a

r
X
i
v
:
h
e
p
-
p
h
/
9
7
0
1
2
9
4
v
2


3

F
e
b

1
9
9
7
UCSD/PTH 9701
hep-ph/9701294
The HQET/NRQCD Lagrangian to order /m
3
Aneesh V. Manohar
Department of Physics, University of California at San Diego,
9500 Gilman Drive, La Jolla, CA 92093-0319
(January 1997)
Abstract
The HQET/NRQCD Lagrangian is computed to order /m
3
. The computa-
tion is performed using dimensional regularization to regulate the ultraviolet
and infrared divergences. The results are consistent with reparametrization
invariance to order 1/m
3
. Some subtleties in the matching conditions for
NRQCD are discussed.
Typeset using REVT
E
X
1
I. INTRODUCTION
Heavy quark eective theory (HQET) [1] and non-relativistic QCD (NRQCD) [2,3] are
two eective theories that describe the interactions of almost on-shell heavy quarks. HQET
describes the interactions of quarks of mass m in which the momentum transfer p is much
smaller than m. The HQET Lagrangian has an expansion in powers of p/m. HQET is
typically applied to hadrons containing a single heavy quark, such as the B meson, in which
p
QCD
, the scale of the strong interactions. The HQET expansion is thus an expansion
in powers of
QCD
/m. NRQCD describes the interactions of non-relativistic quarks, and is
typically applied to

QQ bound states such as the . The NRQCD Lagrangian also has an
expansion in powers of 1/m. The momentum transfer in NRQCD is of order mv, so that the
small expansion parameter in NRQCD is the velocity v. The size of a term in the NRQCD
Lagrangian can be estimated using velocity counting rules [3]. The basic dierence between
HQET and NRQCD can be seen from the rst two terms in the eective Lagrangian,
L = Q

_
iD
0
_
Q +

Q
D
2
2m
Q. (1)
In HQET, the rst term is of order
QCD
, and the second term is of order
2
QCD
/m, whereas
in NRQCD both terms are of order mv
2
. As a result, the quark propagator in HQET is
i/(k
0
+ i), and in NRQCD it is
i
(k
0
k
2
/2m + i)
. (2)
The HQET/NRQCD Lagrangian is computed in this paper to one loop and order 1/m
3
.
Only the terms bilinear in fermions are considered here. There are also four-quark operators
in the eective Lagrangian. Their coecients are order
s
, and can be obtained simply from
tree-level matching.
II. MATCHING CONDITIONS AND POWER COUNTING
The HQET eective theory matching computation is a straightforward generalization
of known results to order 1/m
2
[48]. One can compute diagrams in the full and eective
2
theories, and match to a given order in 1/m. Since the HQET propagator is m independent,
the HQET power counting is manifest one counts powers of 1/m directly from the vertex
factors. This means that graphs with a vertex of order 1/m
r
do not make any contributions
to terms of order 1/m
s
, with s < r to any order in the loop expansion.
The use of NRQCD as an eective eld theory is more subtle. NRQCD with the propaga-
tor Eq. (2) cannot be used as an eective Lagrangian to compute matching corrections, since
the velocity power counting breaks down.
1
The matching conditions for NRQCD should be
computed using the HQET power counting, by expanding in p

/m. After the HQET La-

grangian has been computed, it can be used for computing bound state properties using
the NRQCD velocity power counting rules. In other words, the NRQCD propagator Eq. (2)
should be thought of as the innite series
1
(k
0
k
2
/2m + i)
=
1
k
0
+
k
2
2m(k
0
)
2
+ . . . (3)
where one uses the right hand side inside any ultraviolet divergent Feynman graph. This is
necessary when a cuto such as dimensional regularization is used to regulate the Feynman
graphs. NRQED matching conditions have previously been computed using a momentum
space cuto [9]. In this case, there is no dierence between using the left or right hand sides
of Eq. (3). However, a momentum space cuto can not be used in NRQCD, since it breaks
gauge invariance.
The dierence between using the two forms of Eq. (3) in a loop graph can be illustrated
by a simple example. Consider the integral
_

