Resummation and Renormalization in Eff Ective Theories of Particle Physics

Lecture Notes in Physics 912
Antal Jakovác
András Patkós
Resummation and
Renormalization
in Effective
Theories of
Particle Physics
Lecture Notes in Physics
Volume 912
Founding Editors
W. Beiglböck
J. Ehlers
K. Hepp
H. Weidenmüller
Editorial Board
M. Bartelmann, Heidelberg, Germany
B.-G. Englert, Singapore, Singapore
P. Hänggi, Augsburg, Germany
M. Hjorth-Jensen, Oslo, Norway
R.A.L. Jones, Sheffield, UK
M. Lewenstein, Barcelona, Spain
H. von Löhneysen, Karlsruhe, Germany
J.-M. Raimond, Paris, France
A. Rubio, Donostia, San Sebastian, Spain
M. Salmhofer, Heidelberg, Germany
S. Theisen, Potsdam, Germany
D. Vollhardt, Augsburg, Germany
J.D. Wells, Ann Arbor, USA
G.P. Zank, Huntsville, USA
The Lecture Notes in Physics
The series Lecture Notes in Physics (LNP), founded in 1969, reports new devel-
opments in physics research and teaching-quickly and informally, but with a high
quality and the explicit aim to summarize and communicate current knowledge in
an accessible way. Books published in this series are conceived as bridging material
between advanced graduate textbooks and the forefront of research and to serve
three purposes:
• to be a compact and modern up-to-date source of reference on a well-defined
topic
• to serve as an accessible introduction to the field to postgraduate students and
nonspecialist researchers from related areas
• to be a source of advanced teaching material for specialized seminars, courses
and schools
Both monographs and multi-author volumes will be considered for publication.
Edited volumes should, however, consist of a very limited number of contributions
only. Proceedings will not be considered for LNP.
Volumes published in LNP are disseminated both in print and in electronic for-
mats, the electronic archive being available at springerlink.com. The series content
is indexed, abstracted and referenced by many abstracting and information services,
bibliographic networks, subscription agencies, library networks, and consortia.
Proposals should be sent to a member of the Editorial Board, or directly to the
managing editor at Springer:
Christian Caron
Springer Heidelberg
Physics Editorial Department I
Tiergartenstrasse 17
69121 Heidelberg/Germany
[email protected]
More information about this series at http://www.springer.com/series/5304

Antal Jakovác • András Patkós
Resummation and
Renormalization in
Effective Theories of
Particle Physics
123
Antal Jakovác András Patkós
Institute of Physics Institute of Physics
Roland ERotvRos University Roland ERotvRos University
Budapest, Hungary Budapest, Hungary
ISSN 0075-8450 ISSN 1616-6361 (electronic)

Lecture Notes in Physics
ISBN 978-3-319-22619-4 ISBN 978-3-319-22620-0 (eBook)
DOI 10.1007/978-3-319-22620-0
Library of Congress Control Number: 2015951832
Springer Cham Heidelberg New York Dordrecht London

© Springer International Publishing Switzerland 2016
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.
Printed on acid-free paper
Springer International Publishing AG Switzerland is part of Springer Science+Business Media

(www.springer.com)
Preface
The elective course titled “Finite-Temperature Quantum Fields” has been part of the
curriculum of the master’s degree in physics program at Eötvös University for about
15 years. The original one-semester course was introduced by one of the authors
(A.P.) and extended to a two-semester series in 2013 by the second author (A.J.) in
cooperation with Dr. Zsolt Szép. The aim of the lecturers was and remains to enable
students to acquire the conceptual knowledge and the technical tools of quantum
field theory at a level that allows them to participate in research projects of current
international interest. That success in this goal has been achieved is indicated by
the fact that several of the sections of the present monograph grew out of such joint
work carried out with students attending the course.
The material covered in these notes should help to improve students’ ability to
deal in general with reorganizations of the perturbation series of renormalizable
theories. The specific topics selected reflect the subject area of our own research.
From the middle of the 1990s, thermodynamic changes occurring in the ground
state of the strong and electroweak vacuum has been at the center of our scientific
interest. Although the fundamental theories accounting in principle for all relevant
phenomena are well established, field theories with effective degrees of freedom
(such as the sigma-meson and constituent quarks) were constructed with an eye
to the essential features of the underlying dynamics. These models are able to
account for the qualitative changes and even to reproduce some known results
semiquantitatively.
It was definitely not our goal to transform our handwritten notes into a mono-
graph when we decided to write a book. We hoped to preserve the characteristics of
a university course in which one learns both concepts and “recipes” of quantum field
theory mostly by the detailed technical presentation of relevant examples. The eight
chapters of this book can be divided into four parts, each of a different character.
The historic introductory chapter is followed by two chapters reviewing the basics
of quantum field theory necessary for following the directions of contemporary
research. The next three chapters introduce three different and equally widely
used approaches to improving convergence properties of renormalized perturbation
theory. Finally, the last two chapters discuss some physical features of strong and
v
vi Preface
electroweak matter relying to a large extent on the theoretical tools developed in the
previous chapters. Some frequently used or more complicated formulas are collected
in three appendices.
Most of the presented material is widely known, which explains the relatively
low number of external references. Our citations point either to papers in which
specific results of central interest have appeared or to those works that we found
(very subjectively) the most instructive. We are grateful to the entire community
of theoretical particle physicists investigating characteristic features of strong and
electroweak matter, from whom many ideas reflected in the material of the present
notes originate.
Nevertheless, we would like to express our specific gratitude to a few colleagues
who helped us in our work in the field of the effective models of particle physics
in an essential way. For a very fruitful period of our scientific activity we thank
to Professors F. Karsch (Bielefeld-Brookhaven) and K. Kajantie (Helsinki). Our
research and the present lecture series were shaped in important ways by long-
term collaboration with Zsolt Szép and Péter Petreczky. We thank also Szabolcs
Borsányi, Tamás Herpay, Dénes Sexty, Péter Kovács, Gergely Markó, and Gergely
Fejős.
This book is dedicated to the memory of Professor Péter Szépfalusy (1932–
2015). His perfectionist lectures that we attended as students still represent an ideal
for us. It was our privilege to have participated in joint research projects with Péter.
His results on the application of the 1=N expansion to magnetic systems represented
a constant source of intuition during our collaboration on the temperature-induced
variations of the excitations in the quark–meson theory.
Budapest, Hungary Antal Jakovác

April 2015 András Patkós
Abbreviations
1PI One-particle irreducible

2PI Two-particle irreducible
2PR Two particle reducible
3D Three-dimensional
nPI n-Particle irreducible
BEH-effect Brout–Englert–Higgs effect
BSM Beyond standard model
B Bubble integral (d dimensions)
B3 Bubble integral (3 dimensions)
BF Finite part of the bubble integral (d dimensions)
.0/
Bd Logarithmically divergent part of the bubble integral (d dimen-
sions)
CEP Critical endpoint
CTP Closed time path
DR Dimensional reduction
DS Dyson–Schwinger
ECCP Equivalence class of constant physics
EoM Equation of motion
EoS Equation of state
FAC Fastest apparent convergence
FRG Functional renormalization group
HRG Hadronic resonance gas
HTL Hard thermal loop
IR-divergence Infrared divergence
KMS-condition Kubo–Martin–Schwinger condition
LCP Line of constant physics
LO Leading order
MC simulation Monte Carlo simulation
MS-scheme Minimal subtraction scheme
NLO Next-to-leading order
vii
viii Abbreviations
OMS-scheme On-mass-shell scheme

OPT Optimized perturbation theory
PCAC Partially conserved axial current
PMS Principle of minimal sensitivity
RS Resummation scheme
R/A formalism Retarded/advanced formalism
SB limit Stefon–Boltzmann limit
SM Standard model
SSB Spontaneous symmetry-breaking
QCD Quantum chromodynamics
QED Quantum electrodynamics
QGP Quark gluon plasma
T Tadpole integral in d dimensions
T3 Tadpole integral in 3 dimensions
TF Finite part of the tadpole integral in d dimensions
.2/
Td Quadratically divergent part of the tadpole integral in d dimen-
sions
TCP Tricritical point
Contents
1 Effective Theories from Nuclear to Particle Physics .. . . . . . . . . . . . . . . . . . . . 1

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 9
2 Finite Temperature Field Theories: Review . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 11
2.1 Review of Classical Field Theory .. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 11
2.2 Quantization and Path Integral .. . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 13
2.3 Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 19
2.4 Propagators.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 21
2.5 Free Theories, Propagators, Free Energy .. . . . . . . .. . . . . . . . . . . . . . . . . . . . 25
2.6 Perturbation Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 29
2.7 Functional Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 31
2.7.1 Two-Point Functions and Self-Energies ... . . . . . . . . . . . . . . . . . . . 34
2.7.2 Higher n-Point Functions . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 36
2.8 The Two-Particle Irreducible Formalism . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 37
2.9 Transformations of the Path Integral.. . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 40
2.9.1 Equation of Motion, Dyson–Schwinger Equations . . . . . . . . . . 41
2.9.2 Ward Identities from a Global Symmetry . . . . . . . . . . . . . . . . . . . . 42
2.10 Example: ˚ 4 Theory at Finite Temperature .. . . . .. . . . . . . . . . . . . . . . . . . . 44
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 47
3 Divergences in Perturbation Theory .. . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 49
3.1 Reorganizations of Perturbation Theory .. . . . . . . . .. . . . . . . . . . . . . . . . . . . . 50
3.2 On the Convergence of Perturbation Theory . . . . .. . . . . . . . . . . . . . . . . . . . 52
3.3 Renormalization of the UV Divergences . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 53
3.4 Renormalizability: Consistency of UV Renormalization .. . . . . . . . . . . 56
3.5 Scale-Dependence and the Callan–Symanzik Equations . . . . . . . . . . . . 58
3.5.1 Choice of Scale .. . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 61
3.5.2 Landau Pole, Triviality, and Stability. . . . .. . . . . . . . . . . . . . . . . . . . 62
3.5.3 Stability and Renormalization . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 63
3.6 The Wilsonian Concept of Renormalization . . . . .. . . . . . . . . . . . . . . . . . . . 65
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 68
ix
x Contents
4 Optimized Perturbation Theory.. . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 69

4.1 Infrared Divergences and Their Resummation . . .. . . . . . . . . . . . . . . . . . . . 70
4.2 The Optimization Strategy and the Renormalizability
of the Optimized Series. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 71
4.3 OPT for the 4 -Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 75
4.4 Optimization and Renormalization Schemes . . . . .. . . . . . . . . . . . . . . . . . . . 78
4.5 Optimized Perturbation Theory for the SU.3/L SU.3/R
Symmetric Meson Model . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 81
4.6 OPT for the Three-Flavor Quark–Meson Model .. . . . . . . . . . . . . . . . . . . . 86
4.7 The 2PI Formalism as Resummation and Its Renormalization .. . . . . 88
4.7.1 2PI Resummation as Optimized Perturbation Theory . . . . . . . 89
4.7.2 Perturbative Renormalization of the 2PI Equations . . . . . . . . . 91
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 95
5 The Large-N Expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 97
5.1 The Dyson–Schwinger Equation of the N-Component
Scalar Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 98
5.2 The Large-N Closure and Its Recursive Renormalization . . . . . . . . . . . 102
5.3 Landau Singularity and Triviality .. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 106
5.4 Auxiliary Field Formulation of the O.N/ Model.. . . . . . . . . . . . . . . . . . . . 108
5.5 Renormalization of the O.N/-Model in the Auxiliary
Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 111
5.5.1 Leading-Order Counterterm Action Functional.. . . . . . . . . . . . . 111
5.5.2 Next-to-Leading-Order Counterterm Action
Functional: Effects of the Landau Pole . . .. . . . . . . . . . . . . . . . . . . . 114
5.6 Large-n Approximation in U.n/ Symmetric Models . . . . . . . . . . . . . . . . 124
5.6.1 Auxiliary Fields and Integration over Large
Multiplicity Fields . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 126
5.6.2 Interpretation and Renormalization of V .LO/ .v; x; y0 / . . . . . . . 129
5.6.3 Summary Conclusions of the Analysis . . .. . . . . . . . . . . . . . . . . . . . 134
5.7 The Quark–Meson Theory at Large Nf . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 134
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 138
6 Dimensional Reduction and Infrared Improved Treatment
of Finite-Temperature Phase Transitions . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 139
6.1 The ˚ 4 Theory at Finite Temperature . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 142
6.2 Two-Loop Integration Over the Nonstatic Fields . . . . . . . . . . . . . . . . . . . . 144
6.3 The Effective Three-Dimensional Theory and Its
Two-Loop-Accurate Effective Potential . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 148
6.4 Local Coarse-Grained Effective Theory via Matching . . . . . . . . . . . . . . 150
6.5 Dimensional Reduction of a Gauge Theory
at the One-Loop Level .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 153
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 159
Contents xi
7 Thermodynamics of the Strong Matter . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 161

7.1 The Thermodynamics of the Quark System .. . . . . . . . . . . . . . . 162
7.1.1 The T D 0 Excitation Spectra . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 162
7.1.2 Change of the Ground State at Finite
Temperature and Finite Density . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 165
7.2 The U.3/ U.3/ Meson Model . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 166
7.2.1 Phase Transition in the Three-Flavor
Quark–Meson Model .. . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 168
7.2.2 Dependence of the Nature of the Transition
on the Masses of the Goldstone Fields . . .. . . . . . . . . . . . . . . . . . . . 170
7.3 Equation of State of the Strong Matter . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 171
7.3.1 Characterization of the Excitations .. . . . . .. . . . . . . . . . . . . . . . . . . . 175
7.3.2 Spectral Functions and Thermodynamics . . . . . . . . . . . . . . . . . . . . 177
7.3.3 (In)distinguishability of Particles and the Gibbs Paradox . . . 180
7.3.4 Melting of the Particles and Phenomenology
of the Crossover Regime in QCD. . . . . . . . .. . . . . . . . . . . . . . . . . . . . 182
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 186
8 Finite-Temperature Restoration
of the Brout–Englert–Higgs Effect . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 189
8.1 The Reduced SU.2/ Symmetric Higgs + Gauge Model.. . . . . . . . . . . . . 190
8.2 Optimized Perturbation Theory for the Electroweak
Phase Transition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 193
8.3 Results from the Numerical Simulation
of the Electroweak Model Reduced with One-Loop Accuracy .. . . . . 199
8.4 On the High-Accuracy Determination of the Critical
Higgs Mass. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 202
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 205
A The Spectral Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 207

A.1 Sum Rules .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 207
A.2 Positivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 207
B Computation of the Basic Diagrams . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 209

B.1 Tadpole Integral.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 209
B.2 The Bubble Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 213
B.3 Dimensional Regularization .. . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 218
C Integrals Relevant for Dimensional Reduction . . . . . .. . . . . . . . . . . . . . . . . . . . 221

C.1 Nonstatic Sum-Integrals .. . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 221
C.2 Three-Dimensional Integrals .. . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 223
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 223
Chapter 1
Effective Theories from Nuclear to Particle
Physics
Quantum fields O i .x; t/ have the potential of bringing freely propagating field
quanta, characterized by a set of quantum numbers compressed here into the index
i, to rise or disappear from the ground state. Outgoing particles of radically different
nature from the incoming ones enter the ˇ-decay of a neutron: the neutral constituent
of the nuclei disappears and the positively charged constituent emerges. The idea
that neutrinos are born in the process of decay was at radical variance with earlier
interpretations of the decay that assumed in the neutron the hidden preexistence
of a proton and an electron. Fermi proposed for the mathematical description of
this process the following interaction Hamiltonian [1] containing four different field
operators:
Z

O Fermi
H d3 x N proton .x; t/ N
neutron .x; t//. electron .x; t/ neutrino .x; t/ : (1.1)
The Hamiltonian contains many more potential reactions beyond the ˇ-decay (for
instance, the inverse ˇ-decay) that were discovered in sequence and found to follow
perfectly the statistical characterization predicted by the unique coupling constant
of the Fermi theory.
The variable nature of the nuclear constituents gained ground also in the
interpretation of the charge-invariance of the nuclear binding energy. This led to
the postulation of the isotopic invariance of strong forces, formally accounting
for the fact that the strength of nuclear interaction is invariant under proton $
neutron exchange [2]. The origin of this interaction was modeled by Yukawa
assuming the existence of a mediating quantum field very similar to the photon
mediating electromagnetic interactions (QED) [3]. In analogy with QED, a bilinear
coupling can be assumed of the two-component nucleon field N .x/ and the
quantum force field introduced by Yukawa. The principle of isotopic invariance was
extended by Kemmer [4] also on the mediating field ˛ .x/, requiring its irreducible
© Springer International Publishing Switzerland 2016 1

A. Jakovác, A. Patkós, Resummation and Renormalization in Effective
Theories of Particle Physics, Lecture Notes in Physics 912,
DOI 10.1007/978-3-319-22620-0_1
2 1 Effective Theories from Nuclear to Particle Physics
transformation under the SU.2/I group:

Z
HYukawa N
d 3 xN.x; t/ ˛ N.x; t/ ˛ .x; t/;

N.x/ D p
; ˛ D . C ; ; 0 /: (1.2)
n
(The pseudoscalar nature of the pions was recognized and appropriately taken into
account only later.) Here the generators of the isotopic spin transformations ˛ =2
are introduced in analogy with the SU.2/ group of spatial quantum rotations. The
two components of N fit into the doublet and the three components of a into the
triplet representation of the symmetry group.
By the end of the 1930s, Cassen and Condon, and also Wigner, carefully checked
the consistency of the experimental information on nuclear reactions with the
isotopic SU.2/I symmetry [5, 6]. Also, the mass of the quanta mediating the nuclear
force was estimated by analyzing the relationship between the finite-range force
field of a pointlike pion source and the mass of the scalar field quantum whose
exchange produces the attractive nuclear potential. Everything was in place for the
application of quantum field theory to nuclear phenomena. However, when a radical
turn in world history actually brought nuclear science to the forefront of public
interest, it also temporarily shifted the interest of physicists away from exploring
the quantum field theory of strong interactions.
The return to activity in this field occurred in the early 1950s. In that epoch,
the lowest-mass mesons were discovered, and experimental investigations exhibited
the systematic variation of the nuclear binding energy/nucleon with increasing
atomic number A. The extrapolation A ! 1 led to the notion of nuclear matter
characterized by finite energy density and pressure, and important characteristics
were derived from the corresponding thermodynamic equation of state. The theo-
retical interpretation of the fundamental features of nuclear matter followed a dual
approach.
First of all, some rather specific two-body, three-body, etc., complex interaction
potentials were introduced and used in a generalization of quantum theory of
atomic physics to the nuclear interactions. In competition with this approach, a
quantum-field-theoretic description was introduced by Johnson and Teller [7]. In
their model, the exchange of a scalar quantum (the -field) is responsible for the
strong attraction. Stability against collapse is ensured by the presence of the ! .x/
vector field (recall that same-sign electric charges interacting via the electrostatic
vector field act on each other repulsively). The full classical Lagrangian density is
given by the expression

N
LJT D N.x/ i .@ ig! ! / mN C g .x/ N.x/
1 1 1
.x/ C m2 .x/ ! .x/! C m2! ! .x/! .x/: (1.3)
2 4 2
1 Effective Theories from Nuclear to Particle Physics 3
The significance of this proposition can be appreciated by neglecting for a moment

the effect of ! and solving the sigma-nucleon theory on the level of the mean field
approximation, characterized by a homogeneous nonvanishing expectation value of
the -field.
The corresponding quantum equations in the Heisenberg picture are the follow-
ing:

N
.Cm2 /.x/g N.x/N.x/ D 0; i @ mN C g .x/ N.x/ D 0: (1.4)
The mean field approximation corresponds to
< .x/ >D ˙; meff D mN g ˙: (1.5)
It is remarkable that the mean field modifies the nucleon mass appearing in the
second equation of (1.4). The plane-wave modes of the Dirac field with this modified
mass fill the Fermi sphere of the nuclear matter constituents up to the momentum
kNF determined by the nucleon density of the medium:
Z
d3 k 2 F3

N D 4 .kNF k/ D k ; (1.6)
.2/3 3 2 N
where the degeneracy factor 4 D .2Sz C 1/.2I3 C 1/ reflects the spin and isospin
doublet nature of the nucleons. Next, one takes the expectation value of the first
equation of (1.4) and finds the following gap equation for the mean field amplitude
N k D meff =E.k/; E2 .k/ D k2 C m2eff , which can be
with the help of the relation hNNi
seen to follow directly for the plane wave solutions of the Dirac equation:
Z F
2g kN
k2
m2 ˙ D meff dk : (1.7)
2 0 .k2 C m2eff /1=2
For nonzero density (kNF ¤ 0), the ground state of the system is characterized by a
nonzero -condensate. This condensate is not spontaneously formed; it is generated
by the nonzero nucleon density.
It required five more years after this pioneering investigation before Y. Nambu
raised the possibility of the spontaneous formation of a scalar condensate at vanish-
ing baryon density, following the Ginzburg–Landau scenario of superconductivity.
Here this condensate might emerge from the breakdown of the (approximate)
chiral invariance of NN-interactions. According to his proposition, the mass of the
nucleons would fully result from this symmetry-breaking [8].
With vanishing value for the starting nucleon mass, the Lagrangian of the
nucleons decomposes into a sum of two chiral terms:
N
LN D N.x/i N N
@ N.x/ D NL .x/i @ NL .x/ C NR .x/i @ NR .x/; (1.8)
where
1 5 1 C 5
NL D N; NR D N: (1.9)
2 2
On the two chiral components one is allowed to perform two independent SU.2/
symmetry transformations, the Lagrangian (1.8) has an SU.2/L SU.2/R sym-
N
metry. A nonzero mass term of almost all elementary fermions, mN N.x/N.x/ D
mN .NN L NR C NN R NL /, however, explicitly violates this symmetry, and it therefore
cannot be an exact symmetry of nature.
The symmetry operation can be rearranged into a direct product of an axial
and a vector transformation in the following way. One parameterizes the unitary
transformations as
UL D eiLj ; UR D eiRj ;
j j
(1.10)
which gives for the nucleon field the following transformation rule:

1 5 iLj j 1 C 5 iRj j
N.x/ D eiVj Ci5 Aj N.x/;
j j
N.x/ ! e C e (1.11)
2 2
with Vj D .Lj C Rj /=2; Aj D .Rj Lj /=2. A mass term described by

the quadratic form NN i Mij Nj transforms under a combined infinitesimal vector–axial
transformation as

N
ıLmass D iN.x/ ıVj Œ j ; M ıAj Œ j ; M C N.x/: (1.12)
The and C indices of the square brackets refer to commutators and anticom-
mutators respectively. If the isotopic symmetry is exact (M I), only the axial
symmetry is broken. This means that a mechanism that produces mass to the
nucleons leads to the symmetry-breaking pattern SUV .2/ SUA .2/ ! SUV .2/.
An explicit relativistic field-theoretic mechanism of the chiral symmetry break-
down can be constructed by generalizing the Johnson–Teller model of nucleon–
meson interactions with the help of a meson multiplet M with well-defined left–right
transformation properties and defined as

M D .x/ C i .x/ C .a0j .x/ C ij .x// j ; M ! UL M.x/UR : (1.13)
(The conventional notation for the new meson fields is employed.) The invariant
meson dynamics is easily rewritten in terms of the component fields:
1 2
LM D Tr @ M .x/@ M.x/ 2 M .x/M.x/ Tr.M .x/M.x//
4 16
1h
D @ @ C @ @ C @ a0j @ a0j C @ j @ j
2
i 2
2 . 2 C 2 C a20j C j2 / 2 C 2 C a20j C j2 ; (1.14)
4
and it obviously has an orthogonal O.8/ symmetry. The generalization of the

Yukawa interaction reads

LYukawa D g NN L .x/M.x/NR .x/ C NN R .x/M .x/NL .x/ : (1.15)
This is the matrix version of the linear -model.

The stability of the meson sector requires > 0, but the coefficient of the
quadratic term can be negative (“wrong sign mass term” 2 < 0). Then there
is a spontaneously selected direction in the eight-dimensional field space along
which the classical value is nonzero. This direction is chosen conventionally as the
-axis: class D ˙, and the Yukawa interaction provides mass for the nucleon,
mN D g˙. The expansion of the meson Lagrangian around this classical ground
state leads to massive excitations and seven massless mesons, which illustrates
the consequences of Goldstone’s theorem [9].
Though the idea of symmetry-breaking revolutionized the thinking on the
generation of mass hierarchies in particle physics and was eventually honored with
the Nobel Prize of the year 2008, this model is problematic, because the number of
light (massless) particles is greater than what nature displays. There are two ways
out of this dilemma.
Gell-Mann and Lévy simply omitted half of the mesons, reducing in this way
the symmetry of the meson sector to O.4/. Their version of the -model has the
following Lagrangian density [10]:
N 1
L D N.x/i @ N.x/ C .@ /2 C .@ j /2 2 . 2 C j2 /
2
2
N
. C j2 /2 gN.x/ .x/ C i5 j j .x/ N.x/: (1.16)
4
In this case, one has three massless pions resulting from the spontaneous symmetry-
breaking and
2
min D ; m2 D 22 : (1.17)
2
Another alternative to the reduction to the right multiplet structure proceeds via
the complete elimination of the mesons. The one-meson exchange between the
nucleons results in a scalar but nonlocal 4-fermion interaction that displays the
original SU.2/L SU.2/R symmetry:
Z Z Z
SD N
dxN.x/i
@ N.x/ g
2
dx dy.NN L .x/NR .x//K.x y/.NN R .y/NL .y//;
(1.18)
where K.x y/ is the invariant nonlocal kernel. Its local variant is the origi-
nal Nambu–Jona-Lasinio model [11], which follows the example of the Fermi-
interaction of the ˇ-decay:
Z Z
SNJL D N
dxN.x/i
@ N.x/ g
2
dx.NN L .x/NR .x//.NN R .x/NL .x//
Z
N g2 N N

D dxN.x/i @ N.x/ .N.x/N.x//2 .N.x/ 2
5 N.x// : (1.19)
8
In the original investigation, the authors showed that for strong enough g2 , a nucleon
N
condensate hNNi ¤ 0 is formed in the ground state of the model that generates mass
for the nucleons and provides three massless Goldstone modes that are identified
with the pions in the chiral limit (with vanishing explicit symmetry-breaking term).
A complementary approach is to eliminate just the heavy modes. Then the meson
matrix is nonlinearly parameterized:
M.x/ D S.x/ei j .x/=2F S.x/UŒ:

j
(1.20)
In the expression of the potential energy of (1.14), only the invariant S.x/ shows up,
while the kinetic density of the meson sector looks like
1 1
Lkin D @ S.x/@ S.x/ C S2 .x/tr@ U @ U: (1.21)
2 4
The effect of the nucleons comes from the Yukawa term (1.15). In the broken-
symmetry phase, one has
ˇ
S.x/ D s0 C s.x/; g.NR˛ NN L / D BM ˛ˇ ; (1.22)
where M is a diagonal matrix. Neglecting the effect of the fluctuation s.x/ in the
kinetic density, one recognizes that the dynamics of the massive s.x/ decouples, and
one arrives at the nonlinear dynamics of the Goldstone modes:
s20
L D tr@ U @ U Btr M U C U M : (1.23)
4
This is the SU.2/ SU.2/ symmetric nonlinear -model with the second term
reflecting the explicit symmetry-breaking effect from the condensate of fermions.
This model has been analyzed in great detail, also including terms containing higher
derivatives of the Goldstone fields [12]. It is one of the most popular effective
models of strong interactions, which can be mapped exactly onto the quantum
chromodynamic theory at low enough energies.
All these effective models built on the principle of approximate chiral symmetry
have been extended to three flavors. In the context of nuclear matter, the most
advanced formulation was provided in the framework of the Walecka model [13]. It
generalizes the Johnson–Teller model by including the baryon octet and also some
members of the decuplet. In addition to the scalar and pseudoscalar meson octets,
one also includes the vector and axial-vector resonances.
For applications to particle physics, it is more appropriate to replace the nucleons
by u; d; s quarks in the baryon sector of these models. It is the smallness of
the Lagrangian mass parameters of these quarks as compared to QCD , the scale
parameter of QCD, that clarifies the rather accurate realization of the chiral
symmetry. The emerging quark–meson model was introduced in 1973 by Chan and
Haymaker [14]:
1 2
L.M/ D tr.@ M @ M 2 M M/ f1 tr.M M/ f2 tr.M M/2
2

g det.M/ C det.M / C x x C y y : (1.24)
The first term in the last line is ’t Hooft’s determinant term [15], which represents
in the effective model the UA .1/ axial anomaly. The last two terms correspond to
explicit symmetry-breaking. Also, in this version of the effective model, effort is
actually invested to include higher meson resonances [16].
Till now, we have followed the order of discovery of the effective models of
strong interactions. It is worthwhile to add two remarks on the role played by
the effective scalar-spinor models in the electroweak sector of the standard model
(SM). At least one major question still remains open after the discovery of the
scalar excitation of the symmetry-breaking field. This is the triviality of the Higgs
sector, in other words, the presence of a nonintegrable ultraviolet singularity in the
momentum-dependence of the self-coupling of the Higgs field. Estimates on the
location of the so-called Landau pole are based on nonperturbatively resummed
perturbative contributions to the momentum-dependence of this coupling.
An escape from this limitation in the applicability of SM could be the Landau
singularity preempted by some substantial deviation in the momentum-dependence
of the self-coupling from the behavior suggested by the (resummed) weak coupling
perturbation series. A candidate scenario, called asymptotic safety, introduces a
radical change in the scale-dependence of the couplings by the existence of a fixed
point below the location of the would-be singularity, which would attract the flow
of the couplings in the ultraviolet. The larger the mass of a field, the stronger its
Yukawa coupling to the Higgs field. Therefore, it is a natural first step to investigate
the ultraviolet behavior of the couplings of a reduced Higgs–top–bottom Yukawa

model:
Zh
SD N L i @ QL C NtR i @ tR C bN R i @ bR
tr.@ ˚ @ ˚/ C U.tr.˚ ˚// C Q
x
i
N L ˚bR C bN R ˚ QL / C iht .NtL ˚C bR C NtR ˚C QL / ;
C ihb .Q (1.25)
where
a
t
Qa D ; (1.26)
ba
where a denotes the color of the quarks. Simpler variants of this model were shown
to possess a UV-safe fixed point in space-time dimensions less than 4 [17–19]. At
present, intensive search is being conducted on UV-safe fixed points in systems
mimicking SM in four dimensions; see [20] and references therein.
Another approach is to question the existence of an intrinsically elementary
Higgs particle and assume that a new strong interaction creates scalar bound states
with the quantum numbers of the Higgs multiplet that would dissolve at a scale
below the Landau pole, in this way erasing any trace of the would-be triviality. This
scale refers then to the breakdown of the SU.2/L U.1/Y symmetry. As a workable
mechanism for this scenario, at the end of 1980s several authors rephrased in the
electroweak context the idea of Nambu explaining the origin of the pions as a result
of nucleon–antinucleon condensation [21–23]. Again, it was natural to consider an
effective four-fermion interaction due to an unspecified new interaction in the top–
bottom (t; b) family resulting at scale in quark–antiquark condensation, at the
same time leaving out the elementary Higgs field from the SM:
g2new N aI a b bI
LSM;noHiggs D LSM;kin C .QL tR /.NtR QL /: (1.27)
2
Here a; b refer to the color, and I D .1=2; 1=2/ are the weak isospin indices of the
quarks. The invariance of the additional term under the global SU.3/c SU.2/L
U.1/Y is obvious. The scale is much higher than in the Nambu–Jona-Lasinio
model (not lower than the electroweak scale). The term LSM;kin contains the gauge-
invariant kinetic terms of all fermions and nonabelian vector fields, but the Higgs
sector is assumed to emerge dynamically. A composite doublet field
gnew a
˚ I .x/ D Nt .x/QaI
L .x/ (1.28)
Maux R
References 9
is introduced, together with an auxiliary mass that ensures its correct scaling
dimension on the addition of an appropriately chosen constraint term to the
Lagrangian density:
LSM;noHiggs ! LSM
h gnew N aI a i h gnew b bI i
D LSM;kin Maux ˚ I C QL tR Maux ˚ I C Nt Q
R L
gnew N
D LSM;kin Maux QL ˚tR C NtR ˚ QL Maux
2
˚ ˚: (1.29)

The new strong interaction is taken into account on the level of the loop
corrections of the heavy quarks arising from the new piece of the Lagrangian that
generate both a gauge-invariant kinetic term and a stabilizing quartic potential in
addition to the wrong sign coefficient of the mass term (the result does not depend
on Maux ). At this point, one can go over to the standard BEH treatment of symmetry-
breaking. It turns out that a purely radiatively generated top mass in the right
ballpark requires the choice 1013 GeV. Unfortunately, from this approach a
very robust Higgs to top mass relation emerges, mH 2mt , which in the light
of experiments is incompatible with nature. The scheme has been taken over to
models giving more freedom for tuning the couplings, where the condensation of
techniquarks is responsible for the breakdown of the SU.2/L U.1/Y symmetry (for
comprehensive reviews see [24, 25]).
References
1. E. Fermi, Ric. Sci. 2, 2 (1933); Z. Phys. 88, 161 (1934)

2. W. Heisenberg, Z. Phys. 77, 1 (1932)
3. H. Yukawa, Proc. Phys.-Math. Soc. Jpn. 17, 48 (1935)
4. N. Kemmer, Proc. R. Soc. Lond. A166, 127 (1938)
5. B. Cassen, E.V. Condon, Phys. Rev. 50, 846 (1936)
6. E.P. Wigner, Phys. Rev. 51, 106 (1937)
7. M.H. Johnson, E.Teller, Phys. Rev. 98, 783 (1955)
8. Y. Nambu, Phys. Rev. Lett. 4, 380 (1960)
9. J. Goldstone, Nuovo Cimento 19, 154 (1961)
10. M. Gell-Mann, M. Lévy, Nuovo Cimento 16, 705 (1960)
11. Y. Nambu, G. Jona-Lasinio, Phys. Rev. 122, 345 (1961); ibid. Phys. Rev. 124, 246 (1961)
12. J. Gasser, H. Leutwyler, Ann. Phys. 158, 142 (1984); Nucl. Phys. B 250, 465 (1985)
13. B.D. Serot, J.D. Walecka, Adv. Nucl. Phys. 16, 1 (1986)
14. L.H. Chan, R.W. Haymaker. Phys. Rev., D 7, 402 (1973)
15. G. ’t Hooft, Phys. Rev. D 14, 3432 (1976); ibid. 18, 2199 (1978)
16. D. Parganlija, P. Kovács, Gy. Wolf, F. Giacosa, D.H. Rischke, Phys. Rev. D 87, 014011 (2013)
17. B. Rosenstein, D.Warr, S.H. Park, Phys. Rev. Lett. 62, 1433 (1989)
18. K. Gawedzki, A. Kupiainen, Phys. Rev. Lett. 55, 363 (1985)
19. C. de Calan, O.A. Faria de Vega, J. Megnen, R. Seneor, Phys. Rev. Lett. 66, 3233 (1991)
20. H. Gies, R. Sonderheimer, Eur. J. Phys. C 75, 68 (2015)
21. Y. Nambu, New theories in physics, in Proceedings of the XI International Symposium on

