SPRINGER BRIEFS IN PHYSICS

Cesareo A. Dominguez

Quantum
Chromodynamics
Sum Rules
SpringerBriefs in Physics

Series editors
B. Ananthanarayan, Centre for High Energy Physics, Indian Institute of Science,
Bangalore, India
Egor Babaev, Amherst, MA, USA
Malcolm Bremer, Bristol, UK
Xavier Calmet, Department of Physics and Astronomy, University of Sussex,
Brighton, UK
Francesca Di Lodovico, London, UK
Pablo D. Esquinazi, Institute for Experimental Physics II, University of Leipzig,
Leipzig, Germany
Maarten Hoogerland, Auckland, New Zealand
Eric Le Ru, Kelburn, Wellington, New Zealand
Hans-Joachim Lewerenz, Pasadena, CA, USA
James Overduin, Towson, MD, USA
Vesselin Petkov, Montreal, QC, Canada
Charles H.-T. Wang, Department of Physics, The University of Aberdeen,
Aberdeen, UK
Andrew Whitaker, Belfast, UK
Stefan Theisen, Max-Planck-Institut für Gravitationsphys, Potsdam, Germany
More information about this series at [Link]
Cesareo A. Dominguez

Quantum Chromodynamics
Sum Rules

123
Cesareo A. Dominguez
Centre for Theoretical and Mathematical
Physics, Department of Physics
University of Cape Town
Cape Town, South Africa

ISSN 2191-5423 ISSN 2191-5431 (electronic)

SpringerBriefs in Physics
ISBN 978-3-319-97721-8 ISBN 978-3-319-97722-5 (eBook)
[Link]

Library of Congress Control Number: 2018949890

© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2018

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made. The publisher remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
To Pavel Baikov, Konstantin Chetyrkin,
and Johann Kühn
for obtaining higher order QCD results
allowing for precision determinations from
QCD Sum Rules
Preface

This book is intended for readers with a good knowledge of quantum field theory,
in general, and quantum chromodynamics (QCD), in particular. It is addressed to
readers planning to start research in QCD in the framework of sum rules. Currently,
there are two major approaches to obtain information in QCD, i.e. lattice QCD
(LQCD) and QCD sum rules (QCDSR). The latter is the subject matter of this book.
It deals with the current state-of-the-art formulation of QCDSR in the complex
squared energy plane, called finite energy sum rules (FESR). This allows for a
relation between QCD and hadronic physics following from Cauchy’s residue
theorem in that plane. As a result, current FESR determinations of a plethora of
QCD and hadronic parameters rival in precision with those from LQCD. This
healthy competition is extremely beneficial for our understanding of the strong
interactions at the most elementary level.
This book is not a review of past work on QCDSR. The pioneering formulation
of QCDSR in the framework of integral transforms, e.g. Laplace and Hilbert, while
having played a fundamental role in the development of the subject, is currently no
match for the precision achieved from FESR. In addition, and most importantly, the
FESR parameter related to quark-gluon deconfinement (at finite temperature) has
recently been shown to be related to the Polyakov loop of LQCD. This brings these
two approaches into a beneficial partnership.
The topics discussed in this book concern mostly QCD at zero temperature.
A last chapter on finite temperature QCDSR has been kept short, as there is a recent
comprehensive review on this subject. The extension of QCDSR to include
hadronic/QCD matter in the presence of very strong magnetic fields is not covered
here. This new research direction is currently in a state of flux, so the reader is
advised to consult the literature.

Cape Town, South Africa Cesareo A. Dominguez

vii
Acknowledgements

The author wishes to thank his QCD sum rule collaborators: Alejandro Ayala, Jose
Bordes, Pietro Colangelo, Marcelo Loewe, Giuseppe Nardulliy , Nasrallah
Nasrallah, Nello Paver, Jose Peñarrocha, Eduardo de Rafael, J. Cristobal Rojas,
Karl Schilcher, Joan Sola, Hubert Spiesberger, and Cristian Villavicencio.
Special thanks are due to Marcelo Loewe for reading the manuscript, and to
Alexes Mes and Jed Stephens for providing Eq. (3.31).
This work was supported in part by the University of Cape Town, (South
Africa), and by the Alexander von Humboldt Foundation (Germany).

ix
Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2 Operator Product Expansion in QCD . . . . . . . . . . . . . . . . . . . . . . . 9
Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3 Renormalization Group Equation . . . . . . . . . . . . . . . . . . . . . . . . . . 13
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4 Integration in the Complex s-Plane . . . . . . . . . . . . . . . . . . . . . . . . . 21
5 Determination of the QCD Strong Coupling . . . . . . . . . . . . . . . . . . 25
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
6 Hadronic Spectral Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
7 QCD Chiral Sum Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
8 Determination of the Gluon Condensate . . . . . . . . . . . . . . . . . . . . . 45
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
9 Quark Masses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
10 Corrections to the GMOR Relation . . . . . . . . . . . . . . . . . . . . . . . . 69
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
11 Anomalous Magnetic Moment of the Muon . . . . . . . . . . . . . . . . . . 73
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
12 QCD Sum Rules at Finite Temperature . . . . . . . . . . . . . . . . . . . . . 81
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

xi
xii Contents

13 Summary and Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
Appendix A: Dressed Propagators and Selected Integrals . . . . . . . . . . . . 91
Appendix B: Current Correlators in QCD . . . . . . . . . . . . . . . . . . . . . . . . 95
Appendix C: Light-Quark Pseudoscalar Current Correlator
in QCD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
Appendix D: QCD Integrals of Light-Quark Pseudoscalar Current
Correlator in QCD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
Appendix E: QCD Thermal Space-like Spectral Functions . . . . . . . . . . . 111
Chapter 1
Introduction

The theory of Quantum Chromodynamics (QCD), i.e. its Lagrangian together with
its main features, was first proposed by Harald Fritzsch and Murray Gell-Mann in
1972 [1], almost a year before the discovery of asymptotic freedom [2] in QCD. The
Lagrangian is

1 i μν
LQCD = i ψ̄a (x) γμ ∂ μ ψa (x) − m0 ψ̄a (x) ψa (x) − Fμν (x)Fi (x)
4
− g G iμ (x) ψ̄a (x) γ μ λiab ψb (x) , (1.1)

where a = 1, 2, Nc = 3 is the SU(3)-colour index, i = 1, 2, ...8 ≡ Nc2 − 1, ψ(x) are

the quark fields, G iμ is the gluon field, λi the SU(3) Gell-Mann matrices, and Fμν
i

the gluon tensor

i
Fμν (x) ≡ ∂μ G iν (x) − ∂ν G iμ (x) − gfijk G jμ (x) G kν (x) , (1.2)

where the fijk are proportional to the commutator of the SU(3) Gell-Mann matrices,
[λi , λj ] = 2 i fijk λk .
In addition to the exact SU (3)c colour symmetry, with Nc2 − 1 massless gluons,
there is a non-trivial underlying symmetry hierarchy, incorporating several concepts
from Current Algebra, a pre-QCD attempt to understand strong interactions [3].
To begin with, a label must be introduced to differentiate the quark fields
according to flavour, i.e. ψ A (x), where A = up, down,...etc., and the colour label
is to be understood. Considering the light-quark sector, up-, down-, and strange-
quarks, there are two different types of symmetries in QCD, a flavour symmetry
and a chiral symmetry, depending on whether it manifests itself in the states (irre-
ducible representations of the symmetry group) or not. The former is a classifica-
tion symmetry, the latter a dynamical one with rich consequences. Starting with the
flavour symmetry in the limit mu = md = ms = 0, i.e. SU (3)f , it is realized in the
© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2018 1
C. A. Dominguez, Quantum Chromodynamics Sum Rules,
SpringerBriefs in Physics, [Link]
2 1 Introduction

Wigner-Weyl mode. This means that physical states (hadrons) are classified accord-
ing to the irreducible representations of the group, with the vacuum sharing the
Lagrangian symmetry. The first symmetry-breaking step is to make ms = 0, with
mu = md = 0, thus breaking SU (3)f , but preserving SU (2)f . The next step involves
mu = md = 0, which still respects SU (2)f , as the divergence of the vector current
still vanishes, i.e. ∂ μ Vμ (x) ∝ (md − mu ) = 0. In the final step mu = md = 0 breaks
SU (2)f down to U(1). This symmetry (eightfold-way [4]) allows for the classification
of hadronic states into multiples, leads to mass formulas, and led to the prediction
of the − baryon of mass 1686 MeV, discovered in 1964 [5].
Turning to chiral symmetries, an axial-vector current must be considered together
with the vector current, i.e. Aμ (x) = ψ̄(x)γ5 γμ ψ(x). This current cannot generate
a group by itself, as the commutator of two axial-vector currents transforms as a
vector current. Hence, the chiral symmetry group becomes SU (3) × SU (3), with
∂ μ Aμ (x) ∝ (ms + mud ), and mud = mu + md . The question is how is this chiral sym-
metry realized. A Wigner-Weyl realization would imply, among other things, the
existence of quasi-degenerate parity doublets not seen in the spectrum. The alterna-
tive, a Nambu-Goldstone (NG) realization of chiral symmetry, implies a massless
NG-boson, plus a plethora of dynamical relations among hadronic quantities. The
NG-boson is identified with the electrically charged kaon, after SU (3) × SU (3)
breaking, and with the charged pion after the breaking of SU (2) × SU (2).
One of the major consequences of a NG realization of chiral symmetry is the
Goldberger-Treiman relation (GTR) [6, 7]
√
2 fπ gπNN = (Mp + Mn ) gA , (1.3)

where fπ = 92.28 ± 0.07 MeV [8] is the pion decay constant

√
0|Aμ (0)|π(p) = i 2 fπ pμ , (1.4)

gπNN  13 is the strong coupling, and gA  1.3 the weak beta decay constant. The
GTR is currently satisfied at the 1% level. Alternatives to this realization of chiral
symmetry and pattern of its breaking, proposed in the past, have not survived stringent
tests from π − −π scattering [9].
Another crucial consequence of chiral symmetry is the Gell-Mann–Oakes–Renner
relation (GMOR), first obtained in the framework of Current Algebra [10], prior to
QCD, and now understood as a consequence of the NG realization of chiral symmetry.
An outline of its derivation is important at this stage, as it requires the introduction
of the concept of current correlators, the fundamental objects in the QCD sum rule
(QCDSR) programme.
The starting point is the concept of a current correlator in momentum space,
ψ5 (q2 ), defined in QCD as the Fourier transform

ψ5 (s ≡ −q2 ) = i d 4 x eiqx < 0| T (j5 (x) j5 (0)) |0 > , (1.5)
1 Introduction 3

where < 0| is the physical vacuum and the current density j5 (x) is chosen as

j5 (x) = (md + mu ) : d (x) i γ5 u(x) : , (1.6)

with mu,d the light quark masses, and a convenient change in notation for the quark
fields. Notice that the current densities, j5 (x), are actually the QCD axial-vector
current divergences

∂ μ Aμ (x)|QCD = (mu + md ) : d̄ (x) i γ5 u(x) : . (1.7)

Current correlators have also an alternative representation in terms of hadronic

degrees of freedom, obtained by using hadronic, instead of QCD fields. For instance,
in this particular case of ψ5 (s), the lowest hadronic state is the pion, so that differ-
entiating Eq. (1.4) gives
√
∂ μ Aμ (x)|HAD = 2 fπ Mπ2 φπ (x) , (1.8)

where φπ (x) is the pion field, and hadronic excited states of the pion can be neglected
at this stage.
The pseudoscalar current correlator, Eq. (1.5), satisfies the following low energy
theorem, first obtained in the SU (2) × SU (2) Current Algebra framework [10], valid
in QCD at leading order in the quark masses

ψ5 (0) = −(mu + md ) 0|ū u + d̄ d |0 + O(m2u,d ) , (1.9)

where the quark masses and fields depend on the renormalization scale, but their
product, as above, does not!
An important consequence of the NG realization of SU (2) × SU (2) is that in the
symmetry limit the pion mass squared vanishes as the (light) quark mass

Mπ2 = B mq , (1.10)

and the pion decay constant squared is proportional to the quark-condensate, which
only vanishes at finite temperature

1
fπ2 = − q̄ q . (1.11)
B
These two relations have important consequences in phenomenology. In partic-
ular, in the extension of the QCDSR programme to finite temperature, as well as
strong magnetic fields.
So far one has two different representations of the same object, the current cor-
relator within QCD, and in the hadronic sector. The essential question is how to
relate them. The answer, the first pillar of QCDSR, is provided by considering the
complex squared energy s-plane, and invoking Cauchy theorem, as first proposed in
4 1 Introduction

Fig. 1.1 Integration contour

in the complex squared Im(s)
energy s-plane. The
discontinuity across the real
axis brings in the hadronic
spectral function, while
integration around the circle
involves the QCD correlator.
The radius of the circle is s0 ,
the onset of QCD

Re(s)

[11] (see also [12]). In QCD there are no singularities in this plane, except on the
real positive s-axis, where poles correspond to stable hadrons, and on the second
Riemann sheet corresponding to resonances of certain width , with  ∝ 1/τ , and
τ their lifetime. Next, an integration contour in this plane, a circle of radius |s0 |, is
considered as in Fig. 1.1. While QCD is not valid on the real axis, it is expected to
hold everywhere else on the circle, provided the radius |s0 | is large enough. Cauchy
theorem then relates the Physics on the real axis to that on the circle. This approach,
named quark-hadron duality, leads to QCD Finite Energy Sum Rules (FESR). Since
QCD is not valid on the real axis, it has been customary to multiply spectral functions
by kernels vanishing on the real axis (pinched kernels) [13, 14], in order to quench
this contribution. This procedure is usually quite satisfactory.
In detail, if (s) is some meromorphic correlation function in the complex s-plane,
and P(s) a meromorphic integration kernel, Cauchy theorem states
 
1
(s) P(s) ds = Ri , (1.12)
2πi C(|s0 |) i

where Ri are the residues at the poles of (s)P(s). Splitting the integration range
into the circle and the real axis gives the FESR
  s0 
1 1
ds (s)QCD P(s) + ds Im (s)HAD P(s) = Ri . (1.13)
2π i C(|s0 |) sth π i
1 Introduction 5

The low energy theorem, Eq. (1.9), can be used together with this FESR, for
(s) ≡ ψ5 (s)/s, and the hadronic spectral function, Eq. (1.8), to obtain the Gell-
Mann-Oakes-Renner (GMOR) relation [10]

ψ5 (0) ≡ − (mu + md ) 0|ūu + d̄ d |0 = 2 fπ2 Mπ2 (1 − δπ ) , (1.14)

valid to leading order in the quark masses and in the hadronic spectral function, and
δπ encapsulates higher order hadronic corrections. These are at the level of 7% in
SU (2) × SU (2) [15], and 50% in SU (3) × SU (3) [16].
The GMOR relation, Eq. (1.14), is unique in the sense that an expression for
the quark condensate is obtained from first principles. Other vacuum condensates,
entering the Operator Product Expansion of current correlators at short distances, do
not share this feature. They must be determined from the FESR themselves, or from
Lattice QCD (LQCD).
In the framework of FESR the crucial issues are the value of the radius, |s0 |, its
impact on predicted quantities, and the so-called duality violations due to the fact that
QCD is not valid on the positive real axis. Regarding |s0 |, its uncertainty impacts on
results at a very reasonable level, and combines favourably with other uncertainties
in QCD and hadronic parameters. Concerning duality violations, these are unknown
by definition [17, 18], thus requiring specific models. However, it is expected that
the use of pinched kernels [13, 14], together with large enough radii in the complex
squared energy plane, quenches substantially their importance.
Concerning the contour integral around the circle of radius s0 in Eq. (1.13), there
are two different procedures to perform this integration [19]. Both make use of the
Renormalization Group Improvement (RGI) of the perturbative expansion of the
QCD correlator. In the first one, called Fixed Order Perturbation Theory [20], the
contour integral is performed followed by the RGI. The second procedure, Contour
Improved Perturbation Theory (CIPT) [21], requires RGI to be performed first, fol-
lowed by the contour integration. It is to be noticed that in FOPT the strong coupling
is frozen at s = |s0 |, while in CIPT αs (s) is to be integrated around the circle. The lat-
ter is usually performed by solving numerically the renormalization group equation
for αs (s) at each point on the circle.
Historically, QCDSR were first developed in the framework of the Laplace trans-
form [22], involving an ad-hoc parameter needed for dimensional reasons in the
Laplace exponential kernel. This free parameter comes in addition to the threshold
for perturbative QCD, s0 , needed as a cut-off in the integrals. With s0 having a clear
physical interpretation, it is unfortunate for it to be exponentially suppressed in this
approach. A serious drawback of this technique is that at and beyond next-to-next
to leading order in perturbative QCD the Laplace transform leads to a Volterra-type
function μ(x, β) [23], a fact discovered only when higher orders in perturbation the-
ory were first determined [24]. This turns higher order perturbative corrections into
a cumbersome exercise (see e.g. [25]).
In any case, a large number of applications of these sum rules were developed over
the years, as reviewed in [26], leading to results affected by uncontrollable systematic
6 1 Introduction

uncertainties. For a critical detailed discussion of these sum rules see [27]. As high
precision was eventually required in some instances, e.g. quark mass determinations,
this method has been superseded by the FESR technique.

References

1. H. Fritzsch, M. Gell-Mann, in Proceedings of the XVI International Conference on High

Energy Physics, Chicago, vol. 2, eds by J.D. Jackson, A. Roberts (1972), pp. 135.
arXiv:hep-ph/0208010
2. H.D. Politzer, Phys. Rev. Lett. 30, 1346 (1973) (Phys. Rep. 14 C, 274 (1974); D. Gross, F.
Wilczek, Phys. Rev. Lett. 30, 1343 (1973))
3. S.B. Treiman, R. Jackiw, B. Zumino, E. Witten, Current Algebra and Anomalies, World Sci-
entific, Singapore (1985). (See also: J. Bernstein, Elementary Particles and their Currents (W.
H. Freeman & Co., San Francisco, 1968)
4. The “Eightfold Way” was never published in a journal. For a historical account see e.g. G.
Zweig, in Proceedings of the Conference in Honour of Murray Gell-Mann’s 80th Birthday, eds
by H. Fritzsch, K.K. Phua (World Scientific, Singapore, 2011)
5. V.E. Barnes et al., Phys. Rev. Lett. 12, 204 (1964)
6. H. Pagels, Phys. Rep. 16, 219 (1975)
7. C.A. Dominguez, Phys. Rev. D 7, 1252 (1973). (ibid. D 16, 2320 (1977); Riv. Nuovo Cim.
8N6, 1 (1985))
8. K.G. Patrignani et al., Particle Data Group. Chin. Phys. C 40, 100001 (2016)
9. J. Stern, “Light quark masses and condensates in QCD”, in Mainz 1997, Chiral Dynamics:
Theory and Experiment, eds. A. Bernstein, D. Drechsler and Th. Walcher [hep-ph/9712438];
J. Stern, in Chiral Dynamics: Theory and experiment III, eds. by A.M. Bernstein, J.L. Goity,
and U. Meissner. World Scientific, Singapore, 2001
10. M. Gell-Mann, R.J. Oakes, B. Renner, Phys. Rev. 175, 2195 (1968)
11. R. Shankar, Phys. Rev. D 15, 755 (1977)
12. A. Bramon, E. Etim, M. Greco, Phys. Lett. B 41, 609 (1972). (M. Greco, Nucl. Phys. B 63,
398 (1973); E. Etim, M. Greco, Lett. Nuovo Cim. 12, 91 (1975))
13. K. Maltman, Phys. Lett. B 440, 367 (1998)
14. C.A. Dominguez, K. Schilcher, Phys. Lett. B 581, 193 (2004)
15. J. Bordes, C.A. Dominguez, P. Moodley, J. Penarrocha, K. Schilcher, J. High Energy Phys.
1005, 064 (2010)
16. J. Bordes, C.A. Dominguez, P. Moodley, J. Penarrocha, K. Schilcher, J. High Energy Phys.
1210, 102 (2012)
17. M. Gonzalez-Alonso, A. Pich, J. Prades, Phys. Rev. D 81, 074007 (2010)
18. A. Pich, A. Rodriguez-Sanchez, Mod. Phys. Lett. A 31, 1630032 (2016)
19. M. Jamin, J. High Energy Phys. 09, 058 (2005)
20. A.A. Pivovarov, Z. Phys. C 53, 461 (1992)
21. F. LeDiberder, A. Pich, Phys. Lett. B 286, 147 (1992)
22. M.A. Shifman, A.I. Vainshtein, V.I. Zakharov, Nucl. Phys. B 147, 385 (1979). (ibid., B 147,
448 (1979))
23. A. Erdelyi (ed.), Higher Trascendental Functions, vol. 3 (McGraw-Hill Book Company, Inc.
New York, Toronto, London, 1955)
24. E. de Rafael, in Marseille 1981. In: J.W. Dash (ed.), Theoretical Aspects of Quantum Chromo-
dynamics, Centre de Physique Theorique, Marseille, France, CPT-81/P.1345, 259 (1981)
25. K.G. Chetyrkin, C.A. Dominguez, D. Pirjol, K. Schilcher, Phys. Rev. D 51, 5090 (1995). (K.G.
Chetyrkin, D. Pirjol, K. Schilcher, Phys. Lett. B 404, 337 (1997); C.A. Dominguez, L. Pirovano,
K. Schilcher. Phys. Lett. B 425, 193 (1998))
References 7

26. P. Colangelo, A. Khodjamirian, in At the Frontier of Particle Physics, vol. 3, ed. by M. Shifman
(World Scientific, Singapore 2001), pp. 1495–1576
27. C. A. Dominguez, Analytical determination of QCD quark masses, in Fifty Years of Quarks,
eds. by H. Fritzsch, M. Gell-Mann (World Scientific Publishing Co., Singapore 2015), pp.
287–313
Chapter 2
Operator Product Expansion in QCD

The second of the two pillars of the QCDSR method is the Operator Product Expan-
sion (OPE) of current correlators at short distances, beyond perturbation theory. This
allows for a current correlator to be written in terms of QCD degrees of freedom,
starting with its perturbative expansion, followed by non-perturbative terms. The
latter involve vacuum expectation values of quark and gluon fields, reflecting colour
confinement. An example is that of the quark condensate, discussed in the previous
section. Given some QCD correlation function

(q 2 )| QC D = i d 4 x eiq x < 0| T (J † (x) J (0)) |0 > , (2.1)

where J (x) is some local current built from the QCD fields, the OPE is given by


(q 2 )| QC D = C0 (q 2 , μ2 ) Iˆ + C2N +2 (q 2 , μ2 ) 0| Ô2N +2 (μ2 )|0 , (2.2)
N =0

where μ2 is some renormalization scale and the Wilson coefficients, C2N +2 , depend
on the Lorentz indices and quantum numbers of J (x) and of the local gauge invariant
operators Ô N built from the quark and gluon fields. These operators are ordered by
increasing dimensionality, and the Wilson coefficients, calculable in PQCD, fall off
by corresponding powers of −q 2 . Hence, the OPE implies a factorization of short
distance effects, encapsulated in the Wilson coefficients, and long distance dynamics
from the vacuum condensates. The term C0 Iˆ stands for the purely perturbative
contribution, currently known for some correlators up to order O(αs6 ), with αs ≡
αs (q 2 ) the strong coupling. Figure 2.1 illustrates this contribution to next-to-leading
order. In QCD there are no gauge-invariant operators of dimension d = 2, other than
light-quark mass terms, O(m q2 ), usually negligible.

© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2018 9

C. A. Dominguez, Quantum Chromodynamics Sum Rules,
SpringerBriefs in Physics, [Link]
10 2 Operator Product Expansion in QCD

x y

x y x y

Fig. 2.1 Schematic representation of the purely perturbative QCD term, C0 Iˆ in Eq. (2.2), with a
quark loop and with one-gluon exchange

Fig. 2.2 Schematic representation of the QCD quark-condensate, dimension d = 3 term in the
OPE, 0|q̄(0) q(0)|0, originating from on-shell light quarks (zero-momentum) interacting with the
QCD vacuum. Large momentum flows through the bottom propagator

Next, there is no analytic, first-principles approach to determine the vacuum con-

densates in the OPE, Eq. (2.2), except for the quark condensate. This is known from
the GMOR relation discussed in Chap. 1. In order to see the emergence of the vacuum
condensates let us consider the quark propagator

i i
S F ( p) = =⇒ , (2.3)
 p−m  p − m + ( p 2 )

where ( p 2 ) encapsulates a correction due to confinement, which is not calculable

analytically from first principles. Nevertheless, this correction is expected to peak at
the quark mass-shell. For light quarks this would be p  0, leading to these quarks
condensing into a condensate, 0|q̄(0) q(0)|0, as shown in Fig. 2.2
In the case of the gluon propagator
i i
D F (k) = =⇒ , (2.4)
k2 k 2 + (k 2 )

the correction term, (k 2 ), peaks at k  0, as the gluon is massless. Confinement

information is then parametrized by 0|αs G  μν |0, the gluon condensate, as
 μν · G
shown in Fig. 2.3.
2 Operator Product Expansion in QCD 11

Fig. 2.3 Schematic

representation of the QCD
gluon-condensate, d = 4
term, in the OPE. Large
momentum flows through
the quark propagators

Higher dimensional condensates can be constructed from the quark and gluon
fields in the QCD Lagrangian, but their values are affected by very large uncertainties.
Due to this limitation one normally considers FESR up to dimension d = 4, thus
avoiding these higher-dimensional condensates.
In the heavy-quark sector, i.e. charm and bottom, the heavy-quark condensate,
while present, is unrelated to symmetry. Instead, it is related to the gluon condensate
through [1]
  1  αs 2   1
Q̄ Q = − G +O 2 . (2.5)
12 m Q π mQ

Reference

1. M. A. Shifman, A. I. Vainshtein, V. I. Zakharov, Nucl. Phys. B 147, 385 (1979). (ibid., B 147,
448 (1979))
Chapter 3
Renormalization Group Equation

The momentum dependence of the QCD coupling and of the quark masses is needed
in most applications of QCDSR. The QCD strong coupling, αs ≡ g 2 /4π, with g in
Eq.(1.1), satisfies the renormalization group equation

d
s as (s) = β[as (s)] , (3.1)
ds
where as ≡ αs /π, and the beta function is

β[as (s)] = − b N as (s) N +2 . (3.2)
N =0

The expansion coefficients, b N , are [1–3]

 
1 2
b0 = 11 − n f . (3.3)
4 3
 
1 38
b1 = 102 − n f . (3.4)
16 3
 
1 2857 5033 325 2
b2 = − nf + n . (3.5)
64 2 18 54 f
  
1 149753 1078361 6508
b3 = 4 + 3564 ζ(3) − + ζ(3) n f
4 6 162 27
  
50065 6472 1093 3
+ ζ(3) n 2f + n . (3.6)
162 81 729 f

© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2018 13

C. A. Dominguez, Quantum Chromodynamics Sum Rules,
SpringerBriefs in Physics, [Link]
14 3 Renormalization Group Equation

1 8157455 621885 88209
b4 = + ζ(3) − ζ(4) − 288090 ζ(5)
45 16 2 2
 
336460813 4811164 33935 1358995
+nf − − ζ(3) + ζ(4) + ζ(5)
1944 81 6 27
 
25960913 698531 10526 381760
+ n 2f + ζ(3) − ζ(4) − ζ(5)
1944 81 9 81
 
630559 48722 1618 460
+ n 3f − − ζ(3) + ζ(4) + ζ(5)
5832 243 27 9
 
1205 152
+ n 4f − ζ(3) . (3.7)
2916 81

The values of the Riemann theta-function, ζ(n), are ζ(2) = π6 , ζ(3) =

2

1.2020569032, ζ(4) = π90 , ζ(5) = 1.0369277551, ζ(6) = 945 π6

4
, ζ(7) = 1.0083492774,
π8
ζ(8) = 9450 .
It is important to notice that there is an alternative definition of the coefficients
b N in the literature, i.e. b̄ N = − 21 b N +1 . Another important observation is that the
coefficients bn are usually referred to as βn in the literature. This could lead to
confusion as the symbol β would appear on both sides of Eq. (3.2).
A very useful way to solve Eq. (3.1) is by Taylor-series developing as (s) about
some reference scale s = s ∗ , at which as (s ∗ ) is well known. For instance, as is known
with high precision at the tau-lepton scale [4]

αs (Mτ2 ) = 0.328 ± 0.013 . (3.8)

The expansion of as (s) in terms of as (s ∗ ) reads [5]

as (s) = as (s ∗ ) + [as (s ∗ )]2 (−b0 η) + [as (s ∗ )]3 (−b1 η + b02 η 2 )

 5 
+ [as (s ∗ )]4 − b2 η + b0 b1 η 2 − b03 η 3
2
 3 13 
+ [as (s ∗ )]5 − b3 η + b12 η 2 + 3b0 b2 η 2 − b02 b1 η 3 + b04 η 4
 2 3
∗ 6 7 7 35
+ [as (s )] −b4 η + b0 b1 η + b0 b3 η − b0 b12 η 3 − 6b02 b2 η 3
2 2
2 2 6

77 3 4
+ b b1 η − b05 η 5 , (3.9)
12 0

where η ≡ ln(s/s ∗ ). This procedure avoids the use of the scale QCD in the expression
3 Renormalization Group Equation 15

1
αs (s) ∝ ··· , (3.10)
ln [s/2QC D ]

as  QC D is affected by a large uncertainty [6]. Equation (3.9) can be used e.g. to

obtain as at the light-quark sector scale using its very precise value at the Z-boson
scale, as (s ∗ = M Z2 ).
Next, the renormalization group equation for the quark masses, m q (t), with t =
ln(−q 2 /μ2 ), is given by

1 d 
m q (t) = γ[as (t)] , = −as (t) γ N as (t) N , (3.11)
m q (t) dt N =0

where the coefficients γi are [1–3, 7, 8]

γ0 = 1 (3.12)

 
1 202 20
γ1 = − nf (3.13)
16 3 9
   
1 2216 160 140 2
γ2 = 1249 − + ζ3 n f − n (3.14)
64 27 3 81 f

  91723
1 4603055 135680
γ3 = + ζ(3) − 8800 ζ(5) + −
256 162 27 27
34192 18400   5242 800
− ζ(3) + 880 ζ(4) + ζ(5) n F + + ζ(3)
9 9 243 9
  332 64  
160
− ζ(4) n 2F + − + ζ(3) n 3F . (3.15)
3 243 27
16 3 Renormalization Group Equation

1 99512327 46402466 698126
γ4 = 5 + ζ(3) + 96800 ζ(3)2 − ζ(4)
4 162 243 9
231757160 150736283
−− ζ(5) + 242000 ζ(6) + 412720 ζ(7) + n f −
243 1458
12538016 75680 2038742 49876180
− ζ(3) − ζ(3)2 + ζ(4) + ζ(5)
81 9 27 243
638000 1820000 1320742 2010824
− ζ(6) − ζ(7) + n 2f + ζ(3)
9 27 729 243
46400 166300 264040 92000
+ ζ(3)2 − ζ(4) − ζ(5) + ζ(6)
27 27 81 27
91865 12848 448 5120
+ n 3f + ζ(3) + ζ(4) − ζ(5)
1458 81 9 27

260 320 64
+ n 4f − − ζ(3) + ζ(4) . (3.16)
243 243 27

In order to obtain an expansion of the quark mass in terms of inverse powers of

L ≡ ln(μ2 /− q 2 ), needed in applications of FESR, one has to integrate Eq. (3.11)
after first solving Eq. (3.1). The power series expansion of the latter is

1 b1 ln L 1 b2   b
2
as (μ2 ) = 1− 2 + 2 2 12 ln2 L − ln L − 1 +
β0 L b0 L b0 L b0 b0
1 b3  5 1
+ 3 3 13 − ln3 L + ln2 L + 2 ln L −
b0 L b0 2 2
  
b1 b2 1 b3 1
− 3 2 ln L + +O 4 . (3.17)
b0 2 b0 L

Integrating Eq. (3.11) one finds

γ0 /β0
m q (μ2 ) = m̃ as (μ2 ) exp c1 as (μ2 ) + c2 as2 (μ2 ) + c3 as3 (μ2 ) + · · · , (3.18)

where m̃ is an integration constant, and the coefficients ci are

γ1 γ0 b1
c1 = − 2 , (3.19)
b0 b0

1 γ2 γ1 b1 γ0  b2 
c2 = − 2 + 2 1 − b2 , (3.20)
2 b0 b0 b0 b0
3 Renormalization Group Equation 17

1 γ0 b1 b2 b1  b12 
c3 = 2
− − b2 − b3
3 b0 b0 b0 b0
  
γ1 b1 2
b1 .γ2 γ3
+ 2 − b2 − + . (3.21)
b0 b0 b02 b0

γ0 /β0
The next step is to expand the overall factor in Eq. (3.18), as (μ2 ) . For this
purpose one invokes the expansion

λ(λ − 1) 2 λ(λ − 1)(λ − 2) 3

(1 + x)λ = 1 + λ x + x + x + ··· , (3.22)
2! 3!
which leads to
 1 γ0 /b0  d1 d2 d3 
[as (μ2 )]γ0 /b0 = 1+ + 2 + 3 ··· , (3.23)
b0 L L L L

where
γ0 b1
d1 = − ln L , (3.24)
b03

γ0 b12  2  b
2 γ0 b12  γ0 
d2 = ln L − ln L − 1 + + − 1 ln2 L , (3.25)
b03 b02 b0 5
2 b0 b0

γ0 b13  5 1 b1 b2 1 b3
d3 = − ln3 L + ln2 L + 2 ln L − − 3 2 ln L +
b04 b03 2 2 b0 2 b0
γ0 b1  γ0  b2   b
2
− 5 −1 1
ln 2
L − ln L − 1 + ln L
b0 b0 b02 b0
1 γ0 b13  γ0  γ
0

− 7
−1 − 2 ln3 L . (3.26)
6 b0 b0 b0

The integration constant m̃ is now redefined as

m̃ → m̂ (2 b0 )γ0 /b0 (3.27)

where m̂ is the so-called invariant mass. The quark mass m q (μ2 ) then becomes
18 3 Renormalization Group Equation
 c 1
mˆq 1 c1 e1  c2  1
m q (μ ) =  γ0 /b0
2
1+ + d1 + + c2 + 1 2
1 b0 L b0 2 b0
2
L
1
d1 c1 c1 e2 (c2 + c12 /2)  b1 
+ d2 + + + d1 − 2 ln L
L2
b0 b0 b02 b02
 c13  1 e1 c1 1  1 
+ c3 + c1 c2 + + d1 c1 + d2 + d3 3 + O 4 , (3.28)
6 b03 b0 b0 L L

where the coefficients ci are given in Eqs. (3.19)–(3.21), the di in Eqs. (3.24)–(3.26),
and
b1
e1 = − ln L , (3.29)
b02

1 b12  2  b2
e2 = 2 2
ln L − ln L − 1 + . (3.30)
b0 b0 b0

It should be noticed that the coefficient γ4 is now known

 (see Eq. (3.16)). The
reader is encouraged to determine the last term of order O L14 in Eq. (3.28).
Finally, in analogy with Eq. (3.9) the running quark mass can also be expressed
in terms of its value at some scale s = s ∗ according to


1
m̄(s) = m̄(s ∗ ) 1 − a(s ∗ ) γ0 η + a 2 (s ∗ ) η − 2 γ1 + γ0 (β0 + γ0 ) η
2

1 3 ∗  
− a (s ) η 6 γ2 − 3 β1 γ0 + 2 (β0 + γ0 ) γ1 η + γ0 (2 β02 + 3 β0 γ0 + γ02 ) η 2
6

1 4 ∗
+ a (s ) η − 24 γ3 + 12(β2 γ0 + 2β1 γ1 + γ12 + 3 β0 γ2 + 2 γ0 γ2 ) η
24
 
− 4 6 β02 γ1 + 3 γ02 (β1 + γ1 ) + β0 γ0 (5 β1 + 9 γ1 ) η 2 + γ0 (6 β03 + 11 β02 γ0

+ 6 β0 γ02 + γ03 ) η 3
3 Renormalization Group Equation 19

1 5 ∗ 1 
+ a (s ) η − 120 γ4 + 60 − 7 β1 β2 γ0 + 4 β02 γ3 + β0 (7 β1 γ0 + β3 γ0
120 β0
 
+ 2 β2 γ1 + 3 β1 γ2 + 2γ1 γ2 + 2 γ0 γ3 ) η − 20 3β12 γ0 + β1 (14 β0 + 9 γ0 ) γ1
 
+ 3 (2β0 + γ0 )(β2 γ0 + γ12 + 2 β0 γ2 + γ0 γ2 ) η 2 + 10 12 β03 γ1 + γ03 (3 β1 + 2 γ1 )
 
+ β0 γ02 (13 β1 + 12 γ1 ) + β02 γ0 (13 β1 + 22 γ1 ) η 3 − γ0 24 β04 + 50 β03 γ0

 
∗
+ 35 β02 γ02 + 10 β0 γ03 + γ04 η 4
+ O(a (s ))
6
(3.31)

where η ≡ ln(s/s ∗ ), as ≡ αs /π, and like Eq. (3.9) this expression makes no use of
 QC D .

References

1. P.A. Baikov, K.G. Chetyrkin, J.H. Kühn, Phys. Rev. Lett. 118, 082002 (2017)
2. T. van Ritbergen, J.A.M. Vermaseren, S.A. Larin, Phys. Lett. B 400, 379 (1997)
3. M. Czakon, Nucl. Phys. B 710, 485 (2005)
4. A. Pich, Eur. Phys. J. Web Conf. 137, 01016 (2017)
5. M. Davier, A. Höcker, Z. Zhang, Rev. Mod. Phys. 78, 1043 (2006)
6. K.G. Patrignani et al., Particle Data Group. Chin. Phys. C 40, 100001 (2016)
7. P.A. Baikov, K.G. Chetyrkin, J.H. Kühn, J. High Energy Phys. 10, 076 (2014)
8. T. Luthe, A. Maier, P. Marquard, Y. Schróder, J. High Energy Phys. 1701, 081 (2017)
Chapter 4
Integration in the Complex s-Plane

In the QCD FESR, Eq. 1.13, the integral around the circle of radius |s0 | can be per-
formed in two ways, named Fixed Order Perturbation Theory (FOPT), and Contour
Improved Perturbation Theory (CIPT). In FOPT the strong coupling is frozen on the
circle, i.e. at s = |s0 |, and the renormalization group (RG) improvement, (μ2 = q 2 ),
in all logarithms is performed after integration. In CIPT the RG improvement is per-
formed before integration, and αs (s) is a complex function of s, to be determined by
solving the RG equation at each point on the circle. This is done through a single-step
numerical contour integration, using as input the strong coupling obtained by solving
numerically the RG equation for αs (−s). This technique achieves a partial resum-
mation of higher order logarithmic integrals, and improves the convergence of the
perturbative QCD series. In correlators involving quark masses, e.g. ψ5 (s), Eq. (1.5),
CIPT requires the running quark mass also to be integrated around the circle, after
computing it at each step by solving the corresponding RG equation.
In general, results from both integration procedures depend on the particular
correlation function. In some cases CIPT proves superior to FOPT, but in others
there is no major difference. Hence, it has been customary to use both methods in
applications.
There are some QCD correlation functions that are either asymptotically constant
or that diverge. This would imply subtracted dispersion relations. Alternatively, one
could not consider the correlator itself, but rather its (convergent) derivative(s). This
situation affects the contour integral in the FESR, Eq. (1.13), if one uses CIPT. Hence,
instead of employing the original correlator one can take sufficient derivatives of it.
For instance, the pseudoscalar current correlator ψ5 (q 2 ), Eq. (1.5), diverges asymp-
totically as −q 2 . Hence one needs to consider its second derivative. In this case there
are some useful identities that are required in the FESR, to wit.
One starts from the identity
 
dψ5 (s)
ds g(s) ψ5 (s) = − ds [G(s) − G(s0 )] , (4.1)
|s0 | |s0 | ds
© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2018 21
C. A. Dominguez, Quantum Chromodynamics Sum Rules,
SpringerBriefs in Physics, [Link]
22 4 Integration in the Complex s-Plane

where  s
G(s) = ds  g(s  ) , (4.2)
0

which follows trivially after substituting G(s), Eq. (4.2), into Eq. (4.1). The analytic
function g(s) is some given integration kernel, usually present in the hadronic sec-
tor to quench/enhance contributions to FESR. Repeating this procedure leads to a

formula involving ψ5 (s) ≡ d 2 ψ5 (s)/ds 2 , which is well behaved at infinity
 
d 2 ψ5 (s)
ds g(s) ψ5 (s) = ds [F(s) − F(s0 )] , (4.3)
|s0 | |s0 | ds 2

where
 s  s  s  s0 
  
F(s) = ds [G(s ) − G(s0 )] = ds ds" g(s") − ds" g(s") .
0 0 0 0
(4.4)
Proceeding to the (counter-clockwise) integration in the complex s-plane, the
starting point is the definition

s = −s0 eiφ = s0 ei(φ+π) ≡ s0 ei α , (4.5)

so that
ds = i s0 ei α dα. (4.6)

Typical integrals in the pseudoscalar correlator case are


1 ds N  αs (s0 eiφ )  M
≡I Na ,M (s0 ) s
2πi |s0 | s π
 0   π α M
i αs M
dφ ei N φ dφ ei N φ
s
= (−s0 ) N +
2πi −π π 0 π
 π  α (s eiφ )  M
1
dφ ei N φ
s 0
= (−s0 ) N Re , (4.7)
π 0 π
 0  α (s eiφ )  M  π  α (s eiφ )  M
1 s 0 1 s 0
I N ,M (s0 ) ≡
b
dφ + dφ ,
2πi −π π 2πi 0 π
 π  α (s eiφ )  M
1 s 0
= Re dφ , (4.8)
π 0 π

where the angular integration is performed counter-close-wise, and the last step in
Eqs. (4.7)–(4.8) follows from the requirement that the integrals be real.
Depending on the current correlator, not only the strong coupling, but also the
running quark mass needs to be integrated around the circle on the complex s-plane.
Rearranging Eq. (3.11) one finds
4 Integration in the Complex s-Plane 23

dm q γ(as )
= das (4.9)
mq β(as )

which is the equation to be integrated

  m(q)   as (q)
m(q)
dm γ(as )
= ln = das . (4.10)
m(μ) m m(μ) as (μ) β(as )

If one were to expand the ratio

γ(as )
R(as ) ≡ , (4.11)
β(as )

one looses information on two orders in as . For instance, to keep terms of order
O(as3 ) one needs information on γ5 and β5 , currently unknown. Hence this is not a
good idea. The way out is rather simple, to wit. Starting from the RGE, Eq. 4.9, and
substituting
das (x) = i β(as ) d x , (4.12)

gives
  x 
m q (x) = m q (x0 ) exp i γ[as (x  )] d x  , (4.13)
x0

which will make use of full information up to O(as4 ). In practice, using CIPT one
needs to integrate numerically Eq. (4.13), with x0 = 0 and m q (x0 ) = m q (s0 ) so that
  x 
 
m q (x) = m q (s0 ) ex p − i dx γ J [as (x )] J
. (4.14)
0 J

To complete this section let us consider the numerical integration of the strong
coupling and the quark mass in the complex s-plane. The relation between s and
the angle φ ≡ x ∈ (−π, π) is given in Eq. (4.5), so that the equation to be solved
numerically is
das (x)   N +2
= −i b N as (x) , (4.15)
dx N =0


with the input as (x = 0) = as s0 ei(x+π) |x=0 = as (−s0 ). This differential equation
is easily solved using Euler’s method, to wit.
Given a function f (x, y), with y = y(x), and dy(x)/d x = f (x, y), with initial
condition y(0), then
 h h 
y(i+1) = yi + h f xi + , yi + f (xi , yi ) , (4.16)
2 2
24 4 Integration in the Complex s-Plane

which converges after a few steps. As an example, consider the function

dy(x)
= −2 x y 2 (x) ≡ f (x, y) , (4.17)
dx

with x1 = 0, and y(0) = y(1) = 1. The exact analytic solution is

1
y(x) = . (4.18)
1 + x2

Choosing the point x = 1, gives y(1) = 0.5. Proceeding to iterate Eq. (4.16) with
h = 0.2 and
y1 = 1 (input), (4.19)

gives
h h 
y2 = 1 + 0.2 f , 1+ ∗ 0 = 1 + 0.2 f (0.1, 1) = 0.96 , (4.20)
2 2
 h h 
y3 = 0.96 + 0.2 f 0.2 + , 0.96 + f (0.2, 0.96) = 0.8577 , (4.21)
2 2
· · · y5 = 0.498 . (4.22)

In the case in which the function depends only on y,

dy(x)
= f (y) , (4.23)
dx
the solution is the iteration of
 h 
y(i+1) = yi + h f yi + f (yi ) . (4.24)
2
Chapter 5
Determination of the QCD Strong
Coupling

The idea that hadronic decays of the τ -lepton could provide an ideal laboratory for
studying hadronic weak currents at low and intermediate energy was first proposed
in [1]. In fact, in the light-quark sector the QCD strong coupling can now be deter-
mined with high accuracy from experimental data on τ -lepton decay [2, 3]. After
extrapolation of this value to the Z-boson mass there is excellent agreement with
independent determinations in that region.
The ratio R in τ -decay is defined as

(τ − → ντ hadrons)
Rτ = (5.1)
(τ − → ντ e− ν̄e )

which can be written in terms of QCD vector and axial-vector current correlators as
 m 2τ
ds  s 2  s  
Rτ = 12 π S E W 2
1− 2 1 + 2 2 Im(1) (s) + Im(0) (s) , (5.2)
0 mτ mτ mτ

where S E W = 1.0201 ± 0.0003 is the electro-weak correction [4], and

  
(J ) (s) = |Vuq |2 (J )
uq,V (s) + (J )
uq,A (s) , (5.3)
q=d,s

where Vuq is the Cabibbo-Kobayashi-Maskawa matrix element, J = 0, 1, and

(J ) (J )
uq,V (s) and uq,A (s) are the vector and axial-vector light-quark current correlators


i d 4 xeiq x 0|T (Jμ (x)Jν (0)) = (−gμν q 2 + qμ qν )(1) 2 (0)
V,A (q ) + qμ qν V,A (q ) ,
2

(5.4)
where a simplified notation is implicit, and Jμ (x) stands for the vector, Vμ (x), or
the axial-vector, Aμ (x), currents. In the sequel we concentrate on the non-strange
current correlators, in which case there is no longitudinal term in the vector channel,

© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2018 25

C. A. Dominguez, Quantum Chromodynamics Sum Rules,
SpringerBriefs in Physics, [Link]
26 5 Determination of the QCD Strong Coupling

i.e. (0) (0) 2

V (q ) is absent in Eq. (5.4), while  A (q ) corresponds to the pion pole. In
2

perturbative QCD (PQCD), and for massless quarks, the vector correlator is identical
to the axial-vector and is given by
 1 
−4π 2 V,A (q 2 ) = L + L as + as2 L k2 − L 2 b0
2
  
3 1 3 2 2 1
+ as L b0 + L k3 − L b1 + b0 k2
3 2
 1 1 3 
+ as4 L k4 − L 4 b03 − L 2 b2 + b0 k3 + b1 k2
4 2 2
1 3 
+ L b0 (5 b1 + 6 b0 k2 )
6
 1 3 
+ as5 L k5 − L 2 b3 + 2 b0 k4 + b1 k3 + b2 k2
2 2
 7 1 
+ L 2 k3 b0 + k2 b0 b1 + b2 b0 + b12
3 2
3 2
1 4 2 1 5 4 
− L b0 13 b1 + 12 b0 k2 + L b0 , (5.5)
12 5

where as ≡ αs /π, q 2 < 0, L ≡ ln(−q 2 /μ2 ), bi are given in Eqs. (3.3)–(3.7), k1 =

1, k2 = 1.63982, k3 = 6.37101, and k4 = 49.076.
An alternative procedure involves the Adler function, D(q 2 ) rather than V,A (q 2 )
itself [2, 3]. The Adler function, D(q 2 ), is defined as

d
D(q 2 ) = −q 2 V,A (q 2 ) . (5.6)
dq 2

There is a renormalization group constraint involving D(q 2 )

d
μ2 D[ln(−q 2 /μ2 ), as (μ2 )] = 0 , (5.7)
dμ2

where q 2 < 0. For details on this approach see [2, 3].

Turning to Eq. (5.2), the term proportional to s in Rτ , Eq. (5.2) vanishes.
Hence, there is no contribution from the dimension-four gluon condensate, and Rτ is
essentially given by perturbative QCD, and the strong coupling as becomes a function
of Rτ .
5 Determination of the QCD Strong Coupling 27

The latest determination of αs (Mτ ) in this framework is [3]

αs (Mτ2 ) = 0.328 ± 0.013 , (5.8)

which gives αs (M Z2 ) = 0.1197 ± 0.0015, in good agreement with the electroweak

precision fit [4] αs (M Z2 ) = 0.1196 ± 0.0030, and the world average value αs (M Z2 ) =
0.1181 ± 0.0011 [5].
It is unfortunate that the τ -lepton is not heavier, as it would have allowed for precise
information on the onset of QCD, as well as on potential duality violations. As it
stands, there is only a hint of PQCD starting at, and near, the end-point s  Mτ2 .
However, it has been shown [6] that it is possible to use a suitable QCD-FESR
to effectively extend the kinematical range beyond the end-point of the data up to
s  10 GeV2 . By construction, this FESR suppresses the hadronic contribution along
the real s-axis in the complex s-plane, beyond the kinematical end-point of τ -decay,
s1 = Mτ2 , where there is no longer experimental data.
The starting point is the axial-vector current correlator

i d 4 xeiq x 0|T (Aμ (x)Aν (0))|0 = (−gμν q 2 + qμ qν )(1) (0) 2
A (q ) + qμ qν  A (q ) ,
2

(5.9)
where Aμ (x) = d̄(x)γμ γ5 u(x) is the light-quark axial-vector current. Next, one
invokes the general FESR, Eq. (1.13), which after using the OPE, Eq. (2.2), it becomes
 s0
1
(−) C2N +2 O2N +2  = 4 π
N 2
Im(s)| H AD − s0N +1 M2N +2 (s0 ) ,
ds s N
0 π
(5.10)
where the dimensionless perturbative QCD (PQCD) moments, M2N +2 (s0 ), are

4 π2 s0
1
M2N +2 (s0 ) = ds s N I m (s)| P QC D . (5.11)
s0(N +1) 0 π

For N = 0, C2 O2  = 0 (no condensate of dimension d = 2), and

Im (s)| H AD = 2 f π2 , (5.12)

the FESR Eq. (5.10) gives f π as a function of s0 , shown in Fig. 5.1 for αs (Mτ2 ) =
0.335. Except possibly near the end-point, the prediction is not good. In fact, it is
well known that the Weinberg sum rules, involving the difference of the vector and
the axial-vector correlators are not well saturated by the ALEPH data, unless one
introduces pinched kernels in the FESR [7, 8].
In view of this situation it was proposed in [6] to introduce a polynomial inte-
gration kernel in the FESR, Eq. (5.10), tuned to eliminate the (unknown) hadronic
contribution in the interval s1 − s0 , where s1  Mτ2 , and s0 »Mτ2 . The optimal inte-
gration kernel turns out to be the simple linear function
28 5 Determination of the QCD Strong Coupling

Fig. 5.1 Results for f π from the FESR Eq. (5.10) as a function of s0 . Straight line is the experimental
value of f π

Fig. 5.2 Results for F(s0 ) ≡ f π2 from the FESR Eq. (5.15) in the axial-vector channel as a function
of s0 . Straight line is the experimental value of f π

2s
P(s) = 1 − , (5.13)
s0 + s1

with the requirement  s0

C × P(s) ds = 0 , (5.14)
s1

where C is a constant. With this choice the FESR, Eq. (5.10), becomes
5 Determination of the QCD Strong Coupling 29

Fig. 5.3 Results for F(s0 ) = 0 from the FESR Eq. (5.15) in the vector channel as a function of s0 .
Straight line is the left-hand-side of Eq. (5.15)

 s1 
1 s0 2s0
F(s0 ) = − ds P(s) I m(s)| D AT A + 2 M2 (s0 ) − M4 (s0 )
0 π 4π s 0 + s1

1 2
+ 2 C2 O2  + C4 O4  + (s0 ) , (5.15)
4π s0 + s1

where F(s0 ) ≡ 2 f π2 in the axial-vector channel, and F(s0 ) = 0 in the vector channel.
The pion pole has been separated from the data, and the chiral limit is understood.
The term (s0 ) represents the uncertainty due to assuming the unknown data to be
constant in the interval s1 − s0 . If the onset of PQCD would be at s = s1 then the data
would follow the logarithmic fall-off of PQCD. A detailed discussion of the caveats
of this procedure is given in the original paper [6].
In Fig. 5.2 we show the results for f π from the FESR, Eq. (5.15), to be compared
with Fig. 5.1. The same procedure can be followed in the vector current channel,
where now F(s0 ) = 0, as there is no (hadronically) stable scalar analogue of the
pion. The result for F(s0 ) is shown in Fig. 5.3, exhibiting very good agreement in a
wide range of s0 . In the vector channel there is independent data from the process
e+ e− annihilation into hadrons. This allows for an estimate of the uncertainty (s0 )
in Eq. (5.15). The result is (s0 )|V  (10−5 − 10−4 ) GeV2 , i.e. 1–2 orders of magni-
tude smaller than F(s0 ) (see Fig. 5.3), thus supporting the method. For more details
see [6].

