Perturbative Yang-Mills Ground State in the Temporal Gauge

By Scott Chapman
Research Professor of Physics
Chapman University
ABSTRACT
=====
Second order corrections to the perturbative ground state wave functional and vacuum energy of a Yang-Mills theory
are calculated in the temporal gauge. Using dimensional regularization, the concepts of renormalization and a running
coupling constant are motivated in a spatially gauge invariant way, before the introduction of a scalar product. After
definition of a scalar product, it is shown that the renormalization constants of this approach are just those of the Axial
gauge.
=====
In recent years there has been a renewed interest in studying the perturbative ground state of a Yang-Mills
theory in the Hamiltonian formulation. For example, second order corrections to the ground state wave functional were
recently calculated in the Coulomb gauge [1]. One driver of this renewed interest was an ansatz for the Yang-Mills
vacuum wave functional that reproduces the essential infrared physics of the theory [2-5].
The purpose of this paper is to derive the second order correction to the Yang-Mills perturbative ground state
wave functional in the temporal gauge. In addition to being free of Gribov ambiguities, temporal gauge wave
functionals are interesting since they are spatially gauge invariant and can be derived without the introduction of a
scalar product. Results from calculations made in other gauges can be recovered from temporal gauge results by
defining a scalar product and imposing a spatial gauge condition. In this way, results of this paper are tied back to those
found in the Coulomb and Axial gauges.
Two interesting features of the perturbative ground state arise from this derivation. First, the required equation
for local spatial gauge invariance directly leads to the Ward Identity requirement that two-field terms of the wave
functional must only be transverse at any order of perturbation theory. Second, one notices that the concept of a
running coupling constant with the famous 11/3 N beta function dependence arises naturally in a spatially gauge
invariant way, even before the introduction of a scalar product.
Dimensional regularization is employed to evaluate divergent loop integrals embedded in the two-field terms
and energy density. The Appendix includes useful formulas for dimensionally regularized divergent integrals in three
dimensions. In addition to being used for the present work, the formulas can also be employed to reproduce integrals
from previously published Coulomb gauge papers.
The Hamiltonian for a Yang-Mills theory is given by:

H = d 3 x 12 ia ia + 14 Fija Fija
where Fija = i A aj j Aia + gf

abc

(1)

Aib Acj and ia ( x ) is the canonical momentum of the gauge field Aia ( x ) . Due to

a
working in the temporal gauge ( A0 = 0 ), it is assumed that all fields are evaluated at some constant time t. Since all

fields are always evaluated at that same time, the time dependence of fields is suppressed, and only spatial dependence
is shown explicitly. Calculations are performed in the Schroedinger representation, which is defined by

ia (x ) = i Aia (x ) . Spatial gauge transformations are generated by the color Gauss Law operator:

G a ( x ) = i ia ( x ) gf abc Aic (x ) bi (x ) .

(2)

This operator commutes with the Hamiltonian and must annihilate any physical state in the temporal gauge (in order for
the state to be spatially gauge invariant).

To any order g , one should be able to derive perturbative vacuum wave functionals that simultaneously
satisfy both of the following equations:
n

( )
( )

Hn = En n + O g n+1

G a (x )n = 0 + O g n+1

(3)
(4)

These equations mean that to order g , the wave functional is an eigenstate of the Hamiltonian and locally spatially
gauge invariant. For the remainder of this paper, reference to gauge invariance will mean local spatial gauge invariance
(excluding large, topologically nontrival spatial gauge transformations). Since the above equations do not involve
scalar products, no scalar product is needed when deriving the wave functional.
It is helpful to expand the gauge fields into complex Fourier modes
n

() (

d 3k ~ a
(5)
exp
Aia ( x ) =
A
k
i
k
x .
(2 )3 2 i

~ ~
Since the gauge fields Aia ( x ) are real functions, the Fourier modes must satisfy Aia* k = Aia k . It is convenient to

()

( )

separate the Fourier modes into longitudinal and transverse components:

()

()

()

()

()

~ ~ ~
ATia k = Aia k ALia k ,
(6)
a
From equation (2), it is apparent that the first term of G ( x ) is only nonzero if it acts on longitudinal field components.
~
~
ALia k = ki k j A ja k

It is assumed that one can write the ground state wave functional in the following form:

n = exp O0 + gO1 + g 2O2 + ... + g n On ,

(7)

where each of the operators Om is a polynomial functional involving the gauge fields. By applying equation (4) to this
expansion, one obtains the iterative equation:

[
i

a
i

(x ), Om+1 ] = [ f abc Aic (x ) bi (x ), Om ] .

