Vol201 (2004) Heavy Quark Effective Theory (Andrey Grozin) PDF

Springer Tracts in Modern Physics
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

Springer Tracts in Modern Physics provides comprehensive and critical reviews of topics of current
interest in physics. The following elds are emphasized: elementary particle physics, solid-state
physics, complex systems, and fundamental astrophysics.
Suitable reviews of other elds can also be accepted. The editors encourage prospective authors to
correspond with them in advance of submitting an article. For reviews of topics belonging to the
above mentioned elds, they should address the responsible editor, otherwise the managing editor.
See also springeronline.com
Managing Editor
Solid-State Physics, Editors
Gerhard Hohler
Andrei Ruckenstein
Editor for The Americas
Institut fur Theoretische Teilchenphysik

Universitat Karlsruhe
Postfach 69 80
76128 Karlsruhe, Germany
Phone: +49 (7 21) 6 08 33 75
Fax: +49 (7 21) 37 07 26
Email: gerhard.hoehler@physik.uni-karlsruhe.de
www-ttp.physik.uni-karlsruhe.de/
Elementary Particle Physics, Editors
Department of Physics and Astronomy

Rutgers, The State University of New Jersey
136 Frelinghuysen Road
Piscataway, NJ 08854-8019, USA
Phone: +1 (732) 445 43 29
Fax: +1 (732) 445-43 43
Email: andreir@physics.rutgers.edu
www.physics.rutgers.edu/people/pips/
Ruckenstein.html
Johann H. Kuhn
Peter Wole
Institut fur Theoretische Teilchenphysik

Postfach 69 80
Phone: +49 (7 21) 6 08 33 72
Fax: +49 (7 21) 37 07 26
Email: johann.kuehn@physik.uni-karlsruhe.de
www-ttp.physik.uni-karlsruhe.de/jk
Institut fur Theorie der Kondensierten Materie

Postfach 69 80
Phone: +49 (7 21) 6 08 35 90
Fax: +49 (7 21) 69 81 50
Email: woele@tkm.physik.uni-karlsruhe.de
www-tkm.physik.uni-karlsruhe.de
Thomas Muller
Institut fur Experimentelle Kernphysik
Fakultat fur Physik
Postfach 69 80
Phone: +49 (7 21) 6 08 35 24
Fax: +49 (7 21) 6 07 26 21
Email: thomas.muller@physik.uni-karlsruhe.de
www-ekp.physik.uni-karlsruhe.de
Fundamental Astrophysics, Editor

Joachim Trumper
Max-Planck-Institut fur Extraterrestrische Physik
Postfach 16 03
85740 Garching, Germany
Phone: +49 (89) 32 99 35 59
Fax: +49 (89) 32 99 35 69
Email: jtrumper@mpe-garching.mpg.de
www.mpe-garching.mpg.de/index.html
Complex Systems, Editor

Frank Steiner
Abteilung Theoretische Physik
Universitat Ulm
Albert-Einstein-Allee 11
89069 Ulm, Germany
Phone: +49 (7 31) 5 02 29 10
Fax: +49 (7 31) 5 02 29 24
Email: frank.steiner@physik.uni-ulm.de
www.physik.uni-ulm.de/theo/qc/group.html
Andrey Grozin
Heavy Quark
Effective Theory
With 72 Figures
13
Andrey Grozin
Budker Institute of Nuclear Physics
630090 Novosibirsk, Russia
E-mail: A.G.Grozin@inp.nsk.su
Cataloging-in-Publication Data applied for
Physics and Astronomy Classication Scheme (PACS):

12.39.Hg, 12.38.Bx
ISSN print edition: 0081-3869

ISSN electronic edition: 1615-0430
ISBN 3-540-20692-2 Springer-Verlag Berlin Heidelberg New York
This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned,
specically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on
microlm or in any other way, and storage in data banks. Duplication of this publication or parts thereof is
permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and
permission for use must always be obtained from Springer-Verlag. Violations are liable for prosecution under the
German Copyright Law.
Springer-Verlag is a part of Springer Science+Business Media
springeronline.com
Springer-Verlag Berlin Heidelberg 2004
Printed in Germany
The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in
the absence of a specic statement, that such names are exempt from the relevant protective laws and regulations
and therefore free for general use.
Typesetting: by the author using a Springer LATEX macro package
Cover concept: eStudio Calamar Steinen
Cover production: design &production GmbH, Heidelberg
Printed on acid-free paper
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 B-mesons, with high statistics and low systematic errors. Therefore, a better theoretical understanding of the properties of B-mesons is very important.
An understanding of strong-interaction 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 B-meson the simplest non-trivial 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
Chapter 1 gives an overview of basic experimental facts about mesons

and baryons containing a b or c quark and a qualitative discussion of some of
their properties. Then the HQET Lagrangian (Chaps. 24) and bilinear quark
currents (Chaps. 57) are discussed in detail. Finally, Chap. 8 discuses some
facts and hypotheses about the behaviour of perturbative series in HQET at
high orders.
We shall discuss a few of the most fundamental applications in great
detail, rather than provide a long list of results without derivations. Some
of the material in this book was used in lecture courses, and all calculations
were actually performed in the classroom. Such an explicit approach should
be appropriate for those readers who wish to develop skills for solving similar
problems. Results which cannot be derived during a lecture are in most cases
not included. Readers who want to do non-trivial calculations are advised to
use computer algebra (see, e.g., [8]).
Larger areas which are omitted from this book, owing to space and time
constraints, are inclusive decays of heavy hadrons (see, e.g., [16, 1, 17]) and
their interactions with soft pions (see, e.g., [20, 2]). Remarkable progress has
been made in recent years in both of these areas. Also, we shall not discuss
other eective eld theories, such as non-relativistic QCD (used for heavyquarkantiquark systems) and soft-collinear eective theory (used for decays of hadrons containing a heavy quark producing energetic light hadrons).
These theories are under active development, and a rm knowledge of a
simpler eective theory, HQET, is invaluable for understanding these more
complicated theories.
For more information about applications of HQET, see the book [13] and
the reviews [1, 3, 4, 5, 6, 7, 10, 11, 12, 14, 16, 17, 18, 19].
I am grateful to P.A. Baikov, A.E. Blinov, D.J. Broadhurst, K.G. Chetyrkin, A. Czarnecki, A.I. Davydychev, G.P. Korchemsky, J.G. K
orner, S.A. Larin, T. Mannel, K. Melnikov, S.V. Mikhailov, M. Neubert, T. van Ritbergen, V.A. Smirnov, F.V. Tkachov, N.G. Uraltsev, A.I. Vainshtein, and
O.I. Yakovlev for numerous illuminating discussions of problems discussed in
this book.
Novosibirsk, February 2004
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, SLAC-R508 (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 92-97 (Novosibirsk 1992), hep-ph/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; hep-ph/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. Randjbar-Daemi, 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
Non-Perturbative Particle Theory and Experimental Tests, ed. by M. Jamin,
O. Nachtmann, G. Domokos, S. Kovesi-Domokos (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 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Hadrons with a Heavy Quark . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.1 Mesons and Baryons with a Heavy Quark . . . . . . . . . . . . . . . . . 4
1.2 Semileptonic Decays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
The HQET Lagrangian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.1 The HQET Lagrangian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2 One-Loop Massless Propagator Diagrams . . . . . . . . . . . . . . . . . .
2.3 One-Loop HQET Propagator Diagrams . . . . . . . . . . . . . . . . . . .
2.4 Two-Loop Massless Propagator Diagrams . . . . . . . . . . . . . . . . . .
2.5 Two-Loop HQET Propagator Diagrams . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
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 Heavy-Electron Eective Theory . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
35
35
39
45
50
55
58
The HQET Lagrangian: 1/m Corrections . . . . . . . . . . . . . . . . .

4.1 1/m Corrections to the HQET Lagrangian . . . . . . . . . . . . . . . . .
4.2 On-Shell Renormalization of QCD . . . . . . . . . . . . . . . . . . . . . . . .
4.3 On-Shell Renormalization of HQET . . . . . . . . . . . . . . . . . . . . . . .
4.4 Scattering in an External Gluonic Field in QCD . . . . . . . . . . . .
4.5 Scattering in an External Gluonic Field in HQET . . . . . . . . . .
4.6 Chromomagnetic Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.7 Decoupling of Heavy-Quark Loops . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
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
HeavyLight Currents: 1/m Corrections . . . . . . . . . . . . . . . . . .

6.1 1/m Corrections to HeavyLight Currents . . . . . . . . . . . . . . . . .
6.2 Local Dimension-4 HeavyLight Operators . . . . . . . . . . . . . . . . .
6.3 Bilocal Dimension-4 HeavyLight Operators . . . . . . . . . . . . . . .
6.4 Spin-0 Currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.5 Spin-1 Currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
121
121
126
131
135
140
144
HeavyHeavy Currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.1 HeavyHeavy Current in HQET . . . . . . . . . . . . . . . . . . . . . . . . . .
7.2 Flavour-Diagonal Currents at Zero Recoil . . . . . . . . . . . . . . . . . .
7.3 Flavour-Changing Currents at Zero Recoil . . . . . . . . . . . . . . . . .
7.4 Flavour-Diagonal Currents at Non-Zero Recoil . . . . . . . . . . . . .
7.5 Flavour-Changing Currents at Non-Zero Recoil . . . . . . . . . . . . .
7.6 1/m Corrections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
145
145
152
155
161
162
171
173
Renormalons in HQET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.1 Large-0 Limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.2 Renormalons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.3 Light Quarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.4 HQET Heavy Quark . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.5 On-Shell 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
Our notation mainly follows [2]. Covariant and contravariant components of

4-vectors are related by a = g a , where the indices are in the range 0, 1,
2, 3, and the metric tensor is
1 0 0 0
0 1 0 0
g =
0 0 1 0 .
0 0 0 1
The velocity 4-vector 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 .
We shall use dimensional regularization for calculating loop diagrams: the

spacetime dimensionality is
d = 4 2
(dimension 0 is time-like and d 1 dimensions are space-like). For more
details, see the textbook [1].
The number of quark colours is Nc = 3. We shall write all formulae for
the SU (Nc ) gauge group with an arbitrary Nc . The covariant derivative of
the quark field is
D q = ( iA ) q ,
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 Springer-Verlag 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 ) ,
(tc )ab = if acb .
The quadratic Casimir operators in the fundamental (quark) representation

and the adjoint (gluon) one are ta ta = CF , (ta )bc (ta )cd = CA bd ,
CA = Nc ,
CF =
Nc2 1
2Nc
(see Sect. 3.3).
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
1 Hadrons with a Heavy Quark
The B meson is the hydrogen atom of quantum chromodynamics (QCD),

the simplest non-trivial hadron. In the leading approximation, the b quark
in it just sits at rest at the origin and creates a chromoelectric eld. Light
constituents (gluons, light quarks, and antiquarks) move in this external eld.
Their motion is relativistic; the number of gluons and light quarkantiquark
pairs in this light cloud is undetermined and varying. Therefore, there are
no reasons to expect that a non-relativistic potential quark model would
describe the B meson well enough (in contrast to the meson, where the
non-relativistic two-particle picture gives a good starting point).
Similarly, the b baryon can be called the helium atom of QCD. Unlike in
atomic physics, where the hydrogen atom is much simpler than helium, the B
and b are equally dicult. Both have a light cloud with a variable number of
relativistic particles. The size of this cloud is the connement radius 1/QCD ;
its properties are determined by large-distance nonperturbative QCD.
The analogy with atomic physics can tell us a lot about hadrons with a
heavy quark. The usual hydrogen and tritium have identical chemical properties, despite the fact that the tritium nucleus is three times heavier than
the proton. Both nuclei create identical electric elds, and both stay at rest.
Similarly, the D and B mesons have identical hadro-chemical properties,
despite the fact that the b quark is three times heavier than the c.
The proton magnetic moment is of the order of the nuclear magneton
e/(2mp ), and is much smaller than the electron magnetic moment e/(2me).
Therefore, the energy dierence between the states of the hydrogen atom
with total spins 0 and 1 (hyperne splitting) is small (of the order me /mp
times the ne structure). Similarly, the b-quark chromomagnetic moment
is proportional to 1/mb by dimensionality, and the hyperne splitting between the B and B mesons is small (proportional to 1/mb ). Unlike in atomic
physics, both gross-structure intervals and ne-structure intervals are just
some numbers times QCD , because the light components are relativistic (the
practical success of constituent quark models shows that these dimensionless
numbers for ne splittings can be rather small, but they contain no small
parameter).
In the limit m , the heavy-quark spin does not interact with the gluon
eld. Therefore, it may be rotated at will, without changing the physics. Such

rotations can transform the B and B into each other; they are degenerate
and have identical properties in this limit. This heavy-quark spin symmetry
yields many useful relations among heavy-hadron form factors [13]. Not only
the orientation, but also the magnitude of the heavy-quark spin is irrelevant
in the innite-mass limit. We can switch o the heavy-quark 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 small-size
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.
1.1 Mesons and Baryons with a Heavy Quark

where Q is a heavy quark
Let us consider mesons with the quark contents Qq,
with mass m (c or b), and q is a light quark (u, d, or s). As discussed above,
the heavy-quark spin is inessential in the limit m , and may be switched
o. In a world with a scalar heavy antiquark, S-wave mesons have angular
momentum and parity j P = (1/2)+ ; P -wave mesons have j P = (1/2) and
(3/2) . The energy dierence between these two P -wave states (ne splitting)
is a constant times QCD at m , just like the splittings between these
P -wave states and the ground state; however, this constant is likely to be
small.
has spin and parity sP = (1/2) .
In our real world, the heavy antiquark Q
Q
The quantum numbers in the above paragraph are those of the cloud of light
elds of a meson. Adding the heavy-antiquark spin, we obtain, in the limit
m , a degenerate doublet of S-wave mesons with spin and parity sP = 0
and 1 , and two degenerate doublets of P -wave mesons, one with sP = 0+
and 1+ , and the other with sP = 1+ and 2+ . At a large but nite heavy-quark
mass m, these doublets are not exactly degenerate. Hyperne splittings, equal
to to some dimensionless numbers times 2QCD /m, appear. It is natural to
expect that hyperne splittings in P -wave mesons are less than in the groundstate S-wave doublet, because the characteristic distance between the quarks
is larger in the P -wave case. Note that the 1+ mesons from the dierent
doublets do not dier from each other by any exactly conserved quantum
numbers, and hence can mix. They dier by the angular momenta of the
light elds, which is conserved up to 1/m corrections; therefore, the mixing

angle should be of the order of QCD /m.
Mesons with q = u and d form isodoublets; together with isosinglets with
q = s, they form SU (3) triplets.
The experimentally observed [18] mesons containing the c antiquark are
shown in Fig. 1.1. The energy scale at the left is in MeV, relative to the
2 form a doublet, with the quan 1 and D
lowest-mass meson. The mesons D
tum numbers of the light elds j P = (3/2)+ . The second P -wave doublet
is suspiciously absent. It should be close to the (3/2)+ one; it is not more
dicult to produce these mesons than the (3/2)+ ones. The problem is that
they are too wide, and cannot be cleanly separated from the continuum.
600
2 2+
D
1 1+
D
400
200
0
D
= cd D
0 = cu D
cs
D
s =
Fig. 1.1. Mesons containing c
In the leading approximation, the spectrum of b-containing

mesons is
obtained from the spectrum of c-containing mesons simply by a shift by
antiquark
mb mc . The experimentally observed mesons containing the b
are shown in Fig. 1.2. The spectrum of c-containing mesons is shown by
dashed lines for comparison. It is positioned in such a way that the weighted
average energies of the ground-state doublets coincide, where the 1 meson
has weight 3 and the 0 meson has weight 1. The states B1 and B2 are not
resolved, and are shown by a single line. Experimentally, it is dicult to
measure their masses exactly enough. The hyperne splitting of the groundstate doublet is smaller for B mesons than for D mesons, as expected.
In S-wave Qqq baryons, the light-quark spins can add to give j P = 0+
or 1+ . In the rst case their spin wave function is antisymmetric; the Fermi
statistics and the antisymmetry in colour require an antisymmetric avour
wave function. Hence the light quarks must be dierent; if they are u, d,
then their isospin is I = 0. With the heavy-quark spin switched o, this gives
the 0+ baryon Q with I = 0. If one of the light quarks is s, we have the
isodoublet Q , which forms an SU (3) antitriplet together with Q . With the
600
400
B1,2 1+ , 2+
200
B
0
B 0
B0 = bd
B0s = bs
B+ = bu
Fig. 1.2. Mesons containing b
heavy-quark spin switched on, these baryons have sP = (1/2)+ . In the 1+

case, the avour wave function is symmetric. If the light quarks are u, d, then
their isospin is I = 1. This gives the 1+ isotriplet Q ; with one s quark, we
obtain the isodoublet Q ; and with two s quarks, the isosinglet Q . Together,
they form an SU (3) sextet. With the heavy-quark spin switched on, we obtain
the degenerate doublets with sP = (1/2)+ , (3/2)+ : Q , Q ; Q , Q ; Q ,
Q . The hyperne splittings in these doublets are of the order of 2QCD /m.
Mixing between Q and Q is suppressed both by 1/m and by SU (3). There
is a large number of P -wave excited states; we shall not discuss them here.
The experimentally observed [18] baryons containing a c quark are shown
in Fig. 1.3. The higher states in the rst and the third column are P -wave.
In the third column, the lowest state c is followed by the doublet c , c .
3
2
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
Fig. 1.3. Baryons containing a c quark
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
Experimentally, this ratio is 0.89. This is a spectacular conrmation of the

idea that violations of the heavy-quark spin symmetry are proportional to
1/m. As we shall discuss in Chap. 4, the matrix element 2G depends on
the normalization scale, and hence is not quite the same for D and B; this
produces moderate perturbative corrections to (1.2).
P -wave excited states decay into the ground state emitting a pion. In the
ideal world with an innitely heavy scalar c, the (1/2) P -wave meson decays
into the (1/2)+ ground-state meson plus a pion (having sP = 0 ) with an
orbital angular momentum l = 0 (S-wave); the (3/2) P -wave meson decays
into the ground-state meson plus a pion with l = 2 (D-wave). The pion
momenta p in these decays are rather small, and hence the decay of the
(3/2) meson (whose width is proportional to p5 ) is strongly suppressed.
The decay of the (1/2) meson is not suppressed, its width is proportional
to p . Therefore, in our real world, the 0+ , 1+ doublet mesons are very wide,
and dicult to observe.
Let us consider decays of D1 and D2 , taking the c spin into account, but
still in the limit mc . The widths of D1 and D2 are equal, because the c
spin plays no role in the decay. D1 decays only into D ; the decay into D
is forbidden by angular-momentum conservation. D2 can decay into both D
and D . In order to nd the branching ratio B(D2 D), we shall use a
simple device known as the Shmushkevich factory.
Let us take a sample of D1,2 mesons with random polarizations of c. Then
3/8 of this sample are D1 , and 5/8 are D2 . Now let us wait for a small time dt;
a fraction dt of the sample will decay. The ground-state mesons produced
have randomly polarized c. Therefore, 1/4 of them are D, and 3/4 are D . D
mesons are produced only from D2 :
5
1
B(D2 D) = ,
8
4
and hence [15]
B(D2 D) =
2
,
5
B(D2 D ) =
3
.
5
In the limit mc , D and D are degenerate, and so are D1 and D2 .

In the real world, the pion momenta in these decays dier. The widths are
proportional to p5 , and even rather small momentum dierences produce a
drastic eect. It seems natural to suppose that the predictions of heavy-quark
spin symmetry hold for the coecients in front of p5 . Then

5
(D2 D)
2 p (D2 D)
=
= 2.5 ,
(D2 D )
3 p (D2 D )
while the experimental value is 2.3 0.6 [18]. Formally, the dierence of
p (D2 D)/p (D2 D ) from 1 is a 1/mc correction. We can only hope
that this kinematic 1/mc eect, included in the above estimate, is dominant.
1.2 Semileptonic Decays
Let us now discuss the B-meson leptonic decay constant fB . It is dened

by

0|b 5 u|B(p) = ifB p ,

where the one-particle state is normalized in the usual Lorentz-invariant way:
B(p )|B(p) = 2p0 (2)3 (p p) .
This relativistic normalization becomes nonsensical in the limit mb ,
and in that case the non-relativistic normalization
nr
B(p )|B(p)nr = (2)3 (p p)
should be used instead. Then, for the B meson at rest,

imB fB
0|b 0 5 u|B nr =
.
2mB
Denoting
this matrix element (which is mass-independent 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 ,
from the + and + decays. The branching B(B+ + ) should be of

order 0.5 104 , so that a direct measurement of fB+ at B-factories seems
feasible. Theoretical estimates of fB vary by about a factor 2.

, where W is a virtual W+ which decays
Let us discuss the decay B DW
+
into l l . In the limit mb , mc , it is enough to consider the case
10
c, and W are scalar. We concentrate our attention on decays of B

when b,
with 4-velocity v . Let J be the scalar current which
with 4-velocity v into D
with 4-velocity v by a scalar c with 4-velocity v . With the
replaces a scalar b
non-relativistic normalization of the scalar quark wave functions, these wave
functions are just 1, and the quark decay matrix element is

b = 1.
(1.5)
c J
will be used to denote
The ground-state B meson has sP = (1/2)+ ; D
generically a ground-state 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 S-wave
The transitions of the ground-state (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]:

(1/2)+ J (1/2)+ = (cosh )

u u ,

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 4-velocities
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 non-relativistic 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,
bc weak currents are proportional to the IsgurWise form factor (cosh ),
11
with trivial kinematic coecients. When the current J replaces an innitely

by an innitely heavy c with the same 4-velocity and colour, light
heavy b
elds do not notice it:
(1) = 1 .
(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
Fig. 1.4. The IsgurWise form factor
The B B form factor is also proportional to (cosh ). In this case,

q 2 = 2m2B (1 cosh ). The form factor has a cut in the annihilation channel
from q 2 = 4m2B to +. Therefore, (cosh ) has a cut from cosh = 1 to
(Fig. 1.5). Geometrically speaking, cosh > 1 corresponds to Minkowski
angles between the world lines of the incoming heavy quark and the outgoing
one this is the scattering (or decay) channel. When cosh = 1, the world
line is straight, and there is no transition at all (see (1.8)). When |cosh | < 1,
the angle is Euclidean. When cosh < 1, we have a Minkowski angle again,
but one of the 4-velocities is directed into the past this is the annihilation
channel. At the point cosh = 1, the heavy quark returns back along the
same world line. In fact, the very concept of the IsgurWise form factor is
inapplicable near this point. The HQET picture is based on the fact that
heavy quarks move along straight world lines. If their relative velocity in
the annihilation channel is s , they rotate around each other instead.
The B meson form factor has poles below the threshold corresponding to
mesons with binding energies mb 2s ; its behaviour in this region is not
universal. The concept of the IsgurWise form factor is only applicable at
|cosh + 1| 2s (Fig. 1.5).
12
Fig. 1.5. The complex cosh plane
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)
They are the fractions of the number of B Xc W+ decays with Xc velocity

v , where the hadronic system Xc happens to be a single meson. The decay
(1/2)+ (1/2)+ is S-wave, and therefore the squared matrix element tends
to a constant at 0. The decays (1/2)+ (1/2) and (1/2)+ (3/2)
are P -wave, and therefore the squared matrix elements behave as the relative
velocity squared, 2 . Similarly, decays into D-wave mesons (3/2)+ , (5/2)+ are
D-wave, and behave as 4 at 0. If we choose the mass of the virtual W
larger than mB + mD , we can consider the channel W BD. The squared
matrix elements are given by the formulae (1.10) with an extra factor 2,
because we sum over the B polarizations now, not average. The decays of the
scalar W into (1/2)+ (1/2) , (1/2)+ (1/2)+ , and (1/2)+ (3/2)+ are P -wave,
S-wave, and D-wave. Therefore, their squared matrix elements behave as the
relative velocity to the power 2, 0, and 4, correspondingly. This explains the
behaviour of the formulae (1.10) at cosh 1.
The inclusive decay rate B XcW can be written as F (, cosh )d,
where pX = mX v , v is the Xc 4-velocity, cosh = v v , and = mX mD
is the excitation energy (we are still in the limit mc , where mD = mD ).
The structure function is
F (, cosh ) =
cosh + 1 2
i (cosh )( i )
2
+
(1/2)
cosh 1 2
1/2 (cosh )( i )
2
(1/2)
13
(cosh + 1)2 (cosh 1) 2

3/2 (cosh )( i )
3
(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 D-wave 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)
Its dependence is evident from dimensionality, and the dependence is given

by the well-known QED soft-photon radiation function. It is also known from
classical electrodynamics: this function is the distribution in the radiation
energy when a charge suddenly changes its velocity from v to v .
F (, cosh )
Fig. 1.6. A qualitative sketch of the B Xc decay structure function
The total decay probability is unity:

F (, cosh )d = 1 .
(1.13)
14
This is the Bjorken sum rule [4, 5, 14]. In particular, the decay rate into the
ground-state (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 S-wave 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
D-wave nal state do not contribute in this order, nor do higher S-wave 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)
(this also follows from (1.14)).

We can also consider the inclusive decay of a polarized B. Its structure
function is (we use (1.7) for the spin-(3/2) meson)
u
/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 higher-wave 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 4-velocity v is
v rest frame is E = cosh px sinh . When a b

suddenly transformed into a c with a 4-velocity 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 4-velocity v . Therefore, the average excitation
energy of the Xc and the average squared excitation energy are

E = (cosh
1),
2

2 = 2 (cosh 1)2 + (cosh2 1)

(E )
3
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 ground-state 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)
Let be the minimum P -wave excitation energy. Then, replacing i by in

the left-hand side of (1.24),
we decrease
it. After singling out this factor ,

the remaining sum is 2 2 1/4 (1.16):

1
2 2
.
(1.26)
4
This inequality can also be rewritten as an upper bound on 2 [23]:
2
1
+
.
4 2
(1.27)
Similarly, replacing 2i by i in the left-hand side of (1.25), we decrease

it. After singling out the factor , the remaining sum is (1.24), and [1]

3
1
2 32 2
(1.28)
2
4
(in the second step, the inequality (1.26) has been used). If the lowest resonances in the (1/2) and (3/2) channels with nearly equal energies dominate
in (1.24), (1.25), then the inequalities (1.26), (1.28) should be close to equalities. This is probably the case. Strictly speaking, the minimum excitation
energy is equal to m , because the ground-state meson plus a soft pion
can have the needed quantum numbers. This is small, and the bounds are
very weak. However, the coupling of a soft pion with a heavy meson is small,
and these states contribute little to the sum rules (1.24), (1.25). The rst
important contribution comes from the lowest P -wave resonances.
As you may have noticed, the integral in the Bjorken sum rule (1.13) diverges logarithmically at large , owing to (1.12). The integrals (1.22), (1.23)
diverge even more strongly. Therefore, if we want to take s eects into
account, we have to cut these integrals somehow. On the other hand, the
form factors i (cosh ), 1/2 (cosh ), 3/2 (cosh ) depend on the normalization point , if perturbative eects are taken into consideration. It is natural
to expect that the sum rules are valid, up to O(s ()) corrections, when
is of the order of the cut-o energy (this is discussed in [11] in more detail).
Therefore, the anomalous dimension of the HQET heavyheavy quark current (which describes the -dependence of the form factors) is proportional
to the soft-photon radiation function (1.12).
For more information about inclusive sum rules, see [3, 2, 21, 24].
It is also possible to establish a bound on the IsgurWise form factor at
the cut (Fig. 1.5). The decay rate W BD must be less than the total
decay rate W bc:
References
nl |(cosh )|2 | cosh + 1| Nc ,
17
(1.29)
where nl is the number of light avours, and Nc is the number of colours.

