Volume 201
Managing Editor: G. Hohler, Karlsruhe
Editors: J. Kuhn, Karlsruhe
Th. Muller, Karlsruhe
A. Ruckenstein, New Jersey
F. Steiner, Ulm
J. Trumper, Garching
P. Wole, Karlsruhe
Starting with Volume 165, Springer Tracts in Modern Physics is part of the [SpringerLink] service.
For all customers with standing orders for Springer Tracts in Modern Physics we offer the full
text in electronic form via [SpringerLink] free of charge. Please contact your librarian who can
receive a password for free access to the full articles by registration at:
springerlink.com
If you do not have a standing order you can nevertheless browse online through the table of
contents of the volumes and the abstracts of each article and perform a full text search.
There you will also f ind more information about the series.
3
Berlin
Heidelberg
New York
Hong Kong
London
Milan
Paris
Tokyo
Managing Editor
Gerhard Hohler
Andrei Ruckenstein
Editor for The Americas
Johann H. Kuhn
Peter Wole
Thomas Muller
Institut fur Experimentelle Kernphysik
Fakultat fur Physik
Universitat Karlsruhe
Postfach 69 80
76128 Karlsruhe, Germany
Phone: +49 (7 21) 6 08 35 24
Fax: +49 (7 21) 6 07 26 21
Email: thomas.muller@physik.unikarlsruhe.de
wwwekp.physik.unikarlsruhe.de
Andrey Grozin
Heavy Quark
Effective Theory
With 72 Figures
13
Andrey Grozin
Budker Institute of Nuclear Physics
630090 Novosibirsk, Russia
Email: A.G.Grozin@inp.nsk.su
SPIN: 10791166
56/3141/jl
543210
Preface
Heavy quark physics is one of the most rapidly progressing areas of high
energy physics, both experimentally and theoretically. Experiments at the Bfactories at SLAC and KEK are producing a lot of information about decays
of Bmesons, with high statistics and low systematic errors. Therefore, a better theoretical understanding of the properties of Bmesons is very important.
An understanding of stronginteraction eects in weak decays is necessary
for extracting fundamental electroweak parameters (such as elements of the
KobayashiMaskawa mixing matrix) from experimental data on these decays.
Moreover, investigation of the Bmeson the simplest nontrivial hadron, the
QCD hydrogen atom is interesting in its own right.
In this book, we shall discuss properties of hadrons with a single heavy
quark, b or c; the t quark decays before it can form a hadron, and it is not
interesting for our purposes. In the case of hadrons containing a c quark,
QCD /mc corrections can be rather large. The applicability of the theory for
hadrons containing b is much better. The physics of mesons consisting of a
bc)
is essentially dierent, and will
heavy quark and a heavy antiquark (cc, bb,
not be discussed here. Note that Bc mesons can be, to some approximation,
described by the methods discussed in this book. However, the expansion
parameter mc /mb 1/3 is not very small, and the accuracy would be poor.
Heavy quark eective theory (HQET) is an eective eld theory constructed to reproduce the results of QCD for problems with a single heavy
quark with mass m, expanded to some order (k/m)n , where k m is the
characteristic momentum in the problem. To the leading order in 1/m, it has
symmetries which are not explicit in the original QCD Lagrangian. These
symmetries relate various matrix elements involving heavy hadrons. HQET
considerably simplies lattice simulations with heavy quarks. The considerable progress in the theory of hadrons containing a heavy quark during the
last decade is largely due to HQET.
In this book, we shall discuss the properties of HQET as a quantum eld
theory and the methods used for calculating Feynman diagrams in HQET.
Some knowledge of QCD (see, e.g., the textbook [15]) is needed for understanding the text. However, knowledge of methods of calculation of multiloop
diagrams is not assumed. We shall discuss such methods for QCD and HQET
in parallel (see [9] for more details).
VI
Preface
Andrey Grozin
References
1. I. Bigi, M.A. Shifman, N.G. Uraltsev: Annu. Rev. Nucl. Part. Sci. 47, 591
(1997)
2. R. Casalbuoni, A. Deandrea, N. Di Bartolomeo, R. Gatto, F. Feruglio, G. Nardulli: Phys. Rep. 281, 145 (1977)
References
VII
3. A.F. Falk: The Heavy Quark Expansion of QCD, in The Strong Interaction,
from Hadrons to Partons, ed. by J. Chan, L. DePorcel, L. Dixon, SLACR508 (SLAC, Stanford 1997) p. 43; The CKM Matrix and the Heavy Quark
Expansion, in Flavor Physics for the Millennium, ed. by J.L. Rosner (World
Scientic, Singapore 2001) p. 379
4. J.M. Flynn, N. Isgur: J. Phys. G 18, 1627 (1992)
5. H. Georgi: Heavy Quark Eective Field Theory, in Perspectives in the Standard Model, ed. by R.K. Ellis, C.T. Hill, J.D. Lykken (World Scientic, Singapore 1992) p. 589
6. B. Grinstein: Lectures on Heavy Quark Eective Theory, in High Energy
Phenomenology, ed. by M.A. Perez, R. Huerta (World Scientic, Singapore
1992) p. 161; Annu. Rev. Nucl. Part. Sci. 42, 101 (1992); An Introduction to
the Theory of Heavy Mesons and Baryons, in CP Violation and the Limits of
the Standard Model, ed. by J.F. Donoghue (World Scientic, Singapore 1995)
p. 307; An Introduction to Heavy Mesons, in 6 Mexican School of Particles
and Fields, ed. by J.C. DOlivo, M. Moreno, M.A. Perez (World Scientic,
Singapore 1995) p. 122; Introduction to Heavy Flavors, in Advanced School on
Quantum Chromodynamics, ed. by S. Peris, V. Vento (Universitat Aut`
onoma
de Barcelona, Barcelona 2001) p. 115
7. A.G. Grozin: Introduction to the Heavy Quark Eective Theory, Part 1.
Preprint BudkerINP 9297 (Novosibirsk 1992), hepph/9908366
8. A.G. Grozin: Using REDUCE in High Energy Physics (Cambridge University
Press, Cambridge 1997)
9. A.G. Grozin: Int. J. Mod. Phys. A, to be published; hepph/0307297
10. F. Hussain, G. Thompson: An Introduction to the Heavy Quark Eective
Theory, in Summer School in High Energy Physics and Cosmology, ed. by
E. Gava, A. Masiero, K.S. Narain, S. RandjbarDaemi, Q. Sha (World Scientic, Singapore 1995) p. 45
11. N. Isgur, M.B. Wise: Heavy Quark Symmetry. In Heavy Flavours, ed. by
A.J. Buras, M. Lindner (World Scientic, Singapore 1992) p. 234
12. T. Mannel: Chin. J. Phys. 31, 1 (1993); Heavy Quark Mass Expansion in
Quantum Chromodynamics, in QCD 20 Years Later, ed. by P.M. Zerwas,
H.A. Kastrup (World Scientic, Singapore 1993) v. 2, p. 634; J. Phys. G 21,
1007 (1995); Review of Heavy Quark Eective Theory, in Heavy Quarks at
Fixed Target, ed. by L. Kopke (INFN, Frascati 1997) p. 107; Rep. Prog. Phys.
60, 1113 (1997)
13. A.V. Manohar, M.B. Wise: Heavy Quark Physics (Cambridge University Press,
Cambridge 2000)
14. M. Neubert: Phys. Rep. 245, 259 (1994); Heavy Quark Masses, Mixing Angles,
and Spin Flavor Symmetry, in: The Building Blocks of Creation: from Microfermis to Megaparsecs, ed. by S. Raby, T. Walker (World Scientic, Singapore
1994) p. 125; Int. J. Mod. Phys. A 11, 4173 (1996); Heavy Quark Eective
Theory, in Eective Theories and Fundamental Interactions, ed. by A. Zichichi
(World Scientic, Singapore 1997) p. 98; Heavy Quark Eective Theory, in
NonPerturbative Particle Theory and Experimental Tests, ed. by M. Jamin,
O. Nachtmann, G. Domokos, S. KovesiDomokos (World Scientic, Singapore
1997) p. 39; B Decays and the Heavy Quark Expansion, in Heavy Flavours
II, ed. by A.J. Buras, M. Lindner (World Scientic, Singapore 1998) p. 239;
VIII
15.
16.
17.
18.
19.
20.
Preface
Introduction to B Physics, in ICTP Summer School in Particle Physics, ed.
G. Senjanovic, A.Yu. Smirnov (World Scientic, Singapore 2000) p. 244
M.E. Peskin, D.V. Schr
oder: Quantum Field Theory (Perseus, Reading, MA
1995)
M.A. Shifman: Lectures on Heavy Quarks in Quantum Chromodynamics, in
QCD and Beyond, ed. by D.E. Soper (World Scientic, Singapore 1996) p. 409
N.G. Uraltsev: Heavy Quark Expansion in Beauty and its Decays, in Heavy
Flavor Physics a Probe of Natures Grand Design, ed. by I. Bigi, L. Moroni
(IOS Press, Amsterdam 1998) p. 329; Topics in the Heavy Quark Expansion,
in At the Frontier of Particle Physics: Handbook of QCD, ed. M. Shifman
(World Scientic, Singapore 2001) v. 3, p. 1577
M.B. Voloshin: Surv. High Energy Phys. 8, 27 (1995)
M.B. Wise: New Symmetries of the Strong Interaction, in Particle Physics
the Factory Era, ed. by B.A. Campbell, A.N. Kamel, P. Kitching, F.C. Khanna
(World Scientic, Singapore 1991) p. 222; Heavy Quark Physics: Course,
in Probing the Standard Model of Particle Interactions, ed. by R. Gupta,
A. Morel, E. Derafael, F. David (Elsevier, Amsterdam 1999) v. 2, p. 1051
M.B. Wise: Combining Chiral and Heavy Quark Symmetry, in Particle
Physics at the Fermi Scale, ed. by Y. Pang, J. Qui, Z. Qiu (Gordon and Breach,
Amsterdam 1994) p. 71
Contents
Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
19
19
22
25
27
30
33
Renormalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1 Renormalization of QCD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2 Gluon Propagator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3 Quark Propagator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4 Renormalization of HQET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.5 HeavyElectron Eective Theory . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
35
35
39
45
50
55
58
59
59
63
69
71
76
80
84
89
Contents
HeavyLight Currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1 Bilinear Quark Currents in QCD . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Axial Anomaly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3 The t HooftVeltman 5 and the Anticommuting 5 . . . . . . . .
5.4 HeavyLight Current in HQET . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.5 Decoupling for QCD and HQET Currents . . . . . . . . . . . . . . . . .
5.6 QCD/HQET Matching for HeavyLight Currents . . . . . . . . . . .
5.7 Meson Matrix Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
91
91
95
98
102
103
107
115
119
121
121
126
131
135
140
144
HeavyHeavy Currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.1 HeavyHeavy Current in HQET . . . . . . . . . . . . . . . . . . . . . . . . . .
7.2 FlavourDiagonal Currents at Zero Recoil . . . . . . . . . . . . . . . . . .
7.3 FlavourChanging Currents at Zero Recoil . . . . . . . . . . . . . . . . .
7.4 FlavourDiagonal Currents at NonZero Recoil . . . . . . . . . . . . .
7.5 FlavourChanging Currents at NonZero Recoil . . . . . . . . . . . . .
7.6 1/m Corrections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
145
145
152
155
161
162
171
173
Renormalons in HQET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.1 Large0 Limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.2 Renormalons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.3 Light Quarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.4 HQET Heavy Quark . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.5 OnShell Heavy Quark in QCD . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.6 Chromomagnetic Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.7 HeavyLight Currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.8 HeavyHeavy Currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
175
175
181
183
186
191
194
197
203
209
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
Notation
1 0 0 0
0 1 0 0
g =
0 0 1 0 .
0 0 0 1
The velocity 4vector obeys v v = 1. The unit antisymmetric tensor is
normalized by 0123 = 0123 = 1. The Dirac matrices are defined by
+ = 2g .
For any vector a, we use the notation a
/ = a ; = (i/2)[ , ]. The
matrix 5 is defined by
5 =
i
= i 0 1 2 3 ;
4!
it satisfies
52 = 1 ,
5 + 5 = 0 .
A = gAa ta ,
where the index a varies from 1 to the number of gluons Ng = Nc2 1, and
the generators in the fundamental (quark) representation are normalized by
the condition
A. Grozin: Heavy Quark Eective Theory, STMP 201, 12 (2004)
c SpringerVerlag Berlin Heidelberg 2004
Notation
Tr ta tb = TF ab ,
TF =
1
.
2
The commutator of the generators is [ta , tb ] = if abc tc . The covariant derivative in the adjoint representation is
Dab = ab iAab
,
c c ab
Aab
= gA (t ) ,
CF =
Nc2 1
2Nc
References
1. J.C. Collins: Renormalization (Cambridge University Press, Cambridge 1984) 1
2. M.E. Peskin, D.V. Schr
oder: Quantum Field Theory (Perseus, Reading, MA
1995) 1
rotations can transform the B and B into each other; they are degenerate
and have identical properties in this limit. This heavyquark spin symmetry
yields many useful relations among heavyhadron form factors [13]. Not only
the orientation, but also the magnitude of the heavyquark spin is irrelevant
in the innitemass limit. We can switch o the heavyquark spin, making
it spinless, without aecting the physics. This trick considerably simplies
counting independent form factors, and we shall use it often. Or, if we wish,
we can make the heavy quark have spin 1; it does not matter.
This leads to a supersymmetry group called the superavour symmetry [10, 6]. It can be used to predict properties of hadrons containing a scalar
or vector heavy quark. Such quarks exist in some extensions of the Standard
Model (for example, supersymmetric or composite extensions). This idea can
also be applied to baryons with two heavy quarks. They form a smallsize
bound state (with a radius of order 1/(ms )) which has spin 0 or 1 and is
antitriplet in colour. Therefore, these baryons are similar to mesons with a
heavy antiquark that has spin 0 or 1. The accuracy of this picture cannot be
high, however, because even the radius of the bb diquark is only a few times
smaller than the connement radius.
600
2 2+
D
1 1+
D
400
200
0
D
= cd D
0 = cu D
cs
D
s =
600
400
B1,2 1+ , 2+
200
B
0
B 0
B0 = bd
B0s = bs
B+ = bu
400
200
3
2
3
2
1
2
1
2
3+
2
1+
2
1+
2
1+
2
1+
2
++
+
= cuu
c = c[ud] c
0
+
c c
+
c = csu
0c
0c = css
The c baryon has not yet been observed. Only one baryon with a b quark,
0b , has been discovered so far.
In the leading mb approximation, the masses mB and mB are both
where is the energy of the ground state of the light elds
equal to mb + ,
antiquark. This energy is of the order of
in the chromoelectric eld of the b
QCD . The excited states of the light elds have energies i , giving excited
degenerate doublets with masses mb + i .
antiquark has
There are two 1/mb corrections to the masses. First, the b
an average momentum squared 2 , which is of order of 2QCD . Therefore, it
chromomagnetic moment interhas a kinetic energy 2 /(2mb ). Second, the b
acts with the chromomagnetic eld created by light constituents at the origin,
stays. This chromomagnetic eld is proportional to the angular
where the b
momentum of the light elds j l . Therefore, the chromomagnetic interaction
energy is proportional to
, s = 0,
1
4
sQ j l = [s(s + 1) sQ (sQ + 1) jl (jl + 1)] =
2
1 , s = 1,
4
where s = sQ + j l is the meson spin. If we denote this energy for B by
2G /(2mb ), then for B it will be (1/3)2G /(2mb ). Here 2G is of order of
2QCD . The B, B meson masses with 1/mb corrections taken into account
are given by the formulae
3QCD
2 2G
mB = mb + +
+O
,
2mb
m2b
2
2
3QCD
+
(1/3)
G
mB = mb + +
+O
.
(1.1)
2mb
m2b
The hyperne splitting is
3QCD
22G
mB mB =
+O
.
3mb
m2b
Taking into account mB + mB = 2mb + O(QCD ), we obtain
3QCD
4 2
2
2
mB mB = G + O
.
3
mb
The dierence m2D m2D is given by a similar formula, with mc instead of
mb . Therefore, the ratio
QCD
m2B m2B
=1+O
.
(1.2)
m2D m2D
mc,b
2
,
5
B(D2 D ) =
3
.
5
imB fB
0b 0 5 uB nr =
.
2mB
Denoting
this matrix element (which is massindependent at mb ) by
iF/ 2, we obtain
F
QCD
fB =
1+O
,
(1.3)
mb
mb
and hence
fB
=
fD
QCD
mc
1+O
.
mb
mc,b
(1.4)
As we shall discuss in Chap. 5, the matrix element F depends on the normalization scale, and hence is not quite the same for D and B; this produces
moderate perturbative corrections to (1.4). Lattice simulations and QCD sum
rules show that the 1/mc correction in the formula for fD similar to (1.3) is
of order 100%, so that the accuracy of (1.4) is not high.
Experimentally [18],
= 280 19 28 34 MeV ,
f D+
s
fD+ = 300+180+80
15040 MeV ,
10
b = 1.
(1.5)
c J
will be used to denote
The groundstate B meson has sP = (1/2)+ ; D
generically a groundstate or excited cq meson. It is convenient to work in
motion. Angularthe B rest frame. Let the z axis be in the direction of the D
momentum conservation gives sz = sz . Reection in a plane containing the z
axis transforms a state s, sz into P i2s s, sz . Therefore, the amplitude of
the sz into sz transition is equal to that of the sz into sz transition, up
to a phase factor; an sz = 0 into sz = 0 transition is allowed only when the
naturalness P (1)s is conserved [19]. For example, the transition b c
is described by a single form factor; b c is forbidden by naturalness
(and also suppressed by isospin); b c is described by two form factors
(sz = sz = 0 and 1) [16, 9, 17].
meson into an Swave
The transitions of the groundstate (1/2)+ bq
+
(1/2) cq meson, and into a P wave (1/2) or (3/2) cq meson, are described by one form factor each [13, 8, 14, 7]:
u 5 u ,
(1/2) J (1/2)+ = 1/2 (cosh )
u ,
(1.6)
(3/2) J (1/2)+ = 3/2 (cosh )v u
where cosh = v v , and is the Minkowski angle between the 4velocities
of the B and D.
The Dirac wave function u of the initial (1/2)+ meson satises (/v 1)u = 0
and is normalized by the nonrelativistic condition u
u = 1; the sum over its
two polarizations is
u
u=
1 + v/
.
2
The Dirac wave function u of the nal (1/2) meson has similar properties,
with v instead of v. The RaritaSchwinger wave function u of the spin(3/2)
meson satises (/v 1)u = 0, u = 0, v u = 0, and is normalized by
u
u = 1; the sum over the four polarizations of the meson is
1 + v/
1 + v/
1
2
.
(1.7)
=
u u
g + + v v
2
3
3
2
D
via the vector and axial
All the form factors of B transitions into D,
11
(1.8)
D
decays is about 1.6; a rough
The maximum cosh accessible in B D,
sketch of (cosh ) as extracted from experimental data is shown in Fig. 1.4.
At cosh 1, the IsgurWise form factor behaves as [12]
s
,
(1.9)
(cosh )
cosh2
up to logarithmic factors.
1
0.8
0.6
0.4
0.2
0
1.1
1.2
1.3
1.4
1.5
cosh
12
Squaring the matrix elements (1.6), summing over the nal meson polarizations (using (1.7) for the spin 3/2 meson), averaging over the initial
meson polarizations, and normalizing to the quark decay (1.5), we obtain the
branching ratios
cosh + 1 2
(cosh ) ,
B (1/2)+ (1/2)+ =
2
cosh 1 2
B (1/2)+ (1/2) =
1/2 (cosh ) ,
2
(cosh + 1)2 (cosh 1) 2
3/2 (cosh ) .
B (1/2)+ (3/2) =
3
(1.10)
cosh + 1 2
i (cosh )( i )
2
+
(1/2)
cosh 1 2
1/2 (cosh )( i )
2
(1/2)
13
(3/2)
(1.11)
where the sums run over nal states with the quantum numbers indicated, i
are their excitation energies, the index i is not explicitly shown in the form
factors 1/2 and 3/2 , and the dots mean the contribution of Dwave and
higher states. At = 0, F (, 1) = (). A qualitative sketch of F (, cosh )
as a function of at some xed > 0 is shown in Fig. 1.6. It contains a
and D
), then
peak at = 0 due to the transition into the ground state (D
some peaks due to excited states which become wider when increases, and
then the curve becomes smooth. At QCD , it is given by the perturbative
gluon radiation:
F (, cosh ) =
2CF s ( coth 1)
.
(1.12)
(1.13)
14
This is the Bjorken sum rule [4, 5, 14]. In particular, the decay rate into the
groundstate (1/2)+ meson must not exceed the total rate:
2
.
(1.14)
(cosh )
cosh + 1
At cosh 1, the ground state is rarely produced, and the IsgurWise form
factor (1.9) is much less than the bound (1.14).
The Bjorken sum rule becomes much simpler in the small limit. The
IsgurWise form factor of the transition into the ground state behaves as
(cosh ) = 1 2 (cosh 1) + , and form factors of transitions into
higher Swave mesons as i (cosh ) = 2i (cosh 1) + . Expanding (1.13)
up to linear terms in cosh 1, we obtain
1
1 2
4 2
2
1+
2 +
(1.15)
1/2 (1) +
3/2 (1) (cosh 1) = 1 .
2
2
3
Dwave nal state do not contribute in this order, nor do higher Swave nal
states. Therefore, the slope of the IsgurWise form factor 2 can be expressed
via the form factors of P wave meson production at cosh = 1:
1 1 2
2 2
(1.16)
1/2 (1) +
3/2 (1) .
2 = +
4 4
3
In particular,
2 >
1
4
(1.17)
/v 1 2
v/ + 1 2
u
u
i (cosh )( i ) + u
1/2 (cosh )( i )
2
2
+
(1/2)
(1/2)
cosh + 1
2
+
[2 cosh 1 (2 cosh )
u/v u]
3/2
(cosh )( i )
3
(3/2)
+
Averaging it over polarizations u
/v u cosh , we reproduce (1.11). The
decay rate does not depend on the initial meson polarization. This gives the
Uraltsev sum rule [22]
2
i2 (cosh ) +
1/2
(cosh )
(1/2)+
(1/2)
2
2
(cosh + 1)(2 cosh )
3/2
(cosh ) + = 0 .
3
(3/2)
(1.18)
15
This sum rule becomes much simpler at 0. D and higherwave contributions vanish, and
1+
2
(1)
1/2
Substituting
2 =
4 2
3/2 (1) = 0 .
3
(1.19)
2
3/2
(1) from this sum rule into (1.16), we obtain
3
2
1+
1/2
(1) .
4
(1.20)
In particular,
2 >
3
.
4
(1.21)
This Uraltsev bound is much stronger than the Bjorken bound (1.17). Experimentally, 2 0.8.
More sum rules can be obtained from energy conservation. In the v rest
They have
frame, the light elds in the B have a denite energy E = .
no denite momentum, because they are in the external chromoelectric eld
antiquark. The average of this momentum is p = 0, and
created by the b
the average of its square is p2 = 2 (it is the same as the average squared
momentum of the heavy antiquark, see (1.1)). The light elds energy in the
with a 4velocity v is
v rest frame is E = cosh px sinh . When a b
suddenly transformed into a c with a 4velocity v , the light elds remain
in their original state at rst. After that, the energy E is conserved in the
eld of the c moving with the 4velocity v . Therefore, the average excitation
energy of the Xc and the average squared excitation energy are
E = (cosh
1),
2
(because px = p /3). This gives the Voloshin sum rule [23] and the
BGSUV sum rule [1]:
F (, cosh ) d = (cosh
1) ,
(1.22)
2
(1.23)
F (, cosh )2 d = 2 (cosh 1)2 + (cosh2 1) .
3
The transition into the groundstate meson with = 0 does not contribute
here.
Expanding these sum rules up to linear terms in cosh 1, we obtain,
similarly to (1.15),
16
1 2
4 2
1/2 (1)i +
3/2 (1)i = ,
2
3
1 2
4 2
2
1/2 (1)2i +
3/2 (1)2i = 2 .
2
3
3
(1.24)
(1.25)
1
+
.
4 2
(1.27)
References
17
(1.29)
References
1. I. Bigi, A.G. Grozin, M. Shifman, N.G. Uraltsev, A. Vainshtein: Phys. Lett. B
339, 160 (1994) 15, 16
2. I. Bigi, M.A. Shifman, N.G. Uraltsev: Annu. Rev. Nucl. Part. Sci. 47, 591
(1997) 16
3. I. Bigi, M. Shifman, N.G. Uraltsev, A. Vainshtein: Phys. Rev. D 52, 196 (1995)
16
4. J.D. Bjorken: New Symmetries in Heavy Flavor Physics. In Results and Perspectives in Particle Physics, ed. by M. Greco (Editions Frontieres, GifsurYvette 1990) p. 583 14
5. J.D. Bjorken, I. Dunietz, J. Taron: Nucl. Phys. B 371, 111 (1992) 14
6. C.D. Carone, Phys. Lett. B 253, 408 (1991) 4
7. A.F. Falk: Nucl. Phys. B 378, 79 (1992) 10
8. A.F. Falk, H. Georgi, B. Grinstein, M.B. Wise: Nucl. Phys. B 343, 1 (1990)
10
9. H. Georgi: Nucl. Phys. B 348, 293 (1991) 10
10. H. Georgi, M.B. Wise: Phys. Lett. B 243, 279 (1990) 4
11. A.G. Grozin, G.P. Korchemsky: Phys. Rev. D 53, 1378 (1996) 16
12. A.G. Grozin, M. Neubert: Phys. Rev. D 55, 272 (1997) 11
13. N. Isgur, M.B. Wise: Phys. Lett. B 237, 527 (1990) 4, 10
14. N. Isgur, M.B. Wise: Phys. Rev. D 43, 819 (1991) 10, 14
15. N. Isgur, M.B. Wise: Phys. Rev. Lett. 66, 1130 (1991) 8
16. N. Isgur, M.B. Wise: Nucl. Phys. B 348, 276 (1991) 10
17. T. Mannel, W. Roberts, Z. Ryzak: Nucl. Phys. B 355, 38 (1991) 10
18. Particle Data Group: Eur. Phys. J. C 15, 1 (2000) 5, 6, 8, 9
19. H.D. Politzer: Phys. Lett. B 250, 128 (1990) 10
20. E. de Rafael, J. Taron: Phys. Lett. B 282, 215 (1992) 17
21. N.G. Uraltsev: Heavy Quark Expansion in Beauty and its Decays, in Heavy
Flavor Physics a Probe of Natures Grand Design, ed. by I. Bigi, L. Moroni
(IOS Press, Amsterdam 1998) p. 329; Topics in the Heavy Quark Expansion,
in At the Frontier of Particle Physics: Handbook of QCD, ed. by M. Shifman
(World Scientic, Singapore 2001) v. 3, p. 1577 16
22. N.G. Uraltsev: Phys. Lett. B 501, 86 (2001); J. Phys. G 27, 1081 (2001) 14
23. M.B. Voloshin: Phys. Rev. D 46, 3062 (1992) 15, 16
24. A. Le Yaouanc, L. Oliver, J.C. Raynal: Phys. Rev. D 67, 114009 (2003) 16
In this chapter, we introduce HQET an eective eld theory approximating QCD for problems with a single heavy quark in certain kinematics, to
the leading order in 1/m (Sect. 2.1). The rest of the chapter is about methods used for calculation of one and twoloop Feynman diagrams in HQET.
These methods have much in common with the ones used to calculate loop
diagrams in massless theories, such as QCD with light quarks. Therefore,
for the convenience of readers, we shall recall some wellknown facts about
massless diagrams, too.
(2.1)
where Q is the heavyquark eld, and the dots mean all the terms with light
m)Q gives the
quarks and gluons. The free heavyquark Lagrangian
Q(i/
dependence of the energy on the momentum p0 = m2 + p2 . If we assume
that the characteristic momenta p m, then we can simplify the dispersion
0 0 m)Q. In
law to p0 = m. This law corresponds to the Lagrangian Q(i
our class of problems, the lowestenergy state (vacuum) consists of a single
particle the heavy quark at rest. Therefore, it is convenient to use the
energy of this state m as a new zero level. This means that, instead of the
true energy p0 of the heavy quark (or any state containing this quark), we
A. Grozin: Heavy Quark Eective Theory, STMP 201, 1933 (2004)
c SpringerVerlag Berlin Heidelberg 2004
20
shall use the residual energy p0 = p0 m. Then the onshell heavy quark
has an energy p0 = 0 independently of the momentum. The free Lagrangian
0 0 Q. The spin of the heavy quark at
giving such a dispersion law is Qi
(we can also consider
rest can be described by a twocomponent spinor Q
= Q).
it as a fourcomponent spinor with vanishing lower components: 0 Q
Reintroducing the interaction with the gluon eld by the requirement of gauge
invariance, we arrive at the HQET Lagrangian [4]
+ ,
+ iD0 Q
L=Q
(2.2)
where all lighteld parts (denoted by dots) are exactly the same as in QCD.
The eld theory (2.2) is not Lorentzinvariant, because the heavy quark denes a selected frame its rest frame.
The Lagrangian (2.2) gives the staticquark propagator
p) =
S(
1
,
p0 + i0
0 )(x) ,
S(x)
= S(x
= i(t) .
