1 ()g 4 + , 0 () = 00 + 01 , 1 () = 10 + 11 + 2 12 , etc., and
0 00 /01 . The renormalization group function is (g 2 ) = 0 g 4 +
1 g 6 + . For QCD, we have 00 = (16 2 )1 ( 13
23 NF ), 01 = (16 2 )1 23 ,
2
2
2 1
0 = (16 ) (11 3 NF ), NF = number of flavors. We assume 0 < 0
corresponding to asymptotic freedom. Consequently, the exponent 00 /0 in
eqs. (2) and (3) varies from 13/22 for NF = 0 to 1/10 for NF = 9, and from
1/16 for NF = 10 to 15/2 for NF = 16. We have 0 < 00 /0 < 1 for
NF 9 and 00 /0 < 0 for 10 NF 16; for 00 /0 = 1, our relations (2)
and (3) would require modifications.
2

The essential asymptotic term in eqs.(2) and (3) is the one with the
coefficient CR (g 2 , ), which is not identically zero, although there may be
zero surfaces = 0 (g 2). In case CR should vanish, the C1 term becomes
relevant. Its coefficient is given by
00
1 (0 )
for 0 <
< 1,
0 00
0
00 + 0 11
00
=
for
< 0.
0 00
0

C1 =
C1

(5)

For = 0, the coefficient CR is positive and given by


2

2 00 /0

CR (g , 0) = (g )
0 (x)

exp

g2

dx0 (x),

00
(x, 0)

.
(x)
0 x

(6)

Eq. (6) follows from the exact solution for = 0, with the normalization
2 D(2 , 2 , g, 0) = 1 [1, 2]. This solution can be written in the form
2

g2
g2

k D(k , , g, 0) =

!00 /0

exp

g2

g2

dx0 (x).

(7)

Here g2 = g2 (u, g), u = | k2 | is the effective gauge coupling, with the properties
ln u =

g2 (u,g)

g2

dx 1 (x),

g2(u, g) 1/(0 ln u) +

(8)

for u , in the case of asymptotic freedom with 0 < 0.


In general, CR (g 2, ) satisfies a partial differential equation. In an approximation, where only terms linear in are kept in the expression for the
anomalous dimension (g 2 , ), we obtain
2

2 00 /0

CR (g , ) = (g )

exp

Z

g2

dx0 (x)

2 00 /0 Z 0

dxx00 /0 f (x, g 2 ) .
+ (g )
1
2
0 0
g

(9)

Here
f (x, g 2) = {0 (x) + 0 1 (x)}exp

g2

dy0 (y),

1 (x) = 1 (x)/(x) 01 /0 x,
(x, ) = 0 (x) +
1 (x) + O(2).

(10)

In eq. (9), the normalization for the full solution in the the linear approximation has been used in order to fix an otherwise undetermined coefficient
of CR . This approximation consists of replacing the general anomalous dimension (g 2 , ) by the -linear form 0 (g 2) +
1 (g 2). The corresponding
solution is given by
!

g2

0 (x)
k D(k , , g, ) = exp
dx
2
(x)
g
!)
(
Z x
Z g2
0(y)
1 (x)
, (11)
exp
dy

1+
dx
(x)
(y)
g2
g2
2

with the notation as defined in eqs. (8) and (10).


An important aspect of the leading asymptotic terms for

and k 2 (k 2 , 2 , g, )
k 2 D(k 2 , 2 , g, )
0
is their independence of the gauge parameter , except for the coefficient
CR (g 2 , ). In addition, the exponents of the logarithms in eqs. (2) and
(3) are completely and exactly determined by oneloop coefficients of the
anomalous dimension (g 2 , 0) and of the function.
In view of the asymptotic behavior of D(k 2 , 2 , g, ) as given in eq. (2),
which is valid for all directions in the k 2 plane, we can write the usual unsubstracted dispersion representation
2

D(k , , g, ) =

dk

, 2 , g, )
k 2 k 2

2 (k

(12)

We also have sufficient boundedness for the discontinuity in order to write


a dipole representation
2

D(k , , g, ) =

dk 2

(k 2 , 2 , g, )
,
(k 2 k 2 )2

(k , , g, ) =

k2

dk 2 (k 2 , 2 , g, ).