0
dk
2
(k
2
)
a
(k
2
+ m
2
1
) (k
2
+ m
2
2
)
=

sin a
(m
2
1
)
a
(m
2
2
)
a
m
2
1
m
2
2
. (4)
A typical NRQCD loop integral has the form Eq. (4). The power a increases as one considers
more and more divergent loop graphs in the eective theory. The denominator of a typical
1
I would like to thank M. Luke for extensive discussions on this point. Some of these issues will
be discussed in a future publication. See also [13].
3
NRQCD loop graph has poles at k
2
p
2
, where p is the external momentum, and at k
2
m
2
where m is the quark mass. Thus the scales m
1
and m
2
in Eq. (4) can be taken to be of
order p and m, respectively. One can see immediately that the NRQCD power counting
breaks down. Loop graphs with insertions of higher dimension operators are divergent, and
can be proportional to positive powers of m because of the (m
2
2
)
a
term in Eq. (4). The
positive powers of m from the loop integral can compensate for inverse powers of m in the
coecient, and the entire eective Lagrangian expansion breaks down. Now consider the
same integral, but rst expand
1
k
2
+ m
2
2
=
1
k
2

m
2
2
k
4
+ . . . ,
evaluate the integral, and then resum the series. The answer is
_

0
dk
2
(k
2
)
a
(k
2
+ m
2
1
) (k
2
+ m
2
2
)
=

sin a
(m
2
1
)
a
m
2
1
m
2
2
. (5)
The integral is missing the (m
2
2
)
a
term since it is non-analytic at the origin for a = integer,
which is where the integral is evaluated in 4 dimensions. Equation (5) only has inverse
powers of the high momentum scale m
2
m, and leads to an acceptable eective eld
theory. Thus NRQCD and HQET matching conditions are computed in the same way, and
the two Lagrangians are the same.
III. THE LAGRANGIAN
The continuum NRQED eective Lagrangian at one-loop has previously been com-
puted [9] using a photon mass to regulated the infrared divergences, and a momentum
space cuto. This procedure cannot be used in a non-abelian gauge theory such as QCD.
The kinetic terms in the NRQCD Lagrangian at one-loop have previously been computed
by Morningstar [10] using a lattice regulator. The computations in this letter will be done in
the continuum using dimensional regularization for the infrared and ultraviolet divergences,
and for on-shell external states. This has the advantage that one can freely use the equa-
4
tions of motion to reduce the number of operators in the eective Lagrangian [11]. The most
general eective Lagrangian to order 1/m
3
(up to eld redenitions) is
L = Q

_
iD
0
+ c
2
D
2
2m
+ c
4
D
4
8m
3
+ c
F
g
B
2m
+ c
D
g
[D E]
8m
2
+ic
S
g
(DE E D)
8m
2
+ c
W1
g
{D
2
, B}
8m
3
c
W2
g
D
i
BD
i
4m
3
(6)
+c
p

p
g
DB D+D B D
8m
3
+ ic
M
g
D [DB] + [DB] D
8m
3
+c
A1
g
2
B
2
E
2
8m
3
c
A2
g
2
E
2
16m
3
+ c
A3
g
2
Tr
_
B
2
E
2
8m
3
_
c
A4
g
2
Tr
_
E
2
16m
3
_
(7)
+ic
B1
g
2
(BBE E)
8m
3
ic
B2
g
2
(E E)
8m
3
_
Q,
which is the HQET/NRQCD Lagrangian in the special frame v = (1, 0, 0, 0), and the no-
tation of [9] has been used. The covariant derivative is D

+ igA
a
T
a
= (D
0
, D).
Covariant derivatives in square brackets act only on the elds within the brackets. The
other covariant derivatives act on all elds to the right. The subscripts F, S and D stand
for Fermi, spin-orbit, and Darwin, respectively. The last seven terms in Eq. (6) are not given
in Ref. [9], since they were not required for the computation done there. The last four terms
can be omitted for QED.
In an arbitrary frame, Eq. (6) can be written as
L
v
=