The left-hand side should be the sum over light avours, if SU (3) breaking
is taken into account. The inequality (1.29) is applicable at |cosh + 1| 2s
(outside the shaded region in Fig. 1.5), where the IsgurWise form factor
is given by the free-quark formula.
makes sense, and the decay rate W bc
At |cosh | 1, production of the pair of ground-state mesons is rare, and
the IsgurWise form factor (1.9) is much less than the bound (1.29). The
inequality (1.29) was proposed in [20], where the factor nl was erroneously
omitted. It was used, together with analyticity, to obtain bounds on the
IsgurWise form factor in the physical region.
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, Gif-surYvette 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
2 The HQET Lagrangian
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 two-loop 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 well-known facts about
massless diagrams, too.
2.1 The HQET Lagrangian

Let us consider QCD with a heavy avour Q with mass m, and a number
of light avours. We shall be interested in problems with a single heavy
quark that stays approximately at rest. More exactly, let m be the
characteristic momentum scale. We shall assume that the heavy quark has
a momentum |p| and an energy |p0 m| ; light quarks and gluons
have momenta |ki | and energies |k0i | . Heavy quark eective theory
(HQET) is an eective eld theory constructed to reproduce QCD results for
such problems expanded up to some order in /m. In practice, only the rst
few orders in the 1/m expansion are considered, because the complexity of
the theory grows rapidly with the order.
Let us start from the QCD Lagrangian
D
L = Q(i
/ m)Q +
(2.1)
where Q is the heavy-quark eld, and the dots mean all the terms with light
m)Q gives the
quarks and gluons. The free heavy-quark 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 lowest-energy 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
20
shall use the residual energy p0 = p0 m. Then the on-shell 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 two-component spinor Q
= Q).

it as a four-component 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 light-eld parts (denoted by dots) are exactly the same as in QCD.
The eld theory (2.2) is not Lorentz-invariant, because the heavy quark denes a selected frame its rest frame.
The Lagrangian (2.2) gives the static-quark 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 static-quark 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 half-plane; 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)
v is a four-component spinor obeying the relawhere the static-quark eld Q

v = Q
v , and v is the quark velocity. The momentum p of the heavy
tion v/Q
quark (or any state containing it) is related to the residual momentum p by
p = mv + p ,
|
p | m .
The static-quark propagator is
(2.6)
2.1 The HQET Lagrangian
1
p) = 1 + v/
S(
,
2 p v + i0
21
(2.7)
and the vertex is igv ta .

One can look [7] at how expressions for QCD diagrams tend to the corresponding HQET expressions in the limit m . The QCD heavy-quark
propagator is

p
m(1 + v/) + p
/ 1 + v/ 1
p+m
/
+O
=
=
S(p) = 2
.
(2.8)
p m2
2m
p v + p 2
2 p v
m
v )/2 may be replaced
A vertex ig ta sandwiched between two projectors (1+/
by igv ta (one may insert the projectors at external heavy-quark legs, too).
Therefore, any tree QCD diagram equals the corresponding HQET one up
to O(
p/m) terms. In loops, momenta can be arbitrarily large, and the relation (2.8) can break. Loops with momenta much larger than yield contributions local in the coordinate space, which can be removed by counterterms.
Then loop integrals become convergent, with characteristic loop momenta of
order , and one may use (2.8). Therefore, we can choose a renormalization
scheme of HQET (by tuning the coecients of local counterterms) in such a
way as to reproduce renormalized QCD results (see [11] for details).
The renormalization properties (anomalous dimensions, etc.) of HQET
dier from those of QCD. The ultraviolet behaviour of an HQET diagram is
determined by the region of loop momenta much larger than the characteristic momentum scale of the process , but much less than the heavy-quark
mass m (which tends to innity from the very beginning). It has nothing to
do with the ultraviolet behaviour of the corresponding QCD diagram with
the heavy-quark line, which is determined by the region of loop momenta
much larger than m. In conventional QCD, the rst region produces hybrid
logarithms which are dicult to sum. In HQET, hybrid logarithms become
ultraviolet logarithmic divergences governed by the renormalization group,
with corresponding anomalous dimensions.
The HQET Lagrangian (2.2) possesses the SU (2) heavy-quark spin symmetry [9]. If there are nh heavy-quark elds with the same velocity, it has the
D
SU (2nh ) spinavour symmetry. For example, if we consider the B D,

transitions with v = v, to the leading order in 1/mc,b , then the Lagrangian
has SU (4) spinavour symmetry, which relates all form factors to the B B
form factor at zero momentum transfer, which is equal to 1. If v = v, it has
only SU (2) SU (2) spin symmetry, which allows one to express all form
factors via a single unknown function of v v .
The heavy-quark mass is not uniquely dened. If we shift m by m in (2.6),
then the HQET Lagrangian (2.5) becomes
v i(v D m)Q
v + ,
Lv = Q
(2.9)
where m is called the residual mass term. The most convenient denition of
the heavy-quark mass m is the one which gives m = 0 in (2.9). It gives the
22
HQET propagator having a pole at p v = 0, or the QCD propagator having

a pole at p2 = m2 . This mass is called the pole mass; it is well dened and
gauge-invariant at any order of perturbation theory [10], but it is dicult
to dene it non-perturbatively. However, any mass diering from it by a
constant m (not growing with m) is equally good, and produces the same
physical results [5].
HQET has great advantages over QCD in lattice simulation of heavyquark problems. Indeed, the applicability conditions of the lattice approximation for problems with light hadrons require that the lattice spacing is
much less than the characteristic hadron size and that the total lattice length
is much larger than this size. For simulation of QCD with a heavy quark, the
lattice spacing must be much less than the heavy-quark Compton wavelength
1/m. For the b quark, this is technically impossible at present. The HQET
Lagrangian does not involve the heavy-quark mass m, and the applicability
conditions of the lattice approximation for HQET are the same as in the case
of light hadrons.
2.2 One-Loop Massless Propagator Diagrams

In this section, we shall calculate the one-loop massless propagator diagram
with arbitrary degrees of the denominators (Fig. 2.1),

dd k
G=
.
(2.10)
2
((k + p) i0)n1 (k 2 i0)n2
k+p
Fig. 2.1. One-loop massless propagator diagram
The rst step is to combine the denominators together. To this end, we

write 1/an for Re a > 0 as

1
1
=
ea n1 d
(2.11)
an
(n) 0
(-representation). Multiplying two such representations, we have

1
1
=
ea1 1 a2 2 n1 1 1 n2 2 1 d1 d2 .
an1 1 an2 2
(n1 ) (n2 )
(2.12)
2.2 One-Loop Massless Propagator Diagrams
23
Now we proceed to the new variables 1 = x, 2 = (1 x), and obtain

1
(n1 + n2 ) 1 xn1 1 (1 x)n2 1 dx
=
.
(2.13)
an1 1 an2 2
(n1 ) (n2 ) 0 [a1 x + a2 (1 x)]n1 +n2
This Feynman parametrization is valid not only when Re a1,2 > 0, but also,
by analytical continuation, in all cases where the integral in x is well dened.
Combining the denominators in (2.10) and shifting the integration momentum k k xp, we obtain

(n1 + n2 )
xn1 1 (1 x)n2 1 dx dd k
G=
.
(2.14)
(n1 ) (n2 )
[k 2 + x(1 x)(p2 ) i0]n1 +n2
Now we shall calculate the one-loop massive vacuum diagram, which appears as a sub-expression in many calculations. Performing the Wick rota2
tion k0 = ikE0 and going to the Euclidean space k 2 = kE
, we have (here
d/2
2 / (d/2) is the d-dimensional full solid angle)

d1
dkE
kE
2 d/2
dd k
=
i
(2.15)
2 + m2 )n .
(k 2 + m2 i0)n
(d/2)
(kE
At this point, the meaning of the d-dimensional integral for an arbitrary d
2
, and proceed
becomes completely well dened. Then we use 2kE dkE dkE
2
2
to the dimensionless variable z = kE /m to obtain
d/21
z
dz
i d/2
2 d/2n
(m )
.
n
(d/2)
(z
+
1)
0
The substitution x = 1/(z + 1) reduces this integral to the Euler B-function,
and we arrive at

dd k
(d/2 + n) 2 d/2n
(m )
= i d/2
.
(2.16)
(k 2 + m2 i0)n
(n)
We can now resume the calculation of the integral G (2.14). We shall
assume that p2 < 0, so that production of a pair of on-shell massless particles
is not possible. Using (2.16) with m2 x(1 x)(p2 ) and n n1 + n2 , we
obtain
G = i d/2
(d/2 + n1 + n2 )
(p2 )d/2n1 n2
(n1 ) (n2 )
1
xd/2n2 1 (1 x)d/2n1 1 dx .
This integral is a B-function. Our nal result can be written as

dd k
d/2
(p2 )d/2n1 n2 G(n1 , n2 ) ,
n1 n2 = i
D1 D2
D1 = (k + p)2 ,
D2 = k 2 ,
(2.17)
24
G(n1 , n2 ) =
(d/2 + n1 + n2 ) (d/2 n1 ) (d/2 n2 )

,
(n1 ) (n2 ) (d 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 loop-momentum 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 one-loop propagator
diagram of Fig. 2.1 in the coordinate space is just the product of two propagators. The massless propagators in the p-space and the x-space 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 x-space (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 one-loop 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 one-loop 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)
2.3 One-Loop HQET Propagator Diagrams
25
2.3 One-Loop HQET Propagator Diagrams

Now we shall calculate the one-loop HQET propagator diagram with arbitrary degrees of the denominators (Fig. 2.2),
dd k
.
((k + p)0 i0)n1 (k 2 i0)n2
I=
(2.22)
This diagram depends only on = p0 , and not on p.

k
k + p
Fig. 2.2. One-loop HQET propagator diagram
We cannot use the ordinary Feynman parametrization (2.13), because the

denominators have dierent dimensionalities. Therefore, we make a dierent
change of variables in the double -parametric integral (2.12), 1 = y,
2 = , and obtain the HQET Feynman parametrization

y n1 1 dy
1
(n1 + n2 )
.
(2.23)
n1 n2 =
a1 a2
(n1 ) (n2 ) 0 [a1 y + a2 ]n1 +n2
If the denominator a1 has the dimensionality of energy, and a2 has the dimensionality of energy squared, then the Feynman parameter y has the dimensionality of energy; it runs from 0 to . When combining the denominators
in (2.22), it is slightly more convenient to double the linear denominator;
shifting the integration momentum k k yv, we obtain

y n1 1 dy dd k
n1 (n1 + n2 )
(2.24)
I=2
n +n .
(n1 ) (n2 )
[k 2 + y(y 2) i0] 1 2
We must have a denite sign of i0 in the combined denominator, therefore,
the i0 terms in both denominators must have the same sign.
We shall assume that < 0, so that production of a pair of on-shell
particles is not possible. Using (2.16) with m2 y(y 2) and n n1 + n2 ,
we obtain

(d/2 + n1 + n2 ) n1 1
d/2n1 n2
y
[y(y 2)]
dy . (2.25)
I = i d/2 2n1
(n1 ) (n2 )
0
We proceed to the dimensionless variable z = y/(2). Then the substitution
z + 1 = 1/x reduces this integral to the Euler B-function:
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)
Our nal result can be written as

dd k
= i d/2 (2)d2n2 I(n1 , n2 ) ,
D1n1 D2n2
D1 = (k v + )/ ,
I(n1 , n2 ) =
D2 = k 2 ,
(d + n1 + 2n2 ) (d/2 n2 )
.
(n1 ) (n2 )
(2.27)
If n1,2 are integer, I(n1 , n2 ) is proportional to I1 = (1 + 2) (1 ), the

coecient being a rational function of d.
The integral (2.16) in dd k can have a UV divergence, and it produces
(d/2 + n1 + n2 ) in the numerator of (2.25). However, it is too optimistic
with respect to UV convergence: for example, at n1 = 2, n2 = 1 it has no
pole at d = 4, while the original integral (2.22) is UV divergent. The HQET
Feynman parameter y has the dimensionality of energy and runs up to . It
is natural to expect that this can produce an extra UV divergence. The region
y produces (d+n1 +2n2 ) in the Feynman parametric integral (2.26).
The wrong UV -function is cancelled by the denominator of (2.26), and
n1 + 2n2 determines the correct UV behaviour. The region y 0 produces
the IR (d/2 n2). The rule about the negative/positive sign of d in UV/IR
-functions remains valid.
The one-loop propagator diagram of Fig. 2.2 in the coordinate space is
just the product of two propagators. The HQET propagators in the p-space
and the x-space are related to each other by a Fourier transform:
+
in n1
eit
d
=
t
(t) ,
(2.28)
n
(n)
( i0) 2

(i)n+1 (n + 1)
eit tn dt =
.
(2.29)
( i0)n+1
0
The rst integral is non-zero only at t > 0, when we close the integration
contour downward; for integer n, its value is given by the residue at = i0.
In the second integral, we substitute + i0 for convergence. Multiplying
an HQET propagator (2.28) with degree n1 and a massless propagator (2.19)
with degree n2 , we obtain
22n2 d/2
(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).
2.4 Two-Loop Massless Propagator Diagrams
27
The one-loop diagram (2.27) is an analytic function in the complex

plane with a cut. The cut at > 0, where real pair production is possible,
begins at a branch point at the threshold = 0. This cut comes from the
factor ()2 , which has a discontinuity 2i 2 sin 2. Therefore, the
discontinuity of the one-loop diagram (2.27) at 0 is proportional to
the residue of I(n1 , n2 ) at = 0. The discontinuity of the diagram with
n1 = n2 = 1 can be directly calculated using the Cutkosky rules (2.21), and
1
2i(
p0 ) .

p0 i0
(2.30)

There is only one generic topology of two-loop massless propagator diagrams,
Fig. 2.3a. If one of the lines is shrunk to a point, the diagrams of Figs. 2.3b,c
result. If any two adjacent lines are shrunk to a point, the diagram contains
a no-scale vacuum tadpole, and hence vanishes.
3
4
5
5
1
2
a
2
b
2
c
Fig. 2.3. Two-loop massless propagator diagram
We write down the diagram of Fig. 2.3a as

dd k1 dd k2
d
2 d ni
G(n1 , n2 , n3 , n4 , n5 ) ,
n1 n2 n3 n4 n5 = (p )
D1 D2 D3 D4 D5
D1 = (k1 + p)2 ,
D3 = k12 ,
D2 = (k2 + p)2 ,
D4 = k22 ,
D5 = (k1 k2 )2 .
(2.31)
It is symmetric with respect to (1 2, 3 4), and with respect to (1

3, 2 4). If the indices of any two adjacent lines are non-positive, the diagram
contains a scale-free vacuum subdiagram, and hence vanishes. If n5 = 0, this
diagram is a product of two one-loop diagrams (Fig. 2.3b):
G(n1 , n2 , n3 , n4 , 0) = G(n1 , n3 )G(n2 , n4 ) .
(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
G(0, n2 , n3 , n4 , n5 ) = G(n3 , n5 )G(n2 , n4 + n3 + n5 d/2) .
(2.33)
Of course, the cases n2 = 0, n3 = 0, n4 = 0 follow by symmetry.

When all ni > 0, the problem does not immediately reduce to a repeated
use of the one-loop formula (2.18). We shall use a powerful method called integration by parts [12, 3]. It is based on the simple observation that any integral
of /k1 ( ) (or /k2 ( )) vanishes (in dimensional regularization, no surface terms can appear). From this, we can obtain recurrence relations which
involve G(n1 , n2 , n3 , n4 , n5 ) with dierent sets of indices. By applying these
relations in a carefully chosen order, we can reduce any G(n1 , n2 , n3 , n4 , n5 )
to trivial ones, such as (2.32), (2.33).
The dierential operator /k2 applied to the integrand of (2.31) acts as
n2
n4
n5
2(k2 + p) +
2k2 +
2(k2 k1 ) .
k2
D2
D4
D5
(2.34)
Applying (/k2 )k2 to the integrand of (2.31), we obtain a vanishing integral.

On the other hand, from (2.34), 2k2 k2 = 2D4 , 2(k2 + p) k2 = (p2 )
D2 D4 , and 2(k2 k1 ) k2 = D3 D4 D5 , we see that this dierential
operator is equivalent to inserting
n2
n5
d n2 n5 2n4 +
((p2 ) D4 ) +
(D3 D4 )
D2
D5
under the integral sign (here (/k2 ) k2 = d). Taking into account the
denition (2.31), we obtain the recurrence relation
(d n2 n5 2n4 )G(n1 , n2 , n3 , n4 , n5 )
+ n2 [G(n1 , n2 + 1, n3 , n4 , n5 ) G(n1 , n2 + 1, n3 , n4 1, n5 )]
+ n5 [G(n1 , n2 , n3 1, n4 , n5 + 1) G(n1 , n2 , n3 , n4 1, n5 + 1)] = 0 .
This relation looks lengthy. When using the integration-by-parts method, one
has to work with a large number of relations of this kind. Therefore, special
concise notation has been invented. Let us introduce the raising and lowering
operators
1 G(n1 , n2 , n3 , n4 , n5 ) = G(n1 1, n2 , n3 , n4 , n5 ) ,
(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)
One more relation is obtained by symmetry. Relations of this kind can be

written for any diagram having a triangle in it, when at least two vertices
of the triangle each have only a single line (not belonging to the triangle)
attached.
We can obtain a relation from the homogeneity of the integral (2.31)
with respect to p. Applying the operator p (/p)
to the integral (2.31),
we see that it is equivalent to a factor 2(d
ni ). On the other hand,
explicit dierentiation of the integrand gives (n1 /D1 )(p2 + D1 D3 )
(n2 /D2 )(p2 + D2 D4 ). Therefore,

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)
Three more relations follow from symmetries.

Expressing G(n1 , n2 , n3 , n4 , n5 ) by use of (2.37),
G(n1 , n2 , n3 , n4 , n5 ) =
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 (mirror-symmetric to the previous case). Of
course, if max(n2 , n4 ) < max(n1 , n3 ), it may be more ecient to use the
relation mirror-symmetric 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 n-loop sunset massless diagram are
Gn =
(2n + 1 nd/2) n+1 (d/2 1)

(1 + n) n+1 (1 )
=
. (2.41)
(1 (n + 1))
((n + 1)d/2 2n 1)
Methods of calculation of three-loop massless propagator diagrams are

considered in [3].
2.5 Two-Loop HQET Propagator Diagrams

There are two generic topologies of two-loop HQET propagator diagrams,
Figs. 2.4a,b. If one of the lines is shrunk to a point, the diagrams of
Figs. 2.4c,d,e result. If any two adjacent lines are shrunk to a point, the
diagram contains a no-scale vacuum tadpole, and hence vanishes.
We write down the diagram of Fig. 2.4a as

dd k1 dd k2
d
2(dn3 n4 n5 )
I(n1 , n2 , n3 , n4 , n5 ) ,
n1 n2 n3 n4 n5 = (2)
D1 D2 D3 D4 D5
D1 = (k1 + p) v/ ,
D3 =
k12 ,
D4 =
D2 = (k2 + p) v/ ,
k22 ,
D5 = (k1 k2 )2 .
(2.42)
It is symmetric with respect to (1 3, 2 4). If the indices of any two

adjacent lines are non-positive, the diagram contains a scale-free vacuum
subdiagram, and hence vanishes. If n5 = 0, this diagram is a product of two
one-loop diagrams (Fig. 2.4c):
2.5 Two-Loop HQET Propagator Diagrams
31
4
2
3
5
1
3
5
2
a
b
4
4
5
2
c
2
d
5
1
2
e
Fig. 2.4. Two-loop HQET propagator diagram
I(n1 , n2 , n3 , n4 , 0) = I(n1 , n3 )I(n2 , n4 ) .
(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)
(and similarly for n4 = 0).

When all ni > 0, we apply integration by parts [1]. The dierential operator /k2 applied to the integrand of (2.42) acts as
n4
n2 v
n5
+
2k2 +
2(k2 k1 ) .
k2
D2 D4
D5
(2.46)
Applying (/k2 ) k2 and (/k2 ) (k2 k1 ) to the integrand of (2.42), we

obtain vanishing integrals. On the other hand, from (2.46), k2 v/ = D2 1,
and 2(k2 k1 ) k2 = D3 D4 D5 , we see that these dierential operators
are equivalent to inserting
n2
n5
+
(D3 D4 ) ,
D2
D5
n2
n4
d n2 n4 2n5 +
D1 +
(D3 D5 )
D2
D4
d n2 n5 2n4 +
under the integral sign. Therefore, we obtain the triangle relations

d n2 n5 2n4 + n2 2+ + n5 5+ (3 4 ) I = 0 ,

d n2 n4 2n5 + n2 2+ 1 + n4 4+ (3 5 ) I = 0
(2.47)
(2.48)
32
(two more relations are obtained by (1 3, 2 4)). Similarly, applying the

dierential operator 2(/k2 ) v is equivalent to inserting
2
n4 2
n5 2
n2
+
4 (D2 1) +
4 (D2 D1 ) ,
D2
D4
D5
and we obtain (taking into account the denition (2.42))

2n2 2+ + n4 4+ (2 1) + n5 5+ (2 1 ) I = 0
(2.49)
(there is also the symmetric relation, of course).

We can obtain a relation from the homogeneity of the integral (2.42) with
respect to . Applying the operator (d/d) to the integral (2.42) multiplied
by ()n1 n2 , we see that it is equivalent to a factor 2(d n3 n4 n5 )
n1 n2 . On the other hand, explicit dierentiation of (D1 )n1 (D2 )n2
gives n1 /D1 n2 /D2 . Therefore,

2(d n3 n4 n5 ) n1 n2 + n1 1+ + n2 2+ I = 0 .
(2.50)
This is nothing but the sum of the (/k2 ) k2 relation (2.47) and its mirrorsymmetric (/k1 ) k1 relation.
When trying to nd the most useful combinations of recurrence relations,
it is convenient to manipulate these relations in an algebraic way. We shall
consider the raising and lowering symbols, such as 1 , as operators not commuting with n1 , etc. Traditionally, these operators are written to the right of
the ni factors. Of course, any relation may be multiplied (from the left) by
any ni factors. We can also apply a shift, say n1 n1 1, everywhere in the
relation. This operation can be represented by multiplication by 1 from the
left followed by commuting this operator to the right, if we assume
1 n1 = (n1 1)1 .
(2.51)
The most useful combination of recurrence relations for the integral

I (2.42) is the triangle relation (2.48) minus 1 times the homogeneity relation (2.50):

d n1 n2 n4 2n5 + 1 2(d n3 n4 n5 ) n1 n2 + 1 1

+ n4 4+ (3 5 ) I = 0 .
(2.52)
Expressing I(n1 , n2 , n3 , n4 , n5 ) by use of this relation,
I(n1 , n2 , n3 , n4 , n5 )
=
(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
with integer indices to a combination of integrals with n5 = 0 (Fig. 2.4c,

(2.43)), n1 = 0 (Fig. 2.4d, (2.44)), and n3 = 0 (Fig. 2.4e, (2.45)). Of course,
if max(n2 , n4 ) < max(n1 , n3 ), it may be more ecient to use the relation
mirror-symmetric to (2.52). Thus, any integral I(n1 , n2 , n3 , n4 , n5 ) with integer ni can be expressed as a linear combination of I12 and I2 , the coecients
being rational functions of d. Here the combinations of -functions appearing
in the n-loop sunset HQET diagram are
In = (1 + 2n) n (1 ) = (4n + 1 nd) n (d/2 1) .
(2.54)
We write down the diagram of the second topology, Fig. 2.4b, as

dd k1 dd k2
d
2(dn4 n5 )
J(n1 , n2 , n3 , n4 , n5 ) ,
n1 n2 n3 n4 n5 = (2)
D1 D2 D3 D4 D5
D1 = (k1 + p) v/ ,
D4 =
k12 ,
D5 =
D2 = (k2 + p) v/ ,
k22 .
D3 = (k1 + k2 + p) v/ ,
(2.55)
It is symmetric with respect to (1 2, 4 5). If n4 0, or n5 0, or

two adjacent heavy indices are non-positive, the diagram vanishes. If n3 = 0,
this diagram is a product of two one-loop diagrams (Fig. 2.4c). If n1 = 0 or
n2 = 0, it is a diagram of Fig. 2.4e. When n1,2,3 are all positive, we can insert
1 = D1 + D2 D3
into the integrand of (2.55), and obtain
J = (1 + 2 3 )J .
(2.56)
This reduces n1 + n2 + n3 by 1. Therefore, by applying (2.56) suciently

many times, we can reduce an arbitrary J integral with integer indices to a
combination of integrals with n1 = 0, n2 = 0, n3 = 0.
Methods of calculation of three-loop HQET propagator diagrams are considered in [8, 2].
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
In this chapter, we discuss renormalization of HQET at the leading order

in 1/m. Most of it is identical to the renormalization of QCD, which is
also discussed. The abelian version of HQET (heavy-electron eective theory,
Sect. 3.5) has some unique and beautiful features, which are useful for understanding some aspects of renormalization of the real (non-abelian) HQET.
3.1 Renormalization of QCD

The QCD Lagrangian expressed via the bare (unrenormalized) quantities
(denoted by the subscript 0) is
L=

i
1
1
2
qi0 (iD
/ 0 mi0 )qi0 Ga0 Ga
( A0 ) + ( ca0 )(D0 ca0 ) ,

0
4
2a0
(3.1)
where D0 q0 = ( ig0 Aa0 ta )q0 , a0 is the gauge-xing parameter, c is the

ghost eld, and D0 ca0 = ( ab g0 f abc Ac )cb0 . The renormalized quantities
are related to the bare ones by
qi0 = Zq1/2 qi ,
g0 = Z1/2 g ,
1/2
A0 = ZA A ,
mi0 = Zm mi ,
c0 = Zc1/2 c ,
a0 = Z A a ,
where the renormalization factors have the minimal structure

Z22
Z11 s
Z21 s 2
Z =1+
+
+
+
4
2
(3.2)
(3.3)
It will be explained in Sect. 3.2 why the gauge-xing 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 ] =
36
3 Renormalization
(d 1)/2 = 3/2 . The covariant derivative D0 = ig0 Aa0 ta has

the dimensionality 1, hence the dimensionality of the coupling constant is
[g0 ] = 2 d/2 = .
We dene s to be exactly dimensionless:
g2
s ()
= 2
e
4
(4)d/2
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 renormalization-group 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)
This s () obeys the same renormalization-group equation (3.5) with the

same function (s ). When a bare physical quantity A0 (p2 ) is multiplied
by the corresponding renormalization constant Z 1 (s ()), it becomes nite, with the coecients of the perturbative series depending on L =
1

log (p2 )/2 . If we
A0 by
multiply
Z (s ()), we obtain the same expres2
2
sion with L = log (p )/(f ()) . This expression is also nite at 0;
hence, Z(s ) is the correct renormalization factor in the new scheme. Therefore, the anomalous dimensions (s ) of A are the same in both schemes.
When a renormalized physical quantity is expressed via s (), we may take
the limit 0 in such expression. Therefore, renormalized results expressed
via s () coincide with the results expressed via s (f (0)). Such schemes are
equivalent to just rescaling .
We can consider more general redenitions

2
s ()
s ()
s ()
s ()
+ .
=
1 + c1
+ c2
4
4
4
4
3.1 Renormalization of QCD
37
The term with c1 can be absorbed by a rescaling of ; therefore, we shall

assume c1 = 0. The new s () obeys the renormalization-group equation (3.5)
with a new (s ):
0 = 0 ,
1 = 1 ,
2 = 2 + c2 , . . .
The rst two -function coecients 0,1 are the same in all mass-independent
(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)
Hence, one can obtain (s ) from Z1 , the coecient of 1/ in Z , and vice

versa. Higher poles (1/2 , 1/3 , . . . ) contain no new information; at each order
in s , their coecients (Z2 , Z3 , . . . ) can be reconstructed from lower-loop
results. Up to two loops,

1
s 2
s
+
(3.11)
+ 02 1
Z = 1 0
4
2
4
Since a0 does not depend on , a() obeys the renormalization-group
equation
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)
It is more convenient to present results for anomalous dimensions instead

of renormalization constants, because the anomalous dimensions contain the
same information but are more compact. If we are considering the renormalization of a gauge-invariant quantity, then Z does not depend on a(), and
the only source of its -dependence is s ():
(s ) (s )(s ) 2 (s )(s )
d log Z
1 (s )
=
+
=
+
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 )
When the quantity being considered is gauge-dependent, Z depends on

via both s and a:
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
3.2 Gluon Propagator
Z2 =
1
2
A (s )a
Z1 (s ) ds
1
d0 s 2
= A0 a
+
a
s
8
da 4
39
(3.18)
Equating 2 terms, we can nd Z3 , and so on. Up to two loops, the

renormalization constant is

d0
s 2
1
1 s
+
+
0 (0 + 20 ) + A0 a
21
Z = 1 0
2 4 8
da
4
(3.19)
One can obtain the anomalous dimension (s ) from Z1 , the coecient of 1/
in Z, and vice versa. The coecients Z2 , Z3 , . . . contain no new information;
at each order in s , they can be reconstructed from lower-loop results.