(8)

This equation implies that for semi-simple Lie algebras, any 2-field terms in any of the Om must be transverse; no
longitudinal 2-field terms are allowed. This can be seen as follows: For semi-simple Lie algebras where f

aab

= 0 , any

term involving a single gauge field must have a gauge (e.g. color) index. Since the Om do not have any gauge indices,
none of the Om polynomials can include a term with a single gauge field; all terms of each polynomial must have at least
2 gauge fields. Let us assume that in a 2-field term of some functional Om+1 , at least one of the two gauge fields is
longitudinal. If that were the case, then the left side of (8) would not vanish. The only way to satisfy the equation would
be if Om had a term with just one field, but this is not possible, as argued above. So any 2-field terms in any of the
functionals Om must be comprised only of transverse fields. This applies to all orders of the perturbative expansion.
The zeroth order perturbative ground state and vacuum energy of a Yang-Mills theory are well known. They are
given by

() ( )

~ ~
O0 = 12 d 3 kkATia k ATia k

(9)

d 3 xd 3 k
k,
(2 )3

(10)

E0 = n A

where k = k (a notation that will be used throughout) and n A is the number of adjoint gauge directions in the theory
(e.g. n A = N 1 for SU(N)).
2

From equation (8), the first order functional O1 must involve 3 fields, at least one of which is transverse. The
exact form of O1 can be found by solving the part of the eigenequation (3) in which each term is proportional to g (see
equation (A1) in the Appendix). One finds that

O1 = if abc

()

d 3 kd 3 pd 3 p 3
~a

+
+
k
p
p
k
A
i Tj k
(2 )3 2
~ ~
~ ~
~ ~
~ ~
ATib ( p)ATjc ( p ) ATib ( p)ALjc ( p ) ALib ( p)ATjc ( p ) ALib ( p)ALjc ( p )

+
+
+

k
+
p
+
p
k
+
p
k
+
p
k

E1 = E0

(11)
(12)

satisfies both equations (3) and (8) to first order. As a check, the transverse-only part of O1 matches the corresponding
functional found when expanding the Yang-Mills ground state in the Coulomb gauge [1].
(4 )

The second order functional O2 includes two terms: one involving four gauge fields O2

and one involving two

(2 )

gauge fields O2 .

O2 = O2(2 ) + O2(4 )

(13)

Again, by using equation (8) along with the terms in the eigenequation proportional to g , it is straightforward to solve
2

(4 )

for both O2 and E 2 . The explicit form of O2 is somewhat complicated and is given in the Appendix since it is not
directly relevant to the main discussion of the rest of this paper. As mentioned above, since gauge invariance requires
(2 )

(2 )

any two-field term such as O2 to be comprised of only transverse fields, its final form is simpler. However, O2
involves divergent integrals that need to be regulated. Dimensional regularization is employed for that purpose, and
calculations are performed in d = 3 2 dimensions. In practice, this means that momentum integrals are

[ ]

transformed from d k d k , having mass dimensionality given by d k = 3 2 . In order to ensure that the
argument of the exponent in (7) remains dimensionless, the dimensionalities of the bare gauge fields and bare coupling
3

[~ ( )]

must become Aia k = 2 + and [g ] = . One may define a dimensionless bare coupling g 0 ( ) through the

relation g 0 = g , where is the renormalization scale.