Q
v
_
iD v c
2
D
2

2m
+ c
4
D
4

8m
3
c
F
g

4m
c
D
g
v

_
D

_
8m
2
+ic
S
g
v

_
D

, G

_
8m
2
+ c
W1
g
_
D
2

_
16m
3
c
W2
g
D

8m
3
(8)
+c
p

p
g

_
D

+ D

_
8m
3
ic
M
g
D

_
D

_
+
_
D

_
D

8m
3
+c
A1
g
2
G

16m
3
+ c
A2
g
2
G

16m
3
+ c
A3
g
2
Tr
_
G

16m
3
_
+ c
A4
g
2
Tr
_
G

16m
3
_
ic
B1
g
2

_
G

, G

_
16m
3
ic
B2
g
2

_
G

, G

_
v

16m
3
_
Q
v
,
where
D

= D

v D. (9)
5
The tree level matching conditions can be obtained by integrating out the antiquark com-
ponents, and making a eld redenition to eliminate terms with v D acting on the quark
elds. The standard form of the HQET Lagrangian after integrating out the antiquark
elds is
L
v
=

Q
v
_
iv D + i / D

1
2m + iv D
i / D

_
Q
v
(10)
=

Q
v
_
iv D
1
2m
/ D

/ D

+
1
4m
2
/ D

(iv D) / D

1
8m
3
/ D

(iv D)
2
/ D

_
Q
v
The eld redenition
Q
v

_
1
D
2

8m
2

g

16m
2
+
D

(iv D) D

16m
3
+
gv

16m
3
(11)
i

(iv D) D

16m
3
i
gv

16m
3
_
Q
v
(where the matrices are understood to be P
v
P
v
) can be used to eliminate the time
derivative terms, and put the Lagrangian into the NRQCD form. The result is Eq. (8)
with c
2
= c
4
= c
F
= c
D
= c
S
= c
W1
= c
A1
= c
B1
= 1, and c
W2
= c
p

p
= c
M
= c
A2
= c
A3
=
c
A4
= c
B2
= 0. The c
A
and c
B
terms are quadratic in the eld strengths, and are order g
2
.
The one loop corrections to these terms will not be computed here.
IV. QUARK FORM FACTORS AND MATCHING CONDITIONS
A loop diagram in QCD is a function F({p}, m, , ) where {p} are the external momenta,
m is the quark mass, is the scale parameter of dimensional regularization,and the computa-
tion is done in d = 4 dimensions. As an example, consider the diagram Fig. 1, which gives
radiative corrections to the form factors F
1
(q
2
) and F
2
(q
2
). In dimensional regularization,
the diagram gives the F
1
and F
2
form factors as functions of the form F
1,2
(q
2
/m
2
, /m, ).
The form factor can be expanded as a power series in q
2
/m
2
at xed , followed by the limit
0,
F
1
= F
1
(0)
_
A
0