References

1. K. Schilcher, M.D. Tran, Phys. Rev. D 29, 570 (1984)

2. M. Davier, A. Höcker, Z. Zhang, Rev. Mod. Phys. 78, 1043 (2006)
3. A. Pich, A. Rodriguez-Sanchez, Phys. Rev. D 94, 034027 (2016)
30 5 Determination of the QCD Strong Coupling

4. W.J. Marciano, A. Sirlin, Phys. Rev. Lett. 61, 1815 (1988); E. Braaten, C.-S. Li, Phys. Rev. D
42, 3888 (1990); J. Erler, Rev. Mex. Fis. 50, 200 (2004)
5. K.G. Patrignani et al., Particle data group. Chinese Phys. C 40, 100001 (2016)
6. C.A. Dominguez, N.F. Nasrallah, K. Schilcher, Phys. Rev. D 80, 054014 (2009)
7. K. Maltman, Phys. Lett. B 440, 367 (1998)
8. C.A. Dominguez, K. Schilcher, Phys. Lett. B 581, 193 (2004)
Chapter 6
Hadronic Spectral Functions

Chapters 2–4 have dealt with the main issues related to the QCD correlator entering
the contour integral in the FESR, Eq. (1.13). The next step is to consider the line
integral along the real s-axis. In the interval s = sth − s0 , the line integral requires
hadronic information. This could be experimental data, a parametrization of the
spectral function in terms of hadronic degrees of freedom, or some combination of
both. Regarding the data, the main sources are e+ e− → hadrons, determining the
vector spectral function in the various energy regions, and hadronic decays of the
τ -lepton, determining both the vector and axial-vector spectral functions below the
kinematical end-point s = Mτ2 .
Starting with the pseudoscalar correlator in the up- down-quark sector, ψ5 (q 2 ),
defined in Eq. (1.15), the hadronic spectral function involves the pion pole term
followed by the resonance contribution, i.e.
1 1
Im ψ5 (s) = 2 f π2 Mπ4 δ(s − Mπ2 ) + Im ψ5 (s)|RES , (6.1)
π π
where Im ψ5 (s)|RES , involves the radial excitations of the pion, π(1300) and
π(1800), with masses and widths known from experiment [1]. However, the res-
onance spectral function itself is unknown experimentally, thus leading to a seri-
ous systematic uncertainty. For instance, non-resonant background and resonance
interference, are sources of this uncertainty. In an attempt to reduce this systematic
uncertainty, it was first proposed in [2] to normalize this spectral function at threshold
using the chiral perturbation theory (CHPT) constraint [3]

1 1 Mπ4 1
I m ψ5 (s)|πππ = θ(s) s . (6.2)
π 3 f π2 28 π 4

With this threshold normalization the spectral function involving two pionic radial
excitations becomes

© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2018 31

C. A. Dominguez, Quantum Chromodynamics Sum Rules,
SpringerBriefs in Physics, [Link]
32 6 Hadronic Spectral Functions

1 [BW1 (s) + κ BW2 (s)]

I m ψ5 (s)|RES = I m ψ5 (s)|πππ , (6.3)
π (1 + κ)

where BW1 (0) = BW2 (0) = 1, with

Mi2 (Mi2 + i2 )

BWi (s) = (i = 1, 2) , (6.4)
(s − Mi2 )2 + Mi2 i2

and κ is a free parameter controlling the relative weight of the resonances.

A threshold expression beyond the chiral limit was obtained in [4], corrected for
misprints in [5]

1 1 Mπ4 1
I m ψ5 (s)|πππ = θ(s) I P S (s) . (6.5)
π 9 f π2 28 π 4

where the phase-space integral I P S (s) is given by

 √  
( s−Mπ )2
4Mπ2 1/2 1 1
I P S (s) = du λ (1, u/s, Mπ2 /s) 5 +
1−
2
4Mπ u 2 (s − Mπ2 )2
  2 
4 Mπ
× (s − 3u + 3Mπ2 )2 + 3 λ(s, u, Mπ2 ) 1 − + 20 Mπ4
u
1  
+ 3(u − M 2
π ) − s + 9M 2
π , (6.6)
(s − Mπ2 )

where

 √ 2  √ 2
u + Mπ u − Mπ
λ(1, u/s, Mπ2 /s) ≡ 1− 1− , (6.7)
s s
 √  √ 
2 2
λ(s, u, Mπ2 ) ≡ s − u + Mπ s − u − Mπ . (6.8)

In the chiral limit the phase space integral I P S (s) reduces to the simple expression

lim I P S (s) = 3 s , (6.9)

Mπ2 →0

leading to Eq. (6.2), which is an excellent approximation.

This procedure has been extended to the SU (3) × SU (3) case [6], relevant to the
current correlator, Eq. (1.15), with

j5 (x) = (m s + m ud ) : s(x) i γ5 u(x) : , (6.10)

6 Hadronic Spectral Functions 33

with m s the strange-quark mass, and m ud = (m u + m d )/2. The result is [6]

1 M4 3 I (s)
I m ψ5 (s)| K ππ = θ(s − M K2 ) K2 8 4 , (6.11)
π 2 f π 2 π s(M K2 − s)2

where

  
s
du (s + M K2 ) 
I (s) = (u − M K )(s − u) (M K2 − s) u −
2
M K2 u 2
1 2 3
− (u − M K4 ) (s − u) + (u − M K2 )2 |FK ∗ (u)|2 , (6.12)
8u 4

and

(M K2 ∗ − M K2 )2 + M K2 ∗  2K ∗
|FK ∗ (u)|2 = , (6.13)
(M K2 ∗ − u)2 + M K2 ∗  2K ∗

where K ∗ ≡ K ∗ (892) is the strange vector K π resonance. Notice that the version of
Eq. (6.11) in [6] has misprints.
In order to reduce the model dependency of the hadronic resonance parametriza-
tion it has been proposed [8–11] to choose an integration kernel in the FESR,
Eq. (1.13), such that it vanishes at the peak of each resonance

5 (s) = 1 − a0 s − a1 s 2 , (6.14)

so that 5 (M12 ) = 5 (M22 ) = 0. which fixes a0 and a1 . This simple kernel does
achieve a substantial reduction of the systematic uncertainty arising in the hadronic
sector, as discussed later in Chap. 9. Ultimately, the optimal choice of integration
kernel is most likely application dependent.
Other important hadronic spectral functions needed in applications of QCD sum
rules are the axial-vector and the vector spectral functions in the light-quark sector.
Starting with the axial-vector case, the spectral function was defined in Eq. (5.4).
Concentrating on the longitudinal term, (0) A (q ), the time ordered product of the
2

axial-vector currents can be written as the Fourier transform


d3 p
0|T (Aμ (x)A†ν (0))|0 = e−i p·x 0|Aμ (0)|π( p)π( p)|A†ν (0)|0 ,
2 p0 (2π)3
(6.15)
where
√
0|Aμ (x)|π( p)|x=0 = i 2 f π pμ , (6.16)
34 6 Hadronic Spectral Functions

with f π = 92.28 ± 0.07 MeV the (charged) pion decay constant [1]. Substituting
Eqs. (6.16) into (6.15), followed by the substitution of Eqs. (6.15) into (5.4) gives

 
d3 p
μνAA
= 2 f π2 i d 4 x eiq x e−i p·x μν ( p)
2 p0 (2π)3

2 f π2
= 2 f π2 d 4 xeiq x μν (x) = 2 f π2 μν (q 2 ) = qμ qν , (6.17)
q − Mπ2 + i
2

which after confronting with Eq. (5.4) leads to

2 f π2
(0)
A (q ) =
2
, (6.18)
q 2 − Mπ2 + i

and
1
Im (0)
A (q ) = 2 f π δ(q − Mπ ).
2 2 2 2
(6.19)
π
Continuing with the axial-vector channel, the other correlation function is the
transverse (1)
A (q ), involving the axial-vector three-pion resonance a1 (1260), of
2

mass Ma1 = 1230 ± 40 MeV, and broad width  = 250 − 600 MeV [1]. The main
decay mode is into ρπ, followed by the decay of the rho-meson into two pions. Given
the width of the a1 a Breit-Wigner parametrization is out of the question. Instead a fit
to the ALEPH data in the axial-vector channel [7], from threshold up to s = 2.0 GeV2
is given as [12]

⎡  2 ⎤
1 s − Ma21
Im A (s)|a1 = C f a1 exp ⎣− ⎦
π a21

(0 ≤ s ≤ 1.2 GeV2 ) , (6.20)

⎡  2 ⎤
1 1.2 GeV2 − Ma21
Im A (s)|a1 = C f a1 exp ⎣− ⎦
π a21

(1.2 GeV2 ≤ s ≤ 1.45 GeV2 ) , (6.21)

6 Hadronic Spectral Functions 35
⎡  2 ⎤
1 s − Ma21
Im A (s)|a1 = C f a1 exp ⎣− ⎦
π a21

(1.45 GeV2 ≤ s ≤ Mτ2 ) , (6.22)

where Ma1 = 1.0891 GeV, a1 = 568.78 GeV, C = 0.662 and f a1 = 0.073. This
spectral function is shown in Fig. 6.1 together with the ALEPH data [7].
Turning to the vector channel, (1) V (q ) in Eq. (5.4) with Jμ (x) ≡ Vμ (x) the
2

hadronic representation is obtained as follows. The time ordered product in the

hadronic representation, due to the ρ-meson, is

 
    
0|T (Vμ (x)Vν† (0))|0 | H AD = e−i p·x 0|Vμ (x)|ρ( p, s) ρ( p, s)|Vν† (0)|0 ,
s p
(6.23)
where a shortened notation is to be understood, and

  Mρ2   Mρ2
0|Vμ (0)|ρ( p, s) = μ , ρ( p, s)|Vν† (0)|0 = ν , (6.24)
fρ fρ

where f ρ = 4.97 ± 0.07 from the leptonic decay of the rho-meson [1]. Recalling
 pμ pν
μ ( p, s) ν ( p, s) = −gμν + ≡ μν ( p), (6.25)
s
Mρ2

one has for the vector correlator

Fig. 6.1 Solid curve is the axial-vector (a1 -resonance) spectral function, Eqs. (6.20)–(6.22), fitted
to the ALEPH data [7], shown with error bars the size of the data points. Dotted line is background,
to be ignored here
36 6 Hadronic Spectral Functions

 
Mρ4 d3 p
μν
V
(q 2 ) =i 4
d xe iq·x
e−i p·x μν ( p) (6.26)
f ρ2 2 p0 (2π)3

leading to

μν
V
(q 2 ) = (qμ qν − q 2 gμν ) (1)
V (q )
2
(6.27)

where

Mρ2 1
(1)
V (q ) =
2
. (6.28)
f ρ2 (q 2 − Mρ2 )

Given the relatively narrow width of the ρ-meson, it is customary to approximate

the imaginary part of (1)
V (q ) by a delta-function
2

Mρ2
Im (1)
V (q ) =
2
π δ(q 2 − Mρ2 ) . (6.29)
f ρ2

If more accuracy is needed, the next level is a Breit-Wigner parametrization

1 1 1 Mρ3 ρ
Im(1)
V (s) = , (6.30)
π π f ρ2 s − M 2 2 + M 2  2
ρ ρ ρ

where s ≡ q 2 > 0, and a normalization such that this function reduces to Eq. (6.29)
in the zero-width approximation. This result is not strictly valid near threshold, as
the leading two-pion state coupled to the vector current,
 
0|Vμ (0)|π( p1 )π( p2 ) (6.31)

has not been explicitly taken into account. This matrix element involves a p-wave
threshold factor, and the spectral function can be related to the ratio R(s) in electron-
positron annihilation. For two flavours this relation is
 
5 αs 5 1
R(s) = 1+ + · · · = (8 π 2 ) Im (1) V (s) , (6.32)
3 π 3 π

where attention must be paid to the normalization of the vector correlator, as R(s)
involves the electrically neutral vector current. In the threshold region R(s) is related
to the pion electromagnetic form factor, Fπ (s), through
 3
+ − + − 1 4 m 2π 2
R e e → π π = 1− |Fπ (s)|2 . (6.33)
4 s
6 Hadronic Spectral Functions 37

Substituting this expression into Eq. (6.32) gives a relation between the spectral
function and the pion form factor
 3
3 4 m 2π 2
Im (1)
V (s) = 1− |Fπ (s)|2 , (6.34)
160 π s

which is a more accurate expression than the simple Breit-Wigner Eq. (6.30). The
pion form factor in this expression can be obtained from e+ e− data, or from τ -lepton
decay data in the vector channel [7]. Attention must be paid to the difference in the
currents involved in these two processes, i.e. the former involves electrically neutral
currents, while the latter involves electrically charged currents. The normalization
of these vector current correlators then differs by a factor-two!

References

1. K.G. Patrignani et al., Particle data group. Chin. Phys. C 40, 100001 (2016)
2. C. A. Dominguez, Z. Phys. C 26, 269 (1984)
3. H. Pagels, A. Zepeda, Phys. Rev. D 5, 3262 (1972)
4. C.A. Dominguez, E. de Rafael, Ann. Phys. (NY) 174, 372 (1987)
5. J. Bijnens, J. Prades, E. de Rafael, Phys. Lett. B 226 (1995)
6. C.A. Dominguez, L. Pirovano, K. Schilcher, Phys. Lett. B 425, 193 (1998)
7. M. Davier, A. Höcker, Z. Zhang, Rev. Mod. Phys. 78, 1043 (2006)
8. J. Bordes, C.A. Dominguez, P. Moodley, J. Penarrocha, K. Schilcher, J. High Energy Phys.
1005, 064 (2010)
9. C.A. Dominguez, N.F. Nasrallah, R. Röntsch, K. Schilcher, J. High Energy Phys. 05, 020
(2008)
10. C.A. Dominguez, N.F. Nasrallah, R. Röntsch, K. Schilcher, J. High Energy Phys. 02, 072
(2008)
11. C.A. Dominguez, N.F. Nasrallah, R. Röntsch, K. Schilcher, Phys. Rev. D 79, 014009 (2009)
12. A. Ayala, C.A. Dominguez, M. Loewe, Y. Zhang, Phys. Rev. D 90, 034012 (2014)
Chapter 7
QCD Chiral Sum Rules

This section deals with a few sum rules first obtained as consequences of chiral
symmetry [1], i.e. long before QCD. First, and foremost, the two Weinberg sum
rules (WSR) [2], derived from current algebra and chiral SU (2) × SU (2) symmetry
are given by
∞
1
W1 ≡ ds [ImV (s) − Im A (s)] = 2 f π2 , (7.1)
π
0

∞
1
W2 ≡ ds s [ImV (s) − Im A (s)] = 0 , (7.2)
π
0

where

μν
VV
(q 2 ) = i d 4 x eiq x < 0|T (Vμ (x)Vν† (0))|0 >

= (−gμν q 2 + qμ qν ) V (q 2 ) , (7.3)


μν (q ) = i
AA 2
d 4 x eiq x < 0|T (Aμ (x)A†ν (0))|0 >

= −gμν 1 (q 2 ) − qμ qν  A (q 2 ) , (7.4)

with Vμ (x) =: d̄(x)γμ u(x): is the conserved vector current in the chiral limit,
Aμ (x) =: d̄(x)γμ γ5 u(x): the axial-vector current, and qμ is the four-momentum
carried by the currents. The functions V,A (q 2 ) are free of kinematical singularities,
thus satisfying dispersion relations. In perturbative QCD (PQCD) they are normal-
ized as

© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2018 39

C. A. Dominguez, Quantum Chromodynamics Sum Rules,
SpringerBriefs in Physics, [Link]
40 7 QCD Chiral Sum Rules

(a)

(b)

Fig. 7.1 ALEPH data [3] on the a vector, and b axial-vector spectral functions, ρ(s) ≡ Im V,A (s),
from τ -decay. The dash line is the PQCD prediction

1   
Im V (q 2 ) = Im  A (q 2 ) = 1 + O αs (q 2 ) . (7.5)
4π
A word of warning regarding Eq. (7.5), as it is common in the literature to nor-
malize to an overall factor of 1/8π. The difference stems from whether one employs
the electric charge neutral current instead of the electrically charged current, as in
Eqs. (7.3), (7.4).
In QCD, and in the chiral limit, W1,2 become finite energy sum rules for
s > s0  1 − 3 GeV2 , due to Eq. (7.5), plus the fact that at these energies the leading
vacuum condensate contribution, i.e. the dimension d = 6 four-quark condensate, is
negligible (the gluon condensate is chiral-symmetric). The WSR can then be written
as

s0
1
Wn (s0 ) ≡ ds s n [ImV (s) − Im A (s)] = 2 f π2 δn0 (n = 0, 1) , (7.6)
π
0

where s0  1 − 3 GeV2 is the squared energy beyond which QCD is valid. The
WSR, Eq. (7.6), also follows from Cauchy’s theorem in the complex s-plane, and the
assumption of quark-hadron duality, i.e.
s0 
1 1
ds f (s) Im(s) = − f (s) (s) ds , (7.7)
π 2πi |s|=s0
0

where f (s) = s n to reproduce Eq. (7.6).

7 QCD Chiral Sum Rules 41

Fig. 7.2 The first WSR, Eq. (7.1), (solid squares), together with the modified (pinched) sum rule
(open circles), Eq. (7.8), as a function of s0 , with the integrals computed from the ALEPH data [3].
The straight dotted line is the right-hand-side of Eq. (7.8), i.e. 2 f π2

The ALEPH data on the vector and axial-vector spectral functions, determined
from hadronic decays of the τ -lepton, are shown in Fig. 7.1. The resonance peaks
correspond to the ρ-meson in the vector channel, and to the a1 meson in the axial-
vector channel. Agreement with QCD seems to be beyond the end-point of the
data, although quark-hadron duality violation (DV) cannot be ruled out at this stage.
However, in view of the results discussed in Chap. 5, on the extension of the kinematic
range, one expects PQCD to be valid for s  3 GeV2 . This situation led to the proposal
[4, 5] of introducing pinched integration kernels in the FESR, Eq. (7.7), in order to
quench potential DV. The first WSR, Eq. (7.1), is replaced by a linear combination
of the first and the second WSR, Eqs. (7.1) and (7.2)
∞  s 1
ds 1 − [ImV (s) − Im A (s)] = 2 f π2 , (7.8)
s0 π
0

with the normalization as in Eq. (7.5). Using ALEPH data on the spectral functions
from τ -decay [3] in Eq. (7.8) gives the result shown in Fig. 7.2 (open circles), com-
pared with the original sum rule Eq. (7.1) (solid squares). The impact of the pinched
kernel is quite clear. It should be emphasized, though, that the lack of saturation in
Eq. (7.1) does not necessarily imply DV. It could simply be due to the fact that the
threshold for PQCD lies above the end-point of the τ -data.
Another chiral sum rule, the Das-Mathur-Okubo (DMO) sum rule [6], is obtained
from the difference between the imaginary parts of the vector and the axial-vector
correlator in the up- and down-quark sector, weighted by the kernel 1/s, i.e.
42 7 QCD Chiral Sum Rules

¯
Fig. 7.3 −4 L 10 ≡ (0) as a function of s0 from the pinched DMO sum rule, Eq. (7.11)

 s0
1 ds
¯
(0) = Im V (s) − Im A (s) , (7.9)
π 0 s

with the pion-pole excluded from  A (s). The finite remainder (0) ¯ is related to a
counter-term of the order O( p 4 ) Lagrangian of CHPT [7], L̄ 10 , and can be expressed
as
1 2 2 1
¯
(0) = −4 L̄ 10 = f π rπ  − FA = 0.026 ± 0.001 , (7.10)
3 2

where rπ2  = 0.439 ± 0.008 fm2 is the electromagnetic radius of the pion [8], and
FA = 0.0119 ± 0.0001 is the radiative pion decay constant [9].
The DMO sum rule, Eq. (7.9), is reasonably saturated by the ALEPH data above
s0  1.5 GeV2 . A pinched version of this sum rule, proposed in [5, 10], is


f π2 s0
ds  s 2 1
¯
(0) =2 + 1− Im V (s) − Im A (s) . (7.11)
s0 0 s s0 π

This pinched sum rule is also well satisfied above s0  1.5 GeV2 , as shown in
Fig. (7.3) giving [5]

L̄ 10 = −(6.5 ± 0.1) × 10−3 , (7.12)

7 QCD Chiral Sum Rules 43

a result in very good agreement with an earlier determination [10], L̄ 10 = −(6.43 ±

0.08) × 10−3 , as well as with other determinations from more involved procedures
to deal with potential DV [11, 12], L̄ 10 = −(6.46 ± 0.15) × 10−3 . It also agrees
with lattice QCD determinations, albeit within their larger uncertainties [13, 14].

References

1. V. de Alfaro, S. Fubini, G. Furlan, C. Rossetti, Currents in Hadron Physics (Noth Holland,

Amsterdam, 1973)
2. S. Weinberg, Phys. Rev. Lett. 18, 507 (1967)
3. M. Davier, A. Höcker, Z. Zhang, Rev. Mod. Phys. 78, 1043 (2006)
4. K. Maltman, Phys. Lett. B 440, 367 (1998)
5. C.A. Dominguez, K. Schilcher, Phys. Lett. B 581, 193 (2004)
6. T. Das, V.S. Mathur, S. Okubo, Phys. Rev. Lett. 19, 859 (1967); T. Das, G.S. Guralnik, V.S.
Mathur, F.E. Low, J.E. Young. Phys. Rev. Lett. 18, 759 (1967)
7. S. Scherer, Adv. Nucl. Phys. 27, 277 (2003); J. Gasser, H. Leutwyler, Nucl. Phys. B 250, 465
(1985); G. Ecker, J. Gasser, A. Pich, E. de Rafael, Nucl. Physi. B 321, 311 (1989); G. Amoros,
J. Bijnens, P. Talavera, Nucl. Phys. B 585, 293 (2000) [Erratum ibid. 598, 665 (2001)]
8. S.R. Amendolia et al., Nucl. Phys. B 277, 168 (1986)
9. K.G. Patrignani et al., Particle data group. Chinese Phys. C 40, 100001 (2016)
10. C.A. Dominguez, K. Schilcher, Phys. Lett. B 448, 93 (1999)
11. M. Gonzalez-Alonso, A. Pich, J. Prades, Phys. Rev. D 81, 074007 (2010)
12. M. Gonzalez-Alonso, A. Pich, J. Prades, Phys. Rev. D 82, 014019 (2010)
13. JLQCD collaboration, E. Shintani et al., Phys. Rev. Lett. 101, 242001 (2008)
14. RBC, UKQCD collaboration, P.A. Boyle, L. Del Debbio, J. Wennekers, J.M. Zanotti, Phys.
Rev. D 81, 014504 (2010)
Chapter 8
Determination of the Gluon Condensate

As mentioned in Chap. 2, the gluon condensate of dimension d = 4 is one of the two

leading order terms in the OPE, Eq. (2.2), together with the light-quark condensate.
The gluon condensate was introduced in the pioneer papers on Laplace QCD sum
rules by Shifman et al. [1]. Its numerical value was estimated using experimental
information on e− e+ annihilation in the charmonium channel, with the result [1]
α 
s
G 2 = 0.012 GeV4 , (8.1)
π
with no uncertainty given. This value remained in use in applications for quite some
time. A detailed analysis, also in the charmonium channel, with determined uncer-
tainties, was performed in [2] with the result
α 
s
G 2 = 0.014 ± 0.0044 GeV4 . (8.2)
π
This result was questioned in [3, 4] claiming it underestimates its value by up to
a factor three. British groups [5, 6] confirmed this claim in the framework of two-
dimensional QCD, and a reanalysis in the charmonium channel [7, 8] reached similar
conclusions. Using experimental data on electron-positron annihilation into hadrons,
and Laplace transform QCD sum rules, the gluon condensate was found to be one
order of magnitude smaller than Eq. (8.1) [9]. Also using experimental data on
electron-positron annihilation into hadrons, within FESR [10], a bound was obtained
α 
s
0.07 GeV4 ≤ G 2 ≤ 0.19 GeV4 . (8.3)
π
A determination using the first set of ALEPH data on the vector and axial-vector
current correlators from hadronic decays of the τ -lepton [11], together with FESR,
found [12]

© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2018 45

C. A. Dominguez, Quantum Chromodynamics Sum Rules,
SpringerBriefs in Physics, [Link]
46 8 Determination of the Gluon Condensate
α 
s
G 2 = (0.1 − 0.4) GeV4 . (8.4)
π
This historical account is provided solely for the purpose of highlighting the
difficulties encountered in determining the value of the gluon condensate.
To add to the confusion, and on the purely theoretical domain, there have been
claims that the gluon condensate is affected by renormalon ambiguities, and thus its
numerical value remains in limbo [13]. This is definitely not the case, for a variety
of reasons. First, and foremost, the gluon condensate is a renormalization group
invariant quantity in the operator product expansion of current correlators at short
distances, Eq. (2.2), to be interpreted solely as a parameter of the QCD sum rule
method [14]. Its numerical value is to be obtained only by confronting QCD sum
rules with experimental data. Hence, no further claims are made on its nature/origin.
Second, the renormalon issue is severely model-dependent, in that its value relies
on estimates of extremely high order contributions in perturbative QCD (similar to
ladder diagrams). Clearly, estimates are not determinations.
Regarding the numerical value of the gluon condensate, all the estimates and
determinations mentioned above are rather dated, as the QCD information at the
time of their determination was restricted to next or to next-to-next to leading order
in perturbative QCD, i.e. to O(αs ) or O(αs2 ). Current information is up to O(αs4 ) in the
light-quark region, and O(αs3 ) in the charm- and bottom-quark regions. Furthermore,
the procedures to determine the gluon condensate required the strong coupling, αs ,
dependent on the value of QCD entering the logarithm, i.e.