Solving equations (3) and (8) at second order, one finds

() ( )

g O2

1
k 2 ~a ~a

= d kk + c2 ln 2 ATi k ATi k ,

(14)

E2 =

1
k 2
n A (d 1) d d xd d k

c
1
ln

(2 )d 2 2 ,
2

(15)

(2 )

1
2

11C A g 02
,
2
3(4 )

(16)

where C A is the adjoint Casimir operator for the group (e.g. C A = N for SU(N)), and c2 is a finite numerical constant
that is unimportant for the present work. Note that the famous beta factor associated with the running coupling
constant of a Yang-Mills theory appears in this approach even though a scalar product has not yet been defined.
As an aside, it is worth mentioning that the types of integrals that arise in 3 2 dimensions are quite different
than the types that arise in the more common 4 2 dimensional approaches in covariant gauges. A useful identity for
d = 3 2 dimensional integrals is presented in (A13) of the Appendix. In addition to informing this work, this identity
can also be used as an alternative way to calculate integrals presented in [1].
It is interesting to see how far one can go toward motivating a running coupling before defining a scalar product.
If one defines the following constant,

Z A1 = 1 + c2 ,

(2 )
then it is possible to combine O0 and O2 as follows

(17)

() ( )

k2 ~ ~
O0 + g 2O2(2 ) = Z A1O0 12 d d kk ln 2 ATia k ATia k ,

(2 )

so that the divergent 1 part of O2

(18)
1

has been absorbed into the constant Z A . Since O0 and O1 satisfy equation

1
(8), then Z A O0 and Z A O1 must also satisfy equation (8). This tells us that all of the divergent 1 parts of the third

(3 )
order, three-field operator O3 that have at least one longitudinal field must be incorporated into Z A O1 . If one
1

wanted to absorb all of these divergent terms into renormalized fields and a renormalized coupling, this could be
achieved by making the definitions:

()

()

~
AiRa k = Z A1 2 Aia k

(19)

1 2

g R = Z g 0

(20)

Z = 1 + c2

(21)

with

Equations (20) and (21) motivate the running coupling constant in a spatially gauge invariant way and without the use of
a scalar product.
It should be emphasized that without a scalar product, the running coupling is only motivated, not formally
derived. To formally derive the running coupling, the renormalization procedure must be defined such that the scalar
products associated with any physical observables are finite and not divergent. To make connection with this standard
renormalization procedure, the scalar product will now be defined.
In the temporal gauge, the scalar product
condition

( , ) is defined through the introduction of a canonical gauge

( A) = 0 and its associated Faddeev Popov functional determinant det[G a , b ].

For example the norm of

2 can be defined by:

(2 , 2 ) = DAiRa (x ) [ a ( A)]det[G a , b ]2*2 ,

(22)

Ra
where DAi is the functional integral over renormalized gauge fields at all points of space. So, to evaluate a scalar

a
a
product, one must pick a spatial gauge condition. The Axial gauge = A3 ( x ) = 0 has the advantage that the Faddeev

Popov (FP) determinant is independent of the gauge fields, so it decouples from the integral. In the Axial gauge with no
contributions from FP ghosts, the renormalization procedure presented above in equations (17) to (21) is valid. When
expressed in terms of renormalized quantities, there will be no 1 terms in 2 , so scalar products like that of (22) will
have no divergent 1 dependence. It is no surprise then that expressions for Z A and Z of equations (17) and (21)
are exactly those found in standard Axial gauge renormalization approaches [6] (except that in the present approach,
the finite constant c2 has been included in the renormalization constant for convenience). This agreement provides a
good check of the results of (14), (15) and (16).