UV
+
B
0

IR
+ (A
0
+ B
0
) log

m
+ D
0
_
(12)
+q
2
dF
1
d q
2
(0)
_
A
1

UV
+
B
1

IR
+ (A
1
+ B
1
) log

m
+ D
1
_
+ . . . ,
6
FIG. 1. Non-abelian contribution to the one-loop vertex correction.
and similarly for the F
2
form factor. It is conventional to label as either
UV
or
IR
,
depending on whether the integral is ultraviolet or infrared divergent. Ultraviolet divergences
are cancelled by renormalization counterterms. Infrared divergences cancel when a physically
measurable process is computed. Expanding the form factor in q
2
and then taking the limit
0 gives an expression that is analytic in q
2
, and misses terms which are non-analytic in q
2
.
The non-analytic terms are not needed for the calculation of the coecients in the eective
theory, since the eective Lagrangian is analytic in momentum. The coecients of the
eective Lagrangian are determined by computing (for example) the dierence of F
1
in the
full theory and eective theories. The non-analytic terms in F
1
cancel in the dierence, and
the analytic terms determine the unknown parameters c
F
. . . c
B2
in the eective Lagrangian.
Loop diagrams in HQET are functions F({p}, , ) times powers of the coecients c
i
in the eective Lagrangian, where {p} are the external momenta. All on-shell loop graphs
vanish when expanded in powers of p, followed by 0. This is because the coecient of
any power of p is a dimensionally regulated integral of the form
_
d
d
k
(2)
d
f
_
k
2
, k v
_
. (13)
There is no dimensionful parameter in the integrand, so the integral vanishes. The matching
condition is then trivial: one takes Eq. (12), and throws away the 1/ terms to obtain the
dierence of the graph in the full and eective theory. All the 1/ terms in the dierence are
ultraviolet divergences (which are cancelled by renormalization counterterms), since there
are no infrared divergences in matching conditions. To see this more explicitly, one can
evaluate integrals such as Eq. (13) by breaking them up into the sum of two integrals, one
7
only ultraviolet divergent, and the other only infrared divergent. For example,
_
d
d
k
(2)
d
1
k
4
=
_
d
d
k
(2)
d
_
1
k
2
(k
2
+ m
2
)
+
m
2
k
4
(k
2
+ m
2
)
_
=
1
16
2
_
1

UV

IR
_
= 0, (14)
since
UV
=
IR
= . A given quantity in the eective theory is of the form
A
e
_
1

UV

IR
_
. (15)
There can be no nite parts (the analogs of the (A + B) log /m and D terms in Eq. (12)),
since the net integral is zero. A typical matching condition is of the form
graphs in full theory = graphs in eective theory + c
i
, (16)
where c
i
is a coecient in the eective Lagrangian. Using Eq. (12) and Eq. (15), the
matching condition can be written as
B

IR
+ (A + B) log

m
+ D = A
e
1

IR
+ c
i
. (17)
The 1/
UV
terms are cancelled by the renormalization counterterms in the full and eective
theories, respectively. The coecients in the eective Lagrangian have no infrared diver-
gences. Thus B = A
e
, and
c
i
= (A + B) log

m
+ D, (18)
i.e. c
i
is obtained from Eq. (12) by keeping the nite pieces, and omitting the 1/
UV
and
1/
IR
terms.
The coecients of the Q

(D
2
/2m) Q and Q

(D
4
/8m
3
) Q are xed by the dispersion
relation E
2
= p
2
+ m
2
of QCD, c
2
= c
4
= 1. The other terms, which all contain at least
one power of the gauge eld A

, are obtained by computing the one-loop Q

QA on-shell
scattering amplitude. The wave-function renormalization graph Fig. 2 and the vertex cor-
rection Fig. 3 can be found in many textbooks on quantum eld theory [12]. In dimensional
regularization, one nds that the wave function graph is
8
FIG. 2. One-loop wavefunction renormalization.
FIG. 3. Abelian one-loop vertex correction.
i(p) = iC
2
(Q)

s
4
_
A
_
p
2
_
m + B
_
p
2
_
p /
_
(19)
A
_
p
2
_
=
_
1
0
dx (/2) (4 )
_
m
2
x p
2
x(1 x)
_
/2
(20)
B
_
p
2
_
=
_
1
0
dx (/2) (2 ) (1 x)
_
m
2
x p
2
x(1 x)
_
/2
(21)
where
C
2
(Q) = T
a
T
a
=
4
3
is the Casimir of the quark representation. The on-shell wavefunction renormalization cor-
rection is
Z = C
2
(Q)

s
4
_
_
B
_
m
2
_
+ 2m
2
_
A
p
2
+
B
p
2
_
p
2
=m
2
_
_
= C
2
(Q)

s

_
1
2
UV
+
1

IR
+ 1
3
2
log
m

_
. (22)
The on-shell vertex correction Fig. 3 can be expressed in terms of the form-factors F
1
and F
2
,
igT
a
u (p