S(t)
(2.3)
In the momentum space it depends only on p0 and not on p, because we have
neglected the kinetic energy. Therefore, in the coordinate space, the static
quark does not move. The unit 22 matrix is assumed in the propagator (2.3).
It is often convenient to use it as a 4 4 matrix; in such a case the projector
(1 + 0 )/2 onto the upper components is assumed. The static quark interacts
only with A0 ; the vertex is ig0 ta .
The staticquark propagator in a gluon eld is given by the straight Wilson
line
S(x) = i(x0 )(x)P exp ig A dx
.
(2.4)
Many properties of HQET were rst derived in the course of the investigation
of the renormalization of Wilson lines in QCD. In fact, the HQET Lagrangian
was used as a technical device for investigation of Wilson lines.
Loops of a static quark vanish, because it propagates only forward in time.
In other words, in the momentum space, all poles in the p0 plane lie in the
lower halfplane; closing the integration contours upward, we obtain zero.
The Lagrangian (2.2) can be rewritten in covariant notation [7, 6]:
v iv DQ
v +
Lv = Q
(2.5)

p  m .
(2.6)
1
p) = 1 + v/
S(
,
2 p v + i0
21
(2.7)
(2.9)
where m is called the residual mass term. The most convenient denition of
the heavyquark mass m is the one which gives m = 0 in (2.9). It gives the
22
k+p
Fig. 2.1. Oneloop massless propagator diagram
(2.12)
23
(d/2 + n1 + n2 )
(p2 )d/2n1 n2
(n1 ) (n2 )
1
xd/2n2 1 (1 x)d/2n1 1 dx .
D2 = k 2 ,
(2.17)
24
G(n1 , n2 ) =
(2.18)
where the correct i0 terms are assumed in D1,2 . If n1,2 are integer, G(n1 , n2 )
is proportional to G1 = (1 + ) 2 (1 )/ (1 2), the coecient being a
rational function of d.
The ultraviolet (UV) divergences of the original integral (2.10) are reproduced by the loopmomentum integral (2.16), and are given by the poles
of (d/2 + n1 + n2 ) (with a minus sign in front of d). The infrared (IR)
divergences of the original integral reside in the integral in the Feynman parameter x (2.17), and are given by the poles of (d/2 n1 ) and (d/2 n2 )
(with a plus sign in front of d).
In many cases, calculation of diagrams in the coordinate space can be
simpler than in the momentum space. In particular, the oneloop propagator
diagram of Fig. 2.1 in the coordinate space is just the product of two propagators. The massless propagators in the pspace and the xspace are related
to each other by a Fourier transform:
1
dd p
(d/2 n)
eipx
= i22n d/2
, (2.19)
2
n
d
2
(p i0) (2)
(n)
(x + i0)d/2n
1
eipx
(d/2 n)
dd x = i2d2n d/2
(2.20)
2
n
2
(x + i0)
(n)
(p i0)d/2n
(sanity checks: transform 1/(p2 i0)n to xspace (2.19) and back (2.20),
and you get 1/(p2 i0)n ; take the complex conjugate of (2.19) and rename
x p, and you get (2.20)). Multiplying two propagators (2.19) with degrees
n1 and n2 , we obtain
22(n1 +n2 ) d
(d/2 n1 ) (d/2 n2 )
1
.
(n1 ) (n2 )
(x2 )dn1 n2
Applying the inverse Fourier transform (2.20), we reproduce the result (2.18).
The oneloop diagram (2.18) is an analytic function in the complex p2
plane with a cut. The cut at p2 > 0, where real pair production is possible,
begins at a branch point at the threshold p2 = 0. This cut comes from the
factor (p2 ) , which is equal to (p2 ) ei at the upper side of the cut and
(p2 ) ei at the lower side, and hence has a discontinuity 2i(p2 ) sin .
Therefore, the discontinuity of the oneloop diagram (2.18) at 0 is proportional to the residue of G(n1 , n2 ) at = 0. The discontinuity of the diagram
with n1 = n2 = 1 can be calculated directly using the Cutkosky rules: draw
a cut across the loop indicating the real intermediate state, and replace the
cut propagators by their discontinuities:
1
2i(p2 ) .
p2 i0
(2.21)
25
dd k
.
((k + p)0 i0)n1 (k 2 i0)n2
I=
(2.22)
k + p
Fig. 2.2. Oneloop HQET propagator diagram
26
0
y d/2n2 1 (y 2)d/2n1 n2 dy
= (2)dn1 2n2
(d + n1 + 2n2 ) (d/2 n2 )
.
(d/2 + n1 + n2 )
(2.26)
D2 = k 2 ,
(d + n1 + 2n2 ) (d/2 n2 )
.
(n1 ) (n2 )
(2.27)
(d/2 n2 )
(it)n1 +2n2 d1 (t)
(n1 ) (n2 )
(where x2 = (it)2 ). Applying the inverse Fourier transform (2.29), we reproduce the result (2.27).
27
(2.30)
4
5
5
1
2
a
2
b
2
c
D3 = k12 ,
D2 = (k2 + p)2 ,
D4 = k22 ,
D5 = (k1 k2 )2 .
(2.31)
(2.32)
If n1 = 0 (Fig. 2.3c), the (3, 5) integral (2.18) gives G(n3 , n5 )/(k22 )n3 +n5 d/2 ;
this is combined with the denominator 4, and we obtain
28
(2.33)
n2
n4
n5
2(k2 + p) +
2k2 +
2(k2 k1 ) .
k2
D2
D4
D5
(2.34)
(2.35)
and similar ones for the other indices. Then our recurrence relation can be
written in a shorter and more easily digestible form,
d n2 n5 2n4 + n2 2+ (1 4 ) + n5 5+ (3 4 ) G = 0 .
(2.36)
This is a particular example of the triangle relation. We dierentiate with
respect to the loop momentum running along the triangle 254, and insert
the momentum of the line 4 in the numerator. The dierentiation raises
the degree of one of the denominators 2, 5, 4. In the case of the line 4, we
29
obtain 2D4 in the numerator, giving just 2n4 . In the case of the line
5, we obtain the denominator D3 of the line attached to the vertex 45 of
our triangle, minus the denominators D4 and D5 . The case of the line 2 is
similar; the denominator of the line attached to the vertex 24 of our triangle
is just p2 , and it does not inuence any index of G. Of course, there are
three more relations that can be obtained from (2.36) by symmetry. Another
useful triangle relation is derived by applying the operator (/k2 ) (k2 k1 ):
d n2 n4 2n5 + n2 2+ (1 5 ) + n4 4+ (3 5 ) G = 0 .
(2.37)
2(d n3 n4 n5 ) n1 n2 + n1 1+ (1 3 ) + n2 2+ (1 4 ) G = 0 .
(2.38)
This is nothing but the sum of the (/k2 ) k2 relation (2.36) and its mirrorsymmetric (/k1 ) k1 relation.
Another interesting relation is obtained by inserting (k1 + p) into the
integrand of (2.31) and taking the derivative /p of the integral. On the
one hand, the vector integral must be proportional to p , and we can make
the substitution
(k1 + p) p
D1 D3 p
k1 + p
p
=
1
+
p2
p2
2
in the integrand. Taking /p of this vector integral produces (2.31) with
3
D1 D3
d
ni
1+
2
p2
inserted into the integrand. On the other hand, explicit dierentiation with
respect to p gives
n2
n1
2(k1 + p)2 +
2(k2 + p) (k1 + p) ,
D1
D2
2(k2 + p) (k1 + p) = D5 D1 D2 .
d+
Therefore, we obtain
30
3
1
d + n1 n3 n4 n5 +
d
ni (1 3 )
2
2
+
+ n2 2 (1 5 ) G = 0 .
(2.39)
n2 2+ (5 1 ) + n4 4+ (5 3 )
G,
d n2 n4 2n5
(2.40)
we see that the sum n1 +n3 +n5 is reduced by 1. Therefore, by applying (2.40)
suciently many times, we can reduce an arbitrary G integral with integer
indices to a combination of integrals with n5 = 0 (Fig. 2.3b, (2.32)), n1 = 0
(Fig. 2.3c, (2.33)), and n3 = 0 (mirrorsymmetric to the previous case). Of
course, if max(n2 , n4 ) < max(n1 , n3 ), it may be more ecient to use the
relation mirrorsymmetric to (2.37). The relation (2.39) can also be used
instead of (2.37). Thus, any integral G(n1 , n2 , n3 , n4 , n5 ) with integer ni can
be expressed as a linear combination of G21 and G2 , the coecients being
rational functions of d. Here the combinations of functions appearing in
the nloop sunset massless diagram are
Gn =
k12 ,
D4 =
D2 = (k2 + p) v/ ,
k22 ,
D5 = (k1 k2 )2 .
(2.42)
31
4
2
3
5
1
3
5
2
a
b
4
4
5
2
c
2
d
5
1
2
e
(2.43)
If n1 = 0 (Fig. 2.4d), the (3, 5) integral (2.18) gives G(n3 , n5 )/(k22 )n3 +n5 d/2 ;
this is combined with the denominator 4, and we obtain
I(0, n2 , n3 , n4 , n5 ) = G(n3 , n5 )I(n2 , n4 + n3 + n5 d/2)
(2.44)
(and similarly for n2 = 0). If n3 = 0 (Fig. 2.4e), the (1, 5) integral (2.27) gives
I(n1 , n5 )/(2)n1 +2n5 d ; this is combined with the denominator 2, and we
obtain
I(n1 , n2 , 0, n4 , n5 ) = I(n1 , n5 )I(n2 + n1 + 2n5 d, n4 )
(2.45)
n2 v
n5
+
2k2 +
2(k2 k1 ) .
k2
D2 D4
D5
(2.46)
d n2 n5 2n4 +
(2.47)
(2.48)
32
n4 2
n5 2
n2
+
4 (D2 1) +
4 (D2 D1 ) ,
D2
D4
D5
(2.49)
(2.51)
(2(d n3 n4 n5 ) n1 n2 + 1)1 + n4 4+ (5 3 )
I,
d n1 n2 n4 2n5 + 1
(2.53)
we see that the sum n1 + n3 + n5 is reduced by 1. Therefore, by applying (2.53) suciently many times, we can reduce an arbitrary I integral
References
33
(2.54)
k12 ,
D5 =
D2 = (k2 + p) v/ ,
k22 .
D3 = (k1 + k2 + p) v/ ,
(2.55)
(2.56)
References
D.J. Broadhurst, A.G. Grozin: Phys. Lett. B 267, 105 (1991) 31
K.G. Chetyrkin, A.G. Grozin: Nucl. Phys. B 666, 289 (2003) 33
K.G. Chetyrkin, F.V. Tkachov: Nucl. Phys. B 192, 159 (1981) 28, 30
E. Eichten, B. Hill: Phys. Lett. B 234, 511 (1990) 20
A.F. Falk, M. Neubert, M. Luke: Nucl. Phys. B 388, 363 (1992) 22
H. Georgi: Phys. Lett. B 240, 447 (1990) 20
B. Grinstein: Nucl. Phys. B 339, 253 (1990) 20, 21
A.G. Grozin: JHEP 03, 013 (2000) 33
N. Isgur, M.B. Wise: Phys. Lett. B 237, 527 (1990) 21
A.S. Kronfeld: Phys. Rev. D 58, 051501 (1998) 22
V.A. Smirnov: Applied Asymptotic Expansions in Momenta and Masses
(Springer, Berlin, Heidelberg 2001) 21
12. F.V. Tkachov: Phys. Lett. B 100, 65 (1981) 28
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
3 Renormalization
i
1
1
2
qi0 (iD
/ 0 mi0 )qi0 Ga0 Ga
1/2
A0 = ZA A ,
mi0 = Zm mi ,
c0 = Zc1/2 c ,
a0 = Z A a ,
(3.2)
(3.3)
It will be explained in Sect. 3.2 why the gaugexing parameter a is renormalized by the same factor ZA as the gluon eld A.
Lagrangian has the dimensionality [L] = d, because the action S =
The
L dd x is exactly dimensionless in a spacetime with any d. The gluon kinetic
2
term, which is the g00 term in (1/4)Ga0 Ga
0 , has the structure (A0 ) ;
hence, the dimensionality of the gluon eld is [A0 ] = d/21 = 1. Similarly,
from the quark kinetic terms, the dimensionality of the quark elds is [qi0 ] =
A. Grozin: Heavy Quark Eective Theory, STMP 201, 3558 (2004)
c SpringerVerlag Berlin Heidelberg 2004
36
3 Renormalization
or
s ()
g02
Z (s ())e . (3.4)
= 2
d/2
4
(4)
Here is the renormalization scale, and the factor exp() is included for
convenience (it allows one to get rid of the Euler constant in all equations in
the limit 0 when they are written via s ). This renormalization scheme
is called MS. All the bare quantities, including g0 , are independent. By
dierentiating (3.4), we obtain the renormalizationgroup equation for s ():
d log s
= 2 2(s ) ,
d log
2
s
1 d log Z
s
= 0
+ 1
(s ) =
+ ,
2 d log
4
4
(3.5)
where
0 =
11
4
CA TF nf ,
3
3
1 =
34 2
20
CA 4CF TF nf CA TF nf ,
3
3
(3.6)
and 2,3 are also known [7] (we shall derive 0 in Sect. 4.5). Here nf is the
number of avours.
Sometimes it may be convenient to use other denitions of s () (renormalization schemes). For example, we can dene
s () = s (f ()) so that
()
g02
Z (s ())e f ()2 .
= 2 s
d/2
4
(4)
2
s ()
s ()
s ()
s ()
+ .
=
1 + c1
+ c2
4
4
4
4
37
1 = 1 ,
2 = 2 + c2 , . . .
The rst two function coecients 0,1 are the same in all massindependent
(or minimal) schemes; the higher coecients 2,3,... can be set to arbitrary
values by adjusting the renormalization scheme (i.e., the coecients c2,3,...).
In this book, we shall always use the MS scheme.
The information contained in Z is equivalent to that in (s ). Equation (3.5) can be rewritten as
d log Z
(s ) 2 (s ) 3 (s )
(s )
=
+
=
+
d log s
+ (s )
2
3
Any minimal renormalization constant (3.3) can be represented as
Z1
Z2
Z = exp
+ 2 + ,
(3.7)
(3.8)
where Z1 starts from the order s , Z2 starts from 2s , and so on. Then
dZ1
= (s ) ,
d log s
and
dZ2
= 2 (s ) ,
d log s
dZ3
= 3 (s ) , . . . (3.9)
d log s
s
1 s 2 1 s 3
s
ds
1
Z1 =
(s )
= 0
2
,
s
4 2
4
3
4
0
s
3
ds
1 s 2 2
s
2 (s )
= 02
+ 0 1
+ ,
Z2 =
s
2
4
3
4
0
s
ds
1 s 3
Z3 =
3 (s )
= 03
,
s
3
4
0
...
(3.10)
38
3 Renormalization
da()
+ A (s ())a() = 0 ,
d log
where A =
d log ZA
.
d log
(3.12)
For any renormalization constant (3.3), the corresponding anomalous dimension is dened by
(s ) =
2
s
d log Z
s
= 0
+ 1
+
d log
4
4
(3.13)
+
d log s
2 + (s )
2
22
23
(3.14)
Then
1
dZ2
1
dZ1
= (s ) ,
= (s )(s ) ,
d log s
2
d log s
2
dZ3
1 2
= (s )(s ) , . . .
d log s
2
and
(3.15)
1 s 2 1 s 3
ds
1 s
1
= 0
2
,
s
2 4 4
4
6
4
0
1 s
ds
(s )(s )
Z2 =
2 0
s
2 1
3
1
s
s
= 0 0
+ (0 1 + 1 0 )
+ ,
4 4
6
4
s
s 3
1
ds
1
Z3 =
2 (s )(s )
= 02 0
,
2 0
s
6
4
...
(3.16)
Z1 =
1
2
(s )
log Z
log Z
= (s ) .
A (s )a
log s
a
(3.17)
Equating 0 terms, we see that the rst equation in (3.15) does not change,
(0)
and Z1 is still given by (3.16). Equating 1 terms, we nd Z2 = Z2 + Z2 ,
(0)
where Z2 is given by (3.16) and
Z2 =
1
2
A (s )a
Z1 (s ) ds
1
d0 s 2
= A0 a
+
a
s
8
da 4
39
(3.18)
+ ,
where
0
(p) =
D
(3.20)
1
p p
(1
a
)
g
0
p2
p2
(3.21)
is the free gluon propagator, and the gluon selfenergy (polarization operator) i (p) is the sum of oneparticleirreducible gluon selfenergy diagrams
(which cannot be separated into two disconnected parts by cutting a single
gluon line). The series (3.20) implies the equation
0
0
(p) + D
(p) (p)D (p) .
D (p) = D
(3.22)
40
3 Renormalization
then
A1
A1
p p
p p
.
g 2
+ A1

p
p2
(3.23)
(3.24)
(3.25)
r
D
(p2 ; )
p p
p p
g 2
+ a() 2 2 .
p
(p )
(3.26)
G1
g02 (p2 )
d/2
(d 1)(d 3)(d 4)
(4)
1
3d + 2 (d 1)(2d 7) + (d 1)(d 4) 2 CA
4
+ 4(d 2)TF nf ,
(3.27)
41
1
= 1 + (p2 )
1 (p2 )
via the renormalized quantities s (), a() instead of the bare ones g02 , a0 .
These quantities only appear in the s correction, and we may omit the s
terms in them:
s ()
g02
,
= e 2
d/2
4
(4)
a0 = a() .
(1)n n n
,
(1 + ) = exp +
n
n=2
(3.29)
where
n =
1
,
kn
k=1
2 =
2
,
6
3 1.202 ,
4 =
4
,
90
...
(3.30)
p2 D (p2 ) = 1 +
4
2
3
3
2
9a + 18a + 97
20
CA TF nf + ,
+
36
9
2
p
L = log 2 .
r
This should be equal to ZA (s (), a())p2 D
(p2 ; ), where the renormalization constant ZA (s , a) has the minimal form (3.3), and the renormalized
r
(p2 ; ) is nite at 0. We obtain
propagator D
s 1
4
13
ZA (s , a) = 1
(3.31)
a
CA + TF nf ,
4 2
3
3
2
s 9a + 18a + 97
20
r
CA TF nf
(p2 ; ) = 1 +
p2 D
4
36
9
1
8
13
+
(3.32)
a
CA TF nf L .
2
3
3
42
3 Renormalization
Using (3.19), we arrive at the anomalous dimension of the gluon eld at one
loop,
13
8
s
+
(3.33)
a
A =
CA + TF nf
3
3
4
r
The dependence of the renormalized propagator D
(p2 ; ) (3.32) on (or L)
is determined by the renormalizationgroup equation
r
1
(p2 ; )
dD
r
= A (s (), a())D
(p2 ; ) .
dL
2
(3.34)
43
(see (2.41)).
The gluon propagator with twoloop accuracy,
p2 D (p2 ) =
1
= 1 + 1 (p2 ) + 12 (p2 ) + 2 (p2 ) ,
1 (p2 )
g02 (p2 )
g04 (p2 )2
f
(a
)
+
f2 (a0 ) .
1
0
(4)d
(4)d/2
44
3 Renormalization
g02
s
2 s ()
1
,
=
e
0
4
4
(4)d/2
s
1
a0 = a() 1 A0
,
2
4
and we obtain
p2 D (p2 ) = 1
s () L
s
s df1 (a)
1
1 0
e
e
a
f1 (a()) A0
4
4
2
4
da
2
s
e2L 2 e2 f2 (a) .
+
4
(3.36)
(3.37)
can be easily calculated using (3.29). This bare propagator must be equal to
r
ZA (s ())D
(p2 ; ) at 2 = p2 , where
2
s
s
z1 +
(z20 + z21 ) ,
4
4
2
s ()
s
r
(r1 + r11 + ) +
p2 D
(p2 ; 2 = p2 ) = 1 +
(r2 + )
4
4
ZA (s ) = 1 +
r
(because D
(p2 ; ) must be nite at 0). Equating the coecients of the
powers of in the s term, we obtain
z1 = c10 ,
r1 = c11 ,
r11 = c12 .
45
(3.39)
We see that the condition (3.38) is indeed satised; the gluon eld anomalous
dimension is
s
13
8
A = a
CA + TF nf
3
3
4
2
2a + 11a 59 2
s 2
CA + 2 (4CF + 5CA ) TF nf
+
+ (3.40)
4
4
(see [5] for the threeloop result), and the renormalized propagator at 2 =
p2 is
2
9a + 18a + 97
s ()
20
r
CA TF nf
p2 D
(p2 ; 2 = p2 ) = 1 +
36
9
4
810a3 + 2430a2 + 2817a + 83105
2
+
(2a 3)3 +
CA
2592
2
55
360a + 8659
s
+ 163
.
CF TF nf 83 +
CA TF nf
3
324
4
The propagator for an arbitrary can be found by solving the renormalizationgroup equation (3.34) (this will be discussed in Sect. 4.6). Alternatively,
1
we can multiply (3.36) by ZA
(); the result is nite for all , owing to (3.38).
46
3 Renormalization
(3.41)
1
p + m0
/
= 2
p m0
/
p m20
(3.42)
where
S0 (p) =
is the free quark propagator, and the quark selfenergy (mass operator)
i(p) is the sum of oneparticleirreducible quark selfenergy diagrams
(which cannot be separated into two disconnected parts by cutting a single quark line). The series (3.41) implies the equation
S(p) = S0 (p) + S0 (p)(p)S(p) ,
(3.43)
1
1
.
=
/p m0 (p)
(p)
S01 (p)
(3.44)
(3.45)
1
1
= Zq Sr (p; ) ,
1 V (p2 ) /
p (1 V (p2 ))1 (1 + S (p2 ))m0
(3.46)
g02 (p2 )
d2
G1 a0 .
(4)d/2 (d 3)(d 4)
(3.47)
The oneloop selfenergy vanishes in the Landau gauge a0 = 0. Now we reexpress the oneloop propagator p/S(p) = 1/[1 V(p2 )] via the renormalized
quantities s (), a() and expand it in :
pS(p) = 1 CF
/
s L
e
a(1 + + ) .
4
47
Therefore,
Zq (s , a) = 1 CF a
s
,
4
(3.48)
s
.
4
(3.49)
s
+
4
(3.50)
(3.51)
x
/
(d/2)
.
2 d/2 (x2 + i0)d/2
(3.52)
2 (d/2)
/x
.
d
2
4
(x + i0)d1
(3.53)
48
3 Renormalization
V2a (p2 ) = CF
g04 (p2 )2
8(d 2)2 G2
P.
(4)d
(d 3)(d 4)(d 6)(3d 8)(3d 10)
(3.54)
(d 2)(d 2n)
G(1, n) .
dn1
The same is true for the interior gluon line, with n = 1, as in the oneloop
case (3.47). Therefore, the result contains a20 :
V2b () = CF2
g04 (p2 )2
4(d 2)2 G2
a2 .
d
(4)
(d 4)2 (3d 8)(3d 10) 0
(3.55)
The colour factor of the third diagram can be easily found using the
Cvitanovic algorithm [4] (Fig. 3.7). The gluon exchange between two quark
lines is replaced by exchange by a quarkantiquark pair, minus a colourless
exchange which compensates its coloursinglet part. The correctness of the
coecients in this identity can be checked by closing the upper quark line,
and closing it with a gluon attached (Fig. 3.8). Two example applications
of this algorithm are shown in Fig. 3.9: the rst one is a calculation of CF ,
and the second one shows that the colour factor of the diagram Fig. 3.6c is
CF /(2Nc ) = CF (CF CA /2). The colour factor of the threegluon vertex
if abc is dened via the commutator [ta , tb ] = if abc tc (Fig. 3.10). Therefore,
the colour factor of the diagram Fig. 3.6d is the dierence between those of
Fig. 3.6b and Fig. 3.6c, CF CA /2.
The diagrams in Fig. 3.6c,d can be calculated using the methods of
Sect. 2.4. Collecting all the contributions, we obtain
g04 (p2 )2
(d 2) CF (d 2)a20 + d 6 J1
V2 = CF
d
(4)
+ 2(d 3)(d 6) (3d 8)a20 d 4 J2
CA
+
2(d 3)a0 + d2 10d + 22 J1
d4
49
1
=
2
Nc
1
2
1
Nc
=0
1
=
2
Nc
1
2
1
=
2
=
Nc
Nc2 1
2Nc
1
=
2
Nc
1
2Nc
Fig. 3.10. Threegluon vertex; the curved arrow shows the order of indices in if abc
50
3 Renormalization
1
+ (d 4)(3d 8)(d2 9d + 16)(1 a0 )2
2
2(d 5)(3d 8) (1 a0 ) + 2(9d 114d + 440d 544) J2
2
(3.56)
where
J1 =
G21
,
(d 3)2 (d 4)2
J2 =
G2
(d 3)(d 4)2 (d 6)(3d 8)(3d 10)
2
7
s
+ T F nf
.
2
4
.
0 = Z
1/2 Q
Q
Q
(3.58)
All the renormalization constants in (3.2) are the same as in QCD, where the
Q ,
heavy avour Q is not counted in nf . In order to nd the new constant Z
we need to calculate the heavyquark propagator in HQET.
51
The bare (unrenormalized) staticquark propagator iS()
has the structure (Fig. 3.11)
S0 ()
iS()
= iS0 () + iS0 ()(i)()i
S0 ()(i)()i
S0 () +
+ iS0 ()(i)()i
(3.59)
where S0 () = 1/ is the free HQET propagator, and the static quark self
energy (mass operator) i()
is the sum of oneparticleirreducible HQET
selfenergy diagrams (which cannot be separated into two disconnected parts
by cutting a single heavyquark line). Summing this series, we obtain
S()
=
1
.
()
(3.60)
Fig. 3.11. Structure of diagrams for the heavy quark propagator in HQET
i
i
dd k
ig0 v 2
ig0 v
d
(2)
kv+
k
k k
g (1 a0 ) 2
.
k
After contraction over the indices, the second term in the brackets contains
(k v)2 = (k v + )2 . This factor can be replaced by 2 , because all
integrals without k v + in the denominator are scalefree and hence vanish.
Using the denition (2.27), we obtain
1
g02 (2)12
() = CF
2I(1, 1) + (1 a0 )I(1, 2) ,
2
(4)d/2
and, nally,
2
12
g (2)
()
= CF 0
(4)d/2
I1
A,
d4
A = a0 1
2
,
d3
(3.61)
52
3 Renormalization
Let us also rederive this result in the coordinate space. Using the heavyquark propagator (2.3) and the gluon propagator (3.51), we obtain
0
(x)
= CF g02 D
(vt)v v (t)
= iCF g02
(3.62)
(d/2 1)
(d 3)A(it)2d (t) .
8 d/2
g02 (2)2 I1
2A .
(4)d/2 d 4
(3.63)
s () 2L
e
(3 a() + 4) ,
4
where L = log
2
.
(3.65)
(3.66)
Q = 2CF (a 3)
s
+
4
(3.67)
Note that there is no UV divergence in the Yennie gauge a = 3. The dependence of the renormalized propagator Sr (; ) (3.66) on (or L) is determined by the renormalizationgroup equation
dSr (; )
=
Q (s (), a())Sr (; ) .
dL
(3.68)
53
Twoloop diagrams for the heavyquark selfenergy are shown in Fig. 3.13.
The rst one contains the oneloop gluon selfenergy (3.27); it can be easily
calculated using (2.27), and is proportional to I2 (2.54) and P (3.28):
4
14
2a () = CF g0 (2)
(4)d
(d 2)I2
4P .
(d 3)(d 4)2 (d 6)(2d 7)
(3.69)
1+
The same is true for the interior gluon line, with n = 1, as in the oneloop
case (3.61). Therefore, the result contains A2 :
4
14
2b () = C 2 g0 (2)
F
(4)d
(d 3)I2
A2 .
(d 4)2 (2d 7)
(3.70)
The diagram Fig. 3.13c, after killing one of the heavyquark lines (2.56),
yields I12 or I2 ; it contains A2 for the same reason as for Fig. 3.13b:
4
g0 (2)14
1
2c () = CF CF CA
2
(4)d
2
2I1
I2
(3.71)
A2 .
(d 4)2
(d 4)(2d 7)
54
3 Renormalization
The diagram Fig. 3.13d vanishes in the Feynman gauge, because the threegluon vertex vanishes after contraction with three identical vectors v. It also
vanishes if the longitudinal parts of all three gluon propagators are taken,
and hence contains no (1 a0 )3 term. The result is
g04 (2)14
1
I2
2
2d () = CF CA
I +
A(1 a0 ) .
(4)d
(d 4)2 1 2(2d 7)
(3.72)
Collecting these results together, we obtain the bare heavyquark propagator with twoloop accuracy [1],
g 2 (2)2 I1
g04 (2)4
1
S()
= 1 + CF 0
2A
+
C
F
(4)d
(d 4)2
(4)d/2 d 4
8(d 2)
4A
TF nf I2 + 2A2 CF I2
CA I12
(d 3)(d 6)(2d 7)
d3
2
(d 2)2 (d 5)
+ (d2 4d + 5)A
+
2
(d 3) (d 6) (d 3)(2d 7)
1 2
2
(d 9d + 16)(d 3)A CA I2 .