APPENDIX
A key assumption for this paper is that the functionals Om in the expansion (7) can be expressed as polynomials
in the gauge fields. With that assumption in hand, the eigenequation (3) can be disaggregated to an infinite number of
separate equations, each involving some power of g and some definite number of gauge fields. For example, to
n

derive O1 of equation (11), one needs only to solve the part of (3) involving a single factor of g and three gauge fields.
This equation is

() ()

~
kkATja k ~ a O1 + f abc d 3 x i Aaj (x )Aib (x )Acj (x ) = 0 .
(A1)
ATj k
~a
Due to its ATj k factor, the first term above has at least one transverse field. If the second term above had a part

()

comprised of only longitudinal fields, then it would not be possible to satisfy (A1), and one might have to question the

a
assumption that the Om are polynomials. However, the i A j ( x ) factor in the second term means that it has at least

one transverse field, so (A1) can be satisfied, and it is straightforward to see that (11) does the job. For O1 to be correct,
it must also satisfy (8) to first order, and it is straightforward to verify that it does.

= 0 , the one-field parts of equation (3) vanish at all orders. The zeroFor semi-simple Lie algebras where f
field first order part of equation 3 is just equation (12). One can generalize equation (12) as follows. Equation (8)
aab

implies that the functionals Om will only have terms involving an even number of gauge fields when m is even and an
odd number of gauge fields when m is odd. It follows that the only nonvanishing parts of equation (3) for a given order

g m will be equations with an even number of gauge fields when m is even and an odd number of gauge fields when m is
odd. In particular, this implies in general that

E2 n+1 E2 n = 0

(A2)

for any non-negative integer n.

() ()

~a

3
3
(4 ) 1
k
k
A
d
~ a O2 2 d k ~ a O1 ~ a O1 +
Tj k

ATj k
A j k
A j k

()

( )

1
4

f abc f ade d 3 x Aib (x )Acj (x )Aid (x )Aej (x ) = 0

(A3)

()

~a

If the second two terms included a part comprised solely of longitudinal fields, then due to the ATj k factor in the first
term, it would not be possible to solve (A3), and one might again question the polynomial assumption. Fortunately, the
longitundinal-only parts of the second two terms of (A3) exactly cancel, so with O1 in hand, it is possible to find an
(4 )

(4 )

expression for O2 that satisfies (A3). It should be noted that to also satisfy equation (8), O2
longitudinal-only term that is not specified by (A3), only by (8).

(4 )

must include a

An explicit form of O2 satisfying (A3) and (8) is given below, where the notations k = p q ,

()

~
~
ALb (q ) = qi ALib (q ) and PijT k = ij ki k j are employed.

O2(4 ) = f abc f ade

d 3 qd 3 pd 3 q d 3 p

(2 )

3 ( p + q + q + p )[O200 + O210 + O220 + O211 + O221 + O222 ]

(A4)

1 ~b ~c ~d ~e
ALi (q )ALj ( p )ALi (q )ALj ( p)
4k
k
p ~ ~ ~ ~
O210 = i i ALib (q )ALjc ( p )ALd (q )ATje ( p)
kq qp

O200 =

(A5)
(A6)

] ()

12 ( pk qk ) ij 2 kj pi PklT k k m q m p l 12 qi p j ~ b ~ c ~ d ~ e
O =
+
ATi (q )ATj ( p )AL (q )AL ( p)
k ( p + q )(k + q + p )
p + q

2
2
1 ( pp + qk qk + k ) ij (qj qi q j qi ) (k ij + q j qi ) ~ b ~ c ~ d ~ e
11
O2 =

AL (q )ATj ( p )AL (q )ATi ( p)

12 ( pk qk ) ij PkmT k mj pi + mi q j 12 (q p ) ij km ~ b ~ c ~ d ~ e
21
O2 =
+
ATi (q )ATj ( p )AL (q )ATm ( p)
k+q+ p
k

20
2

()

(A7)

(A8)

(A9)

1
(qi pi )(qi pi ) (q p )(q p) (q p )(q p)

4 jk lm

(q + p + q + p)
2(k + q + p )(k + q + p)
k

( pm qm )kl + ( pk qk )k j
jk
(q + p + q + p) lm (q + p + q + p)
1

O222 =

4k j k l
jl (k + q + p )(k + q + p) ~ b ~ c ~ d ~ e

+ km
+
ATj (q )ATk ( p )ATl (q )ATm ( p)

(
)
(
)
+
+
+
2
+
+
+
q
p
q
p
q
p
q
p

(A10)