The bare (unrenormalized) gluon propagator iD (p) has the structure
(Fig. 3.1)
0
0
0
(p) + (i)D
(p)i (p)(i)D
(p)
iD (p) = iD
0
0
0
(p)i (p)(i)D
(p)i (p)(i)D
(p)
+ (i)D
+ ,
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 self-energy (polarization operator) i (p) is the sum of one-particle-irreducible gluon self-energy 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)
To solve this equation, it is convenient to introduce the inverse tensor A1
of a symmetric tensor A satisfying A1

= . If
A

p p
p p
+ A|| 2 ,
A = A g 2
p
p
Fig. 3.1. Structure of diagrams for the gluon propagator
40
3 Renormalization
then
A1
A1

p p
p p
.
g 2
+ A1
||
p
p2
Using this notation, the equation (3.22) can be rewritten as

1
D
(p) = (D0 )1
(p) (p) .
(3.23)
Owing to the Ward identity (p)p = 0, the gluon self-energy is transverse:

(p) = (p2 g p p )(p2 ) .
Therefore, the gluon propagator is

1
p p
p p
+ a0 2 2 .
D (p) = 2
g 2
p (1 (p2 ))
p
(p )
(3.24)
(3.25)
Its longitudinal part acquires no corrections. The renormalized gluon propr

agator is related to the bare propagator by D (p) = ZA ()D
(p; ). The
gluon eld renormalization constant ZA is constructed to make the transverse
r
(p; ) nite in the limit 0:
part of D
r
(p; )
D
r
D
(p2 ; )

p p
p p
g 2
+ a() 2 2 .
p
(p )
(3.26)
At the same time, it converts a0 to the renormalized gauge parameter a().

This is the reason why the gauge parameter is renormalized (3.2) by the same
constant ZA .
In the one-loop approximation (Fig. 3.2), we can use the methods of
Sect. 2.2 to obtain
(p2 ) =
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)
Fig. 3.2. One-loop gluon self-energy
41
where G1 is dened in (2.41) and = 1 a0 . When dealing with higher-loop

diagrams, it is often convenient to calculate a diagram with a quark-loop
insertion, and then to incorporate the other two insertions in Fig. 3.2 by
means of replacing TF nf with
P = T F nf
3d 2 + (d 1)(2d 7) (1/4)(d 1)(d 4) 2

CA . (3.28)
4(d 2)
Now we re-express the one-loop propagator

p2 D (p2 ) =
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() .
We expand the result in using

(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)
is the Riemann -function. The Euler constant cancels, and we arrive at

13
1
s L
4
e
a
CA TF nf
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 renormalization-group equation
r
1
(p2 ; )
dD
r
= A (s (), a())D
(p2 ; ) .
dL
2
(3.34)
Therefore, the coecient of L in (3.32) is just A0 /2. It is enough to know

r
r
the initial condition D
(p2 ; 2 = p2 ) in order to reconstruct D
(p2 ; )
from (3.34).
It is straightforward to calculate (p2 ) at two loops (Fig. 3.3) by the
methods of Sect. 2.4 using any suitable computer algebra system. The utmost
attention to detail is required: do not forget mirror-symmetric diagrams if
needed; include symmetry factors and minus signs for quark and ghost loops;
use consistent leg orderings for the colour and Lorentz parts of three-gluon
Fig. 3.3. Two-loop gluon self-energy
43
vertices; be careful while decomposing four-gluon vertices into three terms

with dierent colour structures. Every perturbative-QCD practitioner has
performed this tedious calculation at least once, because the result is often
needed for higher loops. Here is the result, for reference:

g04 (p2 )2
2
CA
2(d2 4d 2) (6d3 51d2 + 120d 56)
2 =
(4)d
1
2(d 1)(d 4)(2d2 14d + 25) 2 + (d 1)(d 4)2 (8d 27) 3
8

1
3 4
(d 1)(d 4) J1
16

+ 2(3d5 59d4 + 376d3 1084d2 + 1648d 1248)
2(43d4 475d3 + 1764d2 2388d + 656)
1
+ (d 4)(3d 8)(9d4 118d3 + 547d2 1034d + 608) 2
2

1
2
2
3
(d 1)(d 4) (3d 8)(d 9d + 22) J2
4

+ CF TF nf 8(d 2) (d2 7d + 16)J1 + 4(d 3)(d 6)(d2 4d + 8)J2

+ CA TF nf 2(d 2) 2(d2 5d + 8) + 2(d 4)(4d 13) (d 4)2 2 J1

4 d4 5d3 56d2 + 364d 528

2
(3.35)
+ (d 4)(3d 8)(3d 27d + 56) J2
where
G21
,
(d 1)(d 3)2 (d 4)3
G2
J2 =
3
(d 1)(d 3)(d 4) (d 6)(3d 8)(3d 10)
J1 =
(see (2.41)).
The gluon propagator with two-loop accuracy,
p2 D (p2 ) =
1
= 1 + 1 (p2 ) + 12 (p2 ) + 2 (p2 ) ,
1 (p2 )
has the form

p2 D (p2 ) = 1 +
g02 (p2 )
g04 (p2 )2
f
(a
)
+
f2 (a0 ) .
1
0
(4)d
(4)d/2
Now we are going to re-express it via the renormalized quantities s () and

a(). In the one-loop term, we have to use the expressions including one-loop
corrections,
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)
For simplicity, we set 2 = p2 , i.e., L = 0:

s ()
c10 + c11 + c12 2 +
p2 D (p2 ) = 1 +
4
2
s
c20 + c21 + c22 2 +
+
4

1
d
2
c10 + c11 + c12 + ,
0 + A0 a
2
da
where the expansions
e f1 (a) = c10 + c11 + c12 2 + ,
2 e2 f2 (a) = c20 + c21 + c22 2 + ,
(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 .
Using this and equating the coecients in the 2s term, we obtain

1
d
z20 = c20 0 + A0 a
c10 ,
2
da

1
d
z21 = c21 c10 + 0 + A0 a
c11 ,
2
da

1
d
r2 = c22 c10 + 0 + A0 a
c12 .
2
da
3.3 Quark Propagator
45
The renormalization constant should have the structure (3.19). Therefore,

the leading coecient of the two-loop term must satisfy the self-consistency
condition

1
1
d
c20 =
(3.38)
c10 + 0 + A0 a
c10 .
2
2
da
When calculating the expansions (3.37) for the gluon propagator, it is
convenient to use
G2
= 1 63 3 +
G21
(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 three-loop 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 renormalization-group 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).

The bare (unrenormalized) quark propagator iS(p) has the structure (Fig. 3.4)
Fig. 3.4. Structure of diagrams for the quark propagator
46
3 Renormalization
iS(p) = iS0 (p) + iS0 (p)(i)(p)iS0 (p)

+ iS0 (p)(i)(p)iS0 (p)(i)(p)iS0 (p) + ,
(3.41)
1
p + m0
/
= 2
p m0
/
p m20
(3.42)
where
S0 (p) =
is the free quark propagator, and the quark self-energy (mass operator)
i(p) is the sum of one-particle-irreducible quark self-energy 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)
with the solution

S(p) =
1
1
.
=
/p m0 (p)
(p)
S01 (p)
(3.44)
The quark self-energy has the structure

(p) = p
/V (p2 ) + m0 S (p2 ) .
(3.45)
Therefore, the quark propagator is

S(p) =
1
1
= Zq Sr (p; ) ,
1 V (p2 ) /
p (1 V (p2 ))1 (1 + S (p2 ))m0
(3.46)
and the renormalization constants Zq , Zm are constructed to make Zq (1V )

and Zm Zq (1 + S ) nite.
For a massless quark in the one-loop approximation (Fig. 3.5),
V (p2 ) = CF
g02 (p2 )
d2
G1 a0 .
(4)d/2 (d 3)(d 4)
(3.47)
The one-loop self-energy vanishes in the Landau gauge a0 = 0. Now we reexpress the one-loop 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
Fig. 3.5. One-loop quark self-energy
47
Therefore,
Zq (s , a) = 1 CF a
s
,
4
pSr (p; ) = 1 + CF a(L 1)

/
(3.48)
s
.
4
(3.49)
The anomalous dimension of the quark eld at one loop is

q = 2CF a
s
+
4
(3.50)
UV divergences do not depend on masses; therefore, this result is also valid

for a massive quark. The coecient of L in (3.49) is just q0 /2, as required
by the renormalization-group equation.
Let us also rederive this result in the coordinate space. Transforming the
gluon propagator (3.21) to x-space (2.19), we obtain
0
(x) =
D
i (d/2 1) (1 + a0 )x2 g + (d 2)(1 a0 )x x

.
8 d/2
(x2 + i0)d/2
(3.51)
Similarly, the propagator (3.42) for a massless quark (m = 0) becomes

S0 (x) =
x
/
(d/2)
.
2 d/2 (x2 + i0)d/2
(3.52)
The one-loop light-quark self-energy (Fig. 3.5) is

0
(x) = iCF g02 S0 (x) D
(x) = CF g02 a0
2 (d/2)
/x
.
d
2
4
(x + i0)d1
(3.53)
Transforming it to p-space (2.20), we recover (3.47).

Two-loop diagrams for the quark self-energy are shown in Fig. 3.6. The
rst one contains the one-loop gluon self-energy (3.27); it can be easily calculated using (2.18), and is proportional to G2 (2.41) and P (3.28):
Fig. 3.6. Two-loop quark self-energy
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)
The second one is also recursively one-loop; it is proportional to CF2 G2 . It is

also not dicult to understand its dependence on the gauge parameter a0 .
The quark propagator with the one-loop correction inside the exterior loop
is proportional to (/
k+p
/)/[(k + p)2 ]n with n = 1 + . The resulting integral
is proportional to p
/, and we may take (1/4) Tr of the numerator with p//p2
to nd the coecient. Substituting the gluon propagator, we see that the
exterior loop integral is proportional to
(d 3 + a0 ) [G(1, n) + G(1, n 1)]
+ (1 a0 ) [G(2, n) 2G(2, n 1) + G(2, n 2)]
= a0
(d 2)(d 2n)
G(1, n) .
dn1
The same is true for the interior gluon line, with n = 1, as in the one-loop
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 colour-singlet 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 three-gluon 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
Fig. 3.7. Cvitanovic algorithm
1
2
1
Nc
=0

1
=
2
Nc
1
2
Fig. 3.8. Checking the coecients

1
=
2
=
Nc
Nc2 1
2Nc

1
=
2
Nc
1
2Nc
Fig. 3.9. Sample applications of Cvitanovic algorithm
Fig. 3.10. Three-gluon 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
+ TF nf 8(d 2)(d 4)J2
(3.56)
where
J1 =
G21
,
(d 3)2 (d 4)2
J2 =
G2
(d 3)(d 4)2 (d 6)(3d 8)(3d 10)
(see (2.41)). Now we expand the p

/S(p) = [1V ]1 to g04 and follow the same
procedure as in Sect. 3.2. The self-consistency condition (3.38) is satised,
the quark eld anomalous dimension is

s 2
s
a2 + 8a + 25
+ CF 3CF +
CA 4TF nf
q = 2CF a
+
4
2
4
(3.57)
(see [3] for the four-loop result), and the renormalized massless-quark propagator at 2 = p2 is
s ()
pSr (p; 2 = p2 ) = 1 CF a()
/
4

5
9a2 + 52a + 82
2
a +
+ CF
CF + 3(a + 1)3
CA
8
8
2
7
s
+ T F nf
.
2
4
3.4 Renormalization of HQET

The HQET Lagrangian expressed via the bare (unrenormalized) quantities
is
iv D0 Q
0 + LQCD ,
L=Q
0
.
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 heavy-quark propagator in HQET.
51

The bare (unrenormalized) static-quark 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 one-particle-irreducible HQET
self-energy diagrams (which cannot be separated into two disconnected parts
by cutting a single heavy-quark line). Summing this series, we obtain

S()
=
1
.

()
(3.60)
Fig. 3.11. Structure of diagrams for the heavy quark propagator in HQET
In the one-loop approximation (Fig. 3.12),

()
= iCF
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 scale-free 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
where I1 is dened in (2.54).
Fig. 3.12. One-loop heavy quark self-energy in HQET
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
Transforming this to p-space (2.29), we recover (3.61).

Therefore, with one-loop accuracy, the heavy-quark propagator in HQET
is

S()
= 1 + CF
g02 (2)2 I1
2A .
(4)d/2 d 4
(3.63)
In the coordinate space,

2
2
= S0 (t) 1 + CF g0 (it/2) ()A + ,
S(t)
(4)d/2
S0 (t) = i(t) . (3.64)
Now we re-express S via the renormalized quantities s (), a(). Expanding

the result in up to the nite term, we obtain

S()
= 1 + CF
s () 2L
e
(3 a() + 4) ,
4
where L = log
2
.
Q (s (), a()) Sr (; ), where the renormalizaThis should be equal to Z

tion constant ZQ (s , a) has the minimal form (3.3), and the renormalized
propagator Sr (; ) is nite at 0. We obtain
Q (s , a) = 1 + CF s (3 a) ,
Z
4
s ()
2 [(a() 3)L + 2] .
Sr (; ) = 1 + CF
4
(3.65)
(3.66)
Using (3.19), we arrive at the anomalous dimension
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 renormalization-group equation
dSr (; )
=
Q (s (), a())Sr (; ) .
dL
(3.68)
Therefore, the coecient of L in (3.66) is just

Q0 . It is enough to know the
initial condition Sr (; = 2) in order to reconstruct Sr (; ) from (3.68).
53
Two-loop diagrams for the heavy-quark self-energy are shown in Fig. 3.13.
The rst one contains the one-loop gluon self-energy (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)
The second diagram is also recursively one-loop; it is proportional to CF2 I2 .

It is also not dicult to understand its dependence on the gauge parameter
a0 . The exterior loop contains the heavy-quark line with a non-integer index
n (in fact, n = 1 + 2). The propagator (3.21) of the exterior gluon produces,
as compared with the Feynman gauge a0 = 1, an extra factor
1 a0 I(n, 2) 2I(n 1, 2) + I(n 2, 2)
4
I(n, 1)
1
1
= 1 + (1 a0 )(d 3) = (d 3)A .
2
2
1+
The same is true for the interior gluon line, with n = 1, as in the one-loop
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)
Fig. 3.13. Two-loop heavy quark self-energy in HQET
The diagram Fig. 3.13c, after killing one of the heavy-quark 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 heavy-quark propagator with two-loop 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)
Now we re-express it via s (3.4) and a (3.2) and expand it in . Instead

of (3.39), we now have
I2
= 1 + 42 2 163 3 +
I12
(3.74)
and hence 2 appears in the results. The self-consistency condition (3.38) is

satised; the anomalous dimension of the static-quark eld is [1]
2

3a + 24a 179
s 2
s
32
+ CF
CA + TF nf
+
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 three-loop anomalous dimension has been calculated recently [6, 2];
the dierence (3.76) is not gauge-invariant at three loops. The renormalized
static-quark propagator at = 2 is
3.5 Heavy-Electron Eective Theory
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 three-loop result).

Now we make a short digression into the abelian version of HQET the
heavy-electron eective theory, an eective eld theory of QED describing
the interaction of a single electron with soft photons. It is obtained by setting
CF 1, CA 0, g0 e0 , s . This theory was considered long ago,
and is called the BlochNordsieck model.
Suppose we calculate the one-loop correction to the heavy-electron propagator in the coordinate space. Let us multiply this correction by itself
(Fig. 3.14). We obtain an integral in t1 , t2 , t1 , t2 with 0 < t1 < t2 < t,
0 < t1 < t2 < t. The ordering of primed and non-primed integration times
can be arbitrary. The integration area is subdivided into six regions, corresponding to the six diagrams in Fig. 3.14. This is twice the two-loop correction
to the propagator. Continuing this drawing exercise, we see that the one-loop
correction cubed is 3! times the three-loop correction, and so on. Therefore,
the exact all-order propagator is the exponential of the one-loop term:
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)
In this theory, ZA = 1, because there exist no loops which can be inserted

into the photon propagator. Now we are going to show that Z = 1, too.
To this end, let us consider the sum of all one-particle-irreducible vertex
diagrams, not including the external leg propagators the electronphoton
proper vertex. It has the same structure as the tree-level term: ie0 v , =
where is the sum of all unrenormalized diagrams starting from one
1 + ,
loop. Let us multiply the vertex by the incoming photon momentum q . This
product can be simplied by the Ward identity for the electron propagator
(Fig. 3.15):
i
i
[
p v p v]
p v
p v

= ie0 S0 (
p ) S0 (
p) .
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
Fig. 3.15. Ward identity for the free electron propagator
we can obtain a set of diagrams for

Starting from each diagram for ,
by inserting the external photon vertex into each electron propagator. After
multiplying by q , each diagram in this set becomes a dierence. All terms
cancel each other, except the extreme ones (Fig. 3.16), and we obtain the
Ward identity

) = ( ) ()
(,

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 photon-eld renormalization, we may
replace e0 e, a0 a in the bare propagator (3.77). This propagator is
57

= e0
= e0
Fig. 3.16. Ward identity for the electronphoton vertex
made nite by the minimal (in the sense of (3.3)) renormalization constant,
which is just the exponential of the one-loop term

Q = exp (a 3) ,
(3.80)
Z
4
and the anomalous dimension is exactly equal to the one-loop 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
two-loop anomalous dimension of the eld (3.75).
References
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.
4 The HQET Lagrangian: 1/m Corrections
This chapter is about the rst 1/m correction to the HQET Lagrangian. It
contains two terms: the heavy-quark 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 on-shell quark scattering amplitudes in an external
chromomagnetic eld in QCD and HQET. To this end, methods for on-shell
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 heavy-avour
masses and when crossing a threshold.
4.1 1/m Corrections to the HQET Lagrangian

As discussed in Sect. 2.1, we are interested in processes with characteristic
momenta and energies m. The heavy quark eective theory is constructed to reproduce QCD S-matrix elements expanded up to some order
in /m. There is a folk theorem that any S-matrix having all the required
properties follows from some Lagrangian. Therefore, the HQET Lagrangian
is constructed as a series in 1/m containing all operators having the necessary symmetries, with arbitrary coecients. These coecients are tuned
to reproduce several QCD S-matrix elements expanded to some degree in
/m. We must perform this matching for a sucient number of amplitudes
to x all the coecients in the Lagrangian. After that, we can use this HQET
Lagrangian instead of the QCD one for calculating other amplitudes.

The HQET Lagrangian is not unique, because the heavy-quark eld Q
can be redened. Such eld transformations can be used to eliminate all time
except in the leading term (2.2) [26].
derivatives D0 acting on Q,
The velocity v can be varied by a small amount v /m without violating the applicability of HQET or changing its predictions. This
reparametrization invariance [31] (see also [14, 21]) relates terms of dierent
orders in 1/m.
At the level of 1/m terms, the heavy-quark spin symmetry and the superavour symmetry are violated by interaction of the heavy-quark chromomagnetic moment with the magnetic component of the gluon eld. This leads
60
to hyperne splittings between states which were degenerate in the innitemass limit (such as B and B ), as well as to violation of leading-order 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 dimension-5 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 heavy-quark 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 kinetic-energy 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 kinetic-energy 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
Fig. 4.1. Kinetic-energy vertices
In Sect. 3.4, we considered the sum of all one-particle-irreducible heavy

quark self-energy diagrams i()
in the innite-mass limit. Let us denote
0
2
k (, p ) the sum of all bare one-particle-irreducible selfby i(Ck /(2m))
energy diagrams at the order 1/m. Each of these diagrams contains a sin for
gle kinetic-energy 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 heavy-quark propagators 1/(
p v + i0)
produces insertions i
pi v into each propagator in turn. Variations of the
4.1 1/m Corrections to the HQET Lagrangian
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
No-gluon kinetic vertices (Fig. 4.1) produce i(Ck0 /m)
pi p ; single-gluon kinetic vertices (Fig. 4.1) produce i(Ck0 /m)g0 ta p ; two-gluon kinetic vertices
do not change. Therefore,
k
.
=2
p
v
(4.3)
This is the Ward identity of reparametrization invariance. Taking into ac
k /p2 )p and /v

k /p = 2(
= (d/d)p
count

, we obtain
k
d
.
=
2
p
d
(4.4)
The right-hand side does not depend on p2 , and hence

k (, p2 ) = d() p2 +
k0 () .
(4.5)
This result can also be understood in a more direct way. The momentum
p ows through the heavy-quark line. No-gluon kinetic vertices are quadratic
in it; one-gluon vertices are linear; two-gluon vertices are independent of p .
The p2 term comes from diagrams with a no-gluon kinetic vertex. Terms
linear in p vanish owing to the rotational symmetry. The coecient of p2
in a no-gluon kinetic vertex is iCk /(2m). Therefore, the coecient of p2 in
the sum of all diagrams is the sum of the leading-order HQET diagrams with
a unit operator insertion into each heavy-quark 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. One-loop diagrams for
62
As we discussed at the beginning of this section, the coecients in the

HQET Lagrangian are obtained by equating on-shell scattering amplitudes in
full QCD and in HQET with the required accuracy in 1/m. A prerequisite for
this matching is the requirement that the mass shell itself is the same in both
theories, to the accuracy considered. The mass shell is dened
as the position
of the pole of the full quark propagator. In QCD it is p0 = m2 + p2 , where
m is the on-shell mass (see Sect. 4.2 for more details). To the rst order in
1/m, this means = p2 /(2m). In HQET, the mass shell is the zero of the
denominator of the bare heavy-quark propagator

S(p)
=

()
.

d()
2

p
p + k0 ()
2m
d
Ck0
(4.7)

k0 (0) = 0 at the two-loop 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 residual-mass 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
kinetic-energy operator is not renormalized; its anomalous dimension is zero
to all orders. The coecient of the kinetic-energy operator in the HQET
Lagrangian is exactly unity,
Ck () = 1 ,
(4.9)
to all orders in perturbation theory, owing to the reparametrization invariance!

In the case of a spin-(1/2) heavy quark, there is one more dimension-5
operator, in addition to the kinetic energy the chromomagnetic interaction [19, 20]
+
+ iD0 Q
L=Q
Ck + 2 Cm +
+
Q D Q
Q B Q
2m
2m
(4.10)
or, in covariant notation,

v iv DQ
D2 Q
v + Cm Q
G Q
v + ,
v Ck Q
Lv = Q
2m v
4m v
(4.11)
where G = gGa ta . In the v rest frame, only chromomagnetic components

and Q
yields zero.
of G contribute, because 0i sandwiched between Q
4.2 On-Shell Renormalization of QCD
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 heavy-quark
+ 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
lower-order 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. Chromomagnetic-interaction 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 heavy-quark 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.

Now we are going to discuss the calculation of on-shell propagator diagrams
of a massive quark in QCD. Let us write the one-loop diagram with arbitrary
degrees of the denominators (Fig. 4.4) as
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
k + mv
Fig. 4.4. One-loop on-shell 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 one-loop 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 on-shell integral (4.13) can be cast into the HQET form (4.14) using
the inversion [13] K = K /K 2 . The on-shell 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
on-shell 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 two-loop on-shell 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 two-loop propagator
diagram. The on-shell denominators D1,2 produce the HQET denominators,
65
4
2
3
5
1
2
a
Fig. 4.5. Two-loop on-shell propagator diagrams
just as in the one-loop case. The denominator D5 becomes (K1 K2 )2 =

(K1 K2 )2 /(K12 K22 ). We obtain [13]
M (n1 , n2 , n3 , n4 , n5 ) = I(n1 , n2 , d n1 n3 n5 , d n2 n4 n5 , n5 ) .
(4.17)
However, this relation is not particularly useful for calculation of M (n1 , n2 , n3 ,
n4 , n5 ), because this HQET integral contains two non-integer indices. The
integrals M can be calculated using integration by parts recurrence relations. This is not so easy as in the massless case (Sect. 2.4) and the HQET
case (Sect. 2.5). We shall not discuss the algorithm here; it can be found
in the original literature [24, 11, 10, 23]. The conclusion is that any integral
M (n1 , n2 , n3 , n4 , n5 ) with integer indices ni can be expressed as a linear combination of M12 (Fig. 4.5c) and M2 (Fig. 4.5d), the coecients being rational
functions of d. Here the combinations of -functions appearing in n-loop
on-shell sunset diagrams with a single massive line are
Mn =
(1 + (n 1)) (1 + n) (1 2n) n (1 )
.
(1 n) (1 (n + 1))
(4.18)
The type N (Fig. 4.5b),

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
a single dicult integral N (1, 1, 1, 0, 0) (Fig. 4.5e), with rational coecients.

Instead of using N (1, 1, 1, 0, 0) as a basis integral, it is more convenient to use
the convergent integral N (1, 1, 1, 1, 1). This integral can be expressed via the
3F2 hypergeometric functions with a unit argument and indices depending on
. Several terms of the expansion in are known [9, 10]:
N (1, 1, 1, 1, 1) = I + O() ,
3
I = 2 log 2 (3) .
2
(4.20)
Until now, we have discussed on-shell propagator diagrams with a single

non-zero mass. Starting from two loops, there are also diagrams with loops
of a quark with a dierent mass m (say, c-quark loops in the b-quark selfenergy, or vice versa) (Fig. 4.6). Such diagrams can be reduced [17], using
integration by parts, to two trivial integrals (Figs. 4.7a,b) and two non-trivial
ones (Figs. 4.7c,d). These non-trivial integrals are

1
5
11
I0
2
24
2 2
=
m
+
2(1
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) +
Fig. 4.6. Two-loop on-shell propagator diagram with two masses
Fig. 4.7. Basis on-shell integrals with two masses
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)
The rst three-loop on-shell calculation, that of the electron anomalous

magnetic moment in QED, has been completed recently [28]. A systematic
algorithm for calculation of three-loop on-shell propagator diagrams in QCD,
using integration by parts, was constructed and implemented in [34].
The on-shell renormalization scheme is most convenient for calculation
of on-shell scattering amplitudes. The heavy-quark part of the QCD Lagrangian (3.1) can be rewritten as
0 Q0 + ,
0 (iD
/ 0 m)Q0 + m Q
L=Q
(4.23)
where m is the on-shell 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.
Fig. 4.8. Mass counterterm vertex
The heavy-quark self-energy can be decomposed as

p m) 2 (p2 ) .
(p) = m1 (p2 ) + (/
(4.24)
The bare heavy-quark propagator is then

S(p) =
(1 2
(p2 )) (/
p
1
.
m) + m m1 (p2 )
(4.25)
It has a pole at p2 = m2 if
m = m1 (m2 ) .
(4.26)
The mass counterterm m is determined by this equation, order by order in

os
perturbation theory; Zm
= 1 m/m = 1 1 (m2 ). Then, near the mass
68
shell, 1 (p2 )m/m = 1 (m2 ) (p2 m2 )+ , where 1 (p2 ) = d1 (p2 )/dp2 .