4
(3.73)
(3.74)
Q = 2CF (a 3)
4
6
3
4
(3.75)
Its dierence from the QCD quark eld anomalous dimension (3.57) is gaugeinvariant up to two loops:
127
s 2
s
44
CF
CA 3CF TF nf
+ (3.76)
Q q = 6CF
4
3
3
4
The threeloop anomalous dimension has been calculated recently [6, 2];
the dierence (3.76) is not gaugeinvariant at three loops. The renormalized
staticquark propagator at = 2 is
55
s ()
Sr (; = 2) = 1 + 4CF
4
+ CF 2(a 3)2 2 + 8 CF
(3a 29)(3a + 53)
(a2 a 10)2 +
CA
24
2
76
s
82 +
+
T F nl
3
4
(see [2] for the threeloop result).
0 t1
t2 t
0 t1
t2 t
+
Fig. 3.14. Exponentiation theorem
56
3 Renormalization
2
e0 (it/2)2
S(t) = S0 (t) exp
()A .
(4)d/2
(3.77)
iS0 (
p ) ie0 v q iS0 (
p) = ie0
(3.78)
In the gure, the photon line with the black triangle at the end means that
the vertex is contracted with the incoming photon momentum q (it includes
no photon propagator!); a dot near an electron propagator means that its
momentum is shifted by q.
= e0
S1 ( ) S1 ()
or ( , ) =
. (3.79)
The vertex function is thus also known to all orders. The charge renormalization constant Z is obtained from the requirement that the renormalized ver1/2
1 in (3.79) into
tex function g0 ZA Z
Q is nite. The factor ZQ transforms S
1
Sr and hence makes nite. Therefore, the remaining factor (Z ZA )1/2 = 1
(this is also true in QED). In the BlochNordsieck model, ZA = 1 and hence
Z = 1.
Owing to the absence of charge and photoneld renormalization, we may
replace e0 e, a0 a in the bare propagator (3.77). This propagator is
57
= e0
= e0
made nite by the minimal (in the sense of (3.3)) renormalization constant,
which is just the exponential of the oneloop term
Q = exp (a 3) ,
(3.80)
Z
4
and the anomalous dimension is exactly equal to the oneloop contribution
Q = 2(a 3)
.
4
(3.81)
58
3 Renormalization
Note that the electron propagator is nite to all orders in the Yennie gauge.
What information useful for the real HQET can be extracted from this
simple abelian model? We can obtain the CF2 term in (3.73) by Fourier transformation, without explicit calculation. There should be no CF2 term in the
twoloop anomalous dimension of the eld (3.75).
References
D.J. Broadhurst, A.G. Grozin: Phys. Lett. B 267, 105 (1991) 54
K.G. Chetyrkin, A.G. Grozin: Nucl. Phys. B 666, 289 (2003) 54, 55
K.G. Chetyrkin, A. Retey: Nucl. Phys. B 583, 3 (2000) 50
P. Cvitanovic: Phys. Rev. D 14, 1536 (1979); Group Theory, Part I (Nordita,
Copenhagen 1984) 48
5. S.A. Larin, J.A.M. Vermaseren: Phys. Lett. B 303, 334 (1993) 45
6. K. Melnikov, T. van Ritbergen: Nucl. Phys. B 591, 515 (2000) 54
7. T. van Ritbergen, J.A.M. Vermaseren, S.A. Larin: Phys. Lett. B 400, 379 (1997)
36
1.
2.
3.
4.
This chapter is about the rst 1/m correction to the HQET Lagrangian. It
contains two terms: the heavyquark kinetic energy and the chromomagnetic
interaction. The coecient of the kinetic term is known exactly owing to the
reparametrization invariance. The chromomagnetic interaction coecient is
obtained by matching the onshell quark scattering amplitudes in an external
chromomagnetic eld in QCD and HQET. To this end, methods for onshell
calculations in QCD and HQET, in particular in external elds, are discussed.
The dependence of the chromomagnetic interaction coecient on the renormalization scale is discussed, both within intervals between heavyavour
masses and when crossing a threshold.
60
to hyperne splittings between states which were degenerate in the innitemass limit (such as B and B ), as well as to violation of leadingorder relations
among form factors. First, we are going to discuss the simpler case of a scalar
heavy quark. We shall return to the realistic spin(1/2) case at the end of
this Section.
The only dimension5 operator in scalar HQET that does not contain D0
Therefore, the Lagrangian is
is Q
+ D2 Q.
acting on Q
+ Ck Q
+
+D2Q
+ iD0 Q
(4.1)
L=Q
2m
The additional term is the heavyquark kinetic energy. This Lagrangian leads
to the dispersion law of a free quark p0 = p0 m = Ck p2 /(2m). Therefore,
at tree level, Ck = 1. The Lagrangian (4.1) can be rewritten in a covariant
form:
Ck + 2
+
Q D Qv + ,
Lv = Q
(4.2)
v iv D Qv
2m v
where D = D v(v D). More accurately, this term should be written as
+ D2 Q
(Ck0 /(2m))0k , where 0k = Q
v0 0 v0 is the bare kineticenergy operator.
k ()k (),
This operator is related to the renormalized operator by 0k = Z
and hence the term in the Lagrangian is (Ck ()/(2m))k (), where Ck () =
k ()C 0 .
Z
k
The kineticenergy term gives the new vertices (Fig. 4.1)
Ck0 2
C0
p , i k g0 ta (p + p ) ,
2m
2m
where g = g v v .
i
Ck0 2 a b
g (t t + tb ta )g
,
2m 0
b
p
energy diagrams at the order 1/m. Each of these diagrams contains a sin for
gle kineticenergy vertex (Fig. 4.1). Let us consider the variation of
v v + v with an innitesimal v (v v = 0). There are two sources of
this variation. The expansion of the heavyquark propagators 1/(
p v + i0)
produces insertions i
pi v into each propagator in turn. Variations of the
61
quarkgluon vertices produce ig0 ta v for each vertex in turn. Now let us
k for p p + p with an innitesimal p .
consider the variation of
Nogluon kinetic vertices (Fig. 4.1) produce i(Ck0 /m)
pi p ; singlegluon kinetic vertices (Fig. 4.1) produce i(Ck0 /m)g0 ta p ; twogluon kinetic vertices
do not change. Therefore,
k
.
=2
p
v
(4.3)
k /p2 )p and /v
k /p = 2(
= (d/d)p
count
, we obtain
k
d
.
=
2
p
d
(4.4)
(4.5)
This result can also be understood in a more direct way. The momentum
p ows through the heavyquark line. Nogluon kinetic vertices are quadratic
in it; onegluon vertices are linear; twogluon vertices are independent of p .
The p2 term comes from diagrams with a nogluon kinetic vertex. Terms
linear in p vanish owing to the rotational symmetry. The coecient of p2
in a nogluon kinetic vertex is iCk /(2m). Therefore, the coecient of p2 in
the sum of all diagrams is the sum of the leadingorder HQET diagrams with
a unit operator insertion into each heavyquark propagator in turn. This sum
is just id/d,
and hence we arrive at (4.5) again.
At one loop (Fig. 4.2),
2
k
v
dd k v 2(k v + )k
2
k0 () = iCF g0
(2)d
(k 2 )(k v + )2
k k
g + (1 a0 )
k 2
2
2
g (2)
d1
I1 A 2
= 2CF 0
(4.6)
d/2
d4
(4)
(see (2.54), (3.61)).
k0
Fig. 4.2. Oneloop diagrams for
62
.
d()
2
p
p + k0 ()
2m
d
Ck0
(4.7)
k0 (0) = 0 at the twoloop level. I
In Sect. 4.3, we shall obtain (0)
= 0 and
do not know if this is true at higher orders or not. If not, these equalities can
be restored by adding a residualmass counterterm to the Lagrangian (2.9);
this does not contradict any general requirements. The mass shell is
Ck0
d()
d()
=
p2 .
(4.8)
1
1
d
2m
d
=0
=0
This is correct if Ck0 = Zk1 ()Ck () = 1. The minimal (3.3) renormalizak has to make Ck () nite; here this means Zk = 1. The
tion constant Z
kineticenergy operator is not renormalized; its anomalous dimension is zero
to all orders. The coecient of the kineticenergy operator in the HQET
Lagrangian is exactly unity,
Ck () = 1 ,
(4.9)
Ck + 2 Cm +
+
Q D Q
Q B Q
2m
2m
(4.10)
(4.11)
63
0 0
Again, it is more accurate to write this term as Cm
m = Cm ()m (),
0
0
where m = Zm ()m () is the bare chromomagnetic operator, and Cm
=
1
Zm ()Cm (). The kinetic term in (4.10) does not violate the heavyquark
+ and Q),
spin symmetry (because it contains no spin matrix between Q
while the chromomagnetic term violates it, producing hyperne splittings.
The chromomagnetic interaction gives the new vertices (Fig. 4.3)
0
Cm
g0 ta q ,
2m
0
Cm
g 2 [ta , tb ] .
2m 0
The arguments leading to (4.9) remain valid in the spin(1/2) case. The
coecient of the chromomagnetic interaction Cm () is not related to the
lowerorder term in 1/m by the reparametrization invariance, and can only
be calculated by QCD/HQET matching (Sect. 4.6).
a
q
Fig. 4.3. Chromomagneticinteraction vertices
The HQET Lagrangian at the 1/m2 level was discussed in [30, 5, 22, 8, 6],
and at the 1/m3 level in [33, 4]. At the tree level, one can easily nd as many
terms in 1/m as one wishes, without QCD/HQET matching. This can be done
by the FoldyWouthuysen transformation [26, 3]. Alternatively, one can substitute the lower components of the heavyquark eld from the equation of
motion into the QCD Lagrangian [30], or integrate them out in the functional
integral [32]. These two methods produce the same Lagrangian, which con in correction terms. Such terms can be eliminated by a
tains D0 acting on Q
suitable eld redenition, and the result of the rst method is reproduced.
dd k
= i d/2 md2(n1 +n2 ) M (n1 , n2 ) ,
D1n1 D2n2
D1 = m2 (k + mv)2 i0 ,
D2 = k 2 i0 .
(4.12)
2
After the Wick rotation k0 = ikE0 , k 2 = kE
and transformation to the
dimensionless integration momentum K = kE /m, the diagram becomes
64
k + mv
Fig. 4.4. Oneloop onshell propagator diagram
dd K
= d/2 M (n1 , n2 ) .
(K 2 2iK0 )n1 (K 2 )n2
(4.13)
Now let us compare this diagram with the oneloop HQET diagram I(n1 , n2 )
(2.27). In terms of the dimensionless Euclidean integration momentum K =
kE /(2), it has the form
dd K
= d/2 I(n1 , n2 ) .
(1 2iK0 )n1 (K 2 )n2
(4.14)
The onshell integral (4.13) can be cast into the HQET form (4.14) using
the inversion [13] K = K /K 2 . The onshell denominator K 2 2iK0 =
(1 2iK0 ) /K 2 produces the HQET denominator. The integration measure
d
d1
becomes dd K = K d1 dK d = (K )
dK d = dd K / K 2 . The nal
result is
M (n1 , n2 ) = I(n1 , d n1 n2 )
=
(d n1 2n2 ) (d/2 + n1 + n2 )
.
(n1 ) (d n1 n2 )
(4.15)
This result can also be obtained using the Feynman parametrization (2.13).
If n1,2 are integer, M (n1 , n2 ) is proportional to M1 = (1 + ), the coecient
being a rational function of d.
The inversion interchanges the UV and IR behaviours of an integral.
Therefore, the poles of (d n1 2n2 ) are IR divergences (sometimes called
onshell divergences in this case), and the poles of (d/2 + n1 + n2 ) are UV
divergences. The usual rule about the sign of d in functions applies.
There are two two generic topologies of twoloop onshell propagator
diagrams, shown in Figs. 4.5a,b. We shall start from the simpler type M ,
Fig. 4.5a:
dd k1 dd k2
d 2(d ni )
M (n1 , n2 , n3 , n4 , n5 ) ,
n1 n2 n3 n4 n5 = m
D1 D2 D3 D4 D5
D1 = m2 (k1 + mv)2 ,
D3 = k12 ,
D4 = k22 ,
D2 = m2 (k2 + mv)2 ,
D5 = (k1 k2 )2 .
(4.16)
Using inversion, we can relate this diagram to the HQET twoloop propagator
diagram. The onshell denominators D1,2 produce the HQET denominators,
65
4
2
3
5
1
2
a
(1 + (n 1)) (1 + n) (1 2n) n (1 )
.
(1 n) (1 (n + 1))
(4.18)
dd k1 dd k2
n1 n2 n3 n4 n5
D1 D2 D3 D4 D5
D1 = m2 (k1 + mv)2 ,
= d m2(d
ni )
N (n1 , n2 , n3 , n4 , n5 ) ,
D2 = m2 (k2 + mv)2 ,
D3 = m2 (k1 + k2 + mv)2 ,
D4 = k12 ,
D4 = k22 ,
(4.19)
is more dicult. The integrals N can be expressed [24, 11, 10, 23], using integration by parts, as linear combinations of M12 (Fig. 4.5c), M2 (Fig. 4.5d), and
66
3
I = 2 log 2 (3) .
2
(4.20)
r
+
)
(L
+
L
)
2
log
r
+
+
2 (1 + )
22 4
8
1
3
m 24 2 + 2 log r + 6 + O() ,
1
I1
5
= m4
+ 2(1 + r)2 L+ + 2(1 r)2 L
+
2 (1 + )
22
2
19
2 log2 r +
+ O() ,
(4.21)
2
where r = m /m,
1
1
log2 r log r log(1 + r) 2
2
6
1
1
1
= Li2 (r ) + log r log(1 + r ) ,
L+ = Li2 (r) +
1
1
log2 r + 2
2
6
1 2
1
= Li2 (1 r ) +
6
1
1
2
= Li2 (r) + log r log r log(1 r) + 2
2
3
= Li2 (r1 ) + log r1 log(1 r1 ) (r > 1) ,
1
1
L+ + L = Li2 (1 r2 ) + log2 r + 2
2
12
1
1
= Li2 (1 r2 ) + 2 .
2
12
67
L = Li2 (1 r) +
(r < 1)
(4.22)
(4.23)
where m is the onshell mass (dened as the position of the pole of the full
os
quark propagator), and m = m m0 is the mass counterterm (m0 = Zm
m).
We shall consider it not as a part of the unperturbed Lagrangian, but as a
perturbation. It produces the counterterm vertex (Fig. 4.8) i m.
(4.24)
(1 2
(p2 )) (/
p
1
.
m) + m m1 (p2 )
(4.25)
It has a pole at p2 = m2 if
m = m1 (m2 ) .
(4.26)
68
1
/+m
p
+ ,
1 2 (m2 ) 2m2 1 (m2 ) p2 m2
(4.27)
T (t) =
os
os
then Zm
= 1 T (0) and ZQ
= [1 T (0)]
(4.28)
. We have
1
1
dd k
Tr(/
v + 1) (/
T (t) = iCF g02
p+k
/ + m)
(2)d D1 (t)D2 4m
k k
g + (1 a0 )
,
D2
where D1 (t) = m2 (p + k)2 . In calculating the numerator, we can express
p k via D1 (t) and D2 , and omit terms with D1 (t), because the resulting
integrals contain no scale. Omitting also t2 and higher terms, we obtain
d2
dd k
2
1
T (t) = iCF g02
(1
t)
.
(2)d D1 (t) D2
2m2
Note that this result is gaugeindependent. Now, taking into account D1 (t) =
D1 + (D1 D2 2m2 )t + O(t2 ), we arrive at
T (t) = CF
g02 m2
d1
(1 t) + O(t2 ) .
()
d3
(4)d/2
Therefore,
os
os
Zm
= ZQ
= 1 CF
g02 m2
d1
.
()
d/2
d3
(4)
(4.29)
os
os
The equality Zm
= ZQ
is accidental, and does not hold at higher orders.
Onshell renormalization of QCD at two loops has been performed in [24,
11, 10] (see also [17] for the exact ddimensional contributions of the loops
69
os
of another massive quark). The O(g02 ) term in m = m(Zm
1), found
in (4.29), is necessary when calculating O(g04 ) diagrams containing the counterterm vertex (Fig. 4.8). Threeloop results have been obtained recently [34].
The onshell mass is gaugeinvariant to all orders [27]; the quark eld renoros
is gaugeinvariant at two loops [11] but not at three [34].
malization ZQ
The following method is used for calculation of such diagrams [12, 16]. For
two vectors a and b in ddimensional Euclidean space, the following average
over the directions of a (or b) is given by
d2
n
d
2 2 n/2 0 cos sin
n
d2
(a b) = (a b )
sin
d
0
=
((n + 1)/2)
(d/2)
(a2 b2 )n/2
(1/2)
((d + n)/2)
(4.30)
for even n (positive or negative), and 0 for odd n. In particular, in a (d = 1)dimensional space the righthand side is just (a2 b2 )n/2 , as expected. We can
use this formula for (k v)n in Minkowski scalar integrals, because they are
calculated via Wick rotation. The HQET propagator in Fig. 4.9 produces
just an additional power of k 2 , and we are left with the vacuum diagram of
Fig. 4.10. This diagram has been calculated in [35]:
(m2
k12
i0)n1 (m2
= d m2(dn1 n2 n3 )
dd k1 dd k2
n
k22 i0)n2 [(k1 k2 )2 i0] 3
70
1
3
2
Fig. 4.10. Twoloop vacuum diagram
Using this method, we see that (0)
= 0, and
d()
dd k v v (k)
2
=
iC
g
F 0
d
(2)d (k v)2 (k 2 )2
=0
(k 2 )
k2
dd k
= iCF g02
1
d
2
(2)
(k v)
(k 2 )2
g 4 m4
2(d 1)(d 6)
i
,
= CF TF 0
2 ()
(4)d
(d 2)(d 5)(d 7)
(4.32)
2(d 6)
dd k (k 2 )
g04 m4
i
=
C
T
2 ()
.
F
F
d
2
2
d
(2) (k )
(4)
(d 2)(d 5)(d 7)
(4.33)
os = 1
We nd the onshell HQET quark eld renormalization constant Z
Q
1
(d()/d)
from the requirement that the renormalized propagator
=0
Z
os (see (4.7)) behaves as the free propagator S0 near the mass
Sos = S/
Q
shell:
4
4
os = 1 CF TF g0 mi 2 () 2(d 1)(d 6)
.
(4.34)
Z
Q
(4)d
(d 2)(d 5)(d 7)
os Sos contains both UV and IR
The onshell HQET propagator S = Z
Q
divergences. The UV divergences can be eliminated by dividing it by the
Q (), so that Z
os /Z
Q () contains only IR
MS renormalization constant Z
Q
71
os
divergences. The same is true for the QCD ratio ZQ
/ZQ (). By construction,
HQET does not dier from QCD in the IR region. Therefore,
Q ()
os /Z
Z
Q
os /Z () = nite .
ZQ
Q
(4.35)
= 1 CF TF
2(d 1)(d 6)
g04 m4
i
.
2 ()
(4)d
(d 2)(d 5)(d 7)
(4.36)
k0 (0)
Fig. 4.11. Twoloop diagrams for
72
A 2 /(2a0 ), where D
= ig0 Aa ta . The ghost
gaugexing term D
0
0
term is changed correspondingly:
L=
1
qi0 (iD
/ 0 mi0 )qi0 Ga0 Ga
0
4
1
2
ca )(D ca ) .
D A0 + (D
0
0 0
2a0
(4.38)
Some vertices containing the background eld A0 dier from the ordinary
vertices. In particular, the gaugexing term contains an A0 A20 contribution,
altering the threegluon vertex (Fig. 4.12a) to
2
1
1 2 3
a1 a2 a3
+ k3 k1 + k2
g 3 1
(k2 k3 ) g
g0 f
a0
3
1
1 2
.
(4.39)
+ k1 k2 k3
g
a0
This term contains no A0 A30 contribution, so that the fourgluon vertex does
not change. The ghost term in (4.38) gives the vertices (Figs. 4.12b,c)
g0 f ab1 b2 (k1 + k2 ) ,
ig02 f a1 b1 c f a2 b2 c g 1 2 .
The terms with 1/a0 in the threegluon vertex contain k11 or k22 ; when
they are multiplied by the propagator D0 1 (k1 ) or D0 2 (k2 ), respectively,
they extract the term with a0 from the propagator (3.21), and no terms with
negative powers of the gauge parameter a0 appear.
1 a1
a
a1
a2
k1
k3
a2
2
k2
3
a3
b1
k1
k2
b2 b1
b2
c
73
k+p+q
kq
k
k+p
b
k
a
(4.40)
a
= g0
b
Fig. 4.14. Ward identities for the backgroundeld vertices
74
a
= g0
c
Fig. 4.15. Ward identity for the backgroundeld vertex function
bf (p, 0) =
(p)
p
or bf
(p, 0) =
S 1 (p)
.
p
(4.41)
The Ward identities (4.40), (4.41) for the backgroundeld vertex are very
simple, and are exactly the same as in QED (or in the heavyelectron eective
theory, Sect. 3.5). Multiplying the ordinary threegluon vertex by q gives
an identity which, in addition to the simple dierence shown in Fig. 4.14b,
has additional ghost terms (see, e.g., Fig. 3 in [18]). Therefore, the Ward
identities for the ordinary quarkgluon vertex function are more complicated
than (4.40), (4.41).
The renormalized background eld is related to the bare eld by A0 =
1/2
75
os 0 2
where the renormalized form factors are F1 (q 2 ) = ZQ
F1 (q ), Gm (q 2 ) =
os 0
2
ZQ Gm (q ). The form factors can be singled out by the appropriate projectors.
Let us rewrite (4.42) as
a
u
(p )bf
t u(p) = u
(p )
Fi Ti ta u(p) ,
F1 = F10 (q 2 ) ,
F2 = G0m (q 2 ) ,
T1 =
(p + p )
,
2m
T2 =
[/
q, ]
,
4m
v+
+1 ,
+
/v +
1 q 2 /(4m2 )
2m
4m
m
1
1
G0m (q 2 ) =
Tr bf (/
v + 1)
2(d 2) 4
1
q [/
q, ]
/q
+
1
v
+
.
+
v
/
+
1 q 2 /(4m2 )
2m
q2
m
(4.43)
F10 (q 2 ) =
= 0 + 1
bf
q
+ ,
m
v , q q =
q =
d
v
g
v
.
1
2m
d1
4m2
4m2
Thus we obtain
1
Tr 0 (/
v + 1)v + O(q 2 /m2 ) ,
4
1 1
Tr 0 (/
G0m (q 2 ) =
v + 1)( v )
d14
1
2
Tr 1 (/
+
v + 1)( g + v v )
(d 1)(d 2) 4
F10 (q 2 ) =
+ O(q 2 /m2 ) .
(4.44)
76
The Dirac form factor at q = 0 is unity, owing to the Ward identity (4.41):
1
os
v + 1)v
1 + Tr bf (mv, 0)(/
F1 (0) = ZQ
4
1
os
= ZQ 1 v
Tr (p)(/
v + 1)
p 4
p=mv
os
ZQ
[1 T (0)] = 1
(4.45)
(see (4.28)). The total colour charge of the heavy quark is not changed by
renormalization.
The heavyquark chromomagnetic moment g = Gm (0) (4.44) can be
calculated as follows. First, we apply the projector in (4.44) to the integrand of each diagram, dierentiating all qdependent propagators and vertices (Fig. 4.13), and obtain the bare 0g via scalar integrals (4.15). Then we
os
multiply 0g by ZQ
(4.29). The oneloop result is [19]
g = 1 +
g02 m2 ()
2(d 4)(d 5)CF (d2 8d + 14)CA +
d/2
2(d 3)
(4)
(4.46)
ig0 ta bf
(, q), where and = + q v are the residual energies of the
in the problem, v and q , and hence the vertex function has the structure
bf = s (, , q 2 )v + a (, , q 2 )q ,
s (, , q 2 ) = s ( , , q 2 ) , a (, , q 2 ) = a ( , , q 2 ) .
(4.47)
This function has been calculated, at one loop, in [18]. It obeys the Ward
identity
)
(
bf (, q)q = s ( ) + a q 2 = ()
or (, q)q = S1 ( ) S1 () ,
bf
(4.48)
77
or, for q 0,
d()
s (, , 0) =
.
d
(4.49)
=0
(4.50)
The total colour charge of the static quark is not changed by renormalization.
Now we shall make a short digression, and derive the oneloop renormalization of the QCD coupling constant (3.6) from the ordinary HQET
quarkgluon vertex (, q). When the bare vertex is expressed via the
r , where Z
is a minrenormalized quantities s and a, it should become Z
imal (3.3) renormalization constant, and the renormalized vertex r is nite
1/2
at 0. When g0 = g r Z Z
is multiplied by the externalleg renor1/2
Q Z , it should give a nite matrix element. Therefore,
malization factors Z
A
Q )2 Z 1 . At one loop (Fig. 4.16), the HQET vertex has been
Z
Z = (Z
A
calculated in [18]. It obeys a more complicated Ward identity than does the
k
b
78
CA
k k
dd k
i
i
ig0 v
v = i
ig0 v
g (1 a0 ) 2
2g0
(2)d
kv
k2
k
k k
k g + k g
g (1 a0 ) 2
v .
(i)g0
k
a0
Here 1/a0 cancels, as explained after (4.39):
1
(k v)2
dd k
)
1
(1
a
.
v = ig02 CA
0
(2)d (k 2 )2
k2
Averaging over the directions of v, we can replace (k v)2 by k 2 /d. The UV
1/ pole of the resulting integral is
1
i
i
i
dd k
12
2 =
,
=
kE
dkE =
(2)d (k 2 )2 UV
8 2
(4)2
(4)2
(4.51)
where is the IR cuto, and terms regular at 0 are omitted. Therefore,
Q Z
= 1 + CA a + 3 s +
Z
4 4
ZQ = 1 in the abelian case (Sect. 3.5).
The CF terms cancel here because Z
Using (3.31), we derive Z (see (3.10) and (3.6)); the gauge dependence
cancels, as expected.
There are two kinds of 1/m corrections to the scattering amplitude: diagrams with a single kinetic vertex, and those with a single chromomagnetic vertex. Let us forget for a while that we have obtained the result
Ck = 1 (4.9), and denote the sum of oneparticleirreducible bare vertex
diagrams in the backgroundeld method containing a single kineticenergy
p, q), where p = v + p and p = p + q =
vertex by ig0 ta [Ck0 /(2m)]k (
v + p are the initial and nal residual momenta of the heavy quark;
k = (p + p ) + k (
p, q). The dependence on p and p comes only from
the kineticenergy vertex (Sect. 4.1), which is at most quadratic in them.
Therefore, the vertex function has the structure
k (
p, q) = ks (, , q 2 )(p + p )
2
+ k1 (, , q 2 )p2 + k1 ( , , q 2 )p2
+ k0 (, , q ) v
2
+ k3 (, , q 2 )p2 k3 ( , , q 2 )p2
+
(,
,
q
)
q ,
k2
ks (, ,q 2 ) = ks ( , , q 2 ) ,
k0 (, ,q 2 ) = k0 ( , , q 2 ) ,
k2 (, ,q 2 ) = k2 ( , , q 2 ) .
(4.52)
79
a (, , q 2 )
k3 (, , q 2 ) =
.
(4.53)
(4.54)
g = 1 + CA TF 0
()
(4)d
(d 2)(d 5)(d 7)
ig0 ta
where the sum is over all massive avours (except the heavy avour of our
HQET, of course).
80
(mv + p)
p)uv (
p) .
uv (
m
g
(4.62)
In the oneloop approximation, reexpressing (4.46) via s () (3.4) and expanding in , we obtain
s () 2L
1
m
Z
e
()Cm () = 1 +
2CF +
4
m
L = log .
81
1
+ 2 CA ,
(4.63)
(4.64)
s (m)
+
4
(4.65)
m = 2CA
s
+ ,
4
Cm (m) = 1 + 2(CF + CA )
s (m)
+
4
(4.66)
m = 2CA
2
4
s
s
+ CA (17CA 13TFnf )
+
4 9
4
(4.67)
(4.68)
82
s (m)
s (m)
s ()
=
+
1 + 20 L
4
4
4
in
m (s ()) = 0
s ()
+ 1
4
s ()
4
2
+ ,
Cm () = 1 + (0 L + C1 )
If [s /(4)]L 1, we have to solve the renormalizationgroup equation (4.68) in another way. Dividing this equation by d log s ()/d log (3.5)
(at = 0), we obtain
d log Cm ()
m (s )
= 0.
+
d log s
2(s )
The solution of this equation is
s ()
m (s ) ds
Cm () = Cm (m) exp
.
2(s ) s
(4.70)
s (m)
(s )
0
2(s ) 20
ds
0
=1+
s
20
1
1
0
0
s
+
4
(4.72)
83
K0 (s ) = 1 ,
K1 +2 (s ) = K1 (s )K2 (s ) .
m0 /(20 )
s ()
Km (s ())
s (m)
m0 /(20 )
s ()
s (m)
m1
m0
s (m)
1 s () s (m)
+ ,
1 + C1
4
20
m0
0
4
(4.73)
where
Cm (m) = Cm (m)Km (s (m))
m1
s (m)
m0
1
= 1 + C1 +
+
20
m0
0
4
Note that this function of and m can be presented as a product of a function
of m,
Cm (m)s (m)m0 /(20 ) ,
and a function of ,
Km (s ())s ()m0 /(20 ) .
The fractional power of s ()/s (m) in (4.73) contains all leading logarithms
(s L)n in the perturbative series (4.69); the correction inside the brackets
contains the subleading logarithms. We cannot use the C2 term here until we
know
m2 .