)
_
F
1
_
q
2
_

+ iF
2
_
q
2
_

2m
_
u (p) , (23)
where q = p

p. The form factors are

F
(V )
1
_
q
2
_
=

2
_
C
2
(Q)
1
2
C
2
(ad)
_
_
1

UV
+
1

IR
_
2
q
2
m
2
_
I
_
q
2
/m
2
_
+
_
3
q
2
m
2
_
I
_
q
2
/m
2
_

1
2
J
_
q
2
/m
2
_

_
1
q
2
2m
2
_
K
_
q
2
/m
2
_
1
_
(24)
9
and
F
(V )
2
_
q
2
_
=

2
_
C
2
(Q)
1
2
C
2
(ad)
_
I
_
q
2
/m
2
_
. (25)
where
I
_
q
2
/m
2
_
=
_
1
0
dx
m
2
m
2
q
2
x(1 x)
(26)
J
_
q
2
/m
2
_
=
_
1
0
dx log
m
2
q
2
x(1 x)

2
(27)
K
_
q
2
/m
2
_
=
_
1
0
dx
m
2
m
2
q
2
x(1 x)
log
m
2
q
2
x(1 x)

2
, (28)
and
C
2
(ad) = 3
is the Casimir of the adjoint representation. Expanding to order q
2
/m
2
gives
F
(V )
1
=

_
C
2
(Q)
1
2
C
2
(ad)
_
_
1
2
UV
+
1

IR
+ 1
3
2
log
m

+
q
2
m
2
_

1
3
1

IR

1
8
+
1
3
log
m

__
, (29)
F
(V )
2
=

_
C
2
(Q)
1
2
C
2
(ad)
_
_
1
2
+
q
2
12m
2
_
. (30)
The nal diagram is the non-abelian vertex correction Fig. 1. This is computed in
background eld Feynman gauge, which preserves gauge invariance. The resulting diagram
can also be evaluated in terms of the F
1
and F
2
form-factors,
F
(g)
1
=

s
8
C
2
(ad)
_
1
0
dx
_
1x
0
dy
_
(1 + /2)
_
2q
2
(x + y)
+2m
2
(1 x y) (2 (x + y) + (4 ) (1 x y))
_ _
m
2
(x + y 1)
2
q
2
xy
_
1/2
+(2 ) (/2)
_
m
2
(x + y 1)
2
q
2
xy
_
/2
_
=

s
8
C
2
(ad)
_
2

UV
+
4

IR
+ 4 6 log
m

+
q
2
m
2
_

IR
1 + 3 log
m

_
+ . . .
_
, (31)
F
(g)
2
=

s
4
C
2
(ad) m
2
(1 + /2)
_
1
0
dx
_
1x
0
dy (1 x y)
( + (2 ) (x + y))
_
m
2
(x + y 1)
2
q
2
xy
_
1/2
=

s
8
C
2
(ad)
_
4

IR
+ 6 4 log
m

+
q
2
m
2
_
4

IR
+ 1 4 log
m

_
+ . . .
_
. (32)
10
The total on-shell form factors at one loop are given by
F
1
= 1 Z + F
(V )
1
+ F
(g)
1
(33)
= 1 +

s

q
2
m
2
__

1
3
IR

1
8
+
1
3
log
m

_
C
2
(Q) +
_

5
24
IR

1
16
+
5
24
log
m

_
C
2
(ad)
_
,
F
2
= F
(V )
2
+ F
(g)
2
(34)
=

s

_
1
2
C
2
(Q) +
_
1
2
IR
+
1
2

1
2
log
m

_
C
2
(Q)
_
+

q
2
m
2
_
1
12
C
2
(Q) +
_
1
2
IR
+
1
12

1
2
log
m

_
C
2
(ad)
_
.
The total form-factor F
1
(0) is unity, since gauge invariance is preserved by the background
eld method.
The scattering amplitude for a low-momentum heavy quark o a background vector po-
tential can be computed by expanding Eq. (23), and multiplying by
_
m/E for the incoming
and outgoing quarks. If p is the three-momentum of the incoming quark, p

is the three-
momentum of the outgoing quark, and q = p

p, one nds that the eective interaction

is
igT
a
u

NR
(p

)
_
A
0a
j
0
A
a
j
_
u
NR
(p) , (35)
where
j
0
= F
1
_
q
2
_
_
1
1
8m
2
|q|
2
+
i
4m
2
(p