The second order two-field part of equation (3) is given by

() ()

~a

d
d
(4 )
(2 ) 1
d
k
k
A
~ a O2 2 d q ~ a ~ a O2 = 0 .
Tj k

A j ( q ) A j (q )
ATj k

(A11)

Just as in similar equations discussed earlier, if the second term above contained a purely longitudinal part, then it
(4 )

(2 )

would not be possible to solve the equation. Fortunately, it does not. So with O2 in hand, one can calculate O2 . As
mentioned in the main body of the paper, the calculation is divergent, and dimensional regularization is employed with

d = 3 2 . Taking q as the loop momentum and p as the momentum of the remaining fields, the second term of

(A11) depends on q and q + p in combinations like q + p + q and q + p q . These are different combinations
than those that arise in 4-dimensional calculations, but they can be evaluated using the identities given below.
Define the following quantities:

K = q+ p + q

Q= q+ p q .

and

(A12)

For small , the following identity holds

ddq
2
2
f (K , Q, p ) = 2 (1 + )p dK K p
q q+ p
4p

dQ p 2 Q 2

) f (K , Q, p ) + O( ) . (A13)

This identity can be derived in the following way: First, on the left side insert a factor of one in the form of integrals over
delta functions:

1 = 4 dk 0 k 0 dq 0 q 0
0

dwdv

(2 )

[(

)]

exp i k 02 (qi + pi ) w + i q 02 qi2 v

(A14)

Next, replace factors of q and q + p with q 0 and k 0 , and let v v w . Then, using known integral expressions for

( )

( )

cos ax 2 and sin ax 2 , perform the integrals over d d q and dw. Next, perform the dv integral by using known

expressions for the integrals of v cos(av ) and v sin (av ) . Rewrite the remaining integrals over q 0 and k 0 to be
over k 0 + q 0 and k 0 q 0 , which are equivalent to K and Q by the delta functions. After this, one obtains the right hand
side of equation (A13).
In (A13), it is apparent that any functions f (K , Q, p ) that are odd in Q will vanish in the above integral. The Q
integrals are convergent due to their finite limits. The K integrals, however can be divergent. In particular, for
logarithmically divergent integrals, the following K integral is useful:

2 2 (1 + )

dK
K 2 p2
(K + p )

p2
= ln 2 ln (4 ) + 1 + O( )

(A15)

For linearly and quadratically divergent integrals, the following identities are useful

2 2 (1 + ) dK K 2 p 2
p

2 2 (1 + ) dKK K 2 p 2
p

= 2 p + O( )

(A16)

= 0 + O( ) ,

(A17)

where the usual approach is employed of first taking d to be small enough for the integrals to be convergent, then
analytically continuing d to the value d = 3 2 at the end of the calculation. In addition to determining the form of

O2(2 ) in (14), these identities can also be used to reproduce the integrals I c and I g found in [1].
The second order no-field part of equation (3) is given by

12 d d k ~ a ~ a g 2O2(2 ) = E2 E0
A j k A j k

( ) ()

The form of the second order vacuum energy shown in (15) follows directly from this equation.

REFERENCES
[1] D. Campagnari, H. Reinhardt and A. Weber, Phys. Rev. D80, 25005 (2009), arXiv:0904.3490 [hep-th].
[2] A. Szczepaniak and E. Swanson, Phys. Rev. D65, 025012 (2001).
[3] C. Feuchter and H. Reinhardt, Phys. Rev. D70, 105021 (2004), arXiv:hep-th/0408236 .
[4] C. Feuchter and H. Reinhardt, (2004), arXiv:hep-th/0402106 .
[5] D. Epple, H. Reinhardt, and W. Schleifenbaum, Phys. Rev. D75, 045011 (2007), arXiv:hep-th/0612241 .
[6] H. C. Lee, Chin. Jour. Phys., Vol. 23, No. 2, 90 (1985),
http://psroc.org/cjp/download.php?type=paper&vol=23&num=2&page=90 , (see eq. 6.36).

(A18)