The largest term missing in (4.69) is [s /(4)]3 L3 . The largest term missing in (4.71) is C2 [s /(4)]2 . Comparing these errors, we can estimate the
value of L at which (4.71) becomes a better approximation than (4.69).
The most obvious eect of the chromomagnetic interaction is the hyperne
BB splitting:
1
2
2
mB mB =
Cm ()G () + O
,
(4.74)
3mb
m2b
where 2G () is the matrix element of the chromomagnetic interaction operator. The product Cm ()2G () is, of course, independent, and hence
2G ()
2G
s ()
4
m0 /(20 )
Km (s ()) .
(4.75)
84
The quantity
2G is independent, and hence is equal to 2MS times some
number; we obtain
mB mB =
2
3mb
s (mb )
4
m0 /(20 )
2G + O
Cm (mb )
1
m2b
.
(4.76)
20
m0
0
4
MS
.
(4.77)
+O
mc,b
In the interval between mc and mb , the relevant number of avours is nf = 4:
m2B m2B
=
m2D m2D
s (mb )
s (mc )
9/25
1
7921 s (mc ) s (mb )
+
3750
The experimental value of this ratio is 0.89. The leading logarithmic approximation gives 0.84; the nexttoleading correction reduces this result by 9%,
giving 0.76. The agreement is quite good, taking into account the fact that
the 1/mc correction may be rather large.
85
In particular, the renormalized light elds of full QCD are related to those
of the lowenergy theory by
qi () = q1/2 (, )qi ( ) ,
1/2
A () = A (, )A ( ) ,
(4.78)
up to 1/mc suppressed terms. The coupling constant and gaugexing parameter in the two theories are related by
g() = 1/2 (, )g ( ) ,
a() = A (, )a ( )
(4.79)
(we shall see in a moment why the last coecient is A ). It is more convenient
to calculate the coecients i0 which relate the bare elds and parameters in
the two theories. After that, it is easy to nd the renormalized coecients:
Zq ( ) 0
,
Zq () q
Z ( ) 0
c (, ) = c
,
Zc () c
q (, ) =
( ) 0
ZA
,
ZA () A
Z ( ) 0
(, ) =
.
Z ()
A (, ) =
(4.80)
The transverse part of the bare gluon propagator (3.25) near the mass
shell in full QCD is
D (p2 ) =
os
ZA
,
p2
os
ZA
=
1
.
1 (0)
(4.81)
os
= 1, because all loop diagrams for (0)
In the lowenergy theory, ZA
0
contain no scale. Therefore, the matching coecient A
in the relation A0 =
0
A A0 is
0
=
A
os
ZA
1
.
=
os
ZA
1 (0)
(4.82)
(p)
.
(0) =
2d(d 1) p p
p=0
In the oneloop approximation (Fig. 4.17a), (0) is a combination of oneloop
vacuum integrals (2.16); we obtain
4 g 2 m2
(0) = TF 0 cd/2 () +
3 (4)
(4.83)
86
Therefore,
4 g 2 m2
0
= 1 TF 0 cd/2 () +
A
3 (4)
It is most natural to calculate i (mc , mc ), which contain no large logarithms.
The ratio of the renormalization constants for the gluon eld is
ZA (mc )
1
4 s (mc )
s (mc )
+ = 1 TF
+ ,
(m ) = 1 2 A0 4
ZA
3
4
c
where A0 = (8/3)TF is the contribution of a single avour to the oneloop anomalous dimension (3.33) of the gluon eld. Finally, we obtain the
renormalized decoupling constant
A (mc , mc ) = 1 + O(2s ) .
The dependence on , can be easily restored using the renormalizationgroup. It is not too dicult to calculate the twoloop diagrams (Figs. 4.17b,c).
They reduce to combinations of twoloop massive vacuum integrals (4.31),
and are proportional to 2 (). Using also the twoloop anomalous dimension
of the gluon eld, one can obtain [29]
2
s (mc )
13
+
(4.84)
A (mc , mc ) = 1 (4CF CA )TF
12
4
Similarly, for the lightquark eld (see (3.46)),
q0 =
Zqos
1
.
=
Zqos
1 V (0)
(4.85)
The only twoloop diagram contributing is Fig. 4.18. Expanding the quark
selfenergy (3.45) at m0 = 0 in p up to linear terms, we reduce it to the
integral (4.33), similarly to (4.32) [15, 25]:
dd k (k 2 )
(d 1)(d 4)
V (0) = iCF g02
,
d
(2)d (k 2 )2
g 4 m4
2(d 1)(d 4)(d 6)
1
q0 = [1 V (0)] = 1 + CF TF 0 c d 2 ()
.
(4)
d(d 2)(d 5)(d 7)
(4.86)
87
4
where 0 = (4/3)TF , A0 = (8/3)TF, and q1 = 4CF TF are the
singleavour contributions. Finally, we obtain [15]
2
s (mc )
5
q (mc , mc ) = 1 CF TF
+
(4.87)
6
4
If we consider bquark HQET instead of QCD, nothing changes. When all
characteristic (residual) momenta become much less than mc , cquark loops
shrink to a point. From (4.34), we obtain
0
=
Q
os
Z
2(d 1)(d 6)
g 4 m4
Q
.
= 1 CF TF 0 c d 2 ()
os
(4)
(d
2)(d 5)(d 7)
Z
(4.88)
s (mc )
4
2
+
(4.89)
The decoupling relation for the coupling constant can be derived by considering any vertex in the theory. We shall use the HQET heavyquarkgluon
vertex. Let g0 ta (, q) be the sum of bare oneparticleirreducible heavyquarkheavyquarkgluon vertex diagrams, not including the external propagators. For this vertex,
1
1/2
0
0
A
g0 ,
g0 = Q
88
Fig. 4.19. Twoloop onshell HQET quarkgluon vertex (the diagram b implies
also the mirrorsymmetric one; the diagram c has two orientations of the quark
loop)
1
,
2
0 0
Q
1
Z 0
Z Z 1
= A
.
Z
Z ZA A 0 2
Q
bf (, q) is the backgroundeld vertex (see Sect. 4.5). Using the identities
+ q v)
of Fig. 4.14, it is easy to demonstrate that bf (, q)q = ()
(
and, in particular,
d()
v .
bf (, 0) =
d
At = 0, only the diagram of Fig. 4.9 with a cquark loop contributes to
0
and bf (0, 0)
the heavyquark selfenergy ().
Therefore, Q
= 1 + ,
cancels (here (0, 0) = v ). The term is given by the diagrams
of Fig. 4.19b, where only the 1/a0 terms in (4.39) are substituted for the
threegluon vertex; the two diagrams contribute equally. This diagram can
be easily reduced to (4.33) by averaging over the directions of k:
= CA TF
g04 m4
2(d 1)(d 6)
c
.
2 ()
(4)d
d(d 2)(d 5)(d 7)
(4.90)
References
(mc , mc ) = 1 +
1
(39CF 32CA ) TF
9
s (mc )
4
89
2
+
(4.91)
This means that at > mc , we have s (), whose running is given by the
(s ) function with nf = nl + 1 avours; at < mc , we have s (), whose
running is given by (s ) with nl avours; at = mc ,
s (mc ) = (mc , mc )s (mc ) .
Finally, we shall discuss the decoupling of cquark loops in the bquark
chromomagneticinteraction constant. Both HQET with cloops and the effective lowenergy theory must give identical onshell scattering amplitudes,
up to corrections suppressed by powers of 1/mc . Therefore, from (4.62) we
have
Cm () =
g
Zm ()
C () .
()
g m
Zm
(4.92)
References
1. L.F. Abbot: Nucl. Phys. B 185, 189 (1981); Acta Phys. Pol. B 13, 33 (1982)
71
2. G. Amor
os, M. Beneke, M. Neubert: Phys. Lett. B 401, 81 (1997) 81, 84
3. S. Balk, J.G. K
orner, D. Pirjol: Nucl. Phys. B 428, 499 (1994) 63
90
4. C. Balzereit: Phys. Rev. D 59, 034006 (1999); Phys. Rev. D 59, 094015 (1999)
63
5. C. Balzereit, T. Ohl: Phys. Lett. B 386, 335 (1996) 63
6. C. Bauer, A.V. Manohar: Phys. Rev. D 57, 337 (1998) 63
7. J.D. Bjorken, S.D. Drell: Relativistic Quantum Mechanics (McGraw Hill, New
York 1964) 80
8. B. Blok, J.G. K
orner, D. Pirjol, J.C. Rojas: Nucl. Phys. B 496, 358 (1997) 63
9. D.J. Broadhurst: Z. Phys. C 47, 115 (1990) 66
10. D.J. Broadhurst: Z. Phys. C 54, 599 (1992) 65, 66, 68
11. D.J. Broadhurst, N. Gray, K. Schilcher: Z. Phys. C 52, 111 (1991) 65, 68, 69,
71
12. D.J. Broadhurst, A.G. Grozin: Phys. Rev. D 52, 4082 (1995) 69
13. D.J. Broadhurst, A.G. Grozin: Multiloop Calculations in Heavy Quark Eective Theory, in New Computing Techniques in Physics Research IV, ed. by
B. Denby, D. PerretGallix (World Scientic, Singapore 1995) p. 217 64, 65
14. Y.Q. Chen: Phys. Lett. B 317, 421 (1993) 59
15. K.G. Chetyrkin, B.A. Kniehl, M. Steinhauser: Nucl. Phys. B 510, 61 (1998)
84, 86, 87
16. A. Czarnecki, A.G. Grozin: Phys. Lett. B 405, 142 (1997) 69, 76, 79, 81
17. A.I. Davydychev, A.G. Grozin: Phys. Rev. D 59, 054023 (1999) 66, 68, 71,
76
18. A.I. Davydychev, A.G. Grozin: Eur. Phys. J. C 20, 333 (2001) 74, 76, 77
19. E. Eichten, B. Hill: Phys. Lett. B 243, 427 (1990) 62, 76, 81
20. A.F. Falk, B. Grinstein, M.E. Luke: Nucl. Phys. B 357, 185 (1991) 62
21. M. Finkemeier, H. Georgi, M. McIrvin: Phys. Rev. D 55, 6933 (1997) 59
22. M. Finkemeier, M. McIrvin: Phys. Rev. D 55, 377 (1997) 63
23. J. Fleischer, O.V. Tarasov: Phys. Lett. B 283, 129 (1992); Comput. Phys.
Commun. 71, 193 (1992) 65
24. N. Gray, D.J. Broadhurst, W. Grafe, K. Schilcher: Z. Phys. C 48, 673 (1990)
65, 68
25. A.G. Grozin: Phys. Lett. B 445, 165 (1998) 86, 87
26. J.G. K
orner, G. Thompson: Phys. Lett. B 264, 185 (1991) 59, 63
27. A.S. Kronfeld: Phys. Rev. D 58, 051501 (1998) 69
28. S. Laporta, E. Remiddi: Phys. Lett. B 379, 283 (1996) 67
29. S.A. Larin, T. van Ritbergen, J.A.M. Vermaseren: Nucl. Phys. B 438, 278
(1995) 86, 88
30. C.L.Y. Lee: Aspects of the Heavy Quark Eective Field Theory at Subleading
Order, Preprint CALT681663 (CalTech, Pasadena 1991); revised (1997) 63
31. M.E. Luke, A.V. Manohar: Phys. Lett. B 286, 348 (1992) 59
32. T. Mannel, W. Roberts, Z. Ryzak: Nucl. Phys. B 368, 204 (1992) 63
33. A.V. Manohar: Phys. Rev. D 56, 230 (1997) 63
34. K. Melnikov, T. van Ritbergen: Phys. Lett. B 482, 99 (2000); Nucl. Phys. B
591, 515 (2000) 67, 68, 69, 71
35. A.A. Vladimirov: Theor. Math. Phys. 43, 417 (1980) 69
5 HeavyLight Currents
Until now, we have considered the Lagrangian, its Feynman rules, and the
Green functions and scattering amplitudes following from them. It is often
also important to consider composite operators which do not appear in the
Lagrangian products of elds at coincident points. This division is not
strict. Some operators, such as the electromagnetic and weak currents, are
interesting precisely because they appear (multiplied by electroweak gauge
elds) in the Lagrangian of a wider theory incorporating QCD the Standard
Model. Other composite operators are interesting in their own right, e.g., their
space integrals can be generators of exact or approximate symmetries, or they
appear in operator product expansions, etc.
In this chapter, as well as in the following two, we shall discuss bilinear
quark currents in QCD and HQET, which have numerous applications.
= [1 n ]
(5.1)
(5.2)
where the Zjn are minimal (3.3) renormalization constants. The dependence
of jn () is determined by the renormalizationgroup equation
d
+ jn (s ()) jn () = 0 ,
(5.3)
d log
A. Grozin: Heavy Quark Eective Theory, STMP 201, 91119 (2004)
c SpringerVerlag Berlin Heidelberg 2004
92
5 HeavyLight Currents
where
jn (s ) =
d log Zjn
d log
(5.4)
renormalization factors Zq , it should give a nite matrix element. Therefore, Zjn = Zq Z n . The UV divergences of n (p, p ) do not depend on the
quark masses and the external momenta. Therefore, we may assume that all
quarks are massless, and set p = p = 0. An IR cuto is then necessary in
order to avoid IR 1/ terms.
p
p
a
k
1
d
(1 a0 ) ,
= iCF g02
(2)d (k 2 )3
d
where the averaging /k /
k k 2 /d has been used. In d = 4 2 dimensions,
= 2h(d) ,
where, for the antisymmetrized product (5.1),
d
n .
h(d) = (1)n
2
Using the UV divergence of the integral (4.51), we obtain
(5.5)
(5.6)
Z n = 1 + CF
s
(n 2)2 1 + a .
4
93
(5.7)
s
+
4
(5.8)
The twoloop anomalous dimension for generic n can be derived by calculating n (p, 0) or n (p, p) at two loops, using the methods of Sect. 2.4. Historically, it was rst obtained as a byproduct of a more dicult QCD/HQET
matching calculation [4] (see Sect. 5.6):
s
jn = 2(n 1)(n 3)CF
4
1
s
1
2
2
5(n 2) 19 CF
3(n 2) 19 CA
1+
2
3
4
2
s
1
+
(5.9)
(n 1)(n 15)CF 0
3
4
The corresponding threeloop result has been derived recently [10].
It is not an accident that our results contain a factor n 1. The vector
current has j1 = 0 to all orders. The integral of q 0 q over all space is
the number of quarks minus the number of antiquarks of the avour being
considered. It is an integer, and cannot depend on . The same argument
holds for vector currents with diagonal avour matrices having Tr = 0,
and hence also for q q (the integral of the 0th component of this current
is the generator of the avour symmetry group which replaces q by q ; its
normalization is xed by the commutation relations of this group, and thus
cannot depend on ).
We can understand how this happens using the Ward identity for 1 (p, p ).
Starting from each diagram for (p), we can obtain a set of diagrams for
1 (p, p ) by inserting the current vertex into each possible place. This vertex
must be on the open quark line going through the diagram, because the current changes the quark avour. After contracting with q , we have exactly
the same situation as in Fig. 3.16, and
1 (p, p + q)q = (p) (p + q) ,
(5.10)
In particular, at q 0
1 (p, p) =
(p)
p
or 1 (p, p) =
S 1 (p)
.
p
(5.11)
94
5 HeavyLight Currents
(5.12)
(5.15)
os
Zm
m.
Zm ()
(5.17)
os
os
The UV divergences in Zm
and Zm () cancel; Zm
is IR nite (unlike Zqos ).
Substituting (4.29), we nd at one loop
m()
2
s
= 1 6CF
+
log
+
(5.18)
m
4
m 3
95
s ()
4
m0 /(20 )
Km (s ()) .
(5.20)
Currents (5.1) with n > 4 do not exist in fourdimensional space. Any matrix element involving such a current contains a matrix trace that vanishes
at 0. However, if it is multiplied by an integral containing 1/, a nite
contribution may result. Such operators are called evanescent; we shall not
discuss them here. There are ve nonevanescent currents with 0 n 4.
The currents with n = 4, 3 can be obtained from those with n = 0, 1 by
multiplying them by the t HooftVeltman 5 , discussed in the next section.
(5.21)
96
5 HeavyLight Currents
Tr 5 = d Tr 5 = Tr 5
= (d 8) Tr 5 (d 4) Tr 5 = 0 ,
where we have used = (d 8) + terms with fewer matrices. We have learned that if d = 4 then Tr 5 = 0. Assuming
d = 4, we can show that (d 6) Tr 5 = 0, and so on. All
traces vanish if d is not an even integer. Therefore, an anticommuting 5 is
not usable in dimensional regularization.
A way out was proposed by t Hooft and Veltman. Let us split our ddimensional spacetime into a fourdimensional subspace and the orthogonal
(d 4)dimensional subspace, and dene
i
= i 0 1 2 3 ,
4!
5HV + 5HV = 0 , 5HV 5HV = 0 ,
5HV =
(5.22)
M (p, p )p = 0 .
k + p
k+p
p
k
a
q = p p ,
(5.23)
97
(5.24)
2
(p, p )q = iTF g0
Tr S0 (k + p )/
q 5HV S0 (k + p) S0 (k)
M
(2)d
+ (p p , ) .
Taking into account
S0 (k + p )/
q 5HV S0 (k + p)
k 5HV S0 (k + p)
= 5HV S0 (k + p) + S0 (k + p )5HV 2S0 (k + p )/
at m = 0, where k and k are the components of k in the physical and nonphysical subspaces, we obtain
dd k
M (p, p )q = iTF g02
Tr 5HV S0 (k + p) + S0 (k + p )5HV
d
(2)
2S0 (k + p )/
k 5HV S0 (k + p) S0 (k) + (p p , ) .
The rst two traces vanish, because they contain only one of the two vectors
p, p (this is the basis of the naive proof of (5.24)). The last term contains
k, and its numerator should vanish at 0. But the UV divergence of the
integral can give 1/, producing a nite result. The trace in the numerator
of the last term is
k 5HV (/
k+p
/) /k = Tr p
/ /k 5HV /p /k = 4ik2 p p ,
Tr(/
k+p
/ )/
because we have to retain both p and p , and k
/ anticommutes with the
d4 2
k .
d
We need only the UV 1/ pole of the integral, so we may set the external
momenta to zero and use (4.51). The two diagrams contribute equally, and
we arrive at the anomaly in the axial Ward identity,
M (p, p )q = 8TF
s
i
p p .
4
(5.25)
If the quark is massive, the normal contribution from (5.23) should be added.
98
5 HeavyLight Currents
(5.26)
() = ZA (s ())jHV
() .
jAC
(5.27)
Li Ri = 1 1 , /
p /p/p2 , .
(5.28)
(5.29)
i, =
i R
yi = L
Mij xj ,
j
Mij =
99
1
i Lj 1 Tr R
i Rj ,
Tr L
4
4
i R
i = 1 1 , /
L
p /p /p2 , .
Solving the linear system, we can express the coecients xi via the double
traces yi :
(d 2)(3d 2) 0 (d 2)
x1
y1
1
x2 =
0
2d
2 y2 .
2(d 1)(d 2)
x3
y3
(d 2)
2
1
(5.30)
p:
/
= 1 .
(5.31)
p /p /p2 + (d + 2 4h + 4h2 ) .
+ 4h(d 2h) /
Now we can apply the projector P to the integrand of the oneloop diagram
of Fig. 5.1b, and reduce it to a scalar expression quadratic in h. Then the
integral can be easily calculated (Sect. 2.2), once, and for all currents.
For any current (5.1) with satisfying (5.31) and (5.32), (p, p) =
a (p2 ) and (p, 0) = b (p2 ), where a,b (p2 ) = 1 + a,b (p2 ). At one loop,
d
g 2 (p2 ) G1
1 a0 h 2
a = CF 0
+
(1
a
)
+
a
h
,
(5.33)
2
0
0
d4
2
(4)d/2 d 3
g02 (p2 ) G1
1 a0 h 2
+ 1 a0 h ,
(5.34)
b = CF
2
d4
(4)d/2 d 3
100
5 HeavyLight Currents
dV (p2 )
pV (p2 ) = V (p2 ) 2/pp
/
.
p
dp2
dV (p2 )
= (d 3)V (p2 ) ,
d(p2 )
ZP (s ()) =
ZP,AC () a,b (h = 2 )
.
ZP,HV () a,b (h = 2 + )
(5.35)
At one loop (but not beyond), ZP,AC () = Zj0 () and ZP,HV () = Zj4 ()
coincide (see (5.8)). Therefore, ZP = 1+a,b (h = 2)a,b (h = 2+). Only
the 1/ term contributes, and we obtain, either from (5.33) or from (5.34),
ZP (s ) = 1 8CF
s
+
4
(5.36)
The result for ZA (s ) can be derived either from the longitudinal components
of the axial currents or from the transverse components (see (5.27)):
ZA (s ()) =
ZA,AC () a,b (h = 1 )
ZA,AC () a,b (h = 1 + )
=
,
ZA,HV () a,b (h = 1 )
ZA,HV () a,b (h = 1 + )
(5.37)
ZA (s ) = 1 4CF
s
+
4
101
(5.38)
where j1 = 0 .
(5.39)
Therefore,
ZP (s ) = Kj0 j4 (s ) ,
ZA (s ) = Kj1 j3 (s )
(5.40)
(see (4.72)). The oneloop results (5.36) and (5.38) can be reproduced using
the twoloop anomalous dimensions (5.9). Now we see the reason why the
last term in (5.9), which is not symmetric with respect to n 4 n, is
proportional to 0 . As ZP,A (s ) can be obtained solely from the anomalous
dimensions, they are determined by the UV behaviour of the matrix elements,
and cannot depend on masses and external momenta.
The twoloop results
2
2
s
s
+ CF (CA + 4TF nf )
+ ,
(5.41)
4 9
4
1
s
s 2
ZA (s ) = 1 4CF
+ CF (198CF 107CA + 4TF nf )
+
4 9
4
(5.42)
ZP (s ) = 1 8CF
can be obtained either from the twoloop matrix elements (p, p) (or (p, 0)),
which can be calculated by the methods of Sect. 2.4, or from the threeloop
anomalous dimensions. The threeloop results have been calculated [14] from
the matrix elements only.
() = ZT (s ())jHV
()
Why have we not discussed a similar relation jAC
AC
HV
between the tensor currents j0AC = q0 5 q0 and j0HV = q0 5 q0 ?
has the same anomalous dimenThe reason is the following. The current jAC
AC
sion as the current j2 without 5 . Multiplication of by 5HV is merely
has the
a spacetime transformation, e.g., 5HV 01 = i 23 , and hence jHV
same anomalous dimension, too. Therefore, ZT (s ()) cannot depend on ,
and we conclude that ZT (s ) = 1. This result can also be obtained by the
following argument. Let us choose the axis 3 along p. Then = 5AC 01
has h = , = 5HV 01 has h = , and ZT = a,b (h = )/a,b (h = ).
On the other hand, = 5AC 23 has h = , = 5HV 23 has h = , and
ZT = a,b (h = )/a,b (h = ). Therefore, ZT = 1.
102
5 HeavyLight Currents
0
j0 = q0 Q
(5.43)
103
j = 3CF
4
Twoloop vertex diagrams (Figs. 5.3ck) can be obtained by inserting
the heavylight quark vertex in all possible places along the quark line in
the quark selfenergy diagrams (see Fig. 3.13; the blob in Fig. 5.3c means
the oneloop gluon selfenergy shown in Fig. 3.2). To nd the anomalous
0) (with a massless light quark),
dimension, it is enough to calculate (,
which is IR nite. This quantity contains an even number of matrices, and
can depend only on 0 ; hence, it is just a scalar function of , and we may
0) (Figs. 5.3ck)
take (1/4) Tr of the integrand. The twoloop result for (,
can be calculated using the methods of Sect. 2.5. This has been done in [3];
the result leads to
s
j = 3CF
4
2 2 49
s 2
10
8 2 5
.
+ CF +
CF +
CA + TF nl
3
2
3
6
3
4
(5.45)
104
5 HeavyLight Currents
without this avour. This section continues the discussion of decoupling given
in Sect. 4.7. The currents in the full theory can be expanded in 1/mc , where
the coecients are operators of the eective theory with the appropriate
quantum numbers and dimensionalities:
jn () = jn (, )jn ( ) + O(1/mc ) .
(5.46)
The meaning of the operator expansion (5.46) is that onshell matrix elements
of jn () with external momenta pi mc , after expansion in pi /mc to some
order, coincide with the corresponding matrix elements of the righthand
side. It is natural to match the currents at = = mc , because jn (mc , mc )
contains no large logarithms. Currents at arbitrary normalization scales are
related by
jn0 /(20 )
s ()
jn (, ) = jn
Kjn (s ())
s (mc )
jn0
/(20 )
s ( )
K
(5.47)
(s ( )) ,
jn
s (mc )
where
jn = jn (mc , mc )Kjn (s (mc ))K jn
(s (mc )) ,
105
Fig. 5.4. Twoloop onshell matrix element of a QCD bilinear quark current
n (0, 0) =
iCF g02
(2h)2
1
d
dd k (k 2 )
,
(2)d (k 2 )2
g04 m4
2(d 6)
c
2 ()
(4)d
(d 2)(d 5)(d 7)
(d 2n)2
1 .
n (0, 0) = 1 CF TF
(5.49)
s (mc )
4
2
1
= 1 + CF TF (n 1) [6(n 3) (n 15)]
9
s (mc )
4
2
,
where 0 and jn1 are the singleavour contributions. Using also (4.86),
we arrive at the nal result [12]
2
s (mc )
1
+ (5.50)
jn (mc , mc ) = 1 + CF TF (n 1)(85n 267)
54
4
The vector current has j1 = 1 to all orders. For the vector current with
a diagonal avour matrix , the integral of its 0th component is an integer
the dierence between the numbers of quarks and antiquarks weighted by
the diagonal elements of . This dierence is the same in full QCD and in
the lowenergy eective theory. The same holds for nondiagonal by avour
symmetry. We can also see this explicitly. Multiplying the Ward identity
(p)
= (1 V (0))
1 (0, 0) =
p p=0
1
106
5 HeavyLight Currents
The decoupling for the scalar current q q is closely related to that for
the MS lightquark mass. The quark propagators (3.46) in full QCD and the
lowenergy eective theory should be related by S(p) = q0 S (p). Therefore,
1 + S (p2 )
1 + S (p2 )
m
=
0
(p2 ) m0 .
1 V (p2 )
1 V
Let us dene
0
m
=
m0
,
m0
m (, ) =
m()
0 Zm ( )
=
;
m
m ( )
Zm ()
then
0
m
=
1 V (p2 ) 1 + S (p2 )
(p2 ) 1 + (p2 ) .
1 V
S
Zqos 0 (0, 0)
.
Zqos 0 (0, 0)
(5.51)
89
CF TF
18
s (mc )
4
2
+
(5.52)
This means that at > mc , the running of the lightquark MS masses m()
is governed by the anomalous dimension m (s ) (5.16) and (s ), with nf =
nl + 1 avours; at < mc , the running of m () is governed by m
(s ) and
(s ), with nl avours; and at = mc
m(mc ) = m (mc , mc )m (mc ) .
The currents j4 and j3 dier from j0 and j1 by insertion of 5HV . They
are related to those containing 5AC by (5.26) and (5.27). Inserting 5AC does
not change the decoupling coecient. Therefore,
j4 = j0
ZP
,
ZP
j3 = j1
ZA
,
ZA
(5.53)
107
where ZP,A are given by (5.41) and (5.42) at two loops, and ZP,A
contain nl
instead of nf = nl + 1.
The heavylight current
j in bquark HQET can be considered similarly.
Instead of (5.48), we now have
os 1/2 os 1/2
ZQ
Zq
Zj () (0, 0)
,
(5.54)
j (, ) =
os
os
Zq
Z
Zj ( ) (0, 0)
Q
where (, p) is the proper vertex of the current. Only diagrams with cquark
loops can contribute to (0, 0). For a massless light quark, (0, 0) contains
an even number of matrices, and can depend only on 0 . Therefore, it is
scalar, and we may take (1/4) Tr of an integrand. The only relevant twoloop
diagram is shown in Fig. 5.5:
0) = iCF g02
(0,
dd k (1/4) Tr /k v (k 2 g k k )(k 2 )
= 0.
(2)d
(k 2 )3 k0
(5.55)
(5.56)
Fig. 5.5. Twoloop onshell matrix element of the HQET heavylight current
108
5 HeavyLight Currents
C (, ) = C
s ()
s (m)
s ( )
s (m)
jn0 /(20 )
j0 /(20 )
Kjn (s ())
K
j (s ( )) ,
(5.58)
where
C = C (m, m)Kjn (s (m))K j (s (m))
(see (4.72)). If the heavy avour considered in HQET is, say, b, then there
are no bquark loops in HQET; its s ( ) is the same as in nl avour QCD
(Sect. 4.7), and its running is governed by the nl avour (s ).
os 1/2
The onshell matrix element of j() is M (p, p , ) = (Zqos )1/2 (ZQ
)
1
Z () (p, p ). It should be equal to C (, )M (
p, p , ) + O((
p, p )/m),
j
1 ( )(
os )1/2 Z
(
p, p ) (here
where p = mv + p and M
p, p , ) = (Zqos )1/2 (Z
Q
j
os
Zq contains no bquark loops, see Sect. 4.7). Both matrix elements are UVnite; their IR divergences coincide, because HQET coincides with QCD in
the IR region. Therefore,
C (, ) =
Zqos
Zqos
1/2
os
ZQ
os
Z
Q
1/2
j ( ) (p, p )
Z
p, p
+O
.