p)
_
+ F
2
_
q
2
_
_

1
4m
2
|q|
2
+
i
2m
2
(p

p)
_
,
(36)
and
j = F
1
_
q
2
_
_
1
2m
(p +p

) +
i
2m
q
i
8m
3
_
|p|
2
+|p

|
2
_
q (37)

i
16m
3
_
|p

|
2
|p|
2
_
(p +p

)
1
8m
3
_
|p

|
2
+|p|
2
_
(p +p

)
1
16m
3
_
|p

|
2
|p|
2
_
q
_
+F
2
_
q
2
_
_
i
2m
q
i
16m
3
|q|
2
q
1
16m
3
|q|
2
(p

+p)

1
16m
3
_
|p

|
2
|p|
2
_
q
i
8m
3
_
|p

|
2
|p|
2
_
(p

+p) +
i
8m
3
(p

+ p) (p

p)
_
.
11
Comparing Eqs. (36,37) with the scattering amplitude in the eective theory from the
Lagrangian Eq. (6) gives
c
F
= F
1
+ F
2
(38)
c
D
= F
1
+ 2F
2
+ 8F

1
(39)
c
S
= F
1
+ 2F
2
(40)
c
W1
= F
1
+
1
2
F
2
+ 4F

1
+ 4F

2
(41)
c
W2
=
1
2
F
2
+ 4F

1
+ 4F

2
(42)
c
p

p
= F
2
(43)
c
M
=
1
2
F
2
4F

1
(44)
where
F
i
F
i
(0) , F

i

dF
i
d (q
2
/m
2
)

q
2
=0
(45)
Note that the nine parameters (including c
2
and c
4
) in the eective Lagrangian Eq. (6) are
determined in terms of only three independent constants, F
2
, F

1
, and F

2
, since F
1
= 1.
Reparametrization invariance [14] gives six linear relations among the coecients. This will
be discussed in more detail in the next section.
The explicit expressions for the coecients are obtained using Eqs. (33,34):
c
F
= 1 +

_
1
2
C
2
(Q) +
_
1
2

1
2
log
m

_
C
2
(ad)
_
(46)
c
D
= 1 +

__
8
3
log
m

_
C
2
(Q) +
_
1
2
+
2
3
log
m

_
C
2
(ad)
_
(47)
c
S
= 1 +

_
C
2
(Q) +
_
1 log
m

_
C
2
(ad)
_
(48)
c
W1
= 1 +

__
1
12
+
4
3
log
m

_
C
2
(Q) +
_
1
3

17
12
log
m

_
C
2
(ad)
_
(49)
c
W2
=

__
1
12
+
4
3
log
m

_
C
2
(Q) +
_
1
3

17
12
log
m

_
C
2
(ad)
_
(50)
c
p

p
=

_
1
2
C
2
(Q) +
_
1
2

1
2
log
m

_
C
2
(ad)
_
(51)
c
M
=

__
1
4

4
3
log
m

_
C
2
(Q)
_
7
12
log
m

_
C
2
(ad)
_
(52)
12
FIG. 4. Quark contribution to the vacuum polarization.
FIG. 5. Gluon contribution to the vacuum polarization.
The results for NRQED can be obtained by setting C
2
(ad) = 0 and C
2
(Q) = 1, and agree
with those found in [9], with the replacement
log log 2
5
6
(53)
The dierence in nite parts is because the NRQED integrals were evaluated in Ref. [9]
using a momentum space cuto, instead of using dimensional regularization. The results for
the 1/m operators agree with known results for HQET [4,5]. The 1/m
2
matching conditions
at tree-level, and the dependence at one-loop also agree with known results [68]. Note
that c
F
is independent of in QED. This is easy to see if one computes the renormalization
of the magnetic moment operator in the eective theory in Coulomb gauge, in which all
transverse photon interactions are suppressed by 1/m.
The discussion so far has concentrated on the fermion part of the eective Lagrangian.
There is, in addition, the pure gauge eld part of the eective action. The one-loop correction
to the gluon propagator is shown in Figs. (4) and (5). The gluon diagram is the same in QCD
and in HQET, so the one-loop matching condition is from the quark vacuum polarization
diagram. This gives the eective action [1517]
L =
1
4
d
1
G
A