2π
αs (s0 ) =   + ··· , (8.5)
b0 ln s0 /2QCD

where b0 is given in Eq. (3.3). The value of QCD at the time of the above determi-
nations was in the range QCD  100 − 150 MeV, while currently it is more than
twice as large! [15].
Turning to the present, the most recent QCD-FESR determinations of the gluon
condensate are (i) based on data on e+ e− annihilation to hadrons in the region (1.0 −
5.0) GeV2 using standard FESR [16], and (ii) based on latest ALEPH data on hadronic
spectral functions [17], and (iii) from a novel, non-standard FESR in the charm-
quark region, designed to overcome the drawbacks of standard FESR, and allowing
for unprecedented and trustworthy accuracy [18]. The latter also allows to test the
expectation that the gluon condensate should be scale-independent.
A description of the most recent gluon condensate determination from data on
e+ − e− annihilation into hadrons [16] starts with the following definitions. The
relevant two-point function in this case is the electromagnetic current correlator

μν (q )
EM 2
= i d 4 x eiqx 0|T (JμEM (x) JνEM (0))|0
 
= −q2 gμν + qμ qν EM (q2 ) , (8.6)
8 Determination of the Gluon Condensate 47

where for three flavours

2 1 1
JμEM (x) = ū(x)γ μ u(x) − d̄ (x)γ μ d (x) − s̄(x)γ μ s(x) . (8.7)
3 3 3
Since QCD is flavour blind, and if isospin invariance is exact, then it is convenient
to define a QCD current correlator using any of the quark currents q̄i γμ qi with flavour
i. This leads to
nf
Im EM (q2 ) = Qi2 Im VV (q2 ) , (8.8)
i=1

with Qi the charge of the quark i = u, d , s, . . ., i Qi2 = 2/3 for nf = 3, and

Im VV (q2 ) is the QCD correlator of vector currents of flavour i. The spectral func-
tion, Im EM (s), with s the square energy, is accessible experimentally from data on
e+ e− annihilation into hadrons as follows.
Starting in the hadronic sector, the standard ratio R(s) is

σTOT (e+ e− → hadrons)

R(s) = , (8.9)
σ (e+ e− → μ+ μ− )

where the electromagnetic cross section is

4π αEM
2
σ (e+ e− → μ+ μ− ) = , (8.10)
3s

and αEM = e2 /4π . In QCD the relation between R and the electromagnetic spectral
function is given by
nf  αs 
R(s) = 12 π Im EM (s) = 3 Qi2 1 + + ... . (8.11)
i=1
π

 2
i Qi arises at order O(αs ) and vanishes
3
A singlet contribution proportional to
if one sums over three flavours. In the case of the two-pion final state, dominated by
the ρ-resonance, there is a relation between R and the pion form factor, Fπ(0) (s)
3
1 4m2π 2
2
Re+ e− →π + π − (s) = 1− Fπ(0) (s) . (8.12)
4 s

Turning to the QCD sector, the vector current correlator, V V (q2 ), in Eq. (8.8) is
given by the OPE, Eq. (2.2). Of particular interest here is the purely perturbative term,
as well as the dimension d = 4 non-perturbative contribution. The latter involves the
gluon condensate together with the quark condensate multiplied by the quark masses
48 8 Determination of the Gluon Condensate

π 2 αs  
C4 O4  =  G μν G μν  + 4 π 2 mu ūu + md d̄ d  + ms s̄s , (8.13)
3 π
where αs is the running strong coupling, and in the sequel 0|O2N |0 ≡ O2N  is to
be understood. This condensate is renormalization group invariant to all orders in
PQCD.
The purely perturbative QCD part of Im VV (q2 ) is currently known [19, 20] up
to order O(αs4 )

b1 b2
8π Im VV (s) = 1 + as + F3 + Lμ + as F4 + b1 F3 +
as2 3
Lμ
2 2
 
b2 π2 2 5 3 b3
+ 1 L2μ + as4 k3 − b1 F3 − π 2 b1 b2 + b1 F4 + b2 F3 + Lμ
4 4 24 2 2

b1 3 5 b3
+ b1 F3 + b2 L2μ + 1 L3μ , (8.14)
2 2 4 8

where as ≡ αs (μ2 )/π , Lμ ≡ ln(Q2 /μ2 ), k3 = 49.076 [21], F3 = 1.9857

− 0.1153 nf , and F4 = 18.2427 − π3 ( b21 )2 − 4.2158 nf + 0.0862 n2f , with b1 given
2

in Eq. (3.4). Notice that the normalization factor 8π is due to the nature of the vec-
tor current, Eq. (8.7), entering in the correlator Eq. (8.6) (recall that for electrically
charged currents the standard normalization factor is instead 4π ) . The expression
for the running coupling up to five-loop order is given by [22]
2
αs(4) (s0 ) α (1) (s0 ) αs(1) (s0 ) b2
= s + − ln L
π π π b1
(1) 3 2  
α (s0 ) b2 b3
+ s ln2 L − ln L − 1 +
π 2
b1 b1
(1) 4 3
α (s0 ) b2 5 1
− s ln3 L − ln2 L − 2ln L +
π 3
b1 2 2

b2 b3 b3
+ 3 2 lnL + , (8.15)
β1 b1

with
αs(1) (s0 ) −2
≡ , (8.16)
π β1 L
 
where L ≡ ln s0 /2MS defines the standard MS scale MS = (340 ± 8) MeV for
nF = 3 [15].
The FESR in this channel can be written as
 s0
1
(−) C2N +2 O2N +2  = 8π
N 2
ds sN Im DATA (s) − s0N +1 M2N +2 (s0 ) , (8.17)
0 π
8 Determination of the Gluon Condensate 49

1.0

0.5
GeV4
C4 04

0.0

0.5

1 2 3 4
s0 GeV2
(n =3)
Fig. 8.1 C4 O4  calculated in FOPT using αs (Mτ2 ) = 0.321 ± 0.015, corresponding to MSf =
341 MeV. The smaller uncertainties are obtained assuming no correlations between experiments,
while the larger ones assume 100% correlations for data obtained using the same experimental
facility

where the dimensionless PQCD moments M2N +2 (s0 ) are given by

  
1 ds s N
M2N +2 (s0 ) ≡ −8π 2
VV (s)
2π i C(|s0 |) s0 s0
 s0  
ds s N 1
= 8π 2 Im VV (s) . (8.18)
0 s0 s0 π

Given the expected size of the dimension-four term, C4 O4 , Eq. (8.13), i.e.
C4 O4  = O(10−3 GeV4 ), one expects a sizeable cancellation between the data con-
tribution and PQCD in Eq. (8.17). A posteriori, this turns out to be the source of
the large uncertainty in the result, as shown in Fig. 8.1 using FOPT. The hadronic
data used in the line integral in Eq. (8.17) is described in detail in [16]. It includes
twenty-nine different sources in a variety of hadronic final-states.
The situation with the dimension d = 6 condensate is even worse, as expected.
The results for C6 O6  are shown in Fig. 8.2 in a different vertical scale. The origin
and structure of C6 O6  is discussed at the end of this section.
The conclusion is that this approach is unsuccessful. Even if more accurate data
were available, the partial cancellation between the two terms of comparable mag-
nitude on the right hand side of Eq. (8.17) is a serious challenge.
Another attempt at obtaining the value of the gluon condensate was made in [17],
still from QCD FESR but using instead the ALEPH data base from τ -decay [23].
The relevant correlators are
50 8 Determination of the Gluon Condensate

1
GeV6

0
C6 06

3
1 2 3 4
s0 GeV2
(n =3)
Fig. 8.2 C6 O6  calculated in FOPT using αs (Mτ2 ) = 0.321 ± 0.015, corresponding to MSf =
341 MeV. The smaller uncertainties are obtained assuming no correlations between experiments,
while the larger ones assume 100% correlations for data obtained using the same experimental
facility


μν (q ) = i
VV 2
d 4 x eiqx 0|T (Vμ (x)Vν† (0))|0 (8.19)

= (−gμν q2 + qμ qν ) V (q2 ) ,


μν (q )
AA 2
=i d 4 x eiqx 0|T (Aμ (x)A†ν (0))|0 (8.20)

= (−gμν q2 + qμ qν ) A (q2 ) − qμ qν 0 (q2 ) ,

where Vμ (x) =: ū(x)γμ d (x) :, Aμ (x) =: ū(x)γμ γ5 d (x) :, with u(x) and d (x) the
quark fields, and V,A (q2 ) normalized in perturbative QCD (PQCD) (in the chi-
ral limit) according to

1 PQCD 1 PQCD 1 αs (s)

Im V (s) = Im A (s) = 1+ + ... , (8.21)
π π 4π 2 π

where s ≡ q2 > 0 is the squared energy. Notice the normalization factor due to the
currents being electrically charged. The Lorentz decomposition is used to separate
the correlation function into its J = 1 and J = 0 parts. To the accuracy needed in the
following, the vector current can be assumed to be conserved. The running strong
coupling was obtained from Eq. (3.9) using as input [24] αs (Mτ2 ) = 0.341 ± 0.013.
Instead of using the correlators Eqs. (8.19) and (8.20), the Adler function was
invoked as it is more convenient when using contour improved perturbation theory
(CIPT) to integrate in the complex plane. The Adler function is defined as
8 Determination of the Gluon Condensate 51

d
D(s) ≡ −s (s) , (8.22)
ds
with (s) ≡ V,A (s). Invoking Cauchy’s theorem and after integration by parts the
following relation is obtained
 
s N
1 1 ds  N +1 
ds (s) = s − s0N +1 D(s) . (8.23)
|s|=s0 s0 N + 1 s0N |s|=s0 s

After RG improvement, the perturbative expansion of the Adler function becomes

1  α (−s) m
s
D(s) = Km , (8.24)
4 π2 m=0
π

where K0 = K1 = 1, K2 = 1.6398 , K3 = 6.3710, for three flavours, and K4 =

49.076 [19–22]. The vacuum condensates are determined from the following pinched
FESR
 s0  
N +1 s N 1
C2N +2 O2N +2  = (−) 2 N
4π s0 ds 1 − Im (s)HAD
0 s0 π
 
+ (−)N s0N +1 M0 (s0 ) − MN (s0 ) , (8.25)

1 1  
MN (s0 ) = Km IN +1,m (s0 ) − I0,m (s0 ) , (8.26)
2π (N + 1) m=0

with   m
N
s αs (−s)
IN ,m ≡ i ds , (8.27)
|s|=s0 s0 π

to be computed as explained in Chap. 4.

A key issue with the FESR Eq. (8.25) is the opposite signs of the two terms on
its right-hand-side. In practice these two terms are roughly of the same order of
magnitude, thus leading to results affected by large uncertainties.
The results for C4 O4  in CIPT are shown in Fig. 8.3 for αs (Mτ2 ) = 0.341 from
[24]. The data near the end-point (s0  Mτ2 ) are affected by such uncertainties that
no safe estimate of the gluon condensate can be given there. However, there is a
region s0  2.0 − 2.4 GeV2 where an estimate can be obtained, albeit with a large
error, i.e.
C4 O4 |CIPT = (0.017 ± 0.012) GeV4 , (8.28)

where this value is obtained from results corresponding to the V + A spectral function
at s0 = 2.35 GeV2 . Results from using FOPT are
52 8 Determination of the Gluon Condensate

Fig. 8.3 The dimension d = 4 condensate in CIPT from the FESR, Eq. (8.25), with N = 1 and
αs (Mτ2 ) = 0.341 from [24]. The ALEPH data for the vector (V) (upper black solid triangles), the
axial (A) (lower blue solid squares) and the average 21 (V + A) (middle red solid dots) spectral
function were used

C4 O4 |FOPT = (0.022 ± 0.006) GeV4 . (8.29)

Combining both results leads to the conservative upper bound [17]

C4 O4 |V,A  0.035 GeV4 . (8.30)

It should be mentioned that a very different result was obtained by the ALEPH
collaboration [23] after fitting simultaneously all relevant parameters, obtaining neg-
ative values for the gluon condensate from the axial-vector channel, as well as from
the V + A channel. Given the positivity of C4 O4 , this global fitting procedure seems
unreliable.
An entirely different and successful approach to determine the gluon condensate,
still based on QCD FESR, was proposed in [18]. This leads to a reasonably accurate
value in full agreement with the bound Eq. (8.30). The first consideration is to choose
the charm-quark energy region, where data for the R-ratio in e+ e− → hadrons [25,
26] starts with the two well known narrow-width resonances J /ψ and ψ(2S). Next,
the approach should be such that this contribution to a FESR becomes leading. The
latter excludes the standard FESR, e.g. Eq. (8.17), as it lacks such weighting. The
optimal FESR satisfying this criterion should be such that the gluon condensate does
not become the result of a cancellation between terms of similar magnitude, as it
happens in Eq. (8.17). In order to achieve this one starts by considering the most
general FESR
8 Determination of the Gluon Condensate 53
 s0 
1 1
p(s) Im (s) ds = − p(s) (s) ds + Res[(s) p(s), s = 0] ,
0 π 2π i C(|s0 |)
(8.31)
where p(s) is now a meromorphic function, the integral on the right hand side involves
QCD, provided s0 is large enough, and the left hand side involves the hadronic spectral
function
1
Im (s)|DATA = Rc (s) , (8.32)
12π

with Rc (s) the standard R-ratio for charm production in e+ e− annihilation. The
residue term in Eq. (8.31) would appear if p(s) is singular at some point(s) in the
complex plane, e.g. if p(s) is of the form

1
p(s) = . (8.33)
sN +1
The leading non-perturbative contribution to a FESR involving such a kernel was
¯
obtained in [27], and in [28] using the MS-scheme with the result
  α  
(s)|NPQCD ec2 s 2 αs 
Res , s = 0 =  N +2 π
G aN 1 + b̄N , (8.34)
sN +1 4m̄2c π

where the charm-quark mass, m̄c , and the strong coupling αs depend on the renor-
malization scale μ, and

2N + 2 (4 + N ) (7/2)
aN = − , (8.35)
15 (7/2 + N ) (4)

4
b̄N = bN − (2N + 4) − lm , (8.36)
3

where lm ≡ ln(m̄2c (μ)/μ2 ), μ is a renormalization scale, b0 = 1469/162, b1 =

135779/12960, and b2 = 1969/168. Other values of bi are given in [27, 28]. The
unknown NNLO term will be included as a source of uncertainty in the final
analysis. The QCD parameters are the charm-quark mass mc (μ2 ), the strong cou-
pling αs (μ2 ), and the gluon condensate  απs G 2 . Their numerical values used in
[18] are: αs (MZ2 ) = 0.1183 ± 0.0007 from LQCD [29], to be scaled down to the
charm-quark region using Eq. (3.9), and the charm-quark mass from LQCD [30]
m̄c (3 GeV) = 986.4 ± 4.1 GeV, in good agreement with the most recent QCDSR
determination [31] m̄c (3 GeV) = 987 ± 9 MeV.
There is also a low-energy PQCD contribution to be taken into account, together
with the gluon condensate. The QCD low energy expansion in inverse powers of the
charm-quark mass can be formally written as
54 8 Determination of the Gluon Condensate

3 ec2
PQCD (s) = C n zn , (8.37)
16 π 2 n≥0

where z = s/(4m2c ). The coefficients C n are then expanded in powers of αs (μ)

αs (μ)  (10)  αs (μ) 2  (20) 

C̄n = C̄n(0) + C̄n + C̄n(11) lm + C̄n + C̄n(21) lm + C̄n(22) lm2
π π
αs (μ)  (30)
3

+ C̄n + C̄n(31) lm + C̄n(32) lm2 + C̄n(33) lm3 + . . . . (8.38)
π

where lm ≡ ln(m̄2c (μ)/μ2 ). Up to three loop level the coefficients of C̄n are known
up to n = 30 [32, 33]. At four-loop level C̄0 and C̄1 were determined in [32, 34],
[35], C̄2 is from [33, 36], and C̄3 from [37]. The kernel p(s) is chosen so that no
coefficients C̄4 and above contribute to the Cauchy residue at s = 0.
Notice that the gluon condensate also enters in the contour integral, Eq. (8.31).
However, its contribution to that integral is negligible. Its appearance in the residue,
Eq. (8.34), is the key point of this approach. Indeed, substituting Eq. (8.34) into the
FESR Eq. (8.31) gives

α   2 N +2   s0
s 4m̄c 1
G 2
=  αs
 p(s) Im (s) ds
π ec aN 1 + π b̄N
2
0 π
   
1
+ p(s)(s)ds − Res PQCD (s) p(s) |s=0 , (8.39)
2π i C(|s0 |)

where the last term above is the residue at the singularity from the low energy PQCD
expansion, Eq. (8.37). It is non-zero depending on the value of N in the integration
kernel p(s), Eq. (8.33). A key property of this FESR is that the two integral terms in
brackets, i.e. the data line integral, and the contour QCD integral now have the same
sign !
Solving the renormalization group equation for the strong coupling and for the
quark mass one can obtain their values at any scale s in terms of their values at
any given reference scale s∗ from Eqs. (3.9) and (3.31), e.g. s = s0 . Regarding the
renormalization scale μ, we follow the choice [31–38] μ2 = (3 GeV)2 in the low
energy QCD expansion, and μ2 = s0 in the high energy QCD expansion on the circle
of radius s = |s0 |.
The PQCD correlator entering the contour integral in Eq. (8.39) is given by the
(high energy) expansion
n
αs (μ2 )
(s)|PQCD = ec2 (n) (s) , (8.40)
n=0
π

where ec = 2/3 is the charm-quark electric charge, and

8 Determination of the Gluon Condensate 55

J 2s
5

3
Rs

0
8 10 12 14 16 18 20 22
s GeV2

Fig. 8.4 Experimental data for R(s) [25, 26] together with the optimal integration kernel, Eq. (8.42),
with N = 2 (dash curve), and p(s) = 1/s2 (solid curve) normalized to coincide with the former at
the position of the ψ(2S) peak

i
m̄2c
(n) (s) = (n)
i , (8.41)
i=0
s

with mc ≡ mc (μ) the running charm-quark mass in the MS-scheme. The function
(s)PQCD is known up to next-to-next to leading order in PQCD [21, 39, 40].
Finally, turning to the experimental data in this region, shown in Fig. 8.4, the
analysis followed that of [28, 41], to wit.
For the contribution of the first two narrow resonances PDG data [15] was used,
followed by the open charm region, after√subtraction of the light-quark sector con-
tribution [42]. In the region 3.97 GeV ≤ s ≤ 4.26 GeV only CLEO data [25] was
taken into account, as it is the more accurate.
Next, there are two data sets from BES [26, 43], which were assumed not fully
independent for the error analysis, thus adding errors linearly, rather than in quadra-
ture. These data set, though, is independent from CLEO [25] so that errors were
combined in quadrature. There is a data desert region for s = 25 − 49 GeV2 , fol-
lowed by CLEO data up to s  90 GeV2 , fully compatible with PQCD. For further
details see [18].
Turning to the integration kernel, p(s), Eq. (8.33), the optimal choice is
 s N
0
p(s) = − 1, (8.42)
s
with N ≥ 2. A detailed justification of this choice is given in [18].
56 8 Determination of the Gluon Condensate

Table 8.1 Results for the gluon condensate for the kernel, Eq. (8.42), for N = 2 and its sources of
uncertainty from s0 , αs , mc , the experimental data, and the total uncertainty. Method (a) refers to
using the currently known NLO radiative correction to the residue, Eq. (8.34). Method (b) assumes
that the NNLO correction is as large, and of the same sign as the NLO one
Method  απs G 2  (GeV4 ) Uncertainties (GeV4 )
s0 αs mc DATA T
(a) 0.044 0.0028 0.0003 0.0048 0.0043 0.007
(b) 0.026 0.0016 0.0001 0.0027 0.0024 0.004

The impact of the integration kernel, Eq. (8.42), can be appreciated from Fig. 8.4.
As expected by design, the contribution of the data region to the line integral in
the FESR, Eq. (8.39), enhances the well known first two narrow resonances, J /ψ
and ψ(2S), and quenches substantially the rest. The results for the gluon condensate
together with the various uncertainties, is shown in Table 8.1. The result for the gluon
condensate, after considering all uncertainties, is
α 
s
G 2 = 0.037 ± 0.015 GeV4 , (8.43)
π

which agrees with a LQCD determination [44]  απs G 2  = 0.028 ± 0.003 GeV4 . It is
also consistent with the bound, Eq. (8.30). An independent QCD sum rule determi-
nation in the light-quark region, from an unconventional method, gives [45]
αs 2
 G  = 0.062 ± 0.019 GeV4 , (8.44)
π
in agreement with Eq. (8.43) within very large errors. If taken with caution, this result
from the low energy region together with Eq. (8.43) from the high energy domain
would support the scale independence of the gluon condensate.
Returning to the issue of the dimension d = 6 condensates, there are three possible
structures built from the QCD quark and gluon fields

Ô6 |q = q̄(x) 1 q(x) q̄(x) 2 q(x) , (8.45)

λa
Ô6 |σ = m q̄(x) σμν q(x) G aμν (x) , (8.46)
2

Ô6 |G 3 = fabc G aμν (x) G bνα (x) G cαμ (x) , (8.47)

corresponding to the four-quark condensate, the mixed quark-gluon condensate, and

the three-gluon condensate. The object i in Eq. (8.45) above is a combination of
gamma-matrices, SU(3) λ-matrices, and the tensor σμν ∝ [γμ , γν ], depending on
the correlator (see [1]). None of these condensates shares the status of the quark
condensate, which enters in the Gell-Mann-Oakes-Renner relation. In principle they
8 Determination of the Gluon Condensate 57

Fig. 8.5 The chiral condensate of dimension d = 8 from a pinched FESR [17]

would all contribute to the dimension d = 6 term in the OPE, Eq. (2.2), with a-
priori unknown hierarchy. In the early days of the QCD sum rule technique attempts
were made to relate the four-quark condensate to the square of the quark condensate
through some sort of vacuum saturation ansatz [1]. This was accompanied by hopes
that the other two d = 6 condensates would be numerically negligible. The vacuum
saturation ansatz has been recently shown to break down at next-to-next-to leading
order [46]. Worse still, this ansatz gives the wrong sign for the ratio of the vector
and the axial-vector dimension d = 6 condensates [47, 48], and it underestimates
the d = 6 condensate in the vector channel by a factor 10 [47, 48].
This does not preclude attempts at determining the numerical values of vacuum
condensates of higher dimensionality [17]. What would be lacking is their identifi-
cation in terms of quark-gluon fields. Given the deterioration of uncertainties with
increasing dimensionality, the determination of the dimension d = 8 condensate
appears to be the borderline [17, 49, 50]. An example of the result for the dimension
d = 8 chiral condensate O8  from [17] is shown below in Fig. 8.5.