(13)

For = 0, the dipole representation has been discussed in refs. [5] and [6]
in connection with an approximately linear quarkantiquark potential.
Of particular interest is the situation for 0 < 00 /0 < 1, corresponding
to NF 9 in QCD. There, the function D + 0 k 2 vanishes faster than k 2
for k 2 , and hence we have the sum rule

dk 2 (k 2 , 2 , g, ) =

.
0

(14)

This is the generalization of the superconvergence relation [1, 2]


Z

dk 2 (k 2 , 2 , g, 0) = 0,

which was obtained previously in the Landau gauge. The relation (14) expresses the fact that the coefficient of the asymptotic term proportional to k 2
in the representation (12) is given by /0 . It is not valid for 00 /0 < 0.
The distribution aspects of sum rules like eq. (14), and of the related dispersion representations, have been discussed in refs. [1, 2].
In order to derive the asymptotic properties of the structure function
D, we consider the renormalization group equation for the dimensionless
function R(k 2 /2 , g, ) k 2 D(k 2 , 2 , g, ). We obtain
k2
R 2 , g,

2
k2
= R
,
g,

R
, g,
,
2
2
!

(15)

where we have used the relation


R1 (2 /2 , g, ) = Z3 (2 /2 , g, ),
which follows from the normalization condition (4): R(1, g, ) = 1 at k 2 =
2 < 0 , and where Z3 is the square of the conventional renormalization
2
factor for the gauge field. Further, in eq. (15), g = g( 2 , g) is the
effective


1 2
,
g,

.
(running) gauge coupling parameter, and
is given by
= R
2

We use a mass independent renormalization scheme [11], which is appropriate


for the study of asymptotic limits.
Let us first consider the limit k 2 along the negative, real k 2 axis,
where R(k 2 /2 , g, ) is analytic and real. We set u = |k 2 /2 | and define the
function R(
g 2; g 2 , ) R(u, g, ), with g(u, g) being the effective coupling
defined in eq. (8). From eq. (15), we then obtain the differential equation
(
g2)

R(
g 2 ; g 2 , )
= (
g2,
)R(
g 2 ; g 2, ),

g2

=
(
g 2 ; g 2 , ) R1 (
g 2; g 2 , ).

(16)

For 6= 0, it is more convenient to work with the equation


(
g 2)

=
(
g2,
),

g2

(17)

with
as defined in eq. (16). From the general solution (7) of eq. (16) for
= 0, as well as the solution (11) for 6= 0 in the linear approximation,
we know that R and
have a branch point as a function of g2 at g2 = 0,
13
32 NF
2
.
which is of the form (
g 2 ) , 00 /0 . For QCD, we have = 11
2
N
3 F
We see that || = n/d is rational, with n and d being relative primes. It
is then convenient to uniformize the algebraic branch point by introducing
x = (
g 2)1/d as a uniformization variable.
We consider first the case 0 < < 1, corresponding to NF 9 for QCD.
Here it is convenient to use eq. (16). We define y(x) = (
0 )xn , and
obtain the differential equation
dy
n
= xn1 y 2 d xdn1 (0 + xn y)(xd, 0 + xn y),
dx
0

(18)

where
(g 2 , ) 0 ()

= 0 () + g 2 1 () +
(g 2)
0 g 2
1 () 1
2 0 (), etc.,
0 () =
0
0

(g 2, )

with the definitions given below eq. (4).


6

(19)

In an appropriate finite domain including g 2 = 0, and excluding possible, nontrivial fixed points corresponding to zeroes of (g 2), it is reasonable
to assume that (g 2, ) is continuously differentiable. As far as (g 2) and
(g 2 , ) are represented by power series expansions for g 2 +0, (g 2, ) is
also a power series in g 2 and . Under these circumstances, the r.h.s. of
eq. (18) satisfies the Lipschitz condition for x = 0, and we have exactly one
solution through every point x = 0, y = C. In as far as the r.h.s. of eq. (18)
is also a power series, we obtain the solution in the form of a series:
C2 n
+1
y(x) = C +
x +C
0 (0 ) 0 0 (0 ) xd +
0
1
0
0 (0 )xdn + ,
+
1
!

(20)

where we have separated the terms proportional to C.


For the asymptotic expansion of R(
g 2 ; g 2, ) for g2 +0, eq. (20) implies
!

+ CR (
g 2 ) + CR 01 10 12 02 00 (
g 2)+1 +
R(
g ; g , )
0
0
1 (0 ) 1 2
+
g + ,
(21)
0 0 1
2

with CR = C/02 , and 0 < < 1. This formula is also valid for CR = 0.
The term proportional to C 2 in eq.(20) cancels in the inversion leading to
eq.(21).
For < 0, corresponding to 10 NF 16 for QCD, it is more convenient
to use eq. (16). With = nd , n and d being positive integers which are
relative primes, we uniformize the branch point at g2 = 0 by introducing
again a new variable x so that g2 = xd . Then we define z(x) = xn R(xd ; g 2, ),
and obtain the differential equation
dz
xn
= 0 nxn1 + d xd1 z(xd ,
)
dx
z

(22)

As long as xn z 1 remains bounded around x = 0, we have again the Lipschitz


condition satisfied and obtain the power series solution
z(x) = CR + CR 0 (0)xd +
n
d
+
x +
(0 (0) + 0 0 (0))xd+n + ,
0
0 d + n
7