G
A
+
d
2
m
2
G
A

D
2
G
A
+
d
3
m
2
gf
ABC
G
A

G
B

G
C

+O
_
1
m
4
_
, (54)
with
13
d
1
= 1

3
T (Q) log m
2
/
2
,
d
2
=

60
T (Q) , (55)
d
3
=
13
360
T (Q) ,
where
T (Q) = 1/2
is the index of the quark representation. The identity
0 =
_
2 D

G
A

G
A
+ 2 g f
ABC
G
A

G
B

G
C

+ G
A

D
2
G
A
has been used to eliminate D

G
A

G
A
from Eq. (54).
V. REPARAMETRIZATION INVARIANCE
The coecients of operators in the HQET Lagrangian are constrained by reparametriza-
tion invariance [14]. The reparametrization invariant spinor eld
v
is given by

v
= (w, v)
v
, (56)
where
v
is the conventional heavy quark eld that satises
v /
v
=
v
, (57)
(w, v) is the Lorentz transformation matrix
(w, v) =
1 + w/ v /
_
2 (1 + w v)
, (58)
and
w

=
v

+ iD

/m
|v

+ iD

/m|
. (59)
One needs to choose a particular operator ordering for the covariant derivatives; dierent
orderings are related to each other by eld redenitions.
14
It is simplest to consider the consequences of reparametrization invariance when D

in Eq. (56). Then there is no operator ordering ambiguity, and the eld
v
can be written
as

v
=
_

_
1
2
_
(m + i v)
2
+ (i

)
2
_
m + i v +
_
(m + i v)
2
+ (i

)
2
_
_

_
1/2
(60)

_
m + +i v + i/

+
_
(m + i v)
2
+ (i

)
2
_

v
,
where

v. (61)
If one uses Eq. (60), replaces
v
by the spinor u
NR
e
ipx
, with p
2
= m
2
,
0
u
NR
= u
NR
,
u

NR
u
NR
= 1, and v = (1, 0, 0, 0)), the eld
v
reduces to the spinor ue
ipx
, that satis-
es the Dirac equation p /u = mu, and is normalized so that uu = 1. This shows that
reparametrization invariance will determine the coecients of the 1/m suppressed operators
which are xed by relativistic invariance.
The reparametrization invariant kinetic term is

v
i/
v
=

v
__
(m + i v)
2
+ (i

)
2
m
_

v
. (62)
This is not the same as the terms in the Lagrangian Eq. (6). The reparametrization invariant
eld Eq. (56) does not automatically produce a Lagrangian in the standard NRQCD form.
However, one can convert Eq. (62) to this form by making a eld redenition,

v
=
_
_
_
(m + i v)
2
+ (i

)
2
+ m
m + i v +
_
(i

)
2
+ m
2
_
_
1/2

v
(63)
The kinetic energy term in the primed eld is

v
_
m + i v
_
(i

)
2
+ m
2
_

v
. (64)
which when expanded gives Eq. (6) with c
2
= c
4
= 1. Thus c
2
= c
4
= 1 follows from
reparametrization invariance. The transformation factor in Eq. (63) when applied to on-
shell spinors (instead of elds) reduces to
_
m/E. This is the same as the ux factor for the
incoming and outgoing particles which was included in Eq. (35).
15
To determine the constraints of reparametrization invariance on the eective Lagrangian,
consider Eq. (60) with the gauge elds included, i.e. with D. Expanding to order 1/m
3
gives

v
=
_
1 + A +
i / D

2m
B
_

v
(65)
where
A = 1
(iD

)
2
8m
2
+
(iD

)
2
(iv D)
4m
3
B = 1
iv D
m

3 (iD

)
2
8m
2
+
(iv D)
2
m
2
. (66)
A particular ordering has been chosen for the operators in Eq. (65). A dierent ordering
gives an eective Lagrangian that is related by a eld redenition.
The most general reparametrization invariant Lagrangian is a linear combination of in-
variant terms, such as