Elementary Particle Physics, Katimierz, Poland, eds. by Z. Ajduk, S. Pokorski, A. Trautman.
(World Scientific, Singapore, 1988), pp. 1–10
22. V.A. Miransky, M. Tanabashi, K. Yamawaki, Mod. Phys. Lett. A 4, 1043 (1989)
23. W.A. Bardeen, C.T. Hill, M. Lindner, Phys. Rev. D 41, 1647 (1990)
24. G. Cvetic, Rev. Mod. Phys. 71, 513 (1999)
25. C.T. Hill, E.H. Simmons, Phys. Rep. 381, 235 (2003)
Chapter 2
Finite Temperature Field Theories: Review
This introductory section very concisely summarizes the most important features of
the process of constructing quantum field theories at zero and nonzero temperatures.
There are excellent books and monographs in the literature where the interested
reader can find further details. Without trying to be complete, we give here some
basic references. For general questions on quantum field theory, one can make use
of several books [1–3]. For the more specialized problems of finite-temperature field
theory, a very useful review can be found in [4], and two more recent books are also
available [5] and [6]. It is often useful to think about a field theory as the continuum
limit of a theory defined on a spacetime lattice: an excellent book on this topic was
written by Montvay and Münster [7]. Our review of renormalization in the next
chapter is largely based on the book of Collins [8]. Beyond these basic works, there
are numerous excellent books and papers on more specific subjects that we are going
to quote in the text.
Throughout these notes, field variables are defined in spacetime points x of flat
d-dimensional Minkowskian or Euclidean spacetime M isomorphic to Rd . It is often
useful to have an observer splitting the spacetime into time and space extensions,
denoted by x D .t; x/; here x is a .d 1/-dimensional vector.
2.1 Review of Classical Field Theory
In classical field theory, the basic object is the field A W M ! V, where V

is some vector space. The state of the system at time t is characterized by the
configuration A.t; x/. The dynamics of the system is determined by the action S,
which is a real-valued functional of the field configurations: S W A 7! R. The
classically realizable time evolutions are solutions of the equation of motion (EoM)
manifesting themselves as the extrema of the action ıS=ıA D 0; these solutions are

DOI 10.1007/978-3-319-22620-0_2
12 2 Finite Temperature Field Theories: Review
also called physical configurations. In local field Rtheories, the action can be written
as an integral over the Lagrangian density, S D d d x L . In the fundamental field
theories, the Lagrangian density depends only on the field and its first derivatives:
L .A.x/; @ A.x//. In this case, we can introduce the canonically conjugate field
˘ D @L =@A. P
We can apply to the configurations transformations represented by bijections
U W A 7! A0 . The transformations of the system form a group. We will
consider only pointwise transformations, where the transformed new field at a given
position depends on the original field at another position, i.e., A0 .x0 / D UV .A.x//,
where x0 D UM .x/. The transformation is called internal if UM D 1 (identity);
otherwise, it is an external transformation. We often encounter transformations
determined by continuously varying parameters, i.e., when a certain subgroup of
all transformations forms a Lie group. In this case, the transformation U can be
characterized by a set of parameters c 2 Rn . In the standard representation, the
transformation is written as U.c/ D eica Ta , where Ta are the generators of the
transformation.
The transformation is a symmetry if SŒA0 D SŒA. This also means that if A is an
extremum of S (i.e., realizable motion), then A0 is an extremum, too. In several cases,
we expect that the system exhibits some abstractly defined symmetry group. In this
case, the fields are classified as irreducible representations of the abstract group, and
S is built up from the group invariants constructed from the field products.
An important theorem (Noether’s theorem) states that if U is a continuous
symmetry of the system, then a conserved current associated with it always exists.
The proof can be found in all basic field theory textbooks [8], but for the later use
of its logic of construction, we present a brief overview. Consider a one-parameter
continuous symmetry U , where U0 D 1. The transformed field is A D U .A/,
and the transformed action is S ŒA D SŒA . If the transformation is a symmetry,
then S D S. Now let us consider an infinitesimal but position-dependent parameter
ı.x/. Then two statements can be made. First, the change of the action can be
written as
Z
ıSŒA D Sı ŒA SŒA D d d xj .x/@ ı; (2.1)
since at constant ı, it must be identically zero. Then, after partial integration, we
obtain
ıS
D @ j .x/: (2.2)
ı.x/
On the other hand, Aı is the variation of the original configuration. Therefore, if at
ı D 0 we began from the solution of the equations of motion (EoM), we obtain
ˇ
ıS ˇˇ
D 0: (2.3)
ı.x/ ˇAph
2.2 Quantization and Path Integral 13
These statements together mean that j is a conserved current when evaluated at

solutions of EoM.
The conserved current associated with spacetime translations is the energy–
momentum tensor
@L
T D @ A g L ; (2.4)
@.@ A/
or its symmetrized version [9]. In particular, the energy density and momentum
density read as
P L;
" D A˘ Pi D ˘ @i A (2.5)
(for fields carrying discrete indices, the product of two field variables implies
summation over the indices). Another often used example is the current associated
with a linear internal symmetry. In this case, the infinitesimal transformations can
be written with the help of the generators Tij of the symmetry transformations as
ıAi D iTij Aj ı (where we have written indices explicitly for a more transparent
result). Thus
@L
ji D Tij : (2.6)
@.@ Aj /
2.2 Quantization and Path Integral
The main difference between a classical and a quantum system is the characteri-
zation of each of their states. While in a classical system the state could be fully
described by the configuration of the generalized coordinates and the canonically
conjugated momenta, in the quantum system we introduce a separate Hilbert space
H for the states. The physical states are ŒA 2 H of unit norm j ŒAj D 1.
The transformations of the quantum system form U W H ! H isomorphisms.
These are (anti)linear maps that map a state onto a state, i.e., it conserves the norm;
therefore, the transformation must be (anti)unitary, U D U 1 . In the continuous
case, Uc D eica Ta , just as in the classical case; if U is unitary, then the generator T
is Hermitian, T D T.
Consider now a one-parameter subgroup U D ei T of the linear transforma-
tions. The eigenstates of the transformation are eigenstates of the generator, too: if
T D , then U D ei . The eigenvalues of T are real numbers; those of
U lie on the unit circle. We can also describe the transformation of the system by
transforming instead of the states, the operators O O acting on H (Heisenberg picture)
requiring unchanged values for the inner products
hQ 0 jOj
O 0 i D hQ jO
O 0 j i ! O
O 0 D U OU
O ; (2.7)
O
or infinitesimally ıO D iŒT; Oı.
The generator of a transformation does not change under the effect of the
same transformation T 0 D U TU D T. It makes it possible to identify the
generator of a symmetry transformation as the conserved quantity belonging to this
transformation: in field theory, this is the conserved quantity coming from Noether’s
theorem. Therefore, the generators of the space and time translations are the Noether
charges coming from the energy–momentum tensor, i.e., the conserved momentum
and energy, respectively. After quantization, these become the momentum operator
PO i and Hamilton operator H, O and so the space and time translations are eiPO i xi =„ and
O
eiHt=„ , respectively („ is introduced in order to have dimensionless quantities in the
exponent).
The infinitesimal transformations of the system are interpreted as measurements.
In an eigenstate, we obtain a definite value for the result of the measured quantity;
in other cases, the projection of the arising state onto the eigenstates gives the prob-
ability amplitude of the possible outcomes of the measurement. Two measurements
are not interchangeable if the corresponding generators do not commute.
In particular, in quantum field theories there is an operator that measures the
field at spacetime position x, denoted by A.x/,O and another one that measures the
O
canonically conjugated momentum (˘ .x/). The generator of the space translations
˘i is defined in Rany field theory with the help of the energy-momentum tensor
as Pi D T0i D dy˘.y/@i A.y/ . On the other hand, the fields transform under
an infinitesimal space translation as ıA.x/ D A.x C ıx/ A.x/ D @i Aıxi . When
substituted into the relation expressing the action of an infinitesimal translation on
the field A as .i=„/ıxi ŒPi ; A.x/ D ıxi @i A.x/, it leads to the canonical commuta-
tion/anticommutation relations
O x/; ˘O .t; y/˛ D i„ı.x y/;

ŒA.t; O x/; A.t;
ŒA.t; O y/˛ D 0;
Œ˘O .t; x/; ˘O .t; y/˛ D 0; (2.8)
where ˛ D ˙1 and ŒX; Y˛ D XY ˛YX, i.e., commutator for ˛ D 1 and

anticommutator for ˛ D 1. In the following, „ D 1 units are used. The choice
˛ D 1 applies for bosons, and ˛ D 1 for fermions.1
The eigenvectors of AO (and also those of ˘O ) form an orthonormal basis. In this
basis, we can expand any state, and using the projectors on the eigenstates also any
O
operator. In particular, we can expand the time translation operator eiHt .
For bosonic fields (denoted by ˚.x//, because of the commutation relations (2.8),
there exists a common eigenstate for all field measurement operators at time t, and
also (separately) for all canonical momentum measurement operators:
O x/ j'i D '.t; x/j'i;

˚.t; ˘O .t; x/ ji D .t; x/ji: (2.9)
1
We remark that in .2 C 1/-dimensional field theories, there is a possibility to define particles with
any ˛ with unit length. These are called anyons.
Their inner product, using again (2.8), is

R
d d x .x/'.x/
h'ji D ei : (2.10)
Now we can consider a time interval Œ0; t, which we divide into N equal parts, with
each part of length t D t=N. At each division point Rr Q D 0; : : : ; N (also at the
beginning
R Q and at the end), we insert complete sets 1 D . x d'r .x// j'r i h'r j and
1 D . x dr .x// jr i hr j in alternating order. After a bit of algebra, we obtain
Z
O
iHt
e D D'DeiSŒ'; j'N i h'0 j ; (2.11)
where
Y Y Z
D' D d'r .x/; D D dr .x/; SŒ'; D dt0 Œ 'P H.'; / ;
r;x r;x
h'jHji
'Pr t D 'rC1 'r ; H.'; / D : (2.12)
h'ji
In the case of fermions, the situation is more complicated, because the field
measurement operators O .t; x/ do not commute at a given time. To resolve the
problem, one first introduces the Grassmann algebra G . It is a (graded) complex
algebra generated by the unity and elements ei for which the algebraic product is
antisymmetric ei ej D ej ei . For the physical applications, one also needs a star
operation that connects the generator elements into pairs W e ! e . We can
define a Hilbert space over the Grassmann algebra H.G /, which means that if
j 1 i ; j 2 i 2 H.G / and g1 ; g2 2 G , then g1 j 1 i C g2 j 2 i 2 H.G /. The scalar
product is Grassmann-algebra-valued, h 1 j 2 i 2 G . In the physical applications,
one introduces two Grassmann algebra generators for each spatial position and for
each fermionic component e˛;x and e˛;x . The field operators O .t; x/ and O .t; x/ act
linearly on H.G /, i.e., O .t; x/.g1 j 1 i C g2 j 2 i/ D g1 O .t; x/ j 1 i C g2 O .t; x/ j 2 i,
and similarly for O .t; x/.
Let us consider a single degree of freedom, where we have just two field operators
O and O . We assume, in agreement with most physical applications, that ˘O D iO ,
which implies
ŒO ; O C D 1: (2.13)
The corresponding Grassmann algebra generators are e and e . The complete

algebra is four-dimensional; it has the basis 1, e, e , and ee . We take an element
from the Hilbert space j i 2 H.G / and apply O to it. We obtain
ˇ ˛ ˇ ˛
ˇ 0
D O j i; O ˇ 0
D 0: (2.14)
This means that either j 0 i D 0 itself, or O j 0 i D 0. Therefore, there exists

an element in the Hilbert space that is annihilated by O : this is the vacuum j0i.
Applying O to the vacuum, we obtain
j1i D O j0i; O j1i D 0; O j1i D O O j0i D .1 C O O /j0i D j0i:

(2.15)
Therefore, the Hilbert space is two-dimensional. The basis elements are j0i and j1i.
Now we can define fermionic coherent states: we take a grade-1 Grassmann
element, i.e., D ˛e C ˇe . It is nilpotent, 2 D 0. The product formed
with the help of the conjugated element D ˛ e C ˇ e is a commuting element
of the algebra. The definition of the coherent state is
1
C O 1
j i D e 2 j0i D .1 /j0i C j1i: (2.16)
2
This is an eigenstate of O :
O j i D j0i D j i; (2.17)
because of the nilpotency of . The norm of j i is unity.

Two linearly independent coherent states would form a basis if there existed an
inverse element in the Grassmann algebra. Though this is not the case here, we
can introduce a formal integral over the Grassmann elements of grade 1 (Berezin
integrals), which does the job for us:
Z Z
d D 0; d D 1: (2.18)
Then we obtain
Z
d d j ih j D 1; (2.19)
because
Z

d d .1 /j0ih0j C j1ih0j C j0ih1j C j1ih1j D j0ih0j C j1ih1j D 1:
(2.20)
Similarly, we obtain
Z D E
O
d d jOj O
D Tr O; (2.21)
because
Z h i
O
d d .1 /h0jOj0i O
h1jOj0i O
C h0jOj1i O
h1jOj1i
O
D h0jOj0i O
C h1jOj1i: (2.22)
Note that for the trace, one has to evaluate an antisymmetric expectation value
(2.21).
The case of a single degree of freedom discussed above can be taken over for
more fermionic dynamical variables without any problem. The most important
finding is the integral formula representing the spectral decomposition of the
identity and the trace formula.
Now we can write explicitly the representation of the time evolution operator
with the help of the fermionic integrals. Just as in the bosonic case, we split the
time interval Œ0; t into N parts and insert an identity at each of these division points.
Finally, we obtain
Z
O
eiHt D D D eiSŒ ; j 0 i h N j ; (2.23)
where
Y
N

D D D d nd n; rC1 r D @t r t;
nD1
Z

SŒ
; D dt P HŒ
; ; HŒ
; D h jHj i : (2.24)
We have obtained analogous formulas for both the bosonic and fermionic
theories. Symbolically, we write in common notation
Z
O
eiHt D DAeiSŒA jA0 i hAN j : (2.25)
Once we have a representation for the time evolution operator, we can work out
any n-point correlation function. We are primarily interested in expectation values
of the product of Heisenberg-picture operators:
Tr AO n .tn / : : : AO 1 .t1 /%;

O (2.26)
where %O is the initial density operator defined at time t0 . Using the cyclic property
O D eiH.tt0 / Ae
of the trace and the time evolution of the field operators A.t/ O iH.tt0 / ,
we have
Tr eiH.ti tf / eiH.tf tn / AO n eiH.tn tn1 / : : : eiH.t2 t1 / AO 1 eiH.t1 t0 / %e
O iH.t0 ti / ;
(2.27)
where ti and tf are two arbitrary moments of time, practically ti ! 1 and tf ! 1.
Now we can substitute the path integral representation of the time evolution
operator. Between two time translations, there appear matrix elements of the AO
operators evaluated between states sitting on the first point of the left side and
the next point of the right side of the time evolution operator. We can write (also
introducing a complete system with ji representation in the bosonic case)
hN j˘O j'0 i O 0i

hN j˚j'
D N ; D 0 ;
hN j'0 i hN j'0 i
h O
N j j 0i h N j
O j 0i
D 0; D N: (2.28)
h Nj 0i h Nj 0i
We will also encounter the matrix elements of the density matrix
hAN j%jA
O 0i
%ŒAN ; A0 D ; (2.29)
hAN jA0 i
which can be, of course, a rather complicated functional. In the t ! 0 limit, we
P We have finally, in symbolic notation,
will denote the arguments by A and A.
CZpath
R
Tr AO n .xn / : : : AO 1 .x1 /%O D DA ei C dtL P
An .tn / : : : A1 .t1 /%ŒA; A: (2.30)
P=A
Here P=A means periodic=antiperiodic contours: for bosons we must apply periodic,
for fermions antiperiodic, boundary conditions. In this representation, the time flows
along a closed time path (CTP) C W ti ! t0 ! t1 ! ! tn ! tf ! ti . To
emphasize the order of the time arguments along the contour, we can introduce a
parameterization t./. In the “contour time” , the contour is monotonic and closed.
The representation of the time evolution in fact uses operators that depend on the
contour time, and we obtain contour time ordering in a natural way.
In real time, however, if the time arguments appear without any definite order, the
contour runs back and forth several times. If for a certain section it is monotonic,
those subsections can be merged into a common section; e.g., if t1 < t2 < t3 , then
we can write instead of t1 ! t2 ! t3 simply t1 ! t3 . Moreover, we can always
0 0
plug in a unit operator as eiH.t t/ eiH.t t/ , i.e., we can include a bypass anywhere
in the time chain: instead of tn ! tnC1 , we can write tn ! t0 ! tnC1 . According to
these observations, the time contour can be standardized as ti ! tf ! ti ! !
tf ! ti , where a sufficient number of back-and-forth sections must be allowed for
it to contain (taking into account the ordering) all the operators. This form depends
on the operators only through their number. On the sections where the time flows
as ti ! tf , the contour ordering is time ordering, while along the backward-running
contours, the contour ordering is anti-time-ordering.
To simplify the treatment, one usually considers only two time sections (this
is the minimal choice, since we always have to return to the same time ti ). The
path integral variables are called A.1/ for the contour segment ti ! tf , and A.2/ for
2.3 Equilibrium 19
the section tf ! ti . In the contour ordering, segment 2 always follows segment 1,

independently of the actual time values. On segment 1, the usual action
Z tf
SD dt L (2.31)
ti
is used. On segment 2, where the orientation of the integration is tf ! ti , the action

S can be used with the same ordering of the endpoints as on segment 1. With this
setup, we can study correlation functions like
h i
F D Tr T .AO n .xn / : : : AO k .xk // T.AO k1 .xk1 / : : : AO 1 .x1 //%O ; (2.32)
where T is anti-time-ordering, T is time-ordering. In this simplified setup, F has

the following path integral representation:
Z
.1/ iSŒA.2/ .1/ .1/ P .1/
FD DA eiSŒA A.2/
n .tn / : : : A1 .t1 /%ŒA ; A ; (2.33)
P=A
where DA D DA.1/ DA.2/ . This is the generic path integral representation for
arbitrary initial conditions encoded into the density matrix.
The generating functional of the n-point functions can be written as
Z R
.1/ iSŒA.2/ C
ZŒJ D DA eiSŒA JA
%ŒA.1/ ; AP .1/ ; (2.34)
P=A
R R
where J D .J .1/ ; J .2/ / and JA D .J .1/ A.1/ C J .2/ A.2/ /. Note that ZŒJ D 0 D
Tr %O D 1. An n-point correlation function can be obtained as
ˇ
@n ZŒJ ˇ
ˇ
FD ˇ : (2.35)
.xk1 / : : : @J .x1 / ˇ
.2/ .2/ .1/ .1/
@Jn .xn / : : : @Jk .xk /@Jk1 1 JD0
In the case of fermions, all derivations must operate from the left.
2.3 Equilibrium
While for an arbitrary initial density matrix the above path integral is hardly
accessible for practical use (but see [10]), we can have more definite yet still simple
formulas if we are in equilibrium. There, the form of the density matrix is
1 ˇHO O
%O D e ; Z0 D Tr eˇH : (2.36)
Z0
Note that thermodynamics tells us that Z0 D eˇF , where F is the free energy of the
system.
Now we can exploit that formally, this density matrix is similar to the time
evolution operator with ˇ ! it, t ! iˇ. There we can use the representation
discussed in the previous subsection. Choosing the time arguments (only their
difference matters) ti ! ti iˇ, we have
Z R ti iˇ
O
eˇH D DA ei ti dt L
jAN i hA0 j : (2.37)
In the action, we can introduce the integration variable with the definition t D
ti i, and define new field variables as A.ti i/ D A.3/ ./. Then there appears
tZ
i iˇ Zˇ
SE D i dt LŒA.t/; @t A.t/ D d LŒA.3/ ./; i@ A.3/ ./: (2.38)
ti 0
The corresponding contour section continuously joins the end of segment 2, and
it is labeled as segment 3. The time contour then goes out to complex time values, as
can be seen in Fig. 2.1. For the generating functional, we introduce external currents
for all the sections of the contour. Then it has the form
D R E R
Z R
O
ˇ H J AO .1/ .2/ .3/
ZŒJ D Z0 e JA
D Tr e e D DA eiSŒA iSŒA SE ŒA C JA ; (2.39)
P=A
where now J D .J .1/ ; J .2/ ; J .3/ /. With this definition, in view of the second equality,
the contribution to Z0 D ZŒJ D 0 D Tr eˇHO comes purely from segment 3. The
free energy defined from ZŒJ can be considered a free energy in the presence of an
external time-dependent field.
The consequence of equilibrium is that we have time translation-invariance in
the observables. Therefore, we can put the initial density matrix anywhere in time,
even ti ! 1. But in this case, all propagation between contour sections C1;2 and
C3 is suppressed, since we would need infinitely long propagation. This means that
segment 3 factorizes from segments 1 and 2. The formalism that uses exclusively
Fig. 2.1 Closed time path

contour for equilibrium t
ti C1 tf
C3 C2
t i−iβ
2.4 Propagators 21
segment 3 is called Euclidean, or imaginary, time or the Matsubara formalism. The

theory built on contours 1 and 2 is called real time, or the Keldysh formalism.
If there is a conserved quantity in the system, i.e., if there is an operator NO that
commutes with the Hamiltonian ŒH; O N
O D 0, then we can associate a chemical
potential with this operator. Then the density matrix reads as
1 ˇ.H
O O O O
%O D e N/
; Z0 D Tr eˇ.HN/ : (2.40)
Z0
Formally, the treatment of this system is equivalent to that presented above just
involving a modified imaginary time evolution operator H O0 D H O N,
O or in the
0
path integral formalism with L D L C N.
2.4 Propagators
Let us consider the two-point functions in more detail. Path integration implies
contour time ordering in the integrand, so we have several choices for two arbitrary
local operators A and B:
.ij/
iGAB .t/ D hTC A.i/ .t/B.j/ .0/i (2.41)
(here and below, the distinctive sign O will be omitted from the operators for the sake
of notational simplicity). In terms of operator expectation values, we have
.12/ 1
iGAB .t/ D ˛ Tr eˇH B.0/A.t/
Z0
.21/ 1
iGAB .t/ D Tr eˇH A.t/B.0/
Z0
.11/ 1
iGAB .t/ D Tr eˇH TA.t/B.0/ D .t/ iG21 12
AB .t/ C .t/ iGAB .t/
Z0
.22/ 1
iGAB .t/ D Tr eˇH T A.t/B.0/ D .t/ iG12 21
AB .t/ C .t/ iGAB .t/
Z0
.33/ 1
GAB .t/ D Tr eˇH T A.i/B.0/ D ./ iG21 12
AB .i/ C ./ iGAB .i/:
Z0
(2.42)
In the case of nonzero chemical potential, we should use the replacement H !

.33/
H 0 . Remark: for GAB , as is usual, we do not include the imaginary factor i in the
definition.
The following identity is always true:
.11/ .22/ .12/ .21/

GAB C GAB D GAB C GAB : (2.43)
To avoid the use of nonindependent quantities, it is sometimes useful to change to

another basis, called a retarded/advanced (R/A) basis. We introduce the fields
A.1/ C A.2/
A.r/ D ; A.a/ D A.1/ A.2/ : (2.44)
2
The propagators of the ordered products of A.r/ and A.a/ are expressed in terms of
the previously defined contour-ordered quantities as follows:
.ra/ .11/ .12/

iGAB .t/ D iGAB .t/ iGAB .t/ D .t/.iG21 12
AB .t/ iGAB .t//
.ar/ .11/ .21/
iGAB .t/ D iGAB .t/ iGAB .t/ D .t/.iG21 12
AB .t/ iGAB .t//
.rr/ iG21 12
AB .t/ C iGAB .t/
iGAB .t/ D ; (2.45)
2
.aa/
while GAB D 0. The nonzero combinations are called the retarded, advanced, and
Keldysh propagators, respectively.
As these equations show, all propagators can be expressed through G.21/ and
.12/
G ; this is true also out of equilibrium. Moreover, in equilibrium, these two
apparently independent propagator are also related through the Kubo–Martin–
Schwinger (KMS) relation. Consider the 12 propagator of the A and B operators
and use the cyclic property of the trace to obtain
1 h 0
i 1 h 0 0 0
i
.12/
iGAB .t/ D ˛ Tr eˇH B.0/A.t/ D ˛ Tr eˇH eˇH A.t/eˇH B.0/ :
Z0 Z0
(2.46)
We assume that the operator A has a definite charge with respect to the symmetry,
i.e., ŒN; A D qA. Then this implies eˇN AeˇN D eˇq A. We also use the time
0 0
translation eiHt A.t/eiHt D A.t C t0 /, applicable also to imaginary t0 D iˇ values.
Then we have
0 0
eˇH A.t/eˇH D eˇH eˇN A.t/eˇN eˇH
D eˇq eˇH A.t/eˇH D eˇq A.t iˇ/: (2.47)
Therefore,
.12/
iGAB .t/ D ˛eˇq iG.21/ .t iˇ/; (2.48)
2.4 Propagators 23
which leads in Fourier space to the KMS relation:
.12/
iGAB .!/ D ˛eˇ.!Cq/ iG.21/ .!/: (2.49)
Now all operators can be expressed solely through G.12/ or G.21/ .

We can also define the spectral function % as
.21/ .12/
%AB .x/ D iGAB .x/ iGAB .x/: (2.50)
The spectral function has many advantageous properties, which we summarize in

Appendix A: it satisfies sum rule, and under some conditions, it is a positive definite
function of positive frequencies. Therefore, it is advantageous to choose the spectral
function for the basic quantity in equilibrium thermodynamics, describing two-point
correlations.
From the KMS relation (2.49), we obtain
.12/ .21/
iGAB .k/ D ˛n˛ .k0 Cq/%.k/; iGAB .k/ D .1C˛n˛ .k0 Cq//%.k/; (2.51)
where
1
n˛ .!/ D (2.52)
eˇ! ˛
is the Bose–Einstein (˛ D 1) or Fermi–Dirac (˛ D 1) distribution, respectively.
We note that although this is a discussion of the full interacting theory, the same
distribution function always appears, irrespective of the choice of the operators A
and B. The distribution function satisfies the equality
n˛ .!/ C n˛ .!/ C ˛ D 0: (2.53)
The dynamical information is exclusively contained in the spectral function. The

Keldysh propagator is expressed as

.rr/ 1
iGAB .k/ D C ˛n˛ .k0 C q/ %AB .k/: (2.54)
2
The proportionality function .1=2 C ˛n˛ .!// is a purely (anti)symmetric function

of !.
The retarded and advanced propagators read as
.ra/ .ar/
iGAB .x/ D .t/%AB .x/; iGAB .x/ D .t/%AB .x/: (2.55)
In Fourier space, using .!/ D i=.! C i"/j"!0C , one derives the convolution
Z
.ra/=.ar/ d! %AB .!; k/
iGAB .k/ D ; (2.56)
2 k0 ! ˙ i"
which is of the form of a dispersion relation, called the Kramers–Kronig relation.

The inverse relation is
.ra/
%AB .k/ D Disc iGAB .k/: (2.57)
k0
.ra/
Note that the discontinuity is equal to 2 times the imaginary part of GAB only if
the spectral function is real, i.e., when B D A .
Let us examine the imaginary time propagator, too. Since the imaginary contour
runs in the interval Œ0; ˇ, the range of the argument of the 33 propagator (which is
the difference of two imaginary time values) is 2 Œˇ; ˇ. For negative values,
by the definition (2.42), we have
.33/ .21/ .12/ .33/

GAB . C ˇ/ D iGAB .i iˇ/ D ˛iGAB .i/ D ˛GAB ./: (2.58)
With this relation we can extend the definition of the 33 propagator to the complete
imaginary axis as an (anti)periodic function.
Since in the ˇ > > 0 range, we have G.33/ ./ D iG.21/ .i/, it follows that
Z Z
.33/ d! i.i /! d! e.ˇ /!
GAB .; k/ D e .1 C ˛n˛ .!//%AB .!; k/ D %AB .!; k/:
2 2 eˇ! ˛
(2.59)
We arrive at a particularly simple form if B D A is a bosonic operator and the
spectral function depends only on k D jkj. Then the negative frequency parts can be
mapped to the positive frequency part, and

ˇ
Z1 cosh !
.33/ d! 2
GAA .; k/ D %AA .!; k/: (2.60)
2 ˇ!
0 sinh
2
In the Fourier space, being periodic, G.33/ in general has a discrete spectrum. We
use the following definition for the Fourier transformation:
Zˇ X
f .!n / D d ei!n f ./; f ./ D T ei!n f .!n /: (2.61)
n
0
2.5 Free Theories, Propagators, Free Energy 25
The (anti)periodicity dictates the possible values of the frequency:

2nT bosons
!n D (2.62)
.2 C 1/nT fermions:
The actual form of the Fourier transform reads as
Zˇ Z
.33/ i!n .33/ d! %AB .!; k/
GAB .!n ; k/ D d e GAB .; k/ D : (2.63)
2 ! i!n
0
If we compare it with the Kramers–Kronig relation obtained for the retarded

propagator, we obtain (k stands now for the four-vector of the momentum):
.33/
GAB .i!n ! k0 C i"; k/ D Gra
AB .k/: (2.64)
2.5 Free Theories, Propagators, Free Energy
The path integral can be performed if the action is Gaussian. In order to unify the
treatment of fermionic and bosonic fields, we write the Lagrangian density as
1 T
L .x/ D .x/K .i@/ .x/: (2.65)
2
Here can be either a (multicomponent) bosonic field or a (multicomponent)
fermionic field in the Nambu representation. The Nambu representation in terms
of the original (Dirac) representation can be expressed as

NT 0 KD .i@/
D ; K .i@/ D : (2.66)
KDT .i@/ 0
If the field satisfies some self-adjoint property, D 0 , then the Hermiticity

of the action requires K D 0 K 0 . The bosonic/fermionic nature dictates
K .i@/ D ˛K .i@/.
T
The operator EoM reads as
K .i@/ .x/ D 0; (2.67)
which implies that the spectral function % D % also satisfies the equation
K .i@/%.x/ D 0: (2.68)
In case of general relativistic theories with a single mass shell, there exists a
differential operator D.i@/ (Klein–Gordon divisor [4]), for which
K .i@/D.i@/ D @2 m2 ; (2.69)
the Klein–Gordon operator. Then
%.x/ D D.i@/%0 .x/; %.k/ D D.k/%0 .k/; (2.70)
where %0 .k/ is the spectral function of the one-component bosonic field (Klein–
Gordon field)
.@2 C m2 /%0 .x/ D 0; %0 .t D 0; x/ D 0; @t %0 .t D 0; x/ D iı.x/: (2.71)
This initial value problem has the following solution in real time:
sin !k t
%0 .t; k/ D i ; !k2 D k2 C m2 I (2.72)
!k
in the Fourier space, this implies
2
%0 .k/ D 2 sgn.k0 /ı.k2 m2 / D .ı.k0 !/ ı.k0 C !//: (2.73)
2!k
As an example we give the Klein–Gordon divisor for the fermion field in the
Dirac representation and for the gauge field in R gauges:
k k
D.p/ D p/ C m; D .k/ D g C .1 / : (2.74)
k2
We can also define the Euclidean version of the Klein–Gordon divisors as
kE kE
DE .p/ D ipE C m; DE .k/ D ı .1 / ; (2.75)
kE2
where E D f0 ; i g are self-adjoint 4 4 matrices.

Using the generic relations between the spectral functions and propagators, we
can write in particular
ˇ ˇ
D.k/ ˇˇ D.k/ ˇˇ
G.ra/ .k/ D ; G.ar/ .k/ D ;
k2 m2 ˇk0 !k0 Ci" k2 m2 ˇk0 !k0 i"
iG.12/ .k/ D ˛n˛ .k0 /D.k/%.k/; G.21/ .k/ D .1 C ˛n˛ .k0 //D.k/%.k/;
DE .kE /
G.33/ .kE / D .kE0 D .2n C 1 .˛//T/ (2.76)
kE2 C m2
2.5 Free Theories, Propagators, Free Energy 27
Now let us calculate the path integral in the Gaussian case. By completing the
square in the exponent of (2.39), we have schematically (using the Hermiticity of
the kernel)
i T i T i
K C J D 0 K 0 C J T K 1
J; (2.77)
2 2 2
where 0 D iK 1 J. Then, by introducing a new integration variable ! 0 ,

i T 1
we have ZŒJ D Z0 e 2 J K J . This form suggests that the generating functional
is a Gaussian function of the currents. In the formal derivation, we would have to
take into account carefully the boundary conditions when changing over to the new
variable. But a shortcut can be taken: we know that the second functional derivative
of the generating functional with respect to the currents yields the propagators.
Therefore, consistency requires
R d4 p
i
J T .p/G.p/J.p/
ZŒJ D Z0 e 2 .2/4 ; Z0 D Tr eˇH : (2.78)
We can also evaluate Z0 in the quadratic theory. We should take into account that
it comes entirely from the Euclidean part (segment 3) of the time path. The Gaussian
integral in the case of bosonic/fermionic variables yields
Z R
1
KE
Z0 D D e 2 D .det KE /˛=2 : (2.79)
The free energy is therefore

Z
˛T ˛V d4 p
FD ln det KE D ln det KE .p/: (2.80)
2 2 .2/4
For fermions, up to a constant, the integrand can be expressed through the Euclidean
Dirac kernel as 2 ln det KDE .p/ (see Eq. (2.66)).
To evaluate this expression formally, we derive a differential equation for it, by
shifting first KE ! KE C a, and then differentiating with respect to a. We have for
the free energy density f D F=V,
Z
@f ˛ d4 p ˛
D Tr.KE C a/1 D Tr G.33/
a .x D 0/; (2.81)
@a 2 .2/4 2
.33/
where the a subscript in Ga reminds us that here we have to work with the shifted
kernel. Using the definition of G.33/ through G.12/ and G.21/ (cf. (2.42)), we can
write the symmetric combination

Z
@f ˛ ˛ d4 p 1
D Tr iG.rr/
a .x D 0/ D C ˛n˛ .p0 / %0a .p/ Tr DEa .p/;
@a 2 2 .2/4 2
(2.82)
where we also have taken into account chemical potential.
To proceed, it is worth beginning with a brief discussion of the bosonic and
fermionic cases separately. In the bosonic case, the Klein–Gordon divisor DEa is
a projector, and the trace simply yields the dimension, i.e., the number of bosonic
degrees of freedom Nb . Here a has the meaning of a shift in the squared mass, so we
have
Z
@fb Nb d4 p 1
D C nC .p0 / %0 .p/; (2.83)
@m2 2 .2/4 2
where %0 is the Klein–Gordon spectral function with mass m.