The bare quark propagator (4.25) becomes
S(p) =
1
/+m
p
+ ,
1 2 (m2 ) 2m2 1 (m2 ) p2 m2
(4.27)
where the dots mean terms which are non-singular at p2 m2 . We dene

os
the heavy-quark eld renormalization constant in the on-shell scheme ZQ
by the requirement that the renormalized quark propagator Sos (p) (which is
os
related to S(p) by S(p) = ZQ
Sos (p)) behaves as the free one (3.42) near the

1
os
mass shell. Therefore, ZQ = 1 2 (m2 ) 2m2 1 (m2 ) .
os
os
Now let us explicitly calculate Zm
and ZQ
at the one-loop order. It is
convenient [34] to introduce the function
1
Tr(/
v + 1)(mv(1 + t))
4m

= 1 (m2 ) + 2 (m2 ) + 2m2 1 (m2 ) t + O(t2 ) ;
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 gauge-independent. 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.
On-shell renormalization of QCD at two loops has been performed in [24,
11, 10] (see also [17] for the exact d-dimensional contributions of the loops
4.3 On-Shell Renormalization of HQET
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). Three-loop results have been obtained recently [34].
The on-shell mass is gauge-invariant to all orders [27]; the quark eld renoros
is gauge-invariant at two loops [11] but not at three [34].
malization ZQ
4.3 On-Shell Renormalization of HQET

Now we shall consider the on-shell renormalization of the HQET Lagrangian
(3.58). The HQET mass shell is = 0. All loop diagrams without massive
particles are no-scale and hence vanish. Only diagrams with loops of massive
quarks (of other avours) can contribute. Such diagrams rst appear at two
loops (Fig. 4.9).
Fig. 4.9. Two-loop on-shell HQET propagator diagram
The following method is used for calculation of such diagrams [12, 16]. For
two vectors a and b in d-dimensional 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 right-hand 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
(d/2 + n1 + n3 ) (d/2 + n2 + n3 ) (d/2 n3 ) (d + n1 + n2 + n3 )

.
(n1 ) (n2 ) (d/2) (d + n1 + n2 + 2n3 )
(4.31)
1
3
2
Fig. 4.10. Two-loop 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)
where (k) = (k 2 g k k )(k 2 ) is the massive-quark contribution to

the gluon self-energy, k 2 /(k v)2 = (d 2), and the sum is over all massive
avours. Here, from (4.31),
iCF g02

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 on-shell 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 on-shell 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
4.4 Scattering in an External Gluonic Field in QCD
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)
Q () from on-shell QCD results at two [11]

This fact was used for obtaining Z
and three [34] loops.
The renormalization constant (4.34) is not smooth at mi 0. This discontinuity comes from IR gluon momenta, where HQET does not dier from
QCD. Therefore, the QCD on-shell quark eld renormalization constant has
the same non-smooth behaviour at mi 0 [11, 17]:
os
os (mi )
Z
ZQ
(mi )
Q
=
os (m = 0)
os (mi = 0)
ZQ
i
Z
Q
= 1 CF TF
2(d 1)(d 6)
g04 m4
i
.
2 ()
(4)d
(d 2)(d 5)(d 7)
(4.36)
k0 (0) (see (4.5)). At the two-loop level, it

Now we are going to calculate
is given by the diagrams of Fig. 4.11 (where the second diagram also implies
the mirror symmetric one), with zero external residual momentum ( = 0,
p = 0). We obtain

dd k (k 2 )
k2
2

k0 (0) = iCF g0
d2+
= 0.
(4.37)
(2)d k 2
(k v)2
k0 (0)
Fig. 4.11. Two-loop diagrams for

Now we are going to perform matching for the scattering amplitudes of an
on-shell heavy quark in an external chromomagnetic eld in QCD and in
HQET, with linear accuracy in q/m, where q is the momentum transfer.
It is most convenient to calculate scattering amplitudes in an external
eld using the background eld method [1]. In the QCD Lagrangian (3.1),
72
we substitute A0 A0 + A0 , where A0 is the external eld, and choose the

A 2 /(2a0 ), where D
= ig0 Aa ta . The ghost
gauge-xing 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 gauge-xing term contains an A0 A20 contribution,
altering the three-gluon 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 four-gluon 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 three-gluon 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
Fig. 4.12. Vertices of the interaction with the background eld
The sum of one-particle-irreducible background-eld vertex diagrams not
including the external propagators is the proper vertex ig0 ta bf

(p, q), where

p is the incoming quark momentum, p = p + q is the outgoing quark momentum, and and a are the background-eld gluon polarization and colour

q
k+p
73
k+p+q
kq
k
k+p
b
k
a
Fig. 4.13. One-loop proper vertex
indices. The vertex function is bf

(p, q) = + bf (p, q), where bf (p, q)
contains one-loop (Fig. 4.13) and higher-loop corrections.
Now we are going to derive the Ward identity for bf (p, q)q . Backgroundeld vertices obey the simple identities shown in Fig. 4.14. Here a gluon line
with a black triangle at the end means the contraction of the vertex with the
incoming gluon momentum (as in Fig. 3.15); the colour structures are singled
out as the prefactors. Starting from each diagram for (p), we can obtain
a set of diagrams for bf by attaching the background-eld gluon to each
possible place. For example, starting from the one-loop diagram Fig. 3.5 for
(p), we can attach the external gluon either to the quark line or to the gluon
line, obtaining the diagrams in Fig. 4.13. The results of their contraction with
q are shown in Figs. 4.15a,b. The dierences in the square brackets are equal
to each other. The colour factors combine to give the colour factor of Fig. 3.5
times ta (Fig. 4.15c), owing to the denition of the colour factor of the threegluon vertex (Fig. 3.10). Therefore, we have the Ward identity
bf (p, q)q = (p) (p + q)
(p, q)q = S 1 (p + q) S 1 (p) .

or bf
(4.40)
This identity also holds at higher orders of perturbation theory. To verify

this, one has to derive identities similar to Fig. 4.14 for other backgroundeld vertices. For an innitesimal q, we obtain

= g0
a

= g0
b
Fig. 4.14. Ward identities for the background-eld vertices
74

= g0
a

= g0
c
Fig. 4.15. Ward identity for the background-eld 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 background-eld vertex are very
simple, and are exactly the same as in QED (or in the heavy-electron eective
theory, Sect. 3.5). Multiplying the ordinary three-gluon 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
ZA A. The renormalized matrix element is the proper vertex g0 ta bf

(p, q)
1/2
1/2
1/2
times Zq ZA : Zq ZA Z gta bf (p, q). At q = 0, the factor Zq converts S 1

in (4.41) into Sr1 , making it nite. Therefore, Z ZA = 1, just as in QED (or
in heavy-electron eective theory, Sect. 3.5). In other words, g0 A0 = g A.

The scattering amplitude of the on-shell quark in an external eld is
a
os
u
(p )bf
t u(p), where p2 = p2 = m2 , (/
p m)u(p) = 0, (/
p m)u(p ) = 0.
gZQ
It is UV nite, but may contain IR divergences. It can be expressed via two
scalar form factors. For comparison with HQET, it is most convenient to use
the Dirac and chromomagnetic form factors:

q, ] a
a

0 2 (p + p )
0
2 [/
+ Gm (q )
u
(p )bf t u(p) = u(p ) F1 (q )
t u(p) , (4.42)
2m
4m
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
and calculate the traces

1
Tr bf (/
v + 1)Ti (/
v + q//m + 1) .
4
Solving the linear system for Fi , we obtain
1
1
Tr bf (/
v + 1)
2(d 2) (1 q 2 /(4m2 )) 4

d 2 + q 2 /(4m2 )
q [/
q, ]
/q
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 ) =
There is no singularity in G0m (q 2 ) at q 2 0. We can expand the form factors

in q 2 by expanding
= 0 + 1
bf
q
+ ,
m
splitting q = (q v)v + q (where q v = q 2 /(2m)), and averaging over the

directions of q in the (d 1)-dimensional subspace orthogonal to v:

q2
q2
q2
q2

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 heavy-quark 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 q-dependent 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 one-loop 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)
Setting g0 e0 , CF 1, CA 0, we reproduce the electron magnetic

moment = 1 + /(2) + in QED. It is convergent. The heavy-quark
chromomagnetic moment in QCD (4.46) contains an IR divergence associated
with the colour factor CA . The two-loop correction to (4.46) was calculated
in [16], and the eect of another massive avour (say, the c-loop correction
to the b-quark chromomagnetic moment) was considered in [17].
4.5 Scattering in an External Gluonic Field in HQET

First, let us consider the leading (zeroth) order in 1/m. Let the sum of oneparticle-irreducible bare vertex diagrams in the background-eld method be
ig0 ta bf
(, q), where and = + q v are the residual energies of the
initial and the nal quark; bf

(, q) = v + bf (, q). There are two vectors
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)
On the mass shell = 0, = 0, the renormalized scattering amplitude with

linear accuracy in q is

d
()
os
os
Q
Q
Z
(1 + s (0, 0, 0))
uv (q)v uv (0) = Z
u
v (q)v uv (0)
1
d
=u
v (q)v uv (0) .
=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 one-loop 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 external-leg 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
(, q). If we write the vertex as = bf

+ ,
background-eld vertex bf
then is given by Fig. 4.16b, where only the 1/a0 terms in (4.39) are
Q , we obtain at
substituted for the three-gluon vertex. Multiplying by Z
one loop
r ,
Q Z
v + (nite) + = Z
Q is nite. Therefore, we need only the UV divergence of .
because Z
bf
This divergence does not depend on the external momenta, and therefore, we
may set them to zero. An IR cut-o is then necessary in order to avoid IR
1/ terms. We have
k
b
Fig. 4.16. One-loop HQET proper vertex
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 cut-o, 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 one-particle-irreducible bare vertex
diagrams in the background-eld method containing a single kinetic-energy
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 kinetic-energy 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
Similarly to Sect. 4.1, the variation of the diagrams for bf (

p, q) for v

v + v is equal to that of the diagrams for k (
p, q) for p p + p , if
v = (Ck /m)p . Therefore, the reparametrization invariance ensures that
ks (, , q 2 ) = s (, , q 2 ) ,
s (, , q 2 )
,
k1 (, , q 2 ) =
a (, , q 2 )
k3 (, , q 2 ) =
.
The Ward identity

k (
k (
k (
p, q)q =
p)
p )
(4.53)
(4.54)
ensures, owing to (4.5), that

d()
,
ks (, , q 2 ) + k1 (, , q 2 )( ) + k3 (, , q 2 )q 2 =
d
k0 ()
k0 ( ) .
k0 (, , q 2 )( ) + k2 (, , q 2 )q 2 =
(4.55)
The rst relation here is, owing to (4.53), just the derivative of (4.48) with
respect to .
On the mass shell = 0, = 0, the kinetic-energy correction to the
renormalized scattering amplitude with linear accuracy in q is
C0
Ck0 os
v (q)(p + p ) uv (0) .
ZQ (1 + s (0, 0, 0))
uv (q)(p + p ) uv (0) = k u
2m
2m
(4.56)
Let the sum of one-particle-irreducible bare vertex diagrams in the
background-eld method containing a single chromomagnetic vertex be
0
1 + v/
Cm
1 + v/
[/
q, ]
m (, , q 2 ) ,
(4.57)
4m 2
2
where m (, , q 2 ) = 1 + m (, , q 2 ). Reparametrization invariance does
not relate m to any vertex function of zeroth order in 1/m. We have no
better alternative than a direct calculation. In order to obtain the on-shell
scattering amplitude with linear accuracy in q, we need only the static-quark
chromomagnetic moment
0g = m (0, 0, 0). All loop diagrams for m vanish,
except those with loops of some massive quark. Such diagrams rst appear
at two loops. They can be calculated using the method of Sect. 4.3. The
0
os
renormalized matrix element
g = Z
Q g of the chromomagnetic operator
0
m is [16]

d2 9d + 16
g 4 m4
2
i
,
(4.58)
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
4.6 Chromomagnetic Interaction

Now we are ready to compare the on-shell scattering amplitudes in full QCD
p) are two-component in the v rest
and in HQET. The HQET spinors uv (
frame: v/uv (
p) = 0. The corresponding QCD spinors u(mv + p) are related to
them by the FoldyWouthuysen transformation (see, e.g., [7])

/
p
+ O(
p2 /m2 ) uv (
p) .
(4.59)
u(mv + p) = 1 +
2m
A quick check:

/p
p) = 0 ,
uv (
(m/
v+p
/ m)u(mv + p) = m v/ 1 + (/
v + 1)
2m
because we may set p v = 0 in the 1/m term;
u
(mv + p) u(mv + p) =
(mv + p)
p)uv (
p) .
uv (
m
Expanding the QCD scattering amplitude up to linear terms in q/m and

re-expressing it via HQET spinors, we obtain

q
[/
q, ] a
u
v (q) v +
+ g
t uv (0) .
(4.60)
2m
4m
The HQET scattering amplitude with 1/m accuracy is

[/
q, ] a
q
1
u
v (q) v + Ck ()Zk1 ()
+ Cm ()Zm
()
g
t uv (0) . (4.61)
2m
4m
Both scattering amplitudes (4.60) and (4.61) are renormalized and hence UV
nite. Both may contain IR divergences. By construction, HQET is identical
to QCD in the IR region, so that these IR divergences are the same. For
example, if there are no other massive avours in the theory, all loop corrections to
0g vanish because they contain no scale. These zero integrals contain
UV and IR divergences which cancel. The UV divergences are removed by
1
m
Z
(), and the IR ones match those in g .
k = 1 and Ck () =
Comparing the coecients of q /m, we again see that Z
1 (4.9). Comparing the coecients of [/q , ]/m, we obtain
1 ()Cm () = g .
Z
m
g
(4.62)
In the one-loop approximation, re-expressing (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)
m that makes Cm nite is

The minimal (3.3) renormalization constant Z
m = 1 CA s + ,
Z
4
(4.64)
and the chromomagnetic interaction constant is

Cm () = 1 + 2 (CA L + CF + CA )
s (m)
+
4
(4.65)
Therefore, the anomalous dimension of the chromomagnetic operator and

Cm (m) are [19]
m = 2CA
s
+ ,
4
Cm (m) = 1 + 2(CF + CA )
s (m)
+
4
(4.66)
The two-loop anomalous dimension was calculated in [2] within HQET,

and (a week later) in [16] by QCD/HQET matching:
m = 2CA
2
4
s
s
+ CA (17CA 13TFnf )
+
4 9
4
(4.67)
The anomalous dimension vanishes in QED, where CA = 0. It is most convenient to calculate

2
s (m)
s (m)
+ C2
Cm (m) = 1 + C1
+ ,
4
4
because it contains no large logarithms. Then we can use the renormalizationgroup equation to nd Cm () at m. The two-loop term in Cm (m) was
found in [16].
The scattering amplitude (4.61) does not depend on the arbitrary renormalization scale ; the matrix element
0g of the bare chromomagnetic operator is also -independent. Therefore,
1 ()Cm () = const .
Z
m
Dierentiating this equality with respect to log , we obtain the renormalization-group equation
dCm ()
=
m (s ())Cm () .
d log
(4.68)
If L = log(m/) is not very large, it is reasonable to nd the solution as

a series in s (m). Re-expressing s () via s (m) as
82

s (m)
s (m)
s ()
=
+
1 + 20 L
4
4
4
in
m (s ()) = 0
s ()
+ 1
4
s ()
4
2
+ ,
we obtain the equation

2
s (m)
dCm ()
s (m)
s (m)
+ 0
+ Cm () ,
1 + 20 L
+ 1
dL
4
4
4
which can be solved order by order in s (m):
s (m)
4

(m) 2
0
s
+ 0
0 L2 (1 + C1 0 ) L + C2
+
2
4
(4.69)
Cm () = 1 + (0 L + C1 )
If [s /(4)]L 1, we have to solve the renormalization-group 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)
Subtracting and adding

m0 /(20 ) in the integrand, we can rewrite the solution as

m0 /(20 )
s ()
Km (s ())Km (s (m)) . (4.71)
Cm () = Cm (m)
s (m)
Here we have introduced a useful notation: for any anomalous dimension
(s ), we dene

K (s ) = exp
(s )
0
2(s ) 20
ds
0
=1+
s
20
1
1
0
0
s
+
4
(4.72)
83
This function has the obvious properties

K (s ) = K1 (s ) ,
K0 (s ) = 1 ,
K1 +2 (s ) = K1 (s )K2 (s ) .
The solution (4.71) can also be written as

Cm () = Cm (m)

=
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)
We can write a similar equation for mD mD . The quantities

2G in the bquark and the c-quark HQETs dier by an amount of order [s (mc )/]2 due
to decoupling of c-quark loops (Sect. 4.7). Multiplying (4.76) by mB + mB =
2mb [1 + O(1/mb )] and dividing by a similar D-meson equation, we obtain [2]

/(20 )
s (mb ) m0
m2B m2B
=
m2D m2D
s (mc )

s 2
m1
s (mc ) s (mb )
m0

1
+O
1 C1 +
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 next-to-leading 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.
4.7 Decoupling of Heavy-Quark Loops

Let us consider QCD with nl light avours and a single heavy one, say, c. Processes involving only light elds with momenta pi mc can be described by
an eective eld theory QCD with nl avours. There are 1/mnc -suppressed
local operators in the Lagrangian, which are the remnants of heavy-quark
loops shrunk to a point. This low-energy eective theory is constructed to
reproduce the S-matrix elements of full QCD expanded to some order in
pi /mc . Operators of full QCD can be expanded in operators of the eective
theory; higher-dimensional operators are divided by the appropriate powers
of mc . The coecients of these 1/mc expansions are xed by matching
equating on-shell matrix elements, up to some order in pi /mc . Matching full
QCD with the low-energy QCD is called decoupling; it is clearly presented
in [15], where references to earlier papers can be found. All quantities in the
low-energy theory will be distinguished from those in full QCD by primes.
85
In particular, the renormalized light elds of full QCD are related to those
of the low-energy theory by
qi () = q1/2 (, )qi ( ) ,
1/2
A () = A (, )A ( ) ,
c() = c1/2 (, )c ( ) ,
(4.78)
up to 1/mc -suppressed terms. The coupling constant and gauge-xing 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 low-energy 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)
Multiplying by this coecient also converts a0 into a0 in the longitudinal

part of the propagator.
In the full theory, only diagrams with c-quark loops (Fig. 4.17) contribute.
It is convenient to extract (0) using

1
(p)
.
(0) =
2d(d 1) p p
p=0
In the one-loop approximation (Fig. 4.17a), (0) is a combination of one-loop
vacuum integrals (2.16); we obtain
4 g 2 m2
(0) = TF 0 cd/2 () +
3 (4)
(4.83)
86
Fig. 4.17. Massive loops in gluon self-energy
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 two-loop diagrams (Figs. 4.17b,c).
They reduce to combinations of two-loop massive vacuum integrals (4.31),
and are proportional to 2 (). Using also the two-loop 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 light-quark eld (see (3.46)),
q0 =
Zqos
1
.
=
Zqos
1 V (0)
(4.85)
The only two-loop diagram contributing is Fig. 4.18. Expanding the quark
self-energy (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
Fig. 4.18. Two-loop on-shell massless-quark self-energy
The ratio of the renormalization constants for the quark eld is

2
s (mc )
1
dq0
Zq (mc )
1
=1+
a q1
+
q0 0 + A0
Zq (mc )
4
2
da
4

2
1 s (mc )
+ ,
= 1 + CF TF
4
where 0 = (4/3)TF , A0 = (8/3)TF, and q1 = 4CF TF are the
single-avour contributions. Finally, we obtain [15]

2
s (mc )
5
q (mc , mc ) = 1 CF TF
+
(4.87)
6
4
If we consider b-quark HQET instead of QCD, nothing changes. When all
characteristic (residual) momenta become much less than mc , c-quark 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)
The ratio of renormalization constants is, from (3.75),

2
Q (mc )
Z
s (mc )
4
,
= 1 + 2CF TF 1
(mc )
3
4
Z
Q
and we arrive at [25]
52
Q (mc , mc ) = 1 + CF TF
9
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 heavy-quarkgluon
vertex. Let g0 ta (, q) be the sum of bare one-particle-irreducible heavyquarkheavy-quarkgluon vertex diagrams, not including the external propagators. For this vertex,
1
1/2
0
0
A
g0 ,
g0 = Q
88
Fig. 4.19. Two-loop on-shell HQET quarkgluon vertex (the diagram b implies
also the mirror-symmetric one; the diagram c has two orientations of the quark
loop)
because after multiplication by the three propagators it becomes the Green

0 , and A0 . It is most convenient to set all the external
0 , Q
function of Q
momenta to zero. In the low-energy theory, = v . In full HQET, only diagrams with c-quark loops contribute, and (0, 0) = v . They rst appear
at two loops (Fig. 4.19). We have
0 =
1
,
2
0 0
Q
1
Z 0
Z Z 1
= A
.

Z
Z ZA A 0 2
Q
It is convenient to write (, q) = bf

+ , where bf
(, q) = v +
bf (, q) is the background-eld 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 c-quark loop contributes to
0

and bf (0, 0)
the heavy-quark self-energy ().
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
three-gluon 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)
Taking into account (4.84) and

2

s (mc )
Z (mc ) ZA
(mc )
5
= 1 + CA TF 1
,
Z (mc ) ZA (mc )
6
4
we arrive at the well-known result [29]
(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 c-quark loops in the b-quark
chromomagnetic-interaction constant. Both HQET with c-loops and the effective low-energy theory must give identical on-shell 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)
g is given by (4.58), with just the c-quark contribution.

Here
g = 1, and
m is
The ratio of the renormalization constants Z

2
m (mc )
Z
1
s (mc )
= 1 + (0 0 1 )
+
(mc )
4
4
Z
m

2
s (mc )
2 13
= 1 CA TF

+ ,
(4.93)
3
8
4
where 0 = (4/3)TF and 1 = (4/9) 13CA TF are the single-avour
contributions. Re-expressing (4.58) via s (mc ) and expanding in , we see
that singular terms cancel, and

2
s (mc )
71

+ Cm
(mc ) .
(4.94)
Cm (mc ) = 1 + CA TF
27
4
Therefore, when crossing = mc , we have to adjust Cm () according
to (4.94), and at < mc the renormalization-group running is driven by
the function and the anomalous dimension for nf = 3.
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. Perret-Gallix (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 CALT-68-1663 (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.
5.1 Bilinear Quark Currents in QCD

Now we are going to consider the bilinear quark currents
jn0 = q0 q0 ,
= [1 n ]
(5.1)
in QCD. Here is the antisymmetrized product of n -matrices, and q0 and

q0 are bare quark elds of dierent avours. Currents with q = q have some
peculiarities, because the quark line emerging from the current vertex can
return to the same current again. We shall not discuss such currents here. All
properties that we shall derive for (5.1) are also valid for avour-non-singlet
currents qi0 ij qj0 , where is a avour matrix with Tr = 0, and all qi have
equal masses.
The renormalized currents are related to the bare ones by
1
()jn0 ,
jn () = Zjn
(5.2)
where the Zjn are minimal (3.3) renormalization constants. The dependence
of jn () is determined by the renormalization-group equation

d
+ jn (s ()) jn () = 0 ,
(5.3)
d log
92
where
jn (s ) =
d log Zjn
d log
(5.4)
is the anomalous dimension.

Let the sum of one-particle-irreducible bare diagrams with a current vertex , an incoming quark q with momentum p, and an outgoing quark q
with momentum p , not including the external quark propagators, be the
proper vertex n (p, p ) = + n (p, p ) (Fig. 5.1; here in the right-hand
side is the Dirac matrix). When the vertex is expressed via the renormalized
quantities s (), a(), it should become Z n nr (p, p ), where the renormalized vertex nr (p, p ) is nite in the limit 0. When the proper vertex of
1
n (p, p ) is multiplied by the two external-leg
the renormalized current Zjn
1/2
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 cut-o is then necessary in
order to avoid IR 1/ terms.
p
p
a
Fig. 5.1. Proper vertex of a bilinear quark QCD current
In the one-loop approximation (Fig. 5.1b),

k i/
k
k k
dd k
i/
i
n (0, 0) = CF
ig0 2 2 ig0 2 g (1 a0 ) 2
(2)d
k k
k
k

d
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)
5.1 Bilinear Quark Currents in QCD
Z n = 1 + CF

s
(n 2)2 1 + a .
4
93
(5.7)
The gauge dependence is cancelled by Zq (3.48), and we arrive at

jn = 2(n 1)(n 3)CF
s
+
4
(5.8)
The two-loop 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 by-product 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 three-loop 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) ,
1 (p, p + q)q = S 1 (p + q) S 1 (p) .
(5.10)
In particular, at q 0
1 (p, p) =
(p)
p
or 1 (p, p) =
S 1 (p)
.
p
(5.11)
Multiplying the last equation by Zq transforms S 1 into Sr1 , and hence

makes the left-hand side nite. Therefore, Zj1 = Zq Z 1 = 1.
94
The renormalization of the scalar current q q is closely related to that of

the quark mass m. The MS renormalized mass
1
m() = Zm
()m0
(5.12)
(with a minimal renormalization constant Zm ) is dened in such a way that

the bare quark propagator S(p), when expressed via the renormalized quantities s (), a(), m(), is equal to Zq Sr (p), where the renormalized propagator
Sr (p) is nite in the limit 0 (Sect. 3.1). In order to nd Zm , it is sucient
to consider |p2 | m2 and retain only terms linear in m. We can calculate
[(p)/m0 ]m0 =0 using the identity

S0 (p)
2
= [S0 (p)] .
(5.13)
m0 m0 =0
Starting from each diagram for (p), we obtain a set of diagrams for
[(p)/m0 ]m0 =0 by dierentiating each quark propagator of the considered
avour in turn. After dierentiating a propagator in a quark loop, we obtain
a trace with an odd number of -matrices, which is zero. Therefore, only the
propagators along the open quark line going through the diagram need to
be dierentiated. The resulting diagrams are exactly those for 0 (p, p). We
arrive at

(5.14)
1 + S (p2 )m0 =0 = 0 (p, p) .
As explained after (3.46), the renormalization constant Zm is dened by the
condition that Zq Zm (1 + S ) is nite. Therefore,
1
Zm = Zj0
.
(5.15)
In other words, m() [

q q] = m0 q0 q0 is not renormalized. The anomalous
dimension of the mass is

s 2
s
97
20
+ CF 3CF + CA TF nf
+ (5.16)
m = j0 = 6CF
4
3
3
4
The four-loop result has been obtained recently [5, 18].
1
()m0 is related to the on-shell mass m =
The MS mass m() = Zm
os 1
[Zm ] m0 by
m() =
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
5.2 Axial Anomaly
95
It is most natural to use = m, because the relation between m(m) and

m contains no large logarithms. The two-loop relation between the MS mass
and the pole mass was derived in [11], and the three-loop relation in [15].
Note that the on-shell mass has a meaning only for heavy quarks, because
for light quarks perturbative calculations at p2 = m2 are not possible; the
MS mass has a meaning for both heavy and light quarks.
If we need m() for widely separated from m, we need to solve the
renormalization-group equation

d
(5.19)
+ m (s ()) m() = 0 .
d log
Its solution is (see (4.72))

m() = m
s ()
4
m0 /(20 )
Km (s ()) .
(5.20)
Currents (5.1) with n > 4 do not exist in four-dimensional 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 non-evanescent 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.
5.2 Axial Anomaly

It is not possible to dene a 5 satisfying
5 + 5 = 0
(5.21)
in d-dimensional space. Let us consider the following chain of equalities:

Tr 5 = d Tr 5 = Tr 5 = Tr 5 = d Tr 5
d Tr 5 = 0
(the anticommutativity of 5 was used in the second step, and the trace
cyclicity in the third one). We have learned that if d = 0 then Tr 5 = 0. We
assume that d = 0 and continue:
Tr 5 = d Tr 5 = Tr 5 = (d 4) Tr 5
(d 2) Tr 5 = 0 .
We have learned that if d = 2 then Tr 5 = 0. We assume that d = 2
and continue:
96
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 four-dimensional 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)
where the tensor lives in the four-dimensional subspace. Here lives

in the physical 4-dimensional subspace, and in the non-physical (d 4)dimensional one. This prescription spoils d-dimensional Lorentz invariance.
The inconsistency of the naively anticommuting 5 (5.21) is strikingly
demonstrated by the anomaly in the avour-singlet axial current j = q 5 q
(see, e.g., [7]). Let us consider its matrix element M (p, p ) mn between an
incoming gluon with momentum p and an outgoing gluon with momentum
p (Fig. 5.2). The Ward identities yield
M (p, p )p = 0 ,
M (p, p )p = 0 .
Using the equality

q 5 S0 (p) = 5 S0 (p) + S0 (p )5 + 2mS0 (p )5 S0 (p) ,
S0 (p )/
k + p
k+p
p
k
a
Fig. 5.2. Gluon matrix element of the axial current
q = p p ,
(5.23)
5.2 Axial Anomaly
97
one can, it seems, prove that for a massless quark,

M (p, p )q = 0 .
(5.24)
But this axial Ward identity is wrong!
Let us dene jHV

= (1/2)
q 5HV 5HV q, and calculate (5.24) for
the diagrams of Fig. 5.2 carefully:

dd k

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
physical-subspace matrices /p, p

/ , , . Averaging over orientations of the
physical subspace, we may make the replacement
k2
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
5.3 The t HooftVeltman 5 and the Anticommuting 5

The quark line cannot be closed in matrix elements of avour-nonsinglet currents. Therefore, traces with a single 5 , which can lead to an anomaly, never
appear. There is a general belief that one may use a naively anticommutating
5AC satisfying (5.21), without encountering contradictions. The pseudoscalar
1
1
currents jAC () = ZP,AC
()
q0 5AC q and jHV () = ZP,HV
()
q0 5HV q are related to each other by a nite renormalization:
jAC () = ZP (s ())jHV () ,
2
s
s
+ zP2
ZP (s ) = 1 + zP1
+
4
4
(5.26)
Similarly, the axial currents are related by
() = ZA (s ())jHV
() .
jAC
(5.27)
This is clearly discussed in [14], where references to earlier papers can be

found.
If the operator equalities (5.26), (5.27) are valid, they must hold for all
matrix elements. We can only calculate a few of them, and check the supposed
universality. The vertex functions (p, p) and (p, 0) reduce to propagator
integrals. For massless quarks, they were discussed in Sect. 2.2 at one loop
and in Sect. 2.4 at two loops. We shall discuss these two examples here; more
matrix elements will be discussed in Sects. 5.6, and 7.27.5.
Initially, we make no assumptions about the properties of the matrix ,
and note that the one-loop diagram of Fig. 5.1b for (p, p) or (p, 0) can be
written as
= 1 2 1 2 I 1 2 ;1 2 ,
where I 1 2 ;1 2 is an integral with respect to the loop momentum k. After
integration, it can contain only g and p . Anticommuting the -matrices,
we can rewrite the diagram as

= x1 + x2 /
p /
p/p2 + x3 =
xi Li Ri ,
i
Li Ri = 1 1 , /
p /p/p2 , .
(5.28)
The coecients xi can be found by calculating the traces of -matrices to the

left of and to the right of it, separately. For any term L R, let us dene

R,
L R = 1 Tr LL
1 Tr RR
.
L
4
4
Then
(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:
/
Now we assume that the matrix either commutes or anticommutes with

p = /
/
p,
= 1 .
(5.31)
This means that the components of parallel and orthogonal to p have to

be treated separately. We also assume that is proportional to . It is
convenient to modify the denition of h(d) slightly, and to use

d
n
(5.32)
= 2h(d) , h(d) = (1)n
2
instead of (5.5). Under these conditions, the correction (5.28) is proportional
to :

= x1 + x2 (2h) + x3 (2h)2 .
Substituting the solution (5.30) for xi , we obtain
= P, ,

1
(d 2)(3d 2 4h2 )1 1
P =
2(d 1)(d 2)

p /p /p2 + (d + 2 4h + 4h2 ) .
+ 4h(d 2h) /
Now we can apply the projector P to the integrand of the one-loop 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
where G1 is dened in (2.41). Both matrix elements are IR nite; their UV

divergences are given by (5.7).
A strong check of these results is provided by the Ward identities.
From (5.11), we obtain
1 (p, p) =

dV (p2 )

pV (p2 ) = V (p2 ) 2/pp
/
.
p
dp2
Therefore, for the longitudinal component of the vector current ( = p

/,
= +1, h = 1 d/2), we have
a (p2 ) = V (p2 ) 2(p2 )
dV (p2 )
= (d 3)V (p2 ) ,
d(p2 )
and for its transverse component ( =

= /pp /p2 , = 1, h =
2
2
d/2 1), a (p ) = V (p ). Substituting V (p2 ) (3.47), we obtain two
checks of (5.33). From (5.10), we obtain p 1 (p, 0) = /pV (p2 ), and hence,
for the longitudinal component of the vector current (h = 1 d/2), b (p2 ) =
V (p2 ). This provides a check of (5.34).
Now we can resume our investigation of the relation between matrix elements of currents containing 5HV and 5AC . Multiplying by 5AC does
not change h (5.32), and hence cannot change the matrix element. On the
other hand, multiplying the antisymmetrized product of n -matrices by
5HV means n 4 n and . This is equivalent to the substitution
d 8 d, or in h.
The matrix elements of the bare pseudoscalar currents j0AC (h = 2 )
and j0HV (h = 2 + ) are related by (5.26)
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)
where ZA,AC () = Zj1 () = 1 and ZA,HV () = Zj4 () coincide at one loop.