References

1. M. A. Shifman, A. I. Vainshtein, V. I. Zakharov, Nucl. Phys. B 147, 385 (1979). (ibid., B 147,
448 (1979))
2. B. Guberina, R. Meckbach, R.D. Peccei, R. Rückl, Nucl. Phys. B 184, 476 (1981)
3. J.S. Bell, R.A. Bertlmannn, Nucl. Phys. B 177, 218 (1981)
4. J.S. Bell, R.A. Bertlmannn, Phys. Lett. B 137, 107 (1984)
5. A. Bradley, C. S. Langensiepen, G. Shaw, Phys. Lett. B 102 180, 359
6. D.J. Broadhurst, Phys. Lett. B 123, 251 (1983)
7. J. Marrow, G. Shaw, Z. Phys, C-Particles and Fields 33, 237 (1986)
8. J. Marrow, J. Parker, G. Shaw, Z. Phys. C-Particles and Fields 37, 103 (1987)
58 8 Determination of the Gluon Condensate

9. S.I. Eidelman, L.M. Kurdadze, A.I. Vainshtein, Phys. Lett. B 82, 278 (1979)
10. R.A. Bertlmann, C.A. Dominguez, M. Loewe, M. Perrotet, E. de Rafael, Z. Phys. C-Particles
and Fields 39, 231 (1988)
11. H. Albrecht et al., Z. Phys. C-Particles and Fields 33, 7 (1986)
12. C.A. Dominguez, J. Solá, Z. Phys. C-Particles and Fields 40, 63 (1988)
13. M. Beneke, M. Jamin, J. High Energy Phys. 01, 125 (2013)
14. P. Minkowski, private communication
15. K.G. Patrignani et al., Particle Data Group. Chin. Phys. C 40, 100001 (2016)
16. S. Bodenstein, C.A. Dominguez, S.I. Eidelman, H. Spiesberger, K. Schilcher, J. High Energy
Phys. 01, 039 (2012)
17. C.A. Dominguez, L.A. Hernandez, K. Schilcher, H. Spiesberger, J. High Energy Phys. 03, 053
(2015)
18. C.A. Dominguez, L.A. Hernandez, K. Schilcher, J. High Energy Phys. 07, 110 (2015)
19. S.G. Gorishnii, A.L. Kataev, S.A. Larin, Phys. Lett. B 259, 144 (1991)
20. T. van Ritbergen, J.A.M. Vermaseren, S.A. Larin, Phys. Lett. B 400, 379 (1997)
21. P.A. Baikov, K. Chetyrkin, J.H. Kühn, Phys. Rev. Lett. 101, 012002 (2008)
22. K. Chetyrkin, B.A. Kniehl, M. Steinhauser, Phys. Rev. Lett. 79, 2184 (1997)
23. M. Davier, A. Höcker, Z. Zhang, Rev. Mod. Phys. 78, 1043 (2006)
24. A. Pich, Progr. Part. Nucl. Phys. 75, 41 (2014)
25. D. Cronin-Hennessy et al., CLEO 2009. Phys. Rev. D 80, 072001 (2009)
26. J.Z. Bai et al., BES 2006. Phys. Rev. Lett. 97, 262001 (2006)
27. D.J. Broadhurst et al., Phys. Lett. B 329, 103 (1994)
28. J.H. Kühn, M. Steinhauser, C. Sturm, Nucl. Phys. B 778, 192 (2007)
29. R. Horsley et al., Phys. Rev. D 86, 054502 (2012)
30. C. McNeile et al., PQCD Collaboration. Phys. Rev. D 82, 034512 (2010)
31. S. Bodenstein, J. Bordes, C.A. Dominguez, J. Peñarrocha, K. Schilcher, Phys. Rev. D 83,
074014 (2011)
32. R. Boughezal, M. Czakon, T. Schutzmeier, Phys. Rev. D 74, 074006 (2006)
33. A. Maier, P. Maierhöfer, P. Marquard, Phys. Lett. B 669, 88 (2008)
34. R. Boughezal, M. Czakon, T. Schutzmeier, Nucl. Phys. B (Proc. Suppl.) 160 (2006) 164
35. K.G. Chetyrkin, J.H. Kühn, C. Sturm, Eur. Phys. J. C 48, 107 (2006)
36. A. Maier, P. Maierhöfer, P. Marquard, Nucl. Phys. B 797, 218 (2008)
37. A. Maier, P. Maierhöfer, P. Marquard, A.V. Smirnov, Nucl. Phys. B 824, 1 (2010)
38. K. Chetyrkin et al., Theor. Math. Phys. 170, 217 (2012)
39. K.G. Chetyrkin, R. Harlander, J.H. Kühn, M. Steinhauser, Nucl. Phys. B 503, 339 (1997)
40. P.A. Baikov, K. Chetyrkin, J.H. Kühn, Nucl. Phys. Proc. Suppl. B 135, 243 (2004)
41. K.G. Chetyrkin et al., Phys. Rev. D 80, 074010 (2009)
42. A. Hoang, M. Jamin, Phys. Lett. B 594, 127 (2004)
43. J.Z. Bai et al., BES 2002. Phys. Rev. Lett. 88, 101802 (2002)
44. R. Horsley et al., arXiv: 1205.1659
45. B.V. Geshkenbein, Phys. Rev. D 70, 074027 (2004)
46. A. Gomez Nicola, J.R. Pelaez, J. Ruiz de Elvira, Phys. Rev. D 82, 074012 (2010)
47. J. Bordes, C.A. Dominguez, J. Peñarrocha, K. Schilcher, J. High Energy Phys. 02, 037 (2006)
48. C.A. Dominguez, K. Schilcher, J. High Energy Phys. 01, 093 (2007)
49. M. Gonzalez-Alonso, A. Pich, J. Prades, Phys. Rev. D 81, 074007 (2010)
50. M. Gonzalez-Alonso, A. Pich, J. Prades, Phys. Rev. D 82, 014019 (2010)
Chapter 9
Quark Masses

Light quark masses (up-, down-, strange-) were first considered in the framework of
chiral symmetries and named current algebra quark masses [1–5]. These quantities,
being pre-QCD, lacked the current detailed understanding of e.g. quark-mass running
and renormalization. The mechanism of global SU(3) × SU(3) chiral symmetry
realized in the Nambu-Goldstone fashion, and its breaking down to SU(2) × SU(2),
followed by a breaking down to SU(2), and finally to U(1) was first understood using
the strong interaction Hamiltonian [2, 4, 5]

H (x) = H0 (x) + 0 u 0 (x) + 3 u 3 (x) + 8 u 8 (x) . (9.1)

The term H0 (x) is SU(3) × SU(3) invariant, the 0,3,8 are symmetry breaking
parameters, and the scalar densities u 0,3,8 (x) transform according to the 3 3 ⊕ 3 3
representation of SU(3) × SU(3). Only ratios of these densities yield finite results.
In QCD language, 8 is related to the strange quark mass m s , and 3 to the quark mass
difference m d − m u . The scalar densities are related to products of quark-anti-quark
field operators. For instance, the ratio of SU(3) to SU(2) breaking is given by
√
m s − m ud 3 8
R≡ = , (9.2)
md − mu 2 3

where m ud ≡ (m u + m d )/2. Before QCD many relations for quark-mass ratios

were obtained from hadron mass ratios, as well as from other hadronic infor-
mation, e.g. η → 3π , K l3 decay, etc. [5]. A pioneering determination of R, Eq.
(9.2), from a solution to the η → 3π puzzle proposed in [6], was obtained in [7],
R −1 = 0.020 ± 0.002, in remarkable agreement with a later determination based on
baryon mass splitting [8] R −1 = 0.021 ± 0.003, and with the most recent value [9]
R −1 = 0.025 ± 0.003.
With the advent of CHPT [2, 5, 9, 10], certain quark mass ratios turned out to be
renormalization scale independent to leading order, and could be expressed in terms
© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2018 59
C. A. Dominguez, Quantum Chromodynamics Sum Rules,
SpringerBriefs in Physics, [Link]
60 9 Quark Masses

of pseudoscalar meson mass ratios [5, 11, 12], e.g.

mu M 2 + − M 2 0 + 2Mπ2 0 − Mπ2 +
= K 2 K 2 = 0.56 , (9.3)
md M K 0 − M K + + Mπ2 +

ms M 2 + + M K2 0 − Mπ2 +
= K2 = 20.2 , (9.4)
md M K 0 − M K2 + + Mπ2 +

where the numerical results follow after some subtle corrections due to electromag-
netic self energies [9]. The current values of these ratios from the FLAG group are
[9]
mu
= 0.46 ± 0.03 , (9.5)
md
ms
= 27.43 ± 0.31 . (9.6)
m ud

Knowledge of these quark mass ratios is extremely important in the determination

of individual values of light-quark masses from QCD sum rules. The reason being
that the correlator of the axial-vector divergences, ψ5 (q 2 ), Eq. (1.5), employed for
this purpose involves as an overall multiplicative factor the terms (m u + m d ), or
(m ud + m s ).
Beyond leading order in CHPT things become complicated. At next to leading
order (NLO) the only parameter-free relation is

m 2s − m 2ud M K2 − Mπ2 M K2
Q2 ≡ = . (9.7)
m 2d − m 2u M K2 0 − M K2 + Mπ2

Other quark mass ratios at NLO and beyond depend on the renormalization scale,
as well as on some CHPT low energy constants which need to be determined inde-
pendently [9–11]. After taking into account electromagnetic self energies, Eq. (9.7)
gives [11] Q = 24.3, a recent analysis of η → 3π [10, 11] gives Q = 22.3 ± 0.8,
and the current value from the FLAG Collaboration [9] is

Q = 22.6 ± 0.7 ± 0.6 . (9.8)

The ratios R, Eq. (9.2), and Q, Eq. (9.7), together with the leading order ratios
Eqs. (9.3)–(9.4), will prove useful for a comparisons with QCD sum rule results. An
additional important quark mass ratio involving the ratios Eqs. (9.3)–(9.4) is

ms 2 m s /m d
rs ≡ = = 28.1 ± 1.3 , (9.9)
m ud 1 + m u /m d

where the numerical value follows from the NLO CHPT relation [11], to be compared
with the LO result from Eqs. (9.3)–(9.4), rs = 25.9, and a large Nc estimate [13]
9 Quark Masses 61

rs = 26.6 ± 1.6. The most recent FLAG Collaboration result is [9]

rs = 27.46 ± 0.15 ± 0.41 . (9.10)

In order to go beyond light-quark mass ratios and determine analytically their

individual values, including the charm and bottom quark masses, one needs suit-
able current correlators. In the light-quark sector (up-, down-, strange-quark) the
appropriate correlator is that of the divergence of the axial-vector current ψ5 (q 2 ),
Eq. (1.5). This is quite convenient as its QCD expression involves the square of the
quark masses as overall factor. For the heavy-quark masses the standard procedure
uses the vector current correlator which involves these masses in both the high energy,
as well as the low energy QCD expansions.
Starting with the light-quark masses, the case of the strange quark will be analysed
in full detail following its latest determinations [14–16]. The two pieces of the FESR
involving ψ5 (q 2 ) are the QCD part

1
δ5QC D (s0 ) ≡ − ds ψ5QC D (s) 5 (s) , (9.11)
2πi C(|s0 |)

and the hadronic part

 s0
1
δ5H AD (s0 ) ≡ 2 f P2 M P4 5 (M P2 ) + ds Im ψ5 (s)| R E S 5 (s) , (9.12)
sth π

where 5 (s) is an (analytic) integration kernel to be introduced shortly, the first term
on the right hand side is the pseudoscalar meson pole contribution (P = K ), sth
is the hadronic threshold, and text I m ψ5 (s)| R E S is the hadronic resonance spectral
function. The radius of integration s0 is assumed to be large enough for QCD to be
valid on the circle.
For later convenience the FESR, δ5 (s0 )| H AD = δ5 (s0 )| QC D can be rewritten as

δ5 (s0 )| QC D = δ5 | P O L E + δ5 (s0 )| R E S , (9.13)

where the meaning of each term is self evident. The full expression of ψ5QC D and its
second derivative are given in Appendix C, while the contour integrals, δ5QC D (s0 ),
are given in Appendix D. The quark-mass squared (m s + m q )2 , with m q = (m u +
m d )/2, enters in δ5 (s0 ) QC D as an overall factor. Hence, it is given by the ratio

δ5 (s0 )| H AD
(m s + m q )2 = (9.14)
δ̂5 (s0 )| QC D

where δ̂5 (s0 )| QC D obviously lacks the overall quark mass squared factor. The quark
mass m q stands for the running mass in the M S scheme. If one were to ignore the
hadronic resonance contribution in Eq. (9.12) then the quark mass following from
62 9 Quark Masses

(b)

(c)

(a)

Fig. 9.1 The strange quark mass m s (2 GeV) in the M S scheme taking into account only the kaon
pole with 5 (s) = 1 (curve (a)), and the two Breit-Wigner resonance spectral function with a
threshold constraint from CHPT, Eq. (6.5), with 5 (s) = 1 (curve (b)), and 5 (s) as in Eq. (9.15)
(curve (c)). A systematic uncertainty of some 20 % due to the resonance sector is dramatically
unveiled (difference between curves (b) and (c))

Eq. (9.14) would be a monotonically decreasing function of s0 , thus providing no

information. This is illustrated in Fig. (9.1) for 5 (s) = 1 (curve(a)).
Turning to the integration kernel, 5 (s), entering the FESR its purpose is to
quench the contribution of the hadronic resonances beyond the pole. Historically,
these resonances were predicted long ago [17, 18] as radial excitations of the pseu-
doscalar ground states, i.e. the pion and the kaon. The motivation in [17] was to
account for chiral symmetry breaking corrections in SU (2) × SU (2) (pion case)
and SU (3) × SU (3) (kaon case). The predicted masses of these resonances from
the dual-resonance model agreed well with the experimental values [19], once these
states were observed years later in hadronic interactions. However, information on
the existence and on the mass and width of these resonances is hardly enough to
reconstruct the hadronic spectral function. For instance, potential non-resonant back-
ground, inelasticity and resonance interference are impossible to predict. Hence,
while their presence is required, the lack of relevant information introduces a sys-
tematic uncertainty in the results. In order to try and quench this uncertainty the
integration kernel 5 (s) is introduced such that e.g. it vanishes at the peak of what-
ever functional form is chosen for the resonant states, e.g. a Breit-Wigner form.
Assuming two radial excitations the kernel becomes a quadratic function of the
squared energy,s, i.e.
5 (s) = 1 − a0 s − a1 s 2 , (9.15)

where the constant coefficients a0 and a1 are chosen such that

9 Quark Masses 63

5 (M12 ) = 5 (M22 ) = 0 (9.16)

where M1,2 are the masses of the resonances. An equivalent expression is given by
[16]
5 (s) = (s − M12 )(s − M22 ) , (9.17)

and two other kernels were considered in [16]

5 (s) = (s − M12 )(s − M22 )(s0 − s) , (9.18)

and
5 (s) = (s − a)(s − s0 ) , (9.19)

where the free parameter a was determined by demanding maximal reduction of the
uncertainty in the quark mass.
Regarding the resonance spectral functions, historically they were initially
parametrized by Breit-Wigner forms, albeit without any specific threshold normal-
ization. A procedure to normalize at threshold according to chiral perturbation theory
was eventually introduced in [20], thus reducing this systematic uncertainty. In the
case of the kaon channel the threshold behaviour is described in detail in Chap. 6,
with results given in Eqs. (6.11)–(6.13). It is important to notice the importance of the
sub-channel K ∗ (892) − π . Due to the narrow width of the K ∗ (892) this contribution
is not negligible. The two resonances entering the integration kernel 5 (s) are the
K(1460) and K(1830), both with widths of 250 MeV.
The crucial importance of the kernel, Eq. (9.15), is highlighted in Fig. 9.1. Curve
(b) is the result for the strange-quark mass with no integration kernel, i.e. with
5 (s) = 1, in which case both resonances contribute 100% to the mass. Given the
stability of the result one would have determined m s (2 GeV)  100 − 120 MeV,
with an unknown systematic uncertainty from the hadronic sector. The presence of
the integration kernel leads to curve (c), not only a more stable prediction, but one
unveiling the systematic uncertainty, and a result some 20% lower.
Turning to Eq. (9.14), in order to determine m s one needs information on the
quark mass ratio m q /m s , which is provided by e.g. chiral perturbation theory. All
the ingredients up to this point were used in [15] to find

95 ± 5 MeV
m s (2 GeV) = (9.20)
111 ± 6 MeV ,

corresponding to  QC D = 420 (330) MeV, respectively. While several sources of

uncertainty were considered in [15] to arrive at this result, including a conservative
guess of the unknown six-loop perturbative QCD contribution, the convergence of the
perturbative expansion was not explicitly taken into account. Furthermore, the use of
the strong running coupling αs involving the scale  QC D , Eq. (3.10), is responsible
for a rather large uncertainty in the quark mass. Theses issues were addressed in
64 9 Quark Masses

[14, 16] as follows. The strong running coupling, αs , was determined using Eq.
(3.9) together with its precise value at the Z -boson scale, scaled down to αs (s0 ). The
various integration kernels, Eqs. (9.15)–(9.19) were used with the results contributing
to the final uncertainty. Of essential importance, an analysis of the convergence of
the perturbative QCD (PQCD) expansion was performed as follows. Numerically,
the result for δ5P QC D using the integration kernel Eq. (9.17) with s0 = 4.2 GeV2 , and
√
the renormalization scale μ = s0 , is given by the expansion
 
δ5P QC D = 0.23 GeV8 1 + 2.2 αs + 6.7 αs2 + 19.5 αs3 + 56.5 αs4 , (9.21)

which after replacing a typical value of αs leads to all terms beyond the leading order
to be roughly the same, e.g. for αs = 0.3 the result is

δ5P QC D = 0.23 GeV8 [1 + 0.65 + 0.60 + 0.53 + 0.46] , (9.22)

which does not look at all convergent. To judge from the first five terms, this expansion
is worse behaved than the non-convergent harmonic series. Since the quark mass is
actually proportional to the inverse square-root of δ5QC D , after expanding the latter
gives instead
 
P QC D −1/2
(δ5 ) = 2.08 GeV−4 1 − 1.10 αs − 1.52 αs2 − 2.08 αs3 − 3.21 αs4 , (9.23)

which exhibits a considerably improved convergence, e.g. for αs = 0.3 this expansion
becomes

(δ5P QC D )−1/2 = 2.08 GeV−5 [1 − 0.33 − 0.14 − 0.06 − 0.03] . (9.24)

It is interesting to notice that this expansion of the inverse square-root is equivalent

to the lowest Padè approximant.
A thorough analysis of the various sources of uncertainty can be found in the
original reference [16]. The final result for the strange-quark mass at a scale μ =
2 GeV is
m s (2 GeV) = 94 ± 8 MeV . (9.25)

Turning to the up- and down-quark masses were determined in [21] to five-loop
order in PQCD, and using a quadratic integration kernel of the form Eq. (9.15).
However, the convergence of the PQCD expansion was not analysed as above for
the strange quark mass. This is currently an ongoing project being finalized [22].
Turning to the heavy-quark sector, the charm-quark mass determination [23] is
discussed next (the bottom-quark case is treated similarly in [24]). The starting point
is the choice of the heavy-quark vector current correlator
9 Quark Masses 65

1010

mc 3 GeV MeV 1000

990

980

970
14 16 18 20 22 24
s0 GeV2

Fig. 9.2 The charm-quark mass at a scale μ = 3 GeV, as a function of s0 , for the kernel Eq. (9.30)


μν (q ) = i d 4 x eiq x 0|T (Vμ (x) Vν (0))|0
2

= (qμ qν − q 2 gμν ) V (q
2
), (9.26)

where Vμ (x) = c̄(x)γμ c(x).

For reasons to become clear later the chosen FESR involves an meromorphic
integration kernel, p(s), so that the residue theorem in the complex s-plane is written
as
 s0 
1 1
p(s) I m V (s) ds = − p(s) V (s) ds
0 π 2πi C(|s0 |)
+ Res[ V (s) p(s), s = 0] , (9.27)

where
1
Im V (s) = Rc (s) , (9.28)
12π

with Rc (s) the standard R-ratio for charm production. The low energy expansion of
V (s) around s = 0 is given in Eq. (8.37), and the coefficients C n can be expanded
in powers of αs (μ) as in Eq. (8.38). At short distances the perturbative expansion of
V (s) is given in Eqs. (8.40)–(8.41). The experimental data for Rc (s), used in the
determination of the gluon condensate had been used in this application. Finally, two
integration kernels were chosen in [23]

1
p1 (s) = , (9.29)
s2
and
66 9 Quark Masses

s0 2
p2 (s) = 1 − , (9.30)
s
Given that this correlation function was used in the determination of the gluon
condensate in Chap. 8, it is important to mention that the contribution of this con-
densate in the determination of the charm-quark mass is negligible. In other words
there is no bootstrap.
It turns out that kernel p2 (s), Eq. (9.30) leads to a far more precise value of
the charm-quark mass (Fig. 9.2). Using this kernel, and after taking into account all
possible sources of uncertainty the quark mass at a scale of 3 GeV becomes

m c (3 GeV) = 987 ± 9 MeV , (9.31)

where this scale is chosen to allow for a direct comparison with LQCD results [9], with
which there is good agreement. This result agrees within errors with a determination
using a different integration kernel [25]. For other determinations see [26–29].
The determination of the bottom quark mass [24] follows closely that of the charm
quark so it will not be discussed here. The result is

m b (10 GeV) = 3623 ± 9 MeV , (9.32)

and
m b (m b ) = 4171 ± 9 MeV , (9.33)

in good agreement with LQCD [30].

References

1. S. B. Treiman, R. Jackiw, B. Zumino, E. Witten, Current Algebra and Anomalies (World

Scientific, Singapore, 1985). See also: J. Bernstein, Elementary Particles and their Currents
(W. H. Freeman & Co., San Francisco, 1968)
2. H. Pagels, Phys. Rep. 16, 219 (1975)
3. M. Gell-Mann, R.J. Oakes, B. Renner, Phys. Rev. 175, 2195 (1968)
4. S. Glashow, S. Weinberg, Phys. Rev. Lett. 20, 224 (1968)
5. J. Gasser, H. Leutwyler, Phys. Rep. C 87, 77 (1982)
6. C.A. Dominguez, A. Zepeda, Phys. Rev. D 18, 884 (1978)
7. C.A. Dominguez, Phys. Lett. B 86, 171 (1979)
8. P. Minkowski, A. Zepeda, Nucl. Phys. B 164, 25 (1980)
9. S. Aoki et al., Eur. Phys. J. C 77, 112 (2017)
10. G. Colangelo, S. Lanz, E. Passemar, PoS CD 09, 047 (2009)
11. H. Leutwyler, PoS CD 09, 005 (2009)
12. S. Weinberg, Trans. New York Acad. Sci. 38, 185 (1977)
13. H. Leutwyler, Nucl. Phys. B Proc. Suppl. 64, 223 (1998)
14. C. A. Dominguez, Analytical determination of QCD quark masses, in Fifty Years of Quarks, ed.
by H. Fritzsch, M. Gell-Mann (World Scientific Publishing Co., Singapore, 2015), pp. 287–313
15. C.A. Dominguez, N.F. Nasrallah, R. Röntsch, K. Schilcher, J. High Energy Phys. 05, 020
(2008)
References 67

16. S. Bodenstein, C.A. Dominguez, K. Schilcher, J. High Energy Phys. 07, 138 (2013)
17. C. A. Dominguez, Phys. Rev. D 7, 1252 (1973); ibid. D 16, 2320 (1977); Riv. Nuovo Cim.
8N6, 1 (1985)
18. P. Frampton, Phys. Rev. D 1, 3141 (1970)
19. K.G. Patrignani et al., Particle data group. Chin. Phys. C 40, 100001 (2016)
20. C. A. Dominguez, Z. Phys. C 26, 269 (1984)
21. C.A. Dominguez, N.F. Nasrallah, R.H. Rontsch, K. Schilcher, Phys. Rev. D 79, 014009 (2009)
22. C. A. Dominguez, A. Mes, K. Schilcher, to be published
23. S. Bodenstein, J. Bordes, C.A. Dominguez, J. Peñarrocha, K. Schilcher, Phys. Rev. D 83,
074014 (2011)
24. S. Bodenstein, J. Bordes, C.A. Dominguez, J. Peñarrocha, K. Schilcher, Phys. Rev. D 85,
034003 (2012)
25. S. Bodenstein, J. Bordes, C.A. Dominguez, J. Peñarrocha, K. Schilcher, Phys. Rev. D 82,
114013 (2010)
26. J.H. Kühn, M. Steinhauser, C. Sturm, Nucl. Phys. B 778, 192 (2007)
27. K.G. Chetyrkin et al., Phys. Rev. D 80, 074010 (2009)
28. B. Dehnadi, A.H. Hoang, V. Mateu, S.M. Zerbarjard, American Institute of Physics Conference
Proceedings, vol. 1441 (2012), p. 628
29. A. Signer, Phys. Lett. B 672, 333 (2009)
30. C. McNeile et al., PQCD collaboration. Phys. Rev. D 82, 034512 (2010)
Chapter 10
Corrections to the GMOR Relation

The Gell-Mann-Oakes-Renner (GMOR) relation was introduced in the Introduction,

Chap. 1, Eq. (1.14). In this section it will be discussed in more detail. While originally
obtained in the framework of current algebra and chiral symmetry, it is a low energy
theorem of QCD. It involves the correlator of the divergence of the QCD axial-vector
current, Eq. (1.7), leading to the low energy theorems

ψ5 (0)|du = −(m u + m d )0|ūu + d̄d|0 + O(m 2u,d ), (10.1)

in SU (2) × SU (2), and

ψ5 (0)|qs = −(m s + m q )0|s̄s + q̄q|0 + O(m 2s ), (10.2)

in SU (3) × SU (3), where m q ≡ (m u + m d )/2. Considering the correlator ψ5 (s) in

the complex squared energy s-plane, Fig. 10.1, from Cauchy’s theorem one has

1 ψ5 (s)
ψ5 (0) = ds. (10.3)
2π i s

In the hadronic sector there is a pole on the real axis corresponding to the pion or the
kaon, followed by higher resonances

2 f P2 M P4
ψ5 (s)| H AD = , + ψ5 (s)| Res , (10.4)
MP − s

where f P and M P are respectively the decay constant and the mass of the pion or
kaon, with f K / f π = 1.197 ± 0.006, f π = 92.21 ± 0.14 MeV [1]. Equation (10.3)
then becomes
j
ψ5 (0)|i ≡ −(m i + m j ) 0| q¯i qi + q¯j q j |0 = 2 f P2 M P2 (1 − δ P ), (10.5)

where i, j stand for the corresponding quark flavours, and δ P is the (hadronic) cor-
rection from the resonance contribution. This correction is rather important as it is

© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2018 69

C. A. Dominguez, Quantum Chromodynamics Sum Rules,
SpringerBriefs in Physics, [Link]
70 10 Corrections to the GMOR Relation

related to two of chiral perturbation theory (CHPT) low energy constants, L r8 and
H2r , i.e.