For fermions, we do not have the 1=2 in passing to the Dirac representation DDEa ,
but now a has the meaning of a (linear) mass term. The trace of the kernel is Nf m.
Therefore, the m2 derivative can be written
Z
@ff Nf d4 p 1
D n .p0 / %0 .p/: (2.84)
@m2 2 .2/4 2
The two cases therefore yield a result that can be uniformly summarized as
Z
@f ˛N d4 p 1
D C ˛n .p
˛ 0 / %0 .p/
@m2 2 .2/4 2
Z
N d3 p
D 3
n˛ .!p / C n˛ .!p C / C ˛ ; (2.85)
2 .2/ 2!p
where we used the free Klein–Gordon spectral function. In this expression, the last
term is a constant not depending on the temperature. It has no thermodynamic
meaning,2 and so we omit it. The rest can be integrated, and we use the physical
condition f .m2 ! 1/ D 0 to obtain, eventually,
Z
N d3 p
f D Œf .T; / C f .T; /; f .T; / D ˛T ln.1 ˛eˇ.!p / /:
2 .2/3
(2.86)
2
This vacuum contribution used to be attributed to the Casimir effect, where we measure the energy
difference arising when the volume of the quantization space is changed. It is also worth noting that
fermions and bosons contribute with opposite signs, which means that in supersymmetric models,
the net zero-point contribution is zero.
2.6 Perturbation Theory 29
The two terms of this formula can be associated with the separate contributions
of particles and their antiparticles, respectively. The number of degrees of freedom
equals N=2 for both. The chemical potential of the antiparticles has opposite sign
relative to of the particles.
From the thermodynamic potential, one obtains the other thermodynamic quanti-
ties, too. The grand canonical potential f is connected to the internal energy density
by
f D " Ts n D p: (2.87)
The energy density and number density for one particle degree of freedom have the
expressions
Z Z
@.ˇf / d3 p @f d3 p
"D D !p n˛ .!p /; nD D n˛ .!p /:
@ˇ .2/3 @ .2/3
(2.88)
These expressions are finite unless ˛ D 1 and jj > m (the jj ! m p C 0 case
is still finite). In this pathological case, there appears a pole at p D 2 m2 .
Physically, this phenomenon is connected to the Bose–Einstein condensation: in
response to trying to increase the number density of particles, the system generates
a finite field condensate.
Some limiting cases are the following:
• If T m and D 0, one has, for both the fermionic and bosonic cases,
3=2
mT " 3 m
pDT eˇm ; D C : (2.89)
2 p 2 T
• If T
m; , one has
pb 2 7
4
D ; pf D pb ; " D 3p: (2.90)
T 90 8
2.6 Perturbation Theory
If the action contains nonquadratic terms, then we must rely on some approxima-
tions for the evaluation of the path integral. Under the assumption that the quadratic
(Gaussian) approximation captures the main physical features of the system, only
slight nonlinear modifications are expected. Then one can use perturbation theory.
In conformity with this understanding, the action is split into a quadratic and a
higher-order part,
SŒ D S.2/ŒA C Sint ŒA; (2.91)

and we write formally

Z
Sint ŒA D d d x .g/A1 .x/ : : : Av .x/ (2.92)
for the interaction part. In general, g could contain local operations such as
derivations, which we leave hidden in this overview. We recognize next that the
generating functional (2.39) can be written as an expectation value of the free theory:
D R E
ZŒJ D Z00 eiSint ŒAC JA ; (2.93)
0
where Z00 is the free partition function at J D 0. For perturbation theory, we expand
the exponential in a Taylor series:
1 D R E
X1
ZŒJ D Z00 .iSint ŒA/m e JA : (2.94)
mD0
mŠ 0
Formally, we can also rewrite it as

ı
D R E ı i
R T
ZŒJ D Z00 eiSint Œ ıJ e JA D eiSint Œ ıJ e 2 J GJ : (2.95)
0
For a certain n-point function, we need to apply the required functional deriva-
tives to this expression and eventually put J D 0 (cf. (2.35)). This leads to the
Feynman rules for the computation of the n-point function hTc Ai1 .p1 / : : : Ain .pn /i,
where Ai .p/ are fundamental fields with momentum p and with a generic index (also
including the Keldysh indices) i:
• draw a graph with points and links (lines);
• associate the points where only a single link ends (external legs) to the fields
Aik .pk /, k D 1 : : : ; n, where n is the number of this type of point;
• associate the points with v links to the piece in the interaction Sint containing the
product of v field operators. These points are called interaction vertices;
• associate propagators to the links connecting any two points. A number of
externally determined propagators enter into each vertex and generate virtual
fields propagating along the internal lines of the diagram.
Draw all possible diagrams with n external fields and m vertices, and eventually
evaluate them with help of the following rules:
• external legs force the joining propagator to have momentum pk ;
• a link connecting points with indices i and j is represented by iG.ij/ .q/;
Pv having incoming momenta q1 ; : : : ; qv yields a contribution ig.2/
d
• a vertex
ı. aD1 qa /.
2.7 Functional Methods 31
Fig. 2.2 The tadpole and the

q q
bubble diagrams
k k
k−q
In the last step, one multiplies all these contributions, and an extra factor 1=mŠ
is included in the mth order of the perturbation theory. Finally, one evaluates the
R dd qi
integrals over the momenta flowing through the internal links: ˘i .2/ d .
At the one-loop level, there are two diagrams that are the most important: the
tadpole and the bubble diagrams; cf. Fig. 2.2. The details of their computation can
be found in Appendix B.
2.7 Functional Methods
Perturbation theory can be made more efficient if we realize that the only quantities
that need calculation are the loop integrals arising from the rules for setting up
Feynman diagrams described in the last section. Therefore, it is worth developing a
formalism that concentrates solely on them.
The first step is to introduce WŒJ, the generator of the connected diagrams:
ZŒJ D eWŒJ : (2.96)
Since the generator of the imaginary time diagrams ZŒJ3 can be considered the
partition function in the presence of an external field, it follows that WŒJ3 is the
free energy in the presence of an external field.
The generator W generates the connected diagrams, which can be proved by
differentiating Z with respect to the current Jx as follows. The differentiation yields
a series of diagrams with one external leg. There is a portion of each diagram
connected to the external leg; if we denote by W N the generator of the connected
diagrams, this portion can be computed as ı W=ıJ N x . The remainder, on the other
hand, is the sum of all zero leg diagrams, i.e., just ZŒJ. Therefore, we can write
down the functional differential equation
ıZŒJ N
ı WŒJ
D ZŒJ: (2.97)
ıJx ıJx
N
Its obvious solution is ZŒJ D eWŒJ N
. Thus we have W D W.
The connected n-point function is the (left) derivative of W at the zero external
field. In particular, the propagator is calculated as
@2 WŒJ
iGxy D : (2.98)
@Jx @Jy
Using (2.93), we find that
i ˛ˇV D R E
WŒJ D Jx Gxy Jy Tr ln KE C e JA eiSint ŒA 1 : (2.99)
2 2 conn
Then the free energy reads
ˇ D R E
FŒJ D Jx GE;xy Jy C F0 C ˇ e JA 1 eSE;int ŒA ; (2.100)
2 conn
with F0 denoting the free energy of the free system evaluated in the previous section.
Here W (or the free energy F) is a function(al) of the external current. The external
current can be set by hand. This is the “tool” for controlling the system from outside;
thermodynamically, the currents/sources correspond to the extensive variables. We
may, however, want to express the dependence of the free energy on
@W
AN x ŒJ hAx iJ D ; (2.101)
@Jx
where the subscript J is to emphasize that the expectation value is to be evaluated

in the presence of the external current. This is a dynamical quantity, thermody-
namically corresponding to the intensive variables. In thermodynamics, this type of
change is realized by applying a Legendre transformation, which will be applied
here, too.
Technically, the change of W under an infinitesimal change of J reads as ıW D
ıJx AN x (the ordering is important in the case of fermionic theories). We define the
associated potential via a Legendre transformation:
Z
N D WŒJ
i ŒA Jx AN x : (2.102)
In the Euclidean case, we proceed by the replacement i ! E . Since WE is

N As
the free energy expressed through J, E is the free energy expressed through A.
can be easily seen, ıi D Jx ı AN x , or
@i
D Jx : (2.103)
@AN x
In the fermionic case, it is a right derivative.

The value of thus defined has many interesting properties. Diagrammatically,

hAx iJ is a one-point function. Fixing the value of AN means fixing the value of the
one-point function. Therefore, we must ensure the vanishing of the perturbative
corrections to the one-point function. This requirement is equivalent to the omission
of all diagrams that contain a part connected to the rest by a single line. Such
diagrams are called one-particle reducible diagrams. Those diagrams that remain
connected after cutting a line form the class of one-particle irreducible (1PI)
diagrams. Therefore, can be interpreted as being the generator of the 1PI
diagrams.
The physical expectation values must be computed at zero external current,
which means that the physical value of AN can be determined as a solution of
ˇ
@ ˇˇ
D 0: (2.104)
@AN x ˇAN phys
This is the same EoM as obtained in the classical case, except that one replaces the
classical action by . Therefore, the generator of the 1PI diagrams is also called the
quantum effective action.
In the functional integral representation, we have
R
Z R
Z R
N J AN N N
ei ŒA D eiWŒJ D DA eiSŒAC J.AA/
D Da eiSŒACaC Ja
; (2.105)
where in the last equality, the integration variable is shifted to a by A D AN C a,

and the full quantum field is split into the sum of the expectation value (mean field,
background) AN and the fluctuations around it, a. This means that hai D hAi AN D 0,
i.e., the expectation value of the fluctuations is zero.
In fact, the role of the external current is to ensure that the fluctuations have zero
expectation value. This can equivalently be required by an extra condition as
Z ˇ
N N ˇ
ei ŒA D Da eiSŒaCA ˇˇ : (2.106)
haiD0
This is the compact formulation of the background field method.

From this formula, it is also evident that if we require a D 0 instead of hai D 0,
i.e., no fluctuations are allowed at all, then D S will be the result. This means that
the quantum effective action to leading order is the classical action:
D S C quantum fluctuations: (2.107)
Together with (2.104), this formula further supports the interpretation of as the
quantum effective action.
2.7.1 Two-Point Functions and Self-Energies
Let us investigate, in particular, the second derivative of at the physical value.

The 1PI quantum correction to the two-point function is called the self-energy i˙:
@2 i
ixy.1;1/ D D iKxy i˙xy ; (2.108)
@AN x @AN y
2
where Kxy D @AN@ @SAN is the classical kernel. In the case of fermion fields, we have to
x y
apply one left and one right derivative here, which is indicated by the upper (double)
index of the notation introduced on the left-hand side.
On differentiating (2.103) with respect to J from the left, we obtain
@2 W @2 i
D ıxy : (2.109)
@Jx @Jz @AN z @AN y
The second derivatives of W and are therefore the inverse quantities of each other.
We can write this expression as
Gxz .Kzy ˙zy / D .Kxz ˙xz /Gzy D ıxy : (2.110)
In the Euclidean formalism, we define the self-energy as
.KE C ˙E .p//GE .p/ D ıxy : (2.111)
Since ˙ is the (quantum part of the) 1PI two-point function and G is the inverse of
the full two-point function, (2.110) generates the iterative relation that tells how the
1PI diagrams should be resummed in order to give the full result. To leading order
we have ˙ D 0, and so G0 D K 1 . With it we obtain formally
G D G0 CG0 ˙G D G0 CG˙G0 D G0 CG0 ˙G0 CG0 ˙G0 ˙G0 C : (2.112)
This formula is sometimes called the Dyson–Schwinger (DS) series.

From this formula, a simple rule can be read off how to determine the self-energy.
The first correction to the two-point function that is ı hTC AAi D hTC AAi iG0 can
always be written formally as
ı hTC AAi D iG0 ı hTC AAiamp iG0 D iG0 ˙G0 C ; (2.113)
or in the Euclidean formalism,

˝ ˛
ıhTC A.3/ A.3/ i D G0 ı TC A.3/ A.3/ amp G0 D G0 ˙E G0 C ; (2.114)
which means
˙ D iı hTC AAiamp ; ˙E D ıhTC A.3/ A.3/ iamp : (2.115)
The quantity ı hTC AAiamp is the “amputated” two-point function, where we chop off
the propagators belonging to the external lines.
Although it looks quite simple, (2.112) in fact avoids by resummation a series
of individually increasingly divergent contributions. Namely, the free propagator
exhibits a pole at p2 D m2 on the mass shell. Near the mass shell, we have p2 D
m2 C x, and the free propagator behaves as G0 1=x. Therefore, the nth term
in the DS series is proportional to 1=xn . These terms are therefore more and more
divergent at a finite value of the momentum: this is characteristic of an infrared
(IR) divergence. The DS series resums the most relevant terms and provides the
propagator G free of unphysical divergences. It also may have a pole, but that is
already physical: no further diagrams can change it. This logic will be followed
later in performing the resummation of IR divergent (IR sensitive) series.
The propagator is in general a matrix expressing the correlation of all possible
field components at the same momentum (if we have spacetime translation-
invariance). This holds for the self-energy, too. In particular, if we use the real-time
formalism, then ˙ is a matrix with Keldysh indices 1; 2. Omitting other compo-
nents, the DS equation can be written as
K .p/G.ij/ .p/ D .1/iC1 .ı ij C ˙ .ik/ .p/G.kj/.p//; ; i; j; k 2 f1; 2g; (2.116)
i.e., the sign is C for i D 1 and for i D 2. In the R/A formalism, G.aa/ D 0 implies
˙ .rr/ D 0. The previous equation can be rewritten with help of the 1 Pauli matrix:
0
h 0 0
i
K .p/G.ab/ .p/ D 1ab ı b b C ˙ .b c/ .p/G.cb/ .p/ ; a; b; b0 ; c 2 fr; ag:
(2.117)
This relation shows that the retarded propagator is not mixed with the others:
K .p/G.ra/ .p/ D 1 C ˙ .ar/ .p/G.ra/.p/: (2.118)
The generic relation between the propagators discussed in Sect. 2.4 implies
relations between the self-energies. In particular,
˙ .12/ C ˙ .21/
˙ .11/ D ˙ .ar/ ˙ .12/ D ˙ .ra/ ˙ .21/ ; ˙ .aa/ D (2.119)
2
Moreover,
˙E .!n / D ˙ .ar/ .k0 ! i!n /: (2.120)

The causal behavior of the retarded propagator implies that ˙ .ar/ is either local or at
least causal itself. Therefore, after separating the local contribution ˙0 , one writes
.ar/
for the remainder (˙1 ) a dispersive representation:
Z
.ar/ .ar/ d! Disck0 i˙ .ar/ .k0 ; k/
˙ .ar/ .k/ D ˙0 C ˙1 .k/; ˙1 .k/ D : (2.121)
2 k0 ! C i"
The discontinuity can be expressed through other self-energies, too:
Disc˙ .ar/ D ˙ .ar/ ˙ .ra/ D ˙ .12/ ˙ .21/ (2.122)
Finally, the relation between G.rr/ and the retarded propagator implies that

.aa/ 1
˙ D C ˛n˛ .k0 / Disc i˙ .ar/ (2.123)
2
This means that full knowledge of the retarded self-energy (or equivalently, the Mat-
subara self-energy) is enough to fully reconstruct all the self-energies of the system.
2.7.2 Higher n-Point Functions
Further derivatives of (2.109) exploiting (2.101) provide relations between the 1PI
and the connected n-point functions. In particular,
.3/ .3/
iWijk D iGii0 iGjj0 iGkk0 ii0 j0 k0 ;
.4/ .3/ .3/
iWijk` D iGii0 iGjj0 ii0 j0 k0 iGk0 a iabc iGbk iGc` C : : :
.4/
CiGii0 iGjj0 iGkk0 iG``0 ii0 j0 k0 `0 :
(2.124)
Here we used the notation
.n/ @n ŒA N .n/ @n WŒJ

i1 :::in D ; Wi1 :::in D : (2.125)
@Ai1 : : : @AN in
N @Ji1 : : : @Jin
In the case of fermions, one has to pay attention to the order of differentiation.
These formulas correspond to the tree-level relations between the n-point
functions and the vertices when vertex strengths are derived from the classical
action. The derivatives of the effective action therefore play the role of the classical
vertices: hence the name proper vertices. In contrast to the classical theory, quantum
fluctuations produce nonzero values for arbitrary high n-point functions.
2.8 The Two-Particle Irreducible Formalism 37
2.8 The Two-Particle Irreducible Formalism
In the previous subsection we investigated the generator of the 1PI diagrams, where
the system is forced by a choice of appropriate external current distribution to take a
predetermined value AN D hAi for the expectation value of the field operator. We can
go on with this logic and try to impose other constraints on the path integral. The
next step is to fix the value of the two-point functions as well. Technically, what we
have to do is to assign an external current to the two-point function and determine
its value from the requirement that the exact two-point function coincide with the a
priori chosen expression.
Thus we define an extended generator functional containing both an external
current J and an external bilocal source term R:
Z
1
ZŒJ; R D e WŒJ;R
D DAeiSŒACJx Ax C 2 iRxy Ax Ay ; (2.126)
where we have suppressed the explicit indication of summation/integration. We

want to fix the values of the field expectation value to AN and the connected two-
N
point function to G:
@WŒJ; R
D hAx i D AN x ;
@Jx
@WŒJ; R ˝ ˛
2 D TC Ax Ay D iG N xy C AN x AN y : (2.127)
@iRxy
From these equations, one can in principle determine J and R.

Just as in the 1PI case, we define the extended effective action as
i ŒA; N D WŒJ; R Jx AN x 1 iRxy .AN x AN y C iG

N G N xy /: (2.128)
2
After differentiation (right differentiation in the case of fermi fields), this double
Legendre transform provides us with the relations
N G
@i ŒA; N N G
@i ŒA; N 1
D Jx iRyx AN y ; D iRxy : (2.129)
@AN x @iGN xy 2
In the physical point we should set the external sources to zero, and therefore, we
have
ˇ ˇ
N G
@i ŒA; N ˇ N G
@i ŒA; N ˇ
ˇ D 0; ˇ D 0; (2.130)
@AN x ˇphys @iGN xy ˇphys
which suggests that is indeed the quantum analogue of the extended classical
N obtained with fixed AN function
action. By substituting back the physical value of G
from the solution of the second equation, we obtain the (1PI) effective action:
N D ŒA;
ŒA N GN phys ŒA:
N (2.131)
Since AN and GN are the exact one- and two-point functions, respectively, there
should be no quantum correction to them, and this is true even on the internal
lines of the diagrams. In the case of the predetermined one-point function, this has
had the consequence that the effective action is the generator of 1PI diagrams. In
a similar manner for the case of the exact two-point function, only two-particle-
irreducible (2PI) diagrams can appear in the perturbation theory, which do not fall
apart when we cut two internal lines of a diagram. In other words, we do not have
perturbative self-energy corrections to the propagators representing the lines in the
corresponding Feynman diagrams (called also skeleton diagrams). If we apply this
requirement to the physical value of the propagator, using the fact that on the one
hand, hTC AAiconn equals iG N phys , and on the other hand, we can express it by the
self-energy, we obtain
N 1
G 1 N N
phys D G0 ˙2PI ŒA; Gphys : (2.132)
This is a self-consistent (gap) equation that allows to determine the value of the
physical connected propagator.
In special cases, we can determine ŒA; N G.
N If there is no quantum correction at
all, then the path integral substitutes the solution of the classical EoM, which is the
same as the first equation of (2.129) with ! S. Classically, G N D 0, and so the
classical 2PI effective action reverts to the classical 1PI case, i.e., D S.
In the free case, we can solve the system even in the presence of external sources.
In bosonic theories, in copying the steps leading to (2.78), one arrives at
1 1
WŒJ; R D J i.G1 1
0 C R/ J Tr ln.G1
0 C R/: (2.133)
2 2
This implies
@iWŒJ; R N
D i.G1 1
0 C R/ J D A;
@J
@iWŒJ; R N xy C AN x AN y ; (2.134)
2 D J i.G1 2 1
0 C R/ J C i.G0 C R/
1
D iG
@iR
which means
G1 N 1
0 CRD G ; AN D iGJ;
N
N G
i ŒA; N 1 Tr ln G
N D iSŒA N 1 1 Tr G1 N
0 G: (2.135)
2 2
2.8 The Two-Particle Irreducible Formalism 39
Motivated by the result of the free theory, in the interacting case we write
N G
i ŒA; N 1 Tr ln G
N D iSŒA N 1 1 Tr G1 N N N
0 G C iint ŒA; G; (2.136)
2 2
where in the correction part int , we should use the exact propagator GN and keep
only the 2PI diagrams. The physical value of the propagator, using (2.130), leads to
(2.132) with
N N
˙2PI ŒA; N D 2 @iint ŒA; G :
N G (2.137)
@G N
A peculiarity of the 2PI formalism is that we can access the same correla-
tion functions in different ways. For example, the self-energy can be expressed
from (2.137) as the derivative of the 2PI action with respect to G, N but it is also the
N
second derivative with respect to A. This ambiguity remains true for all higher point
functions. For the exact correlation functions, these definitions yield the same result,
but in the perturbative expansion, usually we find deviations. The derivative of the
1PI effective action, for example, always respects the symmetry of the Lagrangian,
and so it satisfies Goldstone’s theorem in the case of spontaneous symmetry-
breaking (SSB) of a continuous symmetry. This is not true for the solution of the
2PI equation (2.132) if we substitute back the physical value of the background. The
reason is that Goldstone’s theorem is satisfied at each order of perturbation theory as
a result of subtle cancellations between the self-energy and the vertex corrections.
Therefore, when we resum all the self-energy diagrams up to a certain order using
the 2PI formalism and leave the vertex corrections at their tree-level perturbative
value, we easily find a mismatch. We will return to this point later, in Sect. 4.7.
This line of thought can be continued, and we can demand that the correlation
functions up to n-point functions have predetermined values, denoted by AN
N V2 and Vk ; k > 2; in general, we refer to them as Vk (k D 1; : : : ; n)
V1 ; G
(cf. [11]). Similarly as above, we can force the system to provide the required
n-point functions by introducing external currents besides Jx ; Rxy . These will be
.k/
denoted by Rx1 :::xn (k D 1; : : : ; n):
Xn
1 .k/
iS ! iS C R A x : : : A xk : (2.138)
kD1
kŠ x1 ;:::;xx 1
The derivative of WŒR.k/ with respect to these sources will give us the k-point
functions. We require that the fully connected part coincide with the predefined
values.
We can also perform a Legendre transformation to obtain the free energy ŒVk ,
which is the functional of Vk , where we can also express the external sources
R.k/ through Vk . The physical value of the vertices comes from the requirements
ı ŒVk =ıVj D 0. Since we fixed all the correlation functions up to the n-point
functions, the perturbative expansion should contain only n-particle-irreducible

(nPI) diagrams, which remain connected even after cutting n internal lines.
2.9 Transformations of the Path Integral
In this subsection we will perform some changes in the integration variable of the
path integral
Z
ZŒJ D DAeiSŒACJA : (2.139)
Let us consider an infinitesimal transformation Ai ! A0i D Ai C "i Ai ŒA for which

the corresponding Jacobian is unity to linear order in "i (no summation is understood
for i):

@Ai @ Ai Š
ln Jacobian D ln det ıij C "i D Tr "i C O."2 / D 0: (2.140)
@Aj @Aj
The change in the path integral up to leading order in "i reads as

Z Z
0 0
ZŒJ D DA0 eiSŒA CJA D DAeiSŒAC"i Ai CJ.AC"i Ai /
Z
@S
D DAe 1 C i"i
iSŒACJA
C Ji "i Ai
@"i

@S
D ZŒJ 1 C i"i iJi Ai :
@"i
(2.141)
The invariance under this change of variables implies

@SŒA
iJi Ai ŒA D 0; (2.142)
@"i
a true set of relations for any transformation.

One way in which we can exploit this equation is to take its nth derivative with
respect to Ja1 ; : : : ; Jan :
* +
ıS Xn
iJi Ai Aa1 : : : Aan i ıiak Aa1 : : : Aak1 Aak AakC1 : : : Aan D 0:
ı"i kD1
(2.143)
2.9 Transformations of the Path Integral 41
In the physical vacuum, J D 0, and we obtain

X
ıS
n
˝ ˛
Aa1 : : : Aan Di ıiak Aa1 : : : Aak1 Aak AakC1 : : : Aan phys : (2.144)
ı"i phys kD1
We can also use (2.142) to find equations for W and , exploiting that

1 @ @ @ @W
hf .A/i D f ZŒJ D eW f eW D f C : (2.145)
ZŒJ @J @J @J @J
Therefore, we should substitute A ! @J C @W=@J in (2.142) to obtain its functional

expression and set J D 0 at the end.
To have an equation for the effective action, we use (2.103) and
@ @Ay @ @2 W @ @
D D D iGxy (2.146)
@Jx @Jx @Ay @Jx @Jy @Ay @Ay
(in the fermionic case, left differentiation should be applied here). Since iJ D @ =@A
(with right differentiation), we have

@ @ @ @
S A C iG D Ai A C iG : (2.147)
@"i @A @A @Ai
The expressions in the square brackets denote the “variables” of the functionals S
and Ai , respectively. We should emphasize that (2.142) and therefore (2.147) are
true for every transformation. Specifying the form of the transformation leads to
different applications.
2.9.1 Equation of Motion, Dyson–Schwinger Equations
In the simplest application, the transformation is a simple shift of the field value,
that is, A0 .x/ D A.x/ C ".x/. Then A D 1, and the " derivative is the same as the A
derivative. Then we obtain the Dyson–Schwinger equations

@S @S @ @
iJi D 0 or A C iG D : (2.148)
@Ai @Ai @A @Ai
In particular, for the physical vacuum (J D 0), we obtain the EoM

ˇ ˇ
@S ˇˇ @S @ ˇˇ
D0 or A C iG D 0: (2.149)
@Ai ˇphys @Ai @A ˇphys
An illustrative example of the application of these general formulas is presented in

the next subsection. Taking the derivative of the previous equation with respect to
Ja1 ; : : : ; Jan we obtain, in the physical vacuum J D 0,
X
ıS
n
˝ ˛
Aa1 : : : Aan D i ıiak Aa1 : : : Aak1 AakC1 : : : Aan : (2.150)
ıAi kD1
This relation corresponds to (2.144) for the actual transformation.

In the real time formalism we should take into account that the Lagrangian is
L D L ŒA.1/ L ŒA.2/ , and so
ˇ
@S ˇ
iC1 @S ˇ
D .1/ : (2.151)
@A.i/ @A ˇADA.i/
Therefore,
ˇ X
@S ˇˇ ˝ ˛
n
Aa1 : : : Aan D i.1/ iC1
ıiak Aa1 : : : Aak1 AakC1 : : : Aan : (2.152)
@A ˇADA.i/ kD1
2.9.2 Ward Identities from a Global Symmetry
If the transformation A0 D AC"A corresponds to a global symmetry of the system,

then the derivative is (cf. (2.2))
@S
D @ ji : (2.153)
@"i
We remark here that j is the conserved current, which is a functional of the field
variables, while J is the current assigned to sustain a certain field configuration. In
this way, we obtain

˝ ˛ @ @ @
@ ji iJi Ai D 0 or @ j A C iG D Ai A C iG :
@A @Ai @A
(2.154)
For the n-point function, we obtain, from (2.144),
˝ ˛ X
n
˝ ˛
@ ji Aa1 : : : Aan D i ıiak Aa1 : : : Aak1 Ai AakC1 : : : Aan : (2.155)
kD1
This is the generic form of the Ward identities. These are nothing other than the
current conservation equations for the quantum case.
2.9 Transformations of the Path Integral 43
We discuss here three short applications.

1. If we integrate over the variable i (which also contains the spacetime integration),
we obtain
X
n
˝ ˛
0D Aa1 : : : Aak1 Aak AakC1 : : : Aan : (2.156)
kD1
The right-hand side is the total change in the expectation value
0 D hAa1 : : : Aan i : (2.157)
This means that the expectation value of every n-point function is invariant under
the global symmetry transformation. Note that this is true for the complete n-
point function, not just the connected part.
2. In the case of linear continuous transformations, Ai D Tij Aj . Taking the
integrated form of the effective action expression with a constant (i.e., zero-
momentum) A field, we have
@
0D Tij Aj ; (2.158)
@Ai
where one sums over repeated indices. Taking its functional derivative yields
@2 @
Tij Aj C Tik D 0: (2.159)
@Ak @Ai @Ai
@
In the physical point, @A i
D 0, but the field itself can take a (constant) expectation
value in the SSB case. Therefore,
@2
Tij Aj D 0; (2.160)
@Ak @Ai
which is Goldstone’s theorem: the inverse propagator has a zero mode at zero
momentum in the SSB case.
3. Remaining in the class of linearly represented symmetries, we write down the
Ward identity for two fields explicitly, displaying the spacetime indices:
˝ ˛ ˝ ˛
@x j .x/Ai .y/Aj .z/ D iı.x y/Tik Ak .y/Aj .z/ C iı.x z/Tjk hAi .y/Ak .z/i :
(2.161)
We assume that we are in the symmetric case (the SSB case is only formally
more complicated). One introduces for the left-hand side a 1PI proper vertex:
˝ ˛
j .x/Ai .y/Aj .z/ D iGii0 .y y0 / iGjj0 .z z0 / i0 j0 .x; y0 ; z0 /: (2.162)
The Ward identity takes the following form after passage to Fourier space:

Gii0 .p/Gjj0 .q/.ik /i0 j0 .k; p; q/ D Tik Gjk .q/ C Tjk Gik .p/; (2.163)
where k C p C q D 0. Multiplying by the inverse propagators gives us

.ik /ij .k; p; q/ D G1 1
ik .p/Tkj C Gjk .q/Tki : (2.164)
There is a special case, in which the propagation is diagonal (Gij D Gi ıij and
Tij D Tji . Then

.ik /ij .k; p; q/ D Tij G1 1
i .p/ Gj .q/ : (2.165)
This is an often used consequence of the Ward identity.
2.10 Example: ˚ 4 Theory at Finite Temperature
The simplest example for perturbation theory that is appropriate for demonstrating
the mechanisms of finite temperature calculations is the ˚ 4 theory of a one-
component scalar field. The Lagrangian reads as
1
L D ˚.@2 m2 /˚ ˚ 4 : (2.166)
2 24
This is the basic form of the Lagrangian. However, we should use it in other forms
at finite temperature. The free part was discussed in Sect. 2.5; the propagators are
listed in (2.76). Here the Klein–Gordon divisor is 1. The Euclidean and Keldysh
forms of the interactions are respectively
4 4
LE;int D ˚ ; Lint D ˚1 ˚24 ; (2.167)
24 24
while in the R/A formalism, we obtain

Lint D ˚r3 ˚a ˚r ˚a3 : (2.168)
6 24
In the perturbation theory, first we calculate the 1PI correction to the 2-point
function (the self-energy). This is the amputated connected 2-point function, as can
be seen from (2.115). At one loop level in the coordinate space in the Euclidean
formalism we obtain
˝ ˛
˙E .x/ D TC ˚ .3/ .x/˚ .3/ .0/.SE;int / amp D G.33/ .0/ı.x/ D T ı.x/;
2 2
(2.169)
2.10 Example: ˚ 4 Theory at Finite Temperature 45
where T is the tadpole function defined in Appendix B.1. In momentum space, the
self-energy is a constant. Therefore, in the R/A formalism, ˙ .ar/ D ˙ .ra/ D ˙E and
˙ .aa/ D 0. Physically, it is a mass correction. Using the results of Appendix B.1
derived with dimensional regularization, at zero temperature we obtain for the
second derivative of the effective action

.2/ 2 2 m2 1 m2
.p/ D p C m C C E 1 C ln ; (2.170)
32 2 " 42
while in the high-temperature expansion, we have

.2/ 2 2 T 2 mT m2 1 4T 2
.p/ D p C m C C C E ln : (2.171)
24 8 32 2 " 2
In the symmetric phase, the three-point function vanishes; the four-point proper
vertex is therefore the amputated four-point function. The first quantum correction
is of second order in the coupling constant expansion. In the Euclidean formalism,

.4/ 1 2
E .p1 ; p2 ; p3 ; p4 / D TC ˚.p1 /˚.p2 /˚.p3 /˚.p4 /.1 SE;int C SE;int / ;
2 amp
(2.172)
where we should use fields on the Euclidean contour (segment 3). The first term
gives a disconnected piece. For the connected part, we have from momentum
conservation
.4/ .4/
E;conn .p1 ; p2 ; p3 ; p4 / D NE .p1 ; p2 ; p3 ; p4 /.2/4 ı.p1 C p2 C p3 C p4 /;
where
.4/ 2
NE .p1 ; p2 ; p3 ; p4 / D ŒIE .p1 C p2 / C IE .p1 C p3 / C IE .p1 C p4 / ;
2
(2.173)
where IE .k/ is the contribution from the bubble diagram, defined in Appendix B.2.
We can also give the effective potential to this lowest order. We use the
background field method of (2.106), i.e., we shift the field ˚ ! ' C ˚N and omit
linear terms in the fluctuations (and at higher order all contributions to the 1-point
function). We obtain for the shifted Lagrangian
N C 1 '.@2 m2 ˚N 2 /' ˚N ' 3 ˚N 4 :

L D L .˚/ (2.174)
2 2 6 24
The first term gives the classical effective potential. For the first quantum correc-
tions, we need just the free fluctuation part, which will lead to some quantum
correction ı . We could use the formula (2.86) with D 0, ˛ D 1, but
here we should keep the regularized form. We can use that for constant fields,
@ı N In the self-energy one can take p D 0, since it is a tadpole,

D 12 ˙E .p D 0; ˚/.
@˚N 2
and does not depend on the momentum in any way. In ˙, all the ˚-dependence
is through the mass term. Therefore, ˙E .M/ D @M @ı
2 , where M
2
D m2 C 2 ˚N 2 .
Using (2.169), we have
@ı 1
D T .M; T/: (2.175)
@M 2 2
2T4
The integration constant is ı jMD0 D 90
. Finally, we obtain at zero temperature

m2 N 2 M4 M2
ıE .T D 0/ Veff .T D 0/ D ˚ C ˚N 4 C D" C ln ;
2 24 64 2 42
(2.176)
and in the high-temperature expansion,