(23)

with CR 6= 0. In terms of R(
g 2 ; g 2, ), eq. (23) leads to the asymptotic
expansion for < 0 and g2 +0:
1
1
(10 00 )(
g 2)+1 +
0
0
(10 + 0 11 ) 2

+
g +
+
0 0 0 (1 )

R(
g 2 ; g 2, ) CR (
g 2 ) + C R

(24)

For CR = 0, we cannot use eq. (22), but obtain the asymptotic expression
directly from eqs. (20) and (21) :
R(
g 2 ; g 2, )

1 (0 ) 1

+
g2 + .
0 0 0 1 + ||

(25)

This relation corresponds to eq. (21) with CR = 0 and < 0.


With eqs. (21) and (24), we have obtained the asymptotic expressions
2
for R( k2 , g, ) in the limit k 2 along the real axis, provided we can
use g2 (u, g) 1/(0 ln u) + . It remains to consider the limit k 2
in all directions of the complex k 2 plane. We return to eq. (15). Setting
2 = |k 2 |, we find, with 2 < 0 and k 2 = |k 2 |ei for all || :
k2
R 2 , g, = R


!
k2

, ).

2 , g, R(ei , g

(26)

Here g = g(| k 2 |, g) and


= R1 (| k 2 |, g, ). For 0 < 0, the effective
2
remains bounded in this limit,
coupling g2 vanishes for | k2 | , and
as may be seen from eqs. (21) and (24). Because R is analytic in the
cut complex k 2 plane, we can then use the perturbation expansion for the
structure function,
k2
k2
R 2 , g, 1 + g 2 0 ()ln 2

+ O(g 4),

(27)

and write for g2 +0:


R(ei , g, )
1 + g2 0 ()i

+ O(
g 4).

(28)

Eq. (26) expresses the asymptotic limit for k 2 in all directions in terms
of the limit along the negative real k 2 axis. With eqs. (26), (28), (21) and
(24), we finally obtain the limits given in eqs. (2) and (3).
8

A priori, the coefficients C or CR appearing in the solutions of the nonlinear, ordinary differential equations are undetermined constants. However,
because of the normalization condition R(g 2 ; g 2, ) = 1 or (g
2 ; g 2, ) = ,
2
the coefficients become functions of g and , satisfying partial differential
equations in these variables. For CR (g 2 , ), we find the equation
CR (g 2 , ) = R(g 2 ; g 2, )CR (g 2 , )
= R1 (g 2 ; g 2, ),

(29)

and the corresponding differential equation is:


(g 2 )

CR
CR
= (g 2, )
(g 2, )CR .
2
g

(30)

For = 0, and in the linear approximation, we have given CR in eqs.


(6) and (9), which satisfy eq. (30) with = 0 and (g 2 , ) = 0 (g 2 )+
1 (g 2 )
respectively.
Several of the results presented in this paper have been obtained by one
of us (R.O.) in collaboration with W. Zimmermann [12]. It is a pleasure to
thank Wolfhart Zimmermann for his contribution and for many discussions.

References
[1] R. Oehme and W. Zimmermann, Phys. Rev. D21 (1980) 475, 1661.
[2] R. Oehme, Phys. Lett. B252 (1990) 641.
[3] K. Nishijima, in Symmetry in Nature, Festschrift for Luigi A. Radiacati
di Brozolo (Scuola Normale Superiore, Pisa, 1989) pp. 627655.
[4] R. Oehme, Phys. Rev. D42 (1990) 4209.
[5] R. Oehme, Phys. Lett. B232 (1989) 498.
[6] K. Nishijima, Prog. Theor. Phys. 77 (1987) 1035.
9

[7] R. Oehme, Mod. Phys. Lett. A8 (1993) 1533;


N - Newsletter No. 7 (1992) 1.
[8] T. Kugo and I. Ojima, Prog. Theor. Phys. Suppl. No. 66 (1978) 1;
N. Nakanishi and I. Ojima, Covariant Operator Formalism of Gauge
Theories and Quantum Gravity (World Scientific, Singapore, 1990) p.
382.
[9] F. Strocchi, Comm. Math. Phys. 56 (1977) 57;
Phys. Rev. D17 (1978) 2010.
[10] H.J. Bremermann, R. Oehme and J.G. Taylor,
Phys. Rev. 109 (1958) 2178.
[11] S. Weinberg, Phys. Rev. D8 (1973) 3497;
G. tHooft, Nucl. Phys. B61 (1973) 455;
J.C. Collins and A.J. MacFarlane, Phys. Rev. D10 (1974) 1201.
[12] R. Oehme and W. Zimmermann, (1983) unpublished.

10

You might also like