Mπ2 M K2
δπ = 4 (2L r8 − H2r ) and δK = δπ . (10.6)
f π2 Mπ2

The low energy constants L r8 and H2r also enter in the ratio s̄ s/q̄ q through

< s̄s > 8

Rsq ≡ = 1 + 3μπ − 2μ K − μη + 2 (M K2 − Mπ2 )(2L r8 + H2r ), (10.7)
< q̄q > fπ

where

M P2 M P2
μP = ln , (10.8)
32π 2 f π2 νχ2

with νχ the chiral renormalization scale.

In the past there was some debate on the feasibility of determining the hadronic
corrections δπ and δ K using QCDSR. This issue was fully resolved in the affirmative
in [2–4].
The determination of the hadronic corrections δπ and δ K involves Eq. (10.3).
Introducing an analytic kernel 5 (s), and splitting the integration into a line integral
along the positive real semi-axis, and a contour integral around a circle of radius |s0 |,
gives

 
1 s0
5 (s) 1 5 (s)
ψ5 (0)5 (0) = Im ψ5 (s) ds + ψ5 (s) ds, (10.9)
π sth s 2πi C(|s0 |) s

The integration kernel in the pionic channel was introduced in Eqs. (9.15)–(9.16),
and the threshold behaviour in the pionic channel is given in Eq. (6.2), while for the
kaon channel it is given in Eqs. (6.11)–(6.13) (Fig. 10.2).
A detailed analysis of the determination of δπ is given in [5], and the case of δ K
is discussed in [6]. The results for ψ5 (0) in both channels are very stable in the wide
range s0  3.0 − 5.0 GeV2 leading to

δπ = (6.2 ± 1.6)%, (10.10)

δ K = (55 ± 5)%, (10.11)

corresponding to
10 Corrections to the GMOR Relation 71

(a)
(b)

Fig. 10.1 Results for ψ5 (0) in SU (2) × SU (2) in units of 10−4 GeV4 as a function of s0 and using
a two-resonance parametrization. Curve (a) corresponds to αs (Mτ2 ) = 0.335 ( = 365 MeV), and
curve (b) to αs (Mτ2 ) = 0.353 ( = 397 MeV)

Fig. 10.2 Results for ψ5 (0) in SU (3) × SU (3) in units of 10−3 GeV4 as a function of s0 , and using
a two-resonance parametrization. Curve (a) corresponds to αs (Mτ2 ) = 0.335 ( = 365 MeV), and
curve (b) to αs (Mτ2 ) = 0.353 ( = 397 MeV)

ψ5 (0)|du = (3.11 ± 0.04) × 10−4 GeV4 , (10.12)

−3
ψ5 (0)|ud
s = (2.8 ± 0.3) × 10 GeV4 . (10.13)

These results together with Eqs. (10.6)–(10.7) determine the low energy constants
L r8 and H2r

L r8 = (1.0 ± 0.3) × 10−3 , (10.14)

72 10 Corrections to the GMOR Relation

H2r = −(4.7 ± 0.6) × 10−3 . (10.15)

The result for L r8 is in very good agreement with several Lattice QCD determinations
[7], and the unphysical H2r is in good agreement with the estimate [8]

H2r (νχ = Mρ ) = −(3.4 ± 1.5) × 10−3 . (10.16)

Regarding the quark-mass corrections indicated in Eqs. (10.1) and (10.2), they are
negligible in comparison with the hadronic corrections so that they can be safely
ignored.
For a more thorough discussion of this subject the reader is referred to the original
references [5, 6].

References

1. K.G. Patrignani et al., Particle data group. Chin. Phys. C 40, 100001 (2016)
2. K.G. Chetyrkin, C.A. Dominguez, D. Pirjol, K. Schilcher, Phys. Rev. D 51, 5090 (1995); K.G.
Chetyrkin, D. Pirjol, K. Schilcher, Phys. Lett. B 404, 337 (1997). C.A. Dominguez, L. Pirovano,
and K. Schilcher. Phys. Lett. B 425, 193 (1998)
3. D.J. Broadhurst, Phys. Lett. B 101, 423 (1981)
4. H. Leutwyler, Private communication (2013)
5. J. Bordes, C.A. Dominguez, P. Moodley, J. Penarrocha, K. Schilcher, J. High Ener. Phys. 1005,
064 (2010)
6. J. Bordes, C.A. Dominguez, P. Moodley, J. Penarrocha, K. Schilcher, J. High Ener. Phys. 1210,
102 (2012)
7. S. Aoki et al., Eur. Phys. J. C 77, 112 (2017)
8. M. Jamin, Phys. Lett. B bf 538, 71 (2002)
Chapter 11
Anomalous Magnetic Moment
of the Muon

The anomalous magnetic moment of the muon, due to quantum corrections, is related
to the g-factor as
g−2
aμ ≡ = 0. (11.1)
2
The theoretical prediction of the muon anomaly, aμ , currently disagrees with ex-
periment at the level of some 3 σ . The largest uncertainty in the Standard Model
prediction comes from the leading order hadronic contribution, aμHAD,LO [1], to be
discussed in this chapter. The traditional procedure to determine this quantity involves
the e+ e− → hadr ons data, leading to [2]

aμHAD = (693.1 ± 3.4) × 10−10 . (11.2)

After adding the rest of the contributions to the anomaly, leading to the Standard
Model (SM) result, aμSM , the difference with the direct measurement, aμEXP , is

aμEXP − aμSM = (26.8 ± 7.6) × 10−10 , (11.3)

a 3.5 σ discrepancy. This result has prompted a plethora of attempts over the years
to explain this discrepancy beyond the SM. Before entertaining such an extreme
possibility one should attempt to predict the anomaly entirely from theory, i.e. QCD.
This proposition is also supported by the fact that the e+ e− data base leading to Eq.
(11.3) comprises many different experiments, at various facilities, performed over
the years, and involving a large number of different final states. For a detailed and
critical analysis of this data base see [3].
An attempt to account for the hadronic muon anomaly entirely within QCD was
made in [4, 5] and discussed in the sequel.
The standard expression for the lowest order muon anomaly is given by

© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2018 73

C. A. Dominguez, Quantum Chromodynamics Sum Rules,
SpringerBriefs in Physics, [Link]
74 11 Anomalous Magnetic Moment of the Muon

 ∞
α 2E M ds
aμHAD = K (s) R(s), (11.4)
3 π2 sth s

where α E M is the electromagnetic coupling, and the standard R -ratio is


R(s) = 3 Q 2f [8 π Im (s)] , (11.5)
f

where Q f are the quark charges and (s) is the vector current correlator, Eq. (8.6),
normalized to 8 π Im (s) = [1 + αs /π + · · · ] (notice dropped upper label EM).
The integration kernel K (s) in Eq. (11.4) is given by [6]
 1
x 2 (1 − x)
K (s) = dx 2 , (11.6)
0 x + ms2 (1 − x)
μ

where m μ is the muon mass. The traditional approach to determine aμHAD is to split
the integration region in Eq. (11.6) into a low-energy piece from threshold up to
s = s0  (1.8 GeV)2 , and a a high energy contribution from s = s0 to infinity. The
low energy integral involves the data (e+ e− annihilation or tau-lepton decay into
hadrons), while the high energy integral is computed in perturbative QCD (PQCD).
For convenience it is useful to split the contributions to the integral in Eq. (11.4) into
the three main regions, dominated by the light-, charm- and bottom-quark

aμH AD = aμH AD |uds + aμH AD |c + aμH AD |b . (11.7)

While the integration kernel K (s) is known, and given in Eq. (11.6), a fit to it in terms
of a meromorphic function allows using the power of the complex squared energy
plane to determine the anomaly entirely from QCD . For this to rival the data driven
determinations, the fit function must be extremely accurate. This is indeed possible
thanks to the shape of K (s), to wit.
Starting with the light-quark sector, the optimal fit function K 1 (s) is given by

K 1 (s) = 2.257 × 10−5 s + 3.482 × 10−3 s −1

− 1.467 × 10−4 s −2 + 4.722 × 10−6 s −3 , (11.8)

where s is expressed in GeV2 , and the numerical coefficients have the appropriate
units to render K 1 (s) dimensionless. The functions K (s) and K 1 (s) are shown in
Fig. 11.1. The difference between the two functions is less than 0.08 % in the whole
s-range!.
Next, in the charm-quark region s1  M J/ψ
2
≤ s ≤ s2  (5.0 GeV)2 , the optimal
fit function K 2 (s) is
11 Anomalous Magnetic Moment of the Muon 75

0.030

0.025

0.020

0.015
Ks

0.010

0.005

0.000
0.5 1.0 1.5 2.0 2.5 3.0
s GeV2

Fig. 11.1 The exact kernel K (s), Eq. (11.6) (solid line) together with the fit in the light-quark
region K 1 (s), Eq. (11.8), (solid circles). Relative difference between the two is less than 0.08%

a1 a2
K 2 (s) = + 2, (11.9)
s s

where a1 = 0.003712 GeV2 and a2 = −0.0005122 GeV4 , providing an excellent fit,

with K 2 (s) differing from the exact kernel K (s) by less than 0.02% in the whole
range. For the bottom-quark region the fit function is given by

K 3 (s) = 0.003719 GeV2 s −1 − 0.0007637 GeV4 s −2 , (11.10)

differing from the exact kernel, K (s), by less than 0.0005 % in the range Mϒ2 ≤ s ≤
(12 GeV)2 .
The QCD sum rule in the light-quark region is given by
  
s0
ds 1 K 1 (s)
K 1 (s) Im uds (s) = Res uds (s)
sth s π s
 s=0
1 ds
− K 1 (s) uds (s), (11.11)
2πi |s|=s0 s

where the contour integral around the circle of radius s0  (1.8 GeV)2 , is computed
using PQCD. This is known up to five-loop level, Eq. (8.14). The contour integration
can be performed using fixed order perturbation theory (FOPT) or, alternatively,
contour improved perturbation theory (CIPT). In the present case the difference
between the two integration methods turns out to be negligible, as discussed later.
The residues are given in terms of derivatives of the correlator at zero momentum,
which in principle can be determined in LQCD [5]. Hence Eq. (11.4) in this sector
becomes
76 11 Anomalous Magnetic Moment of the Muon

   
K 1 (s)
aμH AD |uds = 8 α 2E M Q i2 Res uds (s)
i=u,d,s
s s=0

1 ds
− K 1 (s) uds (s)|PQCD
2πi |s|=s0 s
 ∞ 
ds 1
+ K (s) Im uds (s)|PQCD , (11.12)
s0 s π

where the last integral above involves the exact integration kernel K (s). Analogous
expressions follow in the charm-quark and bottom-quark sector. An important differ-
ence is that in these cases the residues can be obtained directly from the low energy
expansion of the vector correlator [4]. The result in the charm-quark region is
 
K 2 (s)
Res c (s)| P QC D = 76.1(5) × 10−7 , (11.13)
s s=0

where the error is due to the uncertainty in αs and to the truncation of PQCD. In the
bottom-quark region the residue is
 
K 3 (s)
Res b (s)| P QC D = 6.3 × 10−7 , (11.14)
s s=0

where the error is negligible.

Turning to the integrals in Eq. (11.12) the contour integrals in fixed order pertur-
bation theory and in the three regions are
⎧
 ⎨ 135.3(6) × 10−7
1 ds
K n (s) q (s)|PQCD = 20.3(1) × 10−7 (11.15)
2πi s ⎩
|s|=s0 3.6(2) × 10−7 ,

for n = 1, 2, 3 and q = uds, c, b, respectively. The result from integrating in contour

improved perturbation theory differs by 0.2%. A comparison of these results with
those from using e+ e− annihilation data from the BES Collaboration [7] shows
excellent agreement.
The line integrals in Eq. (11.12) in the three regions are
⎧
 ∞ ⎨ 151.8(1) × 10−7
ds 1
K (s) Im q (s)| P QC D = 20.0(4) × 10−7 (11.16)
s π ⎩
sj 3.4(2) × 10−7

with j = 0, 2, 4 corresponding to the regions q = uds, c, b, respectively. Substitut-

ing the numerical values of these two integrals into Eq. (11.12), and into its corre-
11 Anomalous Magnetic Moment of the Muon 77

sponding expressions for the charm- and bottom-quark regions, the leading order
hadronic anomaly becomes
 
16 2 K 1 (s)
aμH AD = α Res uds (s) + 19.4(2) × 10−10 , (11.17)
3 EM s s=0

where the Cauchy residue in the light-quark sector is discussed below. The individual
contributions in the charm- and bottom-quark regions are fully determined after using
Eqs. (11.13)–(11.14) with the results

aμH AD |c = 14.4(1) × 10−10 , (11.18)

aμH AD |b = 0.29(1) × 10−10 . (11.19)

A crucial vindication of this approach was provided later by LQCD determinations

of these heavy-quark contributions [8, 9]

aμH AD |c = 14.42(39) × 10−10 , (11.20)

aμH AD |b = 0.271(37) × 10−10 , (11.21)

in excellent agreement with Eqs. (11.18)–(11.19).

Turning to the Cauchy residue in Eq. (11.12), given the expression of K 1 (s), Eq.
(11.8), it can be written as
  3
K 1 (s) an d n
Res uds (s) = lim uds (s), (11.22)
s s=0 s→0
n=1
n! ds n

where the an are the coefficients of the s −1 , s −2 and s −3 terms in Eq. (11.8), respec-
tively. It is important to notice that the term proportional to s −1 in K 1 (s), Eq. (11.8),
is positive, while that proportional to s −2 is negative. Hence, there is a partial cancel-
lation among the first two terms in the residue, Eq. (11.22). The first two derivatives
can be calculated in Lattice QCD from the slope of the vector current correlator.
Results from the first determination are [10]
d
(s)uds |s=0 = 0.07190 ± 0.0025 GeV−2 , (11.23)
dq 2

d2
(s)uds |s=0 = 0.136 ± 0.0009 GeV−4 (11.24)
(dq 2 )2

where these values correspond to the definition and normalization of the correlator
as in Eq. (11.5), which differs from that in [10] by a factor 3/4. After substituting
these results into Eq. (11.22) the residue becomes
78 11 Anomalous Magnetic Moment of the Muon
 
K 1 (s)
Res uds (s) = (0.240 ± 0.009) × 10−3 (11.25)
s s=0

so that the total value of the anomaly is

aμH AD = (701 ± 26) × 10−10 , (11.26)

a result in need of considerable improvement, which should eventually be achieved

in the future. In the meantime the residue can be obtained from the leading or-
der hadronic saturation of the vector correlator, plus the Vector Meson Dominance
(VMD) approximation to the pion form factor, as follows
 
d3 p
μν (q 2 ) = i d 4 xeiq x 0|Vμ (x)|π π |π π |Vν† (0)|0
2 p0 (2π )3

+ 0|Vμ (x)|ρ |ρ|Vν† (0)|0 · · · , (11.27)

The neglected terms correspond to multi-pion states. The first term above is loop-
suppressed with respect to the second term, i.e. the single ρ-meson Born term. The
matrix element in the second term above defines the ρ-photon coupling constant f ρ
according to

Mρ2
0|Vμ (0)|ρ( p, s) = μ (11.28)
fρ

where μ is the ρ-meson polarization vector which satisfies the completeness relation
 pμ pν
μ ( p, s) ν ( p, s) = −gμν + ≡ μν ( p). (11.29)
s
Mρ2

Substituting Eq. (11.29) into Eq. (11.27), and after integrating over the three-
momentum one finds

Mρ2 1
μν (q 2 ) = (qμ qν − q 2 gμν ) . (11.30)
f ρ2 (q 2 − Mρ2 + ı )

The electromagnetic form factor of the pion, Fπ (q 2 )|VMD , in VMD is given by

gρππ Mρ2 Mρ2

Fπ (q 2 )|VMD = = (11.31)
f ρ (Mρ2 − q 2 ) (Mρ2 − q 2 )

with Fπ (0) = 1, and gρππ  f ρ from data, so that μν (q 2 ) becomes

1
μν (q 2 ) = − (qμ qν − q 2 gμν ) Fπ (q 2 ). (11.32)
f ρ2
11 Anomalous Magnetic Moment of the Muon 79

The derivative of the pion form factor is related to the root-mean-squared radius so
that
d 1 1  2
(s)uds |s=0 = r = 0.076 GeV−2 , (11.33)
dq 2 6 f ρ2 π

somewhat larger than the LQCD result, Eq. (11.23), and leading to

aμH AD = (775 ± 14) × 10−10 . (11.34)

Subtracting a potential 5% contribution from the second derivative of the vector

correlator (sign is negative for this term) would give

aμH AD = (736 ± 14) × 10−10 , (11.35)

a value considerable larger than that obtained entirely from e+ e− data [2]

aμH AD |e+ e− = (693.1 ± 3.4) × 10−10 . (11.36)

A far better model for the pion form factor is the Dual-QCD∞ model involving
an infinite number of zero width resonances, as in QCD∞ , with masses and cou-
plings given by the dual resonance model [11, 12]. The expression of the light-quark
correlator is given by

1 1 (β − 1/2)
uds (s)|QCD∞ = √ B(β − 1, 1/2 − s/2Mρ2 ), (11.37)
f ρ2 π (β − 1)

where β is a free parameter and B(x, y) is the Euler beta-function. With β = 2.30
one obtains an excellent fit to the data up to q 2 = −10 GeV2 , with a chi-squared
per degree of freedom χ 2 = 1.5 [11], and more importantly, a root-mean-square
pion radius < rπ2 >= 0.436 ± 0.004 fm2 [11] to be compared with the most recent
experimental value [13] < rπ2 >= 0.439 ± 0.008 fm2 . The result for the anomaly
in this model is

aμH AD |QCD∞ = (722 ± 9) × 10−10 , (11.38)

in agreement with the LQCD prediction, Eq. (11.26), but not with the value from
using e+ e− data, Eq. (11.36). This result already incorporates the first and the second
derivatives of the vector correlator.
80 11 Anomalous Magnetic Moment of the Muon

References

1. F. Jegerlehner, Springer Tracts Mod. Phys. 274, 1 (2017)

2. M. Davier, A. Hoecker, B. Malaescu, Z. Zhang, Eur. Phys. J. C 77, 827 (2017)
3. S. Bodenstein, C.A. Dominguez, S.I. Eidelman, H. Spiesberger, K. Schilcher, J. High Energy
Phys. 01, 039 (2012)
4. S. Bodenstein, C.A. Dominguez, K. Schilcher, H. Spiesberger, Phys Rev. D 85, 014029 (2012)
5. C.A. Dominguez, H. Horch, B. Jäger, N.F. Nasrallah, K. Schilcher, H. Spiesberger, H. Wittig,
Phys Rev. D 96, 074016 (2017)
6. S.J. Brodsky, E. de Rafael, Phys. Rev. 168, 1620 (1968)
7. J.Z. Bai et al. BES Coll., Phys. Rev. Lett. 88, 101802 (2002); M. Ablikim et al. BES Coll.
Phys. Lett. B 677, 239 (2009)
8. B. Chakraborty et al. Phys. Rev. D 89, 114501 (2014); J. Koponen et al. Proc. of Sci. LATTICE-
2014, 129 (2015)
9. B. Colquohoun et al., Phys. Rev. D 91, 074514 (2015)
10. Sz. Borsanyi et al. Phys. Rev. D 96, 074507 (2017)
11. C.A. Dominguez, Phys. Lett. B 512, 331 (2001); C.A. Dominguez, T. Thapedi, J. High Energy
Phys. 10, 003 (2004); C.A. Dominguez, R. Röntsch, ibid. 10, 085 (2007)
12. C. Bruch, A. Khodjamirian, J.H. Kühn, Eur. Phys. J. C 39, 41 (2005)
13. S.R. Amendolia et al., Nucl. Phys. B 277, 168 (1986)
Chapter 12
QCD Sum Rules at Finite Temperature

The extension of Finite Energy QCD sum rules (FESR) to finite temperature, and
its many applications, has been reviewed recently in [1]. Hence, this section will be
brief.
Introducing a finite temperature in quantum field theory implies the presence of
a medium, the thermal bath, thus in principle breaking Lorentz covariance. To avoid
this situation, and continue using covariant expressions for current correlators, one
chooses a frame at rest with respect to the medium. This is to be understood in the
sequel.
The starting point is the qualitative expectation on the behaviour of hadronic
spectral functions with increasing temperature. Figure 12.1 shows a typical hadronic
spectral function at T = 0 in the complex squared-energy, s-plane. It comprises a
hadronically stable hadron (pole on the real s-axis), followed by a few resonances
(poles on the second Riemann sheet), with hadronic widths increasing with energy.
The pole gives a spectral function

Im (s)|POLE = fP2 δ(s − −MP2 ), (12.1)

where fP is the hadron-current coupling, and MP its mass. The resonances, if narrow
enough, could be parametrized e.g. by a Breit-Wigner form

MR3 R
Im (s)|RES = fR2 , (12.2)
(s − MR2 )2 + MR2 R2

where fR is the resonance-current coupling entering a correlation function, MR its

mass, and R its (hadronic) width. At finite temperature this spectrum is expected to
change dramatically with increasing T . An example is the pion pole with fP ≡ fπ ,
which will acquire a width (interpreted as absorption in the thermal medium). The
resonance widths should increase monotonically with temperature, and eventually
become so large that there would be no trace of them in the spectrum. As this happens
the hadron-current coupling should decrease with increasing T , approaching zero.
Finally, the onset of perturbative QCD, i.e. the continuum, parametrized by s0 (the
radius of the integration contour in the complex s-plane) will decrease and approach
the origin (Fig. 12.1). This scenario completes the qualitative description of quark-
gluon deconfinement at finite temperature
© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2018 81
C. A. Dominguez, Quantum Chromodynamics Sum Rules,
SpringerBriefs in Physics, [Link]
82 12 QCD Sum Rules at Finite Temperature

Fig. 12.1 Typical hadronic

spectral function comprising
a pole and three resonances.
Curve a is at T=0, while
curve b is at finite
temperature

The starting point of the thermal QCD program is to allow for a temperature
dependence in the quark propagator. There are two different procedures to achieve
this, the Dolan-Jackiw [2], and the Matsubara formalism [3], equivalent at the one-
loop level. The former leads to a straightforward extension of the QCD sum rule
program through a simple correction to the fermion or boson propagator

i 2π
SF (k, T ) = −  |k 0 |/T  (k/ + m) δ(k 2 − m2 ), (12.3)
k/ − m e +1

and an equivalent expression for bosons, except for a positive relative sign between
the two terms above, and the obvious replacement of the Fermi by the Bose thermal
factor. An advantage of this expression is that it allows for a straightforward calcula-
tion of the imaginary part of current correlators, which is the function entering QCD
sum rules. There is an issue at finite T , i.e. the potential presence of a space-like
contribution to the imaginary part (absent at T = 0), to be discussed in Appendix E.
The extension of the QCD sum rule method to finite temperature was first proposed
by Bochkarev and Shaposhnikov [4] in 1986, in the framework of Laplace transform
QCD sum rules. One of the main problems with this kind of sum rules at finite T is
that the role of s0 is exponentially suppressed. While this might not have much impact
at T = 0, it is definitely an undesired feature at T = 0. Since s0 signals the end of
the resonance region, and the threshold for QCD, and given the scenario shown in
Fig. 12.1, its thermal behaviour should be enhanced rather than suppressed. For this
reason thermal Finite Energy QCD sum rules (FESR) were proposed in [5, 6] and
used over the years in a large number of applications (for a recent comprehensive
review see [1]). A major breakthrough regarding the role of s0 (T ) took place recently
with the establishment of a relation between this parameter and the Polyakov loop
of Lattice QCD, which signals deconfinement [7].
A sample of results from thermal FESR is as follows. The T-dependence of the
ratio s0 (T )/s0 (0) from a FESR in the vector (ρ) channel, signalling deconfinement, is
shown in Fig. 12.2 normalized to its T = 0 value (solid curve), together with the same
ratio in the axial-vector channel (dash curve), signalling chiral symmetry restoration.
The 10% difference at the end-point is within the accuracy of the method. The thermal
12 QCD Sum Rules at Finite Temperature 83

Fig. 12.2 The normalized thermal behaviour of s0 (T ) in the vector (ρ)-channel (solid curve), and
in the axial-vector channel (dotted curve) [8]

Fig. 12.3 The normalized thermal behaviour of the ρ-meson width [8]

behaviour of the normalized width of the ρ-meson [8] is shown in Fig. 12.3. The rise
of the width near the deconfinement temperature is rather dramatic.
In the heavy-quark sector, i.e. charm- and bottom-quark region, results from
thermal QCD sum rules turned out quite unexpected [9, 10]. Indeed, the thermal
behaviour of s0 (T ), (T ), and f (T ) clearly show that these states survive at the
critical temperature for deconfinement. In fact, the deconfinement parameter, s0 (T ),
begins to decrease near Tc , but eventually flattens out at T /Tc  0.6 up to T /Tc  1.2
beyond which there is no support to the integrals in the FESR. The width, while ini-
tially increasing substantially with temperature, it begins to decrease at and beyond
Tc . Finally, the coupling remains constant up to close to Tc , and then shoots up.
This scenario is further supported by results in the pseudoscalar (ηc ) and scalar (χc )
84 12 QCD Sum Rules at Finite Temperature