Zj () (
m
p, p )
109
(5.59)
(5.60)
110
5 HeavyLight Currents
mv
0
a
where
x1 = x1 + x2 ,
x2 = x3 + x4 ,
x3 = x5 ,
Li Ri = 1 1 , /
v ,
= 1 ,
= 2h(d) ,
(5.61)
where
d
h(d) = n
,
2
= (1)n+1
(5.62)
for the antisymmetrized product of n matrices. The eect of each contraction is then to produce a factor 2h. Terms with an odd number of contractions necessarily contain /v on the left, which yields an extra when moved
111
to the right, where it merely gives v/uQ = uQ . Thus the result involves only
powers of h:
=u
q x1 + x2 2h + x3 (2h)2 uQ .
Substituting the solution (5.30) for xi , we obtain
= 0 P, ,
1
v + 1)
P =
(d 2)(3d2 2 4h2 )1 (/
2(d 1)(d 2)
+ 4h(d 2h) /v (/
v + 1)
d 2 4h(1 h) (/
v + 1) .
Now we can apply the projector P to the integrand of the oneloop diagram of Fig. 5.6b and reduce it to a scalar expression quadratic in h:
dd k 2(d 1) + (dD2 /m2 + 4)h 2(D2 /m2 + 4)h2
iCF g02
,
=
2(d 1)
(2)d
D1 D2
(5.63)
where terms with D1 in the numerator have been omitted as they yield 0.
The integral can be easily calculated (Sect. 2.2), once, and for all currents.
We obtain the oneloop result [4]
= CF
g02 m2
(1 h)(d 2 + 2h)
.
()
(d 2)(d 3)
(4)d/2
(5.64)
112
5 HeavyLight Currents
situation: one of the quarks is massive, and the momenta are onshell (mv
and 0). From (5.35) and (5.64) we recover (5.36); from (5.37) and (5.64) we
have two new ways to derive (5.38). The twoloop results presented in [4]
conrm (5.36) and (5.38).
At the time when the paper [4] was written, the twoloop anomalous
dimension (5.9) of the generic quark current was not yet known (though all
specic results except the one for n = 2 were known). Therefore, the matching
condition (5.59) was used to nd both the minimal renormalization constant
Zj () and the nite matching coecient C (). Of course, Zj () could have
been obtained from an easier massless calculation, but actually it was a byproduct of a more dicult massive onshell calculation.
At last, we can combine all pieces of (5.59). With twoloop accuracy,
s (m)
3(n 2)2 + (2 )(n 2) 4
C (m, m) = 1 + CF
4
nl
m
2
s
i
. (5.65)
+ CF
CF aF + CA aA + TF ah + TF
al +
4
m
i=1
The oneloop result (following from (5.64)) was rst obtained in [8]. The twoloop calculation was done in [4]. There, a slightly dierent (and less useful)
quantity was obtained. The proper matching coecient (given by (5.59)) was
considered in [12], where an additional term to be added to the results of [4]
was derived. Here, for convenience, the complete result is presented:
317 10
11
2 (n 2)4 + 11(n 2)3 (n 2)3
aF =
24
3
2
253
16
+ 482 I (n 2)2 2(n 2)2 20(n 2)
+
6
3
689
8
32 64
2 + I (n 2) +
812 83 + 12I ,
+
3
3
3
16
43 4
aA = + 2 (n 2)4 2(n 2)3 + (n 2)3
12 3
8
143
9491 52
2 + I (n 2)2 +
(n 2)
+
216
3
3
18
281
29017
4
+
+ 82 I (n 2)
+ 292 + 23 6I ,
18
3
432
82 16
59
2
809 56
ah =
(n 2)2 (n 2) + + 2 (n 2) +
2 ,
54
9
9
3
27
3
38
1745
20
445 8
2
2 (n 2)2 (n 2) + (n 2) +
+ 2 ,
al =
54
3
9
9
108
3
(5.66)
(see (4.20)). The mass correction
(r) =
113
2
4 (1 (r) + 2 (r) 3 (r)) (n 2)2
3
4 (2 (r) 23 (r)) (n 2) 101 (r) 52 (r) 183 (r)
(5.67)
(5.68)
C5AC (, )
HV (, )
C5
C1 (, )
,
C 0 1 2 3 (, )
C5AC 0 (, )
C5HV 0 (, )
C 0 (, )
C 1 2 3 (, )
C5AC 3 (, )
C5HV 3 (, )
C 3 (, )
,
C 0 1 2 (, )
C5AC 0 1 (, )
C5HV 0 1 (, )
C 0 1 (, )
C 1 2 (, )
C5AC 2 3 (, )
C5HV 2 3 (, )
C 2 3 (, )
= 1.
C 0 1 (, )
(5.69)
114
5 HeavyLight Currents
+1
01
+1
2
12
012
+1
123
0123 4
+1
C (m, m)
369
s (m)
1 + 2CF
+ CF CF
+ 152 83 4I
16
4
1351
149
+ CA
32 + 23 + 2I + TF
82
48
9
nl
s 2
95
42 + 41 + 22 123
+ TF
12
4
i=1
255
s (m)
1 2CF
+ CF CF
152 83 + 4I
16
4
871
727
+ CA
+ 52 + 23 2I + TF
242
48
18
nl
s 2
47
+ 42 41 + 22 203
+ TF
12
4
i=1
28
s (m)
1453
173
1 4CF
+ CF CF
2 83 +
I
48
3
3
4
40
6821
133
14
+ 212 + 23
I + TF
2
+ CA
144
3
6
3
nl
2
445
10
28
s
+ 42 41
2
3
+ TF
36
3
3
4
i=1
689
s (m)
1 4CF
+ CF CF
812 83 + 12I
16
4
56
29017
809
+ 292 + 23 6I + TF
2
+ CA
432
27
3
nl
s 2
20
20
10
1745
+
2
1
2 123
+ TF
108
3
3
3
4
i=1
397
28
173
1 + CF CF
2 83 +
I
48
3
3
40
1703
391
14
+ 212 + 23
I + TF
2
+ CA
48
3
18
3
nl
s 2
10
28
143
+ 42 41
2
3
+ TF
12
3
3
4
i=1
s (m)
31
1 + 2CF
+ CF CF
152 83 + 4I
16
4
901
719
+ 52 + 23 2I + TF
242
+ CA
144
18
nl
s 2
125
+ 42 41 + 22 203
+ TF
36
4
i=1
s (m)
1649
1 + 10CF
+ CF CF
+ 152 83 4I
4
16
4021
47
32 + 23 + 2I + TF
82
+ CA
144
3
nl
s 2
317
42 + 41 + 22 123
+ TF
36
4
i=1
115
reconstruct the twoloop results for ZP (5.41) and ZA (5.42) (in two ways).
The nitemass corrections depend only on (n2), and are not aected when
is multiplied by 5HV ; therefore, they cancel in the ratios giving ZP and
ZA , as expected for results obtainable from anomalous dimensions (Sect. 5.3)
which are massindependent.
(5.71)
where the spin wave function M of the meson M is, for the groundstate
mesons,
1 + v/ 5
M = i mM
(5.72)
2
e
/
(for 0+ and 1+ mesons M should be rightmultiplied by 5 ; of course, their
F () diers from the groundstate value). This wave function has the obvious
property
v M = M/
/
v=M
(5.73)
( = 1 for groundstate mesons and +1 for P wave ones). The trace formula (5.71) shows that the HQET quantities corresponding to those in (5.70)
coincide:
116
5 HeavyLight Currents
F ()
fBP () = fB () = fB () = fBT () =
mB
(see (1.3)), as a consequence of the heavyquark spin symmetry. Hence, ratios
of QCD matrix elements are equal to ratios of matching coecients. These
ratios do not depend on the threeloop HQET anomalous dimension.
From the equations of motion,
(5.74)
where m() is the MS running massof the bquark. To the leading order in
1/m, we may replace mB by the onshell bquark mass m, obtaining
C1 (m, m)
fBP (m)
m
=
=
fB
m(m)
C 0 (m, m)
s (m)
= 1 + 4CF
4
121
1111
+ 302 8I + CA
82 + 4I
+ CF CF
8
24
143
+ TF
+ 162
6
nl
s 2
71
.
+ TF
82 + 81 (ri ) + 83 (ri )
6
4
i=1
(5.75)
The oneloop result was derived in Sect. 5.1 (see (5.18)); the twoloop result
was obtained in [11] (a typographical error was corrected in [4]).
The ratio of two observable matrix elements,
0
qQB
m f e
= B B ,
(5.76)
AC
0
mB f B
0
q 5 QB
is independent of the renormalization scale. At the leading order in 1/m, we
obtain [4]
C 1 (m, m)
fB
=
fB
C 0 (m, m)
s (m)
= 1 2CF
4
+ CF
263
1
CF (31 1282 + 16I) + CA
+ 162
3
9
41
4
+ TF + 82
3
3
nl
s 2
4
19
+ TF
42 (ri ) + 83 (ri )
3
3
4
i=1
8
I
3
117
(5.77)
Numerically,
2
fB
2 s (mb )
s
(6.37 + c )
=1
+ O 3s , MS
fB
3
mb
= 1 0.046 0.031 + ,
+ O(s ) ,
F () = F
1+
4
20
j0
0
4
C 0 (mb , mb ) = 1 + C1
3/2
j0
1
2
+ O(s ) + O
+ C1 +
20
j0
0
4
mc,b
(5.78)
j1 (5.45) and 0 , 1 (3.6) at nf = 4, we obtain
(cf. (4.77)). Substituting
j0 ,
6/25
s (mc )
fB
mc
=
1
fD
mb
s (mb )
MS
7
1887 s (mc ) s (mb )
+
.
2 +
+ O(2s ) + O
225
100
mc,b
118
5 HeavyLight Currents
fB
0
q i QB
we have [4]
C 0 1 (m, m)
fBT (m)
=
fB
C 1 (m, m)
307
4
1
4277
702 + 8I + CA
+ 82 I
= 1 + CF CF
3
8
216
3
1
421
+ TF
162
3
18
nl
s 2
205
8
+ TF
+ 2 1 (ri ) 3 (ri )
.
3
144
4
i=1
(5.80)
k
a
b
k
References
119
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
In this chapter, we continue the discussion of heavylight bilinear quark currents, and consider 1/m corrections in the expansion of QCD currents in
HQET, as well as dimension4 HQET operators which appear in these corrections. The 1/m term in the expansion (5.57) for the vector current (and the
axial current containing 5AC ) was rst investigated in [4], where the oneloop
anomalous dimension matrix of dimension4 operators i was found. The
full oneloop corrections to Bi were obtained in [5, 7]. Some general properties of the matching coecients Bi and the anomalous dimension matrix of
i following from reparametrization invariance and the equations of motion
were established in [7], and the twoloop anomalousdimension matrix was
calculated in [1, 2].
Bjm (, ) = C (, )Cm ( )
(6.2)
in the sum in the righthand side of the expansion (5.57), and use the leadingorder Lagrangian (2.5) to calculate its matrix elements. Renormalization of
the bilocal operators (6.1) requires both bilocal counterterms of the same
structure and local counterterms proportional to the local operators i in
A. Grozin: Heavy Quark Eective Theory, STMP 201, 121144 (2004)
c SpringerVerlag Berlin Heidelberg 2004
122
1
( /v /v ) ,
2
(6.3)
(6.4)
(6.5)
iD
,
m
(6.6)
123
which is the gaugecovariant extension of p/m (2.6). One should also use
V instead of Q
v . This eld is obtained by a Lorentz transformation which
Q
rotates v into V. With linear accuracy in 1/m,
1+V
/
iD
/
QV =
Qv = 1 +
Qv .
(6.7)
2
2m
The leadingorder expression
q Q =
1
v + 1 C+ C q/v Q
v
C+ + C q Q
2
2
are not determined by
The coecients of the operators qD Q
and mq qQ
general considerations. These coecients appear rst at the oneloop level.
We shall consider only O(s ) corrections at mq = 0 here, because some
operators are not yet known
anomalous dimensions for mixing with mq qQ
at the nexttoleading order (Sect. 6.3). We calculate the matrix element of
the QCD current from a heavy quark with momentum mv to a massless
quark with momentum p (with p2 = 0), expanded in p/m to the linear term,
and equate it to the corresponding HQET matrix element. In HQET, loop
corrections contain no scale, and hence vanish (except, possibly, massivequark loops, which rst appear at the twoloop level). The QCD matrix
element is proportional to u
(p) (p, mv)u(mv), where (p, mv) is the bare
propervertex function. At one loop, it is given by Fig. 6.1.
If we assume nothing about the properties of , then the term linear in
p has the structure
mv
k + mv
k+p
124
u
(p)
xi Li Ri u(mv) ,
(6.8)
p (1 + v/)/
p . Now we can take these double traces of the
v/
/
p (1 + v/)/
p, /
integrand of Fig. 6.1, and express xi via scalar integrals. Their numerators
involve (k p)n ; putting k = (k v)v + k and averaging over the directions of
k in the (d 1)dimensional subspace, we can express them via the factors
in the denominator.
Now we assume that satises (5.61). Then (6.8) becomes
x1 + (x2 + 2x5 ) 2h + x3 (2h)2 p v u(p) u(mv)
+ [x4 x5 2h] u
(p) /
pu(mv) .
(6.9)
Performing a simple calculation, we obtain
(p, mv) = (0, mv) +
1
g 2 m2
CF 0 d/2 () b1 p v + b2 /p
2m
(4)
(6.10)
The rstorder coecients for the components of the vector current were
calculated in [5, 7].
For = 1, 0 , the bracket in (6.10) becomes b p v, with b1 = b11 and
0
0
b 0 = b1 + 2b2 . For = i , i 0 , it becomes b1 p v i + b2 pi , where
i
125
01 ()M =
Tr
M,
2
1 = qiD Q
(6.12)
2
2
( = 1 for groundstate mesons and +1 for P wave 0+ and 1+ ones). On
the other hand, this matrix element is, from (6.12),
3
F2 () Tr M .
2
Therefore,
F2 () =
F () ,
3
(6.13)
(6.14)
The last formula requires some explanation. The chromomagnetic vertex always has the structure (Sect. 4.1), and there is always a heavyquark
propagator containing (1 + v/)/2 between and this vertex. We can always
insert the projector (1 + v/)/2 to the left of M; therefore, the indices , live
in the subspace orthogonal to v.
Let us consider the 1/m correction to the QCD current matrix element
and mq qQ
0jM , neglecting all s corrections. We may omit all qD Q
operators in the sum in (5.57), because their matching coecients start at
part
the order s . Reparametrization invariance (6.5) tells us that the qDQ
of this sum is just q iD
/ Q. Its matrix element is, from (6.12),
126
1
0
q iD
/ QM
= d 0
jM ,
3
where d is dened by
= d ,
d = 1 2h(d = 4) = 1 2(n 2)
(6.15)
(the properties (5.61) of have been used). The matrix elements of the
bilocal operators (6.1) are
jM ,
0jk M = Gk 0
1
0jk M = d Gm 0
jM ,
3
where d is dened by
1 + v/ 1 + v/
1 + v/
= 2d
,
2
2
2
d =
1 2
d 3 .
2
(6.16)
(6.17)
The numbers d (6.15) and d (6.16) have some interesting properties. Multiplying by 5 does not change them. Multiplying by v/ = 0 changes the
sign of d (6.15), leaving d (6.16) unchanged. For all eight antisymmetrized
products (Sect. 5.6), d = (1)n+1 d ; d = 3 for the four currents that
couple to 0 mesons, and d = 1 for those four that couple to 1 mesons.
In the rest of this chapter, we shall consider perturbative corrections to
the 1/m term of the expansion (5.57). It is sucient to consider two spin0
currents with = 1, 0 (Sect. 6.4) and two spin1 currents with = i ,
i 0 (Sect. 6.5). Replacing q by q5AC and mq by mq does not change the
matching. All the other QCD currents can by obtained by 5AC 5HV , i.e., by
dividing the current by ZP , ZA , or ZT = 1 (Sect. 5.3). Before considering 1/m
corrections to these four currents in detail, we shall discuss renormalization of
the local (Sect. 6.2) and bilocal (Sect. 6.3) dimension4 heavylight operators
which appear in these corrections.
O() = Z 1 ()O0 ,
(6.18)
dO()
+ (s ())O() = 0 ,
d log
127
(6.19)
where
= Z 1 ()
dZ 1 ()
dZ()
=
Z()
d log
d log
(6.20)
Z1
Z2
+ 2 + ,
where Z1 starts from the order s , Z2 from 2s , and so on. The anomalousdimension matrix must be nite at 0, and hence
= 2
dZ1
,
d log s
dZ2
dZ1
= (Z1 (s ))
d log s
d log s
(6.21)
(we consider only gaugeinvariant operators here). One can obtain the anomalousdimension matrix from Z1 , the coecient of 1/ in the matrix Z, and
vice versa. The coecients Z2 , Z3 , . . . contain no new information; at each
order in s , they can be reconstructed from lowerloop results. With twoloop
accuracy, the matrix of renormalization constants Z can be expressed via the
matrix of anomalous dimensions as
2
1 s
1
s
Z = 1 0
+ [(0 + 20 )0 21 ]
+ ,
2 4 8
4
1
1 s
s 2
+ [(0 20 )0 + 21 ]
+
(6.22)
Z 1 = 1 + 0
2 4 8
4
First, let us discuss the renormalization of the local dimension4 operators [1]
= q iD Q
2 = q i D
1 = q iD Q
,
Q,
= iv q /v Q
, = mq q Q
,
3 = q iv D
v Q
/
(6.23)
. When radiative
q Q
for any Dirac matrix . Note that 1 2 = i
corrections to matrix elements of these operators are calculated, the HQET
quark line contains no Dirac matrices, just scalar propagators (2.3) and vertices v . Therefore, no matrices can appear to the right of . If the light
quark is massless, the number of matrices to the left of remains even, if
it was even initially; in the terms proportional to mq , it becomes odd. The
full set of dimension4 operators satisfying these conditions is listed in (6.23).
Therefore, they are closed under renormalization.
128
Two of the operators are full derivatives of dimension3 currents which are
j (5.45): = Z
j , = Z
j ( ). The operrenormalized by Z
30
3
10
20
1
2
j ,
ator 4 with the MS lightquark mass mq is renormalized as 40 = Zm Z
4
where Zm is the mass renormalization factor (Sect. 5.1). The operator 1 is
renormalized as
j 1 + Z
a 2 + Z
b 3 + Z
c 4
10 = Z
(6.24)
j ). Therefore,
(we shall see in Sect. 6.5 that the rst coecient is indeed Z
j
a
b Z
c
Z
Z
Z
b Z
c
j + Z
a Z
0 Z
,
(6.25)
Z =
j 0
0
0
Z
0
0
0 Zj Zm
and
Z 1
1
Z
Za
j
0 Z1 + Z
a
j
=
0
0
0
0
Zb
Zb
Z 1
Zc
Zc
0
j
1
1 Zm
0 Z
j
a,b,c , Z
j , Zm . The anomalousdiwhere Za,b,c can be easily expressed via Z
mension matrix (6.20) is
0
a
b
c
0
a
b
c
=
j +
(6.26)
0 0 0 0 .
0 0 0 m
a,b,c , Za,b,c , Zm can be easily expressed
j , Z
The renormalization constants Z
a,b,c , m using (6.22).
via the anomalous dimensions
j ,
Now let us take = . Then
10 = 10 + 30 ,
20 = 20 + 30 ,
30 = (3 2)20 ,
where
,
1 = i q Q
40 = (3 2)30 ,
,
2 = iv q/v Q
(6.27)
.
3 = mq q Q
(6.28)
j () (), =
Of course, these operators are renormalized as 10 = Z
1
20
j () (), = Z
j ()Zm () (). We obtain
Z
2
30
3
129
j Za + (3 2)Zb ()
1 () = 1 () + Z
2
j Za + (3 2)Zc 3 () ,
+ Zm 1 + Z
2 () = 1 + Zj Za + (3 2)Zb 2 ()
j Za + (3 2)Zc () ,
+ Zm 1 + Z
3
3 () = 32 () ,
4 () = 33 () .
(6.29)
Therefore,
j Za + (3 2)Zb
Z
j Za + (3 2)Zc
and Zm 1 + Z
b = (
a +
a m +
3
3
2
1
s
a0 (
a0 20 )
+ O(3s ) ,
b =
3
4
2
1
s
c = (
a0 m0 ) (
a0 m0 20 )
+ O(3s ) .
3
4
(6.30)
4 () = 33 () .
(6.31)
130
k+p
k+p
a p + Zb p v /
Z
v + Zc mq .
(p, 0) =
iCF g02
/k /k
k
dd k /k + mq
v
2 /p 2 +
g
(2)d
k2
k k
kv
k k
1
2 g (1 a0 ) 2
.
k
k
dd k
(2)d
/v /k /p/k k 2 mq
k
.
kv
(k 2 )3
c mq
Za p
v+Z
+ Zb p v /
s
1
(6p + 2 p v /v 2mq
= CF
).
4
4
Finally, we arrive at
a = 3CF
s
+ ,
4
b = CF
s
+ ,
4
c = CF
s
+ ,
4
131
a = 3CF
4
4 2
s 2
1
41
10
+ CF
5 CF + 2 +
+
CA TF nl
3
3
3
3
4
(6.32)
,
= iv
j , k2 = mq q/v Q
k1 = q iv D Q
k3 = i dx T {
j(0), k (x)}
(6.33)
are closed under renormalization for any Dirac matrix .
(where
j = q Q)
k
j ()k1 () and k20 = Z
j ()Zm ()k2 (). To renormalize
Of course, 10 = Z
k30 , we need to renormalize
j0 and k0 (the later is trivial: Zk = 1, Sect. 4.1).
However, this is not enough. The product of the renormalized operators is
singular at x = 0, and the bilocal operator needs additional local counterterms:
j ()k3 () + Z
ak ()k1 () + Z
bk ()k2 () .
k30 = Z
Therefore,
j 0
Z
j Zm
Z = 0 Z
k
k
Za Z
b
and
0 ,
Zj
Z 1
1
Z
0
j
1
1 Zm
= 0 Z
j
Zak
Zbk
(6.34)
0 ,
Zj1
0 0 0
=
j + 0 m 0 .
ak
bk 0
(6.35)
(6.36)
132
The running of the local operators, of course, does not depend on the bilocal
ones. The renormalized bilocal operator contains an (innite) admixture of
the local operators which compensates its singularity at x = 0. The amount
of the local operators contained in the renormalized bilocal operator varies
with , as shown by the last equation.
k
at one loop. To this end, we take the onshell matrix
Let us calculate
a,b
with p = 0 to q with a small onshell
element u
(p, 0) u of k30 from Q
momentum p, and expand it in p and mq up to the linear term:
(p, 0) = 0 + 1 mq + 1 p +
The rst contribution to the linear term is oneloop (Fig. 6.3). It has a logarithmic UV divergence, and we may retain only UV 1/ poles. According
to (6.34), the result should be equal to
k p v + Z
k mq /
Z
v.
a
b
From Fig. 6.3, we obtain a gaugeinvariant (p, 0); its linear term is
2
/v /k /p/k k 2 mq
dd k
k
.
iCF g02
k
/
(2)d
kv
(k 2 )3 k v
Similarly to the previous calculation, we obtain
k2
1
4
dd k
1
1
.
1 mq + 1 p = iCF g02 p v + mq /v
3
2
(2)d (k 2 )2 (k v)2
Averaging the rst term over the directions of k, we obtain (d 2) 2
(Sect. 4.3). Finally,
ak = 8CF
s
+ ,
4
bk = 4CF
s
+ .
4
ak = 8CF
4
16 2 608
s 2
160
32 2 32
T F nl
+ CF
CF +
CA +
9
3
3
9
9
4
+
(6.38)
k
k+p
k
Fig. 6.3. Matrix element of O20
k+p
133
Finally, we shall discuss bilocal operators with a chromomagneticinteraction insertion m (4.11), again following [2]. The four operators
1
i
m
q
iv
(1
+
v
/
)
(1
+
v
/
)
Q
=
Q
,
1
4
4
i
,
i D /
v (1 + v/) Q
m
2 = q
4
1
,
/
v (1 + v/) Q
m
3 = mq q
4
m
dx T {
j(0), m (x)}
(6.39)
4 =i
are closed under renormalization for any . Note that the indices , live in
= Q.
When we calculate radiative
the subspace orthogonal to v, owing to v/Q
m
corrections to matrix elements of 4 , there is always one chromomagnetic
vertex (with the Dirac structure ) to the right of , and a heavyquark
propagator(s) containing 1 + v/ between and this vertex. If the light quark
is massless, the number of matrices to the left of remains even; in terms
proportional to mq it becomes odd. The only local operators satisfying these
conditions are m
1,2,3 . They are just O3,2,4 (6.23), with
1
= i /
v (1 + v/) .
4
They renormalize among themselves, according to (6.25). The bilocal operam
tor m
4 needs the additional local counterterms 1,2,3 , so that
Zj
0
0
0
c
a Z
0
Z Z +Z
,
Z = b j
j Zm 0
0
0
Z
am
cm Z
j Z
m
m Z
Z
Z
b
1
Z
0
0
0
j
Z Z1 + Z
Zc
0
b j
a
Z 1 =
,
1
0
0
0
Zj1 Zm
1 Z
1
Zbm
Zam
Zcm Z
m
j
0 0 0 0
b
a
c 0
.
=
j +
0 0 m 0
bm
am
cm
m
(6.40)
Do not confuse the anomalous dimension m (5.16) of the MS mass (Sect. 5.1)
with the anomalous dimension
m (4.67) of the chromomagneticinteraction
operator (Sect. 4.6)!
134
m
Let us calculate the anomalous dimensions
a,b,c
at one loop. To this end,
m
with p = 0 to q with a
we take the onshell matrix element of 4 from Q
small onshell momentum p, and expand it in p and mq up to the linear term.
It starts at one loop (Fig. 6.4), and is logarithmically divergent; we retain
only its UV 1/ pole. This should give us
1 m
bm p v + Z
cm mq /v (1 + v/) .
Za ip Z
(6.41)
4
The vertex correction of Fig. 6.4 is obviously gaugeinvariant; its linear part
is
k /p/k k 2 mq )
1
dd k k (/
2
CF g0
(1 + v/) .
2
(2)d
(k 2 )3 k v
am = 2CF
4
4
m
The twoloop calculation of
a,b
was performed in [2]:
s
am = 2CF
4
2 2 110
s 2
4
8 2 22
+
+ CF +
CF +
CA + TF nl
9
3
9
9
9
4
s
bm = 2CF
4
10 2 50
s 2
4
40 2 16
+ CF
+
CF +
CA + TF nl
9
3
9
9
9
4
(6.42)
k
k+p
m
Fig. 6.4. Matrix element of O40
135
j 0
Z
0
0 Z
j
0
j Zm
Z = 0 0 Z
k
ak Z
0 Z
b
m
0 Z1 Z2m
0
0
0 0 0 0 0
0 0 0 0 0
0
0
0 0 m 0 0
j +
0
0 , =
. (6.44)
k
k
0
0
0
Zj 0
a
b
m m
0
0
m
1
2
0 Zj Zm
B51 = Cm C1 ;
B1 = C 0 2C 1 ,
B4 = C 0 ,
B5 = Cm C 0 .
(6.45)
(6.46)
(6.47)
136
1 1 m
m
m 2 ()
m
4 () = 5 () + Zj Zm Z1 (1 )Zj Za + (3 2)Zb
1 Z
m + (1 )Z
1 Z
j Zm Z m + (3 2)Z m 3 () .
+ Z
j
(6.48)
Therefore,
1 m
1 Z
m
j Zam + (3 2)Zbm
Z
Z1 (1 )Z
and
j
1 m
m
j Zm Zam + (3 2)Zcm
1 Z
Z2 + (1 )Z
Z
j
m
must be nite at 0. This allows one to reconstruct
1,2
in (6.44) from
m
1m =
am 3
bm +
1m ,
2m =
am + 3
cm +
2m ,
s 2
1
1
m
m
m
(
m0 20 ) (
a0
a0
+ 5b0
)
a0
+ O(3s ) ,
1m =
2
3
4
2
1 m
s
a0 (2
a0 + m0 3
m0 + 60 )
+ O(3s ) .
(6.49)
2m =
6
4
The nite part, at the nexttoleading order, is
m
4 () = 4 () +
1 m
1 m s ()
m s ()
(
a0 + 5b0
2 ()
3 () . (6.50)
)
2
4
2 a0 4
B2
=
j B2 +
ak B4 +
1m B5 ,
log
B3
= (
j + m ) B3 +
bk B4 +
2m B5 .
log
(6.51)
The running of the coecients B4,5 of the bilocal operators does not depend
on the coecients of the local operators (see (6.46)), as expected (6.45); the
amount of 2,3 contained in the bilocal operators 4,5 is dependent,
and hence the equations (6.51) for B2,3 contain B4,5 in the righthand side.
k +
2 (m) + (other operators) .
= CF b 0
2
4
CF
Taking
0k +
0m = 0 into account, we see that both b should vanish at = 0.
The O() terms of (6.10) give [5, 7]
s (m)
+ ,
4
0
(m)
B2 (m, m) = 12CF s
+
4
137
(6.52)
The initial values B3 (m, m) are also O(s ); we do not consider them here.