2 T 4 m2 N 2 T 2 M 2 M 3=2 T M4 3 4T 2
Veff D C ˚ C ˚N 4 C C D" C ln ;
90 2 24 24 12 64 2 2 2
(2.177)
where D" D 1" C E 32 is a divergent constant. We remark here that in the
spontaneously broken phase, where m2 < 0, the quantum corrections cannot be
interpreted physically for ˚N 2 values yielding M 2 < 0 because of the M 3=2 term.
We may also illustrate the Dyson–Schwinger equations by means of the example
of the ˚ 4 model. We begin with the generator of the 1PI DS equations (2.148):
3 .3/

x.1/ D Kxy ˚y ˚x C 3˚x iGxx C iGxx0 iGxy0 iGxz0 x0 y0 z0 ; (2.178)
6
where K .p/ D p2 m2 . Here the indices of the different quantities refer simply
to spacetime points, x is the free argument, and one has summation on the others.
After differentiating once more with respect to ˚y , we obtain
2
xy.2/ D Kxy ˚x C iGxx ıxy
2
.3/ .3/ .3/

˚x iGxa iGxb iaby C iGxy0 iGxz0 x0 y0 z0 iGxa iGx0 b iaby
2
.4/
iGxx0 iGxy0 iGxz0 x0 y0 z0 y : (2.179)
6
As a specific application, we can reproduce the perturbative results: we go to the
symmetric phase, where ˚ D 0, and we restrict ourselves to the lowest order (i.e.,
we neglect terms O.2 /). We obtain in the Fourier space

.2/ .p/ D p2 m2 iGxx : (2.180)
2
References 47
This form actually agrees with the 2PI equation (2.132) with the self-energy
from (2.169). Therefore, in fact, this equation is a gap equation for the propagator.
References
1. M.E. Peskin, D.V. Schroeder, An Introduction to QFT (Westview Press, New York, 1995)
2. C. Itzykson, J.-B. Zuber, Quantum Field Theory (McGraw-Hill, New York, 1980)
3. J.D. Bjorken, S.D. Drell, Relativistic Quantum Fields (McGraw-Hill, New York, 1965)
4. N.P. Landsmann, Ch.G. van Weert, Real- and imaginary-time field theory at finite temperature
and density. Phys. Rep. 145, 141–249 (1987)
5. M. Le Bellac, Thermal Field Theory (Cambridge University Press, Cambridge, 1996)
6. J.I. Kapusta, C. Gale, Finite-Temperature Field Theory, Principles and Applications (Cam-
bridge University Press, Cambridge, 2006)
7. I. Montvay, G. Münster, Quantum Fields on a Lattice (Cambridge University Press, Cambridge,
1994)
8. J. Collins, Renormalization. Cambridge Monographs for Mathematical Physics (Cambridge
University Press, Cambridge, 1984)
9. L.D. Landau, E.M. Lifshitz, Course of theoretical physics, in The Classical Theory of Fields,
vol. 2 (Butterworth-Heinemann, Oxford, 1975)
10. M. Garny, M.M. Muller, Phys. Rev. D 80, 085011 (2009)
11. J. Berges, Phys. Rev. D 70, 105010 (2004)
Chapter 3
Divergences in Perturbation Theory
Already in the simplest examples where Feynman diagrams involving loops are
calculated, perturbative corrections to the tree-level result turn out to be divergent,
or very large in certain cases. Several of these divergences can be identified in the
example of the one-loop self-energy of ˚ 4 theory. In (2.169), we have seen that " !
0 (dimensional regularization) yields a divergence, while in the case of T 2
m2
or j ln T 2 =2 j
1, the one-loop corrections are much bigger than the tree-level
mass. We have also seen that in fact, the series of 1PR corrections to the propagator
exhibits divergence on the mass shell; see (2.112).
The physical reason for these divergences is that the system interacts with an
infinite (or very large) number of fluctuating modes. The mass shell divergence of
the 1PR diagrams emerges because the particle on the mass shell lives infinitely
long, and so interactions of arbitrarily low intensity are repeated infinitely many
times. In the limit " ! 0, or for ! 1 (cutoff regularization), we include more
and more short-wavelength fluctuations into the system. For T 2
m2 or 2 , the
heat bath provides us the large number of interacting modes.
There are also further examples. At zero temperature, quantum corrections carry
powers of ln m=, and this is also a source of large corrections when the values
of m and are vastly different. The value of usually refers to the scale where
we measure the values of the parameters, while m is the physical mass. If these
are very different, then we must include a large number of fluctuation modes in
comparing the actual physics to the reference scale. A more complicated example
arises when two particles propagate with approximately the same momentum (or in
other words, the corresponding plane waves have nearby wave vectors): this again
gives the possibility for repeated interactions.
These divergences can be physical. For example, the poles of the exact propa-
gator signal the mass shell of a stable particle, the singularities in the derivatives
of the free energy signal phase transitions. The problematic divergences are the
perturbative ones emerging in the calculation of per se finite quantities, which

DOI 10.1007/978-3-319-22620-0_3
50 3 Divergences in Perturbation Theory
become worse and worse as we go to higher order in perturbation theory. These

kinds of divergences are symptoms of the inappropriate organization of perturbation
theory.
3.1 Reorganizations of Perturbation Theory
Let us pause at the concluding sentence of the last paragraph. So far, it has seemed
that perturbation theory is a well-defined and unique procedure, based on the Taylor
expansion of eiSint in (2.94). The open question is, however, how we define Sint and
what the expansion parameter is.
If we want to maintain the as much generality as possible, we should allow the
unperturbed system, which we denote by S0 , to be not necessarily identical to the
quadratic part S.2/ (although in practice, S0 is also taken to be a quadratic expression,
since it is the only case we can solve). We require only two features for S0 : it should
provide a solvable system, and its predictions should be close to the true physical
characterization of the system. If these are fulfilled, then we may hope that the small
deviations of the leading-order predictions from the exact characteristic quantities
can be gradually taken into account by some perturbative series. In that case the
interaction Lagrangian is just Sint D S S0 , and not S S.2/, as was expected in
naive perturbation theory.
This consideration has far-reaching consequences: in fact, the choice of the
unperturbed Lagrangian, and thus the perturbation theory itself, is already an
interpretation of the physical system. We must determine the most important part of
the system, and this can be very dissimilar in various physical situations. The hadron
world and the quark–gluon world are different faces of the strong interactions
relevant in different physical circumstances. For more, see Sect. 3.6.
Once we have fixed the unperturbed Lagrangian, we also should fix the expansion
parameter. It is not necessarily the coupling constant that characterizes
P the nonlinear
part of the action. We can use an arbitrary polynomial of it, like 1 nD1 cn , where
n
we can fix c1 D 1. The coefficients of the polynomial might depend on the scale
of the investigated physical phenomena. We can call this expansion parameter the
renormalized coupling ren . The interaction Lagrangian Sint is then an (infinite)
polynomial in ren .
In fact, we also have the freedom to choose the operators in which we want to
construct the perturbation series. In the simplest case, we can consider the rescaling
of the original operators, which we call wave function renormalization. In principle,
however, we could introduce completely different operators defined in terms of the
original field variables for the perturbation series. Technically, it means that we use
an effective model that is capable of accounting for the true physical observables.
The parameters of this effective model are determined either by direct measurements
or by matching certain observables that are equally easy to calculate in the original
and effective models.
3.1 Reorganizations of Perturbation Theory 51
Finally, we have chosen a splitting of the original action as
S D S0 C Sint .ren /; ren D ren ./: (3.1)
In order to have a well-defined example in mind, we consider the ˚ 4 theory, where

the original (or bare) Lagrangian has the form (2.166). To realize the ideas discussed
above, the simplest, but still rather general, choice is obtained by rescaling the
original fields as ˚ D Z 1=2 ' and choosing S0 quadratic. We write
1 ren 4 ıZ ım2 2 ı 4
L D '.@2 m2ren /' ' C '.@2 /' ' ' ; (3.2)
2 24 2 2 24
where the first term on the right-hand side is identified with S0 . Consistency requires
1 C ıZ D Z; m2ren C ım2 D Zm2 ; ren C ı D Z 2 ; (3.3)
but otherwise, ıZ, ım2 , and ı are arbitrary polynomials in ren ,

1
X 1
X 1
X
ıZ D ıZn nren ; ım2 D ım2n nren ; ı D ın nren ; (3.4)
nD1 nD1 nD2
provided (3.3) is satisfied. In this rewriting of the original expression (2.166), m2ren
and ren are called the renormalized parameters; m2 and are the bare parameters.
The part of the Lagrangian containing the complementary ım2 ; ı; ıZ are called
counterterms, forming the counterterm Lagrangian. The way of choosing the
coefficients of the counterterms is called renormalization scheme. To fix the value
of the renormalized parameters, one chooses an appropriate number of physical
observables: this is called a renormalization prescription. And finally, the whole
procedure is called renormalized perturbation theory.1
All this means that there are several representations for the same bare Lagrangian
from the point of view of perturbation series. All choices of the form of (3.2) that
satisfy (3.3) are permitted. We can thus define an equivalence relation between the
renormalized Lagrangians: two theories are equivalent if they describe the same
bare Lagrangian, or equivalently, the same physics. Application of the equivalence
relation leads to equivalence classes, which may be called equivalence classes of
constant physics (ECCP). If we consider just a one-parameter subclass, we obtain
the line of constant physics (LCP), nomenclature that is often used in the context of
lattice field theory [1].
1
We remark that these notions usually are applied to the procedure of getting rid of the UV
divergences. But as we have seen, the UV divergences are just specific examples the possible diver-
gences of perturbation series, and the complete terminology is applicable to any reorganization of
the perturbation theory.
Although at infinite order, all elements of a given ECCP yield the same
physics, the rates of convergence of these series can be rather different. Since
the renormalized coupling ren is a function of , the bare coupling, starting with
ren D C O.2 /, it follows that the renormalized perturbation series at order
n contains the powers k with k n with the same coefficients, while at higher
order, the coefficients are different. Therefore, any two perturbation series are the
same at the order of the expansion and different at higher orders. Put another way,
any two perturbation series are resummations of each other. We should choose that
representative of the ECCP for which the observable(s) we want to calculate exhibit
the best convergence property. The minimal expectation is the requirement to avoid
the perturbative UV or IR divergences systematically.
3.2 On the Convergence of Perturbation Theory
Based on the different convergence properties of the perturbation theories, we can

find optimal elements of the ECCP showing the fastest convergence. This is the
basic idea behind various popular schemes, such as the renormalization group aided,
resummed, or optimized perturbative techniques.
These techniques, albeit usually not explicitly stated that way, use two generic
features of perturbation series. The first observation is that on the one hand, the
divergences appear at each order in perturbation series, but on the other hand, we
have also the freedom to update the counterterms at each order. This gives the
opportunity to design the counterterms to cancel the divergences. If divergences
come from loop corrections, then at nth order, the necessary counterterms are
proportional to at least nren . The so-defined perturbation theory is then divergence-
free.
In fact, the bare parameters of (3.3) must be in accordance with this requirement;
otherwise, the nonperturbatively defined theory is inconsistent, i.e., there are some
physical quantities that are not finite. Therefore, practically, in perturbation theory
one uses the renormalized perturbation series to properly define the theory.
In this context, we make an additional remark: the divergences are usually
environment-dependent; for example, they depend on the actual momenta we use
in an experiment, the temperature and chemical potential if we are in an equilibrium
system. Then the divergence-free perturbation theory itself should also depend on
the environmental data. The only exception is the elimination of UV divergences,
since the physics at short scales is insensitive to the environmental data set by IR
physics.
Another remark is that the above procedure can be performed for as many
observables, as many free parameters we have in the nth-order counterterms; in
the case of ˚ 4 theory, this number is three, related to the wave function, the mass,
and the self-coupling renormalization. After we have cleared all the divergences in
the appropriately chosen (three) observables, we may hope that it has done the job
3.3 Renormalization of the UV Divergences 53
for all other observables, too. This consistency requirement is not trivial, in fact it
represents, in the strictest sense, the criteria for perturbative renormalizability.
The second generic feature of perturbation series is that we can ensure that a
renormalized perturbation theory P belongs to a given ECCP by choosing the
values of the renormalized parameters. This can be done in two ways. Either we
use the renormalization scheme to directly determine the values of the parameters
from experimental observations. This is sometimes tedious, since the perturbation
theory is chosen by the requirement that it should give the best performance under
special conditions, which usually differ from the circumstances under which the
experiments were done. As a second more practical alternative, therefore one
conveniently chooses another perturbation theory P optimal for computing a
certain set of the experimental observables and determines its renormalized
parameters. Then we should find observables that are reliably calculable in both P
and P . By comparing the calculations, we can determine the relationship between
the desired renormalized parameters of P and those of P . This procedure is called
matching, and usually this is used to give the best estimate for the renormalized
parameters.
3.3 Renormalization of the UV Divergences
Historically, the renormalization program was first developed to treat the UV

divergences that endangered the consistency of the whole perturbation theory. The
observation is that at each order, loop corrections yield divergent integrals even
to physical observables that are clearly finite. Therefore, if we determine the bare
parameters from observables (regularization of the loop integrals is needed for this
step), the bare parameters should also be divergent when we want to remove the
regularization. Unless we have a physical regulator (such as the lattice spacing
in solid-state physics) that ensures that even the bare parameters are small, the
perturbative series behave very badly. For example, the first-order correction to the
mass parameter is proportional to in ˚ 4 theory: if ! 1, the perturbative series
is not convergent at all.
The program of the renormalized perturbation theory, however, helps. We should
not develop any quantity in power series of the divergent bare parameters, but
in powers of the finite (and eventually small) renormalized ones. At each order,
we encounter regularized UV divergences, which, however, are canceled with
systematically chosen counterterms. The part of the counterterms relevant to the
cancellation of the regularized UV divergences is uniquely fixed, but the rest
(sometimes called the “finite part”) can be freely chosen to relate the renormalized
couplings in the simplest possible way to the chosen set of observables.
To see how this works, we take the ˚ 4 theory at the one-loop level.
The coefficients of the first two terms of the effective action were calculated
in (2.169) and (2.173) using the bare action. Using the renormalized form of the
Lagrangian (3.2) and setting T D 0, we obtain

.2/ 2 2 2 ren m2ren 1 m2ren
.p/ D p mren ım1 C E 1 C ln ; (3.5)
32 2 " 42
P
and introduce the factorized form .4/ .pi / D N .4/ .pi /.2/4 ı. pi /

2 3
N .4/ .pi / D ren Cı1 C ren2 C IN0 .p1 C p2 / C IN0 .p1 C p3 / C IN0 .p1 C p4 / ;
32 "
(3.6)
where
Z1
m2ren p2 x.1 x/
IN0 .p/ D E C dx ln : (3.7)
42
0
Here we set ıZ1 D 0 for simplicity (which is a self-consistent assumption).

In principle, .2/ and N .4/ are measurable, so they must be finite. This is made
possible if we choose

ren m2ren 1 32ren 1
ım21 D C C 1 ; ı 1 D C D 1 ; (3.8)
32 2 " 32 2 "
where C1 and D1 are constant terms, finite for " ! 0.
Therefore, .2/ and .4/ read

ren m2ren m2ren
.2/ .p/ D p2 m2ren C1 C E 1 C ln ; (3.9)
32 2 42
2
N .4/ .pi / D ren C ren2 3D1 C IN0 .p1 C p2 / C IN0 .p1 C p3 / C IN0 .p1 C p4 / :
32
(3.10)
These expressions are still not fully determined due to the arbitrary constants C1 and
D1 . More precisely, the way in which they are chosen defines the renormalization
scheme. Popular choices are the following:
• The minimal subtraction (MS) scheme: C1 D D1 D 0.
• The MS scheme: here we optimize for the length of the resulting expression, and
choose C1 D E C 1 C ln 4 and D1 D E C ln 4. Then we have
.2/ ren m2ren m2ren

MS .p/ D p2 m2ren ln 2 ;
32 2
2
NMS .pi / D ren C ren2 INMS .p1 C p2 / C INMS .p1 C p3 / C INMS .p1 C p4 / ;
.4/
32
(3.11)
3.3 Renormalization of the UV Divergences 55
where
Z1
m2ren p2 x.1 x/
INMS .p/ D dx ln : (3.12)
2
0
In the MS and MS schemes, the dimensionless counterterms (ım21 =m2ren ; ı1 ) are
mass-independent.
• The on mass shell (OMS) scheme: we require that .2/ .p2 D m2ren / D 0, which
m2
means that C1 D E C 1 ln 4
ren
2 . Often, it is used with the requirement that
m2
.4/ .pi D 0/ D ren , yielding D1 D E ln 4
ren
2 . Here we have
.2/
OMS .p/ D p2 m2ren ;
2
NOMS .p/ D ren C ren2 INOMS .p1 C p2 / C INOMS .p1 C p3 /
.4/
32

N
C IOMS .p1 C p4 / ; (3.13)
where
Z1
m2ren p2 x.1 x/
INOMS .p/ D dx ln : (3.14)
m2ren
0
These expressions for the 2- and 4-point functions are apparently different,
but they belong to the same ECCP if we determine the renormalized parameters
correctly. For example, we may prescribe the following renormalization conditions
to be satisfied in all schemes:
.2/ .p2 D m2phys / D 0; .4/ .pi D 0/ D phys : (3.15)
This yields in the different schemes for m2ren and ren different expressions in terms
of the “measured” mass and coupling parameters (the indices below explicitly refer
to the corresponding renormalization scheme):
" #
phys m2phys m2phys
m2MS D m2phys E 1 C ln C O.2phys /;
32 2 42
32phys m2phys
MS D phys ln C O.3phys /; (3.16)
32 2 42
and
phys m2phys m2phys

m2MS D m2phys ln C O.2phys /;
32 2 2
32phys m2phys
MS D phys ln C O.3phys /: (3.17)
32 2 2
Moreover,
m2OMS D m2phys ; OMS D phys : (3.18)
After back-substituting these values and using phys as expansion parameter, we find
that the results are equal only up to the next-to-leading order; the higher orders are
different. This leads to the aforementioned consequence: the rate of convergence is
different in the different schemes. The best convergence for these two quantities is
guaranteed in the OMS scheme, and it can be achieved by the choice 2 D m2phys in
MS. In fact, with this special choice, the OMS and the MS schemes are the same at
one loop order. This choice also illustrates how one might get rid of nondivergent
but potentially large logarithms.
3.4 Renormalizability: Consistency of UV Renormalization
In the previous section, we demonstrated at one loop order that in the ˚ 4 model
with appropriate choice of the counterterms, the second and fourth derivatives of the
effective action could be made finite for arbitrary momentum values. The question
still to answer is whether this will be true for every observable and to every order.
In the case of an affirmative answer, we say that the theory is renormalizable.
The proof of renormalizability is not easy, and there are several monographs
dealing with this topic more deeply [2]. Here we just briefly go over the main points
of the arguments and the conclusions.
First we pin down that only loop integrals produce divergences, and so the
divergence structure of 1PR diagrams can be built up from the divergences of 1PI
building blocks. Stating this in another way, the Legendre transformation leading to
W from does not change the UV divergences. So we will consider the divergences
of the 1PI diagrams.
With the method of renormalized perturbation theory, divergences can be made
to vanish of only those n-point functions that have counterterms. In ˚ 4 theory,
therefore, one can arrange only the cancellation of the divergences showing up in 2-
and 4-point functions.
Now consider a contribution to the n-point function, and define its overall
divergence: this is the behavior of the integral corresponding to a certain Feynman
diagram when the momenta on all lines go to infinity at the same rate. The most
3.4 Renormalizability: Consistency of UV Renormalization 57
singular behavior of a diagram comes from the overall divergence. This necessarily
has to be canceled by a counterterm.
The overall divergence of an integral can be determined by simple dimensional
analysis. Let us assume that the renormalized coupling constant has mass dimension
d . For simplicity, we use cutoff regularization. Then consider an observable at Nth
order in perturbation theory, and examine the relative weight of the .N C 1/th-order
contribution to it. Factoring out the value of the Nth-order result shows that the
relative contribution is necessarily linear in , but in a dimensionless combination.
For making it dimensionless, we need #d , where # is some quantity of mass
dimension. Therefore, in general, we must expect the appearance of a contribution
d (its coefficient might be zero in some cases). This shows that if d < 0,
then we get a new divergence of .N C 1/th order. Therefore, above a certain order,
all observables will receive divergent contributions. For d > 0, the divergences
will die out after a certain order. If d D 0, the overall divergence at all orders will
be characterized by the same multiplicative expression.
This leads to the conclusion that for d < 0, we expect overall divergences
in all n-point functions. Then for all of these n-point functions, we would need
counterterms, which means infinitely many terms in the Lagrangian. In this case,
renormalized perturbation theory cannot solve the problem of UV divergences.
These are the perturbatively nonrenormalizable theories.
If d > 0, then just a finite number of diagrams are divergent, and only up to a
certain order of the perturbation theory. Then the full expression of the counterterms
can be determined exactly using perturbation theory. These theories are called
superrenormalizable theories.
If d D 0, as in the ˚ 4 theory in four dimensions, the n-point functions up to a
certain nmax can be divergent, but then they are divergent at each order. These are
the potentially perturbatively renormalizable theories.
The next point that we have to see is that the overall divergent part of the
diagram is a polynomial in the external momenta. To see this, one differentiates
the expression of the diagram’s contribution with respect to one of the external
momenta. There must be at least one internal line where the carried momentum
contains additively the external momentum. Therefore, differentiation with respect
to the external momentum provides a derivative of a propagator, which decreases the
degree of the divergence. After some number of differentiations, the diagram will
be finite: therefore, the overall divergent part is a polynomial in the chosen external
momentum.
This means that an overall divergence can be made to vanish by a local operator
counterterm containing a finite number of derivatives.
A diagram can be divergent also in a way that only a finite subset of lines have
large momenta, with the rest finite.2 The finite subset of lines defines a subdiagram,
which is perturbatively of lower order than the complete diagram. Therefore, if
it is overall divergent, we already have associated a counterterm to it that makes
2
Actually, they need not be finite; only their ratio must vanish in the limit.
it finite. Since also the counterterm is of lower order perturbatively, it always

appears together with the divergent subdiagram in any diagram. Finally, with the
earlier determined counterterms, all the subdivergences vanish, and only an overall
divergence may remain, which then can be canceled by an appropriate counterterm.
This line of reasoning summarizes the main steps of the demonstration that
the renormalized Lagrangian of (3.2) solved with the method of renormalized
perturbation theory is an adequate tool for making all the UV divergences vanish
from the system in the case of renormalizable theories.
3.5 Scale-Dependence and the Callan–Symanzik Equations
In the renormalization process, the introduction of a scale parameter is unavoidable.

It is a direct consequence of the regularization of the logarithmic divergences. Using
a cutoff (or a finite lattice spacing a), the divergence is ln =M (ln.aM/), where
M is some mass parameter of the theory. When we make the cutoff vanish, we still
have ln =M: the parameter must be there, since the log needs a dimensionless
argument. In the case of dimensional regularization, the scale compensates the
change in the dimension of the infinitesimal phase-space volume.
The peculiarity of in the renormalized perturbation theory is that on the one
hand, it is not present in the original bare theory, so it cannot influence the physics;
on the other hand, its value is arbitrary. Since it does not influence the physics,
any choice of yields a theory that is in the same ECCP: therefore, perturbation
theories based on different choices of can be connected by appropriate relations
among the renormalized parameters. This means that if g stands for the renormalized
parameters at scale (in the case of ˚ 4 , these are ren and mren ), and g0 at scale 0 ,
then there exists a relation g0 .g; ; 0 / emerging from the condition that both belong
to the same ECCP. This relation expresses the fact that the renormalized perturbation
theory describes the same physical n-point functions as the bare one.
Taking into account that the field is also rescaled as compared to the bare one,
we can write
.n/ .pi ; g; / D Z n=2 ren

.n/
.pi ; gren ; /; (3.19)
for every n and for arbitrary momentum set fpi g. Clearly, it is enough to consider
a finite number of equations of this kind for the determination of Z and gren .
Renormalizability ensures that with these parameters, all n-point functions will
satisfy (3.19).
An equivalent approach is to compare two renormalized perturbation theories
defined at different scales. A direct consequence of (3.19) is
.n/ .n/
ren .pi ; gren ; / D n=2 ren .pi ; g0ren ; 0 /; (3.20)
3.5 Scale-Dependence and the Callan–Symanzik Equations 59
where D Z=Z 0 is a finite wave function renormalization factor. The result of this
equation is the curves g.2 / and .2 / that characterize the scale-dependence of the
couplings and the wave function renormalization constant for a given ECCP.
One can use this equation in perturbation theory; for example, we can apply this
relation to the four-point function of the ˚ 4 model at one loop order at p D 0 and at
T D 0 in the MS scheme (cf. (3.11)). We use the fact that Z D 1 (there is no infinite
wave function renormalization at one loop order), and therefore,
32ren m2ren 3 0 30 2ren m0 2ren 3

ren C 2
ln 2
C O. / D C 2
ln 0 2 C O.0 ren /; (3.21)
32 ren ren
32
where we have set ren D ren .2 / and 0ren D ren .0 2 /. Up to the desired accuracy,
we obtain
32ren 0 2
0ren D ren C ln 2 C O.3ren /: (3.22)
32 2
We can follow an even cleverer strategy. As we discussed earlier, the matching of

two renormalization schemes should rely on observables that can be computed in
both schemes with high precision. In the example above, the direct matching of the
two schemes is problematic if the scales and 0 differ considerably. But we can
always match theories with scales and C d, and then we can progress to 0
gradually in repeated infinitesimal d steps.
We can implement this strategy by differentiating (3.22) with respect to log 2 ;
since the left-hand side does not depend on 2 , we have
dren .2 / 32ren

2 D C O.3ren / ˇ : (3.23)
@2 32 2
The right-hand side is called the beta function. It characterizes the scale-dependence
of the coupling. We can even solve this differential equation,
0
ren .2 / D ; (3.24)
30 2
1 log
32 2 20
where 0 is the integration constant of the differential equation, ren .20 / D 0 . This
equation describes the running of the renormalized coupling.
This relation is an example of the effect of a partial resummation of perturbation
series (performed this time with the help of the renormalization group equation). On
expanding the denominator into a power series of 0 , one sees that its first two terms
reproduce (3.22). The rest (with n0 ; n > 2) is the result of the so-called leading-log
resummation, corresponding to the infinite nested repetition of the bubble diagram
(see the details in Chap. 5 on the N-component scalar model).
The dependence of the other constants on the scale can be obtained when we
differentiate (3.19) with respect to ln 2 :
d
0 D 2 .n/ .pi ; ; .2 /; m2 .2 //
d2
d h n=2 2 2 .n/ i
D 2 2 Z = ; .2 / R pi ; ; .2 /; m2 .2 /
d
" .n/ .n/ .n/
#
n=2 2 @R @R @R n .n/
D Z C ˇ C ˇm2 C R : (3.25)
@2 @ @m2 2
Here we introduced another beta function characterizing the scale-dependence of

the mass and also the anomalous dimension of the field , :
@m2 @ ln Z
ˇm2 D 2 ; D 2 : (3.26)
@2 @2
The set (3.25) represents the Callan–Symanzik equations [3, 4], or renormalization
group equations (RGE). From the 2-point function (3.11) we can also easily
determine the latter coefficient functions at the one-loop level:
ren m2ren
ˇm2 D ; D 0: (3.27)
32 2
Since 3ˇm2 D m2 ˇ , the combination .m2 /3 = is -independent:

1=3
2 2 .2 /
m . / D m20 : (3.28)
0
When these are known, we can use (3.25) to predict the form of all n-point functions.
We emphasize that the coefficients ˇ ; ˇm2 ; above are accurate only up to one
loop level. Therefore, one cannot continue reliably far away from the weak coupling
regime.
.n/
It is interesting to note that the -independence of R , i.e., the right-hand side
of (3.19), allows one to write a similar equation describing the cutoff dependence
of the unrenormalized 4-point function. This is more appropriate for lattice field
theories where the continuum limit is achieved by tuning the lattice constant a
1 towards zero.
3.5.1 Choice of Scale
Although the scale parameter is arbitrary, the perturbative series built on different
choices of 2 may have different convergence properties. According to the generic
discussion, we should choose the element of the ECCP, in the present case by
choosing the scale, that ensures the fastest convergence.
In practice, this means that we should diminish the effects of the logs in the
perturbative formulas. In case of the 4-point function (3.11), we see that for small
external momentum and finite mass, the best choice is 2 D m2ren (one may also
fine-tune the optimization and apply an appropriate constant); for p D 0, this choice
suppresses all perturbative corrections in the MS scheme. On the other hand, if we
have, for some reason, a vanishing mass, then to avoid the blowup of the perturbative
series, we must choose 2 p2 .
A physically typical situation arises when one measures the parameters of the
model at low momentum and low temperature, and the task is to discover the
behavior of the correlation functions in a physical environment where the modulus
of the momentum is much larger than the mass. If the momentum is small, we should
adapt the scale to the mass, 20 D m2ren . We determine the physical values of ren
and m2ren at this scale. These will be the measured phys and m2phys values. In the case
of large external momentum, another choice is more convenient: 2 D p2 . Then to
a good approximation in that kinematic region, we again obtain .4/ .p/ D ren .p/,
i.e., formally, there is no quantum correction. In order to obtain simple results at
large p2 , we should work with renormalized couplings defined at the scale 2 . To
ensure that the two perturbation theories belong to the same LCP equivalence class,
the two couplings are related to each other by the relation
phys
ren .p/ D ; (3.29)
p2
1 ˇ1 phys ln
m2phys
where ˇ1 is the coefficient of the leading (one-loop) term in ˇ :
3
ˇ1 D : (3.30)
32 2
A generic strategy seems to emerge. We should always choose for the computa-
tion the relevant scale in the system. By doing so, the quantum corrections will be
small, so we can essentially use the classical results, but with 2 -dependent running
values of the couplings.
We should make here a remark concerning the counterterms of the system.
We have mentioned that since the overall divergences are polynomial in the
external momentum, the counterterms depend polynomially on the momenta. The
counterterms depend on ren (cf., for example (3.8)). But we see from (3.29) that ren
actually depends on the momentum logarithmically. One may ask whether there is
a contradiction here.
The answer is that the relations in (3.8) represent counterterms for local operators
irrespective of the choice of the renormalized couplings. We may choose the value
of 2 to be whatever we wish; we may even fix it to the value of the typical
external momentum. But any choice of 2 just defines a specific scheme for the
perturbation series. Therefore, we must not mix momentum dependencies coming
from the optimized choice of the renormalized parameters with the dependency that
originates from the inclusion of nonlocal operators (involving an infinite number of
derivatives). This lesson will be recalled when we use temperature-dependent values
for the renormalized parameters.
3.5.2 Landau Pole, Triviality, and Stability
Let us briefly elaborate further on the running of the renormalized coupling (3.24).
We can rewrite it as
1
.2 / D ; (3.31)
2MS
ˇ1 ln
2
where
1
2MS D 20 e ˇ1 0 : (3.32)
This form is very instructive for drawing physical conclusions. First, let us remark
that instead of a dimensionless initial value 0 at 20 , we have expressed .2 / with
the help of a dimensional parameter MS . This is called dimensional transmutation.
The value of MS specifies the ECCP, so it is a renormalization-group-independent
quantity.
We can also see that .2 / is a monotonically increasing function of 2 . We
have seen that the value of 2 is the typical energy scale of the system. At very high
scale, we resolve the close neighborhood of the (pointlike) source of the ˚-field. If
we move from higher scales (shorter distances) to lower scales (larger distances),
the strength of the interaction decreases: this is the screening effect of the quantum
fluctuations.
If we go with the scale to the UV cutoff , then we reach the source of the ˚ field
as closely as is allowed in the regularized system. There the value of the coupling is
the largest; it is the bare coupling D ren .2 D 2 /. At every lower resolution,
the value of the coupling is smaller.
Obviously, one cannot choose the maximal value of 2 arbitrarily large, because
a pole will be hit at 2 D 2MS with a .2 2MS /1 divergence of the coupling.
This singularity differs substantially from the divergences of the perturbation theory.
It is a physical effect, called the Landau pole. At the Landau pole, the quartic
coupling changes sign. Since this coupling is the coefficient of the highest power
in the potential, with this change the potential becomes unstable.
Assigning any finite value to the coupling at some 2 always leads to a Landau-
pole singularity. The only scalar field theory in which we can choose arbitrary
values of the scale is one in which MS ! 1. This is, however, possible only
by choosing .2 / D 0 for every finite 2 . One arrives at the conclusion that the
˚ 4 -theory is trivial: the only consistent ˚ 4 -theory from which the cutoff can be
fully eliminated, extending in this way its validity to arbitrarily high momentum, is
a free and noninteracting theory.
The perturbative input into the renormalization group equation assumes that
is small. One might hope that its increase will necessitate the inclusion of
further (higher) contributions into its beta function, altering the conclusion based
on the leading-log approximation. However, nonperturbative studies in the four-
dimensional theory have arrived at unchanged conclusions.
The present leading-order perturbative analysis repeats itself without any con-
ceptual change for every field theory that has a beta function starting with a positive
coefficient in front of some power of the coupling. This is the case for quantum
electrodynamics but not quantum chromodynamics (QCD). In quantum electrody-
namics, the charges are screened, which is consistent with the intuition based on
classical arguments. In QCD, however, the screening effect of fermionic degrees
of freedom is counterbalanced by the gluonic contribution, leading eventually to
antiscreening.
3.5.3 Stability and Renormalization
The influence of the quantum fluctuations on the stability of the effective potential
depends critically on the renormalization process. Considerable insight can be
gained by studying the effect of the fermionic fluctuations coupled to a scalar with
Yukawa coupling on the effective potential of the scalar field [5]. Let us consider
the Euclidean model
1
LŒ; ; N D N .x//@ .x/ C .@m .x//2 C U.
D 2 =2/ C h.x/ N .x/ .x/;
2
(3.33)
and
2
U.
/ D m2
C
: (3.34)
2
Integrating over the Fermi field in a constant D v background, one obtains the
following formal expression for the effective potential (one-loop contribution):
Z
1 detŒ C M 2 M2
Ueff .
0 / D U.
0 / ln D U.
0 / 2 ln 1 C 2 ;
2Vd detŒ p p
(3.35)
where M 2 D 2h2
0 and Vd is the d-dimensional quantization volume.
A sharp four-dimensional cutoff is applied, and the integral can be done
analytically:

M 2 2 1 4 2 4 M2
Ueff .
0 / D U.
0 / C M ln 1 C C ln 1 C :
16 2 16 2 M2 2
(3.36)
One obtains the divergent pieces of the fermionic contribution by expanding it for
large 2 =M 2 :
M 2 2 M4 2 M4
2
C 2
ln 2 C C O.M 2 =2 /: (3.37)
8 16 M 32 2
The conventional renormalization consists in canceling the divergent terms by
separating appropriate divergent pieces from the bare parameters as follows:
h2 2 h4 2
m2 D m2R C ; D R ln : (3.38)
4 2 2 2 2
Here in the argument of the logarithm, the squared mass m2 is changed to the square
of an arbitrary renormalization scale . After cancellation of the divergences, one
obtains
R 2 h4 2 2e1=2 h2
0
Ueff .
0 ; / D m2R
0 C
0
ln (3.39)
2 16 2 0 2
for the finite and -independent effective potential. It is obvious that for a suffi-
ciently large value of
0 , the fermionic contribution destabilizes the renormalized
potential.
However, if one keeps an arbitrary finite cutoff , then using D .2 /, but
performing the mass renormalization as before, one writes, without expanding in
powers of M 2 =2 ,
2
2 1 2 4 M M2
Ueff D m2R
0 C
0 C M 4
ln 1 C C ln 1 C :
2 16 2 M2 2 2
(3.40)
3.6 The Wilsonian Concept of Renormalization 65
The effect of the fermionic fluctuations is now positive definite without any sign of
instability. The lesson is that no instability occurs by keeping the cutoff finite though
arbitrarily high: if one uses positive values of .2 /, one cannot cross over to the
unstable regime.
We can also interpret this scenario in a different way: the instability of the
potential reflects the fact that the one-loop correction to the coupling is larger
than its tree-level value. This, in fact, signals a nonconvergent behavior of the
perturbation theory. This situation begs for some resummation; up to the present
order, one can envisage several possibilities. For example, we may assume that the
2
asymptotics of the potential change from
02 to
0 , where is some constant
with mass-squared dimension. The value of can be read off the requirement to
reproduce (3.39) in an expansion in the powers of the small parameter . Using
2

0
02
02 ln
0 =, we obtain
h4 2
D ; D ; (3.41)
8 2 R 2e1=2 h2
where both quantities are finite positive numbers. Then we have
R 2
Ueff .
0 ; / D m2R
0 C
0 ; (3.42)
2
which is a stable potential. The consistency of this resummation can and should be
checked against higher-order perturbative and nonperturbative calculations.
The two solutions of the stability problem are different, and they yield different
potentials, too. However, both lead to stable potentials, and they differ from each
other only at high values of
0 , where the perturbation theory is no longer reliable.
3.6 The Wilsonian Concept of Renormalization
The exploration of the universal scaling features at large distances of continuous

phase transitions led L. Kadanoff to propose coarse graining of statistical models
defined on regular lattices by applying block-spin transformations [6]. Wilson’s
generalization [7] of mapping a system with characteristic upper momentum scale
onto a system of lower maximal momentum k to continuum field theories is easiest
to formulate in terms of Feynman’s path integral:
Z R

ZŒJ D N Œ˘q< d.q/eSE ŒC Jq q
Z
Z R
ˇˇ
S
E ŒC Jq q ˇ
DN Œ˘q<k d.q/ Œ˘k<q< d.q/e ˇ
ˇ
J.q/D0;k<q<
Z Rk
0
Œ˘q<k d.q/eSE ŒC Jq q
k
N : (3.43)
Here SEk denotes the classical action of a Euclidean field theory depending on Fourier
components with maximal wave number k. No restriction is put on the highest power
of the fields allowed to appear in this action. All terms obeying the symmetry of the
model can be present. Information concerning the microscopic action, valid at the
highest momentum , just provide the initial conditions for the gradual integration
over the field components of lower and lower momenta. P
The evolution of a local potential given initially as U D M n
nD1 gn can be

characterized by the mapping gn D gn .fg g/, usually called a renormalization
k k
group (RG) transformation. Pictorially, it is represented as a trajectory in the

coupling space, where each point corresponds to a theory potentially reachable via
the subsequent scale transformations. Important points are the sinks and the sources,
the fixed points where the theory is scale-invariant. If an initial theory is located near
a fixed point, it might be attracted to it when the momentum scale is lowered. Such
fixed points are called IR (infrared) points. If the theory is repelled away from the
fixed point, then such a fixed point is of UV (ultraviolet) type. The direction(s)
in the operator space along which the theory is repelled from a fixed point are
called relevant; those along which it is attracted to the fixed point are classified
as irrelevant. The Lagrangian for which one keeps only the relevant operators can
be understood as an effective theory valid in a neighborhood of the fixed point.
A particularly important fixed point is the origin of the coupling space, which
corresponds to a free massless theory. Perturbation theory deals with the dynamics
of models defined in a neighborhood of this point. The coupling .k2 / of the one-
component scalar theory approaches this point with decreasing k. It moves away
when one increases the maximum momentum scale, but it cannot be extended ad
infinitum, because it reaches the Landau pole. On the other hand, with increasing
momentum, QCD moves toward the origin (and moves away toward the infrared),
which is a feature called asymptotic freedom.
The above brief characterization of the RG transformation is illustrated in
Fig. 3.1, where a conjectured RG structure of quantum gravity is displayed in a two-
Fig. 3.1 Conjectured

renormalization group
0.03
trajectories of quantum
gravity in the
two-dimensional coupling
space consisting of Newton’s 0.02
gravitational constant (g) and UV
g
Einstein’s cosmological
constant () (from [10])
0.01
G IR
0
-0.4 -0.2 0 0.2 0.4
λ
3.6 The Wilsonian Concept of Renormalization 67
dimensional coupling space, where in addition to Newton’s gravitational constant

(g), a scale-dependent cosmological constant () is also included. It is well known
that toward the ultraviolet, g is increasing, which is displayed by the (renormalized)
trajectory starting from the origin and conjecturally ending in an ultraviolet stable
fixed point (signaled by the inscription “UV”). Trajectories starting with nonzero
are attracted to this trajectory. There is a third fixed point in the figure, which is
IR-attractive and corresponds to the dominance of the cosmological constant term
of the gravitational action at the largest distance scales.
The original scheme of the gradual integration, which used a sharp value for the
maximal momentum, has been generalized by several authors. The most popular is
the so-called Wetterich–Morris equation [8, 9], which is the easiest to implement for
finding the k D 0 effective action dressed with the effect of quantum fluctuations
at all wavelengths. A recent detailed pedagogical introduction can be consulted in
[10].
The functional integration is restricted to the momentum range above some scale
k by modifying the propagator of the field .k/:
Z Z
1 1
Sfree Œ D .q/ q2 C m2 .q/ ! .q/ q2 C m2 C Rk .q/ .q/;
2 q 2 q
(3.44)
where the infrared regulator function Rk .q/ obeys the limiting behaviors
lim Rk .q/ > 0; lim Rk .q/ D 0: (3.45)

q2 =k2 !0 k2 =q2 !0
The first feature “freezes” the IR fluctuations by increasing the inertia of the modes
below the actual value of k. There is no change in the ultraviolet, and therefore,
one effectively integrates layer by layer near the moving scale k. The most popular
choice is the so-called linear regulator:
Rk .q/ D .k2 q2 /.k2 q2 /; (3.46)
which for k ! blocks all quantum fluctuations. Different functional choices

of Rk correspond to the freedom of choosing different renormalization schemes in
perturbation theory.
The rate of change of the generating functional of the 1PI Feynman diagrams is
solely due to the IR regulator function and is easily shown to obey the following
equation (t D ln.k=/) [11]:
Z Z
1 ı 2 WŒJ ıWk ŒJ ıWk ŒJ
@t Wk ŒJ D @t Rk .x y/ C : (3.47)
2 x y ıJ.x/ıJ.y/ ıJ.x/ ıJ.y/
By the conventional Legendre transform of Wk , one obtains the functional equation

describing the scale-dependence of the effective action k . The simplest form is
obtained for a slightly modified effective action defined as

Z Z
1
k Œ D Wk ŒJ C J .q/.q/ .q/Rk .q/.q/; (3.48)
q 2 q
which goes over to the conventional effective action when k ! 0. (The index
associated with the source J reminds one that the specific J appearing in this
relation is just the source distribution needed to sustain the field configuration .x/.)
This effective action functional obeys the following equation:

h i1
1 .2/
@t k Œ D Tr .@t Rk / k C Rk ; (3.49)
2
with the second functional derivative of k Œ˚ appearing on the right-hand side. In

integrating this equation from k D down to k D 0, the effective action picks up
the effect of all quantum fluctuations and fully determines the infrared physics of
the model. A similar, but less explicit, functional equation was proposed earlier by
Polchinski [12].
The renormalization procedure in terms of the functional renormalization group
(FRG) is equivalent to including gradually the effect of all quantum fluctuations into
the effective action. Therefore, if one starts at k D with a stable theory, its stability
is maintained toward the infrared. There is no question of “sending” to infinity,
and therefore no Landau singularity will arise. The procedure can be applied to any
theory, even those that are not renormalizable perturbatively near the Gaussian fixed
point. The only valid question is the sensitivity of the infrared physics to the initial
value of the momentum . In other words, one might question whether the value of
the maximal momentum scale belongs to the physical data set of the model.
References
1. I. Montvay, G. Münster, Quantum Fields on a Lattice (Cambridge University Press, Cambridge,

1994)
2. J.C. Collins, Renormalization: An Introduction to Renormalization, the Renormalization
Group and the Operator-Product Expansion (Cambridge University Press, Cambridge, 1984)
3. C.G. Callan, Phys. Rev. D 12, 1541 (1970)
4. K. Symanzik, Commun. Math. Phys. 18, 27 (1970)
5. H. Gies, C. Gneiting, R. Sondenheimer, Phys. Rev. D 89, 045012 (2014)
6. L.P. Kadanoff, Physics (N.Y.) 2, 263 (1966)
7. K.G. Wilson, Rev. Mod. Phys. 55, 583 (1983)
8. C. Wetterich, Phys. Lett. B 352, 90 (1993)
9. T.R. Morris, Int. J. Mod. Phys. A 9, 2411 (1994)
10. S. Nagy, Ann. Phys. 350, 310 (2014)
11. J. Berges, N. Tetradis, C. Wetterich, Phys. Rep. 363, 223 (2002)
12. J. Polchinski, Nucl. Phys. B 231, 269 (1984)
Chapter 4
Optimized Perturbation Theory
In Chap. 3, we discussed how the different schemes belonging to the equivalence

class of the same physics help us to resum the ultraviolet (UV) divergences coming
from loop corrections, thus making it possible to define a consistent perturbation
theory. The aim of this chapter is to present methods for improving the convergence
properties of the perturbation theory. In particular, the influence of new energy
scales appearing when one embeds a field theory into a specific environment
might require critical analysis of the formal perturbation series, since infrared (IR)
divergences might be generated by the new scale.
For speeding up the convergence of perturbation theory of specific observables,
in particular avoiding IR sensitivity of the perturbation series, an adequate method is
a rearrangement of the series, as suggested first by Stevenson [1]. Since then, it has
been exploited abundantly in calculating cross sections, mass spectra, etc., in various
models. The optimization of the renormalization group equations to incorporate
environmental effects was first mentioned in [2].
The chapter begis with a discussion of the origin of IR singularities. It is followed
by the presentation of the general optimization strategy and the question of the
renormalizability of optimized series following the logic outlined in papers by
Chiku and Hatsuda [3, 4]. At this stage, a static rearrangement of the perturbation
series is realized, which means redefining the renormalized mass parameters and
couplings in a momentum-independent manner. It will be introduced here through
the example of the three-flavor linear sigma model of Chan and Haymaker [5]. Next,
the detailed implementation of the strategy will be illustrated with the example
of the one-component 4 -theory. The version used for actual phenomenological
investigations of the three-flavor quark–meson effective theory will be worked
out at the level of the one-loop approximation. In the second half of the chapter,
an enlarged perturbative framework will be introduced for situations in which a
static resummation is not satisfactory, and a momentum-dependent resummation is

DOI 10.1007/978-3-319-22620-0_4
70 4 Optimized Perturbation Theory
needed. This is accomplished by the two-particle-irreducible (2PI) approximation,

which is understood in our interpretation also as a specific example of resummation.
4.1 Infrared Divergences and Their Resummation
IR divergences, or IR sensitivity,1 appear under the influence of new characteristic

energy scales, determined usually by some external (environmental) effect such
as the transferred momentum in a scattering process, the temperature in a finite
temperature medium, or an external field. This implies that IR divergences are
not universal, and therefore, the way they are treated cannot be universal either.
Moreover, IR singularities might reflect a physical phenomenon such as the
diverging correlation length at a second-order phase transition or the on-mass-shell
singularity of a propagator.
There are two approaches to dealing with IR divergences. The first uses a
fixed renormalization scheme such as MS and “discovers” within this scheme the
occurrence of IR sensitive contributions. Then one attempts the identification of the
source of the divergences at each order of perturbation theory, and eventually resums
the sensitive terms. One must proceed carefully, however, in this case, since in order
to maintain UV consistency, the corresponding counterterm diagrams should also
be summed.
In the other approach, one makes use of the freedom of choosing an optimally
selected representative of the perturbation theories belonging to the same equiva-
lence class of constant physics, which is free of IR divergences. In practice, one first
identifies the IR divergence occurring at some level of perturbation theory. Then one
chooses an appropriate counterterm to cancel this IR divergent part, just as one does
in the case of UV divergences. Finally, one should make sure that the perturbation
theory defined with this extra subtraction belongs to the desired equivalence class
by choosing the renormalized parameters appropriately. This can be achieved by
fulfilling the renormalization prescriptions (as in the case of UV divergences) or
comparing the expressions to some reference renormalization scheme or directly
to the bare theory. It is important to emphasize that the countercouplings cancel-
ing the IR-sensitive effects will depend on the environmental parameters, since
the perturbatively computed IR-sensitive terms themselves depend on external
parameters.
Let us discuss the question of consistency of the described second procedure
here in this introduction. One might ask whether by choosing an environmentally
(temperature, external field, external momentum) dependent mass term, we spoil
the consistency of perturbation theory. The criticism can be stated simply that
one loses the universal nature of the UV renormalization by making use of a
1
Often, IR problems appear only as enhanced contributions and not actual divergences. Neverthe-
less, we will call them “IR divergences.”
4.2 The Optimization Strategy and the Renormalizability of the Optimized Series 71
counterterm m2 log , defined with m depending on the temperature. The answer

is that the consistency of the perturbation theory does allow a choice among
the representatives of the equivalence class of constant physics characterized by
temperature (environment) dependence of the renormalized parameters. This does
not mean that the divergences generated in perturbation theory (expressed through
the bare parameters) are temperature-dependent, just as by choosing an exter-
nal momentum (scale-)dependent coupling g.p/ we do not generate momentum-
dependent counterterms. If the counterterms are proportional to the parameters
of the renormalized Lagrangian, then the renormalized perturbation theory works
consistently, irrespective of the chosen value of the renormalized parameters. The
perturbation theory is truly inconsistent if terms like 1" log p and 1" T are produced
directly (independent of the renormalized couplings), but this is not the case for the
renormalizable theories.
4.2 The Optimization Strategy and the Renormalizability

of the Optimized Series
The general strategy of the optimized perturbation theory (OPT) will be described
by means of the phenomenologically relevant example of the three-flavor linear
sigma model (cf. Eq. (1.24)):
1 2
L.M/ D tr.@ M @ M 2 M M/ C f1 tr.M M/ C f2 tr.M M/2
2

C g det.M/ C det.M / C x x C y y C Lct .2 ; f1 ; f2 ; g/; (4.1)
where the 3 3 matrix M is defined with the Gell-Mann matrices i ; i D 1; : : : ; 8

completed with a ninth generator normalized in the same way: tra b D 2ı ab . The
ninth matrix is denoted by 0 . Note the “wrong” sign of the mass parameter 2 ,
which implies already at tree level the existence of a symmetry-breaking ground
state. It is physically more transparent to perform a rotation in the “0-8”-plane and
work with the nonstrange component x C ix and the strange component y C iy
instead of 8 C i8 and 0 C i0 . In terms of its components, the matrix M is
expressed as
7
1 X 1 p
MDp .i C ii /i C p diag.x C ix ; x C ix ; 2.y C iy //: (4.2)
2 iD1 2
The Lagrangian density depends on 2 ; f1 ; f2 ; g, which are the renormalized

coupling parameters, and 1 ; 2 , which characterize the strength of the
explicit symmetry-breaking. The counterterm Lagrangian Lct depends on the
renormalized couplings and ensures the order-by-order finiteness of the perturbative

computations. p
The N-point functions on a scalar background M0 D diag.x; x; 2y/ can be
derived from the effective quantum action defined symbolically as
Z n 1Z h
ŒM0 D ı ln DM exp d4 x L.M C M0 ; 2 ; f1 ; f2 ; g/
ı
io
C x .x C x/ C y .y C y/ : (4.3)
The parameter ı is introduced for bookkeeping the number of loops in the

perturbative expansion. Therefore, it is considered a small parameter in the course
of the loop expansion. On completing the computation at some finite m-loop level,
one sets ı D 1.
The general strategy of optimization begins with the splitting of the original
renormalized couplings in an arbitrary way,
2 D m2 m2 ; fi D i i .i D 1; 2/; g D ; (4.4)
and we assume that the difference between the original and the arbitrary new
parameters is of higher loop order:
i ; ; m2 ı: (4.5)
This means that the contributions proportional to these quantities appear at higher
loop order than the contributions involving only m2 ; i ; .
Since the splitting is arbitrary, no observable can depend on any of the newly
introduced parameters in the exact solution of the theory:
@O @O @O
D D D 0: (4.6)
@m2 @i @
At any finite order, however, there might still appear a nontrivial dependence on
these parameters. Then the preceding equations are understood as nontrivial opti-
mization conditions. Actually, one is free to optimize the perturbative computation
for each coupling with a different observable (O ) and at a different level (N ) of the
loop expansion:
.N / .N /
.N /
@Om m @Oi i @O
D D D 0: (4.7)
@m2 @i @
These conditions represent the principle of minimal sensitivity (PMS), which (if
a solution exists) self-consistently determine the newly introduced parameters.
Another version is the requirement not to produce any further correction at some
4.2 The Optimization Strategy and the Renormalizability of the Optimized Series 73
order n of the perturbation series (the principle of fastest apparent convergence

(FAC)):
Om;nm D Oi ;ni D O;n D 0: (4.8)
In practical applications, one has at one’s disposal only a limited perturbation

series (mostly only the first quantum corrections to the tree-level expressions of the
different observables). Below, the FAC conditions will be worked out for the 3-flavor
linear sigma model. Before we proceed to it, a simple formal argument will be given
for the renormalizability of the OPT valid for any model that is assumed to possess
renormalizable series for the observables in terms of the conventional perturbation
theory.
Let us consider the renormalized effective action determined in the conventional
perturbation theory at coupling order n. It is the sum of contributions from several
m-loop diagrams, up to the maximal n-loop contribution:
X
n
.m/
R;n D R : (4.9)
mD0
The renormalized m-loop contribution is the sum of a (divergent) term .m/ not
containing any contribution from the counter-Lagrangian and a compensating term
C.m/ that makes it finite. Both of them can be expanded in terms of the renormalized
couplings:
1
X t1 CtX
2 Ct3 Dn
.m/ .m/ .m/
R D .2 /l f1t1 f2t2 gt3 l;ti C Cl;ti : (4.10)
lD0 ti
This equality expresses the renormalizability of the model in the framework of the
conventional perturbation theory by stating that the sum of the two terms in the
last bracket on the right-hand side is finite. The presence of infinite powers in the
expansion in 2 is easy to accept when we take into account the way 2 appears in
the propagators of various fields.
Now we substitute (4.4) and rearrange the sum (4.9), taking into account that each
extra factor of m2 ; i ; increases the loop order. One can write, with the help
of the binomial coefficients Cqt , the following expression for the s-loop contribution
to the nth-order effective action:
1 t1 CtX
X 2 Ct3 Dn X
ti X
l
.s/ t q1 t2 q2 t3 q3
R D Cqt11 Cqt22 Cqt33 Cpl .m2 /lp 11 2
lD0 ti qi D0 pD0

.s/ .s/
.m2 /p .1 /q1 .2 /q2 .ı /q3 l;ti C Cl;ti : (4.11)
If we count powers of ı, this expression contains terms of s; s C 1; s C 2; : : :

loop order. If s < l, a subsum will contribute to the .l < n/-loop contribution of
.s/
R;n . One has to choose from each R , s D 0; 1; : : : ; n, those terms for which
the equality s C p C q1 C q2 C q3 D l is satisfied. Denoting the sum of these
.s/
contributions by R .ı l /, one obtains the contributions to R;ın , which represents
the series rearranged in powers of ı and complete up to power ı n :
X
n X
n
.s/
R;ın D R .ı l /: (4.12)
sD0 lDs
Since the original expression was finite, this remains true also after the series has
been rearranged, which proves the renormalizability of the optimized perturbation
series.
If we think about all of this less formally, this statement is intuitively obvious,
since the original procedure of the renormalization is completed before the opti-
mization conditions are imposed. These conditions do not change by reshuffling the
series and imposing the optimization conditions. Of course, there is no guarantee
that one will find a physically reasonable solution for the corresponding self-
consistent equations.
It should be emphasized also that if one has several renormalized parameters,
which are usually fitted to some phenomenological information, then the number
of observables for which optimization conditions are prescribed could be higher
than the exact number of the couplings involved in the optimization (but less
than the number of renormalized parameters to be fitted). Once the values of the
couplings are determined with the help of this modified set of conditions, the
phenomenological (“measured”) values of the observables replaced in the fit by the
extra optimization conditions can now be confronted with the “predictions” based
on values found with the help of a larger number of optimization conditions. This
attitude was taken in the phenomenologically motivated investigations of the three-
flavor linear sigma model [6].
The above cumbersome expression for the terms of the optimized series becomes
much more transparent if one applies only a partial procedure, for instance if only
the mass parameter is optimized. Then for the rest of the couplings, one computes
the expansion in terms of the original renormalized quantities. The introduction of
an improved (optimized) mass parameter is almost unavoidable when models with
spontaneous symmetry-breaking are investigated. In these models, the renormalized
squared mass becomes negative, and in the perturbative treatment of the quantum
fluctuations, tachyonic modes will show up, presenting an obstacle to the evaluation
of some loop integrals. By optimizing the mass parameter, one can put aside this
difficulty. The example of the 4 -theory serves as a detailed illustration for the mass
optimization; it will be discussed in the next section.
4.3 OPT for the 4 -Theory 75
4.3 OPT for the 4 -Theory
The OPT strategy will be illustrated using the example of mass-parameter optimiza-
tion [7]. The reorganization of the perturbation series is based on the following
rearrangement of the Lagrangian density:
1 C ıZ m2 C ım2 2 C ı 4
LŒB ; m2B ; B D .@m B /2 C B C B
2 2 24
1 C ıZ m2 C M 2 2 C ı 4
D .@m B /2 C B C B
2 2 24

M 2 C ım2 C ıM 2 2 ıM 2 2
C B C : (4.13)
2 2 B
In the spirit of the discussion in the previous section, one assumes that the
countercouplings ım2 ; ı are known as Taylor series in the renormalized couplings
(m2 ; ) to all orders from the condition of canceling the divergences showing up in
subsequent orders of the conventional perturbation theory. In an OPT computation,
the mass parameter M 2 D m2 C M 2 is used in the perturbative propagators.
The effective counterterm is written in the last line of the equation. In the first
set of brackets, in addition to the combination M 2 C ım2 , a third term, also
ıM 2 , appears. It will be determined by requiring the finiteness of the one-loop
contribution calculated with optimized mass parameter. The last term has to be taken
into account only at the two-loop level.
The one-loop expression of the self-energy reads as

˙1loop D M 2 C T .M 2 / M 2 C ım21 C ıM12 ; (4.14)
2
where the lower index “1” refers to the loop level of the actual calculation.
The expressions of the T D 0 tadpole integral and of the conventional 1-loop
countermass calculated with the help of an -expansion in dimension d D 4
are the following (see Eq. (B.6) in the appendix):

2M2 1 M2
.tadpole/ D T .M / D C E 1 C ln ;
16 2 42

1
ım21 D m2 C E 1 ln.4/ : (4.15)
32 2
Clearly, the divergences of these two terms would cancel each other if there were
no mass optimization. Now it is the “duty” of the intermediate 1-loop counterterm
ıM12 to eliminate the mismatch:

1
ıM12 D M 2
C E 1 ln.4/ : (4.16)
32 2
If M 2 (note that its value is not fixed, since no optimization requirement yet
applies) depended on the environment (it would be proportional to T 2 in case of
the thermal mass), then the result after its subtraction would change arbitrarily
with changing environment (temperature), which would endanger the comparability
of results obtained, for instance, at different temperatures. This danger can be
eliminated if the renormalized couplings and also the counterterms are related to
quantities defined in a unique environment-independent renormalization scheme
(more explicit discussion follows below).
Now the question is whether this conveniently chosen counterterm (the last
term of the rearranged Lagrangian (4.13)) cancels in the next loop order. The
Feynman diagrams contributing at two-loop order appear in Fig. 4.1. Denoting
the contributions by names with direct reference to the corresponding Feynman
diagrams, one can write
2 2
˙2loop D .double scoop/ C .setsun/
4 6
ı1
C .mass insertion/ C .tadpole/
2 2
ıZ2 p2 C ım22 ıM12 C ıM22 : (4.17)
In the first line, one writes the unrenormalized two-loop contributions. In the second
line, the contribution arising from inserting the effective one-loop mass counterterm
on a tadpole appears together with the one-loop self-coupling counterterm con-
tribution. In the third line, the first two terms are known from the conventional
perturbation series. The third one was determined at one loop of the perturbation
theory with shifted mass, and now has to be compensated. The last term is again
determined by the requirement of canceling fully any remaining mass divergence at
the two-loop level.
Fig. 4.1 Feynman diagrams

contributing to the self-energy
at the two-loop level, called
“doublescoop,” “setsun,” and
“mass-insertion,” respectively
4.3 OPT for the 4 -Theory 77
The contributions from the new diagrams are calculated with the dimensional
regularization in d D 4 " and have the following expressions:

.mass insertion/ D M 2 C ım21 C M12

1 1 M2
C E C ln C I F
16 2 42 massins

3M 2 1 1 3 M2
.sunset/ D E C ln
.4/4 2 2 2 42
p2 1 3 1
4
2
T .M 2 / C Isunset
F
;
4.4/ 16

M2 1 1 M2
.double scoop/ D 2E 1 C 2 ln
.4/4 2 42
1 1 F
2
T .M/ C M 2 Imassins
F
C Idoublescoop
F
: (4.18)
16
Only the divergent parts are needed in explicit form; the finite parts are denoted
F
below by Imassins ; Isetsun
F
; Idoublescoop
F
, respectively. The conventional two-loop level
counterterm contributions to the countercouplings are the following:

2 1 1
ım22 Dm 2
.1 ln.4// C finite
2.4/4 2

32 1 1 1
ı1 D C E ln.4/ ; ıZ2 D : (4.19)
32 2 4.4/4
Substitution into (4.17) leads to the following remainder in the divergent contribu-
tions:

2 1 1
˙2;div D M 2 .1 ln.4// C ıM22 : (4.20)
2.4/4 2
The cancellation of this remainder leads to the determination of ıM22 :

2 1 1
ıM22 D M 2 .1 ln.4// C finite: (4.21)
2.4/4 2
Comparing with the expression of ım22 , one recognizes that it is the same expression
except that the substitution m2 ! M 2 was made (the arbitrary finite piece is
conveniently added to ensure the coincidence). This remark was first made in [8].
The results of the first two loop orders is generalized with the following
conjecture: The series expansion of the complementary counterterm ıM 2 coincides
with the series of ım2 D zm m2 computed in the conventional perturbation series.

The only change is the replacement m2 ! M 2 , resulting in ıM 2 D zm M 2 .
The conjecture can be verified based on the following arguments. At some loop
order l, any diagram D.M 2 / whose subdivergences were already renormalized will
contribute an overall divergence proportional to M 2 whose coefficient is @M2 Ddiv .
This quantity itself does not depend on M 2 ; it is the same divergence that would
be produced if the diagram were computed with the original mass parameter m2 .
.l/
These divergences add together into zm M 2 , the negative of the original mass
renormalization factor. In the modified perturbation theory, at this loop order one
.l/
has by the above conjecture the shifted mass counterterm: ım2l C ıMl2 D zm M 2 ,
canceling the sum of the overall divergences.
Another type of divergence appears from the diagrams of the previous pertur-
bative order, on one of the internal lines of which a new mass insertion is applied.
If the coupling M 2 is put on any internal propagator D, it introduces into the
contribution of the corresponding diagram the replacement D ! M 2 D2 . But
this is just the result of the action of @M2 on the diagram multiplied this time by
.l/
M 2 . The sum equals M 2 zm , and this is what will be canceled at l-loop order
by the supplementary counterterm ıMl2 .
No optimization condition was imposed at this stage in order to make it clear
that the order-by-order renormalizability of the theory in terms of the rearranged
perturbation series is a feature independent of this requirement. Therefore, the
feature of renormalizability is valid for any choice of M 2 . Now one can return
to the 1-loop term of the renormalized self-energy computed with shifted mass term
and require that it remain at its tree-level value:
F 2
˙1loop D M 2 C T .M / M 2 D M 2 : (4.22)
2
This FAC condition produces the gap equation determining the optimal M 2 .
4.4 Optimization and Renormalization Schemes
After this carefully detailed analysis, it is instructive to see in a most direct way an
algorithm for designing optimized estimates at one-loop order for both the 2-point
and the 4-point couplings: m2 ; .
The one-loop expression of the effective potential computed with M 2 D m2 C
v 2 =2 (the squared mass on the background v) is the following:
1 2 2 1 ı 4
VD m v C v 4 C V1loop .M 2 / C ım2 v 2 C v ; (4.23)
2 24 2 24
4.4 Optimization and Renormalization Schemes 79
where V1loop comes from (2.176) at zero and from (2.177) at high temperature.
It gives for the second and fourth derivatives the following expressions:
d2 V ı 2
D M 2 C ım2 C 0
v C V1loop .M 2 / C 2 v 2 V1loop
00
.M 2 /;
dv 2 2
d4 V
D C ı C 32 V1loop
00
.M 2 / C O.3 v 2 /: (4.24)
dv 4
The FAC condition for requires the O.2 / quantum correction to vanish (the
correction to the tree-level value is pulled down to O.3 //,
d4 V
D C O.3 v 2 / $ ı D 32 V 00 .M 2 /; (4.25)
dv 4
which is used in turn in the optimization condition of M 2 :
d2 V 1
D M2 $ ım2 D V1loop
0
.M 2 / C 2 v 2 V1loop
00
.M 2 /: (4.26)
dv 2 2
What was done here was to choose specific finite parts of the counterterms; i.e.,
we have chosen the renormalization scheme. Since both countercouplings depend
on the background, we have a specific optimized scheme for each background value.
We have still to ensure that all these schemes belong to the same equivalence class
of constant physics. Here we solve this problem by comparison to a common “refer-
ence” renormalization scheme, for instance the MS scheme, where the renormalized
2- and 4-point couplings are obtained by simply subtracting the divergent parts of
the respectively contributing quantum expressions:
ˇ
ıMS D 32 V1loop
00
.M 2 /ˇdiv ;
ˇ 1 ˇ
ım2MS D V1loop
0
.M 2 /ˇdiv C 2 v 2 V1loop
00
.M 2 /ˇdiv : (4.27)
2
Since the bare couplings are the same, .m2 C ım2 D m2MS C ım2MS /, and the same
kind of invariance is valid also for B , one obtains
1
MS D 32 VMS
00
.M 2 /; m2 m2MS D VMS
0
.M 2 / 2 v 2 VMS
00
.M 2 /;
2
(4.28)
where VMS D V1loop V1loop jdiv .