Fig. 12.4 Hadronic width of J /ψ as a function of T /Tc . The sum rules have no support beyond a
certain temperature, normally for T /Tc  1. Notice the exceptional behaviour for T /Tc  1

charmonium channels [11], as well as in vector bottonium (ϒ), and pseudoscalar

bottonium (ηc ) [10].
Later determinations from Lattice QCD in the bottonium channel [12, 13] have
confirmed this survival. These results put to rest three decades of popular belief in the
opposite scenario, i.e. charmonium early melting [14]. In Fig. 12.4 the behaviour of
the hadronic width of the J /ψ is shown as a function of T /Tc . The thermal behaviour
of the other two parameters, the deconfinement parameter s0 (T ), and the coupling,
fV (T ), is also indicative of survival, i.e. s0 (T ) does not approach zero, and fV (T )
grows substantially with increasing temperature.
The last issue concerns di-muon production in heavy-ion collisions in the energy
region around the ρ-meson resonance. The theoretical issues of this process received
some attention [18, 19] well before experiments were conducted [15, 16]. In this
energy region the ρ-meson is expected to play a fundamental role. It actually enters
the di-muon production rate through the pion electromagnetic form factor, and was
considered in [19], albeit with hadronic parameters independent of temperature.
The temperature dependence of the pion form factor was first proposed in [20], and
rediscovered much later in [21, 22]. While the expression for the thermal width of the
ρ-meson in [20] was somewhat crude, the conclusion was not, i.e. that there would
be an important change in the di-muon production rate. More recently, given that
the ρ-meson parameters at finite T had been obtained from FESR [1, 8], it became
possible to obtain a parameter-free prediction of this rate. This is shown in Fig. 12.5.
Given that the ρ-meson thermal parameters were already determined from the FESR
in the light vector channel, the parameter-free prediction is exceptional.
References 85

Fig. 12.5 The dimuon invariant mass distribution in In+In collisions in the region of the ρ-meson
using pre-determined values of thermal parameters from QCDSR [1, 8] (solid curve). Dash curve is
for all ρ-parameters independent of T . Data is from [15, 16]. Results are for μ = 0. Finite chemical
potential results change slightly in off-peak regions (see [17])

References

1. A. Ayala, C.A. Dominguez, M. Loewe, Adv. High Energy Phys. 2017, 9291623 (2017)
2. L. Dolan, R. Jackiw, Phys. Rev. D 9, 3320 (1974)
3. M. Le Bellac, Thermal Field Theory, (Cambridge University Press, United Kingdom, 1996)
4. A.I. Bochkarev, M.E. Shaposhnikov, Nucl. Phys. B 268, 220 (1986)
5. C.A. Dominguez, M. Loewe, Phys. Lett. B 233, 201 (1989)
6. A. Barducci, R. Casalbuoni, S. de Curtis, R. Gatto, G. Pettini, Phys. Lett. B 240, 429 (1990)
7. J.P. Carlomagno, M. Loewe, Phys. Rev. D 95, 036003 (2017)
8. A. Ayala, C.A. Dominguez, M. Loewe, Y. Zhang, Phys. Rev. D 86, 114036 (2012)
9. C.A. Dominguez, M. Loewe, J.C. Rojas, Y. Zhang, Phys. Rev. D 81, 014007 (2010)
10. C.A. Dominguez, M. Loewe, Y. Zhang, Phys. Rev. D 88, 054015 (2013)
11. C.A. Dominguez, M. Loewe, J.C. Rojas, Y. Zhang, Phys. Rev. D 83, 034033 (2011)
12. G. Aarts et al., J. High Ener. Phys. 03, 084 (2013)
13. G. Aarts et al., J. High Ener. Phys. 12, 064 (2013)
14. T. Matsui, H. Satz, Phys. Lett. B 178, 416 (1986)
15. R. Arnaldi et al., NA60 collaboration. Phys. Rev. Lett. 96, 162302 (2006)
16. S. Damjanovic, NA60 collaboration. Eur. Phys. J. 61, 711 (2009)
17. A. Ayala, C.A. Dominguez, L.A. Hernandez, M. Loewe, A.J. Mizher, Phys. Rev. D 88, 114028
(2013)
18. G. Domokos, J.I. Goldman, Phys. Rev. D 32, 1109 (1985)
19. J. Cleymans, J. Fingberg, K. Redlich, Phys. Rev. D 35, 2153 (1987)
20. C.A. Dominguez, M. Loewe, Z. Phys. C, Particles & Fields, 4(9), 423 (1991)
21. V.L. Eletsky, B.L. Ioffe, J.I. Kapusta, Nucl. Phys. A 642, 155 (1998)
22. V.L. Eletsky, B.L. Ioffe, J.I. Kapusta, Eur. Phys. J. A 3, 381 (1998)
Chapter 13
Summary and Outlook

In this book the current method of QCD sum rules, Finite Energy Sum Rules (FESR),
has been introduced and reviewed. This approach is based on the relation between
QCD and hadronic physics, established by invoking Cauchys residue theorem in the
complex squared-energy s-plane [1]. In this plane QCD information is present on
the circular contour of radius s = |s0 | (see Fig. 1.1, Sect. 1), while hadronic physics
is formulated along the positive real semi-axis. The objects considered in this plane
are current correlators, e.g. Eq. (1.5), having both a QCD as well as a hadronic
representation in terms of their respective quantum fields. This method complements
Lattice QCD, achieving comparable precision in many important instances.
High precision applications are mostly restricted to two-point functions, as they

are currently known up to O(α∫ ) in perturbative QCD. In contrast, there is no
information available on radiative corrections to three-point functions. Even at next-
to-leading order, O(αs ), these corrections would lead to considerable improvement
of current predictions of form factors. For instance, the electromagnetic form fac-
tors of the proton, F1,2 (q 2 ), determined from a three-point function sum rule, shows
only the electric form factor, F1 (q 2 ), to agree well with data [2]. In contrast, the
form factor F2 (q 2 ) disagrees with data by a factor two. The reason could well be
the absence of radiative corrections. Nevertheless, it would be important to under-
stand why three-point function sum rules for mesons, instead of baryons, do lead
to reasonable predictions, e.g. for the pion electromagnetic form factor [3, 4], the
axial-vector coupling of the nucleon [5], the ρ − π π strong interaction coupling [4],
and the strong ωρπ coupling [6].
A topic of current interest, omitted here due to space limitations, is that of the
leptonic decay constants of heavy pseudoscalar mesons, such as the D-, Ds -, B-,
Bs - and Bc -mesons. A pioneering determination using Hilbert moment QCD sum
rules [7] was followed by many determinations over the years (for recent results
see [8, 9] and references therein). The leptonic decay constant of the Bc meson
remains a contentious issue, as results from QCD sum rules differ substantially
among each other.
Current state of the art determinations from FESR include the values of the light
and heavy quark masses (except the top-quark). Improvement in precision could be
achieved by employing series convergence methods. For instance, it has been shown

© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2018 87

C. A. Dominguez, Quantum Chromodynamics Sum Rules,
SpringerBriefs in Physics, [Link]
88 13 Summary and Outlook

that Padè approximants do improve convergence of the perturbative series [10].

However, there is more than this particular choice in the literature to be explored.
A topic of current high importance is that of the leading hadronic contribution
to the anomalous magnetic moment of the muon, g − 2, as discussed in Chap. 11.
The QCD sum rule approach allows for an entirely theoretical prediction of this
quantity, provided the first two derivatives of the vector current correlator at the
origin are known. It is important to stress that these derivatives can be obtained
analytically entirely from QCD for the charm- and bottom-quark regions. There is
perfect agreement between the QCD sum rule value of the anomaly in these regions
and LQCD results. A model independent determination in the light-quark sector is
only possible in the framework of Lattice QCD. Current precision in this framework
[11] is not enough to draw definite conclusions regarding the need for Physics beyond
the Standard Model (SM). It is only through relations between the vector current
correlator and the pion electromagnetic form factor that these derivatives can be
estimated. Results from this approach at the moment are consistent with the SM, but
disagree with results from using data on electron-positron annihilation into hadrons.
Potential problems with this data base have been pointed out in Sect. 8, based on the
analysis of [12].
An important extension of the QCD sum rule method is to consider finite temper-
ature, as first proposed in [13], and briefly discussed in Sect. 12. A recent comprehen-
sive review [14] should be consulted. Perhaps the most successful and unexpected
result was the prediction of the survival of charmonium and bottonium states at, and
beyond the critical temperature for deconfinement [15, 16]. This prediction was later
confirmed by Lattice QCD results in the bottom-quark sector [17, 18]. Such a sce-
nario contradicts and lays to rest a decades long popular expectation of charmonium
melting at or even below the critical temperature [19]. Occam’s razor applies.
It is also possible to formulate QCD sum rules at finite temperature as well as
baryon chemical potential [20]. This allows for an exploration of the QCD phase
diagram and the study of the phase transitions of chiral symmetry restoration and
deconfinement. Regarding the latter it is important to highlight a recent result [21]
showing a relation between the QCD sum rule deconfinement parameter s0 (T ) (the
onset of perturbative QCD) and the Polyakov loop, i.e. the LQCD deconfinement
parameter. This unexpected relation provides a fundamental validation of the thermal
QCD sun rule method.
The current research direction of finite temperature QCD is to combine it with the
presence of external strong magnetic fields. At present there is a substantial amount of
information in the framework of hadronic models, e.g. Linear Sigma Model, Nambu-
Jona-Lasinio Model, etc. (for reviews see e.g. [22, 23]). On the theory side the first
QCD sum rule attempt at including a magnetic field was done at zero temperature
[24]. The addition of temperature dependence is currently work in progress.
For background material on QCD sum rules the reader may consult the textbooks
[25, 26].
References 89

References

1. R. Shankar, Phys. Rev. D 15, 755 (1977)

2. H. Castillo, C.A. Dominguez, M. Loewe, J. High Ener, Phys. 0503, 012 (2005)
3. B.L. Ioffe, A.V. Smilga, Nucl. Phys. B 216, 373 (1983)
4. C.A. Dominguez, M.S. Fetea, M. Loewe, Phys. Lett. B 406, 149 (1997)
5. C.A. Dominguez, M. Loewe, C. van Gend, Phys. Lett. B 460, 442 (1999)
6. C.A. Dominguez, M. Loewe, Phys. Lett. B 481, 295 (2000)
7. C.A. Dominguez, N. Paver, Phys. Lett. B 197, 423 (1987)
8. M.J. Baker, J. Bordes, C.A. Dominguez, J. Peńarrocha, K. Schilcher, J. High Ener. Phys. 07,
032 (2014)
9. W. Lucha, D. Melikhov, S. Simula, Phys. Rev. D 88, 056011 (2013)
10. S. Bodenstein, C.A. Dominguez, K. Schilcher, J. High Ener. Phys. 07, 138 (2013)
11. C.A. Dominguez, H. Horch, B. Jäger, N.F. Nasrallah, K. Schilcher, H. Spiesberger, H. Wittig,
Phys Rev. D 96, 074016 (2017)
12. S. Bodenstein, C.A. Dominguez, S.I. Eidelman, H. Spiesberger, K. Schilcher, J. High Energy
Phys. 01, 039 (2012)
13. A.I. Bochkarev, M.E. Shaposhnikov, Nucl. Phys. B 268, 220 (1986)
14. A. Ayala, C.A. Dominguez, M. Loewe, Adv. High Energy Phys. 2017, 9291623 (2017)
15. C.A. Dominguez, M. Loewe, J.C. Rojas, Y. Zhang, Phys. Rev. D 81, 014007 (2010)
16. C.A. Dominguez, M. Loewe, Y. Zhang, Phys. Rev. D 88, 054015 (2013)
17. G. Aarts et al., J. High Ener. Phys. 03, 084 (2013)
18. G. Aarts et al., J. High Ener. Phys. 12, 064 (2013)
19. T. Matsui, H. Satz, Phys. Lett. B 178, 416 (1986)
20. A. Ayala, A. Bashir, C.A. Dominguez, E. Gutierrez, M. Loewe, A. Raya, Phys. Rev. D 84,
056004 (2011)
21. J.P. Carlomagno, M. Loewe, Phys. Rev. D 95, 036003 (2017)
22. V.A. Miranski, I.A. Shovkovy, Phys. Rep. 576, 1 (2015)
23. J.O. Andersen, W.R. Naylor, A. Tranberg, Rev. Mod. Phys. 88, 025001 (2016)
24. A. Ayala, C.A. Dominguez, L.A. Hernandez, M. Loewe, J.C. Rojas, C. Villavicencio, Phys.
Rev. D 92, 016006 (2015)
25. P. Pascual, R. Tarrach, QCD: Renormalization for the Practitioner, Springer-Verlag, Brlin
Heidelberg (1984)
26. W. Greiner, A. Schäfer, Quantum Chromodynamics (Springer-Verlag, New York, Brlin, Hei-
delberg, 1995)
Appendix A
Dressed Propagators and Selected Integrals

The quark propagator in an external non-Abelian gauge field is given by [1]

αβ δ αβ g  λa αβ G aκλ 
φκλτ κλ

S F ( p, m) = − p φ  γ γ
τ 5 + mσ
p−m
 2 2 ( p 2 − m 2 )2
g 2 αβ 2 p 2 + mp
+ δ G m 2 , (A.1)
12 ( p − m 2 )4

where m is the quark mass, α, β = 1, 2, Nc are colour indexes, a = 1, 2, 3...8 the

SU (3)c index, g is the strong quark-gluon coupling, and G aμν the gluon field tensor.
In the massless limit this becomes
p + m αβ g  λa αβ a 1
S F ( p, m)|m→0 = 2
αβ
δ − G κλ 4 pφ φκλτ γτ γ5 . (A.2)
p − m2 2 2 p

In coordinate space, and in the massless limit, the quark propagator is given by

αβ δ αβ  γ μ · xμ i m 1 x φ g  λa αβ a φκλτ
S F (x, m)|m→0 = + − G κλ  γτ γ5 .
2π 2 x4 2 x2 8π 2 x 2 2 2
(A.3)
The quark condensate in an external non-Abelian gauge filed is

 β  1 αβ  i  
0|q̄iα (x) q j (0)|0 = δ δi j + m q x μ (γμ ) ji + O(m q2 ) 0|q̄ q|0
12 4
δ αβ  i  2 i  λa
+ − x δi j + m q x (γμ ) ji 0|q̄ σ μν i g G aμν q|0 ,
μ
(A.4)
12 16 6 2
where 0|q̄ q|0 is the standard quark condensate, the second term in the expansion
is the dimension d = 5 mixed quark-gluon condensate, and the dimension d = 6
four-quark condensate has been omitted. The suggestion of relating the four-quark
condensate to the square of the ordinary quark condensate (vacuum saturation) [2]
has no theoretical support [3]. Furthermore, the numerical value of this dimension

© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2018 91

C. A. Dominguez, Quantum Chromodynamics Sum Rules,
SpringerBriefs in Physics, [Link]
92 Appendix A: Dressed Propagators and Selected Integrals

d = 6 term in the OPE, as determined from FESR in the vector channel, has some
100% uncertainty [4]. The situation is better for the d = 6 chiral condensate (V-A),
known with some 10% uncertainty [5], albeit an order of magnitude smaller than the
quark condensate at relevant comparable energy scales.
The following space-time integrals enter in current correlators:

1 4 π2 i
d 4 x eiq·x = − (A.5)
x2 q2

xμ 8 π2 μ
d 4 x eiq·x = q (A.6)
x2 (q 2 )2

xμ xν 8 π2 i
d 4 x eiq·x = 2 3 (−q 2 g μν + 4 q μ q ν ) (A.7)
x 2 (q )

xμ 2 π2 μ
d 4 x eiq·x = q (A.8)
x4 q2

xμ xν 2 π2 i
d 4 x eiq·x = (−q 2 g μν + 2 q μ q ν ) (A.9)
x4 (q 2 )2

1 π 2 (−)n 1 ln(−q 2 /μ2 )

d 4 x eiq·x = i 2n−4 (n > 1) (A.10)
(x )
2 n 2 (n) (n − 1) (q 2 )2−n

xμ π 2 (−)n μ (n − 2) ln(−q 2 /μ2 )

d 4 x eiq·x = q (n > 2) (A.11)
(x 2 )n 22n−5 (n) (n − 1) (q 2 )3−n

xμxν π 2 (−)n+1 (n − 2)
d 4 x eiq·x = i
(x 2 )n 22n−5 (n) (n − 1)
 μ ν
q q ln(−q 2 /μ2 )
× g μν + 2 (n − 3) 2 (n > 2) (A.12)
q (q 2 )3−n

xαxβ xρ π2 (n − 2)(n − 3) 2 n−4

d 4 x eiq·x = (−)n+1 (q )
(x 2 )n 22n−6 (n) (n − 1)
 (n − 4) α β ρ 
× q α g βρ + q β g αρ + q ρ g αβ + 2 q q q ln(−q 2 /μ2 ). (A.13)
q2

Dimensional regularized integrals, with dimension D = 4 + 2 :

d D k (k 2 )a = 0, (A.14)
Appendix A: Dressed Propagators and Selected Integrals 93

valid for all a, except a = D/2. Hence, for consistency one defines

d D k (k 2 )−D/2 = 0. (A.15)

(k 2 )a (a + D/2) (b − a − D/2)

d Dk = π D/2 (M 2 )(a−b+D/2)
(k 2 + M 2 )b (D/2) (b)
(A.16)
d 4k 1 −i  
= ln(−q /μ ) − 2
2 2
(A.17)
(2π)4 k 2 (k − q)2 (4π)2

d 4k kμ −i 1 μ  
= q ln(−q 2
/μ2
) − 2) (A.18)
(2π)4 k 2 (k − q)2 (4π)2 2

d 4k kμ kν i  2 μν  ln(−q 2 /μ2 ) 2 
= q g −
(2π)4 k 2 (k − q)2 (4π)2 12 9
 ln(−q 2 /μ2 ) 13 
+ q μq ν − + (A.19)
3 18

d 4k 1 −i   −q 2  m 2  m 2 
= ln + 2 ln
(2π)4 k 2 [(k − q)2 − m 2 ] (4π)2 μ2 q −q 2
 m 2  m 2 
+ 1 − 2 ln 1 − 2 − 2 (A.20)
q q

d 4k kμ −i 1 μ   −q 2  m 2  m2 
= q ln + 2 −
(2π)4 k 2 [(k − q)2 − m 2 ] (4π)2 2 μ2 q2 q2
 m2   m 2
m 4  m 2
m 2 
× ln + 1 − 2 2 + 4 ln 1 − 2 + 2 − 2 . (A.21)
−q 2 q q q q

1 dD p 1
I (q, a, b) ≡
μ 2 (2π) ( p + iη) [( p − q)2 + iη]b
D 2 a

i  q 2 
(2 − a + ) (2 − b + )
= − q −2(a+b−2)
(4π) 2 4πμ 2 (a)(b)(4 − a − b + 2)
× (a + b − 2 − ) . (A.22)

1 dD p pμ
I μ (q, a, b) ≡
μ2 (2π) D ( p 2 + iη)a [( p − q)2 + iη]b
i  q 2 
(3 − a + ) (2 − b + )
= − q −2(a+b−2) q μ
(4π)2 4πμ2 (a)(b)(5 − a − b + 2)
× (a + b − 2 − ) . (A.23)
94 Appendix A: Dressed Propagators and Selected Integrals

1 dD p pμ pν
I μν (q, a, b) ≡
μ2 (2π) D ( p 2 + iη)a [( p − q)2 + iη]b
i  q 2 
(3 − a + ) (3 − b + )
= − q −2(a+b−2) g μν q 2
(4π)2 4πμ2 2 (a) (b) (6 − a − b + 2)
(4 − a + ) (2 − b + )
× (a + b − 3 + ) + q μ q ν
(a) (b) (6 − a − b + 2)

× (a + b − 2 + ) . (A.24)

In some applications of QCDSR one uses the imaginary part of current correlators,
which involve integrals of delta-functions. A sample list is provided below (for more
details see [6]).
Definitions
1
I1 ≡ dp0 θ( p0 ) δ( p 2 − m 2 ) = , E = + (p2 + m 2 ). (A.25)
2E
w(a, b, c) ≡ [a 2 + b2 + c2 − 2(ab + ac + bc)]1/2
= [(a + b − c)2 − 4ab]1/2
√ √ √ √
= [a − ( b + c)2 ]1/2 [a − ( b − c)2 ]1/2 . (A.26)

Integrals

I2 ≡ d 4 q θ( p 0 − q 0 ) δ[( p − q)2 − m 2 ] δ(q 2 − M 2 )

π 1
= θ( p 0 ) θ[ p 2 − (M + m)2 ] 2 w( p 2 , M 2 , m 2 ). (A.27)
2 p

I3 ≡ d 4 q θ(q 0 ) θ( p 0 − q 0 ) δ[( p − q)2 − m 2 ] δ(q 2 − M 2 ) q μ

π 1
= θ( p 0 ) θ[ p 2 − (M + m)2 ] 4 w( p 2 , M 2 , m 2 ) p μ
4 p
× ( p2 + M 2 − m 2 ) (A.28)

I4 ≡ d 4 q θ(q 0 ) θ( p 0 − q 0 ) δ[( p − q)2 − m 2 ] δ(q 2 − M 2 ) q μ q ν

π 1 pμ pν
= θ( p 0 ) θ[ p 2 − (M + m)2 ] 4 w( p 2 , M 2 , m 2 ) ×
6 p p2
g μν
× [( p 2 + M 2 − m 2 )2 − p 2 M 2 ] − [w( p 2 , M 2 , m 2 )]2 . (A.29)
4
Appendix B
Current Correlators in QCD

In this Appendix it is shown how to obtain a current correlator to leading order in

perturbative QCD, as well as to leading order in the power corrections due to the
quark condensate and the gluon condensate.
Starting with the light-quark, electrically charged, vector-current correlator in the
chiral limit (m u,d = 0), identical to the axial-vector current correlator in this limit, it
is defined as
μν
V
(q 2 ) = i d 4 x eiq·x 0|T (Jμ (x)Jν† (0))|0 , (B.1)

where
Jμ (x) =: d̄ia (x) γμ |i j u aj (x), (B.2)

The time-order product in Eq. (B.1) is given by

τ ≡ T (Jμ (x)Jν† (0)) = d̄ia (x) γμ |i j u aj (x) ū bk (0) γν |kl dlb (0). (B.3)

Contracting the quark fields generates the quark propagators, S F (x) so that τ
becomes
τ = (−1)i S F(d) (−x)|li γμ |i j δ ab i S F (x)| jk δ ab γν |kl , (B.4)

where the Wick-sign, (-1), is due to the ordering of the quark fields. After contracting
δ ab δab = Nc (the number of colours) the trace becomes

τ = Nc T r [S F(d) (−x)γμ S F (x) γν ]. (B.5)

Invoking Eq. (A.3) for the quark propagator in the massless limit, the correlator
becomes
d 4 x iq·x x α x β
μν
V
(q 2 ) = Nc T r (γμ γα γν γβ ) ] e . (B.6)
4 π4 x8

© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2018 95

C. A. Dominguez, Quantum Chromodynamics Sum Rules,
SpringerBriefs in Physics, [Link]
96 Appendix B: Current Correlators in QCD

Using Eq. (A.12) with n = 4 and performing the trace gives

1
μν
V
(q 2 ) = − ln(−q 2 /μ2 ) (qμ qν − q 2 gμν ). (B.7)
4 π2
Often in the literature the electrically neutral vector current is considered

1
Vμ (x) = : [ū(x) γμ u(x) − d̄(x) γμ d(x)]. (B.8)
2

This choice leads to an extra factor of 21 in Eq. (B.7), and is the source of much
confusion.
Next, the light-quark pseudoscalar current correlator ψ5 (q 2 ), Eq. (1.5), involves
the time-ordered product

τ ≡ T [∂ μ Aμ (x) ∂ ν A†ν (0)] = −i 2 d̄ia (x)(γ5 )i j u aj (x) ū bk (0)(γ5 )kl dlb (0) , (B.9)

where the overall minus sign is due to the Wick-sign. This trace then becomes
 
τ = −i 2 Nc T r S F(d) (−x) γ5 S F(u) (x) γ5 . (B.10)

In momentum-space this is given by

τ = − i 2 Nc 4 (m u m d − k1 · k2 ), (B.11)

where the term of order O(m q2 ) can be safely neglected (fully justified after comparing
this perturbative results with the power corrections due to the quark and the gluon
condensates).
The correlator ψ5 (q 2 ) is

d 4 k2 (k1 · k2 ) i(k2 −k1 )·x

d 4 k1
ψ5 (q 2 ) = 4 i 3 Nc (m u + m d )2 d 4 x eiq·x e .
(2π)4
(2π)4 k12 k22
(B.12)
After performing the first (space-time) integral, and making use of the emerging
delta-function δ(q + k2 − k1 ) to integrate over e.g. k2 one has

d 4 k (k 2 − k · q)
ψ5 (q 2 ) = −4 i Nc (m u + m d )2 . (B.13)
(2π)4 k 2 (k − q)2

The first integral above vanishes, while the second integral is given in Eq. A.18
so that  3 
ψ5 (q 2 ) = (m u + m d )2 (−q 2 ) ln(−q 2 /μ2 ). (B.14)
8 π2
Appendix B: Current Correlators in QCD 97

Current information on the PQCD expression of ψ5 (q 2 ) beyond the leading order

is given in Appendix C.
Next, the quark condensate contribution to ψ5 (q 2 ) is obtained as follows. Starting
with the time ordered product, Eq. (B.17), for each flavour a quark-anti-quark pair
combines to make a quark propagator while  the other pair  interacts with the QCD
vacuum giving rise to a quark condensate, q aj (x) q̄kb (0) , giving
    
τ = i 3 δ ab (γ5 ))i j (γ5 ))kl S Fd (−x)|li u aj (x)ū bk (0) + S Fu (x)| jk d̄ia (x)dlb (0) . (B.15)

From Eq. (A.3) the quark propagator is

1 γ α xα 1 mq
S F (x) = +i 2 2 (B.16)
2π 2 x 4 4π x
which after substitution in the trace leads to
    
τ = i T r S Fd (−x) γ5 ū a (0)u a (x) γ5 + S Fu (x) γ5 d̄ a (x) d a (0) γ5 , (B.17)

where
q 1 γμ x μ mq 1
S F (x) = + i 2 2, (B.18)
2π x
2 4 4π x
and
1 i 
q̄ a q a  = 1 + mq γμ . (B.19)
12 4
After substituting the above two expressions in the trace, Eq. (B.17), and sub-
sequently substituting the trace in Eq. (1.5), and performing the coordinate space
integral the result for ψ5 (q 2 ) is

(m u + m d )2  
ψ5 (q 2 )|q̄q = 2
m d ūu + m u d̄d . (B.20)
2q

In practical applications ūu = d̄d is a very good approximation. In fact, SU(2)

vacuum symmetry breaking is negligible in comparison with explicit symmetry
breaking (m d /m u  2).
The quark condensate contribution to the vector current correlator, Eq. (B.1), is
obtained in a similar fashion, with the result

1
μν
V
(q 2 )|q̄q = (m u + m d )q̄ q (qμ qν − q 2 gμν ), (B.21)
q4

where vacuum symmetry, ūu  d̄d, has been used.