v
_
v / + i / D/m
_

v
,

v
, etc. The eective Lagrangian ob-
tained in this way is not in the form Eq. (8), but it can be converted into that form by eld
redenitions that preserve v /
v
=
v
. One nds by a straightforward (but not very enlight-
ening computation) that the eective Lagrangian is a linear combination of the invariant
linear combinations
iv D + O
2
+ O
4
+ O
F
+ O
D
+ O
S
+ O
W1
,
2O
F
+ 4O
D
+ 4O
S
+ O
W1
+ O
W2
+ 2O
p

p
O
M
,
O
W1
+ O
W2
, (67)
2O
D
+ O
W1
+ O
W2
O
M
,
O
A1
, O
A2
, O
A3
, O
A4
, O
B1
, O
B2
,
up to terms of order 1/m
4
. Here O
F
, etc. are the operator coecients of c
F
, etc. in Eq. (8).
The above linear combinations imply the constraints
c
2
= 1,
c
4
= 1,
16
c
S
= 2c
F
1, (68)
c
W2
= c
W1
1,
c
p

p
= c
F
1,
2c
M
= c
F
c
D
.
which are satised by Eqs. (38),(46).
VI. CONCLUSIONS
The HQET/NRQCD Lagrangian has been computed to one loop and order 1/m
3
, and
has been shown to be reparametrization invariant to order /m
3
. The original form of
the NRQCD propagator Eq. (2) cannot be used to compute the eective Lagrangian by
matching to QCD. Instead, one must treat the propagator as an innite series, and resum
the series after doing the loop integral. As a result, the matching computations for NRQCD
and HQET are the same. It is straightforward to obtain the eective Lagrangian (in the
one-quark sector) to higher orders in 1/m by expanding the form factors F
1,2
and the spinors
in the computation of Eqs. (35),(36) to higher orders. No further Feynman graphs need to
be evaluated.
ACKNOWLEDGMENTS
I am indebted to M. Luke for numerous discussions about matching conditions in non-
relativistic eective eld theories. This work was supported in part by a Department of
Energy grant DOE-FG03-90ER40546,
17
REFERENCES
[1] N. Isgur and M.B. Wise, Phys. Lett. B232 113 (1989); Phys. Lett. B237 527 (1990).
[2] W.E. Caswell and G.P. Lepage, Phys. Lett. B167, 437 (1986).
[3] G.T. Bodwin, E. Braaten, and G.P. Lepage, Phys. Rev. D51, 1125 (1995).
[4] E. Eichten and B. Hill, Phys. Lett. B243, 427 (1990).
[5] A.F. Falk, B. Grinstein, and M.E. Luke, Nucl. Phys. B357, 185 (1991).
[6] C. Balzereit and T. Ohl, Phys. Lett. B386, 335 (1996).
[7] C. Bauer, Diplom Thesis, Karlsruhe 1996 (unpublished).
[8] B. Blok, J.G. Korner, D. Pirjol, and J.C. Rojas, hep-ph/9607233.
[9] T. Kinoshita and M. Nio, Phys. Rev. D53, 4909 (1996).
[10] C. Morningstar, Phys. Rev. D48, 2265 (1993), D50, 5902 (1994).
[11] H.D. Politzer, Nucl. Phys. B172, 349 (1980).
[12] See for example, S. Pokorski, Gauge Field Theories, Cambridge University Press, 1987.
[13] M.E. Luke and A.V. Manohar, hep-ph/9610534.
[14] M.E. Luke and A.V. Manohar, Phys. Lett. B286, 348 (1992).
[15] P. Cho and E.H. Simmons, Phys. Rev. D51, 2360 (1995).
[16] V. Novikov, M. Shifman, A. Vainshtein, and V. Zakharov, Fort. Phys. 32, 585 (1984).
[17] S. Balk, J.G. Korner, and D. Pirjol, Nucl. Phys. B428, 499 (1994).
18