We see that the renormalized parameters of the optimized scheme belonging to
the same equivalence class of constant physics depend on the background, and they
contain all powers of couplings defined in the reference MS scheme. Therefore,
from the point of view of the reference scheme, the result is nonperturbative. It
can be obtained only after the resummation of an infinite set of Feynman diagrams.
For this reason, one can baptize these particular renormalization schemes RS, or
resummation schemes.
So far, we have done “full optimization,” i.e., we required that no corrections
whatsoever come to certain quantities. With a somewhat more moderate attitude,
one can perform only a partial optimization. A typical example for that is to take
into account only the leading temperature effect in the one-loop mass correction
(see Chap. 6). This leads to the following definition of the optimized mass:
2
M 2 D m2MS C T ; (4.29)
24
while the coupling is the same as in MS. This is the thermal mass approximation,
and it nicely catches several important physical effects, such as the continuous
restoration of symmetry with increasing temperature from a low-temperature
spontaneous symmetry-breaking ground state.
In fact, if m2MS < 0, then the classical T D 0 effective potential exhibits a
minimum at a finite value of v:
6m2MS
V 0 .v0 / D 0 ) v0 D : (4.30)

With the mass optimization of (4.29), m2MS is replaced by M 2 , and we obtain that at
24m2MS
Tc2 D ; (4.31)

the optimized mass is zero, signaling a second-order phase transition. For T > Tc ,
the value of the squared mass turns into the positive range, and one recovers the
symmetric phase.
Let us remark that one can arrive at the thermal mass using the MS scheme
without optimization, directly resumming an infinite series of diagrams. It turns
out (see Chap. 6) that the relevant diagrams are the daisy diagrams, where each
propagator line is decorated by a sequence of tadpole diagrams following each other
like petals of a daisy and the leading-temperature contribution from each tadpole
corrects multiplicatively the expression of the propagator.
The compensation of the T D 0 wrong sign mass by a perturbatively first-
order mass correction T 2 is of dubious validity. The implementation of the
full optimization relation (4.28) might ensure the optimal convergence of the
perturbation series and at the same time also revise the false conclusion of first-order
symmetry-restoring transition drawn from the daisy resummation (see (6.41)). The
application of this step is equivalent to the so-called super-daisy resummation.
Let us return therefore to (4.28). It is easier to solve these equations using the
high-temperature expansion of the integrals contributing to the derivatives of the
potential. Moreover, we substitute on the right-hand side of the first () equation by
4.5 Optimized Perturbation Theory for the SU.3/L SU.3/R Symmetric. . . 81
MS C O.3 /, and omit the corrections. Then we find at high temperatures,
MS p
D ; m2 D U 2 C W 2 W; (4.32)
3 c2 T 2 T
1 MS2 log C 2 C MS
32 4 16M
where
" #
1 2 MS T 2 2MS v 2 MS T
UD mMS C C ; WD ; (4.33)
24 64 2 16
c2 T 2
and D 1 32 MS C
2 log 42 . Here cC is a constant defined in (B.24). For the scale-
dependence of these equations, we can recover the solution of the renormalization

group equations, cf. (3.24), but the complete expressions have additional terms. Near
Tc given in (4.31) and at v D 0, the mass will satisfy M 2 D m2 0, and then
other types of behavior will be dominant. In this regime, we obtain the following
(tri)critical behavior:
2 16m
mD .T Tc /; D : (4.34)
3 Tc
This means that the second and fourth derivatives change sign simultaneously. We
have to admit that this is still not the correct Wilson–Fisher critical point of the
one-component real field. in this approximation, we rather obtain the O.N/ critical
exponents valid for N ! 1.
Several variants of the OPT have been proposed and used in computing the
thermodynamic characterization of various model field theories (for a recent redis-
covery, see [9]). In the next section, the lowest-order mass parameter optimization
is worked out in a model that was successfully applied to the study of the phase
structure of strongly interacting matter.
4.5 Optimized Perturbation Theory for the SU.3/L SU.3/R

Symmetric Meson Model
The propagators of the scalar and pseudoscalar mesons are computed at the one-
loop level for the model (4.1) [6, 10]. Only the squared mass parameter 2 is
replaced by m2 , to be found from the FAC condition imposed on the pion and
kaon propagators. Additional equations will be invoked to determine all seven free
parameters of the model. The mass of the heavier pseudoscalar mesons in the -
sector and all scalar mesons are predicted once these parameters are fixed.
Fig. 4.2 Diagrams contributing to the self-energies of the ; K; fields at one-loop. The fields
propagating on the internal lines have tree-level masses. The last diagram in each line corresponds
to taking into account the “optimizing counterterm” m2 (from [6])
The propagators of the pion and the kaon are parameterized as
D .p/1 D Z .p2 C m2 C ˙F .p2 ; mi ; //;

DK .p/1 D ZK .p2 C m2K C ˙KF .p2 ; mi ; //; (4.35)
where m2 and m2K are the tree-level masses computed with the optimally chosen
mass parameter m2 :
m2 D m2 C 2.2f1 C f2 /x2 C 4f1 y2 C 2gy;

p
m2K D m2 C 2.2f1 C f2 /.x2 C y2 / C 2f2 y2 2x.2f2 y g/: (4.36)
The self-energies contain the loop contributions depicted in Fig. 4.2 ( is the
renormalization momentum scale, mi is the set of masses of the fields i ; i , which
couple via the three-point vertex to either or K in the bubble integral). The self-
energies are renormalized by including appropriate tree-level contributions from the
counterterm Lagrangian:
1 2
Lct .M/ D ı2 tr.M M/ C ıf1 tr.M M/ C ıf2 tr.M M/2
2

C ıg det.M/ C det.M / ; (4.37)
with

1 2 g2 2
ı2 D 2
.5f1 C 3f2 / 2 C m2 ln 2 2
ln 2 ;
2 2
ıf1 D 21 2 .13f12 C 12f1 f2 C 3f22 / ln 2 ; ıf2 D 12 .3f1 f2 C 3f22 / ln 2 ;
1 2
ıg D 2 2
3g.f2 f1 / ln 2 : (4.38)
The replacement 2 ! m2 is reflected only in the divergence of the tadpole. This
counter-Lagrangian cancels the divergences of the tadpoles and the bubble integrals.
The FAC conditions reduce the form (4.35) of the pion and kaon propagators to
their optimized tree-level form:
1 1
D .p/ D ; DK .p/ D ; (4.39)
p2 C m2 p2 C m2K
which is ensured partly by an appropriate choice of Z and ZK . Both the pion and
the kaon self-energies are expanded up to terms linear in p2 C m2pole around the
respective squared pole masses:
@˙F ˇˇ
˙F .p2 ; mi ; / ˙F .m2 ; mi ; / C ˇ .p2 C m2 /;
@p2 m2
@˙KF ˇˇ
˙KF .p2 ; mi ; / ˙KF .m2K ; mi ; / C ˇ .p2 C m2K /: (4.40)
@p2 m2K
Choosing the finite wave function renormalisation coefficients as
@˙F ˇˇ @˙KF ˇˇ
Z1 D 1 C ˇ ; ZK1 D 1 C ˇ : (4.41)
@p2 m2 @p2 m2K
The expressions of the inverse propagators corresponding to Fig. 4.2 change to
D .p/1 D p2 C m2 C Z ˙F .m2 ; mi ; /;

DK .p/1 D p2 C m2K C ZK ˙KF .m2K ; mi ; /: (4.42)
The second step in imposing the optimization conditions is to require the vanishing
of the one-loop contribution to the masses of the pion and the kaon, which leads to
the following two equations:
0 D Z ˙F .m2 ; mi ; / D ZK ˙KF .m2K ; mi ; /: (4.43)
These equations are written more explicitly by separating the true one-loop contri-
butions and the result of the optimizing mass shift:
Z ˙F .m2 ; mi ; / D Z ˙.1loop/F .m2 ; mi ; / m2 2 D 0;

.1loop/F
ZK ˙KF .m2K ; mi ; / D ZK ˙K .m2K ; mi ; / m2 2 D 0: (4.44)
The optimization conditions take explicitly the form of self-consistent equations for
the pion/kaon masses when one uses the tree-level mass expressions of the pion
and the kaon to express m2 and substitute the respective expressions into the FAC
condition (4.44):
p
m2K D 2 C 2.2f1 C f2 /.x2 C y2 / C 2f2 y2 2x.2f2 y g/
.1loop/F
C ZK ˙K .m2K ; mi .mK /; /;
m2 D 2 C 2.2f1 C f2 /x2 C 4f1 y2 C 2gy
C Z ˙.1loop/F .m2 ; mi .m /; /: (4.45)
The masses of all other mesons appearing in the one-loop contributions to the
self-energies are computed at tree level with m2 . (The explicit expression for the
remaining pseudoscalars will be given below; the masses of the scalar excitations
are not listed here.) However, m2 should be expressed everywhere with the help
of either the pion or the kaon mass from (4.36), depending on which of the self-
consistent equations is under consideration.
These one-loop level self-consistent equations display very much the same
structure as the leading-order large-N gap equation of the O.N/-model, which
determines the mass of the Goldstone excitation.
The determination of the coupling parameters 2 ; f1 ; f2 ; g; x ; y ; m2 and the
values of the condensates x; y needs five more equations beyond the four equations
expressing the FAC condition for the pion and the kaon propagators. In the meson
mass spectra, one has reliable information in the pseudoscalar 0 sector. At the
one-loop level, the 2 2 matrix describing the mixing spectra (choosing the x; y
coordinates in place of the 0; 8-components) is the following:
!
p2 C m2 xx C ˙ Fxx .k; mi ; / m2 xy C ˙ Fxy .k; mi ; /
D 1 .k/ D ; (4.46)
m2 xy C ˙ Fxy .k; mi ; / p2 C m2 yy C ˙ Fyy .k; mi ; /
with the tree-level entries
m2 xx D m2 C 2.2f1 C f2 /x2 C 4f1 y2 2gy;

m2 yy D m2 C 4f1 x2 C 4.f1 C f2 /y2 ;
m2 xy D 2gx; (4.47)
and the one-loop self-energy components calculable by the Feynman diagrams of

Fig. 4.2. The roots p2 D M 2 ; M 20 of the determinant of this matrix can be used in
the fit. Actually, one selects only the lighter one of them.
In addition to the pseudoscalar mass spectra, one can exploit the Ward identities
of the model, which relate the condensates to specific combinations of the inverse
propagators at vanishing momentum:

x x
x D Z1 D1
.0/x; y D ZK1 D1
K .0/ p C y Z1 D1
.0/ p : (4.48)
2 2
As an external source of information, one also invokes the PCAC relations, which
relate the strength of explicit symmetry-breaking to the weak decay amplitudes
f ; fK of the pion and the kaon:
p 1=2 1
x D Z1=2 f M2 ; y D 2ZK fK MK2 p Z1=2 f M2 : (4.49)
2
Here the capital letters MK ; M are used for the physical mass values of the pion and
kaon. These values are substituted everywhere in the above equations for m ; mK .
The fitting procedure begins by expressing 2 from the gap equation of the pion
(the second equation of (4.45)) and g from the tree-level mass expression of the
kaon rewritten with the help of the pion mass:
p p
m2K D m2 2 .g 2f2 y/ . 2y x/: (4.50)
These expressions are used for finding the values of f1 ; f2 ; x; y by solving numerically
the set of nonlinear equations consisting of the gap equation of the kaon (the first
equation of (4.45)), the mass formula for the smaller mass in the 0 -sector, and the
combined Ward identities (4.48) and PCAC-relations (4.49) from which x ; y was
eliminated. Finally, the external fields x ; y are extracted from the PCAC relations
after the finite wave-function renormalization factors have been evaluated.
The couplings and also the masses are actually functions of the renormalization
scale . The sensitivity of the renormalized couplings to is natural, since they
do not represent physical observables. The sensitivity to is a consequence of the
finite perturbative order of the calculation. One can plot the dependence of the order
parameters of the chiral symmetry-breaking and of the one-loop estimate for the
heavier eigenvalue in the 0 sector, and choose a range of values where this
dependence is minimal.
It is important to note that the procedure described above does not rely on any
data from the scalar meson sector, contrary to a typical feature of the tree-level
or other perturbative determinations of the meson spectra. In this way, one avoids
a major source of systematic inaccuracy because of the uncertain experimental
situation in the scalar sector. The common experience is that the fully predicted
scalar spectrum is rather insensitive to .
A detailed presentation of the T ¤ 0 phenomenology based on this model can be
found in Chap. 7.
4.6 OPT for the Three-Flavor Quark–Meson Model
The completion of (4.1) by a density accounting for the interaction of the constituent
quarks with the mesons is achieved by adding to it
LQ D qN a .x/.m @m ı ab gF M5ab .x//qb .x/;

7
1 X
M5ab .x/ D p .l .x/ C i5 l .x//ab
l
2 lD1
1 a1 b1
C .ı ı C ı a2 ı b2 /.x .x/ C i5 x .x//
2
p
Cı a3 ı b3 2.y .x/ C i5 y .x// : (4.51)
Here l , l D 1; : : : ; 7 stand for the standard Gell-Mann matrices of the SU.3/ group,
while the diagonal part corresponds to the pure nonstrange–strangepparameterization
arising from redefining the contributions proportional to ab 8 and 2=3ı ab .
With the new coupling gF one has to determine the values of nine physical
parameters and one optimization mass. In order to simplify the presentation, the
finite wave-function renormalization constants are not included in the discussion
that follows [10] and is outlined below. The ten equations one needs include, with
one exception, all those used in the pure meson model in [6].
If we assume a homogeneous condensate in x .x/ D x and y .x/ D y, the
constituent quark masses are generated as
gF gF
Mu D Md D x; Ms D y: (4.52)
2 2
In the parameter fitting, one makes use of the estimates for the constituent masses
arising in the additive quark model:
MN 1
Mu D Md D ; Ms D .M C M˙ / 2Mu : (4.53)
3 2
In this way, one has two relations among three parameters (x; y; gF ). These are used
in the parameter fitting together with the PCAC relation x D f (the equation for fK
can be left out, since the quark sector provides two equations at the expense of only
one new parameter).
The one-loop equations of state are used to determine the strengths of the explicit
symmetry-breaking:
x D x.2 C 2gy C 2.2f1 C f2 /x2 C 4f1 y2 /

X gF
C t˛x i ˛j TbF .ij/ C .Tf F .u/ C Tf F .d//;
˛˛
2
i j
4.6 OPT for the Three-Flavor Quark–Meson Model 87
y D y.2 C 4f1 xy C 4.f1 C f2 /y2 C 4f1 x2 / C gx2

X gF
C t˛y i ˛j TbF .ij/ C p Tf F .s/: (4.54)
˛i ˛j 2
The renormalized bosonic tadpole contribution of the ˛ D ; sector (allowing for

mixing between the mesons of types i and j) is given as TbF .˛i ˛j / with the algebraic
coefficients t˛z i ˛j (of mass dimension) in the z D x; y directions. Similar contributions
describing the quark tadpoles Tf F .q/ D 4Mq TbF .Mq / are now added with simple
algebraic coefficients.
The five additional equations included in the determination of the couplings
and condensates are the formulas for the ; K; masses and the two optimization
constraints. The masses of the kaon and the are the pole masses calculated in
the same way as in the pure meson theory, just including the contribution of the
fermion loops to the self-energy of these fields. The pion mass is given by the tree-
level expression calculated with the optimal mass, and it provides also one FAC
condition:
m2 D m2 C 2.2f1 C f2 /x2 C 4f1 y2 C 2gy; m2 C 2 D ˙ .p D 0; mi .m /; Mu /:

(4.55)
Here instead of its pole mass, the pion mass is estimated by evaluating the self-
energy contribution at vanishing momentum. The reason is that the optimization
equation defined with the pole mass turned out not to have a solution for the
interesting range of the parameters. The self-energy has the previous contribution
from the mesons and also a u-quark contribution, which at p D 0 can be reduced
to an expression proportional to the quark tadpole (expressed here with a bosonic
tadpole integral calculated with the quark mass):
2
˙.f
F
/ .p D 0/ D 2gF Tb .Mu /:
F
(4.56)
Similar equations are obtained for the kaon, which can be used at the pole:
p
m2K D m2 C 2.2f1 C f2 /.x2 C y2 /.x2 C y2 / C 2f2 y2 2x.2f2 y g/;
m2 C 2 D ˙K .p2 D MK2 ; mi .m /; Mu ; Ms /: (4.57)
The fermionic contribution comes from a bubble integral composed with a u and an
s-quark propagator:
h
2 2 2
˙K.f
F
/ .p D m K / D g F Tb .Mu / C Tb .Ms /
F F
i
C .m2K C .Mu Ms /2 /BbF .p2 D m2K ; Mu ; Ms / : (4.58)
The fifth equation is the mass formula for the lighter excitation in the 0 sector,
to be calculated from the mixing mass matrix of (4.46). Here again, the fermionic
contribution to the matrix elements has to be given:

˙ Fxx .f / .p/ D g2F 2TbF .Mu / p2 BbF .p; Mu ; Mu / ; ˙ Fxy .f / D 0;

˙ Fyy .f / .p/ D g2F 2TbF .Ms / p2 BbF .p; Ms ; Ms / : (4.59)
The values of the input parameters for finding the couplings of the model were
chosen as
m ŒMeV D 138; mK ŒMeV D 495:01; m ŒMeV D 947:8;

f ŒMeV D 93; Mu ŒMeV D 313; Ms ŒMeV D 530: (4.60)
The parameterization to be used for phenomenology was found by minimizing

the deviation of the predicted masses for 0 and in the scalar sector (a0 ; f0 ; )
from their observed values. This can be tuned with two free parameters: M0b , the
bosonic, and M0f , the fermionic renormalization mass scales. The best fit obtained
numerically in [10] led to the choice
M0b ŒMeV D 520; M0s ŒMeV D 1210:9: (4.61)
With this input, one “predicts” the following physical data:
m ŒMeV D 614:2; Mf0 ŒMeV D 1210:9; fK ŒMeV D 125:23: (4.62)
This parameterization is used in Chap. 7 for studying the phase diagram of the
quark–meson model at finite temperature and finite baryonic density.
4.7 The 2PI Formalism as Resummation and Its

Renormalization
We have seen earlier, in Sect. 2.8, that the 2PI formalism is appropriate for solving
a field theory by beginning with a prescribed propagator of the theory. The value
of the prescription is demonstrated by (2.132): there, the right-hand side is the
exact propagator calculated in the theory, while the left-hand side is the predefined
propagator. The theory is consistent if the two definitions are the same.
We can approach the 2PI formalism also from another viewpoint, as a resum-
mation of a certain subset of Feynman diagrams. In fact, the 1PI formalism has
already provided a resummation: (2.112) demonstrates that the iterative evolution
of the self-energy is equivalent to a summation of an infinite series of diagrams. In
the case of the 2-point functions, the self-energy-diagrams (the Dyson–Schwinger
4.7 The 2PI Formalism as Resummation and Its Renormalization 89
series) are summed. For higher-order diagrams, the 1PI formalism in general sums
all diagrams with one-particle-reducible skeleton structure.
The algorithm of the 2PI formalism can be interpreted also as a resummation: it
sums all the self-energy diagrams including that of the internal lines. In fact, (2.132)
requires that the propagator representing a line contributing to ˙2PI be already the
exact one. There is no need to include further self-energy corrections. It is exactly
this property that allows one to take into account only two-particle irreducible
diagrams in int Œ˚N ; G
N of (2.130). This aspect will be discussed in Sect. 4.7.1
However, the formal definition of the 2PI resummation is not enough. We should
define a theory that shows good convergence properties against the potential UV
divergences, i.e., it should renormalized. The renormalization issue is far from trivial
in the 2PI case, exactly because a 2PI skeleton in fact resums an infinite number
of diagrams in the original perturbation theory. For consistency, the counterterm
diagrams should also have been included. In the literature, there are thoroughly
detailed accounts on the renormalizability of the 2PI resummation, providing
useful recipes on how to perform practically the resummation [11–19]. The main
conceptual points will be reviewed in Sect. 4.7.2.
4.7.1 2PI Resummation as Optimized Perturbation Theory
A peculiarity of the 2PI resummation is the fact that it represents a momentum-

dependent reorganization of the perturbation series, since the full propagator
should reproduce the exact one, not just a single value taken for a fixed value
of the momentum. We have seen that momentum-independent reorganizations
(optimization) are obtained by appropriately choosing a momentum-independent
finite part of the counterterms, eventually ensuring that we remain in the same
equivalence class of constant physics. Momentum-dependent resummation means
that (the finite part of) the counterterms must be momentum-dependent, and the
new kind of terms (and counterterms) in the 2PI effective action simply reflect this
new kind of optimization.
Let us illustrate this line of thought by the example of the ˚ 4 model. We
generalize the renormalized form of the Lagrangian (3.2) as
1 ren 4 1 ı 4
L D ' K .i@/' ' C 'ıK .i@/' ' ; (4.63)
2 24 2 24
where we allow rather general momentum-dependence in K and ıK .

As we will demonstrate in the next subsection, the 2PI renormalization is possible
because the overall divergences are local independently of the propagators used on
the internal lines, at least within a certain limit (for example, we must not change
the scaling dimension of the propagator in the far UV). This means that (4.63) will
provide us a theory that can be renormalized in exactly the same way as we did for
the choice of the free kernel K0 D @2 m2 . This means that we obtain ıZ, ım2 ,
and ı in the same way as before, but of course, the values of the divergent terms
may be rather different. In particular, at the one-loop level,
Z
2ren d4 p 2
ı D G .p/ ; (4.64)
2 .2/4 div
where G D K 1 . We should remark here that this ı, in accordance with the usual
expectations based on perturbation theory, is just the (overall) divergent part of the
bubble diagram.
In this way, we see that for every choice of K and ıK , provided that this latter
contains the appropriate local terms that cancel the UV divergences, we obtain a
well-defined renormalized perturbation theory.
We have still to ensure that we are in a fixed equivalence class of constant physics.
Since the physics is fully determined by the bare action, it is enough to reproduce
it. In particular, we have
K .p/ C ıK .p/ D Zp2 m2B : (4.65)
Since K is finite, this equation tells us that ıK must have a divergence structure
of ıZp2 ım2 , where ıZ and ım2 are determined by consistency. Therefore, we can
write
K .p/ C ıKfin .p/ D p2 m2ren ; ıKdiv D ıZp2 ım2 : (4.66)
There are still some free parameters, , m2ren , and ren . These must be determined
by the usual renormalization scheme prescriptions, for example requiring that
.2/ .p2 m2ren ; T D 0/ D p2 m2ren and .4/ .pi D 0; T D 0/ D ren .
The above analysis is valid for any (reasonable) choice of the kernel. The
possibility that we can work consistently with any tree-level propagator widens the
potential of the optimization. In particular, in (4.22), the FAC could be achieved by
ensuring that the mass counterterm and the self-energy correction cancel each other.
This is impossible if ˙ is explicitly momentum-dependent, as in (4.44), where we
must choose a specific momentum, e.g., p2 D m2 , and require the self-energy to
vanish at this point. But in a 2PI framework, when we can choose ıK freely, we
can require that the complete self-energy vanish for all momenta:
Š
˙.K ; p/ D ˙diagram .K ; p/ ıK .p/ D 0; (4.67)
which means that ıK D ˙diagram .K ; p/, where ˙diag D ˙div C ˙fin is the self-
energy expression coming from the diagrams. On substituting it into the consistency
equation (4.66), we obtain
K .p/ C ˙fin .K ; p/ D p2 m2ren (4.68)
and ˙div .K ; p/ D ıZp2 ım2 , which is, in fact, just the definition of ıZ and ım2 .
We see that the momentum-dependent optimization condition (4.67) leads to

the 2PI equation (2.132). It is also true that all diagrams that contain self-
energy insertions (i.e., the 2PR diagrams) are canceled by the counterterm diagram
containing ıK . So this optimization process is equivalent to the 2PI resummation.
Then the strategy for a finite-temperature 2PI calculation can be the following:
we start from some initial choice for the kernel and determine the counterterms order
by order at zero temperature, just as in an ordinary perturbation theory, applying
some renormalization prescription: for example, we choose the finite parts of the
subtracted divergences for which (4.68) is satisfied with D 1 and a given mren .
Then with the same counterterms, we determine the finite-temperature self-energy:
this will be finite, since the counterterms do not depend on temperature. Finally, we
use (4.68) to update the kernel until the recursion is closed.
4.7.2 Perturbative Renormalization of the 2PI Equations
A separate discussion of the perturbative renormalization of the 2PI formalism is

necessary, since the coefficients in front of various terms of the 2PI effective action
are not measurable quantities. These terms contain integrals to be performed, and
therefore also counterterms are needed that render the result of the integration finite
when performed with the physical propagators. The 2PI action is in some sense
halfway between the original action and the fully physical 1PI action, in which the
coefficients are measurable n-point functions.
The most important feature that makes renormalization possible at all is that the
overall UV divergences of the 2PI skeleton diagrams, after subtracting subdiver-
gences, are local. The reason is that the only property of the propagator needed for
the demonstration of the relevant theorem (Weinberg’s theorem) within perturbation
theory is that differentiation with respect to its momentum lowers the polynomial
power of its asymptotic behavior. This is in fact true for every smooth enough
function. Then from Weinberg’s theorem, it directly follows that the degree of the
overall divergence decreases when we differentiate with respect to any momentum
external to some diagram; therefore, the overall divergences are all local. And so
after we have subtracted the subdivergences, the remainder can be made finite
with local counterterms. Because of the generality of Weinberg’s theorem, all these
statements remain true when we evaluate a diagram with a generic propagator
instead of the tree-level (perturbative) one.
This means that the counterterm renormalization works also in the case of
the 2PI resummation. These counterterms are, however, necessarily different from
the operators renormalizing the 1PI perturbation theory. After this conceptual
clarification, the remaining problem is to determine the counterterms.
Let us examine in detail why the logic of the determination of counterterms is
different from the conventional perturbative procedure. The basic reason is that the
operators in the 2PI action (2.132) themselves represent an iterative structure, not
just the couplings that determine the respective strengths of the contributions to
˙2PI . Since the same counterterms should appear, the counteroperators are also
summed iteratively. To see the problem more clearly, we discuss here the 2PI
equations at vanishing classical field. The 2PI action reads (cf. (2.136))
1 1
i2PI ŒG D Tr ln G Tr G1
0 G C iint ŒG; (4.69)
2 2
with G0 given by the free theory. Then the corresponding 2PI equations can be
written as
ıint ŒG
G1 D G1
0 ˙2PI ŒG; ˙2PI ŒG D 2i : (4.70)
ıG
The parameters of G0 are the bare parameters: these are responsible for the
cancellation of the overall divergences of the self-energy. This means that the ˙2PI
self-energy must be free of subdivergences, at least when evaluated at the solution
of (4.70). In fact, this condition is enough to determine the counterterm structure of
int [20].
The condition for the cancellation of the subdivergences can be made explicit
with the help of a trick. We exploit that if we change the free propagator by an
amount ıG0 that does not affect the UV behavior, then the overall divergences of
the Feynman diagrams are not affected. But the finite parts of the variation are
multiplied by the possible subdivergences. Therefore, if the ˙2PI determined from
the theory is free from subdivergences, the change due to the variation remains finite.
Therefore, one assumes at the start that a propagator G has been found that
solves (4.70) with a fixed G0 and that the emerging solution is well renormalized.
Now let us change G0 ! G0 CıG0 , where ıG0 is sufficiently (at least exponentially)
suppressed in the UV. Expanding the 2PI equation (4.70) around G, we obtain the
modified solution G C ıG. It is important to note that asymptotically, to leading
order, G p2 and ıG p4 up to logarithmic terms. Now let us examine the
change of ˙2PI :
˙2PI ŒG C ıG D ˙2PI ŒG C ˙ .0/ C ˙ .r/ ; (4.71)
where
Z
1 ı˙2PI ŒG
˙ .0/ D ıG; D2 (4.72)
2 ıG
and ˙ .r/ functionally depends at least quadratically on ıG. By the starting assump-
tions, the modified theory should also be correctly renormalized, which implies that
˙ .0/ and ˙ .r/ are finite.
Here ˙ .r/ is explicitly finite: due to the 2PI nature, every subdiagram contains
at least three legs, which by dimensional analysis asymptotically behaves as p2 .
Therefore, integrating it by two factors of ıG p4 results in a convergent integral.
The critical term is ˙ .0/ , which can contain extra subdivergences, even when
contains only overall divergences. This is because ıG p4 , while p0 (up to
logarithmic
R corrections). Therefore, there is a source of divergences in the integral
ıG itself. To localize the subdivergences, we use the splitting ıG D ıG.0/ C
ıG.r/ , where ıG.0/ is proportional to p4 at large momenta, while ıG.r/ is at most
p6 . The crucial observation here is that ıG.0/ must come from the self-energy ˙ .0/ :
ıG.0/ D G˙ .0/ G: (4.73)
Otherwise, we do not get p4 corrections, when one subtracts eq. (4.70) written
down for G from the same equation written for GCıG. Substituting back into (4.72),
we obtain
Z
1
˙ .0/ D G˙ .0/ G C ıG.r/ : (4.74)
2
This defines a self-similar structure, similar to the Bethe–Salpeter equations. Its

solution reads
Z
.0/ 1
˙ D VıG.r/ ; (4.75)
2
where
Z
1
V D C GVG: (4.76)
2
To prove this statement, we substitute back the desired form:

Z Z Z Z
1 .r/ Š1 1 1
VıG D G˙ .0/ G C ıG.r/ D C GVG ıG.r/ ;
2 2 2 2
(4.77)
where in the last term, we changed the order of the two integrations. Factoring out
ıG.r/ , we see that we indeed have the desired equation.
The main point in (4.75) is that it is finite if V is finite. The reason is that V
remains p0 , but ıG.r/ p6 , so the integral does not produce extra divergences.
Therefore, we have arrived at the main result: the 2PI effective action is free of
subdivergences if V defined in (4.76) is finite.
The above proof in fact is appropriate for proving that it is enough to make V
finite at zero temperature; the self-energy remains free of subdivergences also at
finite temperatures. Namely, the change at finite temperature is only in the free
propagator G0 ! G0 C ıGT , where ıGT is exponentially suppressed at large
momenta. Therefore, what was said above can be repeated exactly to conclude that
the finite temperature corrections remain finite.
Finally, we demonstrate how counterterms can be used to make V finite. We

separate the first nontrivial term from int :
1 1 0
i2PI ŒG D Tr ln G Tr G1
0 GC .Tr G/2 C i2PI
0
ŒG: (4.78)
2 2 8
(Truncating at this term constitutes the Hartree approximation.) Then
0
ı˙2PI ŒG
D 0 C 0 ; 0 D 2 : (4.79)
ıG
Substituting back into the equation of V (4.76), we obtain
Z
1
V D 0 C 0 C .0 C 0 /GVG: (4.80)
2
The above equation can be made finite only if 0 is only overall divergent; this will
0
lead to additional conditions for the couplings in 2PI . The overall divergence of 0
is momentum-independent, and so we can make it finite by defining the divergent
part of the coupling 0 at any specific momentum. Practically, one chooses the
momentum q where the renormalization of the coupling is defined (subtraction
point). There we can demand V.q / D ren , and so we have formally
Z
0 1
ren 0 GVG
0 D Z2 : (4.81)
1
1C 2 GVG
We emphasize here again that 0 is not only responsible for making the 0
contribution finite; it also accounts for the extra divergences caused by the explicit
integrals in the formula.
In particular, if 0 D 0, then V D ren , and we obtain
Z
1 1 1
D C G2 ; (4.82)
0 ren 2
which is the same equation as (4.64).

This line of thought can be generalized to systems in which we take into account
the background field as well. We will not discuss it here; the interested reader will
find details in [19]. The result is that we have another Bethe–Salpeter-like equation
for
@2 ˙ŒG; ˚
Q D ; (4.83)
@˚ 2
References 95
namely
Z
1
VQ D Q C Q
GVG: (4.84)
2
Attention! There is no mistake on the right-hand side: there it is V that appears, not
Q If one can make both V and VQ finite, the theory will be free of subdivergences.
V.
In certain truncations, one has V D V, Q but in a generic truncation, this is not the
case. This is closely related to the fact that in a given truncation,
1 ı 3 2PI ı 2 2PI
¤ : (4.85)
2 ı˚ 2 ıG ıGıG
As a consequence, different counterterms are needed for different diagram topolo-
gies, in particular, the leading terms in 2PI :
0 2
i2PI Œ˚; G D .Tr G/2 C ˚ 2 Tr G C iint
0
Œ˚; G: (4.86)
8 4
Here 0 comes from (4.81), and 2 from an analogous equation with the change
! .Q Thus for a generic truncation, 0 ¤ 2 .
References
1. P.M. Stevenson, Phys. Rev. D 23, 2916 (1981)

2. D. O’Connor, C.R. Stephens, Int. J. Mod. Phys. A 9, 2805 (1994)
3. S. Chiku, T. Hatsuda, Phys. Rev. D 58, 076001 (1998)
4. S. Chiku, Prog. Theor. Phys. 104, 1129 (2000)
5. L.H. Chan, R.W. Haymaker, Phys. Rev. D 7, 402 (1973)
6. T. Herpay, Zs. Szép, Phys. Rev. D 74, 025008 (2006)
7. A. Jakovác, Zs. Szép, Phys. Rev. D 71, 105001 (2005)
8. P. Petreczky, F. Karsch, A. Patkós, Phys. Lett. B 401, 69 (1997)
9. J.-P. Blaizot, N. Wschebor, Phys. Lett. B741, 310 (2014)
10. P. Kovács, Zs. Szép, Phys. Rev. D 75, 025015 (2007)
11. H.v. Hees, J. Knoll, Phys. Rev. D 65, 025010 (2002)
12. H.v. Hees, J. Knoll, Phys. Rev. D 65, 105005 (2002)
13. H.v. Hees and J. Knoll, Phys. Rev. D 66, 025028 (2002)
14. J.P. Blaizot, E. Iancu, U. Reinosa, Phys. Lett. B 568, 160 (2003)
15. J.P. Blaizot, E. Iancu, U. Reinosa, Nucl. Phys. A 736, 149 (2004)
16. F. Cooper, B. Mihaila, J.F. Dawson, Phys. Rev. D 70, 105008 (2004)
17. E. Calzetta, B.L. Hu, Phys. Rev. D 35, 495 (1987)
18. A. Arrizabalaga, J. Smit, Phys. Rev. D 66, 065014 (2002)
19. J. Berges, S. Borsányi, U. Reinosa, J. Serreau, Ann. Phys. 320, 344 (2005)
20. A. Patkós, Zs. Szép, Nucl. Phys. A 811, 329 (2008)
Chapter 5
The Large-N Expansion
In conventional perturbation theory, one computes physical observables in power

series of some coupling strengths, which in effective theories of strong interactions
leads to large higher-order corrections. This situation begs for alternative nonpertur-
bative methods. One of the most efficient propositions is an expansion in the inverse
powers of the number of degrees of freedom representing the dynamical variables.
In connection with the chiral symmetry-breaking, the inverse of the number of light
(nearly massless) quarks and correspondingly the inverse of the number of light
(pseudo-Goldstone) mesons provide examples of such acceptably small parameters.
In the first case, the solution of UL .2/ UR .2/ or UL .3/ UR .3/ symmetric theories
is attempted by departing from the large-n solution of the UL .n/ UR .n/ symmetric
theory. In the second, the solution of an O.N/ symmetric effective Gell-Mann–Lévy
meson theory is continued down to N D 4 with the help of a power series in 1=N.
In this chapter, we shall first give a technically detailed account of constructing
the renormalized leading-order (LO) and next-to-leading-order (NLO) solutions
of a scalar field with N components supplied with O.N/-symmetric dynamics.
Our discussion is largely self-contained, although several aspects, in particular
applications to critical phenomena, are exhaustively covered by recent reviews and
monographs [1–3]. A less comprehensive analysis of the application of the saddle-
point approximation to the solution of UL .n/UR .n/ symmetric scalar field theories
will also be presented. The chapter closes with a review of the large-N behavior of
the quark–meson model, where the O.N/-symmetric meson theory is supplemented
with quark degrees of freedom interacting with meson fields via Yukawa coupling.
The renormalization process described in this chapter will serve as the basis of the
phenomenological applications to be presented in Chap. 7.