98 Appendix B: Current Correlators in QCD

The next power correction in the OPE is the gluon condensate. Let us consider
the correlation function in the light-quark vector channel, Eq. (B.1). Contracting it
with g μν projects the scalar function V (q 2 )

g μν μν
V
(q 2 ) = −3 q 2 V (q 2 ), (B.22)

where V (q 2 ) is explicitly given by

i3 d 4 k1 −ik1 ·x
V (q 2 ) = (−1)2 (γμ )i j (γ μ )kl d 4 x eiq·x e
3 q2 (2π)4
d 4 k2 ik2 ·x ab
× e S jk (k1 ) Sliba (k2 ), (B.23)
(2π)4

where the additional factor i arises from the Wick sign. After integrating over space-
time V (q 2 ) becomes

i3 d 4 k1 −ik1 ·x d 4 k2 ik2 ·x
V (q 2 ) = (γμ )i j (γ μ )kl (2 π)4 e e
3 q2 (2π)4 (2π)4
αβ βα
× S jk (k1 ) Sli (k2 ) δ (4) (q + k2 − k1 ). (B.24)

The dressed quark propagator, Eq. (8.18), in the chiral limit and only up to order
O(G aκλ ) is given by

αβ k
jk αβ g  λa αβ G aκλ φ φκλτ

S jk (k) = δ − k  (γτ γ5 ) jk . (B.25)
k2 2 2 k4
After substituting this quark propagator into Eq. (B.24) gives

i3  g 2  λa  λb  
μ
V (q 2 )|G 2  = (2π) 4
(γμ )i j (γ ) kl G aκλ G bρδ φκλτ
3 q2 2 2 2
φ
d 4 k1 d 4 k2 k1 k2ν (4)
× νρδτ (γτ γ5 ) jk (γα γ5 )li δ (q + k2 − k1 ). (B.26)
(2π)4 (2π)4 k14 k24

The gluon condensate term becomes proportional to 0|G 2 |0 >≡ G 2  after
using the relation
 δ ab   
G aκλ G bρδ = gκρ gλδ − gκδ gλρ G 2 . (B.27)
96
The trace of the six gamma-matrices gives a factor −8 gτ α , so that the index α in
the epsilon tensor becomes τ , and the contraction of the two epsilon tensors with the
metric term in parenthesis above reduce to
 
φκλτ νρδτ gκρ gλδ − gκδ gλρ = −12 gφν . (B.28)
Appendix B: Current Correlators in QCD 99

The term λa λb δ ab = 16, so that putting all together V (q 2 ) becomes

g2  2  d 4 k k μ (k − q)μ
V (q 2 )|G 2  = −i G . (B.29)
3 q2 (2π)4 k 4 (k − q)4

The integral is found in Eqs. (A.23)–(A.24), and after substitution the vector
correlator is given by
1 1  αs 2
V (q 2 )|G 2  = G , (B.30)
12 q 4 π

where αs ≡ g 2 /(4 π).

There is an alternative procedure to obtain this result using the free quark propa-
gators in the loop and coupling two gluon lines ending in the vacuum in all possible
(three) ways. This calculation is far more lengthy than the one using the “dressed”
quark propagator, Eq. (B.25), as above.
In the case of the heavy-quark vector current correlator, Eq. (B.1), now with
Jμ (x) = Q̄(x)γμ Q(x), where Q = c, b, the gluon condensate term in the OPE is
obtained as above, except that the quark mass must be retained in the dressed quark
propagator. This is given in Eq. (A.1). There are two types of contributions remaining
after performing the traces. These are (i) the product of the two terms linear in G μν ,
and (ii) the product of the quadratic term, G 2 , and the quark term. The former involves
the integral
d 4k k μ (k − q)μ
I1 (q 2 ) = (B.31)
(2π)4 (k 2 − m 2Q )2 [(k − q)2 − m 2Q ]2

and the latter involves

d 4k k μ (k − q)μ − 2 k 2
I2 (q 2 ) = , (B.32)
(2π)4 (k 2 − m 2Q )4 [(k − q)2 − m 2Q ]

which are given in Eqs. (A.23)–(A.24). The result for the gluon condensate contri-
bution to V in the heavy-quark sector is [1]

1 1  αs 2 3(1 + a)(1 − a)2

V (q 2 )|G 2  = G
48 q 4 π a2
√
 a + 1
1 3 a2 − 2 a + 3
× √ ln √ − . (B.33)
2 a a−1 a2

Another important correlator in the heavy-quark sector is the pseudoscalar current

correlator, defined by

5 (q 2 ) = i d 4 x eiq·x 0|T (J5 (x) J5† (0))|0, (B.34)

100 Appendix B: Current Correlators in QCD

where
J5 (x) = Q̄(x) i γ5 Q(x), (B.35)

with Q(x) a heavy-quark (charm/bottom) field. The gluon condensate contribution

to 5 (q 2 ) can be obtained following the procedure for the vector current correlator.
The result is given by

1 1  αs 2 3(1 + 3a)(1 − a)2

5 (q 2 )|G 2  = G
48 4 m 2Q π a2
√
 a + 1
1 9 a2 + 4 a + 3
× √ ln √ − . (B.36)
2 a a−1 a2

Finally, the gluon condensate contribution to the pseudoscalar current correlator,

ψ5 (q 2 ), Eq. (1.5) is obtained along the same lines as for the vector current correlator.
The result is
1  αs 2
ψ5 (q 2 )|G 2  = −(m u + m d )2 G . (B.37)
8 q2 π

Regarding radiative corrections to current correlators, the first order in αs is the

only one doable by hand, and even then its derivation involves a couple of pages of
intermediate steps. Hence, only the general outline is presented here. Considering the
electrically charged, light quark, vector current correlator, Eq. (B.1), at the one-loop
level with one-gluon exchange
 
μν
V
(q 2 ) = i 3 d 4 x d 4 y d 4 z eiq·x 0|T Vμ (x) Vν† (0)L I (y) L I (z) |0 , (B.38)

where the quark-gluon interaction Lagrangian is

 αβ
λa
L I (y) = g ū α (y)|m γ λ |mr u β (y)|r G aλ (y), (B.39)
2

and similarly for L I (z). Substituting the Lagrangians into Eq. (B.38) gives rise to
three terms corresponding to the three possible ways of connecting quarks with a
gluon in the loop, i.e. an up-quark line dressed with a gluon, a down-quark line
dressed with a gluon, and an up- and a down-quark lines exchanging a gluon. In
the chiral limit the first two give the same answer, so only two terms remain. After
integration over the variable x, there remain two four-momentum integrals leading
to the simple result

1  αs 
μν
V
(q 2 ) = (−gμν q 2 + qμ qν ) ln(−q 2
/μ2
) 1 + . (B.40)
4 π2 π
Appendix B: Current Correlators in QCD 101

Radiative corrections to the heavy-quark vector current correlator are important

in applications, e.g. in the determination of the heavy-quark masses of the charm and
the bottom quark. These have already been discussed in Chap. 9.
Appendix C
Light-Quark Pseudoscalar Current Correlator
in QCD

The pseudoscalar current correlator in the up-down-quark sector ψ5 (q 2 ), Eq. (1.5),

is given by

ψ5 (q ) = (m̄ u + m̄ d )
2 2
− q 2 0 (q 2 ) + (m̄ u + m̄ d )2 2 (q 2 )

Cq   C4 O4   1 
− ( m¯ u + m¯d ) q̄ q + + O , (C.1)
−q 2 −q 2 q4

where m̄ q stands for the quark mass in the MS-bar renormalization scheme. In the
strange-quark case one must perform the replacement (m̄ u + m̄ d ) ⇒ (m̄ q + m̄ s ),
with m̄ q ≡ (m̄ u + m̄ d )/2. The coefficient Cq above is Cq = 1/2. Of particular impor-
tance is to notice the asymptotic behaviour of ψ5 (q 2 ), as it diverges quadratically

lim |−q 2 →∞ ψ5 (q 2 ) ∼ q 2 . (C.2)

This implies that dispersion relations for ψ5 (q 2 ) are to involve two subtractions.
In the framework of FESR this is usually of no major concern as integrals are over a
finite range. Otherwise, the alternative is to consider the second derivative of ψ5 (q 2 ),
to be discussed at the end of this Appendix.
The perturbative QCD function 0 (q 2 ) is known up to order O(αs4 ) [7–10], to
wit.
1  
0 (q 2 ) = − 12 + 6L + a s A 1 (q 2
) + a 2
A 2 (q 2
) + a 3
A 3 (q 2
) + a 4
A 4 (q 2
) ,
16 π 2 s s s
(C.3)

where L ≡ ln(−q 2 /μ2 ), as ≡ αs (−q 2 )/π, and the Ai (q 2 ) terms are given by

© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2018 103
C. A. Dominguez, Quantum Chromodynamics Sum Rules,
SpringerBriefs in Physics, [Link]
104 Appendix C: Light-Quark Pseudoscalar Current Correlator in QCD

131
A1 (q 2 ) = − + 34 L − 6 L 2 + 24 ζ(3), (C.4)
2
 65 10801   11 
A2 (q 2 ) = 4 n F ζ(3) − n F − 117 ζ(3) + L+ n F − 106 L 2
4 24 3
 n 19  3
F
+ − + L + constants, (C.5)
3 2
 4781 475  221 4
A3 (q 2 ) = C1 L − 6 − ζ(3) L 2 + 229 L 3 − L , (C.6)
18 8 16
where
4748953 π4 91519 715
C1 = − − ζ(3) + ζ(5), (C.7)
864 6 36 2
and finally

5
A4 (q 2 ) = Hi L i , (C.8)
i=1

where the coefficients Hi contain a very large number of terms, e.g. H1 has
twenty-three terms!. Hence, we only list their numerical values [10]: H1 = 33532.3,
H2 = −15230.645111, H3 = 3962.454926, H4 = −534.0520833, and H5 =
24.17187500.
The remaining terms in Eq. (C.1) are
1   
2 (q 2 ) = − 12 + 12 L + a s − 100 + 64 L − 24 L 2
+ 48 ζ(3) , (C.9)
16 π 2
1 7 
Cq = + − L as , (C.10)
2 3
1   
C4 O4  = − as G μν , G μν 1 + O(as ) , (C.11)
8
where the radiative correction, O(as ), is not known. The current correlator in the
strange-quark case follows from the above results after replacing m̄ d by m̄ s , and m̄ u
by (m̄ u + m̄ d )/2.
Returning to the issue of the asymptotic behaviour of ψ5 (q 2 ), in FOPT there is
no difference between using in FESR the function ψ5 (q 2 ) or its second derivative,
ψ5 (q 2 ). This is due to the identity, Eq. (4.3), plus the fact that the hadronic term of
the FESR involving ψ5 (q 2 ) is identical to the one involving its second derivative,
ψ5 (q 2 ). This is not necessarily the case in CIPT. Hence, one might need the PQCD
expression of ψ5 (q 2 ). Up to order O(αs4 ), after Renormalization Group Improvement
(RGI), i.e. setting the renormalization scale μ2 = −q 2 , it is as follows [7–10]

(m u + m d )2 1 
ψ5 (q 2 )| P QC D = − K m asm , (C.12)
16 π 2 s m=0
Appendix C: Light-Quark Pseudoscalar Current Correlator in QCD 105

where the coefficients K m are

K 0 = 6, (C.13)

K 1 = 22, (C.14)

65 10801  11 
K 2 = 4 n f ζ(3) − n f − 117 ζ(3) + +2 n f − 106 , (C.15)
4 24 3
6163613 109735 815  46147 524
K3 = − ζ(3) + ζ(5) + − + ζ(3)
864 36 2 81 3
50   15511  49349 1755
− 5 ζ(4) − ζ(5) n f + − 2 ζ(3) n 2f − + ζ(3)
3 1944 12 2
 11651   275 1 
+ − 59 ζ(3) n f + − + ζ(3) n 2f . (C.16)
36 54 3
K 4 = C41 + 2 C42 , (C.17)

where
10811054729 3887351 458425 2 265
C41 = − ζ(3) + ζ (3) + ζ(4)
82944 54 72 3
373975 4125 178045  1045811915
+ ζ(5) − ζ(6) − ζ(7) + −
72 16 128 62208
5747185 2865 2 9131 41215
+ ζ(3) − ζ (3) − ζ(4) + ζ(5)
864 8 96 72
2875 665   220313525 11875
+ ζ(6) + ζ(7) n f + − ζ(3) + 5 ζ 2 (3)
48 12 373248 72
25 5015   520771 65 1
+ ζ(4) − ζ(5) n 2f + − + ζ(3) + ζ(4)
16 72 93312 72 24
5 
+ ζ(5) n 3f = 56824.55903 − 8725.6816 n f + 328.69544 n 2f
3
− 2.722464877 n 3f , (C.18)

which for n f = 3 gives: C41 = 33532.27. The second term in Eq. (C.17) is given by

49573615 535759 30115  56935973

C42 = − + ζ(3) − ζ(5) +
1152 32 16 10368
243511 1115   6209245 500
− ζ(3) + 5 ζ(4) + ζ(5) n f + − + ζ(3)
144 8 31104 9
25   985 5 
− ζ(5) n 2f + − ζ(3) n 3f , (C.19)
6 486 9
which for n f = 3 gives C42 = − 15230.646.
Turning to the non-perturbative expression of ψ5 (q 2 ), the gluon condensate con-
tribution is given by
106 Appendix C: Light-Quark Pseudoscalar Current Correlator in QCD

1 1  αs 2
ψ5 (q 2 )|G 2  = − (m̄ u + m̄ d )2 2 3 G . (C.20)
4 (q ) π

In principle, there is an issue with the gluon condensate contribution to light-

quark correlators, related to the removal of logarithmic quark-mass singularities [11].
However, this turns out to be numerically unimportant, given the size of uncertainties
from other sources, so that the above expression can be safely used. The next power
correction is that involving the quark condensate, given by

(m̄ u + m̄ d )2   1 7 αs 
ψ5 (q 2 )|q̄ q = 2 ( m¯ u + m̄ d ) q̄ q + . (C.21)
(q 2 )3 2 3 π

The basic CIPT integrals are formally given in Eqs. (4.7)–(4.8), and will be dis-
cussed in the next Appendix.
Appendix D
QCD Integrals of Light-Quark Pseudoscalar
Current Correlator in QCD

Starting with integration in FOPT, the contour integral in the FESR for the QCD
light-quark current correlator involving an (analytic) integration kernel of the form

5 (s) = 1 − a0 s − a1 s 2 , (D.1)

is given by 
1
δ5 (s0 )| QC D ≡ − ds 5 (s) ψ5 (s)| QC D , (D.2)
2πi C(|s0 |)

m 2 (s0 ) s02 s3 s4
δ5 (s0 )|1−L O O P = C01 − a0 0 − a1 0 , (D.3)
16π 2 2 3 4

m 2 (s0 ) αs (s0 )  s2 s03 s04 

δ5 (s0 )|2−L O O P = C 11
0
− a 0 − a 1
16π 2 π 2 3 4
 s2 s 3
s 4
− 2 C12 0 − a0 0 − a1 0 , (D.4)
4 9 16
  2 
m 2 (s0 )  αs (s0 ) 2 s0 s03 s04
δ5 (s0 )|3−L O O P = C 21 − a 0 − a 1
16π 2 π 2 3 4
 2  2  2 
s0 s03
s04
s0 π 1
− 2 C22 − a0 − a1 − 6 C23 −
4 9 16 2 6 4
    
s 3
π 2
1 s 4
π 2
1
− a0 0 − − a1 0 − , (D.5)
3 6 9 4 6 16

© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2018 107
C. A. Dominguez, Quantum Chromodynamics Sum Rules,
SpringerBriefs in Physics, [Link]
108 Appendix D: QCD Integrals of Light-Quark Pseudoscalar Current Correlator in QCD
  2 
m 2 (s0 )  αs (s0 ) 3 s0 s03 s04
δ5 (s0 )|4−L O O P = C31 − a0 − a1
16π 2 π 2 3 4
 2   
s0 s 3
s 4
s 2
π 2
1
− 2 C32 − a0 0 − a1 0 − 6 C33 0 −
4 9 16 2 6 4
   
s03 π 2 1 s04 π 2 1 s2
− a0 − − a1 − + 24 C34 0
3 6 9 4 6 16 2
 2      
π 1 s03 π 2 1 s04 π 2 1
× − − a0 − − a1 − . (D.6)
6 4 9 6 9 16 6 16
  2 
m 2 (s0 )  αs (s0 ) 4 s0 s03 s04
δ5 (s0 )|5−L O O P = C 41 − a 0 − a 1
16π 2 π 2 3 4
 2  2  2 
s0 s03
s04
s0 π 1
− 2 C42 − a0 − a1 − 6 C43 −
4 9 16 2 6 4
   
s 3
π 2
1 s 4
π 2
1 s2
− a0 0 − − a1 0 − + 24 C44 0
3 6 9 4 6 16 4
 2  3  2  4  2 
π 1 s π 1 s π 1
× − − a0 0 − − a1 0 −
6 4 9 6 9 16 6 16
  
s02 π 4 π2 1 s03 π 4
+ 120 C45 − + − a0
2 120 24 16 3 120
   
π2 1 s04 π 4 π2 1
− + − a1 − − , (D.7)
54 81 4 120 96 256

where m̄ ≡ m̄ u + m̄ d , and the constants Ci j above, for three quark flavours, are:
C01 = 6, C11 = 34, C12 = −6, C21 = −105 ζ(3) + 9631/24, C22 = −95,
C23 = 17/2, C31 = 4748953/864 − π 4 /6 − 91519 ζ(3)/36 + 715 ζ(5)/2, C32 =
−6 [4781/18 − 475 ζ(3)/8], C33 = 229, C34 = −221/16, C41 = 33532.26, C42 =
−15230.6451, C43 = 3962.45493, C44 = −534.052083, C45 = 24.1718750. The
leading non-perturbative contributions are due to the gluon and the light-quark con-
densates, which give
 
m 2 (s0 ) αs 2 αs (s0 ) 11
δ5 (s0 )|<G 2 > =  G  1+ + 2 a0 s0 + a1 s02 , (D.8)
8 π π 2
 
αs (s0 ) 14
δ5 (s0 )|<q̄q> = m 2 (s0 ) m ud q̄ q 1 + + 2 a0 s0 + a1 s02 , (D.9)
π 3

where ūu = d̄d ≡ q̄q is a very good approximation. The case of the strange-
quark mass is obtained as above after redefining m → (m/2 + m s ), .
Appendix D: QCD Integrals of Light-Quark Pseudoscalar Current Correlator in QCD 109

Turning to CIPT, integrating Eq. (C.12) gives


m̄ 2  ds    ᾱ (s) n
4
1 s
δ5 (s0 )| RG I
= K n F(s) − F(s 0 ) , (D.10)
P QC D
16 π 2 n=0 2πi C(|s0 |) s π

where

4
F(s) = bN s N , (D.11)
N =1

with
 s2 s3  1 a0 a1
b1 = − s0 − a0 0 − a1 0 , b2 = , b3 = − , b4 = − . (D.12)
2 3 2 6 12
Finally, Eq. (D.10)becomes

m̄ 2   1 ds N  ᾱs (s)  M 1
δ5 (s0 )| RG
P QC D =
I
K M b N s − F(s0 )
16 π M
2
N
2πi |s0 | s π 2πi
    
ds ᾱs (s) M m̄ 
2 
× = KM b N I Na M − F(s0 )I Nb M , (D.13)
|s0 | s π 16 π M
2
N

where the two basic integrals, I Na M ≡ I Na M (s0 ) and I Nb M ≡ I Nb M (s0 ), were introduced
in Eqs. (4.7)–(4.8).
Appendix E
QCD Thermal Space-like Spectral Functions

This Appendix deals with the QCD expression of the so-called scattering term for
a vector current correlation function of non-zero, equal mass quarks. It is actually
a space-like contribution. Other current correlators may be treated similarly. The
a
vector current is Vμ (x) =: Q (x) γμ Q a (x) :, where Q(x) is a quark field of mass
m Q , and a is the colour index. In the time-like region one has
 
aμν (q 2 ) ≡ (−gμν q 2 + qμ qν ) a (q 2 ) = −i 3 Nc d 4 x eiq x T r γμ S F (x) γν S F (−x) , (E.1)

where S F (x) is the quark propagator in space-time, and Nc = 3. Transforming the

propagators to momentum-space, performing the integrations, and taking the imag-
inary part of (q 2 ) gives
+v
3 1
I m a (q 2 ) = d x (1 − x 2 ) = v(3 − v 2 ), (E.2)
16 π −v 8π

where the variable v ≡ v(q 2 ) is given by

 1/2
4 m 2Q
v(q ) =
2
1− . (E.3)
q2

Notice that the normalization factor of a (q 2 ) for massless quarks is I m a (q 2 ) =

1/(4 π) (Fig. E.1).
The extension to finite T can be performed using the Dolan-Jackiw thermal prop-
agator, Eq. (12.3), in Eq. (E.1), to obtain.
+v    
3 |q|x + ω |q|x − ω
I m a (q 2 , T ) = d x (1 − x 2 ) 1 − n F − nF . (E.4)
16 π −v 2T 2T

© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2018 111
C. A. Dominguez, Quantum Chromodynamics Sum Rules,
SpringerBriefs in Physics, [Link]
112 Appendix E: QCD Thermal Space-like Spectral Functions

Fig. E.1 The complex

energy, ω-plane, showing the
central cut around the origin
(scattering term), extending
between ω = −|q|, and
ω = |q|. The standard
(time-like) annihilation
right-hand and left-hand cuts
 
2 1/2 are
at ω = ± |q|2 + ωth
not shown (ωth is some
channel dependent mass
threshold)

In the rest frame of the thermal medium, |q| → 0, this reduces to

+v
3
I m a (ω, T ) = d x (1 − x 2 ) [1 − 2 n F (ω/2T )]
16 π −v
3 +v  ω 
= d x (1 − x 2 ) tanh . (E.5)
16 π −v 4T

Proceeding to the scattering term, the equivalent to Eq. (E.4) is

∞    
3 |q|x + ω |q|x − ω
I m s (q 2 , T ) = d x (1 − x 2 ) n F − nF ,
8π v 2T 2T
(E.6)
where the integration limits arise from the bounds in the angular integration in
momentum space. Notice that this term vanishes identically at T = 0, and the overall
multiplicative factor is twice the one in Eq. (E.4). Next, the term in brackets in Eq.
(E.6) becomes a derivative
∞
3 ω d
I m s (q 2 , T ) = d x (1 − x 2 ) n F (y) (E.7)
8π T v dy

where y = |q| x/(2 T ). This expression reduces to

  ∞
3 ω |q| v 8T2
I m  (q , T ) =
s 2
−n F (1 − v 2 ) + y n F (y) dy .
4 π |q| 2T |q|2 |q| v/2T
(E.8)
Appendix E: QCD Thermal Space-like Spectral Functions 113

In the limit |q| → 0 this result becomes

 
3 ω m  2 T2 ∞
Q
I m  (q , T ) =
s 2
lim 2
mQ nF + 2 y n F (y) dy .
π |q|→0 |q|3 T mQ m Q /T
ω→0
(E.9)
After careful performance of the limit, in the order indicated, the singular term
ω/|q|3 above becomes a delta function

ω 2
lim = δ(ω 2 ), (E.10)
|q|→0
ω→0
|q| 3 3

and the final result for the scattering term is

 
2 2 m  2 T2 ∞
Q
I m  (ω, T ) = m Q δ(ω ) n F
s 2
+ 2 y n F (y) dy . (E.11)
π T mQ m Q /T

Depending on the correlator, the limiting function, Eq. (E.10), could instead be
less singular in |q|, in which case the scattering term would vanish identically.

References

1. V.A. Novikov, M.A. Shifman, A.I. Vainshtein, V.I. Zakharov, Fortschritte der Physyk 32, 585
(1984)
2. M.A. Shifman, A.I. Vainshtein, V.I. Zakharov, Nucl. Phys. B 147, 385 (1979), ibid., B 147,
448 (1979)
3. A. Gomez Nicola, J.R. Pelaez, J. Ruiz de Elvira, Phys. Rev. D 82, 074012 (2010)
4. S. Bodenstein, C.A. Dominguez, S.I. Eidelman, H. Spiesberger, K. Schilcher, J. High Energy
Phys. 01, 039 (2012)
5. C.A. Dominguez, L.A. Hernandez, K. Schilcher, H. Spiesberger, J. High Energy Phys. 03, 053
(2015)
6. H. Pietschmann, Weak Interactions-Formulae, results, and Derivations, Springer Verlag/Wien
(1983)
7. S.G. Gorishnii, A.L. Kataev, S.A. Larin, Phys. Lett. B 259, 144 (1991)
8. K. Chetyrkin, B.A. Kniehl, M. Steinhauser, Phys. Rev. Lett. 79, 2184 (1997)
9. P.A. Baikov, K.J. Chetyrkin, J.H. Kühn, Phys. Rev. Lett. 96, 012003 (20016)
10. K.J. Chetyrkin, Private communication
11. K.G. Chetyrkin, C.A. Dominguez, D. Pirjol, K. Schilcher, Phys. Rev. D 51, 5090 (1995); K.G.
Chetyrkin, D. Pirjol, K. Schilcher, Phys. Lett. B 404, 337 (1997). C.A. Dominguez, L. Pirovano,
K. Schilcher. Phys. Lett. B 425, 193 (1998)