In the leadinglogarithmic approximation (LLA),
Cm ( ) = xm0 ,
x=
s ( )
s (m)
C (m, ) = xj0 ,
1/(20 )
mq ( )
= xm0 ,
mq (m)
(6.53)
(see (4.71), (5.58), (5.19)). The coecients B (, ) for = 1, 0 are [4, 7],
from (6.45) and (6.51),
B1 (, )
B5 (, )
B4 (, )
=
=
1
,
= xm0 ,
C (, )
C (, )
C (, )
m
m
B2 (, )
+ 3
b0
a0
k
m0
x
=
log
x
1
,
a0
C (, )
m0
k
m
b0
B3 (, )
a0
m0
m0
m0
x
.
=
(x
1)
+
x
C (, )
m0
m0 m0
(6.54)
(6.55)
(j v)v + j
and substitute the expansions (5.57) with (6.43) and (6.45).
The matrix element of (6.55) from a heavy quark with momentum mv to an
onshell light quark with momentum p reads
0
C 1 (, )
B2 (, )
1
+2
1
pv
mC 0 (, ) 1 +
2m
C 0 (, )
C 0 (, )
0
B3 (, )
mq ( ) + r
+
C 0 (, )
1 B21 (, )
pv
= m()C1 (, ) 1 +
2m C1 (, )
1
B3 (, )
+
2 mq ( ) + r
,
C1 (, )
138
where
j ( )Q
qi dx T {
j( ), (k + Cm ( )m ( ))x } Q p q
r=
,
q
j( )Q
(6.56)
B21 (, ) B2 (, )
=2
C1 (, )
C 0 (, )
C 1 (, )
1
C 0 (, )
,
B31 (, ) B3 (, )
= 2.
C1 (, )
C 0 (, )
(6.57)
Note that (6.11) is just the oneloop case of the rst result.
Now we shall discuss Bmeson matrix elements; therefore, we replace q by
and the denitions
j = q5AC Q,
q5AC and set mq = 0. The leading current is
of i (6.43) are changed accordingly. The matrix element of
j is
(6.58)
0
j()B = i mB F () ;
see (5.71). When taking matrix elements of 1,2 (), we replace the derivative
acting on the whole operator by the dierence between the momenta of the
, and obtain
states, so that i v
() .
(6.59)
01 ()B = 02 ()B = i mB F
The operator 4 () is just k2 () with = 5AC (Sect. 6.3), and the rst
formula in (6.14) gives
dGk ()
=
k (s ()) ,
d log
dGm0 ()
+
m (s ())Gm0 () =
m (s ()) .
d log
139
(6.62)
m
m s ()
Gm0 () = G
+ 0 ,
4
m0
(6.63)
fB
mB
1
C ( ) + Gk ( ) + Cm ( )Gm0 ( ) ,
1+
2m
(6.64)
where = 1, 0 , and
C1 ( )
B 1 (, )
=1 2
,
C1 (, )
0
C ( )
C 1 (, ) B2 (, )
=12
.
C 0 (, )
C 0 (, )
fB
4
mB
k
m
1
(m)
0 log s
1+ 0
1+
2m
20
4
m0
/(20 )
s (m) m0
+ Gk + Gm
.
(6.65)
4
The NLO results can be found in [2, 3].
The matrix element of (6.55) is (5.74). Substituting (6.64) and using the
relation (6.57), we obtain
fBP ()
C1 (, )
=
1+
.
fB
C 0 (, )
m
The ratio of the quark masses is given by (6.56); naturally, it contains no 1/m
corrections with Bmeson hadronic parameters; it is just a series in s (m).
Therefore, we recover mB = m + .
140
i1 = qiDi Q
2
= i0
i3 = q i D i Q
, i4 = q i D 0 i Q
ji , i5 = mq
ji ,
!i
"
!i
"
i6 = i dx T
j (0), k (x) , i7 = i dx T
j (0), m (x) . (6.66)
Note that
.
i1 i3 = i i qQ
(6.67)
B1 = 2C 0 , B2 = B6 = C i , B7 = Cm C i ,
i 0
i 0
i 0
1 i 0
B1 = B2 = B6 = C i 0 , B7 = Cm C i 0 .
2
(6.68)
00
a
b
c 0 0
0 0 0 0 0 0 0
0 0
b
c 0 0
a
(6.69)
=
j +
0 0 0 0 0 0 0 .
0 0 0 0 m 0 0
0 0 0
ak
bk 0 0
00
3m
4m
5m 0
m
The operators (6.23) with = 1 are
1 = i1 ,
2 = i3 ,
3 = i4 ,
4 = i5 ;
k2 = i5 ,
k3 = i6 .
Therefore, the relevant parts of (6.69) can be taken from (6.26) and (6.36).
The bilocal operator i6 = k3 (with = i ) needs counterterms proportional
to k1 = i4 and k2 = i5 (see (6.35)), but not i3 . The operators i2,4 are
renormalized multiplicatively with
j , which determines the second and the
fourth row. The same holds for i1 i3 (6.67), and hence the rst and the
141
third row coincide. Furthermore, the form of B1,2,6,7 (see (6.68)) xes columns
1, 2, 6, and 7. In particular, the (nonmixing) anomalous dimension 11 of
i1 is
j ; therefore, the fact that the element 11 of the matrix (6.26) is
j
follows from reparametrization invariance (this proves the statement made in
Sect. 6.2).
i
i
The operators m
1,2,3,4 (6.39) with = are related to 3,4,5,7 :
i
m
10 = (1 )(1 + 2)40 ,
i
i
i
m
20 = (1 2)30 + 40 + 50 ,
i
m
30 = (1 )(1 + 2)50 ,
i
m
40 = 70 .
(6.70)
We have
i
1 1 m
m i
m
4 () = 7 () + Zj Zm Z3 + (1 2)(Zj + Za )Za 3 ()
1 Z
m + Z
j + (1 2)Z
b Z m
+ Zj1 Z
m
4
a
j Zbm i4 ()
+ (1 )(1 + 2)Z
1 Z
m + Z
j Zm + (1 2)Z
c Z m
+ Zj1 Z
m
5
a
j Zm Z m i () .
+ (1 )(1 + 2)Z
c
5
(6.71)
3,4,5
in (6.69) from
a,b,c
in (6.40):
3m =
am +
3m ,
4m =
bm +
4m ,
5m =
cm +
5m ,
2
s
m
a0
(
a0
m0 + 20 )
+ O(3s ) ,
3m =
4
1
1 m
s 2
m
m
m
(
a0 +
a0 a0
4 =
b0 ) (
m0 20 )
+ O(3s ) ,
2
3
4
2
1 m
s
a0 (2
a0 + m0 3
m0 + 60 )
+ O(3s ) .
5m =
6
4
(6.72)
5 () .
+
2 a0 4
i
m
a0
m
3 () = 6 ()
(6.73)
B3
= (
j +
a ) B3 +
a B1 +
3m B7 ,
log
142
B4
ak B6 +
=
j B4 +
b B1 + B3 +
4m B7 ,
log
B5
= (
j + m ) B5 +
c B1 + B3 +
5m B7 .
bk B6 +
log
(6.74)
g02 m2
i
10 20 + 50 + 60
() b,2 i30 + b,1 O40
d/2
(4)
m
s (m)
10
=
a0
CF b,2
i3 (m)
4
2
m
20
0k +
i
b0
+ CF b,1
4 (m) + (other operators) .
2
CF
B3 (m, m) = 4CF
The initial values B5 (m, m) are also O(s ); we do not consider them here.
In the LLA, the coecients B (, ) for = i , i 0 are [4, 7],
from (6.68) and (6.74),
1 B1 (, )
B (, )
B7 (, )
B6 (, )
=
= 2
= 1,
= xm0 ,
C (, )
2 C (, )
C (, )
C (, )
m
B3 (, )
a0
a0
a0
m0
x
,
=
2
x
1
+
x
C (, )
a0
m0
m
B4 (, )
a0
1
k
a0
x
=
log
x
1
a0
C (, )
3
a0
m0
m0
m
1
x
a0
a0
1
m
+
+
b0
.
3
a0
m0
m0
k
m
B5 (, )
b0
a0
1
m0
a0
m0
x
=
(x
1)
x
C (, )
m0
3
a0
m0
+
m
1 (
a0 m0 )
a0
xm0 xm0
.
3
a0
m0
m0 m0
(6.76)
143
0
j ()B = i mB F ()ei ,
(6.77)
where e is the B polarization vector. The operators i2,4 are full derivatives:
()ei .
0i2 ()B = 0i4 ()B = i mB F
(6.78)
The matrix elements of i1 and i3 are equal, owing to (6.67). Let us dene [6]
F2 ()
Tr
01 ()M =
M,
2
(6.79)
s
1
+ O(2s ) .
R(s ) = 1 + a0
3
4
(6.80)
Finally [3],
i
()R(s ())ei .
mB F
0i1 ()B = 0i3 ()B =
3
(6.81)
(6.82)
0i5 ()B = i mB F ()Gk ()ei .
The second formula, together with (6.73), gives
i
0i7 B =
mB F ()Gm1 ()ei ,
3
Gm1 () = Gm () + Rm1 (s ()) ,
1 m
m s
Rm1 (s ) = (
+ 3
+
b0
)
2 a0
4
Taking the matrix element of (5.57), we obtain
(6.83)
144
fB
fBT ()
C (, )F ( )
mB
1
1
1+
,
C ( ) + Gk ( ) Cm ( )Gm1 ( )
2m
3
(6.84)
where = i , i 0 , and
B (, )
C ( ) = 1 4
C (, )
1
C 0 (, )/C 1 (, )
2
1
3
B3 (, )
C (, )
R(s ( )) .
fB
fT
=
j0 /(20 )
C F
mB
0k
1
s (m) 1
0m
log
+
1+
1
2m
20
4
3
m0
m0 /(20 )
1
(m)
s
k G
m
+G
.
3
4
s (m)
4
(6.85)
References
1. G. Amor
os, M. Neubert: Phys. Lett. B 420, 340 (1998) 121, 127, 130
2. T. Becher, M. Neubert, A.A. Petrov: Nucl. Phys. B 611, 367 (2001) 121, 131,
132, 133, 134, 137, 139, 142, 144
3. F. Campanario, A.G. Grozin, T. Mannel: Nucl. Phys. B 663, 280 (2003); Erratum: Nucl. Phys. B 670, 331 (2003) 137, 138, 139, 142, 143, 144
4. A.F. Falk, B. Grinstein: Phys. Lett. B 247, 406 (1990) 121, 137, 142
5. M. Golden, B. Hill: Phys. Lett. B 254, 225 (1991) 121, 124, 136
6. M. Neubert: Phys. Rev. D 46, 1076 (1992) 125, 143
7. M. Neubert: Phys. Rev. D 49, 1542 (1994) 121, 122, 124, 136, 137, 142
7 HeavyHeavy Currents
In this chapter, we discuss heavyheavy quark currents. They have many interesting applications. Matrix elements of the vector and axial b c currents
describe exclusive semileptonic B decays, which provide one of the ways to
measure the CKM matrix element Vcb . Matrix elements of the electromag and DD
production in e+ e annihilation.
netic b and ccurrents describe BB
+ Q
J0 = Q
v 0 v0 ,
cosh = v v ,
(7.1)
v and Q
v are scalar staticquark elds with velocities v and v . This
where Q
current is clearly discussed in [9]. Let the sum of all oneparticleirreducible
vertex diagrams with this current (not including external leg propagators) be
The oneloop vertex function in the coordinate space (Fig. 7.1) is
= 1 + .
0
t ; cosh ) = iCF g02 D
(t,
(vt + v t )v v (t)(t )
g02
(d/2 1)(t)(t )
8 d/2
(1 + a0 )x2 cosh + (d 2)(1 a0 )(t + t cosh )(t + t cosh )
,
(x2 )d/2
= CF
vt
v t
(7.2)
146
7 HeavyHeavy Currents
; cosh ) =
(,
t ; cosh ) .
dt dt eit+i t (t,
(7.3)
When the vertex function (, ; cosh ) is expressed via the renormal r (, ; cosh ), where Z
is
ized quantities s (), a(), it should become Z
minimal and r (, ; cosh ) is nite at 0. The renormalization constant
. The UV divergences of (, ; cosh ) do
Q Z
J = Z
of the current (7.1) is Z
not depend on the residual energies , , and we may set them to zero. An IR
cuto is then necessary to avoid IR 1/ terms. At one loop, substituting (7.2)
into (7.3) and proceeding to the variables
t=
1+
,
2
t =
1
,
2
we obtain
0; cosh ) = CF
(0,
+1
d
1
g02
16 d/2
d
1
2
0
d
d3
2 2
+/2
1a
(cosh ) = 1 CF s
Z
cosh 2
d 2 coth +
4 /2
sinh
s
= 1 CF
(2 coth + 1 a) .
4
(7.5)
is multiplied by Z
Q (3.65), the gauge dependence cancels, and
When this Z
we obtain the anomalous dimension [12]
J (s ; cosh ) = 4CF
s
( coth 1) .
4
147
(7.6)
(7.7)
In fact, this Ward identity holds to all orders. Let us consider an arbitrary
+ t ) in the coordinate space. The vertices along the heavydiagram for (t
quark line have times t0 < t1 < < tn1 < tn , t0 = t, tn = t , and
integration with respect to t1 , . . . , tn1 is performed. The integrand is an
integral over the coordinates of all vertices not belonging to the heavyquark
t ; = 0) obtained by inserting the
line. Now consider all diagrams for (t,
heavyheavy current vertex with = 0 at time 0 into all the possible places
along the heavyquark line. These diagrams have the same integrand, and
integration regions t0 < t1 < < ti1 < 0 < ti < < tn1 < tn
(i = 1, . . . , n). These regions span the whole integration region of the original
diagram. Therefore, the sum of this set of vertex diagrams is equal to the
original selfenergy diagram.
Applying the Fourier transformation (7.3) to (7.7), we obtain
; = 0) = ( ) ()
(,
1 ( ) S1 ()
S
.
or (, ; = 0) =
(7.8)
This can be proved directly, in the same way as for the QED Ward identity
(see Fig. 3.16). We start from an arbitrary diagram for (),
and construct
a set of diagrams for (, ; = 0) by inserting the heavyheavy vertex
into all propagators along the heavyquark line in turn (Fig. 7.2; a dot near a
heavyquark propagator means that its residual energy is shifted by ).
Each diagram is a dierence, because
i
+ i ) iS(
+ i ) iS(
+ i ) =
+ i ) .
i
S(
iS(
All terms cancel each other, except the extreme ones (Fig. 7.2), and we arrive
at (7.8). In particular,
; = 0) = d()
(,
d
dS1 ()
or (, ; = 0) =
.
d
148
7 HeavyHeavy Currents
=
i
=
i
=
J (s ; = 0) = 0 ,
(7.9)
to all orders in s .
The heavyheavy current has especially simple properties in heavyelectron eective theory (Sect. 3.5). Let us consider the full (nonamputated)
After singling out the obvious functions, it
Q
+ , and J.
Green function of Q,
can be written as G(t, t ; cosh ) (Fig. 7.3). The exponentiation argument (see
Fig. 3.14) holds for this heavyquark world line with an angle, too. Therefore,
e20
F (t, t ; cosh ) ,
G(t, t ; cosh ) = (t)(t ) exp
(4)d/2
(7.10)
where F (t, t ; cosh ) is just the oneloop correction. Let us divide (7.10) by
149
v t
vt
+ t )(t)(t )
G(t, t ; = 0) = iS(t
at t > 0, t > 0:
G(t, t ; cosh )
+ t )
iS(t
e20
(F
(t,
t
;
cosh
)
F(t,
t
;
=
0))
, (7.11)
= exp
(4)d/2
G(t, t ; cosh ) =
where F (t, t ; cosh ) is the oneloop correction which has the J vertex inside
(shown in Fig. 7.1); corrections to the external legs cancel here. If this ratio
is reexpressed via the renormalized quantities (this is trivial, because in this
J Gr (t, t ; cosh ), where
theory e = e0 and a = a0 ), it should be equal to Z
Gr (t, t ; cosh ) is nite at 0. Therefore,
J = exp (f (cosh ) f ( = 0)) ,
Z
4
e F (t, t ; cosh ) = f (cosh ) + O() .
The anomalous dimension is exactly equal to the oneloop contribution,
J (cosh ) =
( coth 1) .
(7.12)
g02
g04
F
+
C
(C
F
+
T
n
F
)
+
F
A A
F l l
(4)d
(4)d/2
(7.13)
(by the nonabelian exponentiation theorem [7, 6]). If the colour factors of
all twoloop diagrams with two gluons attached to the heavyquark line were
equal to CF2 (as in the abelian case), they would produce the F 2 term in the
expansion of the exponential. In the nonabelian case, the colour factors of
some diagrams also contain a nonabelian part CF CA , which should be taken
into account separately (these parts contribute to FA ). The ratio (7.11) can be
150
7 HeavyHeavy Currents
written similarly. Only the diagrams with the J vertex inside the correction
(Fig. 7.4) contribute to FA and Fl (the diagrams of Figs. 7.4b,d should be
taken with the nonabelian part of their colour factors, CF CA ). Therefore,
the renormalization constant has a similar structure:
s
(f (cosh ) f ( = 0))
Z = exp CF
4
2
s
CA (fA (cosh ) fA ( = 0))
+ CF
4
+ TF nl (fl (cosh ) fl ( = 0)) + .
(7.14)
In particular, this means that the twoloop anomalous dimension contains no
CF2 term.
These diagrams were calculated in [9] (except the easiest one, Fig. 7.4a
with the quarkloop correction, which gives fl (cosh ) and was found in [8]).
The diagrams of Figs. 7.4a,b,d can be calculated straightforwardly in the coordinate space, similarly to the oneloop case discussed above (for Fig. 7.4a,
this is slightly easier than the momentumspace method used in [9], see
Sect. 8.4). The diagram of Fig. 7.4c is more dicult; it was calculated in
the momentum space in [9]. The result is
J (s ; cosh ) = 4CF
s
( coth 1)
4
151
268 2 2
( coth 1) + 8
9
3
+ 4 coth ( coth + 1) Li2 (1 e2 ) Li2 (1 e2 )
8 coth2 Li3 (1 e2 ) + Li3 (1 e2 )
sinh
coth 1
log
d
8 sinh 2
sinh
sinh2 sinh2
0
2
80
s
TF nl ( coth 1)
+
(7.15)
9
4
+ CF CA
2
,
6
(7.16)
the result (7.15) can be rewritten via just one dilogarithm any of Li2 (1e2 ),
Li2 (1 e2 ), or Li2 (e2 ).
At small angles,
0 (s )2 + O(4 ) ,
J (s ; cosh ) =
376 8 2
s 2
s
4
80
0 (s ) = CF
+ CF CA
T F nl
+
3
4
27
9
27
4
(7.17)
At large angles,
J (s ; cosh ) =
(s ) + O(0 )
(7.18)
268 4 2
s 2
s
80
+ CF CA
T F nl
+
(s ) = 4CF
4
9
3
9
4
(7.19)
It is remarkable that one of the nest perturbativeHQET papers [9] was
written several years before the HQET gold rush of 199091. The oneloop
result (7.6) is, perhaps, the anomalous dimension that has been known for
the longest time it follows from classical electrodynamics (Sect. 1.2).
152
7 HeavyHeavy Currents
1
1
+
Gi (cosh , , )i ( ) +
Gi (cosh , , )i ( )
2mb i
2mc i
+
(7.20)
i = ,
v ,
/
/
v ,
/v /v .
(7.21)
H3A = 2(H3 H4 ) .
(7.22)
First we shall consider the simple case v = v, = 0; the general case will
be discussed in Sects. 7.4 and 7.5. For satisfying (5.5), only one leadingorder HQET current J with the same appears in (7.20). should commute
we
with /v ; otherwise, by inserting two (1 + v/)/2 projectors around it in J,
see that the leading term vanishes. The HQET current J, and hence the
matching coecient H, does not depend on .
Q.
We begin with the simplest case of avourdiagonal currents J = Q
The onshell matrix element of the QCD current is
153
os 1
Q, p = mvJQ, p = mv = ZQ
Zj u
(mv, mv)u ,
where (p , p) = + (p , p) is the bare propervertex function, and the onos
shell heavyquark eld renormalization constant ZQ
has been calculated in
Sect. 4.2. For satisfying (5.5), u (mv, mv)u = u u (m2 ), (m2 ) = 1 +
p = 0J
Q,
p = 0 ,
(m2 ). This QCD matrix element should be equal to H Q,
where the HQET onshell matrix element is
os
p = 0J
Q,
p = 0 = ZQ
u
v (0, 0)uv
Q,
J ( = 0) = 1). Both matrix elements
(here HQET has nl = nf 1 avours; Z
are UV nite, but may contain IR divergences, which are the same because
HQET was designed to reproduce QCD in the IR region. At this point p = 0,
os vanish (except those
p = 0, and all loop corrections in (0, 0) and Z
Q
containing loops of other massive avours; such contributions rst appear at
two loops).
The oneloop diagram for u
u is shown in Fig. 7.5. It has the structure
xi Li Ri u , Li Ri = 1 1 , ,
u
(/
v can always be anticommuted to u or u
). Taking the double traces with
i R
i = (1 + v/) (1 + v/) ,
L
(1 + v/) (1 + v/) ,
(1 + v/) (1 + v/) ,
we see that the matrix M 1 is the same as in (5.30). Therefore, this diagram
can be calculated, once, and for all , using the projector
P =
1
(d 2)(3d 2 4h2 )(1 + v/) (1 + v/)
2(d 1)(d 2)
+ 4h(d 2h) (1 + v/) (1 + v/)
+ (d + 2 4h + 4h2 ) (1 + v/) (1 + v/)
(see Sect. 5.3). The scalar integrals can be calculated using (4.15). The result
is obviously gaugeinvariant:
(m2 ) = CF
g02 m2
2(d 2) + (d 3)h(d 4 + 2h)
.
()
d/2
(d 2)(d 3)
(4)
(7.23)
154
7 HeavyHeavy Currents
os
Multiplying (m2 ) by ZQ
, we obtain
os
ZQ
(m2 ) = 1 CF
g02 m2
(1 h)(d 2 + 2h)
+
()
d/2
d2
(4)
(7.24)
The matrix element of the vector current can be obtained from the Ward
identity (5.10) (see (4.25)):
p m) + m m1 (p2 ) ,
1 2 (p2 ) (/
p
u
(mv, mv)u = 1 2 (m2 ) 2m2 1 (m2 ) v u
u ,
(p, p) =
os
ZQ
u (mv, mv)u = v uu
(7.25)
to all orders. The matrix = v/ has h = d/2+1; substituting this into (7.24)
we see that our oneloop result satises this requirement.
Recall that we are only considering matrices commuting with 0 = v/
(in the v rest frame), for which u u does not vanish. Such matrices have
d
h(d) = n
,
(7.26)
2
where = (1)n+1 for an antisymmetrized product of n matrices, and
= (1)n for such a product multiplied by 5AC . Comparing the matrix
elements (7.24) for = 1 (h = 2 ) and 5AC 5HV (h = 2 + ), we reproduce (5.36); for = 0 (h = 1 + ) and 5AC 5HV 0 (h = 1 ), we
reproduce (5.38); for = 5AC (h = 1 ) and 5HV (h = 1 + ), we
reproduce (5.38) again; and for = 5AC 0 (h = ) and 5HV 0 (h = ),
we reproduce ZT = 1 (Sect. 5.3).
The product (7.24), when expressed via the renormalized s (), should
be equal to Zj (s ())H() (which does not depend on ). Its UV divergence
reproduces (5.8). It is most natural to perform matching at = = m. The
matching coecient is
H (m) = 1 + CF
s (m)
(n 2)(n ) +
4
(7.27)
This calculation can be done at two loops; the calculation of the necessary
integrals was discussed in Sect. 4.2. The diagrams can be obtained from those
for the quark selfenergy (Fig. 3.6) by inserting the current vertex into all
propagators along the quark line. For other , the matching coecient can
be found using the renormalization group.
The integral of the 0th component of the QCD vector current J over all
space is the number of heavy quarks (minus antiquarks). The same is true
. Therefore,
for the HQET current Jv
H 0 () = 1
(7.28)
155
exactly, for all . Technically, this follows from the Ward identity (7.25). For
the axial current [1],
s (m)
H5AC (m) = 1 2CF
4
16
8
29
143
82 + I
+ CF CF + 322
+ CA
3
3
9
3
2
s (m)
920
28
+ TF
642 + TF nl
+
9
9
4
(7.29)
(see (4.20)).
= CF
(7.30)
r (r) r1 (r1 ) 1
r + r (3 2) ,
1r
2
(7.31)
156
7 HeavyHeavy Currents
0 = mb mc ,
(7.32)
s (0 )
L
H (0 ) = 1 CF n(n 4)L coth n(3n 10) + (n 2)
2
4
+
(7.33)
(where L = log r). This result reproduces (7.27) at r 1. For the vector
and axial currents, the matching coecients do not depend on :
s (0 )
L
L
H 0 = 1 + 6CF
coth 1
+ ,
2
2
4
L
L 4 s (0 )
H5AC = 1 + 6CF
coth
+
(7.34)
2
2
3
4
When mc mb , the vertex can be expanded in mb mc . Owing to the
mb mc symmetry, the result can be written as a series in = (mb
mc )2 /(mb mc ):
s (0 )
1
1 2
+
+ O(2s ) ,
H 0 = 1 + CF
2
20
4
s (0 )
1 2
1
H5AC = 1 + CF 2 + +
+ O(2s ) .
(7.35)
2
20
4
The twoloop correction can be calculated straightforwardly, because the resulting twoloop integrals are singlemass onshell ones (Sect. 4.2). However,
calculating many terms of such expansions requires ecient computeralgebra
programming. For the vector and axial currents, the twoloop matching coefcients expanded up to 4 have been calculated in [1]. For the vector current,
the expansion starts from O(), in accord with (7.28).
These twoloop matching coecients have also been calculated exactly [2]
(and independently checked in [5]). The results are rather complicated: they
contain trilogarithms. The nl contributions to (7.34) are simple:
s 2
L
L
4
coth 1
,
CF TF nl
3
2
2
4
2
L 44 s 2
CF TF nl 5L coth
;
(7.36)
3
2
3
4
see Sect. 8.8.
157
Until now, we have matched the QCD currents J onto HQET, where both
the b and the c are considered static, in a single step, at a scale somewhere
between mb and mc . This method is good when mc mb : it gives us the exact
dependence of the matching coecients on r = mc /mb . These coecients
are known at two loops; all terms of order (s /)n with n 3 are neglected.
When r 1 (mc mb ), the matching coecients are [2]
s (0 )
H 0 = 1 + CF (3L 6)
4
9
41
2
L + 162
+ CF CF
L 16I 243 + 582 + 29
2
2
5
23
+ CA
42 +
L + 8I 243 2
6
3
26
1147
+ TF 4L2 L 372 +
3
18
2
4
s (0 )
2
+ O(3s ) ,
+ T F nl L +
3
3
4
s (0 )
H5AC = 1 + CF (3L 8)
4
9
53
46
154
32
2
L + 162
+ CF CF
L I 243 + 2 +
2
2
3
3
3
49
332
16
42 +
L + I 243 + 152
+ CA
6
3
9
79
971
+ TF 4L2 14L 2 +
3
18
2
88
s (0 )
10
+ O(3s ) ,
(7.37)
+ T F nl L +
3
9
4
where L = log r, nl includes neither the b nor the c, and s (0 ) has nf =
nl + 2 avours (this quantity is somewhat articial because 0 is below mb ).
We may want to sum leading powers of L. In other words, when the scales mb
and mc are widely separated, it is desirable to take into account the running
of s () and of the currents between these two scales correctly.
This can be achieved by a twostep matching. First we match J onto the
eective theory in which the bquark is considered static while the cquark is
still dynamic (HQET1) at the scale = mb :
j() +
J(mb ) = C (mb , )
1
Bi (mb , )i () + ,
2mb i
(7.38)
where
j = c b. Some of the dimension4 operators i have matrix elements
proportional to mc , so that this is an expansion in r = mc /mb . There are
ni = nf 1 = nl + 1 avours in this intermediate theory. Then we match it
158
7 HeavyHeavy Currents
onto the eective theory in which the c is also considered static (HQET2)
at the scale = mc :
j(mc ) = E(mc )J ,
... ,
(7.39)
c b, and terms of the order QCD /mc are neglected. Therefore,
where J =
r
Bi (mb , mc )Fi (mc ) + (7.40)
H (mb , mc ) = C (mb , mc )E(mc ) +
2 i
The matching coecients C (mb , mb ) for all currents have been considered in Sect. 5.6. The running is given by (5.58):
C (mb , )
j0
= C (mb , mb )x
x=
s ()
s (mb )
j0
1 +
20
1/(20 )
j1
1
j0
0
,
(7.41)
/v (/
k + mc /v + mc )
dd k
2
1 (mc v, 0) = 1 iCF g0
.
(2)d [m2c (k + mc v)2 ] (k v)(k 2 )
We may replace /k by k v /v , because the integral of k vanishes. In the
integrals
dd k
,
n
2
2
[mc (k + mc v) ] 1 (k v)n2 (k 2 )n3
the denominators are linearly dependent. We can multiply the integrand by
159
m2c (k + mc v)2 + 2mc k v
1=
k 2
until one of the quark denominators disappears. Integrals without the massivequark denominator are scalefree and vanish; those without the HQET denominator have been discussed in Sect. 4.2. We obtain
g 2 m2
d1
1 = 1 CF 0 cd/2 ()
,
(d
3)(d 5)
(4)
(7.42)
or
1/2
(Zcos )
g 2 m2
1
d1
1 = 1 CF 0 cd/2 ()
.