DOI 10.1007/978-3-319-22620-0_5
98 5 The Large-N Expansion
5.1 The Dyson–Schwinger Equation of the N-Component

Scalar Theory
The classical Euclidean action of the perturbatively renormalizable O.N/-symmetric

theory of the N-component scalar field j ; j D 1; 2; : : : ; N has the following
expression:
Z
1 2
Scl Œj D dx j .x/. C m2 /j .x/ C j .x/j .x/ : (5.1)
2 24N
We shall investigate the question of the emergence of a symmetry-breaking

minimum that is taken into account in the following parameterization of the j -
field:
p
j .x/ ! . Nv C .x/; ˛ .x//: (5.2)
The scaling of the quartic term in the original action and here in the shifted part
of 1 corresponds to the extensive nature of the effective action, that is, the fact
that the potential energy density in the large-volume (thermodynamic) limit is
proportional to the number of the degrees of freedom of the system. After this shift,
one obtains
Z h1
Scl Œ; ˛ D dx .x/. C m2 /.x/
2
1 1 p
C ˛ .x/. C m2 /˛ .x/ C m2 Nv 2 C 2 Nv
2 2
p 2 i
C Nv 2 C 2 Nv C .x/2 C ˛2 .x/ : (5.3)
24N
The classical field equations of and ˛ are found by taking the appropriate
functional derivatives:

ıScl 2 p 2 v
D C m C v .x/ C Nv m C v C p 3 2 C ˛2
2 2
ı.x/ 2 6 6 N

C 3 C ˇ2 ; (5.4)
6N

ıScl v 2
D C m2 C v 2 ˛ .x/ C p ˛ C ˛ C ˇ2 ˛ :
ı˛ .x/ 2 3 N 6N
(5.5)
The derivative of the quantum action with respect the field variables (the quantum
equations of motion) can be obtained from these two equations by applying the
5.1 The Dyson–Schwinger Equation of the N-Component Scalar Theory 99
construction rule (2.148) for generating the n-point functions of some generic field
theory in the one-particle-irreducible formalism:

ı ıScl ı
D ˚A C G˚A ˚B ; (5.6)
ı˚U ı˚U ı˚B
where ˚U stands for the field variables, and U includes all discrete and continuous
(space-time) indices of the field. The quantity
1 ˇ
ı2 ˇ 1
G˚A ˚B D ˇ .2/ ˚A ˚B (5.7)
ı˚U ı˚V ˚A ˚B
is the ˚A ˚B element of the inverse of the second functional derivative “matrix” of

the effective action. Repeated appearance of an index means summation/integration
over it. The use of this rule for the present model is illustrated by working out the
transmutation of a specific classical product into an expression of the quantum field
theory:

ı ı
.x/˛ .y/ ! .x/ C G j .x; z/ ˛ .y/ C G˛ k .y; w/ 1
ıj .z/ ık .w/

ı
D .x/ C G j .x; z/ ˛ .y/ D .x/˛ .y/ C G˛ .x; y/; (5.8)
ıj .z/
where 1 denotes the identity element of the functional space. An important relation,
repeatedly used below, gives the functional derivative of a propagator:
.2/
ıGi l .x; y/ ıj j .z1 ; z2 /
D Gi j1 .x; z1 / 1 2
Gj2 l .z2 ; y/
ık .z/ ık .z/
D Gi j1 .x; z1 /j1 j2 k .z1 ; z2 ; z/Gj2 l .z2 ; y/; (5.9)
where
ı3
j1 j2 k .z1 ; z2 ; z/ D : (5.10)
ıj1 .z1 /ıj2 .z2 /ık .z/
With these remarks, one obtains the two fundamental quantum equations of the
model (with the simplified notation G˛ j D G˛j ):

ı p
D C m2 C v 2 C Nv m2 C v 2
ı.x/ 2 6
v
C p 3 2 C 3G .x; x/ C ˛2 C G˛˛ .x; x/
6 N
3
C C 3G .x; x/ G k .x; z1 /G j .x; z2 /j l k .z2 ; z3 ; z1 /Gl .z3 ; x/
6N
2
C C 2˛ G˛ .x; x/ C G˛˛ .x; x/
6N ˛

G˛k .x; z1 /G˛j .x; z2 /j l k .z2 ; z3 ; z1 /Gl .z3 ; x/ : (5.11)

ı v
D C m2 C v 2 ˛ C p .˛ C G ˛ .x; x//
ı˛ .x/ 6 3 N
2
C ˛ C ˛ G C 2G˛
6N

G k .x; z1 /G j .x; z2 /j l k .z2 ; z3 ; z1 /Gl ˛ .z3 ; x/

C ˛ ˇ2 C ˛ Gˇˇ .x; x/ C 2ˇ Gˇ˛ .x; x/
6N

Gˇk .x; z1 /Gˇj .x; z2 /j l k .z2 ; z3 ; z1 /Gl ˛ .z3 ; x/ : (5.12)
The equation that determines the vacuum expectation value of the variable
is obtained by equating (5.11) to zero, at the same time setting all fields and
nondiagonal propagator components (G˛ ; ) on the right side equal to zero:

p 2 2 v
0 D Nv m C v C p .3G .x; x/ C .N 1/G .x; x//
6 6 N

G .x; z1 /G .x; z2 / .z2 ; z3 ; z1 /G .z3 ; x/
6N

C G .x; z1 /G .x; z2 /˛ ˛ .z2 ; z3 ; z1 /G .z3 ; x/ : (5.13)
Here further simplified notation is introduced for the diagonal propagator elements,
G˛ˇ D G ı ˛ˇ ; G D G , and the discrete summation is performed using ı ˛˛ D
N 1.
The Dyson–Schwinger equations for higher n-point functions of the theory are
obtained by taking appropriate further functional derivatives of (5.11) and (5.12).
The final result of a lengthy but straightforward sequence of calculational steps also
involves the fourth functional derivatives of . For these higher n-point functions,
one assumes the tensorial structure dictated by the corresponding functional deriva-
tives of the classical action Scl .
In particular, for the two-point functions, it will be shown that the leading order
(LO) of the 1=N hierarchy self-consistently required for the 3- and 4-point functions
naturally truncates the infinite set of coupled n-point equations within the subset
of the two-point functions and one three-point function. For this, one needs the
large-N scaling of the related 3- and 4-point functions, which can be naturally
5.1 The Dyson–Schwinger Equation of the N-Component Scalar Theory 101
assumed to coincide with the leading N-dependence of their classical (tree-level)

expressions:
v
.x; y; z/ D p ı.x y/ı.x z/;
class
3 N
v
˛ˇ
class
.x; y; z/ D p ı.x y/ı.x z/ı ˛ˇ ;
3 N

.x; y; z; w/ D
class ı.x y/ı.x z/ı.x w/;
N

˛ˇ .x; y; z; w/ D
class
ı.x y/ı.x z/ı.x w/.ı ˛ˇ ı C ı ˛ ı ˇ C ı ˛ ı ˇ /;
3N

˛ˇ .x; y; z; w/ D
class ı.x y/ı.x z/ı.x w/ı ˛ˇ : (5.14)
3N
The 3- and 4-point functions with any other index set consistently vanish. In view of
this, we introduce the following scaled 3- and 4-point functions, which are assumed
to be independent of N at LO:
v v
.x; y; z/ D p sss.x; y; z/; ˛ˇ .x; y; z/ D p ı ˛ˇ s .x; y; z/
N N
1
.x; y; z; w/ D ssss.x; y; z; w/;
N
1
˛ˇ .x; y; z; w/ D ss ı ˛ˇ .x; y; z; w/;
N
1
˛ˇ .x; y; z; w/ D .x; y; z; w/.ı ˛ˇ ı C ı ˛ ı ˇ C ı ˛ ı ˇ /: (5.15)
N
One obtains, for the elements of the inverse propagator matrix,

2
G1
˛ .x; y/ D G1
.x; y/ı
˛
D C m C v ı.x y/ı ˛
2
6
v
p G .x; z1 / ˛ .z1 ; z2 ; y/G .z2 ; x/
3 N
h ˛
C ı ı.x y/G .x; x/
6N
G .x; z1 /G .x; z3 / ˛ .z1 ; z2 ; z3 ; y/G .z2 ; x/
CG .x; v1 /G .x; v2 / .v1 ; v2 ; v3 /G .v3 ; z1 / ˛ .z1 ; z2 ; y/G .z2 ; x/
i
C2G .x; z1 / .z1 ; z2 ; y/G .v3 ; z2 /G .v1 ; x/ ˛ .v1 ; v2 ; v3 /G .v2 ; x/
h ˛
C ı .N C 1/ı.x y/G .x; x/
6N
G .x; z1 /G .x; z2 /ˇ˛ˇ .z1 ; z2 ; z3 ; y/G .z3 ; x/
CG .x; v1 /G .x; v3 /ˇˇ .v1 ; v2 ; v3 /G .v2 ; z1 / ˛ .z1 ; z2 ; y/G .z2 ; x/
i
C2G .x; z1 /ˇ .z1 ; z2 ; y/G .v3 ; z2 /G .x; v1 /ˇ˛ .v1 ; v2 ; v3 /G .v2 ; x/ :
(5.16)

2
G1
.x; y/ D C m 2
C v ı.x y/
2
v
p .3G .x; z1 / .z1 ; z2 ; y/G .z2 ; x/
6 N
CG .x; z1 /˛˛ .z1 ; z2 ; y/G .z2 ; x//
h
C 3G .x; x/ı.x y/ G .x; z1 /G .x; z2 / .z1 ; z2 ; z3 ; y/G .z3 ; x/
6N
i
C3G .x; z1 /G .z2 ; v1 / .z1 ; z2 ; y/G .x; v2 / .v1 ; v2 ; v3 /G .v3 ; x/
h
C .N 1/G .x; x/ı.x y/
6N
G .x; z3 /G .x; z1 /˛ ˛ .z1 ; z2 ; z3 ; y/G .z2 ; x/
CG .x; v3 /G .x; v1 /˛ ˛ .v1 ; v2 ; v3 /G .v2 ; z1 / .z1 ; z2 ; y/G .z2 ; x/
i
C2G .x; z1 /˛ˇ .z1 ; z2 ; y/G .z2 ; v3 /G .x; v1 /˛ˇ .v1 ; v2 ; v3 /G .v2 ; x/ :
(5.17)
5.2 The Large-N Closure and Its Recursive Renormalization
Using the scaled quantities, one can find in the large-N limit the LO form for
(5.13), (5.16), and (5.17) determining the 1- and 2-point functions:

p
0D Nv m2 C .v 2 C G .x; x// ;
6

1 2 2
G .x; y/ D C m C .3v C G .x; x// ı.x y/
6
5.2 The Large-N Closure and Its Recursive Renormalization 103
v 2
G .x; z1 / s .z1 ; z2 ; y/G .z2 ; x/;
6

2
G1
.x; y/ D C m 2
C .v C G .x; x// ı.x y/: (5.18)
6
An important feature of the large-N approximation is that it satisfies Goldstone’s

theorem [4], which can be seen by comparing the equation of state with
2
m2 D m2 C .v C G .x; x//; (5.19)
6
the mass term in the pion propagator. In p the presencepof an explicit symmetry-
breaking term in the Lagrangian density, . Nv C /h N, the squared pion mass
is given as h=v. The same relation reappears also in the quark–meson model of
Sect. 5.7 in Eq. (5.158). (The N-scaling of the symmetry-breaking term is again
determined by its extensive nature.)
The other observation is that despite considerable simplifications, this large-N
set is not yet closed; one still needs an equation providing the large-N expression of
s . The equation for ˛ can be obtained by returning to an intermediate stage
in deriving (5.16) for G1˛ , where all terms of the corresponding second functional
derivative of are still kept. One has to take the additional functional derivative of
p to .z/ and extract its large-N asymptotic form. After
this expression with respect
scaling out the factor v= Nı ˛ on both sides, as required by the second scaling rule
in (5.15), one arrives at the following LO equation:

s .x; y; z/ D ı.x y/ı.x z/ G .x; z1 / s .z1 ; z2 ; z/G .z2 ; x/ı.x y/:
3 6
(5.20)
This is a self-consistent equation for the 3-point function, which means that at
LO in the large-N approximation, the set of equations consisting of (5.18) and (5.20)
closes in a natural way.
By the translational invariance of the theory, s depends only on the coordinate
differences x y and x z, which implies momentum conservation. Moreover, the
proportionality of the right-hand side of (5.20) to ı.x y/ implies also
s .x y; x z/ D Q s .x z/ı.x y/; (5.21)
which leads after Fourier transformation and the separation of the delta function,
expressing momentum conservation, to the simple equation
Z

Q s .q/ D Q s .q/ G .p/G .p C q/: (5.22)
3 6 p
The determination of the finite 3-point function as a solution of this equation

is a simple and transparent application of the iterative renormalization procedure
[5]. A perturbation series in powers of is generated for Q s by the iterative
solution of (5.22). Since one recognizes the logarithmic divergence of the “bubble”
integral formed with the pion propagators at each step of the iteration, one has to
renormalize the coupling in order to keep Q s finite. Separating the finite (BF )
.0/
and the divergent parts (Bd ) of the bubble integral and splitting the coupling into
renormalized (R ) and counterterm (ı) piece yields the equation in the following
form:

R C ı 1 .0/
Q s .q/ D 1 Q s .q/.BF C Bd / ;
3 2
Z
.0/
B.q/ D G .p/G .p C q/ D BF .q/ C Bd : (5.23)
p
The goal of the perturbation theory is to determine both Q s .q/ and ı in power
series of R :
1
X 1
X .n/
Q s .q/ D nR Q.n/
s .q/; ı D cd nC1
R : (5.24)
nD1 nD1
The iteration begins with the classical value
1
Q.1/
s .q/ D : (5.25)
3
The first iteration consists in substituting this value into the second term of the right-
hand side and comparing the two sides at the level of 2R ,
2
.1/ R 2R F .0/
2R Q.2/
s .q/ D cd .B C Bd /; (5.26)
3 18
which gives
1 F .1/ 1 .0/
Q.2/
s .q/ D B .q/; cd D B : (5.27)
18 6 d
It is worthwhile following in detail the second iteration, since it provides insight
into the cancellation of the so-called subdivergences. The equality at level 3R takes
the form
1 .2/ 1 F .0/ .1/
Q .3/ .q/ D cd .B C Bd /.Q .2/ .q/ C cd Q .1/ .q//
2 6
1 .2/ 1 .0/2
D cd .B BF2 /: (5.28)
2 108 d
5.2 The Large-N Closure and Its Recursive Renormalization 105
.0/
The absence of the mixed product term BF Bd exemplifies the cancellation of
subdivergences, and one easily identifies
2
1 1 F 2 .2/ 1 .0/
Q.3/
s .q/ D B ; cd D B : (5.29)
3 6 6 d
Further, one guesses and proves by complete induction the following recurrence
relations:
1 F Q .n1/ .n/ 1 .0/ .n1/
Q.n/
s .q/ D B s .q/; cd D B c : (5.30)
6 6 d d
Since these geometric series can be summed, one ends up with the resummed and
renormalized nonperturbative result
R 1 R .0/ 1
Q s .q/ D ; ı D B : (5.31)
3 1 C 6R BF .q/ 6 d 1 R B .0/
6 d
The term “resummation” is rightly used, since the result is the sum of the infinite
pion bubble series (see Fig. 5.1) contribution to the 3-point function emerging from
the original quartic Lagrangian density by the shift (5.2).
The same results can be obtained also by a more formal (more intuitive?)
procedure when one directly solves (5.22) as
1
Q s .q/ D
3 1 C .B .q/ C B .0/ /
F
6 d
1 1 1 1
D D 1 1
; (5.32)
3 1 C 1 B .0/ C 1 B F .q/ 3 R
C B F .q/
6
6 d 6
1 1 1 .0/
D C Bd : (5.33)
R 6
The LO solution is completed by finding the explicitly finite equation for G1

starting with the formal expression in (5.18). Using the first equation of (5.18)
together with (5.20), one obtains
2 Q
G1
.x; y/ D ı.x y/ C v s .x y/: (5.34)
π π π π π
s s s s
= + + + ...
π π π π π
Fig. 5.1 The resummed bubble series contributing to the s vertex in the large-N limit
When we apply (5.21) and (5.31), this propagator assumes a rather simple finite
expression in Fourier space without any need for further renormalization:
2 Q R v 2 1
G1 2 2
.q/ D q C v s .q/ D q C : (5.35)
3 1 C 6R BF .q/
The last step is to renormalize the equation of state. The tadpole integral of a
zero-mass pion propagator is a pure quadratically divergent quantity:
Z
.2/
G .x; x/ D G .p/ D Td : (5.36)
p
With the help of this representation, one can easily write the following for the
equation of state:

p 6 2 .2/
p 6 2
0D Nv m C v 2 C Td D Nv mR C v 2 ; (5.37)
R
which requires the counterterm relation

m2 .2/
ı C Td D 0: (5.38)

5.3 Landau Singularity and Triviality
The formally perfect analysis of the renormalization reveals its limitation when
.0/
one analyzes (5.33), employing the explicit expression of Bd obtained with cutoff
regularization:
.0/ 1 2
Bd D 2
ln 2 ; (5.39)
16 M0
where is the maximal momentum at which the integral defining B.q/ in Eq. (5.23)
is cut off, and M0 is the normalization scale. One recognizes R ./ D by
choosing M0 D . Using this, the relation defining the renormalized coupling can
be considered a function describing the -dependence of R .M0 /:
R ./
R .M0 / D R ./ 2
: (5.40)
1C 96 2
ln M 2
0
5.3 Landau Singularity and Triviality 107
Renormalizability in the strict sense means finding results independent of the

regularization, which requires sending to infinity. But then one obtains (keeping
R ./ fixed)
lim R .M0 / D 0; (5.41)

!1
which means that the coupling characterizing the pointlike source is completely
screened at every finite scale. The feature revealed by this analysis is called
triviality. The same phenomenon was found in the one-loop analysis of the one-
component scalar theory in Sect. 3.5.2. Here, however, it is an exact statement in
the N ! 1 limit. Only asymptotically free theories escape this consequence of
the scale-dependence of couplings. Quantum electrodynamics and the Higgs sector
of the standard model share this feature, and these sectors can be used only as
effective theories up to some maximal momentum (cutoff) value. In such theories,
renormalization turns into a weaker statement: one expects that physical results are
not sensitive to a sufficiently high choice of . However, the fulfillment of such
an expectation can be checked only numerically by choosing at physically relevant
scales phenomenologically “dictated” values for the couplings.
The presence of a maximal allowed momentum in the theory is signaled by a
singularity: the Landau pole. It shows up in the momentum-dependence of the
coupling when one continues the theory from the Euclidean to the Minkowski
metrics with the help of the following rules:
q2E ! q2M ; BEF .qE / ! BM

F
.qM /; G1.E/
! iG1.M/
; (5.42)
resulting in
R 1
Q.M/
s .qM / D : (5.43)
3 1 R q2M
96 2
ln M02
This expression obviously has a pole located at the imaginary momentum

2 =
q2M D M02 e96 R
; (5.44)
which by analyticity in the complex qM -plane puts a limit on the range of

convergence in momentum. The larger the value of R .M0 /, the lower this upper
limit.
The singularity shows up also in the physical (Minkowskian) spectra of the -
channel. It leads to the existence of a purely imaginary (tachyonic) zero of the
equation
R 2 1
iG1.M/
D q2M v D 0; (5.45)
3 1 R q2M
96 2
ln M02
which should be separated (to lie much higher in absolute value) from the domain
of physical momenta described by the effective theory. In this sense, the energy
scale of the Landau pole represents an upper bound on the masses of the physically
meaningful excitations. If one identifies the O.N/ model with the Higgs sector of
the standard model, one obtains an upper bound on the mass of the Higgs boson [6].
In the case of the strong interaction, a similar bound is produced for the mass of the
-meson. This issue will be discussed in applications of the linear -model to the
phenomenology of strong interactions.
5.4 Auxiliary Field Formulation of the O.N/ Model
An alternative formulation of the model (5.1) is sometimes more convenient for

developing the large-N expansion. It transforms formally the term depending
quartically on and ˛ into a quadratic dependence with the help of the so-
called Hubbard–Stratonovich transformation. The price to pay is the introduction
of a new field (
.x/),
N which is a quadratic composite of the original variables and
can be thought of as describing in a pointlike approximation some two-particle
(bound) state of them. The extended classical action has the following form after
the shift (5.2):
Z
1 2 1 2
SŒ; ˛ ;
;
N v D d4 x @m .x/ C @m ˛ .x/
2 2
2 p p p
m
C 2 .x/ C ˛2 .x/ C 2 Nv.x/ C Nv 2 Nh. Nv C /
2
r
1 2 1 2 2
p 2

N .x/ C
.x/
N .x/ C ˛ .x/ C 2 Nv.x/ C Nv : (5.46)
2 2 3N
The variable
.y/
N has no dynamics; the variation of S with respect to
N leads to a
local algebraic expression for it, which, after being substituted back, reproduces the
original action. Applying the rule (5.6) to the extended action, one arrives at the
corresponding set of 1- and 2-point equations. The 1-point equations for and
N
can be written as follows
p (setting in the final result the nonzero background for the
auxiliary field
D
N =3N):
r

Nv h C m2 C
C vG
.x; x/ D 0;
3
3N 1 2

C Nv C .N 1/G .x; x/ C G .x; x/ D 0: (5.47)
2
5.4 Auxiliary Field Formulation of the O.N/ Model 109
A second functional derivation gives the equations that determine the 2-point
functions:
r
1 1
G .x; y/ D D .x; y/ G
A .x; z/GB .w; x/AB .z; w; y/;
3N
r
1 1
G
.x; y/ D D
.x; y/ G
A .x; z/GB .w; x/AB
.z; w; y/;
3N
G1 1

.x; y/ D D
.x; y/
r h i
1
G A .x; z/GB .w; x/ C G˛A .x; z/GB˛ .w; x/ AB
.z; w; y/;
2 3N
r
1 ˛ˇ 1
G˛ˇ .x; y/ D ı D .x; y/ G
A .x; z/GB˛ .w; x/ABˇ .z; w; y/: (5.48)
3N
The indices A; B run through the set ; ˛ ;
, and one has to integrate/sum over all
repeated space coordinates/indices. The tree-level propagators appearing here have
the following expressions in Fourier space:
r

D1
.k/ D D1
.k/
2
D k CM ; 2
D1

.k/ D 1; D1

.k/ D v: (5.49)
3
Here we introduced an abbreviated notation for the permanent combination
m2 C
D M 2 : (5.50)
From (5.46), one finds that only

and
assume nonzero tree-level values,
r

.x; y; z/ D ı.x y/ı.x z/;
˛ˇ .x; y; z/ D ı ˛ˇ
.x; y; z/;
3N

D
: (5.51)
Since all other 3-point functions vanish at this level, in an iterative sense their exact
value is also compatible with zero. On the other hand, for the nonzero vertices, one
can argue that in (5.48) it is sufficient to use the above tree-level expressions if one
is interested in the determination of the propagators with O.1=N/ accuracy.
The argument goes as follows. A first type of typical quantum contribution to

has the following structure:
r

G
C GDA CD
GB AB ; (5.52)
3N
which iteratively contributes at level O.N 3=2 /, and is therefore negligible at the
required accuracy. The other typical contribution is of the form
r

G
A GB AB
: (5.53)
3N
The iterative determination of the 4-point function begins with the contribution of
terms that contain exclusively lower n-point functions (that is, propagators and 3-
point functions). As one can trace by further derivations of (5.48), the iterative
determination of any 4-point function then pcontains at least the functional product of
three 3-point functions plus the explicit 1= N factor. Therefore, the N-dependence
of this contribution can be estimated as N 2 . This makes the second type of
contribution also negligible.
This argument allows us to write the Dyson–Schwinger equations completely
explicitly with O.1=N/ relative accuracy using (5.51):
G1 2
.x; y/ D . C M /ı.x y/

G
.x; y/G .y; x/ C G

.x; y/G
.y; x/ ;
3N
r

G1

.x; y/ D vı.x y/ G
.x; y/G .y; x/;
3 3N

G1

.x; y/ D ı.x y/ ŒG .x; y/G .y; x/

6N
C.N 1/G .x; y/G .y; x/ ;

G1 2
.x; y/ D . C M /ı.x y/ G
.x; y/G .y; x/: (5.54)

3N
These equations form a closed set together with the two equations of (5.47), which
demonstrates that the 1=N expansion provides a natural closure also at the NLO.
Before proceeding to a discussion of the renormalization of these equations, we
remark that one can construct an effective action from which these equations follow
by considering v;
; G , G
, G
, and G to be independent variational variables:
1 3N 2
eff Œv;
; G ; G D NM 2 v 2
Nhv
2 2R
Z
1
C .N 1/.ln G1 1
.k/ C D .k/G .k// C Tr ln G
1
.k/ C Tr D 1 .k/G .k/
2 k
Z Z
R
G .k/G .k C p/G
.p/ C .N 1/G .k/G .k C p/G
.p/
12N k p

C 2G
.k/G
.p/G .p C k/ C Œv;
; G ; G : (5.55)
5.5 Renormalization of the O.N/-Model in the Auxiliary Formulation 111
Here the matrices

! 0 q 1
R
D1

.k/ D1

.k/
1 v
D 1 .k/ D D @q 3 A;
D1

.k/ D1
.k/
R
k2 C M 2
3 v

G
.k/ G
.k/
G D (5.56)
G
.k/ G .k/
are introduced. The second represents the components of the exact propagator
matrix in the
sector; the first is just the tree-level expression of its inverse
matrix. Here the part of the action depending on the renormalized coupling is
separated from the counteraction depending on the countercouplings. The
countercouplings themselves have to be determined as functions of the renormalized
parameters. This action coincides with the NLO part of the two-particle irreducible
(2PI) action obtained by introducing explicit bilocal sources conjugated to the
quadratic combinations of the field variables in addition to the source terms linear
in the fields and constructing the functional Legendre transform with respect to
both (see Sects. 2.8 and 4.7). The effective action was obtained in this way in
[7].
5.5 Renormalization of the O.N/-Model in the Auxiliary

Formulation
In this section, the counterterm Œv;

; G ; G displayed in the effective
action (5.55) will be constructed, which provides on variational differentiation
the appropriate subtractions renormalizing (5.47) and (5.54) of the 1- and 2-point
functions. The task will be solved in two steps. One starts at LO, and using its
results, one can proceed to the determination of the NLO piece of . The
treatment follows the constructive strategy of [8, 9]. For notational simplicity
below, the index “R” is omitted from the symbols of the renormalized coupling and
the mass parameter appearing in (5.55).
5.5.1 Leading-Order Counterterm Action Functional
The respective LO parts of the equations of (5.47) in momentum representa-

tion are reproduced by taking the derivative of (5.55) with respect to v and
and keeping only terms proportional to N. It will be sufficient to assume that

.LO/
.LO/ Œv;
; G ; G D
Œ
:
ıeff
D N.vM 2 h/ D 0;
ıv
Z .LO/
ıeff 3N N ı
D
C v 2 C G .k/ C D 0: (5.57)
ı
2 k ı
Only the second equation requires a counterterm contribution, since it has the
divergent piece originating from the pion tadpole. It is a modified expression
of (5.36), valid for nonzero pion mass. From here on, a unified notation of
the divergent (strongly cutoff-dependent) quantities will be used throughout the
.2/ .0;1/
chapter. The quadratically divergent expression Td is denoted by Id , and the
.0/ .0;2/
logarithmically divergent Bd by Id . The complete set of divergent quantities
relevant for the renormalization will be listed in Eq. (5.75). With the new notation,
the finite part of the pion tadpole is defined after subtracting the quadratic and the
logarithmic divergences:
Z
.0;1/ .0;2/
G .k/ D TF C Id C .M02 M 2 /Id : (5.58)
k
The divergence due to the pion tadpole is compensated by the following

contribution to the counterterm action,
N h .0;1/ 2
.0;2/ i

.LO/ Œ
D
Id m M02 C Id ; (5.59)
2 2
resulting in a finite equation that determines
(called a saddle-point equation):
3 1 2

C v C TF D 0: (5.60)
2
The meaning of M 2 is determined by the LO expression of the pion propagator:
ıeff 1 h
D N G1 1
.k/ C D .k/ D 0; M2 D M 2 D : (5.61)
ıG .k/ 2 v
The last equality, which exploits the equation of state, expresses Goldstone’s
theorem in the presence of the explicit symmetry-breaking source h.
The leading-order propagators in the coupled
sector are determined
by the NLO piece of (5.55) using the above LO solution of the model in
the auxiliary formulation. Since its result for G is necessary for the com-
plete comparison with the LO solution obtained with the approach based on
the directly derived Dyson–Schwinger equations, we discuss this sector in this
subsection.
The inverse propagator matrix to leading order can be read off from (5.54)
keeping only O.N 0 / terms on the right-hand side:
0 q 1
.0;2/
1 .B
F
.k/ C I / v
1
.k/ D @ 6 q 3 A:
d
GLO (5.62)
2 2
3 v k C M
Clearly, the cancellation of the logarithmic divergence in G1

requires the following

additional piece into the counteraction functional:
Z
.0;2/

ŒG
D I G
.k/: (5.63)
12 d
Then the renormalized propagator matrix is given as

0 q 1
1 2 2
k C M v
GLO .k/ D .k2 C M 2 / 1 C BF .k/ C v 2 @ q 3 A;
6 3
3 v 1 6 BF .k/
(5.64)
which leads for G to the same expression as (5.35), generalized to nonzero pion
mass. The common denominator signals that the and the composite
2 C
2 fields have degenerate spectra (in the broken symmetry phase), a phenomenon
sometimes called hybridization.
Some further insight can be gained into the working of the large-N expansion by
rederiving the inverse propagator matrix of the
sector in an alternative way.
Let us return to (5.46). Recognizing that it is formally quadratic in ˛ .x/, one is
allowed to perform the functional integral over this multiplet in an inhomogeneous
.x/
N background. In this integration, however, one has to include the part of the
counteraction depending on the pion fields. Therefore, in this part of the action one
has to make use of the bare couplings. In order to avoid confusion, we append to
them explicitly the lower index B. From the auxiliary field, the constant piece is
transferred into the effective pion mass term
r
B

N !
N0 C
;
N MB2 D m2B C
N0 ; (5.65)
3
yielding in this way the effective theory of the

sector:
Z "
1 4
p p
Seff Œ;
;
N v D .@m .x//2 Nh. Nv C .x//
d x
2
r ! #
1 B p .
N C
.x//
N 2
2 2 2 0
C MB C
.x/
N .x/ C 2 Nv.x/ C Nv
2 3N 2
r !
N 1 B
C Trx log C MB2 C
.x/
N : (5.66)
2 3N
If we take the derivative of this effective action with respect to v, the equation of
state tells us that the combination MB2 is finite: MB2 D M 2 . This is reflected already
in the notation if we rewrite the last term, relying on the basic features of the
logarithm, as
r !
N1 2 B
Trx log C M C
.x/
N
2 3N
" 1 r !n #
N1 2
X .1/nC1 B
D Trx log C M C D .x; z/
.z/
N : (5.67)
2 nD1
n 3N
The elements of the inverse

propagator matrix are constructed by computing
the second functional derivative of (5.66) with respect to the appropriate pairs of
1
the field variables. Not surprisingly, we arrive at the matrix GLO as given in (5.62)
with the replacement ! B , which makes the
element finite. It automatically

contains the additional O.N 0 / contribution to the propagator matrix in the
-
sector relative to its tree-level expression. This rederivation of (5.64) reveals an
important advantage of using the auxiliary field: the contribution of a single pion-
loop integral to the self-energy of the auxiliary field modifies the -propagator in
the same way as does an infinite set of iterated pion bubbles. Also, (5.67) explicitly
displays the large-N hierarchy of the different n-point functions of the composite
field
,
N which was important in the argument of the previous section for the natural
truncation of the Dyson–Schwinger equations at O.1=N/ order.
5.5.2 Next-to-Leading-Order Counterterm Action Functional:

Effects of the Landau Pole
The NLO part of the counterfunctional .NLO/ will be determined as the sum of
three terms:
.NLO/ ŒG ; v;
D ŒG ; v;
C v Œv;
C
.NLO/ Œ
: (5.68)
The piece cancels the divergence of the NLO part of the pion propagator.
It contributes also to the expressions of ı .NLO/ =ıv and ı .NLO/ =ı
. The
next piece is determined by the requirement of suppressing the divergences of
the equation of state. It contributes to ı .NLO/ =ı
, but has no feedback on
.NLO/
ı .NLO/ =ıG . Finally,
depends only on
. Therefore, it plays no role in
canceling divergences in the other two equations. The three additive contributions
are determined in this sequence.
5.5.2.1 Determination of ı ŒG ; v;
The only 2-point function that has to be renormalized at the O.1=N/ level for
the determination of .NLO/ is the equation of G . The Fourier transform of its
expression in (5.54) reads
Z
2 ı
G1
.k/ D D1
.k/ G
.p/G .p C k/ C : (5.69)
3N p N ıG .k/
Here in the last term, is the only piece of the full counterterm action that
contributes to the pion propagator. It will be found by “integrating” the divergent
O.1=N/ contribution with respect to G .k/. In the integrand of the second term
on the right-hand side, one is allowed to replace both propagators by their N 0
accurate expressions:
1
D .q/ D ;
q2 C M 2
.p/
D
.q/ D 2
v2
: (5.70)
1C 6 B .q/
F C 3 v D .q/ 1C 3 .p/D .p/
Note that .p/ is defined here. After substitution, one can expand the integrand in
powers of .p/D :
Z X1 n
1 .p/ 2 2 ı
G1 1
.k/ D D .k/ C .p/D .p C k/ v D .p/ C :
3N p nD0
3 N ıG .k/
(5.71)
Since among the terms of the expansion, arbitrarily high powers of .p/ appear, it
is clear that one is not allowed to send the upper limit of the momentum integral to
infinity. The cutoff value should be chosen below the location (5.44) of the Landau
singularity. The sensitivity of the different terms is illustrated by Fig. 5.2, which
displays the cutoff dependence of several integrals from the set
Z
.n;m/
I D n .p/Dm
.p/ (5.72)
p
4.000
3.500 I¯2,1
I¯2,2 /5
3.000
I 1,1 /Λ2
2.500
2.000
1.500
1.000
0.500
0.000
1.004
1.002
1.000
0.998
0.996 λ = 80 2 − I¯F2,1 2 − I¯F2,2 I¯3,3
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Λ/Λp
Fig. 5.2 The dependence of the “divergent” integrals (upper part) and subtracted or convergent
integrals (lower part) on the cutoff measured in proportion to the estimated location of the
Landau pole (p ). The integrals are defined in (5.72). One applies ! 1 in the finite bubble
integrals BF .p/ and B0F .p/, with the exception of the dashed curve corresponding to j D k D 2,
for which a cutoff regularization is applied. The bar on I j;k indicates that the integrals are scaled
by the value taken at the inflection point (inflection of I 1;1 =2 when j D k D 1) and, for the
sake of the presentation, also by an additional factor in the case of j D k D 2. We set M D 1,
M0 =M D 2, and used D 65 for the coupling, except where indicated (from [9])
up to the Landau pole for various choices of n; m for a pion mass much below the
cutoff (M 2 2 ).
From Fig. 5.2, it is obvious that the integrals that are termed “convergent” by a
simple power counting (m > 2) become insensitive to the cutoff well below the
Landau pole, while the integrals with m D 1; 2 display clear numerical evidence
for divergence. The graphs in the lower part demonstrate that with appropriate
subtractions, these integrals are made cutoff-insensitive. Such subtractions are
applied in the consistent renormalization procedure described below. For this reason,
the conventional classification of the integrals into “divergent” and “convergent”
classes will be employed below, although one is not allowed to send the cutoff
beyond the Landau singularity. One expects that the NLO correction to the pion
propagator will become insensitive to the actual value of the cutoff momentum
below the Landau pole if one subtracts appropriate expressions from the integrals of
the n D 0; 1 terms of (5.71). This is a mathematically less strict but pragmatic
understanding of the renormalization [9], which represents an intuitive attempt
to carry over the results of perturbative renormalization theory to this cutoff
theory.
Before proceeding to the construction of the counterterms, one has to investigate

whether there is any prospective dependence of the divergences on the external
momentum of the pion propagator. Consider first the difference
Z X1 n
Z
1 k2 C 2pk
.p/.D .p C k/ D .p// D .p/ 2 2 : (5.73)
p p p C M 2 nD1 p C M2
Any contribution to the possible wave function renormalization should show up in

terms proportional to k2 . These contributions are contained in the n D 1; 2 terms of
the above expansion. Terms of the integral proportional to k2 are
Z
1
.p/ Œ4.pk/2 k2 .p2 C M 2 /; (5.74)
p .p2 C M 2 /3
from which the term in the integrand proportional to k2 M 2 is convergent, while

by the O.4/ invariance of the Euclidean integrand, one is allowed to make the
replacement 4.pk/2 ! k2 p2 . This proves that the expression of the pion propagator
displays only momentum-independent divergence.
The subtractions will be made independent of the dynamical variables of the
system (recall that M 2
) by subtracting a combination of the following divergent
integrals, each depending on a single mass scale, the renormalization scale M0 :
Z Z Z
.0;1/ .0;2/
Id D D0 .p/; Id D D20 .p/; Ia.1;1/ D 0 .p/D0 .p/
p p p
Z Z
Ia.2;2/ D 20 .p/D20 .p/; Ia.2;I/ D 20 .p/D20 .p/B0F .p/; (5.75)
p p
where
Z
1 .0;2/
D0 .p/ D ; B0F .p/ D D0 .q/D0 .p C q/ Id ;
p C M02
2
q