The 1/ term in (5.33) and (5.34) does not change when h h, and
ZA (s ) = 1 4CF
s
+
4
101
(5.38)
The nite renormalization constants ZP,A (s ) can also be obtained from

the anomalous dimensions of the currents. Dierentiating (5.26) and (5.27),
we have
d log ZP (s )
j0 (s ) j4 (s )
,
=
d log s
2(s )
d log ZA (s )
j1 (s ) j3 (s )
,
=
d log s
2(s )
where j1 = 0 .
(5.39)
Therefore,
ZP (s ) = Kj0 j4 (s ) ,
ZA (s ) = Kj1 j3 (s )
(5.40)
(see (4.72)). The one-loop results (5.36) and (5.38) can be reproduced using
the two-loop 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 two-loop 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 two-loop matrix elements (p, p) (or (p, 0)),
which can be calculated by the methods of Sect. 2.4, or from the three-loop
anomalous dimensions. The three-loop 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
5.4 HeavyLight Current in HQET

Now we are going to consider the current
j() = Zj1 ()

j0 ,
0

j0 = q0 Q
(5.43)
The proper vertex (, p) =

in HQET with a scalar heavy quark Q.

/. When the vertex is expressed
1 + (, p) (Fig. 5.3) can contain 0 and p
r (, p),
via the renormalized quantities s (), a(), it should become Z
where the renormalized vertex r (, p) is nite at 0. When the proper
vertex of the renormalized current Z1 (, p) is multiplied by the externalj
1/2 1/2
Zq ZQ ,
leg renormalization factors

it should give a nite matrix element.
1/2 Z
. The UV divergences of (,
p) do not depend
j = Zq1/2 Z
Therefore, Z
Q
on the masses and external momenta, and we may set these masses and
momenta to zero. An IR cut-o is then necessary to avoid IR 1/ terms.
In the one-loop approximation (Fig. 5.3b),

dd k /k v g (1 a0 )k k /k 2
2

(0, 0) = iCF g0
(2)d
(k 2 )2 k0

d
d k 0 /k (1 a0 )k0
= iCF g02
.
(2)d
(k 2 )2 k0
Fig. 5.3. Proper vertex of the heavylight HQET current
5.5 Decoupling for QCD and HQET Currents
103
Here /k = k0 0 k ; the term containing k yields 0 after integration:

1
dd k
2

(0, 0) = iCF g0 a0
.
(2)d (k 2 )2
This integral contains no HQET denominator; its UV divergence is given
by (4.51). We obtain
= 1 + CF a s .
Z
4
1/2
1/2 (3.65), and we
The gauge dependence is cancelled by Zq (3.48) and Z
Q
arrive at
s
.
(5.44)
j = 3CF
4
Two-loop 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 self-energy diagrams (see Fig. 3.13; the blob in Fig. 5.3c means
the one-loop gluon self-energy 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 two-loop 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)
This anomalous dimension was calculated independently in [13, 9], by more

cumbersome methods. The three-loop term has been calculated recently [6].
0 with a spin-(1/2) heavy quark is
The vertex of the current
j0 = q0 Q

[(1 + 0 )/2] (, p). Its anomalous dimension does not depend on the Dirac
matrix . Therefore, the factors relating the currents containing 5AC and
5HV are equal to 1. This is also evident from the fact that any renormalized

() is the
where M
matrix element of
j() is equal to [(1 + 0 )/2] M(),
corresponding renormalized matrix element for a spin-0 heavy quark; it is
nite, and hence we may take the limit 0, where 5HV and 5AC are the
same thing.

Now we are going to discuss the relation between light-quark currents in
QCD with a heavy avour (say, c) and in the eective low-energy theory
104
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 on-shell 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 right-hand
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 )) ,
K is dened by (4.72), and K involves the nl -avour -function (s ).

1
()n (p, p ).
The on-shell matrix element of jn () is Mn (p, p , ) = Zqos Zjn

It should be equal to jn (, )Mn (p, p , ) + O((p, p )/mc ), where
1
Mn (p, p , ) = Zqos Zjn
( )n (p, p ). Both matrix elements are UV-nite;
their IR divergences coincide, because both theories are identical in the IR
region. Therefore,

( ) n (p, p )
Zqos Zjn
p, p
+
O
jn (, ) = os
.
(5.48)
Zq Zjn () n (p, p )
mc
Any on-shell momenta p, p can be used; it is easiest to set p = p = 0, thus
excluding power-suppressed terms in (5.48).
All loop diagrams for n (0, 0) contain no scale, and n (0, 0) = , the
antisymmetrized product of n -matrices. Only diagrams with c-quark loops
contribute to n (0, 0); the only relevant two-loop diagram is shown in Fig. 5.4.
It is equal to

dd k /k /k (k 2 g k k )(k 2 )
n (0, 0) = iCF g02
(2)d
(k 2 )4

dd k ( /k /k k 2 )(k 2 )
= iCF g02
,
(2)d
(k 2 )3
where (k 2 ) is the c-quark loop contribution to the gluon self-energy. Averaging over the directions of k and using the denition (5.5) of h, we obtain
105
Fig. 5.4. Two-loop on-shell matrix element of a QCD bilinear quark current
n (0, 0) =
iCF g02
and using (4.33),

(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)
From (5.9), we derive the ratio

Zjn (mc )
1
(m ) = 1 + 4 (jn0 0 jn1 )
Zjn
c
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 single-avour 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 low-energy eective theory. The same holds for non-diagonal by avour
symmetry. We can also see this explicitly. Multiplying the Ward identity

(p)
= (1 V (0))
1 (0, 0) =
p p=0
1
by Zqos = [1 V (0)] , we obtain just = . Taking account of the fact

that Zj1 = 1, (5.48) yields j1 = 1.
106
The decoupling for the scalar current q q is closely related to that for
the MS light-quark mass. The quark propagators (3.46) in full QCD and the
low-energy 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
This equation should hold for all m mc , p mc . It is easiest to set m = 0,

1
and use (5.14). Then, setting p = 0 and using Zqos = [1 V (0)] , we obtain
0
m
=
Zqos 0 (0, 0)
.
Zqos 0 (0, 0)
Recalling (5.48) and (5.15), we nally arrive at

1
m (, ) = j0
(, ) .
(5.51)
In other words, m()[

q q] = m ( )[
q q] does not vary with and , and
does not change when one goes from the full theory to the low-energy one.
The MS mass decoupling constant is
m (mc , mc ) = 1
89
CF TF
18
s (mc )
4
2
+
(5.52)
This means that at > mc , the running of the light-quark 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)
5.6 QCD/HQET Matching for HeavyLight Currents
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 b-quark 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 c-quark
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 two-loop
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)
From (5.45), we derive the ratio

2
j (mc )
Z
5
s (mc )
= 1 + CF TF 1
+
(mc )
6
4
Z
j
Using also (4.86) and (4.88), we arrive at [12]

2
s (mc )
89

+
j (mc , mc ) = 1 + CF TF
36
4
(5.56)
Fig. 5.5. Two-loop on-shell matrix element of the HQET heavylight current

In physical applications, we are interested in matrix elements of QCD operators. However, we want to utilize the simplications introduced by HQET.
108
Therefore, we expand the QCD operators in 1/m; the coecients of these

expansions are HQET operators with the appropriate quantum numbers and
dimensionalities:

1
1

j() = C (, )
j( ) +
B (, )i ( ) + O
.
(5.57)
2m i i
m2
The meaning of the operator expansion (5.57) is that matrix elements of j(),
in situations amenable to HQET treatment, after expansion to a given order
in 1/m, coincide with the corresponding matrix elements of the right-hand
side of this equation. Here we shall discuss the leading order; 1/m terms are
discussed in Chap. 6, and 1/m2 terms in [1].
For example, let us consider the decay of a heavy quark into a light quark
with energy m via a heavylight weak current. The matrix element in
QCD depends on two widely separated large scales m and the renormalization scale (if the current has a non-zero anomalous dimension). For
no choice of can we get rid of large logarithmic corrections. When we go
to HQET, all m-dependence is isolated in the matching coecient of the
heavylight current C . The HQET matrix element knows nothing about m,
and depends only on and , where the -dependence is determined by the
anomalous dimension of the HQET heavylight current. If is chosen to be
of the order of , then there are no large logarithmic corrections.
It is natural to perform the matching at = = m, where the matching
coecients C contain no large logarithms. Currents at arbitrary normalization scales are related by
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 b-quark 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 on-shell 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 b-quark 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)
Any on-shell momenta p, p can be used; it is easiest to set p = p = 0, thus

excluding power-suppressed terms in (5.59).
The ratio q0 = Zqos /Zqos (4.85) was considered in Sect. 4.7. At one loop,
it is just 1; at two loops, it contains the b-loop contribution (4.86) (with
m = mb instead of mc ). The on-shell renormalization constant of the HQET
os was considered in Sect. 4.3. It is just 1 if all quarks lighter
quark eld Z
Q
than b are considered massless; the two-loop mc correction is given by (4.34).
os
The on-shell renormalization constant of a massive-quark eld in QCD, ZQ
,
was considered in Sect. 4.2. The one-loop result is given by (4.29); the twoloop result has been obtained in [2]. The MS renormalization constant of
j () was considered in Sect. 5.4. It is the
the HQET heavylight current Z
same for all -matrix structures ; the two-loop result is given by (5.45).
The MS renormalization constant of QCD quark currents was considered in
Sect. 5.1. For the antisymmetrized product of n -matrices, the two-loop
result is given by (5.9). The HQET proper vertex (0, 0) = 1 if all light
quarks are considered massless. At two loops, only the diagram of Fig. 5.5
with a c-quark loop can contribute. However, it vanishes; see (5.55).
We have only to calculate (mv, 0) (Fig. 5.6). The tree diagram (Fig. 5.6a)
just gives
0 = uq uQ ,
where is the Dirac matrix in the current. Initially, we make no assumptions
about its properties. The one-loop diagram (Fig. 5.6b) can be written as a
sum of terms of the form
u
q 1 . . . l 1 . . . r uQ I 1 ...l ;1 ...r ,
(5.60)
where I is some integral over the loop momentum, l is even, and l + r 4.

After the integration, I 1 ...l ;1 ...r can contain only g and v . The resulting
contractions of pairs of -matrices on the left, and of pairs on the right,
merely produce additional terms of the form (5.60), with smaller values of
l + r. Before performing the remaining contractions, one may anticommute
-matrices, so as to arrange things such that /v occurs only on the extreme
left or on the extreme right, with the contracted indices in between occurring
in opposite orders on the left and right of . The additional terms arising
from the anticommutations have fewer -matrices, with l remaining even.
Repeating this procedure for all values of l + r, from 4 down to 0, we may
cast the diagram of Fig. 5.6b in the form
v ) + v/ (x3 + x4 /v ) + x5 ] uQ
= uq [ (x1 + x2 /

= uq
xi Li Ri uQ ,
i
110
mv
0
a
Fig. 5.6. Proper vertex of the heavylight QCD current
where
x1 = x1 + x2 ,
x2 = x3 + x4 ,
x3 = x5 ,
Li Ri = 1 1 , /
v ,
(because /v uQ = uQ ). The coecients xi can be found by calculating the

double traces (5.29)

i R
i, ,
yi = L
i R
i = 1 (/
L
v + 1) , /v (/
v + 1) , (/
v + 1) .
Solving the linear system, we obtain (5.30).
Now we assume, similarly to (5.31) and (5.32),
v = /
/
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 one-loop 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 one-loop result [4]
= CF
g02 m2
(1 h)(d 2 + 2h)
.
()
(d 2)(d 3)
(4)d/2
(5.64)
This on-shell vertex is gauge-invariant. The two-loop contribution (Figs. 5.6c

k) has been calculated in [4]. It is a sum of terms of the form (5.60) with
l + r 8. Repeating the above discussion, we see that the integrand of
each diagram can be reduced to a scalar expression quartic in h. Only the
diagrams of Figs. 5.6e and 5.6h contain enough -matrices on each side of
to generate h3 and h4 . Scalar integrals are calculated by the method discussed
in Sect. 4.2. The diagram of Fig. 5.6i and the heavy-quark loop contribution
in Fig. 5.6c are of type N (4.19); they contain the non-trivial basis integral
of Fig. 4.5e. The other diagrams are of type M (4.16), and can be expressed
via -functions (Figs. 4.5c,d). If there is an additional massive avour (c in
b-quark HQET), its loop in Fig. 5.6c yields a contribution which depends on
mc /m. This contribution is quadratic in h. It has been taken into account
in [4].
In Sect. 5.3, we discussed the relation between quark currents containing
5AC and 5HV . The operator relations (5.26) and (5.27) were supposed to
be universal. We have checked them for massless quarks by calculating two
o-shell matrix elements. Now we are in a position to check them in the a
112
situation: one of the quarks is massive, and the momenta are on-shell (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 two-loop results presented in [4]
conrm (5.36) and (5.38).
At the time when the paper [4] was written, the two-loop 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 on-shell calculation.
At last, we can combine all pieces of (5.59). With two-loop 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 one-loop 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)
is the dierence between the contribution of diagrams with a quark loop

with mass mi and the corresponding massless contribution ((0) = 0). This
dierence is nite at = 0, and hence depends only on h|=0 = (n 2). It
can be expressed via dilogarithms:
1 (r) = (1 + r)L+ (r) (1 r)L (r) + log2 r + 2 ,
2 (r) = r(1 r2 )L+ (r) + r(1 r2 )L (r) + 2r2 (log r + 1) ,

3
3 (r) = r3 (1 + r)L+ (r) + r3 (1 r)L (r) r2 log r +
,
2
(5.68)
where L (r) are dened in (4.22).

There are eight distinct kinds of in four-dimensional spacetime. In the
v rest frame, they are antisymmetrized products of zero to three matrices i ,
and the same products times 0 . For each current, we obtain the matching
coecient C (m) by choosing the appropriate values of n and (dened
in (5.62)) in (5.66) and (5.68), see Table 5.1. Typical matrices are presented
in the rst column; the results are, of course, the same for other spatial
components.
Multiplying by 5AC does not change h (5.61), and hence cannot change
C (). On the other hand, multiplying by 5HV leads to n 4 n and
, and thus aects C (). The renormalized HQET currents containing
5HV and 5AC coincide with each other (Sect. 5.4). The renormalized QCD
currents are related to each other by (5.26) and (5.27). Therefore,
ZP (s ()) =
ZA (s ()) =
=
ZT (s ()) =
=
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)
Two matching coecients are equal: C 0 1 (, ) = C 2 3 (, ); indeed, for

n = 2 the results (5.66) and (5.68) do not depend on . From (5.66), we can
114
Table 5.1. Matching coecients C (m)
+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
5.7 Meson Matrix Elements
115
reconstruct the two-loop results for ZP (5.41) and ZA (5.42) (in two ways).
The nite-mass 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 mass-independent.

Weak currents contain 5AC . After eliminating the currents containing 5HV ,
we are left with four essentially dierent currents, having = 1, 0 , , 0 ,
and another four currents with an extra 5AC . When anticommutes with
0 ( = 1), the current has a non-zero matrix element between the vacuum
and a ground-state (0 or 1 ) meson:

0|(
q5AC Q) |B = imB fBP () ,

0|
q 5AC Q|B = ifB p ,

0|
q Q|B = imB fB e ,

0|(
q Q) |B = fBT ()(p e p e ) ,
(5.70)
where e is the B polarization vector. When commutes with 0 ( = +1),
the current has a non-zero matrix element between the vacuum and a P -wave
0+ or 1+ meson. These matrix elements are given by similar formulae with
q q5AC .
The matrix elements (in the relativistic normalization) of the corresponding HQET currents are compactly represented by the trace formula [16]:

|M = F () Tr M ,
0|(
q Q)
2
(5.71)
where the spin wave function M of the meson M is, for the ground-state
mesons,

1 + v/ 5
M = i mM
(5.72)
2
e
/
(for 0+ and 1+ mesons M should be right-multiplied by 5 ; of course, their
F () diers from the ground-state value). This wave function has the obvious
property
v M = M/
/
v=M
(5.73)
( = 1 for ground-state 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
F ()
fBP () = fB () = fB () = fBT () =
mB
(see (1.3)), as a consequence of the heavy-quark spin symmetry. Hence, ratios
of QCD matrix elements are equal to ratios of matching coecients. These
ratios do not depend on the three-loop HQET anomalous dimension.
From the equations of motion,

i q5AC Q = m() q5AC Q .

Taking the matrix element, we obtain [4]

0|(
q 5AC Q) |B
mB
fBP ()

,
=
=
fB
m()
0|
q 5AC 0 Q|B
(5.74)
where m() is the MS running massof the b-quark. To the leading order in
1/m, we may replace mB by the on-shell b-quark 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 one-loop result was derived in Sect. 5.1 (see (5.18)); the two-loop result
was obtained in [11] (a typographical error was corrected in [4]).
The ratio of two observable matrix elements,

0|
qQ|B
m f e

= B B ,
(5.76)
AC
0
mB f B
0|
q 5 Q|B
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 + ,
where c 0.18 is the correction due to mc = 0. The convergence of the

perturbation series is very slow.
Let us now consider the ratio fB /fD (1.4). We have

MS
1
fB =
C 0 (mb , mb )F (mb ) + O
,
mb
mb
where (5.66)
s (mb )
+ O(2s ) , C1 = 2CF .
4
The HQET matrix element F () has the anomalous dimension
j (5.45); its
-dependence is given by the solution of the renormalization-group equation,

/(20 )
s () j0
j1
1 s ()
j0

2
+ O(s ) ,
F () = F
1+
4
20
j0
0
4
C 0 (mb , mb ) = 1 + C1
3/2
where F is -independent, and hence is just a number times MS . Similar

formulae hold for fD . Therefore,

/(20 )

s (mb ) j0
fB
mc
=
1
fD
mb
s (mc )

MS
j1
s (mb ) s (mc )
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
The next-to-next-to-leading-order correction has been calculated recently [6].

For the last ratio,

0|(
q i 0 Q) |B
fBT ()

,
(5.79)
=
fB
0|
q i Q|B
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)
The one-loop term vanishes at = = m, but will of course appear at any

other , . If one wants fBT () at a scale widely separated from m, one
needs to use the three-loop anomalous dimension [10] of the QCD current
q q in conjunction with (5.80).
We can look at the calculation of QCD matrix elements via matching
with HQET from a slightly dierent point of view. Let us return to the
example of a heavy quark with mass m decaying into a light quark with energy
= k v m (Figs. 5.7a,b). The one-loop QCD correction to this process
(Fig. 5.7b) is a complicated diagram with two widely separated large scales.
The decay amplitude (Figs. 5.7a,b) can be written as a matching coecient
times an HQET amplitude. The matching coecient depends only on m,
and is given by the QCD diagrams (Figs. 5.7a,b) at the threshold = 0.
The HQET matrix element (Figs. 5.7c,d) depends only on . Therefore, the
one-loop QCD diagram of Fig. 5.7b at m is just the sum of its value
mv
k
a
b
k
Fig. 5.7. The decay Q q in QCD and HQET
References
119
at the threshold ( = 0) and the HQET contribution (Fig. 5.7d). This is a

simple case of the threshold expansion, which is discussed in the book [17].
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
C. Balzereit, T. Ohl: Phys. Lett. B 398, 365 (1997) 108

D.J. Broadhurst, N. Gray, K. Schilcher: Z. Phys. C 52, 111 (1991) 109
D.J. Broadhurst, A.G. Grozin: Phys. Rev. D 52, 4082 (1995) 93, 111, 112,
116, 118
K.G. Chetyrkin: Phys. Lett. B 404, 161 (1997) 94
K.G. Chetyrkin, A.G. Grozin: Nucl. Phys. B 666, 289 (2003) 103, 118
J.C. Collins: Renormalization (Cambridge University Press, Cambridge 1984)
96
E. Eichten, B. Hill: Phys. Lett. B 234, 511 (1990) 112
V. Gimenez: Nucl. Phys. B 375, 582 (1992) 103
J.A. Gracey: Phys. Lett. B 488, 175 (2000) 93, 118
N. Gray, D.J. Broadhurst, W. Grafe, K. Schilcher: Z. Phys. C 48, 673 (1990)
95, 116
A.G. Grozin: Phys. Lett. B 445, 165 (1998) 105, 107, 112
X.-D. Ji, M.J. Musolf: Phys. Lett. B 257, 409 (1991) 103
S.A. Larin: The Renormalization of the Axial Anomaly in Dimensional
Regularization, in Quarks-92, ed. by D.Yu. Grigoriev, V.A. Matveev,
V.A. Rubakov, P.G. Tinyakov (World Scientic, Singapore 1993), p. 201; Phys.
Lett. B 303, 113 (1993) 98, 101
K. Melnikov, T. van Ritbergen: Phys. Lett. B 482, 99 (2000); Nucl. Phys. B
591, 515 (2000) 95
M. Neubert: Phys. Rev. D 45, 2451 (1992) 115
V.A. Smirnov: Applied Asymptotic Expansions in Momenta and Masses
(Springer, Berlin, Heidelberg 2001) 119
J.A.M. Vermaseren, S.A. Larin, T. van Ritbergen: Phys. Lett. B 405, 327
(1997) 94
6 HeavyLight Currents: 1/m Corrections
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 dimension-4 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 one-loop
anomalous dimension matrix of dimension-4 operators i was found. The
full one-loop 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 two-loop anomalous-dimension matrix was
calculated in [1, 2].
6.1 1/m Corrections to HeavyLight Currents

In the right-hand side of the 1/m expansion (5.57), the i are all local
dimension-4 HQET operators with the appropriate quantum numbers. Matrix elements of this right-hand side are calculated according to the Feynman
rules of the HQET Lagrangian (4.11) with 1/m corrections. To the rst order in 1/m, we can either take a 1/m-suppressed operator in (5.57) and use
only vertices of the leading-order HQET Lagrangian (2.5), or take the leading HQET current in (5.57) and use a single vertex from the 1/m terms in
the Lagrangian (4.11). Alternatively, we can include the bilocal operators
integrals of time-ordered products

j(0), k (x)} , jm = i dx T {
j(0), m (x)}
(6.1)
jk = i dx T {
with the coecients
Bjk (, ) = C (, ) ,
Bjm (, ) = C (, )Cm ( )
(6.2)
in the sum in the right-hand 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
122
this sum. Therefore, renormalization-group equations should be written for

the full set of local and bilocal operators.

or q
The local dimension-4 operators i have the structure qDQ
D Q.
If
this is importhe light-quark mass cannot be neglected, there is also mq qQ;
tant if the light quark is c in the b-quark HQET (see Sects. 7.3 and 7.5).
To nd their coecients, we need to match matrix elements of the QCD
current and its HQET expansion (5.57) from an on-shell heavy quark to an
on-shell light quark, where either the heavy-quark residual momentum p or
the light-quark momentum p is non-zero, with linear accuracy in p/m, p /m,
and mq /m. There is no need to calculate the QCD vertex (p, 0) with an
on-shell p = mv + p, because it can be obtained from (mv, 0) by a Lorentz
transformation. Therefore, the coecients Bi of the operators i with D
can be obtained [7] from the leading-order matching coecients
acting on Q
C (reparametrization invariance).
Let be a combination of matrices not depending on v. This Dirac
matrix can be decomposed as
= + + ,
1
( /v /v ) ,
2
(6.3)
where + commutes with /v , and anticommutes. The matrix element of the

renormalized QCD current q Q from the heavy-quark state with momentum
mv to the light-quark state with momentum 0 is (see (5.59))

1
1
q /v u(mv) .
C+ + C uq u(mv) +
C+ C u
2
2
(6.4)
In this equation, we may substitute v v + p/m:

1
C+ + C uq u(mv + p)
2

1
+
v+p
//m) u(mv + p) .
q (/
C+ C u
2
Using the FoldyWouthuysen transformation (4.59), we obtain the leading
term (6.4) plus

1
/uv + C+ C u
v p
/ + 2/
p ) uv .
q p
q (/
C+ + C u
4m
terms in the sum in (5.57) are
Therefore, the qDQ

1
v + 1 C+ C q (/
v .
C+ + C q iD
/Q
v iD
/ + 2iD
/ )Q
2
2
(6.5)
This derivation can be reformulated in an operator form. In order to

construct reparametrization-invariant objects, one should use not v but
V =v+
iD
,
m
(6.6)
123
which is the gauge-covariant 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 leading-order expression
q Q =

1
v + 1 C+ C q/v Q
v
C+ + C q Q
2
2
is not reparametrization-invariant. It should be replaced by the invariant

combination

iD
/
1
Qv
C+ + C q 1 +
2
2m

iD
/
1
iD
/
C+ C q /
Qv .
v+
+
1+
2
m
2m
v are not reparametrization-invariant separately,
Operators of the form qDQ
and can appear only in this combination. Therefore, we arrive at (6.5).

are not determined by
The coecients of the operators qD Q
and mq qQ
general considerations. These coecients appear rst at the one-loop 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 next-to-leading 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 two-loop level). The QCD matrix
element is proportional to u
(p) (p, mv)u(mv), where (p, mv) is the bare
proper-vertex 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
Fig. 6.1. Proper vertex of the heavylight QCD current
124
u
(p)

xi Li Ri u(mv) ,
(6.8)
with Li Ri = p v 1 1, p v /v , p v , 1 /p, v/ /p. The

coecients xi can be obtained, by solving a linear system, from the double
j and of -matrices to the right of
traces of -matrices to the left of with L
with Rj , where Lj Rj = v//p (1+/

v ), /p (1+/
v ) , /v /p (1+/
v ) ,
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)
(see (5.64)), where

(d 2)(d 8) (d 5)(d 4 + 2h)h
,
(d 2)(d 3)(d 5)
d2h
.
b2 = 2
(d 2)(d 3)
b1 = 2
The rst-order 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
b1 = b1 and b2 = 2b2 for = i , b1 = b1 + 2b2 and b2 = 2b2

for = i 0 . A strong check is provided by the Ward identity: contracting
the vertex function (p, mv) (for = ) with the momentum transfer
(mv p) , we obtain
(p, mv)(mv p) = m (p, mv) + (mv) ,
where (p, mv) is the scalar vertex (for = 1), and is the heavy-quark
self-energy (Sect. 4.2). At the rst order in p, this leads to

(6.11)
b 1 b 0 = 2 c i c 0 ,
where c is the coecient in (5.64).
125
HQET matrix elements of dimension-4 operators between the vacuum

and ground-state 0 and 1 mesons (or P -wave 0+ and 1+ mesons) can be
compactly represented in the trace formalism [6], similarly to (5.71). The

most general form of the matrix element of an operator with D acting on Q
is

F2 ()
0|1 ()|M =
Tr
M,
2

1 = qiD Q
(6.12)
(see (6.23)). If we now take = , this operator becomes

= i q Q

qiD
/ Q
(we shall see in Sect. 6.2 that this is no longer true at the s level), and its
matrix element is
F ()
F ()
Tr v/ M =
Tr M
2
2
( = 1 for ground-state 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)
up to an s correction, which will be obtained in Sect. 6.2.