2
d5
(4)
s (mc )
+
4
(7.43)
s (mb )
(mc )
H 0 (mb , mc ) = xj0 1 2CF
4CF s
4
4
j1
s (mb ) s (mc )
j0
+ ,
+
1
20
j0
0
4
s (mb )
(mc )
H5AC (mb , mc ) = xj0 1 4CF
4CF s
4
4
j0
j1
s (mb ) s (mc )
+ .
+
1
20
j0
0
4
(7.44)
If we express s (mc ) via s (mb ) with twoloop accuracy using the ni avour
function, then use s (mb ) = s (mb ) + O(3s ), and nally express s (mb )
via s (0 ) using the nf avour function, then we reproduce all leading and
nexttoleading logarithms in (7.37), i.e., the oneloop term and the L2 and
L terms at two loops.
Now we shall discuss the O(r) term in (7.40) in the LLA [3]. It is sucient
to know the matching coecients Fi (mc ) at the tree level. In the case of spin0 currents (Sect. 6.4), the only operators producing O(mc ) contributions are
2,3 . In 3 , we may replace the MS mass mc (mc ) by the pole mass mc :
3 (mc ) = mc
j(mc ) .
(7.45)
160
7 HeavyHeavy Currents
2 (mc ) = mc
j(mc ) .
(7.46)
m
m
a0
+ 3
b0
) m0
m0 (
r
k
x
log x +
1
a0
= xj0
2
m0 (
m0 m0 )
m
m
k
a0 + 3
b0
b0
m0
+
1)
(x
+
m0
m0
m0
3/25
5 32
8
s (mb )
= ry 2 log y + y 3 3y 4 , y =
. (7.47)
3
9
9
s (mc )
i4 (mc ) = i5 (mc ) = mc
ji (mc ) .
(7.48)
B3,4,5
(mb , mc )
The coecients
in the LLA are given by (6.76) (with the
upper sign), and the O(r) term in H5AC (mb , mc ) is
i
r i
B4 (mb , mc ) B5 (mb , mc )
2
m
m
r
1
m + 3
a0
b0
k
a0
xm0 1
= xj0
log x +
+ a0
2
3 m0
m0
m0
m
m
1
a0
1
a0
a0
m
k
2
+
+
b0 b0
(xm0 1)
3
m0 (
m0
a0 ) 3
m0 m0
m0
m0
3
2 14
8
8
= ry 2 log y + y 3 + y 4 .
(7.49)
3
9 27
27
How can results of these two approaches be combined? We start from the
singlestep matching results (7.34), with the twoloop corrections [2]. These
results are expressed via s (0 ), and contain all powers of r. Then we take the
twostep matching results (7.44), and subtract from them the terms of order
s (0 ) and 2s (0 ), which are already accounted for in the singlestep results.
By adding this dierence, we take into account an innite sequence of radiative corrections of order O(r0 ) with leading and nexttoleading powers of
the logarithm L. Similarly, we take the O(r) twostep matching results (7.47)
and (7.49), and subtract from them the terms of order s (0 ) and 2s (0 ),
which are already accounted for. By adding this dierence, we take into account corrections of order O(r) with leading powers of L at all orders in
s .
161
,
(a2 k 2 )3
k = k + m(xv + x v ) ,
a2 = m2 (xv + x v )2 = m2 (x2 + x2 + 2xx cosh ) .
We calculate the loop integral using (2.16), and substitute x = (1 + z)/2
and x = (1 z)/2; the integration is trivial (a2 = m2 2 a+ a , a =
cosh(/2) z sinh(/2)):
g 2 m2
= CF 0 d/2 (1 + )
(4)
+1
a+ a h 2
(1 z 2 )h
cosh
1
dz
+
+
1+
2(1 )
8(1 )
(1 2) 1 2
1 (a+ a )
+1
(1 + z 2 )h
1
dz
1
+
(/
v + /
v)
2 1 (a+ a )1+ 4(1 )
1 2
+1
2
h
dz (1 z )
.
(7.50)
/
v /
v
8(1 ) 1 (a+ a )1+
Here the function h(d) (5.5) is given by (7.26). The checks with ZP , ZA , and
ZT are passed as for = 0. At = 0, a+ = a = 1, and (7.23) is reproduced.
It is not dicult to calculate the integrals with the accuracy needed for
the terms nite at 0:
+1
(cosh + 1)
dz
2 + O(2 ) ,
= 2 2
sinh
1 (a+ a )
+1
dz
+ F ()
+ O(2 ) ,
=2
1+
sinh
1 (a+ a )
F () = Li2 (e ) Li2 (e ) log (2 (cosh + 1)) ,
+1
2
z 2 dz
4
1
+
+ O() .
=
1+
(a
a
)
cosh
1
sinh
sinh
+
1
(7.51)
162
7 HeavyHeavy Currents
2
2
.
2
6
os
It is most natural to perform matching at = = m. Using ZQ
(4.29),
J (7.6), we see that the 1/ poles cancel in Hi :
ZJ (5.8), and Z
s (m)
cosh + 1
H1 = 1 + H4 + CF
3
(n 2)2 + 2(n 2)
4
sinh
2
+
+ 4( coth 1) + 2F () coth ,
sinh
(n 2)
s (m)
H2 = H3 = H4 CF
+1 ,
4 sinh
2
(n 2)
s (m)
H4 = CF
1
.
(7.52)
4
sinh
2(cosh 1)
s (m)
3
+ 3 coth 4 + F () coth ,
H1V = 1 + CF
4
sinh
(m)
s
H2V = H3V = CF
,
4 sinh
s (m) 4
,
H1A = H1V CF
4 sinh
2
3
s (m)
A
A
H2 = H3 = CF
1 +
. (7.53)
4
cosh 1 sinh
sinh
(7.54)
163
dd k
(2)d
k + mc (1 x )/
v mb x/v + mc ) (/
k + mb (1 x)/
v mc x /v + mb )
(/
,
(a2 k 2 )3
k = k + mb xv + mc x v ,
= 2iCF g02
dx dx
a = cosh
L
L
+ z sinh
.
2
2
+
1+ 2(1 )
(a
a
)
8(1
)
(1
2)
+
1
cosh L + z sinh L
+
1 2
L
+1
e (1 + z)2 h
1z
dz
1
+
/
v
2 1 (a+ a )1+
4(1 )
1 2
+1
L
2
e (1 z) h
1+z
dz
1
+
/
v
2 1 (a+ a )1+
4(1 )
1 2
+1
2
h
dz (1 z )
.
/
v /
v
8(1 ) 1 (a+ a )1+
= CF
(7.55)
Here the function h(d) (5.5) is given by (7.26). The checks with ZP , ZA , and
ZT are passed as before. At mc = mb (L = 0), this expression coincides
with (7.50). At = 0,
L
L
+ z sinh ,
2
2
and (7.30) is reproduced.
The case q 2 = 0 ( = L) is also easy:
a+ = a = cosh
a+ = cosh L + z sinh L ,
a = 1 ,
(1 ) sinh L
1 a+
+1
2L
dz
2 sinh L
2
1+ = sinh L = sinh L + O( ) ,
a
1
+
(7.56)
164
7 HeavyHeavy Currents
+1
1
+1
sinh(1 )L cosh L sinh L
z dz
2
=
1
a1+
sinh2 L
+
2
(L coth L 1) + O() ,
=
sinh L
2
sinh(2 )L 2 cosh L sinh(1 )L
z 2 dz
=
3
2
1
a1+
sinh L
+
2
cosh L sinh L
+
2 cosh L
=
(L coth L 1) + O() .
sinh2 L
s (0 )
(2 L coth L)(n 2)2 + 2(n 2)
4
+ 2(2L coth L 1) ,
2
+ O(2 ) ,
=
2
cosh L cosh
1 (a+ a )
+1
dz
+ F (L, )
+ O(2 ) ,
=2
1+
(a
a
)
sinh
1
L
e eL
e e
F (L, ) = Li2
Li2
e e
e e
sinh[(L + )/2]
,
+ (L + ) log
sinh
+1
z dz
L sinh sinh L
+ O() ,
=2
1+
sinh (cosh L cosh )
1 (a+ a )
+1
2
z 2 dz
sinh2 L + sinh2
=
1+
cosh L cosh sinh cosh L cosh
1 (a+ a )
2L sinh L
+ 2 + O() .
cosh L cosh
165
i
Hi (, ) = H
s ()
s (0 )
s ( )
s (0 )
jn0 /(20 )
J0 /(20 )
Kjn (s ())
K
J (s ( )) ,
i = Hi (0 , 0 )Kjn (s (0 ))K ( (0 ))
H
J
s
(7.58)
os
J (7.6), we see that the 1/
(see (4.72)). Using ZQ
(4.29), ZJ (5.8), and Z
poles cancel in Hi (0 , 0 ):
s (0 )
L sinh L sinh
H1 = 1 + H4 + CF
3
(n 2)2
4
cosh L cosh
L sinh L sinh
+ 2 coth 4 + 2F (L, ) coth ,
+ 2(n 2) + 2
cosh L cosh
s (0 )
sinh L + cosh eL cosh2
H2 = H3 (L L) = CF
4
sinh (cosh L cosh )
(n 2)
eL cosh 1
+L
eL
cosh L cosh
2(cosh L cosh )
1
cosh eL
+
+L
,
sinh
cosh L cosh
s (0 )
(cosh L cosh 1) L sinh L sinh
1+
H4 = CF
4
sinh (cosh L cosh )
(n 2)
.
(7.59)
2(cosh L cosh )
4
sinh (cosh L cosh )
L sinh L
4 + 2F (L, ) coth ,
+
cosh L cosh
166
7 HeavyHeavy Currents
sinh L
s (0 )
(eL 1) sinh L
L
=
L) = CF
4
(cosh L cosh )2
sinh
1
(2eL + 3 + 3eL )(eL 1)
L
L
+
+ L(e + 2) + e 1
cosh L cosh
2 sinh
(eL 2)
+
,
sinh
s (0 ) 4
,
H1A = H1V CF
4 sinh
sinh L
s (0 )
(eL + 1) sinh L
H2A = H3A (L L) = CF
L
4
(cosh L cosh )2
sinh
L
L
L
1
(2e 3 + 3e )(e + 1)
L
L
+
+ L(e 2) + e + 1
cosh L cosh
2 sinh
(eL + 2)
+
.
(7.60)
sinh
H2V
H3V (L
The easiest way to see that H1V (7.60) (with two dilogarithms) is indeed
equivalent to the result in [10] (with three dilogarithms) is to subtract the two
results, dierentiate the dierence with respect to and L (these derivatives
simplify to 0), and compare the values at 0 and L 0.
As discussed in Sect. 7.3, the singlestep matching of QCD currents onto
HQET with both the b and the c considered static is good when r 1, because it gives the exact dependence of the matching coecients (for example,
at one loop, as in (7.59) and (7.60)) on r. At r 1, it is better to use the
twostep matching. We are considering the QCD current J = c b, where
satises (5.5) ( may be the antisymmetrized product of n matrices,
possibly multiplied by 5AC ; see (7.26)). First we match it onto HQET1,
the eective theory with a static b and dynamic c, at mb . To this end, we
decompose into components (6.3) commuting and anticommuting with v/:
J(mb ) = C+ (mb , )
j+ () + C (mb , )
j ()
1
Bi (mb , )i () + ,
+
2mb i
j = c
1/
v
bv
2
(7.61)
(7.62)
167
H1,2 (mb , mc ) =
/v (/
k + mc /v + mc )
dd k
2
1 (mc v , 0) = 1 iCF g0
.
(2)d [m2c (k + mc v )2 ] (k v)(k 2 )
Using both the usual (2.13) and the HQET (2.23) Feynman parametrization,
this vertex can be rewritten as
k + mc (1 x)/
v y/v + mc )
dd k /v (/
1 + 4iCF g02 dx dy
,
d
2
(2)
(k + a2 )3
k = k + mxv + yv ,
a2 = (mxv + yv)2 .
dy
0
dx
0
y 2mc (1 x) cosh mc x
.
(y 2 + m2c x2 + 2mc xy cosh )1+
z + 2 cosh 1 cosh
+
.
1 2
168
7 HeavyHeavy Currents
2
(mc ) 2 coth + 1
= 1 + CF s
+ 2 coth + 2
4
sinh
+ Li2 (1 e2 ) Li2 (1 e2 ) coth .
j )(Zcos )1/2 are nite [10]:
The matching constants E = (ZJ /Z
(mc )
2
E (mc ) = 1 + CF s
2 coth
4
sinh
+ Li2 (1 e2 ) Li2 (1 e2 ) coth .
(7.64)
The upper sign is for commuting with v/, and the lower sign is for
anticommuting with v/. At = 0, E+ (mc ) reproduces (7.43).
169
Collecting all the pieces together, we have the following [10] for
at O(r0 ):
(mb )
V
j0
H1 = x
1 4CF s
4
s (mc )
+
1
,
20
j0
0
4
s (mb )
2 s (mc )
(mb )
H1A = xj0 1 4CF s
4
s (mc )
j0
1
+
,
20
j0
0
4
s (mb )
2 s (mc )
+
(7.65)
, H3A = 0 .
H2A = 2CF xj0
4
sinh 4
HiV,A (mb , mc )
1
Li2 1 e2 Li2 1 e2 + L + O(r) ,
2
r = eL .
Bi (mb , )i () +
B (mb , )i () .
v
2mb
i
i
Only 2,3 (6.43) and
3,4,5 (6.66) produce O(mc ) contributions; at the tree
cosh ,
,
5 (mc ) = mc J (mc ) J(mc )v
(7.66)
170
7 HeavyHeavy Currents
where J =
cv bv , J =
cv bv . For the axial current J = c5AC b, the
operators i and
are
obtained from (6.43) and (6.66) by c c5AC and
i
mc mc , and the matching coecients remain unchanged:
AC v
/
Bi 5
= Biv/ ,
AC
Bi 5
= Bi .
Therefore,
H1V,A (mb , mc ) H1V,A (mb , mc )
O(r 0 )
r
B4 (mb , mc ) cosh B5 (mb , mc )
=
2
5 5 2
2 3
2 4
8
2
= ry
log y + y y + y cosh
3
9 9
27
27
1 5 2 4 3 2 4
y + y y
,
3 9
9
9
H2V,A (mb , mc ) H2V,A (mb , mc ) 0
O(r )
r v/
2 4
4 10 2 44 3
2
y + y y cosh
= ry
9
9
27
27
5 2 20 3 25 4
y + y y
,
9
9
9
r
5 2 2 3
V,A
2
H3 (mb , mc ) = B3 (mb , mc ) = ry
1 + y y
2
3
3
(7.67)
(see (7.47)). At 0, the O(r) term in H1V + H2V + H3V (7.47) reproduces (7.47), and H1A (7.47) reproduces (7.49).
The results of these two approaches can be combined [10]. We take the
singlestep matching results (7.58) and (7.60), express them via s (0 ), and
retain only terms up to s (0 ). These results contain all powers of r. Then we
take the twostep matching results (7.65), and subtract from them the terms
of order s (0 ), which are already accounted for. By adding this dierence,
we take into account an innite sequence of radiative corrections of order
O(r0 ) with leading and nexttoleading powers of the logarithm L. Similarly,
we take the O(r) twostep matching results (7.67), and subtract from them
the terms of order s (0 ), which are already accounted for. By adding this
dierence, we take into account corrections of order O(r) with leading powers
of L at all orders in s .
171
iD
,
mb
V + = v
iD
.
mc
/
iD
/
bV = 1 + iD
cV =
cv 1
.
bv ,
2mb
2mc
(7.68)
(7.69)
iv D iv D
.
mb
mc
(7.70)
+ 2 (H1 + H2 /v + H3 /v + H4 /v /v ) iv D bv
1
/ (H1 + H2 /v + H3 /v + H4 /v /v )
cv (i) D
2mc
/
+ 2 (H3 + H4 /v ) (i) D
+ 2(i)v D (H1 + H2 /v + H3 /v + H4 /v /v ) bv ,
(7.71)
172
7 HeavyHeavy Currents
cv Dbv ,
cv D bv are not reparametrizationinvariant sepaerators of the form
rately, and can only appear in this combination.
In particular, for the vector current (7.22), we have [11]
J = H1V J1 + H2V J2 + H3V J3
1 V
V
V
H + H2V
+
2 + H3 3 + 2H2 4
2mb 1 1
V
V
+ 2H1V
5 + 2H2 6 + 2H3 7
1 V
V
V
H + H2V
+
2 + H3 3 + 2H3 4
2mc 1 1
V
V
+ 2H1V
5 + 2H2 6 + 2H3 7
+ (bilocal terms) + O(1/m2c,b ) ,
(7.72)
where
J = c b ,
J1 = cv bv , J2 =
cv v bv , J3 =
cv v bv ,
=
cv iD
/ bv , =
cv v iD
/ bv , =
cv v iD
/ bv ,
1
cv iDbv
=
cv iv Dbv = v J1 ,
cv (i) D
/ bv ,
cv (i) Dv
/ bv ,
1 =
2 =
cv (i) D bv ,
4 =
cv (i) D
/ v bv ,
3 =
References
173
1
1
V
H1 (cosh ) J1 +
+
2mb 2mc
2 + 24
2
V
+ H2 (cosh ) J2 +
+
2mb
2mc
3
3 + 2
4
V
+ H3 (cosh ) J3 +
+
2mb
2mc
+ (bilocal terms) + O(1/m2c,b ) ,
(7.74)
References
A. Czarnecki: Phys. Rev. Lett. 76, 4124 (1996) 155, 156
A. Czarnecki, K. Melnikov: Nucl. Phys. B 505, 65 (1997) 156, 157, 160
A.F. Falk, B. Grinstein: Phys. Lett. B 247, 406 (1990) 159, 169
A.F. Falk, B. Grinstein: Phys. Lett. B 249, 314 (1990) 161, 162
J. Franzkowski, J.B. Tausk: Eur. Phys. J. C 5, 517 (1998) 156
J. Frenkel, J.C. Taylor: Nucl. Phys. B 246, 231 (1984) 149
J.G.M. Gatheral: Phys. Lett. B 133, 90 (1983) 149
G.P. Korchemsky: Mod. Phys. Lett. A 4, 1257 (1989) 150
G.P. Korchemsky, A.V. Radyushkin: Nucl. Phys. B 283, 342 (1987) 145, 150,
151
10. M. Neubert: Phys. Rev. D 46, 2212 (1992) 162, 165, 166, 168, 169, 170
11. M. Neubert: Phys. Lett. B 306, 357 (1993) 171, 172
12. A.M. Polyakov: Nucl. Phys. B 164, 171 (1980) 146
1.
2.
3.
4.
5.
6.
7.
8.
9.
8 Renormalons in HQET
It is well known that perturbative series do not converge. They are asymptotic
series, i.e., the dierence between the exact result and its approximation up
L
to the order L
s , divided by s , tends to 0 in the limit s 0. The largeorder
behaviour of various perturbative series has attracted considerable attention
during recent years. Most of the results obtained so far are modeldependent:
they are derived in the large0 limit, i.e., at nf . There are some
hints that the situation in real QCD may be not too dierent from this limit,
but this cannot be proved. However, a few results are rigorous consequences
of QCD. They are based on the renormalization group; see [14, 4]. In this
chapter, we shall discuss some simple technical methods used for calculations
in the large0 limit. Several applications in HQET will be considered in
detail. Many more applications are discussed in the excellent review [2], where
additional references can be found.
L1
aLn nnf
L=1 n=0
g02
(4)d/2
L
,
(8.1)
L1
L=1 n=0
aLn 0n
g02
(4)d/2
L
.
(8.2)
176
8 Renormalons in HQET
This regime is called large0 limit, and can only hold in QCD with nf .
Note that it has nothing in common with the largeNc limit, because we
cannot control the powers of Nc in the coecients aLn .
There is some empirical evidence [7] that the twoloop coecients a21 0 +
a20 for many quantities are well approximated by a21 0 . It is easy to nd a21
from a diagram with a quarkloop insertion into a gluon propagator in the
oneloop correction. Then we can estimate the full twoloop coecient as
a21 0 = a21 (nf (11/4)(CA /TF )). This is called naive nonabelianization [7].
Of course, there is no guarantee that this will hold at higher orders. We can
only hope that higher perturbative corrections are mainly due to the running
of s ; in this respect, gluonic contributions behave as 33/2 avours, and
QCD with nf = 3 or 4 avours is not too dierent from QCD with
avours.
It is easy to nd the coecients aL,L1 of the highest degree, 0L1 ,
at L loops. They are determined by the coecients aL,L1 of nfL1 , i.e.,
by inserting L 1 quark loops into the gluon propagator in the oneloop
correction. We shall assume for now that there is only one gluon propagator
and there are no threegluon vertices at one loop. The bare gluon propagator
with L 1 quark loops inserted is (see (3.27))
1
p p
(L1)
D
(p) =
+
g
p2
(p2 )1+(L1)
L1
4
g02 D()
T F nf
e
,
3
(4)d/2
5
(8.4)
D() = 6e (1 + )B(2 , 2 ) = 1 + +
3
This propagator looks like the free propagator in the Landau gauge a0 = 0,
with a shifted power of p2 (and an extra constant factor). Let the oneloop contribution to A0 be (a1 + a1 a0 ) g02 /(4)d/2 . If we calculate the oneloop contribution in the Landau gauge a1 with the denominator of the gluon
propagator equal to (p2 )n instead of just p2 and call it a1 (n), then for all
L > 1,
aL,L1 =
D()
e
L1
a1 1 + (L 1) .
(8.5)
3 2
P = 0
(a0 + 3) 1 (a0 + 3)
8 +
(8.6)
3
3
2(1 )
2
177
0 s
4
(8.7)
d log Z
=
d
+
(8.8)
1
.
1 + /
(8.9)
2
.
0 log(/MS )
(8.10)
The perturbative series (8.2) can be rewritten (in the Landau gauge) via
the renormalized quantities:
L
1 F (, L)
A0 = 1 +
+O
,
(8.11)
0
L
+
02
L=1
where
F (, u) = ue a1 (1 + u )2u D()u/1 .
(8.12)
If a = 0, the term a1 a/0 +O(1/02 ) should be added (the dierence between
a0 and a is O(1/0 )).
We can expand (8.11) in the renormalized s , or in (8.7), using
L L
2
3
(L)2
(L)3
+
1L +
+
3!
178
8 Renormalons in HQET
F (, u) =
Fnm n um ;
(8.13)
n=0 m=0
dZ1
d log
(see (3.15)). Collecting terms with 1 in the quadruple sum for A0 , we obtain
for 0 Z1
F00 2 (F10 + F01 ) + 3 (F20 + F11 + F02 )
4 (F30 + F21 + F12 + F03 ) +
1
+ 2 (F10 + 2F01 ) 3 (F20 + 2F11 + 4F02 )
2
3
+ 4 (F30 + 2F21 + 4F12 + 8F03 ) +
2
1 3
+ (F20 + 3F11 + 9F02 ) 4 (F30 + 3F21 + 9F12 + 27F03 ) +
3
1
+ 4 (F30 + 4F21 + 16F12 + 64F03 ) +
4
+
= F00
2
3
4
F10 +
F20
F30 +
2
3
4
1
= 2 F (, 0) + O
.
0
02
(8.14)
179
1
+ 3 (F30 + 3F21 + 9F12 + 27F03 )
3
4 (F40 + 3F31 + 9F22 + 27F13 + 81F04 ) +
1
+ 4 (F40 + 4F31 + 16F22 + 64F13 + 256F04 ) +
4
+
2
3
4
F20 +
F30
F40 +
2
3
4
+ F01 + 2 F02 + 2 3 F03 + 6 4 F04 +
= F10
1
1
u/ F (0, u) F (0, 0)
+
+O
du e
,
0 0
u
02
A() = 1 +
(8.15)
where = 0 s ()/(4).
The renormalizationgroup equation
(s )
d log A()
=
d log s
2(s )
can be conveniently solved as
A() = A
0 /(20 )
s ()
s (0 )
K (s ())
(8.16)
d
(s )
F (, 0) F (0, 0)
.
Therefore,
1
A = 1 +
0
du e
0
u/(s (0 ))
1
F (0, u) F (0, 0)
+O
.
u
02
0
180
8 Renormalons in HQET
F (0, u) =
e5/6
m
2u
F (u) ,
(8.17)
(8.18)
F (u) F (0)
.
u
(8.20)
Here,
eu/ =
e5/6 MS
m
2u
.
(8.21)
sL uL1
L=1
1
1
A = 1 +
cL L + O
,
(8.22)
0
02
L=1
L1
d
S(u)
.
(8.23)
cL = (L 1)! sL =
du
u=0
cL uL1
,
(L 1)!
(8.24)
L=1
8.2 Renormalons
181
8.2 Renormalons
The Laplace integral (8.19) is not well dened if the Borel image S(u) has
singularities on the integration path the positive halfaxis u > 0. At the
rst order in 1/0 , S(u) typically has simple poles. If
S(u) =
r
+ ,
u0 u
(8.25)
where the dots mean terms regular at u = u0 , and u0 > 0, then the integral (8.19) is not well dened near u0 . One way to make sense of this integral
is to use its principal value: to make a hole [u0 , u0 + ] and take the limit
0. However, if we make, for example, a hole [u0 , u0 + 2] instead, we
shall obtain a result which diers from the principal value by the residue of
the integrand times log 2. Therefore, the sum of the perturbative series (8.19)
contains an intrinsic ambiguity of the order of this residue. This ambiguity
is equal to
r
reu0 /
=
A =
0
0
e5/6 MS
m
2u0
.
(8.26)
(L 1)!
uL
0
(8.27)
(see (8.23)). The series (8.25) is, clearly, divergent. Using the Stirling formula
for the factorial, we can see that the terms of this series behave as
L
cL r
L
eu0
L
at large L. The best one can do with such a series is to sum it until its
minimum term, and to assign it an ambiguity of the order of this minimum
term. The minimum happens at L u0 / loops, and the magnitude of
the minimum term is given by (8.26). This is another way to look at this
182
8 Renormalons in HQET
1
s ( 0 )
d
A = 1 +
w( )
+O
.
(8.28)
2
0
0
This looks like the oneloop correction, but with the running s under the
integral sign. The function w( ) has the meaning of the distribution function
of gluon virtualities in the oneloop diagram; it is normalized to the coecient
of s /(4) in the oneloop correction. Inside the 1/0 term in (8.28), we may
use the leadingorder formula for the running of s :
s ( 0 ) =
s (0 )
= s (0 )
( log )n .
1 + log
n=0
Substituting this expansion into (8.28), we see that this representation holds
if w( ) is related to the coecients of the perturbative series cL by
d
w( )( log )L1 .
cL =
(8.29)
0
Therefore, S(u) (8.24) becomes
S(u) =
d
w( ) u .
183
(8.30)
In other words, S(u) is the Mellin transform of w( ). Therefore, the distribution function w( ) is given by the inverse Mellin transform:
1
w( ) =
2i
u0+i
du S(u) u ,
(8.31)
u0 i
/ (/
k+p
/)
dd k 14 Tr p
(2)d [(k + p)2 ] (k 2 )n
k k
g +
.
k 2
184
8 Renormalons in HQET
(1 + u) (1 u) (2 )
D()u/1 N (, u) .
F (, u) =
e
p2
(2 + u ) (3 u )
(8.34)
The rst function in the numerator, with a positive sign in front of u,
comes from the rst function in the numerator of (2.18), with a negative
sign in front of d, and its poles are UV divergences. The second function
in the numerator, with a negative sign in front of u, comes from the second
function in the numerator of (2.18), with a positive sign in front of d, and
its poles are IR divergences. For V (p2 ), we obtain
N (, u) = CF (3 2)(u ) .
(8.35)
+O
=
.
30 B(2 + , 2 + ) (3 + ) (1 )
02
In the general covariant gauge, the oneloop term proportional to a (3.50)
should be added:
s
(1 + (2/3))
1
2a +
+O
q = CF
4
B(2 + , 2 + ) (3 + ) (1 )
02
35 2
s
5
= CF
.
(8.36)
2a + 3 1 + +
4
6
36
This perturbative series for q has a radius of convergence equal to the distance from the origin to the nearest singularity, which is situated at = 5/2;
in other words, it converges for  < 5/2. It reproduces the leading 0 terms
in the twoloop result (3.57).
The renormalized expression for p/S(p) is given by (8.15). If we factor out
its dependence as in (8.16), then the corresponding renormalizationgroup
invariant is given by (8.19) with
N (0, u)
N (0, 0)
1
3CF
S(u) =
=
u (1 + u)(1 u)(2 u)
2
(1 + u)(1 u)(2 u)
185
(8.37)
(here p2 plays the role of m). The pole at u = 1 comes from the rst
function in the numerator of (8.34), and is a UV renormalon; the poles at
u = 1, 2 come from the second function, and are IR renormalons (Fig. 8.1a).