The matrix elements of the bilocal operators (6.1) are [6]

F ()Gk ()
0|jk ()|M =
Tr M ,
2

F ()Gm ()
1 + v/
0|jm ()|M =
Tr
M .
12
2
(6.14)
The last formula requires some explanation. The chromomagnetic vertex always has the structure (Sect. 4.1), and there is always a heavy-quark
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
0|j|M , 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
/ Q|M
= d 0|
j|M ,
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

j|M ,
0|jk |M = Gk 0|

1
0|jk |M = d Gm 0|
j|M ,
3
where d is dened by

1 + v/ 1 + v/
1 + v/
= 2d
,
2
2
2
d =

1 2
d 3 .
2
Collecting all contributions, we arrive at

0|j()|M
1
1
1

= C () +
d + Gk + d Gm .
2m
3
3
0|
j()|M
(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 spin-0
currents with = 1, 0 (Sect. 6.4) and two spin-1 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) dimension-4 heavylight operators
which appear in these corrections.
6.2 Local Dimension-4 HeavyLight Operators

Let us consider a set of operators Oi closed under renormalization, and write
them as a column matrix. Then the renormalized operators O() and the
bare ones O0 are related by
O0 = Z()O() ,
O() = Z 1 ()O0 ,
(6.18)
where Z is a matrix of renormalization constants. Its 0s term is the unit

matrix, and the corrections have the minimal structure (3.3). The operators
O() obey the renormalization-group equation
dO()
+ (s ())O() = 0 ,
d log
127
(6.19)
where
= Z 1 ()
dZ 1 ()
dZ()
=
Z()
d log
d log
(6.20)
is the anomalous-dimension matrix. The matrix of renormalization constants

can be written as
Z =1+
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 gauge-invariant operators here). One can obtain the anomalous-dimension 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 lower-loop results. With two-loop
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 dimension-4 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 dimension-4 operators satisfying these conditions is listed in (6.23).
Therefore, they are closed under renormalization.
128
Two of the operators are full derivatives of dimension-3 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 light-quark 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 anomalous-diwhere 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
must be nite at 0. This allows one to reconstruct

b,c from
a :
1
1
b ) ,
c = (
c ) ,
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)
The nite parts, at the next-to-leading order, are

s ()
1
1
s ()
a0
2 () + (
a0 m0 )
3 () ,
1 () = 1 () +
3
4
3
4

s ()
1
1
s ()
a0
a0 m0 )
3 () ,
2 () = 1 +
2 () + (
3
4
3
4
3 () = 32 () ,
4 () = 33 () .
(6.31)
Note a fact which seems to contradict our intuition: at mq = 0, the bare

operators are equal, 10 = 10 , but the renormalized ones are not, 1 () =
1 ()! A renormalized operator is dened not only by the corresponding bare
operator, but also by the set of operators being renormalized together. In our
case, additional counterterms in (6.29) yield a nite contribution, because of
the O() term in (6.27).
Let us calculate the anomalous dimensions
a,b,c at one loop. To this end,
with p = 0
we take the on-shell matrix element u
(p, 0) u of 10 from Q
to q with a small on-shell momentum p, and expand it in p and mq up to the
linear term:
(p, 0) = 0 + 1 mq + 1 p +
The tree matrix element does not depend on p or mq . The rst contribution to
the linear term is one-loop (Fig. 6.2; the third diagram, not shown here, does
not depend on p or mq ). The linear terms 1 mq + 1 p have a logarithmic
130

k
k
k+p
k+p
Fig. 6.2. Matrix element of O10
UV divergence, and we may retain only UV 1/ poles (it also is IR-divergent,

and we need an IR cut-o). According to (6.24), the result should be equal
to
a p + Zb p v /

Z
v + Zc mq .
From Fig. 6.2,
(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
The longitudinal part of the gluon propagator vanishes when multiplied by

the second bracket, and the result is gauge-invariant. The term linear in p
and mq is
iCF g02
dd k
(2)d

/v /k /p/k k 2 mq
k
.

kv
(k 2 )3
In the numerator, we decompose k = (k v)v +k . Terms with an odd number

2
of k s vanish; we average over the directions of k using k
k = [k
/(d
1)]g . The HQET denominator cancels. We may use four-dimensional Dirac

algebra, and anticommute p/ to the left, where it gives mq in the on-shell matrix element. Using the UV divergence of the resulting integral given by (4.51),
we obtain
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
in accord with (6.30). The two-loop calculation was performed in [1]:
6.3 Bilocal Dimension-4 HeavyLight Operators
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)

Next we shall discuss bilocal operators with a kinetic-energy insertion k
(4.11), following [2]. The three operators

,
= 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)
The renormalization-group equations for these operators are

dk1 ()
+
j (s ())k1 () = 0 ,
d log
dk2 ()
+ (
j (s ()) + m (s ())) k2 () = 0 ,
d log
dk3 ()
+
j (s ())k3 () +
ak (s ())k1 () +
bk (s ())k2 () = 0 .
d log
(6.37)
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 on-shell matrix
Let us calculate
a,b
with p = 0 to q with a small on-shell
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 one-loop (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 gauge-invariant (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
The two-loop calculation of

ak was performed in [2]:
s
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
k+p
133
Finally, we shall discuss bilocal operators with a chromomagnetic-interaction 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 heavy-quark
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 chromomagnetic-interaction
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 on-shell matrix element of 4 from Q
small on-shell 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 gauge-invariant; 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
Now we decompose k = (k v)v + k and average over the directions of k .

The HQET denominator cancels. The Dirac algebra may be done in four
dimensions; we anticommute /p to the left (where it gives mq ), and take into
account the fact that the indices , live only in the subspace orthogonal to
v. At one loop, there are not enough -matrices to the left of to give the
third structure in (6.41). Using the UV divergence (4.51), we obtain
i
s
CF
(p v + p /v ) (1 + v/) .
4
4
Therefore, the anomalous dimensions are
s
s
+ ,
bm = 2CF
+ ,
cm = O(2s ) .
am = 2CF
4
4
m
The two-loop 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
6.4 Spin-0 Currents
135
6.4 Spin-0 Currents

First, we shall consider the three-scalar QCD currents with = 1, 0 (in the
the 1/m correction
v rest frame). The leading term in (5.57) contains
j = qQ;
contains the ve operators

, 2 = q i
= i0
= i q Q
1 = qiD
/Q
D0 Q
j , 3 = mq
j,

4 = i dx T {
j(0), k (x)} , 5 = i dx T {
j(0), m (x)} .
(6.43)
j 1 , 20 = Z
j 2 ,
The rst three operators are renormalized as 10 = Z
k

30 = Zj Zm 3 . The bilocal operator 4 is 3 (6.33) with = 1; it needs
counterterms proportional to the local operators k1 = 2 and k2 = 3 :
j 3 + Z
ak 2 + Z
bk 3 .
30 = Z
The bilocal operator 5
j 0
Z
0
0 Z
j
0
j Zm
Z = 0 0 Z
k
ak Z
0 Z
b
m

0 Z1 Z2m
also needs a counterterm proportional to 2,3 , and
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
Do not confuse m with

m !
From (6.2) and (6.5),
B11 = B41 = C1 ,
0
B51 = Cm C1 ;
B1 = C 0 2C 1 ,
B4 = C 0 ,
B5 = Cm C 0 .
(6.45)
The renormalization-group equations for these coecients,

B (, )
= T (s ( ))B (, )
log
(6.46)
(where the superscript T means the transposed matrix), x columns 1, 4, and

5 in the anomalous-dimension matrix (6.44).
The operators m
1,2,3,4 (6.39) with = 1 are related to 2,3,5 (6.43):
m
m
10 = (1 )(3 2)20 ,
20 = (1 ) (20 + 30 ) ,
m
30 = (1 )(3 2)30 , m
40 = 50 .
We have
(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
a,b,c in (6.40) and (6.42):
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 next-to-leading order, is
m
4 () = 4 () +
1 m
1 m s ()
m s ()
(
a0 + 5b0
2 ()

3 () . (6.50)
)
2
4
2 a0 4
The unknown coecients B2,3

(, ) for = 1, 0 are obtained by solving
the renormalization-group equations
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 right-hand side.
(m, m) are obtained by matching at = =

The initial conditions B2,3
m. We write down the sum in (5.57) at mq = 0 via the bare operators:
g02 m2
()b 20 10 + 40 + 50
(4)d/2

0m s (m)
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]
6.4 Spin-0 Currents
s (m)
+ ,
4

0
(m)
B2 (m, m) = 12CF s
+
4
137
B21 (m, m) = 8CF
(6.52)
The initial values B3 (m, m) are also O(s ); we do not consider them here.
In the leading-logarithmic 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)
The next-to-leading-order (NLO) results at mq = 0 can be found in [2, 3].

0
1
and B2,3
can be derived. The QCD vector
An exact relation between B2,3
current and the scalar current are related by the equations of motion
i j0 = i j = (m0 mq0 ) j0 = (m() mq ()) j() ,
(6.55)
where m() is the MS running mass of the heavy quark. We separate j =
(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
on-shell 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
q|i dx T {
j( ), (k + Cm ( )m ( ))x } |Q p q|

r=
,
q|
j( )|Q
At the leading order in 1/m, this yields

and j = q Q.
m
C1 (m, m)
=
;
m()
C 0 (m, m)
(6.56)
see (5.74) and (5.75). At the rst order, we obtain [3]

0
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 one-loop case of the rst result.
Now we shall discuss B-meson 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)
0|1 ()|B = 0|2 ()|B = i mB F
The operator 4 () is just k2 () with = 5AC (Sect. 6.3), and the rst
formula in (6.14) gives

0|4 ()|B = i mB F ()Gk () .

(6.60)
Note that jk () and jm () in (6.14) should be understood as k2 () and
m
3 () (Sect. 6.3). Therefore, we obtain from (6.50)

0|5 ()|B = i mB F ()Gm0 () ,

1 m
m s
+ 5
+
b0
)
Gm0 () = Gm () + Rm0 (s ()) , Rm0 (s ) = (
2 a0
4
(6.61)
The hadronic parameters Gk (), Gm0 () obey the renormalization-group
equations
6.4 Spin-0 Currents
dGk ()
=
k (s ()) ,
d log
dGm0 ()
+
m (s ())Gm0 () =
m (s ()) .
d log
139
(6.62)
Their solution in the LLA is

0k
()
k
,
log s
Gk () = G

20
4

m0 /(20 )
m
m s ()
Gm0 () = G
+ 0 ,
4
m0
(6.63)
k,m are -independent and thus are just some (non-perturbative)

where G
numbers times MS .
Taking the matrix element of (5.57), we obtain
P
C (, )F ( )
fB ()
=
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 (, )
Substituting the solutions of the renormalization-group equations, we arrive

at the explicitly , -independent LLA results

P
/(20 )
s (m) j0
C F
fB
=
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 B-meson hadronic parameters; it is just a series in s (m).
Therefore, we recover mB = m + .
140
6.5 Spin-1 Currents

Similarly, for the three-vector currents with = i , i 0 the leading term
and the 1/m correction contains the seven
in (5.57) contains
ji = q i Q,
operators

i
i
, i = qiD
Q
=
i
Q
,
q
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)
From (6.2) and (6.5),

i
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)
The anomalous-dimension matrix of the dimension-4 operators (6.66) has the

structure
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 ;
the operators (6.33) with = i are

k1 = i4 ,
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
6.5 Spin-1 Currents
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 (non-mixing) 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)
These three coecients must be nite at 0. This allows one to reconstruct

m
m
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)
The nite part, at the next-to-leading order, is

s () i
1 m
m s () i
3 () + (
a0 +
4 ()
b0
)
4
2
4

1 m s () i
5 () .
+
2 a0 4
i
m
a0
m
3 () = 6 ()
(6.73)
The unknown coecients B3,4,5

(, ) for = i , i 0 are obtained by
solving the renormalization-group equations
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)
The initial conditions B3,4

(m, m) are obtained by matching at = = m.
We write down the sum in (5.57) at mq = 0 via the bare operators:

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
The values of b,i at = 0 have to cancel these anomalous dimensions. The

O() terms of (6.10) give
i
s (m)
(m)
+ , B4 (m, m) = 4CF s
+ ,
4
4
i 0
i 0
(m)
(m)
B3 (m, m) = 2CF s
+ , B4 (m, m) = 6CF s
+
4
4
(6.75)
i
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
The NLO results at mq = 0 can be found in [2, 3].
(6.76)
6.5 Spin-1 Currents
143
Now we shall discuss B -meson matrix elements, and therefore set mq = 0.

According to (5.71), the leading HQET matrix element is

i
0|
j ()|B = i mB F ()ei ,
(6.77)
where e is the B polarization vector. The operators i2,4 are full derivatives:

()ei .
0|i2 ()|B = 0|i4 ()|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
0|1 ()|M =
M,
2
(6.79)
for the matrix element of 1 (6.23). If we take = , then

3
0|1 ()|M = F2 () Tr M .
2
Taking into account (6.31) and

1
() Tr M ,
0|1 ()|M = 0|2 ()|M = F
2
we obtain, at the next-to-leading order,
1
()R(s ()) ,
F2 () = F
3
s
1
+ O(2s ) .
R(s ) = 1 + a0
3
4
(6.80)
Finally [3],

i
()R(s ())ei .
mB F
0|i1 ()|B = 0|i3 ()|B =
3
(6.81)
The operator i6 () is just k2 () with = i (Sect. 6.3), and the rst

formula in (6.14) gives

(6.82)
0|i5 ()|B = i mB F ()Gk ()ei .
The second formula, together with (6.73), gives

i
0|i7 |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 ( )) .
Substituting the solutions of the renormalization-group equations, we arrive

at the explicitly , -independent LLA results

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)
The ratio fB /fB is

/(20 )
C 1
s (m) m0
0m
fB
2
Gm
1
=
1+
. (6.86)
fB
3m
m0
4
C 0
The NLO results can be found in [2, 3].
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 c-currents describe BB
7.1 HeavyHeavy Current in HQET

Now we are going to consider the current
1 (cosh )J0 ,
J = Z
J

+ Q
J0 = Q
v 0 v0 ,
cosh = v v ,
(7.1)
v and Q
v are scalar static-quark elds with velocities v and v . This
where Q
current is clearly discussed in [9]. Let the sum of all one-particle-irreducible
vertex diagrams with this current (not including external leg propagators) be
The one-loop 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
x2 = t2 + t2 + 2tt cosh

(see (3.51)). In the momentum space,
0
vt
v t
Fig. 7.1. One-loop heavyheavy vertex

(7.2)
146
; 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
cut-o 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
(1 + a0 ) cosh (c s ) + (d 2)(1 a0 )(c4 s4 2 )

,
(c2 + s2 2 )d/2
where c = cosh(/2), s = sinh(/2), and the upper limit T provides an IR

cut-o. We need only the 1/ UV pole given by the integral

T

1
T 2
1+2
=
+ O(0 ) .
d
=
(7.4)

2
2
0
UV
We may put d = 4 in the rest of the formula:

+1
(cosh ) = 1 CF s
Z
d
4 1

cosh
1
(1 + a)
2
2
2 cosh (/2) 1 tanh2 (/2)

1 2 tanh4 (/2)
+ (1 a)
2 .
1 2 tanh2 (/2)
Changing the integration variable = tanh / tanh(/2), we obtain

+/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)
The heavyheavy current with v = v ( = 0) is not renormalized. The

0 over the whole space is the full number of heavy quarks,
integral of Jv
which is always 1 and does not depend on . We can see how this happens.
At one loop, the vertex (7.2) at = 0 becomes just the HQET quark selfenergy (3.62):
t ; = 0) = i(t
+ t )(t)(t ) .
(t,
(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 heavy-quark
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 heavy-quark 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 self-energy 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 heavy-quark line in turn (Fig. 7.2; a dot near a
heavy-quark 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
Q transforms S1 into Sr1 , and hence makes nite.

Multiplying (7.8) by Z

Therefore, ZJ ( = 0) = 1, or
148
=

i
=

i
=

Fig. 7.2. Ward identity for the heavyheavy vertex at = 0
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 (non-amputated)
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 heavy-quark 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 one-loop correction. Let us divide (7.10) by
149
v t
vt
Fig. 7.3. Green function with insertion of the heavyheavy current
+ 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 one-loop correction which has the J vertex inside
(shown in Fig. 7.1); corrections to the external legs cancel here. If this ratio
is re-expressed 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 one-loop contribution,
J (cosh ) =
( coth 1) .
(7.12)
This is the well-known soft-photon radiation factor.

In the real HQET, the Green function can be written as
G(t, t ; cosh ) = exp CF

g02
g04
F
+
C
(C
F
+
T
n
F
)
+
F
A A
F l l
(4)d
(4)d/2
(7.13)
(by the non-abelian exponentiation theorem [7, 6]). If the colour factors of
all two-loop diagrams with two gluons attached to the heavy-quark 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 non-abelian case, the colour factors of
some diagrams also contain a non-abelian part CF CA , which should be taken
into account separately (these parts contribute to FA ). The ratio (7.11) can be
150
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 non-abelian 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 two-loop anomalous dimension contains no
CF2 term.
Fig. 7.4. Diagrams contributing to the exponent
These diagrams were calculated in [9] (except the easiest one, Fig. 7.4a
with the quark-loop 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 one-loop case discussed above (for Fig. 7.4a,
this is slightly easier than the momentum-space 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
It has been written here in a form explicitly even in . It can be rewritten

via just one trilogarithm using
Li3 (1 e2 ) + Li3 (1 e2 )
4
2
+ 3 ,
= Li3 (e2 ) 22 log(1 e2 ) 3
3
3
but then this symmetry is not obvious. Similarly, using
Li2 (1 e2 ) + Li2 (1 e2 ) = 22 ,
Li2 (1 e2 ) + Li2 (e2 ) = log(1 e2 ) +
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)
to all orders in s , as argued in [9]; with two-loop accuracy,

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 perturbative-HQET papers [9] was
written several years before the HQET gold rush of 199091. The one-loop
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
7.2 Flavour-Diagonal Currents at Zero Recoil

In physical applications, we are interested in matrix elements of QCD curc0 b0 . Their anomarents, such as b c transition currents J() = Zj1 ()
lous dimensions are the same as for light quarks (Sect. 5.1). When the initial
and nal quark momenta are p = mb v + p, p = mc v + p , with small residual
momenta p and p , the QCD currents can be expanded in 1/mb and 1/mc , the
coecients being HQET heavyheavy currents with the appropriate quantum
numbers:

J() =
Hi (cosh , , )Ji ( )
i
1
1
+
Gi (cosh , , )i ( ) +
Gi (cosh , , )i ( )
2mb i
2mc i
+
(7.20)
In HQET, s ( ) is the same as in QCD with nl = nf 2 avours. The

J
-dependence of Ji is governed by the HQET anomalous dimension
(Sect. 7.1). The coecients Hi , Gi , Gi , . . . are found by matching equating
on-shell matrix elements of the left-hand and right-hand sides.
For a generic Dirac matrix satisfying (5.5), there are four leading HQET
currents in (7.20):

Hi
cv ibv + ,
c b =
i
i = ,
v ,
/
/
v ,
/v /v .
(7.21)
In particular, for the vector and the axial currents, we have

cv H1V + H2V v + H3V v bv + ,
c b =
H1V = H1 H2 H3 (2 cosh + 1)H4 ,
H2V = 2(H2 + H4 ) , H3V = 2(H3 + H4 ) ;

c5AC b =
cv 5AC H1A + H2A v + H3A v bv + ,
H1A = H1 + H2 + H3 + (2 cosh 1)H4 ,
H2A = 2(H2 H4 ) ,
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 avour-diagonal currents J = Q
The on-shell matrix element of the QCD current is
7.2 Flavour-Diagonal Currents at Zero Recoil
153

os 1
Q, p = mv|J|Q, p = mv = ZQ
Zj u
(mv, mv)u ,
where (p , p) = + (p , p) is the bare proper-vertex function, and the onos
shell heavy-quark 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 = 0|J|
Q,
p = 0 ,
(m2 ). This QCD matrix element should be equal to H Q,
where the HQET on-shell matrix element is

os
p = 0|J|
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 one-loop 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 gauge-invariant:
(m2 ) = CF
g02 m2
2(d 2) + (d 3)h(d 4 + 2h)
.
()
d/2
(d 2)(d 3)
(4)
Fig. 7.5. Proper vertex of the QCD current at one loop
(7.23)
154
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 one-loop 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 self-energy (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)
7.3 Flavour-Changing Currents at Zero Recoil
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)).

Now we shall discuss a more challenging case of avour-changing QCD currents J = c b, but still at v = v.
The one-loop vertex (Fig. 7.5) contains integrals

dd k
.
n
n
1
2
[mb (k + mb v)2 ] [m2c (k + mc v)2 ] 2 (k 2 )n3
The denominators are linearly dependent. We can multiply the integrand by

mb m2c (k + mc v)2 mc m2b (k + mb v)2
1=
,
(mb mc )(k 2 )
thus lowering n1 or n2 , until one of these denominators disappears. The
remaining integrals are single-mass (4.12). The vertex is, of course, symmetric
with respect to mb mc :
(mc /mb ) m12
(mb /mc )
g02
() m12
c
b
,
d/2
d3
mb mc
(4)
(1 h)(d 2 + 2h)
d1
r
.
(r) =
d5
d2
= CF
(7.30)
This result reproduces (7.23) at mc mb and (5.64) at mc 0. The checks

with ZP , ZA , and ZT are passed as before.
The on-shell matrix element expressed via the renormalized s () is

s () mb mc
1
1/2
(Zbos Zcos ) = 1 + CF
4
2
1 2

r (r) r1 (r1 ) 1
r + r (3 2) ,
1r
2
(7.31)
156
where r = mc /mb . It is most convenient to perform matching at
0 = mb mc ,
(7.32)
when the matching coecients are symmetric with respect to mb mc .

The UV divergence of (7.31) (see (7.30)) once more gives us the anomalous
dimension (5.8). The nite part gives the matching coecient, which does
not depend on :

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 two-loop correction can be calculated straightforwardly, because the resulting two-loop integrals are single-mass on-shell ones (Sect. 4.2). However,
calculating many terms of such expansions requires ecient computer-algebra
programming. For the vector and axial currents, the two-loop 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 two-loop 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 two-step matching. First we match J onto the
eective theory in which the b-quark is considered static while the c-quark is
still dynamic (HQET-1) at the scale = mb :
j() +
J(mb ) = C (mb , )
1
Bi (mb , )i () + ,
2mb i
(7.38)
where
j = c b. Some of the dimension-4 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
onto the eective theory in which the c is also considered static (HQET-2)
at the scale = mc :
j(mc ) = E(mc )J ,
i (mc ) = mc Fi (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
s (mb ) s ()

+
4

,
(7.41)
where s , , and the anomalous dimension

j (5.45) involve ni avours.
Now we are going to nd E(mc ) (it does not depend on the Dirac matrix ).
os )1/2 Z
1
To this end, the HQET-1 on-shell matrix element (Zcos Z
j
b
u
1 (mc v, 0)uv from
b with p = 0 to c with p = mc v is equated to E times the

os u
HQET-2 on-shell matrix element Z
= 0 to c with
Q v 2 (0, 0)uv from b with p

p = 0 (in this theory, the number of avours is nl = ni 1; ZJ ( = 0) = 1). All
loop corrections in HQET-2 vanish (unless there is a massive avour lighter
uv 1 .
than c). For any , in HQET-1 we have u1 (mc v, 0)uv = u
The one-loop diagram for the vertex correction is shown in Fig. 7.6:

/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
Fig. 7.6. Proper vertex of the HQET-1 current at one loop
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 scale-free 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)
1 , and we obtain the matching coecient

The 1/ pole is cancelled by Z
j
E(mc ) = 1 4CF
s (mc )
+
4
(7.43)
Collecting all the pieces together, we obtain
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 two-loop 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
next-to-leading logarithms in (7.37), i.e., the one-loop 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 ) .
In 2 , we may replace i D by the c-quark momentum mc v:
(7.45)
160
2 (mc ) = mc
j(mc ) .
(7.46)
For the vector current c 0 b, the coecients B2,3

(mb , mc ) are given by (6.54)
(the lower sign, with = mc and ni avours), and the O(r) term in
H 0 (mb , mc ) is

0
r 0
B2 (mb , mc ) + B3 (mb , mc )
2

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 )
For the axial current c5AC i b, the leading HQET current is

ji = c5AC ib,
and the subleading operators are obtained from (6.66) by the substitution c
ji . With this denition, the anomalous-dimension
c5AC , except i5 = mc
matrix (6.69) is the same as in Sect. 6.5. Only i3,4,5 can produce O(mc )
contributions:
i3 (mc ) = 0 ,
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
single-step matching results (7.34), with the two-loop corrections [2]. These
results are expressed via s (0 ), and contain all powers of r. Then we take the
two-step matching results (7.44), and subtract from them the terms of order
s (0 ) and 2s (0 ), which are already accounted for in the single-step results.
By adding this dierence, we take into account an innite sequence of radiative corrections of order O(r0 ) with leading and next-to-leading powers of
the logarithm L. Similarly, we take the O(r) two-step 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 .
7.4 Flavour-Diagonal Currents at Non-Zero Recoil
161
7.4 Flavour-Diagonal Currents at Non-Zero Recoil

Now we are going to consider avour-diagonal currents at = 0 [4]. The onos 1
shell QCD matrix element ZQ
ZJ ()
u( + (mv , mv))u should be equal

1

os
( )
0) = 0, if there are no
to Z
uv [ i Hi (, )i ] uv (ZQ = 1 and (0,
J
other massive avours). The one-loop vertex correction (mv , mv) (Fig. 7.5)
is obviously gauge-invariant. Using the Feynman parametrization (2.13), we
have

(/
k + m/v + m) (/
k + m/v + m)
dd k
2
= iCF g0
d
2
2
(2) (k )(k 2m v k)(k 2 2m v k)

dd k
= 2iCF g02 dx dx
(2)d

k + m(1 x )/
v mx/v + m) (/
k + m(1 x)/
v mx /v + m)
(/
,
(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
The dilogarithms here are not independent:

Li2 (e ) + Li2 (e ) =
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)
At 0, H1 + H2 + H3 + H4 reproduces (7.27). In particular, we obtain the

following [4, 10] from (7.22) for the vector (n = 1, = 1) and axial (n = 1,
= 1) currents:

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
7.5 Flavour-Changing Currents at Non-Zero Recoil

Now we shall discuss a more challenging case of avour-changing QCD currents J = c b. The momentum transfer squared is
q 2 = (mb v mc v )2 = 2mb mc (cosh L cosh )
L
mb
L+
sinh
, L = log
.
= 4mb mc sinh
2
2
mc
(7.54)
In the decay channel, varies from 0 to L (this case corresponds to q 2 = 0);

> L is the scattering channel.
The one-loop vertex correction (mc v , mb v) (Fig. 7.5) is obviously gaugeinvariant:

dd k (/
k + mc /v + mc ) (/
k + mb /v + mb )
2
= iCF g0
(2)d (k 2 )(k 2 2mb v k)(k 2 2mc v k)
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
a2 = (mb xv + mc x v )2 = m2b x2 + m2c x2 + 2mb mc xx cosh .

We calculate the loop integral using (2.16), and substitute x = (1 + z)/2,
x = (1 z)/2, and
a2 = mb m c 2 a+ a ,
a = cosh
L
L
+ z sinh
.
2
2
The integration is trivial:

g02 (mb mc )
(1 + )
(4)d/2

+1
(1 z 2 )h
cosh
a+ a h 2
dz
+

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 ,
and the integrals can be calculated:

+1
dz
2 sinh(1 )L
= 2 [1 (L coth L 1)] + O(2 ) ,
=
(1 ) sinh L
1 a+
+1
2L
dz
2 sinh L
2
1+ = sinh L = sinh L + O( ) ,
a
1
+
(7.56)
164
+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
The matching coecients at = = 0 are

H1 = 1 + H4 + CF
s (0 )
(2 L coth L)(n 2)2 + 2(n 2)
4

+ 2(2L coth L 1) ,
s (0 ) 2LeL + (e2L 3) sinh L

H2 = H3 (L L) = CF
(n 2)
4
8 sinh3 L

LeL sinh L
+
,
sinh2 L
s (0 ) L sinh L cosh L
(n 2) .
(7.57)
H4 = CF
4
4 sinh3 L
At L 0, H1 + H2 + H3 + H4 reproduces (7.27).
In general, the integrals are

+1
L sinh L sinh
dz
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
The limiting cases considered previously are reproduced. In particular,

F (L, L) = 0, F (L, ) = 2[L(cosh L + 1)/(2 sinh L) 1] + O(2 ) at 0,
and F (0, ) = F () (7.51).
It is most natural to perform matching at = = 0 (7.32). For any
other , , the matching coecients are determined by the renormalization
group:
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 )
In the avour-diagonal case (L = 0), the result (7.52) is reproduced; at

q 2 = 0 ( L), the formula (7.57) follows. At 0, H1 + H2 + H3 + H4
reproduces (7.33).
In particular, for the vector and axial currents, we have [10]
s (0 ) 2 cosh L(cosh + 1) cosh (3 cosh + 2) + 1

V
H1 = 1 + CF
4
sinh (cosh L cosh )

L sinh L
4 + 2F (L, ) coth ,
+
cosh L cosh
166

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 single-step 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
two-step 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 HQET-1,
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)
(see (5.57)). The running is given by (5.58), with ni avours in HQET-1.

Finally, we match these operators onto HQET-2, where both the b and the
c are considered static, at mc :
j (mc ) = E (mc )J (mc ) + O(1/mc ) ,

i (mc ) = mc
Fij (mc )Jj (mc ) + O(m0c ) .
j
(7.62)
167
We obtain (see (7.20))

1
[C+ (mb , mc )E+ (mc ) C (mb , mc )E (mc )] + O(r) ,
2
(7.63)
H3,4 (mb , mc ) = O(r) .
H1,2 (mb , mc ) =
With this method, we obtain only a few terms in expansion in r = mc /mb ;

on the other hand, the leading (and next-to-leading, etc.) logarithmic terms
(in L = log r) can be summed.
1 c0 bv0 onto HQET-2, we
In order to match the HQET-1 current
j=Z
j
calculate its on-shell matrix element from
b with residual momentum p = 0
os )1/2 u
1 (Z os Z
1 (mc v , 0)uv . This matrix
to c with momentum mc v : Z
c
j
b
element should be equal to E times the corresponding matrix element of the
os )1/2
1 (cosh )(Zos Z
cv 0 bv0 , which is Z
HQET-2 current J = ZJ1 (cosh )
c
b
J

u
v 2 (0, 0)uv . All loop corrections in HQET-2 vanish: 2 (0, 0) = 1, Zcos =
os = 1 (unless there is a massive avour lighter than c; then corrections will
Z
b
start from two loops). The HQET-1 vertex 1 can only contain -matrices to
the left of , owing to the HQET Feynman rules. We can anticommute v/ to
the left, where it disappears. If either commutes or anticommutes with v/,
then u
1 (mc v , 0)uv = u uv (m2c ). Therefore, just one HQET-2 current
appears, as shown in (7.62).
At one loop (Fig. 7.6),

/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 .
/ term vanishes, and we obtain the

Now we take the k integral (2.16); the k
following for :
g02
1 + 2CF
(1 + )
(4)d/2
dy
0
dx
0
y 2mc (1 x) cosh mc x
.
(y 2 + m2c x2 + 2mc xy cosh )1+
After the substitution y = mc xz, the x integration is trivial:

g 2 m2
dz
= 1 + 2CF 0 cd/2 (1 + )
2 + 1 + 2z cosh )1+
(z
(4)
0

z + 2 cosh 1 cosh
+
.
1 2
168
When = 0, the integral is easily calculated; must commute with /v (the

upper sign), and (7.42) is reproduced.
We need two integrals with respect to z. The rst one is convergent, and
needed with O() accuracy:

dz
2
(z + 1 + 2z cosh )1+
0

1
+
Li2 (1 e2 ) Li2 (1 e2 ) + O(2 ) .
=
sinh
2
It has been written in a form explicitly even in ; owing to (7.16), it can be
expressed via a single dilogarithm. The second integral,

z dz
,
2 + 1 + 2z cosh )1+
(z
0
is UV divergent at = 0. We split the integration region into two parts: from
0 to A
1 and from A to . In the rst region, we may neglect :
A
z dz
= log A coth .
2
0 z + 1 + 2z cosh
In the second region, we may neglect 1 as compared with z:

dz
1
log A .
=
1+2
2
A z
In the sum, log A cancels, as expected:

z dz
1
coth + O() .
=
2
1+
(z + 1 + 2z cosh )
2
0
Substituting these integrals, we nd the vertex
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 )
+ CF 2 coth + Li2 (1 e2 ) Li2 (1 e2 ) coth

2
4

j1
s (mb ) s (mc )
j0
+
1
,
20
j0
0
4

s (mb )
2 s (mc )
H2V = 2CF xj0

, H3V = 0 ,
4
sinh 4
(mb )
H1A = xj0 1 4CF s
4

s (mc )
+ CF 2 tanh + Li2 (1 e2 ) Li2 (1 e2 ) coth

2
4

j1
s (mb ) 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 )
These results, expanded up to O(s ), reproduce HiV,A (mb , mc ) (7.58) (also

expanded up to O(s )) with HiV,A (0 , 0 ) (7.60), where
F (L, ) =

1
Li2 1 e2 Li2 1 e2 + L + O(r) ,
2
r = eL .
At 0, H1V + H2V + H3V (7.65) reproduces H 0 (7.44), and H1A (7.65)

reproduces H5AC (7.44).
Now we shall discuss the O(r) terms in the LLA [3]. For the vector current
J = c b, the 1/mb term in (7.61) is

v/
1
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
level, we may replace i D by mc v :

c ) cosh , 3 (mc ) = mc J(m
c) ,
2 (mc ) = mc J(m

v cosh ) ,
3 (mc ) = mc J(mc )(v

cosh ,
4 (mc ) = mc J (mc ) J(mc )v

,
5 (mc ) = mc J (mc ) J(mc )v
(7.66)
170
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/
B2 (mb , mc ) B3 (mb , mc ) B4 (mb , mc ) cosh

=
2

B3v/ (mb , mc ) B5 (mb , mc )

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
single-step 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 two-step 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 next-to-leading powers of the logarithm L. Similarly,
we take the O(r) two-step 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 .
7.6 1/m Corrections
171
7.6 1/m Corrections

There are two kinds of operators i in (7.1): local (having the structure
cv Dbv ), and bilocal,

i dx T Ji (0), bk,m (x) ,
where bk,m are the b-quark kinetic energy and chromomagnetic interaction.
cv Dbv ) or bilocal, containing ck,m .

Similarly, the operators i are local (
The coecients of the bilocal operators are just the leading-order coecients
Hi times the coecients of k,m in the Lagrangian (1 and Cm ).
cv Dbv and

cv D bv are completely
The coecients of the local operators
determined by reparametrization invariance [11]. The leading-order expression (7.21) is not reparametrization-invariant. Instead of v and v , one should
use the operators (6.6)
V =v+
iD
,
mb
V + = v
iD
.
mc
cv , one should use (6.7)

Instead of bv and

/
iD
/
bV = 1 + iD
cV =
cv 1
.
bv ,
2mb
2mc
(7.68)
(7.69)
Finally, the argument v v of Hi should be replaced by

V V + = v v +
iv D iv D
.
mb
mc
(7.70)
Then (7.21) becomes a reparametrization-invariant combination

cv [H1 + H2 /

v + H3 /v + H4 /v /v ] bv

1
+
cv (H1 + H2 /v + H3 /v + H4 /v /v ) iD

/
2mb
+ 2iD
/ (H2 + H4 /v )

+ 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
where the Hi are derivatives of Hi with respect to the argument v v . Op
cv Dbv ,
cv D bv are not reparametrization-invariant 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 v iv Dbv = v J2 ,

6 =
cv v iv Dbv = v J3 ,

7 =

cv (i) D
/ bv ,
cv (i) Dv
/ bv ,
1 =
2 =

cv (i) D bv ,
4 =
cv (i)v D bv = iv J1 ,

5 =
cv (i)v Dv bv = iv J2 ,

6 =
cv (i)v Dv bv = iv J3 .

7 =

cv (i) D
/ v bv ,
3 =
Similar formulae hold for the axial current.

If we are interested in matrix elements of J from the ground-state Bmeson to the ground-state D or D -meson, we can easily calculate the matrix
elements of operators which are full derivatives. Their eect is to replace
cosh = vB vD in the argument of HiV in the leading-order terms by the
new variable

1
1
+
(7.73)
(vB vD 1) .
cosh = vB vD +
mc
mb
With this improved angular variable, the vector current J is equivalent to
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)
when only matrix elements between ground-state mesons are considered. Of

course, similar formulae hold for the axial current.
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 large-order
behaviour of various perturbative series has attracted considerable attention
during recent years. Most of the results obtained so far are model-dependent:
they are derived in the large-0 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 large-0 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.
8.1 Large-0 Limit

Let us consider a perturbative quantity A such that the tree diagram for it
contains no gluon propagators. We can always normalize the tree value of A
to be 1. Then the perturbative series for the bare quantity A0 has the form
A0 = 1 +
L1

aLn nnf
L=1 n=0
g02
(4)d/2
L
,
(8.1)
where L is the number of loops. This series can be rewritten in terms of

0 = (11/3)CA (4/3)TFnf instead of nf :
A0 = 1 +
L1

L=1 n=0
aLn 0n
g02
(4)d/2
L
.
(8.2)
Now we are going to consider 0 as a large parameter such that 0 s 1,

and consider only a few terms in the expansion in 1/0 s :

0 g02
1
1
A0 = 1 + f
+O
.
(8.3)
d/2
0
02
(4)
176
This regime is called large-0 limit, and can only hold in QCD with nf .
Note that it has nothing in common with the large-Nc limit, because we
cannot control the powers of Nc in the coecients aLn .
There is some empirical evidence [7] that the two-loop coecients a21 0 +
a20 for many quantities are well approximated by a21 0 . It is easy to nd a21
from a diagram with a quark-loop insertion into a gluon propagator in the
one-loop correction. Then we can estimate the full two-loop 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 one-loop
correction. We shall assume for now that there is only one gluon propagator
and there are no three-gluon 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)
Only the one-loop contribution a10 contains the additional gauge-dependent

term a1 a0 .
The large-0 limit, as formulated above, does not correspond to summation of any subset of diagrams. If we include not only quark loops, but also
gluon and ghost ones (Fig. 3.2), then (4/3)TFnf in (8.4) is replaced by

4
CA
3 2
P = 0
(a0 + 3) 1 (a0 + 3)
8 +
(8.6)
3
3
2(1 )
2
8.1 Large-0 Limit
177
(see (3.28)). Summing these diagrams yields a gauge-dependent result. This

gauge dependence is compensated (for a gauge-invariant A0 ) by other diagrams, which have more complicated topologies than a simple chain, and
are impossible to sum. In the gauge a0 = 3, the one-loop running of s
is produced by one-loop insertions into the gluon propagator (Fig. 3.2) only,
without vertex contributions. Summation of chains of one-loop insertions into
the gluon propagator in this gauge is equivalent to the large-0 limit.
In the large-0 limit, 1 0 , 2 02 , etc. Therefore, the -function is
equal to
=
0 s
4
(8.7)
(this term is of order 1) plus O(1/0 ) corrections. At the leading order in

1/0 , the renormalization-group equation
d log Z
=
d
+
(8.8)
(see (3.7)) can be explicitly integrated:

Z =
1
.
1 + /
(8.9)
To the leading order in 1/0 ,

s () =
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!
(here (x)n = x(x + 1) (x + n 1) = (x + n)/ (x) is the Pochhammer

symbol). In the applications we shall consider, F (, u) is regular at the origin:
178
F (, u) =
Fnm n um ;
(8.13)
n=0 m=0
however, I know no general proof of this fact. Substituting these expansions

into (8.11), we obtain a quadruple sum expressing A0 via the renormalized
quantities.
The bare quantity A0 is equal to ZA, where both Z and A have the form
1+O(1/0 ). Therefore, we can nd Z 1 with 1/0 accuracy just by retaining
all terms with negative powers of in this quadruple sum. The renormalized
A 1, with 1/0 accuracy, is given by the terms with 0 . It is enough to nd
Z1 , the coecient of 1/ in Z, in order to obtain the anomalous dimension
= 2
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
Therefore, the anomalous dimension is [13]

1
= 2 F (, 0) + O
.
0
02
(8.14)
Collecting terms with 0 in the quadruple sum for A0 , we obtain for

0 (A 1)
(F10 + F01 ) 2 (F20 + F11 + F02 ) + 3 (F30 + F21 + F12 + F03 )
4 (F40 + F31 + F22 + F13 + F04 ) +
1
+ 2 (F20 + 2F11 + 4F02 ) 3 (F30 + 2F21 + 4F12 + 8F03 )
2
3
+ 4 (F40 + 2F31 + 4F22 + 8F13 + 16F04 ) +
2
8.1 Large-0 Limit
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
Therefore, the renormalized quantity is [6]

0
1
F (, 0) F (0, 0)
d
0

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 renormalization-group 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)
(see (4.72)), where

A = A(0 )K (s (0 )) .
At the rst order in 1/0 , we obtain from (8.14)
1
K (s ) = 1 +
0
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
Let us suppose that m MS is the characteristic hard scale in the

quantity A. Then F (, u) contains a factor (/m)2u . When we take the limit
0, the factor D()u/1 in (8.12) becomes exp [(5/3)u]. Therefore,
180

F (0, u) =
e5/6
m
2u
F (u) ,
F (u) = ua1 (1 + u)m2u .
(8.17)
It is most convenient to use

0 = e5/6 m
(8.18)
In the rest of this chapter, will mean 0 s (0 )/(4).

in the denition of A.
This renormalization-group invariant is

1
1
A = 1 +
du eu/ S(u) + O
,
(8.19)
0 0
02
where
S(u) =
F (u) F (0)
.
u
(8.20)
Here,
eu/ =
e5/6 MS
m
2u
.
(8.21)
If we substitute the expansion

S(u) =
sL uL1
L=1
into the Laplace integral (8.19), we obtain the renormalized perturbative

series

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
Therefore, S(u) can be obtained from A (8.22) by using

S(u) =

cL uL1
,
(L 1)!
(8.24)
L=1
which is called the Borel transform.

We see that the function F (, u) (8.12) contains all the necessary information about the quantity A at the order 1/0 . The anomalous dimension (8.14)
is determined by F (, 0), and the renormalization-group invariant A (8.19)
(which gives A() (8.16)) by F (0, u). These formulae are written in the Landau gauge a = 0; if a = 0, additional one-loop terms from the longitudinal
part of the gluon propagator (3.21) should be added.
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 half-axis 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)
These renormalon ambiguities are commensurate with 1/m power corrections

contributions of matrix elements of higher-dimensional operators to the
quantity A. The full result for the physical quantity A must be unambiguous.
Therefore, if one changes the prescription for handling the integral across
the renormalon singularity at u = u0 , one has to change the values of the
dimension-2u0 matrix elements accordingly. This shows that renormalons can
only happen at integer and half-integer values of u, corresponding to the
dimensionalities of the allowed power corrections. The largest ambiguity is
associated with the renormalon closest to the origin.
The renormalon pole (8.25) yields a contribution to the coecients cL of
the renormalized perturbative series (8.22) equal to
cL = r
(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
renormalon ambiguity. The fastest-growing contribution to cL comes from

the renormalon closest to the origin.
Note that renormalons at u0 < 0 give sign-alternating factorially-growing
coecients (8.27). For such series, the integral (8.19) provides an unambiguous denition of the summation called the Borel sum.
Renormalon singularities can result from either UV or IR divergences of
the one-loop integral. Suppose that the integral behaves as d4 k/(k 2 )nUV at
k , so that the degree of its UV divergence (at d = 4) is UV = 4 2nUV .
When we insert the renormalon chain, the power changes: nUV nUV + (L
1) = nUV + u if = 0 (which is the case when one is calculating S(u)). This
integral can have a UV divergence only at u 2 nUV = UV /2. Therefore,
UV renormalons can be situated at UV /2 and to the left. Only quantities A
with power-like UV divergences at one loop have UV renormalons at positive
u. The divergence at u = 0 is the usual UV divergence of the one-loop integral,
which is eliminated by renormalization; renormalized quantities have no UV
renormalon at u = 0.

Similarly, if the one-loop integral behaves as d4 k/(k 2 )nIR at k 0
(where k is the virtual gluon momentum), so that the degree of its IR
divergence is IR = 2nIR 4, S(u) can have an IR divergence only at
u 2 n = IR /2, and IR renormalons can be situated at IR /2 and
integer and half-integer points to the right of this point. Quantities described
by o-shell diagrams have nIR = 1, and their IR renormalons are at u = 1
and to the right.
We can get a better understanding of the physical meaning of renormalons
if we rewrite (8.19) in the form [11]

1
s ( 0 )
d
A = 1 +
w( )
+O
.
(8.28)
2
0
0
This looks like the one-loop 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 one-loop diagram; it is normalized to the coecient
of s /(4) in the one-loop correction. Inside the 1/0 term in (8.28), we may
use the leading-order 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
8.3 Light Quarks

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
where u0 should lie in the gap between the IR and UV renormalons.

For < 1, we can close the integration contour to the right. If S(u) has
IR renormalons ri /(ui u), then

ri ui .
(8.32)
w( ) =
IR
The leading term at small is given by the leftmost IR renormalon. If our

quantity A is IR nite at one loop, all ui > 0, and w( ) 0 at 0. Similarly, for > 1, we can close the contour to the left. If the UV renormalons
are ri /(u ui ), then

w( ) =
ri ui .
(8.33)
UV
The leading term at large is given by the rightmost UV renormalon. If A

is UV nite at one loop, all ui < 0, and w( ) 0 at .
All virtualities (including small ones) contribute to (8.28). The behaviour
of the distribution function w( ) in the small-virtuality region 0 is
determined by the IR renormalon closest to the origin. The integral (8.28)
is ill-dened, just like the original integral (8.19). The one-loop s (8.10)
becomes innite at = (e5/6 MS /m)2 (the Landau pole), and we integrate
across this pole. This happens at small ; substituting the asymptotics (8.32)
of the distribution function at small virtualities, we see that the residue at
this pole, given by the IR renormalon nearest to the origin, is again equal
to (8.26).
8.3 Light Quarks

First, we shall discuss the massless-quark propagator at the order 1/0 . The
one-loop expression for V (p2 ) (Fig. 3.5) in the Landau gauge with the gluon
denominator raised to the power n = 1 + (L 1) is
CF
a1 (n)
=i 2
p
(4)d/2
/ (/
k+p
/)
dd k 14 Tr p
(2)d [(k + p)2 ] (k 2 )n

k k
g +
.
k 2
184
Using the one-loop integrals (2.18), we can easily nd the function F (, u)

(8.12). Such functions, for all o-shell massless quantities, have the same
-function structure, resulting from (2.18) with n2 = 1 + u :
2 u
(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)
At one loop (L = u/ = 1), the Landau-gauge self-energy vanishes (see

(3.47)); at L = 2, the 0 term in the two-loop result (3.56) is reproduced.
1

p , with 1/0 acThe massless-quark propagator S(p) = (1 V (p2 ))/
curacy, in the Landau gauge, is equal to 1//p times (8.11), where F (, u) is
given by (8.34) and (8.35). Terms with negative powers of in the expression
for p
/S(p) via renormalized quantities form Zq . The anomalous dimension is
given by (8.14):

1
N (, 0)
+O
=
.
30 B(2 + , 2 + ) (3 + ) (1 )
02
In the general covariant gauge, the one-loop 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 two-loop 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 renormalization-group
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)
8.3 Light Quarks
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 light-quark 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 self-energy, like that of any o-shell 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 gauge-invariant
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
Fig. 8.1. UV renormalons (black squares) and IR renormalons (black circles) in

the light-quark self-energy (a); the virtuality distribution function (b)
Now we shall discuss light-quark 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
where h is dened by (5.32) (if a0 = 0, the one-loop term proportional to a0

from (5.34) should be added). If L = u/ = 1, the one-loop result (5.34) is
reproduced. For the longitudinal vector current (h = 1 d/2), the result can
be obtained from V (p2 ) using the Ward identity (Sect. 5.3). The result for
a (p2 ) is a little more complicated, and provides no new insight; we omit it.
The anomalous dimension of the current is
jn =
d log Z n
+ q ,
d log
where the derivative of Z n is given by (8.14). We arrive at [7]

1
(n 1)(3 n + 2)
s
2
+O
jn = CF
3
4 B(2 + , 2 + ) (3 + ) (1 )
02

n 15
13n 35 2
s
+ .
= 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 two-loop 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 large-0 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)
They reproduce the leading 0 terms of (5.42) and (5.41).
8.4 HQET Heavy Quark

Now we turn to the HQET static-quark propagator. The one-loop expression

for ()/
(Fig. 3.12) in the Landau gauge with the gluon denominator
raised to the power n = 1 + (L 1) is

v v
a1 (n)
iCF
k k
dd k
=
+
g
.
2
(2)d k v + (k 2 )n
k 2
(4)d/2
187
Using the one-loop integrals (2.27), we can easily nd the function F (, u)

(8.12). Such functions, for all o-shell HQET quantities, have the same function structure, resulting from (2.27) with n2 = 1 + u :

2u
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)
If a0 = 0, the one-loop term proportional to a0 from (3.61) should be added

to ().
This result reproduces (3.61) at L = u/ = 1 and the 0 term
from (3.73) at L = 2.

1

, with 1/0 accuracy,
The static-quark propagator S()
= ()
in the Landau gauge, is equal to 1/ times (8.11), where F (, u) is given
by (8.42) and (8.43). Terms with negative powers of in the expression for

Q . The anomalous dimension is
S()
via renormalized quantities form Z
given by (8.14):

1
N (, 0)
+O
.
=
60 B(2 + , 2 + ) (2 + ) (1 )
02
In the general covariant gauge, the one-loop 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 renormalization-group
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
(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 static-quark self-energy 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 self-energy, like that of any o-shell
quantity, is IR = 2, and the IR renormalons are at u 1.
1
b
Fig. 8.2. Renormalons in the o-shell HQET self-energy (a) and in the on-shell
heavy-quark self-energy (b)
Here we encounter a radically new situation: a UV renormalon at u > 0.

This renormalon leads to an ambiguity ()/
= (r/0 )e5/5 MS /(2),
where r = 4CF is the residue of S(u) at u = 1/2. If we change the prescription
for handling the pole at u = 1/2, we have to change the zero-energy level
of HQET. Therefore, the HQET residual meson energy has an ambiguity

= ()
of order MS /0 [3] (see also [5]):
= 2CF e5/6
MS
.
0
(8.46)
The structure of the leading UV renormalon at u = 1/2 can be investigated beyond the large-0 limit [1]. The renormalization-group 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
Fig. 8.3. Integration contour
We have to use a formula for s () that is more precise than the one-loop
one (8.10). The renormalization-group 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)
The UV renormalon ambiguity must be equal to MS times some

number:
= N0 0 ,
0 = 2CF e5/6
MS
.
0
(8.50)
The normalization factor N0 is only known in the large-0 limit:

N0 = 1 + O(1/0 ) ;
In general, it is just some unknown number of order 1. Comparing (8.48)
with (8.50), we conclude that at u 1/2,
190
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 self-energy with a kinetic-energy insertion
easily calculated in the large-0 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 heavy-quark 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 one-loop diagram (Fig. 5.3b),
as well as this one-loop 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 gauge-invariant, 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 cut-o). 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,
8.5 On-Shell Heavy Quark in QCD
i
22n1 d/2
191
(d/2 n) (2n 1)x2 g + (d 2n)x x

,
(n + 1)
(x2 + i0)d/2n+1
instead of (3.51). Following the derivation in Sect. 7.1, we nd the one-loop

coecient

2n+2d +/2
(d/2 n 1) i
d
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)
This anomalous dimension vanishes at = 0 (Sect. 7.1). It reproduces the nl

term in (7.15).

Now, we turn to the on-shell mass and wave-function renormalization of a
heavy quark in QCD at the order 1/0 . To this end, we need to calculate the
function T (t) (4.28) with linear accuracy in t. Closely following the derivation
in Sect. 4.2 (with the gluon denominator raised to the power n = 1+(L1)),
we obtain

a1 (n)
v + 1) (/
p+k
/ + m)
k k
dd k Tr(/
= iCF
g +
(2)d
4mD1 (t)D2n
D2
(4)d/2

d
1
d k
(d 2)D2
2
= 2iCF
(1 t) + O(t ) .
1
(2)d D1 (t)D2n
4m2
192
Expanding 1/D1 (t) up to the linear term in t and using the one-loop onshell integrals (4.15), we nd a1 (n) and F (, u) (8.12). The functions F (, u),
for all on-shell 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 on-shell 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 renormalization-group 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 self-energy 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 on-shell quark self-energy 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
At 0, w( ) , and at , w( ) 1/ , according to the positions

of the nearest IR and UV renormalons.
w( ) =
Fig. 8.4. Virtuality distribution function
The IR renormalon ambiguity of the on-shell mass is the following [3],

from the residue of S(u) at the leading IR renormalon u = 1/2:
m = 2CF e5/6
MS
.
0
(8.62)
The meson mass (1.1) is a measurable quantity, and must be unambiguous.

In HQET, it is an expansion in 1/m. Its leading term, m, is a short-distance
is a long-distance
quantity a parameter of QCD. The rst correction, ,
quantity, determined by the meson structure at the connement scale. However, the MS regularization scheme contains no strict momentum cut-os. As
a result, the on-shell mass m also contains a contribution from large distances,
where perturbation theory is ill-dened. This produces the IR renormalon
ambiguity (8.62), which is suppressed by 1/m as compared with the leading
term. Likewise, contains a contribution from small distances, which leads
to the UV renormalon ambiguity (8.46). These ambiguities compensate each
other in the physical quantity the meson mass. In other words, in MS the
194
separation of the short- and long-distance contributions is ambiguous, even

though the full result is not.
This cancellation should hold beyond the large-0 limit. Therefore [1],
S(u) =
2CF N0
(1/2 u)1+1 /(20 )
[1 + O (1/2 u)] ,
(8.63)
where the power is exact. The coecients in the perturbative series

L
s (0 )
m
1
=1+
cL
m
0
4
L=1
at L 1 are, according to (8.23),

cn+1 = 21+a 2CF N0 (20 )n (1 + a)n [1 + O(1/n)] ,
a=
1
.
202
From the Stirling formula, (n + 1 + a) = na n![1 + O(1/n)], and we arrive at

2
cn+1 = 4CF N0 n! (20 )n (20 n)1 /(20 ) [1 + O(1/n)] .
(8.64)
This result is model-independent.