We can also see this from power counting (Sect. 8.2). The lightquark selfenergy seems to have a linear UV divergence. However, the leading term of
the integrand at k , k
/ /(k 2 )2 , yields 0 after integration, owing to Lorentz
invariance. The actual UV divergence is logarithmic, UV = 0, and UV renormalons can exist only at u 0. The UV divergence at u = 0 is removed by
renormalization, and the UV renormalons are at u < 0. The index of the IR
divergence of the selfenergy, like that of any oshell quantity, is IR = 2,
and the IR renormalons are at u 1. The power corrections to the lightquark propagator form an expansion in 1/(p2), therefore, IR renormalons
can appear only at positive integer values of u. For gaugeinvariant
quantities,
the rst power correction contains the gluon condensate G2 of dimension 4,
and the rst IR renormalon is at u = 2. The quark propagator is not gaugeinvariant, and a renormalon at u = 1 is allowed. The virtuality distribution
function (8.31) is
1
12 , < 1 ,
w( ) = 3CF 21 1 3
,
>1
6
(Fig. 8.1b).
2
b
Now we shall discuss lightquark currents (5.1). By repeating the calculation of the vertex function (p, 0) = b (p2 ) (Sect. 5.3) with the denominator of the gluon propagator raised to the power n = 1 + (L 1), we obtain
b (p2 ) in the Landau gauge; it has the form of (8.11) and (8.34), with
Nb (, u) = CF [2 u + 2h(u h)] ,
(8.38)
186
8 Renormalons in HQET
d log Z n
+ q ,
d log
+ .
= 2CF (n 1) n 3 +
4
6
12
(8.39)
This perturbative series converges for  < 5/2. It reproduces the leading 0
terms in the twoloop result (5.9). In particular,
1
1 + (2/3)
s
+O
m = j0 = 2CF
.
4 B(2 + , 2 + ) (3 + ) (1 )
02
(8.40)
The expressions for ZP,A (Sect. 5.3) in the large0 limit can be obtained
from (5.39):
d
4 CF
3 0 0 B(2 + , 2 + ) (3 + ) (1 )
13 2
s
1
= 1 4CF
1 + + ,
4
12
36
ZA = 1
2
.
ZP = ZA
(8.41)
iCF
k k
dd k
=
+
g
.
2
(2)d k v + (k 2 )n
k 2
(4)d/2
187
u (1 + 2u) (1 u)
D()u/1 N (, u) . (8.42)
F (, u) =
e
2
(2 + u )
The rst function in the numerator, with a positive sign in front of u,
comes from the rst function in the numerator of (2.27), with a negative
sign in front of d, and its poles are UV divergences. The second function
in the numerator, with a negative sign in front of u, comes from the second
function in the numerator of (2.27), with a positive sign in front of d, and
its poles are IR divergences. For ()/,
we obtain
N (, u) = 2CF (3 2) .
(8.43)
+O
.
=
60 B(2 + , 2 + ) (2 + ) (1 )
02
In the general covariant gauge, the oneloop term proportional to a (3.67)
should be added:
1
s
1 + (2/3)
Q = CF
2a
+O
4
B(2 + , 2 + ) (2 + ) (1 )
02
s
10
(8.44)
= CF
2(a 3) 8 + 2 + .
4
3
This perturbative series converges for  < 5/2. It reproduces the leading 0
terms in (3.75).
The renormalized expression for S()
is given by (8.15). If we factor out
its dependence as in (8.16), then the corresponding renormalizationgroup
invariant is given by (8.19) with [3]
N (0, 0)
(1 + 2u) (1 u)
N (0, u) +
(2 + u)
2u
(1 + 2u) (1 u)
1
= 6CF
+
(2 + u)
2u
S(u) =
(8.45)
188
8 Renormalons in HQET
(here 2 plays the role of m). The rst function, with a positive sign
in front of u, produces UV renormalons, while the second one, with a negative sign, produces IR renormalons (Fig. 8.2a). We can understand this from
power counting (Sect. 8.2). The staticquark selfenergy has a linear UV divergence which is not nullied by Lorentz invariance: UV = 1. This is the
same divergence as that of the Coulomb energy of a point charge in classical
electrodynamics. Therefore, the UV renormalons are situated at u 1/2.
The index of the IR divergence of the selfenergy, like that of any oshell
quantity, is IR = 2, and the IR renormalons are at u 1.
1
b
Fig. 8.2. Renormalons in the oshell HQET selfenergy (a) and in the onshell
heavyquark selfenergy (b)
MS
.
0
(8.46)
The structure of the leading UV renormalon at u = 1/2 can be investigated beyond the large0 limit [1]. The renormalizationgroup invariant
corresponding to S()
is now written as
1
4
u
(8.47)
1+
du S(u) exp
0 0
0 s (0 )
instead of (8.19), where the exact s is used in the exponent, 0 = 2e5/6 ,
and O(1/02 ) is absent. The singularity of S(u) at u = 1/2 becomes a branch
point, so that
S(u) =
r
1+a
(1/2 u)
+ ,
189
with a cut from 1/2 to +, instead of a simple pole. The renormalon am
biguity of ()/
is dened, as before, as the dierence between the integrals (8.47) below and above the real axis divided by 2i:
r 1
du
4
u
=
exp
0 2i C (1/2 u)1+a
0 s (0 )
a
r
2
0 s (0 )
=
(2) exp
(8.48)
20 (1 + a)
0 s (0 )
4
(Fig. 8.3; we have used (a) (1 + a) = / sin(a)). But this must be just
some number times MS , and cannot depend on !
1
2
We have to use a formula for s () that is more precise than the oneloop
one (8.10). The renormalizationgroup equation (3.5) is solved by separation
of variables:
2
ds
1 s
2
+
O(
)
=
d log ,
1
s
0
2s
0 4
1
2
s ()
+
+ O(s ) = log
log
,
0 s () 202
4
MS
and hence
MS
= exp
2
0 s ()
s ()
4
1 /(202 )
[1 + O(s )] .
(8.49)
0 = 2CF e5/6
MS
.
0
(8.50)
190
8 Renormalons in HQET
S(u) =
where
N0
4CF N0
1+1 /(202 )
(1/2 u)
[1 + O (1/2 u)] ,
1
/(202 )
= N 0 1 + 2 0 1
.
20
(8.51)
(8.52)
The result for the power is exact; the normalization cannot be found within
this approach.
k0 (Fig. 4.2) can also be
The selfenergy with a kineticenergy insertion
easily calculated in the large0 limit. In the Landau gauge, raising the gluon
denominator in (4.6) to the power n = 1 + (L 1), we obtain (8.42) with
N (, u) = 2CF (3 2)2 2 ,
(8.53)
and hence
k0 () = 3 .
(8.54)
This leads to a UV renormalon ambiguity of the heavyquark eld renormalization constant [9],
ZQ =
3
.
2 m
(8.55)
Let us also discuss the heavylight current in HQET. If the light quark
is massless, we may take (1/4) Tr of any diagram for (, 0). All diagrams
with insertions into the gluon propagator of the oneloop diagram (Fig. 5.3b),
as well as this oneloop diagram itself in the Landau gauge, vanish owing to
the transversality of the gluon propagator; see (5.55). Therefore, to the rst
order in 1/0 , (, 0) = 1, and
j = (
Q + q )/2 in the Landau gauge. This
anomalous dimension is gaugeinvariant, and [7]
1
1 + (2/3)
s
+O
j = CF
4 B(2 + , 2 + ) (3 + ) (1 )
02
35
s
5
= 3CF
(8.56)
1 + 2 + ,
4
6
36
from (8.36) and (8.44). This perturbative series converges for 0 s  < 4. It
reproduces the leading 0 terms in (5.45). Note that
j = j0 /2 (8.39) at the
rst order in 1/0 .
Finally, we discuss the heavyheavy current in HQET. We consider the
vertex (7.3) at = = 0, integrated over the region t + t < T (T acts as
an IR cuto). Changing the power of p2 in the denominator of the gluon
propagator in the Landau gauge from 1 to n produces, in the coordinate
space,
i
22n1 d/2
191
T cosh
a1 (n) = CF
2n+2d
(n + 1)
2
2
/2 cosh
d
d/2 n
+ n 1 coth +
cosh 2
2
sinh
(this coecient becomes real in the Euclidean space T iTE ). Therefore (8.12),
2u
i
(1 u)
F (, u) = CF
e D()u/1
T cosh
(2 + u )
2
2
+/2
d
1u
sinh 2 .
(2 + u 2) coth +
2u
sinh
/2 cosh
is (8.14)
The anomalous dimension corresponding to Z
=
2(1 + ) coth + 1
s
1
CF
.
3
4 B(2 + , 2 + ) (2 + ) (1 )
In order to obtain
J =
+
Q , we add (8.44):
coth 1
s
2
CF
3
4 B(2 + , 2 + ) (1 + ) (1 )
1 2
s
5
= 4CF
1 + + ( coth 1) .
4
3
3
J =
(8.57)
192
8 Renormalons in HQET
Expanding 1/D1 (t) up to the linear term in t and using the oneloop onshell integrals (4.15), we nd a1 (n) and F (, u) (8.12). The functions F (, u),
for all onshell quantities, have a common function structure, resulting
from (4.15) with n2 = 1 + u :
2u
(1 + u) (1 2u)
D()u/1 N (, u) .
F (, u) =
e
(8.58)
m
(3 u )
The rst function in the numerator, with a positive sign in front of u, comes
from the second function in the numerator of (4.15), with a negative sign
in front of d, and its poles are UV divergences. The second function in the
numerator, with a negative sign in front of u, comes from the rst function
in the numerator of (4.15), with a positive sign in front of d, and its poles
are IR divergences. For T (t), we obtain [7]
N (, u) = 2CF (3 2)(1 u) 1 (1 + u )t + O(t2 ) .
(8.59)
os
The onshell mass renormalization constant Zm
= m0 /m = 1 T (0) with
1/0 accuracy is given by (8.11) and (8.58) with N (, u) equal to minus (8.59)
at t = 0. Retaining only terms with negative powers of , we obtain the MS
os
contains no IR divergences).
mass renormalization constant Zm (because Zm
Using (8.14), we reproduce the mass anomalous dimension (8.40). Retainos
ing terms with 0 , we obtain Zm
/Zm () = m()/m in the form (8.15). As
usual, it is convenient to express m() via the renormalizationgroup invariant m
(8.16). Then the ratio [3]
1
1
m
,
du eu/ S(u) + O
=1+
m
0 0
02
(u) (1 2u)
1
S(u) = 6CF
(1 u)
.
(8.60)
(3 u)
2u
The rst function, with a positive sign in front of u, produces UV renormalons, while the second one, with a negative sign, produces IR renormalons
(Fig. 8.2b). We can understand this from power counting (Sect. 8.2). The
QCD quark selfenergy has a logarithmic UV divergence (UV = 0), and
hence the UV renormalons are situated at u < 0. The index of the IR divergence of the onshell quark selfenergy is IR = 1, and the IR renormalons
are at u 1/2.
The ratio (8.60) can be represented [11] in the form (8.28). For > 1, the
distribution function is given by the sum (8.33) over the UV renormalons at
u = n, n = 1, 2, 3, . . . :
n
(n + 1) (2n)!
1
w( ) = 6CF
n! (n + 2)!
n=1
1
2
4
6
2
= CF 1 2 1
1+
.
2
193
For < 1, the distribution function is the sum (8.32) over the IR renormalons
at u = n + 1/2 (n = 0, 1, 2, . . . ) and at u = 2:
(2n 1) (2n 1)!! (2n 5)!! n+1/2
1
2
w( ) = CF 3
2
(2n)!
(4)n
n=0
1
= CF (2 ) (4 + ) + 2
2
(both of these series are easily summed using the Newton binomial expansion). Finally, we obtain (Fig. 8.4)
1
CF (2 ) (4 + ) + 2 6( 1) .
(8.61)
2
MS
.
0
(8.62)
194
8 Renormalons in HQET
2CF N0
[1 + O (1/2 u)] ,
(8.63)
s (0 )
m
1
=1+
cL
m
0
4
L=1
a=
1
.
202
(8.64)
(8.65)
195
In order to calculate the diagrams in Fig. 8.5c, we need the triangle quark
loop with linear accuracy in q; it is a combination of oneloop propagator
integrals (2.18). Again, all terms in the sum over l from 0 to L 2 are equal,
and the summation just gives a factor L 1:
Nc (, u) = CA
10 4u 28 + 9u + 233 42 u 63 u
1 .
2(1 )
Also, we must include the oneloop onshell quark wavefunction renormalization contribution (8.65) multiplied by the Born scattering amplitude (which
is just 1). Finally, we arrive at [9]
N (, u) = CF NF (, u) + CA NA (, u) ,
NF (, u) = 4u(1 + u 2u) ,
2u
NA (, u) =
(2 + 3u 5 6u + 22 + 42 u) .
2(1 )
(8.66)
L1l
L1
a
L2l
196
8 Renormalons in HQET
m = CA
3
2 24(1 + ) (2 + ) (1 )
02
(8.67)
1 2
s
13
= CA
1 + + .
2
6
2
It reproduces the leading 0 term of the twoloop result (4.67),
s
s
1 + (130 25CA )
+ .
m = CA
2
24
The perturbative series (8.67) converges for 0 s  < 4.
The renormalizationgroup invariant Cm corresponding to Cm () (see
(8.16)) has the form (8.19), with [9]
1
CA
(u) (1 2u)
.
S(u) =
4u(1 + u)CF + (2 u)(2 + 3u)CA
(3 u)
2
u
(8.68)
The renormalon poles coincide with those in Fig. 8.2b. Taking the residue at
the leading IR pole u = 1/2 and comparing with (8.46), we obtain
7 CA
.
(8.69)
Cm = 1 +
8 CF
m
In physical quantities, such as the mass splitting mB mB , this IR renormalon ambiguity is compensated by UV renormalon ambiguities in the matrix
elements in the 1/m correction. Detailed investigation of this cancellation allows one to nd the exact nature of the singularity of S(u) at u = 1/2: it
is a branch point, a sum of three terms with dierent fractional powers of
1/2 u, where the powers are known exactly, but the normalizations are
known only in the large0 limit. The largeL asymptotics of the perturbative series for Cm can be found. The results have been obtained in [9]. We
shall not discuss them here, because they require the use of 1/m2 terms in
the HQET Lagrangian. A similar analysis of bilinear heavylight currents
will be presented in the next section.
We can rewrite Cm in the form (8.28), with [9]
w( ) = CF wF ( ) + CA wA ( ) ,
2 + 4 + 2
2 ,
wF ( ) = 2
(4 + )
14 + 5
wA ( ) =
5 ( 1)
4
(4 + )
(8.70)
(Fig. 8.6; these formulae can be derived in the same way as for (8.61)).
197
HQET
matrix
elements
j , i , . . .
The QCD matrix element j contains no renormalon ambiguities, because
the operator j has the lowest dimensionality in its channel. In schemes without strict separation of large and small momenta, such as MS, this procedure
articially introduces IR renormalon ambiguities into matching coecients
and UV renormalon ambiguities into HQET matrix elements. When calculating the matching coecients C, . . . , we integrate over all loop momenta,
including small momenta. Therefore, the matching coecients contain, in
addition to the main shortdistance contributions, also contributions from
large distances, where the perturbation theory is illdened. The latter contributions produce IR renormalon singularities, which lead to ambiguities
(MS /m)n in the matching coecients C, . . . Similarly, HQET matrix elements of higherdimensional operators i , . . . contain, in addition to the
main largedistance contributions, also contributions from short distances,
which produce several UV renormalon singularities at positive u. The latter
contributions lead to ambiguities of the order nMS times lowerdimensional
matrix elements (e.g.,
j ). These two kinds of renormalon
ambiguities have
to cancel in physical fullQCD matrix elements j (8.71) [12].
Let us consider the QCD/HQET matching coecients C () (5.59).
Closely following the derivation in Sect. 5.6 (with the gluon denominator
raised to the power n = 1 + (L 1)), we obtain for (mv, 0) (5.63)
a1 (n)
(4)d/2
198
8 Renormalons in HQET
iCF
=
2(d 1)
At the rst order in 1/0 , (mv, 0) has the form of (8.11) and (8.58) with [7]
N (, u) = CF 2 u + 2uh 2h2 ,
(8.72)
1
,
02
(8.73)
.
(8.74)
2u
Comparing the residue at the leading IR renormalon u = 1/2 with (8.46),
we obtain the ambiguity of the matching coecient [7],
1 15
+ (n 2) 2(n 2)2
.
(8.75)
C () =
3 4
m
The matching coecients for the currents containing 5AC and 5HV have
identical S(u) and C ; they dier only by K (s ) in (8.16). From (8.72)
and (5.69), we can trivially reproduce (8.41). Taking into account (5.75),
C1 ()
m
=
,
m()
C 0 ()
the result (8.72) also reproduces the corresponding formula for m/m(),
namely (8.59) with t = 0.
The ratio fB /fB (5.76) is given by (8.19), with [12]
S(u) = 4CF
(1 + u) (1 2u)
.
(3 u)
199
(8.76)
This ratio can be rewritten in the form (8.28). Summing the series (8.32)
and (8.33) over the residues of S(u), we obtain [11]
2
w( ) = CF (1 + ) (4 + ) (3 + )
(8.77)
3
(Fig. 8.7).
Now we
turn to the matrix elements of the dimension4 HQET operators
04,5 B (see (6.43)). The UV contributions to these matrix elements are
independent of the external states, and we may use quark states instead
of the hadron states used in (6.60), (6.61). By dimensional analysis, the UV
renormalon ambiguities of the matrix elements of 4,5 are proportional to
times the matrix element of the lowerdimensional operator
j with the same
external states. We consider a transition from an oshell heavy quark with
residual energy < 0 to a light quark with zero momentum, this is enough to
ensure the absence of IR divergences. For 4 , all loop corrections to the vertex
function (see Fig. 8.8) vanish. The kineticenergy vertices contain no Dirac
matrices, and we may take (1/4) Tr on the lightquark line; this yields k at
the vertex, and the gluon propagator with insertions is transverse. There is
one more contribution: the matrix element F of
j should be multiplied by the
1/2
heavyquark wavefunction renormalization ZQ , which has a UV renormalon
(8.78)
200
8 Renormalons in HQET
k
k + v
(8.79)
(8.81)
k + v
In the full QCD matrix elements (6.17), the IR renormalon ambiguities (8.75) of the leading matching coecients C are compensated, at the
order 1/0 , by the UV renormalon ambiguities of the subleading matrix elements (8.46) and Gk,m (8.78), (8.81). This cancellation must hold
beyond the large0 limit. The subleading matrix elements are controlled
by the renormalization group. The requirement of cancellation allows one to
investigate the structure of the leading IR renormalon singularity of C [8].
m =
G
201
m
2 0
;
m0
(8.82)
0
Gk = N1 0 , Gm = N2 2
2
m0
(see 8.50). The normalization factors N1,2 are known only in the large0
limit:
Ni = 1 + O(1/0 ) ;
in general, they are just some unknown numbers of order 1. Using (8.49), we
can represent the UV renormalon ambiguities of the 1/m corrections in (6.65)
and (6.85) as exp[2/(0 s (0 ))] times a sum of terms with dierent fractional powers of s (0 )/(4). It is convenient to replace log[s (0 )/(4)] by
[(s (0 )/(4)) 1]/, and take the limit 0 at the end of the calculation.
In order to cancel this ambiguity, we must have a branch point
ri
(8.83)
S (u) =
1+ai
i (1/2 u)
instead of a simple pole (8.74). Then (see (8.48))
ai
1
2
ri
0 s (0 )
exp
.
C =
0
0 s (0 ) i (1 + ai )
4
(8.84)
The requirement for cancellation of the ambiguities in (6.65) gives the following for = 1, 0 :
S (u) =
1+ /(2 2 )
0
(1/2 u) 1
k
1
1/2 u
m
0 log
1+ 2
1 + 0 N0
20
0
20
m0
3
m
N1 + 2 0
N2 (1/2 u)m0 /(20 ) .
2
m0
CF
CF
2
(8.85)
202
8 Renormalons in HQET
1/2 u
1
1
m
+
N0
1+ 2
1 0
0
20
3
m0
3 1
0m
m0 /(20 )
N1
.
(8.86)
N2 (1/2 u)
2
2
3
m0
0k
20
log
Here
1
/(202 )
1 + 2 0 1
,
20
1
m0
/(202 )m0 /(20 )
N2 = N2 1 + 2
0 1
20
20
N1 = N1
(see (8.52)). The nexttoleading terms were derived in [8]. In the large0
limit, the simplepole behaviour with the residue (8.75) is reproduced.
The asymptotics of the perturbative coecients cL at L 1 is determined
by the renormalon singularity closest to the origin. Similarly to (8.64), we
obtain, for = 1, 0 ,
0k
m
n
1 /(202 )
cn+1 = 2CF n! (20 ) (20 n)
log(20 n) 1 + 0
N0
20
m0
3
0m
m0 /(20 )
N1 + 2
,
(8.87)
N2 (20 n)
2
m0
and for = i , i 0 ,
2
0k
1
m
log(20 n) +
N0
1 0
20
3
m0
3
1
0m
m0 /(20 )
N1
.
N2 (20 n)
2
2
3
m0
(8.88)
CF
4
3 (1/2 u)1+1 /(202 )
m
m
/(20 )
1 0
N0 2 0
N2 (1/2 u) m0
, (8.89)
m0
m0
(8.90)
203
(d 2n 1) (d/2 + n + 1)
(d n 1)
(8.91)
R1
1 + 2u
1 2u
2
(1 u u + u)R0 ,
+ (3 2)
(8.93)
1 + 2u
where
R0 = cosh
L
,
2
R1 =
for the onshell QCD matrix element (which is equal to the matching coecient, because all loop corrections in HQET vanish). The corresponding
anomalous dimension (8.14) reproduces jn (8.39), because
J = 0 at = 0.
The function S(u) (8.20) for the matching coecient is
4 u2
(u) (1 2u)
S(u) = CF 2
(n 2)2 u(n 2)
R1
(3 u)
1 + 2u
(1 2u)(1 u u2 )
(n 1)(n 3)
R0
+3
(8.94)
1 + 2u
u
with
0 = e5/6 mb mc
204
8 Renormalons in HQET
(8.95)
+ (cosh L + z sinh L)u(2 u )
(3 2)(1 u)(1 + u )(ru + ru ) ,
u +1
dz
N2 (, u) = CF
2 1 (a+ a )1+u
L
e (1 + z)2 h
(1 2u) + (1 z)(2 u ) ,
N3 (, u) = CF
u
2
+1
dz
(a+ a )1+u
eL (1 z)2 h
(1 2u) + (1 + z)(2 u ) ,
2
hu(1 2u) +1 dz (1 z 2 )
N4 (, u) = CF
.
1+u
4
1 (a+ a )
205
(8.97)
(u) (1 2u)
Ni (0, 0)
Ni (0, u)
,
(3 u)
2u
have a leading IR renormalon pole at u = 1/2, thus producing the ambiguities (8.26)
Hi =
Ni (0, 1/2)
3CF
mb mc
in the matching coecients. It is easy to calculate Ni (0, 1/2) using the integrals
+1
+1
dz
z dz
4 cosh(L/2)
4 sinh(L/2)
,
.
=
=
3/2
3/2
cosh
+
1
cosh + 1
(a
a
)
(a
a
)
+
+
1
1
We obtain [12]
1
3
1
1
H1 =
+
,
cosh + 1 4
mc
mb
1
1
H2 =
, H3 =
,
cosh + 1 2mc
cosh + 1 2mb
H4 = 0 . (8.98)
206
8 Renormalons in HQET
(v v D + v v D) cosh v v D v v D
sinh2
v cosh v
,
cosh + 1
v cosh v
i D
.
cosh + 1
The longitudinal
of the local 1/mb,c contribution (7.71) to the QCD
part
matrix element J is easily derived by this substitution. It clearly has a UV
Matrix elements of operators with
renormalon ambiguity proportional to .
transverse derivatives cannot be written as matrix elements of the leading
currents Ji times some scalar factors, they require new, independent form
factors. Therefore, they contain no UV renormalon ambiguities, which whould
need to have the form of times lowerdimensional matrix elements of Ji .
The above derivation is exact (and not only valid in the large0 limit). At
the rst order in 1/0 , we may replace H1 by 1, H2,3,4 by 0, Hi by 0. The
contribution of the local subleading operators to the ambiguity of J in this
approximation is [12]
1
1
1
J1
+
1
cosh + 1
mc
mb
2
1
J2 +
J3
.
(8.99)
cosh + 1 2mc
2mb
Now we turn to bilocal subleading operators, and consider the operator
i dx T J1 (0), kc (x)
with the insertion of the cquark kinetic energy. This operator appears in the
expansion (7.20), with a coecient H1 . The oneloop vertex (Fig. 8.10) with
the gluon denominator raised to the power n is
2
v
CF
k
dd k
a1 (n)
=i
k
2mc
(2)d
k v +
(4)d/2
v
k k
g +
,
(k v + )(k v + )(k 2 )n
k 2
where k = k (k v ) v . We are interested in the UV renormalon at u =
1/2; therefore, to make the subsequent formulae shorter, we shall calculate
207
F (u) (8.17) instead of the full function F (, u), and omit terms regular at
u = 1/2. We also set = for simplicity, and obtain
1
d4 k
CF
2u
u(2)
F (u) = i
2
2mc
(k v )(k 2 )1+u
4
cosh
d k
+
+ ,
2 (k v )(k v )2 (k 2 )u
where the dots mean integrals without linear UV divergences at u = 0 (and
hence they have no UV renormalon singularity at u = 1/2), and 2 plays
the role of m in the denition (8.17).
The rst integral is trivial (see (2.27)):
(1 + 2u) (1 u)
i
d4 k
.
=2
2 (2)1+2u
2
1+u
(k v )(k )
(1 + u)
For the second integral, we use the HQET Feynman parametrization (2.23):
i
d4 k
1+2u
I = 2 (2)
(k v )(k v )2 (k 2 )u
(3 + u)
i
= 8 2 (2)1+2u
(u)
y dy dy d4 k
d
dz
I = 2u
2u ,
1+u
2
( + 1)
0
1
cosh (/2) z 2 sinh2 (/2)
208
8 Renormalons in HQET
where
0
1u d
(1 + 2u) (2 u)
.
=
1+u
( + 1)
(1 + u)
1 + u(1 u) cosh
+1
dz
2u
cosh2 (/2) z 2 sinh2 (/2)
+ ,
(8.100)
+1
dz
1
3/2 = cosh + 1 .
cosh (/2) z 2 sinh (/2)
2
.
2 cosh + 1 2mc
The contribution of the bilocal operator containing the bquark kinetic energy
contains mb instead of mc . Therefore, the contribution
of all bilocal operators
with kineticenergy insertions to the ambiguity of J is [12]
1
1
2 cosh + 1
1
1
+
mc
mb
J1 .
2
(8.101)
References
209
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
Index
212
Index
Feynman
gauge 5354
parametrization 2223, 64, 161, 167
HQET 25, 167, 207
rules 2021, 60, 63
FoldyWouthuysen transformation
63, 80, 122
form factor 911
chromomagnetic 7476
Dirac 7476
IsgurWise 1011
gauge
a = 3 177
Feynman (a = 1) 5354
Landau (a = 0) 46, 176177, 180,
183187, 190191
Yennie (a = 3) 52, 58
heavy quark
kinetic energy 7, 60
propagator 2021, 5055, 186188
selfenergy 5055, 6061, 7071, 87,
186187
spin symmetry 34, 21, 5960,
115116
homogeneity relation 29
HQET 32
hyperne splitting 35, 78, 8384
inclusive decays 1213
integration by parts 2829, 6567
HQET 3132
inversion 6465
IsgurWise form factor 1011
isospin 5
Lagrangian
HQET 20, 50, 60, 6263
QCD 19, 3536, 67, 7172
Landau
gauge 46, 176177, 180, 183187,
190191
pole 183
Larins relation 2930
lattice simulations 22
leading logarithms 8283, 137, 139,
142, 144
mass shell
1920, 62
matching
Smatrix elements 59, 80, 194195
dimension4 operators 159160,
169170
quark currents
dynamicdynamic static
dynamic 107109, 122124,
136137, 142, 157, 166, 197198
dynamicdynamic static
static 152157, 159162, 164166,
168170, 203205
staticdynamic staticstatic
157159, 166168
Mellim transform 182183
partial fractions 33, 155, 158159
propagator
electron 5556
gluon 3945, 85, 176177
heavy quark 2021, 5055, 186188
quark 21, 4550, 6769, 183185
renormalization
constants 35, 3739, 50, 126136,
141
coupling constant 3637, 74, 7778,
189
gluon eld 4145, 74
group 3739, 8183, 126127,
131132, 135136, 138139, 141142,
165, 179
heavyheavy current 146151
heavylight current 102103
heavyquark eld 7071, 190
mass 46, 6769, 9395, 116, 128,
192194, 198
onshell 6769, 191194
quark current 9193
quark eld 4647, 50, 6869, 194
scalar current 9394
scale 36
vector current 93
renormalon ambiguity 181183, 189
cancellation 193194, 196197,
200202, 205, 208
infrared 193, 196, 198, 204205
ultraviolet 188190, 199200, 206,
208
Index
reparametrization invariance 5962,
7879, 122123, 141, 171173
residual mass 2122, 62
selfenergy
gluon 3945, 70, 8586
heavy quark 5055, 6061, 7071,
87, 186187
quark 4550, 6769, 8687, 183184,
191192
Shmushkevich factory 8
sum rule
BGSUV 1516
Bjorken 1314
Uraltsev 1415
Voloshin 1516
superavour symmetry 4, 5960
triangle relations
HQET 3132
Uraltsev sum rule
2829
1415
213
52, 58