1
os
= [1 T (0)] at the rst order
Our calculation of T (t) also yields ZQ
in 1/0 . It has the form of (8.11) and (8.58) with
NZ (, u) = 2CF (3 2)(1 u)(1 + u )
(8.65)
(see (8.59)). If we retain only negative powers of , we should obtain

os = 1). Therefore, calculating
Zq ()/ZQ (), according to (4.35) (because Z
Q
the corresponding anomalous dimension by use of (8.14), we obtain

1
(1 + ) (1 + (2/3))
s
+O
q
Q = 2CF
.
4 B(2 + , 2 + ) (3 + ) (1 )
02
This dierence is gauge-invariant at the 1/0 level; it agrees with (8.36)
and (8.44). If we retain terms with 0 , we obtain the nite combination
os
ZQ
ZQ ()/Zq () of the form (8.15); the corresponding renormalization-group
invariant (8.19) has [12]

(u) (1 2u)
1
(1 u2 )
S(u) = 6CF
.
(3 u)
2u

Now we shall discuss [9] the chromomagnetic-interaction coecient Cm ()
(Sects. 4.44.6). It is dened by matching the on-shell scattering amplitudes
195
in an external chromomagnetic eld in QCD and HQET at the linear order

in the momentum transfer q. All loop diagrams in HQET contain no scale
and hence vanish. The QCD amplitude at the rst order in 1/0 is given by
the L-loop diagrams with L 1 quark loops (Fig. 8.5). The results have the
form (8.58). The diagram of Fig. 8.5a is calculated in the standard way, and
gives
Na (, u) = (2CF CA )(3 + 2u u2 5 + 3u 2u2 + 22 22 u) .
We must sum the diagrams of the type shown in 8.5b over l from 0 to L 1.
All terms in the sum are equal, so that the summation just gives a factor
L = u/:
u
Nb (, u) = CA (5 2u 3) .
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 one-loop 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 one-loop on-shell quark wave-function 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)
The sum is regular at the origin = u = 0, unlike the separate contributions.

It reproduces the known results for L = u/ = 1 and 2 (Sect. 4.4).
L1l
L1
a
Fig. 8.5. Quark scattering in an external gluon eld
L2l
196
Now we can easily nd the anomalous dimension

m and Cm () with 1/0
accuracy. The anomalous dimension (8.14) is [9]

1
s
(1 + 2) (5 + 2)
+O
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 two-loop result (4.67),

s
s
1 + (130 25CA )
+ .
m = CA
2
24
The perturbative series (8.67) converges for 0 |s | < 4.
The renormalization-group 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 large-0 limit. The large-L 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)).
8.7 HeavyLight Currents
197
Fig. 8.6. Distribution functions wF (dashed line) and wA (solid line)

Hadronic matrix elements of QCD operators, such as quark currents j, can
be expanded in 1/m (see (5.57)),

1
1
,
(8.71)
j =C
j +
Bi i + O
2m
m2
to separate short-distance contributions the matching coecients
C, Bi ,
. . . , from long-distance contributions
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 short-distance contributions, also contributions from
large distances, where the perturbation theory is ill-dened. The latter contributions produce IR renormalon singularities, which lead to ambiguities
(MS /m)n in the matching coecients C, . . . Similarly, HQET matrix elements of higher-dimensional operators i , . . . contain, in addition to the
main large-distance 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 lower-dimensional

matrix elements (e.g.,
j ). These two kinds of renormalon
ambiguities have

to cancel in physical full-QCD 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
iCF
=
2(d 1)
dd k 2(d 1) + (dD2 /m2 + 4)h 2(D2 /m2 + 4)h2

.
(2)d
D1 D2n
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)
where (5.62) h = (n 2 + ). In order to obtain the renormalized matrix

os 1/2
(mv, 0), we must add NZ (, u)/2 (8.65).
element ZQ
os =
To the accuracy considered, (0, 0) = 1 and Zqos = Zqos = Z
Q
os 1/2
1. Retaining only negative powers of in ZQ
(mv, 0), we obtain
j (), according to (5.59). The corresponding anomalous dimenZj ()/Z
sion (8.14) is
s 2 + 2(n 2 )2 + (3 + 2)(1 + )
jn
+O
j = CF
12
B(2 + , 2 + ) (3 + ) (1 )

1
,
02
(8.73)
in agreement with (8.39) and (8.56). Retaining 0 terms, we obtain C ()

in the form (8.15). The corresponding renormalization-group invariant (8.16)
has the form (8.19), with [7]

(u) (1 2u)
S(u) = CF
5 u 3u2 + 2u(n 2) 2(n 2)2
(3 u)

5 2(n 2)2
.
(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).
Fig. 8.7. Virtuality distribution function for fB /fB
Now we
turn to the matrix elements of the dimension-4 HQET operators
0|4,5 |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 lower-dimensional operator
j with the same
external states. We consider a transition from an o-shell 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 kinetic-energy vertices contain no Dirac
matrices, and we may take (1/4) Tr on the light-quark 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

heavy-quark wave-function renormalization ZQ , which has a UV renormalon
ambiguity (8.55). This gives (3/4)(/m)F

as the ambiguity of the matrix
element of 4 . This must be equal to F Gk /(2m), and we obtain [12]
3
Gk () =
2
(8.78)
200
k
k + v
Fig. 8.8. Matrix element of 4
(this ambiguity is -independent at the rst order in 1/0 ).

For 5 , a straightforward calculation of the diagram in Fig. 8.9 gives the
bare matrix element of the usual form (8.11) and (8.42), with
0
N (, u) = 2d CF Cm
(8.79)
(see (6.16)). The renormalization-group invariant matrix element has the

form (8.19) with 0 = 2e5/6 and

1
(1 + 2u) (1 u)
S(u) = 2d CF Cm (0 )
+
.
(8.80)
m
(2 + u)
2u
Taking the residue at the pole u = 1/2, we nd a UV renormalon ambiguity
times the matrix element of

j. Using (6.17), we
of (d /3)Cm (0 )(/m)
obtain [12]
Gm () = 2
(8.81)
(again, -independent at this order).

k
k + v
Fig. 8.9. Matrix element of 5
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 large-0 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].
In the large-0 limit,

k = 3 ,
G
2
m =
G
201

m
2 0
;
m0
see (8.78), (8.81), (6.63). In general, these ambiguities must be equal to MS

times some numbers:

3
0m
(8.82)
0
Gk = N1 0 , Gm = N2 2
2
m0
(see 8.50). The normalization factors N1,2 are known only in the large-0
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
A similar requirement for (6.85) gives the following for = i , i 0 :

S (u) =
CF
2
(1/2 u)1+1 /(20 )
(8.85)
202

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 next-to-leading terms were derived in [8]. In the large-0
limit, the simple-pole 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
cn+1 = 2CF n! (20 )n (20 n)1 /(20 )

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)
The O(1/n) corrections were calculated in [8].

For the ratio fB /fB (6.86), the Borel image of the perturbative series is
S(u) =
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
and the asymptotics of the coecients is

2
8
cn+1 = CF n! (20 )n (20 n)1 /(20 )
3

0m
0m
m0 /(20 )
1
N0 2
N2 (20 n)
.
m0
m0
(8.90)
8.8 HeavyHeavy Currents
203

First we consider the leading matching coecients for the currents c b at
= 0 at the order 1/0 [10]. By repeating the calculation in Sect. 7.3, we
can easily calculate the one-loop coecient for the vertex function (Fig. 7.5)
with the gluon denominator raised to a power n:
a1 (n) = CF
(d 2n 1) (d/2 + n + 1)
(d n 1)
(mc /mb ) md2n1

(mb /mc )
md2n1
c
b
,
mb mc
2h2 + (d 2n 2)h d + n + 1
d1
r+
,
(r) =
d 2n 3
dn1
(8.91)
generalizing (7.30). Adding the on-shell wave-function renormalization (8.65)

for the b and c, we obtain

u
2
(1 + u) (1 2u)
F (, u) =
D()u/1 N (, u) ,
e
(8.92)
mb mc
(3 u )

N (, u) = 2CF (n 2)2 u(n 2) + 2(n 2) u

4 u2 4 2u2
R1
1 + 2u

1 2u
2
(1 u u + u)R0 ,
+ (3 2)
(8.93)
1 + 2u
where
R0 = cosh
L
,
2
R1 =
sinh [(1 2u)L] /2

,
sinh(L/2)
for the on-shell 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
(see (8.18)). There is no pole at u = 1/2, the leading IR renormalon is at

u = 1. Therefore, the IR renormalon ambiguity of the matching coecients
at = 0 is (MS /mc,b )2 /0 .
For the vector and axial currents [10],

(u) (1 2u) 1 u u2
R1 + (1 2u)R0
(3 u)
1 + 2u

L
L
3
= 6CF
coth 1
1 u + ,
2
2
2
(u) (1 2u)
S5AC (u) = 2CF
(1 + 2u) (3 u)

(3 u + u2 )R1 + (1 2u)(1 u u2 )R0

L
5
L
L coth 6 u + .
= CF 3L coth 8
2
2
2
S 0 (u) = 6CF
(8.95)
The matching coecients do not depend on , , and are given by (8.19):

L
3
s (0 ) L
H 0 = 6CF
coth 1
1 + ,
4
2
2
2

5
L
s (0 )
L
L coth 6 + .
H5AC = CF
3L coth 8
4
2
2
2
(8.96)
Re-expressing s (0 ) via s ( mb mc ), we recover (7.36).

Now we consider the general case = 0 [12]. Closely following the derivation in Sect. 7.5, we obtain Hi , which have the form (8.11), without the
leading 1 for H2,3,4 ; the functions Fi (, u) have the form (8.92), with

+1
dz
N1 (, u) = CF
1+u
1 (a+ a )

(1 z 2 )h
u(1 2u)
a+ a h2 (1 2u)
4
+ (1 + u )(2 u ) cosh

+ (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)
At one loop (u = ), the result (7.55) is reproduced. At = 0, the integrals

can be easily calculated (see (7.56)), and N1 +N2 +N3 +N4 reproduces (8.93).
J ; see (8.39)
The anomalous dimension (8.14) corresponding to (8.97) is jn
and (8.57). The functions (8.20),
Si (u) =
(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)
They do not depend on the Dirac matrix in the current. As expected,

H1 + H2 + H3 + H4 vanishes at = 0.
In matrix elements of QCD currents (7.20), these IR renormalon ambiguities in the leading matching coecients Hi must be compensated by
UV renormalon ambiguities in matrix elements of the subleading operators
i , i . There are two kinds of subleading operators local and bilocal.
First we consider local operators, whose coecients are completely xed by
reparametrization invariance (Sect. 7.6). These local operators contain either
D or D . Let us decompose these derivatives into components in the (v, v )

plane and orthogonal to this plane. The projection of D onto the longitudinal plane is
206
(v v D + v v D) cosh v v D v v D
sinh2
and similarly for D . All operators containing longitudinal derivatives can

be rewritten, using the equations of motion, as full derivatives of the leading
currents Ji . When we are interested in matrix elements from a ground-state
meson to a ground-state meson, we may make the replacement
v ) Ji .
i Ji (v
In this case, projecting onto the longitudinal plane means
iD
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 lower-dimensional matrix elements of Ji .
The above derivation is exact (and not only valid in the large-0 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 c-quark kinetic energy. This operator appears in the
expansion (7.20), with a coecient H1 . The one-loop 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
Fig. 8.10. Kinetic-energy insertions into the c-quark line
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
[k 2 2yv k 2y v k 2(y + y )]3+u

y dy dy
= 8u(2)1+2u
1+u .
[y 2 + y 2 + 2yy cosh 2(y + y )]
The substitution y = (2)(1 z)/2, y = (2)(1 + z)/2 gives

1u d (1 + z) dz
I = 2u

1+u .
cosh2 (/2) z 2 sinh2 (/2) + 1

The substitution cosh2 (/2) z 2 sinh2 (/2) = then leads to a factored
form
1u
+1
d
dz
I = 2u

2u ,
1+u
2
( + 1)
0
1
cosh (/2) z 2 sinh2 (/2)
208
where

0
1u d
(1 + 2u) (2 u)
.
=
1+u
( + 1)
(1 + u)
Collecting all contributions, we obtain (8.20)

2 (1 + 2u) (1 u)
S(u) = CF
mc
(1 + u)

1 + u(1 u) cosh
+1
dz
2u
cosh2 (/2) z 2 sinh2 (/2)
+ ,
(8.100)
where the dots mean terms regular at u = 1/2.

The residue at the pole u = 1/2 can be obtained using

+1
dz
1
3/2 = cosh + 1 .
cosh (/2) z 2 sinh (/2)
2
The corresponding UV renormalon ambiguity is given by (8.26), with 2 inc /2 (8.55),

stead of m. Adding also the external-line renormalization eect Z
we obtain

1
1
.
2 cosh + 1 2mc
The contribution of the bilocal operator containing the b-quark kinetic energy
contains mb instead of mc . Therefore, the contribution
of all bilocal operators
with kinetic-energy insertions to the ambiguity of J is [12]
1
1
2 cosh + 1
1
1
+
mc
mb

J1 .
2
(8.101)
Matrix elements of bilocal operators with a c- or b-quark chromomagnetic

insertion cannot be represented as matrix elements of the leading currents Ji
times scalar factors they require new, independent form factors. Therefore,
they have no UV renormalon ambiguities, which whould need to be equal to
times lower-dimensional
matrix elements.

i (8.98) of the QCD matrix element
J
Summing
the
ambiguity
H
i

J due to the IR renormalon in the leading matching coecients Hi at
u = 1/2, the ambiguity (8.99) due to the UV renormalon in the local subleading operators at u = 1/2, and the contribution of the bilocal subleading
operators (8.101), we see that the ambiguities cancel, at the rst order in
1/0 , for any Dirac structure of the current J [12].
References
209
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
M. Beneke: Phys. Lett. B 344, 341 (1995) 188, 194

M. Beneke: Phys. Rep. 317, 1 (1999) 175
M. Beneke, V.M. Braun: Nucl. Phys. B 426, 301 (1994) 187, 188, 192, 193
M. Beneke, V.M. Braun, N. Kivel: Phys. Lett. B 404, 315 (1997) 175
I. Bigi, M.A. Shifman, N.G. Uraltsev, A.I. Vainshtein: Phys. Rev. D 50, 2234
(1994) 188
D.J. Broadhurst: Z. Phys. C 58, 339 (1993) 179
D.J. Broadhurst, A.G. Grozin: Phys. Rev. D 52, 4082 (1995) 176, 186, 190,
192, 198
F. Campanario, A.G. Grozin, T. Mannel: Nucl. Phys. B 663, 280 (2003); Erratum: Nucl. Phys. B 670, 331 (2003) 200, 202
A.G. Grozin, M. Neubert: Nucl. Phys. B 508, 311 (1997) 190, 194, 195, 196
M. Neubert: Phys. Lett. B 341, 367 (1995) 203, 204
M. Neubert: Phys. Rev. D 51, 5924 (1995) 182, 192, 199
M. Neubert, C.T. Sachrajda: Nucl. Phys. B 438, 235 (1995) 194, 197, 198,
199, 200, 204, 205, 206, 208
A. Palanques-Mestre, P. Pascual: Commun. Math. Phys. 95, 277 (1984) 178
G. Parisi: Phys. Lett. B 76, 65 (1978) 175
Index
-function 3637, 7778, 177

5 matrix 1, 9596, 98, 100101,
106107, 111115, 126, 154155, 186,
198
analytical properties of diagrams 24,
2627
anomalous dimension 3739, 126127,
178
chromomagnetic operator 8081,
133, 135, 195196
electron eld 5658
gluon eld 3739, 4145
heavyheavy current 146151, 191,
205
heavylight current 103, 190
heavylight dimension-4 local
operators 127131
heavylight operators with chromomagnetic insertion 133134, 136,
141
heavylight operators with kineticenergy insertion 132133
heavy-quark eld 52, 5455, 187
mass 9395, 133, 135, 186, 192
quark current 9293, 101, 112, 186,
203, 205
quark eld 4647, 50, 184, 194
axial anomaly 9697
background eld 7172
BGSUV sum rule 1516
bilocal operators 121122, 125,
131135
Bjorken sum rule 1314
BlochNordsieck model 55, 148149
Borel transform 180
chromomagnetic
form factor 7476

interaction 3, 78, 6263, 76, 7981,
8384, 194196
colour factors 48
coordinate space 24, 26, 47, 52, 5556,
145150, 190191
Cutkosky rules 24, 2627
Cvitanovic algorithm 48
decay constant 89, 116118, 139,
143144, 198199, 202
decoupling 8485
chromomagnetic coecient 89
coupling constant 8789
gluon eld 8586
heavylight current 107
heavy-quark eld 87
mass 105106
quark current 103105
quark eld 8687
scalar current 105106
vector current 105
dimensional regularization 1, 23
Dirac form factor 7476
divergences
infrared 24, 26, 64, 7071, 74, 76, 80,
104, 108, 130, 153, 182, 184185, 187,
192
ultraviolet 24, 26, 64, 7071, 7778,
80, 9293, 97, 103, 129130, 132, 146,
153154, 156, 159, 165, 182, 184185,
187188, 192, 206207
Euclidean space 23, 63, 69, 191
evanescent operators 95
exponentiation theorem 5556,
148150
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
self-energy 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
S-matrix elements 59, 80, 194195
dimension-4 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
heavy-quark eld 7071, 190
mass 46, 6769, 9395, 116, 128,
192194, 198
on-shell 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
self-energy
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
vacuum diagram 23, 6970

vertex
electronphoton 56
heavyheavy current 145146,
152155, 161165, 190191, 203205
heavylight current 102103, 190,
197198
HQET 7679, 8788
light-quark current 9293, 104105,
185186
quarkgluon 7274
staticheavy current 158159,
167168
virtuality distribution function
182183, 185, 192193, 196, 199
Voloshin sum rule 1516
Ward identity 56, 7374, 76, 7879,
93, 100, 105, 124, 137139, 147148,
154155, 186
Wick rotation 23, 63
Yennie gauge
52, 58

160 Physics with Tau Leptons
By A. Stahl 2000. 236 gs. VIII, 315 pages
161 Semiclassical Theory of Mesoscopic Quantum Systems
By K. Richter 2000. 50 gs. IX, 221 pages
162 Electroweak Precision Tests at LEP
By W. Hollik and G. Duckeck 2000. 60 gs. VIII, 161 pages
163 Symmetries in Intermediate and High Energy Physics
Ed. by A. Faessler, T.S. Kosmas, and G.K. Leontaris 2000. 96 gs. XVI, 316 pages
164 Pattern Formation in Granular Materials
By G.H. Ristow 2000. 83 gs. XIII, 161 pages
165 Path Integral Quantization and Stochastic Quantization
By M. Masujima 2000. 0 gs. XII, 282 pages
166 Probing the Quantum Vacuum
Pertubative Effective Action Approach in Quantum Electrodynamics and its Application
By W. Dittrich and H. Gies 2000. 16 gs. XI, 241 pages
167 Photoelectric Properties and Applications of Low-Mobility Semiconductors
By R. Knenkamp 2000. 57 gs. VIII, 100 pages
168 Deep Inelastic Positron-Proton Scattering in the High-Momentum-Transfer Regime of HERA
By U.F. Katz 2000. 96 gs. VIII, 237 pages
169 Semiconductor Cavity Quantum Electrodynamics
By Y. Yamamoto, T. Tassone, H. Cao 2000. 67 gs. VIII, 154 pages
170 dd Excitations in Transition-Metal Oxides
A Spin-Polarized Electron Energy-Loss Spectroscopy (SPEELS) Study
By B. Fromme 2001. 53 gs. XII, 143 pages
171 High-Tc Superconductors for Magnet and Energy Technology
By B. R. Lehndorff 2001. 139 gs. XII, 209 pages
172 Dissipative Quantum Chaos and Decoherence
By D. Braun 2001. 22 gs. XI, 132 pages
173 Quantum Information
An Introduction to Basic Theoretical Concepts and Experiments
By G. Alber, T. Beth, M. Horodecki, P. Horodecki, R. Horodecki, M. Rtteler, H. Weinfurter,
R. Werner, and A. Zeilinger 2001. 60 gs. XI, 216 pages
174 Superconductor/Semiconductor Junctions
By Thomas Schpers 2001. 91 gs. IX, 145 pages
175 Ion-Induced Electron Emission from Crystalline Solids
By Hiroshi Kudo 2002. 85 gs. IX, 161 pages
176 Infrared Spectroscopy of Molecular Clusters
An Introduction to Intermolecular Forces
By Martina Havenith 2002. 33 gs. VIII, 120 pages
177 Applied Asymptotic Expansions in Momenta and Masses
By Vladimir A. Smirnov 2002. 52 gs. IX, 263 pages
178 Capillary Surfaces
Shape Stability Dynamics, in Particular Under Weightlessness
By Dieter Langbein 2002. 182 gs. XVIII, 364 pages
179 Anomalous X-ray Scattering
for Materials Characterization
Atomic-Scale Structure Determination
By Yoshio Waseda 2002. 132 gs. XIV, 214 pages
180 Coverings of Discrete Quasiperiodic Sets
Theory and Applications to Quasicrystals
Edited by P. Kramer and Z. Papadopolos 2002. 128 gs., XIV, 274 pages
181 Emulsion Science

Basic Principles. An Overview
By J. Bibette, F. Leal-Calderon, V. Schmitt, and P. Poulin 2002. 50 gs., IX, 140 pages
182 Transmission Electron Microscopy of Semiconductor Nanostructures
An Analysis of Composition and Strain State
By A. Rosenauer 2003. 136 gs., XII, 238 pages
183 Transverse Patterns in Nonlinear Optical Resonators
By K. Staliunas, V. J. Sanchez-Morcillo 2003. 132 gs., XII, 226 pages
184 Statistical Physics and Economics
Concepts, Tools and Applications
By M. Schulz 2003. 54 gs., XII, 244 pages
185 Electronic Defect States in Alkali Halides
Effects of Interaction with Molecular Ions
By V. Dierolf 2003. 80 gs., XII, 196 pages
186 Electron-Beam Interactions with Solids
Application of the Monte Carlo Method to Electron Scattering Problems
By M. Dapor 2003. 27 gs., X, 110 pages
187 High-Field Transport in Semiconductor Superlattices
By K. Leo 2003. 164 gs.,XIV, 240 pages
188 Transverse Pattern Formation in Photorefractive Optics
By C. Denz, M. Schwab, and C. Weilnau 2003. 143 gs., XVIII, 331 pages
189 Spatio-Temporal Dynamics and Quantum Fluctuations in Semiconductor Lasers
By O. Hess, E. Gehrig 2003. 91 gs., XIV, 232 pages
190 Neutrino Mass
Edited by G. Altarelli, K. Winter 2003. 118 gs., XII, 248 pages
191 Spin-orbit Coupling Effects in Two-dimensional Electron and Hole Systems
By R. Winkler 2003. 66 gs., XII, 224 pages
192 Electronic Quantum Transport in Mesoscopic Semiconductor Structures
By T. Ihn 2003. 90 gs., XII, 280 pages
193 Spinning Particles Semiclassics and Spectral Statistics
By S. Keppeler 2003. 15 gs., X, 190 pages
194 Light Emitting Silicon for Microphotonics
By S. Ossicini, L. Pavesi, and F. Priolo 2003. 206 gs., XII, 284 pages
195 Uncovering CP Violation
Experimental Clarication in the Neutral K Meson and B Meson Systems
By K. Kleinknecht 2003. 67 gs., XII, 144 pages
196 Ising-type Antiferromagnets
Model Systems in Statistical Physics and in the Magnetism of Exchange Bias
By C. Binek 2003. 52 gs., X, 120 pages
197 Electroweak Processes in External Electromagnetic Fields
By A. Kuznetsov and N. Mikheev 2003. 24 gs., XII, 136 pages
198 Electroweak Symmetry Breaking
The Bottom-Up Approach
By W. Kilian 2003. 25 gs., X, 128 pages
199 X-Ray Diffuse Scattering from Self-Organized Mesoscopic Semiconductor Structures
By M. Schmidbauer 2003. 102 gs., X, 204 pages
200 Compton Scattering
Investigating the Structure of the Nucleon with Real Photons
By F. Wissmann 2003. 68 gs., VIII, 142 pages
201 Heavy Quark Effective Theory
By A. Grozin 2004. 72 gs., X, 213 pages

Vol201 (2004) Heavy Quark Effective Theory (Andrey Grozin) PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Vol201 (2004) Heavy Quark Effective Theory (Andrey Grozin) PDF

Uploaded by

Copyright:

Available Formats

Springer Tracts in Modern Physics

Springer Tracts in Modern Physics

Solid-State Physics, Editors

Institut fur Theoretische Teilchenphysik

Elementary Particle Physics, Editors

Department of Physics and Astronomy

Institut fur Theoretische Teilchenphysik

Institut fur Theorie der Kondensierten Materie

Fundamental Astrophysics, Editor

Complex Systems, Editor

Cataloging-in-Publication Data applied for

Physics and Astronomy Classication Scheme (PACS):

ISSN print edition: 0081-3869

Chapter 1 gives an overview of basic experimental facts about mesons

Novosibirsk, February 2004

Hadrons with a Heavy Quark . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

The HQET Lagrangian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

The HQET Lagrangian: 1/m Corrections . . . . . . . . . . . . . . . . .

HeavyLight Currents: 1/m Corrections . . . . . . . . . . . . . . . . . .

Our notation mainly follows [2]. Covariant and contravariant components of

We shall use dimensional regularization for calculating loop diagrams: the

(tc )ab = if acb .

The quadratic Casimir operators in the fundamental (quark) representation

(see Sect. 3.3).

1 Hadrons with a Heavy Quark

The B meson is the hydrogen atom of quantum chromodynamics (QCD),

A. Grozin: Heavy Quark Eective Theory, STMP 201, 317 (2004)

1 Hadrons with a Heavy Quark

1.1 Mesons and Baryons with a Heavy Quark

1.1 Mesons and Baryons with a Heavy Quark

light elds, which is conserved up to 1/m corrections; therefore, the mixing

Fig. 1.1. Mesons containing c

In the leading approximation, the spectrum of b-containing

1 Hadrons with a Heavy Quark

Fig. 1.2. Mesons containing b

heavy-quark spin switched on, these baryons have sP = (1/2)+ . In the 1+

Fig. 1.3. Baryons containing a c quark

1.1 Mesons and Baryons with a Heavy Quark

1 Hadrons with a Heavy Quark

Experimentally, this ratio is 0.89. This is a spectacular conrmation of the

In the limit mc , D and D are degenerate, and so are D1 and D2 .

1.2 Semileptonic Decays

Let us now discuss the B-meson leptonic decay constant fB . It is dened

0|b 5 u|B(p) = ifB p ,

B(p )|B(p)nr = (2)3 (p p)

should be used instead. Then, for the B meson at rest,

from the + and + decays. The branching B(B+ + ) should be of

1.2 Semileptonic Decays

1 Hadrons with a Heavy Quark

c, and W are scalar. We concentrate our attention on decays of B

(1/2)+  J (1/2)+ = (cosh )

bc weak currents are proportional to the IsgurWise form factor (cosh ),

1.2 Semileptonic Decays

with trivial kinematic coecients. When the current J replaces an innitely

Fig. 1.4. The IsgurWise form factor

The B B form factor is also proportional to (cosh ). In this case,

1 Hadrons with a Heavy Quark

Fig. 1.5. The complex cosh plane

They are the fractions of the number of B Xc W+ decays with Xc velocity

1.2 Semileptonic Decays

(cosh + 1)2 (cosh 1)  2

Its dependence is evident from dimensionality, and the dependence is given

Fig. 1.6. A qualitative sketch of the B Xc decay structure function

(1/2)+ J (1/2)+ = (cosh )

with trivial kinematic coecients. When the current J replaces an innitely

(cosh + 1)2 (cosh 1) 2