0 .p/ D
: (5.76)
1C 6 B0 .p/
F
The divergence of the n D 1 term of (5.71) is trivially seen to be given by the

same integral with the index replaced everywhere by 0. One promptly identifies
the divergent integral with one of the divergent expressions listed in (5.75):
Z Z
v2 ˇ v2 v 2 .2;2/
2
.p/D2 .p/ˇdiv D 20 .p/D20 .p/ D I : (5.77)
9N p 9N p 9N a
The divergence analysis of the n D 0 terms makes use of the following iterative
series of D and .p/ in terms of D0 and 0 .p/:
D .p/ D D0 .p/ C .M02 M 2 /D0 .p/D .p/;

1
.p/ D 0 .p/ C 0 .p/.p/.B0F .p/ BF .p//: (5.78)
6
Direct analytic integration allows one to determine the difference B0F .p/ BF .p/,
which one needs here with O.1=p2/ accuracy:
B0F .p/ BF .p/

1 2 2 2 M2 2 2
D 3.M M0 / M ln 2.M M0 /B0
F
.p/ D0 .p/: (5.79)
8 2 M02
It is obvious that the iteratively produced difference between the quantities defined
with different mass scales are O.p2 / in both cases. For finding the divergent piece
to be subtracted from the n D 0 term of the expansion, it is sufficient to approximate
the differences with the first terms of the respective iterations and write
Z Z
1 ˇ 1 1
.p/D .p/ˇdiv D 0 .p/ C 20 .p/.B0F .p/ BF .p//
3N p 3N p 6
ˇ
D0 .p/ C .M02 M 2 /D20 .p// ˇdiv : (5.80)
In the product, one keeps only the divergent terms

Z Z(
1 ˇ 1
.p/D .p/ˇdiv D 0 .p/D0 .p/ C .M02 M 2 /0 .p/D20 .p/
3N p 3N p
)
1 3 1 M2
C 20 .p/D20 .p/ 2 2
2B0 .p/ .M M0 /
F 2
M ln 2 : (5.81)
6 8 2 8 2 M0
Since the counterterm should be a polynomial of the variables, one reexpresses the
last term in the square brackets with the help of the explicit expression for the finite
part of the tadpole integral:
1 2 M2 1
M ln D TF .M 2 M 2 /
16 2 M02 16 2 0
Z
.0;1/ .0;2/ 1
D G .p/ Id .M02 M 2 /Id .M 2 M 2 /; (5.82)
p 16 2 0
which is nothing but the relation (5.58). The final form to be used in the determina-
tion of is the following:
Z
1 ˇ
.p/D .p/ˇdiv
3N p
"
1 .1;1/ 1 2 2 .2;I/ .2;2/ 2 2 1 1
D Ia C .M0 M /Ia C Ia .M0 M /
3N 2 24 2
Z #
1 .2;2/ .0;1/ 2 2 .0;2/
Ia G .p/ Id .M0 M /Id : (5.83)
3 p
The sum of the two contributions (5.77) and (5.83) determines the necessary
counterterm
ŒG ; v;

Z (
1 1
D G .q/ Ia.1;1/ C .M02 M 2 /Ia.2;I/
6 q 2
2
v 1 1
C Ia.2;2/ C .M02 M 2 /
3 24 2
Z )
1 .2;2/ 1 .0;1/ .0;2/
Ia G .q/ Id .M02 M 2 /Id : (5.84)
3 2 q
5.5.2.2 Determination of v Œv;
One can proceed further with the equation of state:

r Z
ı ıv ı
D N.M 2 v h/ C D
.p/ C C D 0: (5.85)
ıv 3 p ıv ıv
Using the explicitly known expressions for D

and , one has to determine a
new piece of the counterterm action, v , not allowing
R any feedback on the pion
propagator equation, that does not depend on T q D .q/. The equation to be
integrated is the following:
Z X1 n ˇ
1 .p/ 2 ˇ ıv 1
0D .p/D .p/ v D .p/ ˇ C2 C T Ia.2;2/ :
3 p nD1
3 div ıv 9
(5.86)
The cancellation condition of the divergences can be written down instantly, since
the first term on the right-hand side exactly coincides with the expression already
analyzed in connection with the pion propagator:
(
2ıv 1 .2;2/ 1 v 2 .2;2/ 1
0D 2
C T I C Ia C Ia.1;1/ C .M02 M 2 /Ia.2;I/
ıv 9 a
3 3 2
h
1 1
C Ia.2;2/ .M02 M 2 /
24 2
)
1 .0;1/ 2 2 .0;2/
i
T Id .M0 M /Id : (5.87)
3
The mutual cancellation of the two terms proportional to T is important from two
aspects. First, it ensures the independence of v on G . Second, it demonstrates
.2;2/
the cancellation of a subdivergence TF Ia , the presence of which would be
dangerous, since T is environment- (temperature- and/or density-) dependent. The
F
integration of the resulting form of the condition is very simple; it gives

"
v 4 .2;2/ v 2 1
v Œv;
D I Ia.1;1/ C .M02 M 2 /Ia.2;I/
36 a 6 2
#
1 1 1 .0;1/
.0;2/
CIa.2;2/ C Id C .M02 M 2 /Id : (5.88)
24 2 3
It is interesting to note that a counterterm v 4 is also generated, although the

(renormalized) coefficient of this term in the auxiliary formulation of the model is
fixed to zero. This is not forbidden, since this piece is fully compatible with the
symmetry of the system.
.NLO/
5.5.2.3 Determination of Œ
The final task is the renormalization of the saddle point equation that determines
the actual value of the composite field
. Both previously determined counterterms
contribute to the divergence cancellation condition to be satisfied for the NLO-
renormalization of ı =ı
:
Z ˇ Z ˇ
1 ˇ 1 ˇ
0D G.NLO/
.k/ ˇ C .D .k/ D .k// ˇ
2 k div 2 k div
.NLO/
ıŒ C v ı
C C : (5.89)
ı
ı
From (5.84) and (5.88), one directly obtains

ıŒ C v 1 2 1 .2;I/ 1 1 1 .0;2/
D v C T Ia C Ia.2;2/ C I ;
ı
6 2 24 2 3 d
(5.90)
which has to be canceled by the divergences of the first two terms for a self-
consistent determination of
, since this functional can depend on neither on
v 2 nor T .
The divergence of the second term on the right-hand side of (5.89) is found rather
easily by exploiting the explicit expression of D from (5.64):
Z ˇ 1 Z n ˇ
1 ˇ 1X .k/ 2 ˇ
.D .k/ D .k// ˇ D D .k/ v D .k/ ˇ
2 k div 2 nD1 k
3 div
Z ˇ
v2 2 ˇ v 2 1 .2;2/ 1 .2;I/
D D .k/.k/ˇ D I C Ia : (5.91)
6 k div 6 a 6
The divergence analysis of the first term on the right-hand side of (5.89) begins
by finding a more explicit representation for the NLO correction of the tadpole
integral of the pion, to be calculated from the definition of the renormalized NLO
self-energy:
1 .NLO/
G.NLO/1
.p/ D D1
.p/ C ˙ .p/;
N
Z
1 1
˙.NLO/ .p/ D G
.k/D .p C k/ tdiv C Ia.2;2/ .v 2 C T / : (5.92)

3 k 3
The expression completing the integral in the square brackets ensures the finiteness
.NLO/
Rof ˙ . Here tdiv is a shorthand notation for the part of the divergence of
2
k G
.k/D .k/ that is independent of both v and T :
1
tdiv D Ia.1;1/ C .M02 M 2 /Ia.2;I/
2

1 1 1 .0;2/ 1 .0;1/
C Ia.2;2/ .M 2 M02 / C I C I : (5.93)
24 2 3 d 3 d
.NLO/1
Calculating the reciprocal of the expression of G with O.1=N/ accuracy
gives
Z Z
1 1
G.NLO/
.k/ D D2 .k/˙.NLO/ .k/
2 k 2 k
Z Z
1
D D2 .k/ D
.p/D .p C k/
6 k p

1 1
C B .0/ tdiv Ia.2;2/ .v 2 C T / : (5.94)
6 3
One obtains the complete divergence by analyzing the divergence of its first term
(the second belongs fully to the divergent part). One can interchange the k- and
p-integrations and use the analytic expression of the internal integral:
Z
1 1
D2 .k/D .p C k/ D BF .p/ BF .0/ (5.95)
k p C 4M 2
2 8 2
with
1 M2
BF .0/ D 2
ln 2 : (5.96)
16 M0
For the relevant integral, one arrives at a representation very convenient for the
divergence analysis:
Z ˇ
1 1 1 ˇ
D
.p/ BF .p/ BF .0/ ˇ

6 p p2 C 4M 2 8 2 div
Z ˇ Z ˇ
1 ˇ 1 1 ˇ
D D
.p/D .p/BF .p/ˇ D
.p/D .p/ BF .0/ C ˇ

6 p div 6 p 8 2 div
Z ˇ
1 3M 2 1 ˇ
.p/ 2 B F
0 .p/ B
F
.0/ ˇ : (5.97)
6 p .p C M02 /2 8 2 div
In the first integral, one has to expand D
up to the term linear in v 2 to obtain

Z ˇ
1 ˇ 1
D
.p/D .p/BF .p/ˇ D v 2 Ia.2;I/

6 p div 18
.0;1/ .0;2/ 1 1 .2;2/
C Id C .M02 M 2 /Id tdiv C I T : (5.98)
3 a
The second integral of (5.97) is dangerous, since it is proportional to BF .0/, which
depends logarithmically on M 2 . However, it is multiplied just by the integral giving
the divergence of ˙ .NLO/ . Therefore, it is canceled by a part of the last term of (5.94).
We obtain in this way
Z ˇ
1 1 ˇ
D
.p/D .p/ B .0/ C

F
ˇ
6 p 8 2 div

1 1 .2;2/ 2
C B .0/ tdiv Ia .v C T /
6 3

1 .0;2/ 1 1 .2;2/ 2
D Id tdiv I .v C T / : (5.99)
6 8 2 3 a
In the third integral of (5.97), again the dangerous factor BF .0/ appears. However,
it is now multiplied by M 2 and can be replaced using its relationship to the tadpole
integral:
.0;1/ .0;2/ 1
M 2 BF .0/ D T C Id C .M02 M 2 /Id C .M 2 M02 /: (5.100)
16 2
Using this, the last integral produces the following divergent pieces:
Z ˇ
1 3M 2 1 ˇ
.p/2 C M 2 /2
B F
0 .k/ B F
.0/ 2
ˇ
6 p .p 0 8 div

.0;2/ 1 1 .2;I/
D 3M 2 Id 2 Ia.2;2/ I
6 a

1 1 .2;2/ 1 .2;I/
C Ia C Ia
2 6
2
M .0;1/ 2 2 .0;2/ 1 2 2
T C I C .M 0 M /I C .M M0 / : (5.101)
8 2 d d
16 2
One now adds together the contributions to the divergence of the two integrals
appearing in the saddle-point equation that are proportional to v 2 C T , and
which can be extracted from Eqs. (5.91), (5.98), (5.99), and (5.101). One finds
that the sum exactly cancels the contributions arising from the
-derivatives of
Cv in (5.90). This means that the realization of the renormalization program
.NLO/
consistently ends with
:
.NLO/
ı
.0;1/ 2 2 .0;2/ 1 .0;2/ 1 6
D Id .M0 4M /Id Id tdiv
ı
6 8 2

1 1 .2;2/ 1 .2;I/
Ia C Ia
2 6

.0;1/ .0;2/ 1 6M 2
Id C .M02 M 2 /Id C .3M 2
M 2
0 / C : (5.102)
16 2
.NLO/
It is easy to see that
is also a quadratic polynomial in M 2 , as was the case
.LO/
for
.
5.5.2.4 Summary and Discussion
The LO+NLO counterterm functional that has to be added to (5.55) has a rather
simple functional form:
2
1 2
ŒG ; G
; v;
D .v C T / Ia.2;2/
6
1 .N/ .0/ .N/ .0/
.v 2 C T /tdiv C .Nc1 C c1 /
C .Nc2 C c2 /
2 ; (5.103)
6
where tdiv was defined in (5.93), and the somewhat complicated divergent coeffi-
.N/ .0/ .LO/ .NLO/
cients ci and ci can be found from
C
. When one combines this
expression with the renormalized couplings multiplying the corresponding function-
als in (5.55), one obtains an expression with the bare couplings for the renormalized
effective action. There are several couplings with vanishing renormalized value, still
receiving a nonzero counterterm. An important example is given by the first term
of (5.103), .v 2 C T /2 .
One can eliminate easily the auxiliary field
from the effective action by
completing its quadratic action into a perfect square and substituting for the
propagators G
; G
their expressions through G . In this way, one arrives at a
form that coincides with the 2PI action of the O.N/ model in the original variables
[10].
Even though the renormalized action is given in its 2PI appearance, one has to
understand that the renormalized action constructed above is not renormalized as
an approximate 2PI solution. In the strict 1=N expansion, the subsequent terms are
determined hierarchically, while in the 2PI approach, even in the large-N limit, the
propagators are determined self-consistently [11].
An important consequence of the fact that the counteraction functional depends
only on the sum v 2 C T is that the same subtraction is applied to the equation
of G1 .p/ and v. It ensures that Goldstone’s theorem is also obeyed in the theory
renormalized at the NLO level, since it is obviously satisfied in the unrenormalized
expression.
5.6 Large-n Approximation in U.n/ Symmetric Models
The linear sigma model is formulated in terms of the matrix M with complex scalar
field elements
a
M.x/ D .sa .x/ C i a .x// ; (5.104)
2
5.6 Large-n Approximation in U.n/ Symmetric Models 125
where sa . a / represents the (pseudo)scalar meson multiplet. The matrices a =2,

a D 0; 1; : : : ; n2 1, are the generators of the U.n/ unitary group (0 being
proportional to the identity) satisfying the relations
a b D 2.dabc C if abc /c ; tra b D 2ı ab : (5.105)
A theory invariant under the global chiral transformation defined with the unitary
transformations UL=R ,
M.x/ ! UL .n/M.x/UR .n/; (5.106)
can be constructed with the help of the independent local chiral invariants
Il D tr.M .x/M.x//l ; l D 1; 2; : : : n 1: (5.107)
In view of the requirement of perturbative renormalization in the weak coupling

regime, one truncates the invariants to be included into the Lagrangian density at
l D 2. The following Lagrangian density displaying explicitly the dependence on
the bare couplings will be investigated below at large n in Euclidean space-time:
1
LD tr.@m M .x/@m M.x// C m2B trM .x/M.x/
2
g1B 2 g2B
C 2 trM .x/M.x/ C tr.M .x/M.x//2 : (5.108)
n n
The determinant term standing for the UA .1/ anomaly in the case n D 3 (see (4.1))
becomes irrelevant for n ! 1. One can therefore discard it. It is convenient for
the discussion below to write this Lagrangian explicitly in terms of the component
fields:
1 a p
LD Œs .x/. C m2B /sa .x/ C a .x/. C m2B / a .x/ 2n2 hs0
2
g1B 2 g2B a
C 2 .sa .x//2 C . a .x//2 C .U .x//2 ; (5.109)
4n 2n
where
1 abc b
U a .x/ D d .s .x/sc .x/ C b .x/ c / f abc sb .x/ c .x/: (5.110)
2
5.6.1 Auxiliary Fields and Integration over Large Multiplicity

Fields
The nonlinearity in the meson fields can be bypassed by introducing two auxiliary
fields:
r r 2
1 g1B a 2 a 2
2 1 a g2B a
L D X .s / C . / Y U : (5.111)
2 2n2 2 n
At the same time, one applies a symmetry-breaking shift

p p p
s0 ! s0 C 2n2 v; U a ! U a C 2 nvsa C n 2nv 2 ı a0 : (5.112)
This shift corresponds to the expected spontaneous chiral symmetry-breaking

pattern of the UL .n/ UR .n/ symmetry.
The resulting shifted Lagrangian is the starting point for integrating over the
O.n2 / multiplicity fields a ; sa ; Y a , which will produce the LO effective potential
of the theory:
r
1 a g1B a
LD s .x/ C m2B C X s .x/
2 2n2
r
g1B p
C a .x/ C m2B C X 2
a
.x/ C m2B n2 v 2 C m2B 2n2 s0 v
2n
r
1 g2B a p p
.X 2 C .Y a /2 / C Y a U C 2 g2B Y a sa v C Y 0 2g2B nv 2
2 n
p p p
C2s0 X g1B v C X 2n2 g1B v 2 2n2 hs0 2n2 hv: (5.113)
The saddle-point approximation applied to the O.N/ model can be attempted here as
well, but as we shall see shortly, the nonlinear interaction induced in the Y a multiplet
places an obstacle in the way of finding the exact LO effective potential.
Since this action contains also a term linear in the pion fields in addition to the
quadratic piece,

p r
1 b 2g1B bc g2B a abc
.x/ C m2B C X ı C Y d ı.x y/ c .y/
2 n n
r
g2B a abc b
Y f s .x/ c .x/; (5.114)
n
one can apply the formal rules for the Gaussian functional integration by completing
this expression into a perfect square. As a result, one obtains the following
contributions to the effective action:
1
Tr log.D1
/ .x y/
bc
2
Z Z
g2B
sn .x/ f mnb Y m .x/Dbc
.x y/f Y .y/ sq .y/;
pqc p
(5.115)
2n x y
where
.D1
/ .x y/
bc
p r
2g1B bc g2B a abc
D C m2B C X ı C Y d ı.x y/: (5.116)
n n
The next step is the sa -integration (a ¤ 0), in which the last term of (5.115)
should also be included. Here one notes again that (5.113) contains a linear piece
p
(2 g2B vY a sa ). Therefore, the structure of the result is similar to the result of the
-integration:
Z Z
1
Tr log.D1 / bc
.x y/ 4g 2B v 2
Y n
.x/D nq
.x y/Y q
.y/ ; (5.117)
2 s
x y
s
with
g2B mnb m
.D1 nq 1 nq
s / .xy/ D .D / .xy/ f Y .x/Dbc
.x y/f Y .y/ :
pqc p
(5.118)
2n
The last step, the integration over the auxiliary multiplet Y a , cannot be performed
exactly, since the a - and sa -integrations induce essential nonlinearities into its
effective action. This is why no exact solution can be derived for this model in the
large-n limit. The best one can do is to expand the effective action up to quadratic
terms in Y a and take the corresponding one-loop result as an approximation.
Before proceeding with the expansion, we separate from X and Y 0 a homoge-
neous constant piece and form with them a universal basic propagator:
p p
2g1B 2g2B 0
D1
M .x
2
y/ D C mB C XC Y ı.x y/
n n
2

C M ı.x y/: (5.119)
Expanding the expressions of the “tracelogs” around this massive inverse propaga-
tor, one obtains the following contributions:
ı2 g2B
Tr log.D1
/D DM .z z0 /DM .z0 z/ı ef ; (5.120)
ıY e .y/Y f .z0 / 2
where one has to exploit the algebraic relation
2 1
nX
dfbc decb D n.1 C ı e0 /ı ef : (5.121)
0
Similarly, one obtains
ı2
Tr log.D1
s //
ıY e .y/Y f .z0 /
g2B
D DM .z z0 /DM .z0 z/ dfbc decb C 2f ebc f fbc
2n
3g2B
D DM .z z0 /DM .z0 z/ı ef ; (5.122)
2
where one exploits
2 1
nX 2 1
nX
2
fbc ecb
d d D n 1 C 1 2 ı e0 ı ef ; f ebc f fbc D n.1 ı e0 /ı ef :
1
n 1
(5.123)
The resulting quadratic action is of the following form:
Z Z nX
2 1
1
Y a .x/ Œı.x y/ C 2g2B DM .x y/DM .y x/
2 x y aD1

C4g2B v 2 DM .x y/ Y a .y/: (5.124)
After the approximate (Gaussian) Y a -integral is performed, one obtains the

following estimate for the n2 part of the effective potential per unit four-
dimensional volume V .LO/ D .LO/ =.V4 n2 /:
1 2 1 2
V .LO/ Œv; x; y0 M 2 v 2 2hv x y
4g1B 4g2B 0
Z
1 k2 C M 2
C log 2
2 k k C M02

.k2 C M 2 /.1 C 2g2B BM .k// C 4g2B v 2
C log ; (5.125)
.k2 C M02 /.1 C 2g2B B0 .k//
where one has introduced the rescaled quantities

p p
2g1B 2g2B 0
xD X; y0 D Y ; (5.126)
n n
and also
Z Z
BM .k/ D DM .p/DM .p C k/: B0 .k/ D D0 .p/D0 .p C k/; (5.127)
p p
with DM .k/ D .k2 C M 2 /1 . Using the rescaled mean field condensates, one arrives
at the following simple expression for the common effective mass scale:
M 2 D m2B C x C y0 : (5.128)
The introduction of the arbitrary constant denominators in the arguments of

logarithms realizes the subtraction of the quartic divergences. It achieves that only
quadratic and logarithmic divergences remain in the expression of the unrenormal-
ized potential. The separation and subtraction of the remaining divergences will
be presented in the next subsection after an elaborate discussion of the approximate
nature of the proposed approximate large-n solution with the help of the exact large-
n leading-order Dyson–Schwinger equations of the system.
5.6.2 Interpretation and Renormalization of V .LO/ .v; x; y0 /
The leading-order effective potential of the homogeneous background condensates

displayed in (5.125) can be interpreted with the help of the Dyson–Schwinger
equations obtained at leading order in the large-n expansion following the standard
procedure. The full set has been derived in [12] on the assumption that the necessary
3-point functions are given by the tree-level expressions obtained from (5.113). It
is sufficient here to consider the coupled set of the 2-point equations governing the
fields a ; Y a ; sa , a ¤ 0. The propagators are diagonal in the flavor index:
G a b .x y/ D G .x y/ı ab ; GY a Y b .x y/ D GY .x y/ı ab ;
GY a sb .x y/ D GYs .x y/ı ab ; Gsa sb .x y/ D Gs .x y/ı ab ; (5.129)
and their equations read as follows:
G1 1 2
.x y/ D Gs .x y/ D . C M /ı.x y/
g2B ŒGY .x y/ .G .y x/ C Gs .y x//

CGYs .y x/GYs .x y/ ;
G1
Y .x y/ D ı.x y/ g2B Gs .y x/G .x y/
g2B
ŒG .x y/G .y x/
2
CGs .x y/Gs .y x/ ;
1 p
GYs .x y/ D 2 g2B vı.x y/
g2B GYs .x y/ ŒGs .y x/ G .y x/ : (5.130)
The sequence of functional integrations presented in the previous subsection

corresponds to a particular approximate solution of the set of DS equations:
G1 1 2 1
.x y/ D Gs .x y/ D . C M /ı.x y/ DM .x y/;
G1
Y .x y/ D ı.x y/ 2g2B DM .y x/DM .x y/;
p
G1
Ys .x y/ D 2 g2B vı.x y/: (5.131)
Comparing the exact and the approximate equations, it is clear that the approximate
nature of the latter equations comes from neglecting the “back-reaction” of the Y a
field on the scalar and pseudoscalar propagators and also on themselves.
One can build a 2PI-like theory in which the 2-point functions of the
a ; Y a ; sa ; a ¤ 0 fields appear as dynamical variables in addition to v; x; y0 ,
and this leads to the set (5.131):

1 2PI 2 2 1 x2 y20
V 2 D m v 2hv C C C .x C y0 /v 2
n2 n 4 g1 g2
1h i
C Tr log G1 1 1 1
C DM G C Tr log G.s;Y/ C D.s;Y/ G.s;Y/
2
1
C 2 Vct2PI Œx; y0 ; G ; G.s;Y/ : (5.132)
n
Here we use the renormalized couplings and collect the contributions containing
the countercouplings in Vct2PI . The counterterm potential should compensate the
quadratic and logarithmic divergences showing up when the expressions in the
second line are evaluated.
After taking the variation of this potential functional with respect to G and the
propagator matrix of the .s; Y/-sector G.s;Y/ , one obtains
G1 1
.k/ D DM .k/;
!
.0;2/ p
1 1 1 2g2 .BMF
.k/ C Id / 2 g2 v
G.s;Y/ .k/ D D.s;Y/ .k/ D p : (5.133)
2 g2 v k2 C M 2
On substituting these expressions back into Vn2PI

2 , one recovers the integral contribu-
tions to (5.125).
The cancellation of all divergences of the effective potential with the help of
appropriate counterterms automatically ensures the finiteness of the equation of
state and of the saddle point equations (their formal expressions above should be
complemented with the corresponding derivatives of the counterterm potential).
The only nontrivial task in (5.125) is to separate the divergences of the term
Z
1 .k2 C M 2 /.1 C 2g2B BM .k// C 4g2B v 2
log
2 k .k2 C M02 /.1 C 2g2B B0 .k//
Z
1 M 2 M02 2g2B .BM .k/ B0 .k//
D log 1 C C
2 k k2 C M02 1 C 2g2B B0 .k/
!
4g2B v 2 2g2B .BM .k/ B0 .k//.M 2 M02 /
C 2 C : (5.134)
.k C M02 /.1 C 2g2B B0 .k// .k2 C M02 /.1 C 2g2B B0 .k//
The expansion of the logarithm up to quadratic terms contributes to the divergences

of the integral. The relevant terms are the following, after some minor algebraic
simplification:
Z(
1 M 2 M02 2g2B .BM .k/ B0 .k// 4g2B v 2
C C
2 k k2 C M02 1 C 2g2B B0 .k/ .k2 C M02 /.1 C 2g2B B0 .k//
" 2 2
1 M 2 M02 2g2B .BM .k/ B0 .k//
C
2 k2 C M02 1 C 2g2B B0 .k/
2
4g2B v 2
C 2 2
.k C M0 /.1 C 2g2B B0 .k//
#)
8g2B v 2 .M 2 M02 / 16g22B v 2 .BM .k/ B0 .k//
C 2 C : (5.135)
.k C m20 /2 .1 C 2g2B B0 .k// .k2 C M02 /.1 C 2g2B B0 .k//2
In order to find the divergent contributions explicitly, one has to expand B.k/
BM .k/ B0 .k/ with k4 accuracy:
B.k/ D 2.M 2 M02 /D0 B0F .k/

.0;1/ .0;2/ 1
C T Id C .M 2 M02 /Id .M 2
M 2
0 / D0
4 2

2 2 .0;1/ 2 2 .0;2/ 1
2.M M0 / T Id C .M M0 /Id C .M C 11M0 / D20
2 2
32 2
2.M 4 C M 2 M02 2M04 /D20 B0F .k/: (5.136)
The divergent pieces of all integrals appearing in (5.135) can be obtained with some
straightforward but somewhat tedious algebra. Divergent terms appearing in some
integrands can be absorbed into gN 2 , defined with the relation
2g2B 2Ng2
g .k/ D D : (5.137)
1 C 2g2B B0 .k/ 1 C 2Ng2 B0F .k/
Note that gN 2 is not necessarily finite; it just helps to define the divergent functions
occurring in the subtractions in full analogy with the O.N/ case. Below, it
follows the list of the divergent integrals expressed with the help of the standard
functions (5.75), in which the replacement =6 ! 2Ng2 has to be done:
Z Z
M 2 M02 ˇˇ .0;1/
ˇ
ˇ .1;1/
ˇ D .M 2 M02 /Id ; 2v 2 g .k/D0 .k/ˇ D 2v 2 Id ;
k k2 C M02 div div
(5.138)
Z ˇ
ˇ 2 2 .0;1/ 1 .1;1/
g .k/.BM .k/ B0 .k//ˇ D 2.M M0 / Id I
k div 2Ng2 d

2 2 .2;I/ 1 .2;2/
2.M M0 / Ia C I
2Ng2 a

.0;1/ .0;2/ 1
T Id C .M 2 M02 /Id C .M 2
C 11M 2
0 /
32 2

4 2 2 4 .0;2/ 1 .2;I/ 1 .2;2/
2.M C M M0 2M0 / Id I I
2Ng2 a .2Ng2 /2 a

.0;1/ 2 2 .0;2/ 1
C T Id C .M M0 /Id .M M0 / Ia.1;1/ ; (5.139)
2 2
4 2
Z
1 .M 2 M02 /2 ˇˇ 1 .0;2/
2 2 ˇdiv
D .M 2 M02 /2 Id ;
2 2
k .k C M0 / 2
Z ˇ
4 ˇ
2v 2g .k/D20 .k/ˇ D 2v 4 Ia.2;2/ ; (5.140)
div
Z ˇ
1 ˇ
2g .k/.BM .k/ B0 .k//2 ˇ
2 k div

.0;2/ 1 1
D 2.M 2 M02 / Id Ia.2;I/ I .2;2/
gN 2 .2Ng2 /2 a

2 2 .0;1/ 2 2 .0;2/ 1
2.M M0 / T Id C .M M0 /Id .M M0 / Ia.2;I/
2 2
4 2
2
1 .0;1/ .0;2/ 1
T Id C .M 2 M02 /Id .M 2
M0
2
/ Ia.2;2/ ; (5.141)
2 4 2
Z ˇ
ˇ 1
2v 2 .M 2 M02 / g .k/D20 .k/ˇ D 2v 2 .M 2 M02 / Ia.2;I/ C Ia.2;2/ ;
k div 2Ng2
(5.142)
Z ˇ
2 ˇ
2v 2g .k/D0 .k/.BM .k/ B0 .k//ˇ D 4v 2 .M 2 M02 /Ia.2;I/ ;
k div

.0;1/ .0;2/ 1
2v 2 T Id C .M 2 M02 /Id .M 2
M0
2
/ Ia.2;2/ : (5.143)
4 2
On collecting the various combinations of the dynamical variables occurring in the

above singular integrals, one obtains for the counterterm potential a parameteriza-
tion with the following nonzero countercouplings:
1 2
Vct2PI =n2 D ım20 v 2 C ıx x C ıy y20 C ıxy xy0 C v 2 .ıyv y0 C ıxv x C ıv4 v 4 /
4
C .ım2 C ıx x C ıy y0 /T C ı8 T2 : (5.144)
It is worth remarking that in this expression, there are other counterterm contribu-
tions besides those that have nonzero renormalized coefficients. For instance, one
obtains a counterterm proportional to the “figure-8” Feynman diagram, the squared
pion tadpole. The divergent coefficients of x2 and y20 determine the bare couplings
corresponding to 1=g1 and 1=g2 , respectively.
The corresponding approximate solution for v; x; y0 is simply given by the
extremal point of V 2PI . All equations are made automatically finite by the contri-
bution of the derivatives of the counterpotential:
Z
1 dVn2PI
2 DM .k/ 1 dVct2PI
D 0 D 2.M 2 v h/ C 4Ng2 v C ;
n2 dv k 1C 2Ng2 BM
F
.k/ C 4Ng2 v 2 DM .k/ n2 dv
(5.145)
Z
1 dVn2PI
2 1 2 1
D 0 D x C v C
n2 dx 2g1 k k 2 C M2
Z
1 2 2 dBM .k/ 1 dV 2PI
C 4N
g 2 v D M .k/ C 2N
g 2 2
C 2 ct ;
k 1 C 2N
g2 BM .k/ C 4Ng2 v DM .k/
F 2 dM n dx
(5.146)
Z
1 dVn2PI
2 1 2 1
D 0 D y0 C v C
2
n dy0 2g2 k k CM
2 2
Z
1 2 2 dBM .k/
C 4Ng2 v DM .k/ C 2Ng2
k 1 C 2Ng 2 BMF
.k/ C 4Ng2 v 2 DM .k/ dM 2
1 dVct2PI
C : (5.147)
n2 dy
By the structure of the counterterm contributions, one recognizes that
g 2 x D g 1 y0 : (5.148)
5.6.3 Summary Conclusions of the Analysis
The lengthy, rather technical analysis of this section followed the steps of the suc-
cessful large-N solution of the O.N/ symmetric meson model. The key difference
is that in addition to the scalar auxiliary field X, one has an auxiliary multiplet
Y a . Although one can perform the sa and a integrations, the resulting effective
action for the Y a field is not quadratic. Therefore, a Gaussian integration over it
does not represent the exact solution in the n ! 1 limit. It is just an approximate
solution. A nonzero X-background would lead only to the breakdown of the O.2n2 /
symmetry [13]. Therefore, it is tempting to search for other possible symmetry-
breaking patterns. An analysis performed in the above approximation revealed
several apparently new minima, which, however, turned out to be related by unitary
transformations to the pure X-background [14].
5.7 The Quark–Meson Theory at Large Nf
The effective theory approach introduces independent fields for every excitation
of the ground state that can be considered stable at the characteristic time scale
of the investigated phenomena. Certainly, with the change of the energy scale, the
weight of the presence of different excitations changes. In QCD, at short distances
the quark degrees of freedom dominate, while at large spatial scales, only meson
excitations are observable. The study of the dynamical change in the relative weight
is an interesting research direction that can be realized, for instance, using functional
renormalization group equations [15].
In particular, at finite temperature, the entities contributing independently to the
thermodynamic potential should be considered independent thermodynamic degrees
of freedom. At those scales where both quarks and their mesonic composites are
active, it is fully legitimate to introduce an effective model in which quarks and
5.7 The Quark–Meson Theory at Large Nf 135
mesons are treated on an equal footing. This question will be further elaborated in
Chap. 7.
In the present section, the large-N renormalization features of the combined
quark–meson model are investigated, which is a necessary precondition for the
consistent treatment at finite temperature/density. The quark–meson theory is a
completion of the O.N/ model (5.1) with a Yukawa interaction describing the
interaction of Nf quark flavors with the meson fields. The meson fields originally are
components of the matrix field M transforming in the fundamental representation
of the UL .Nf / UR .Nf / group. For the study of the quark–meson dynamics, only
the Nf2 1 light pion fields ( a ) and the scalar order parameter field () are kept,
which together form a vector of dimension N D Nf2 . The O.N/-invariant dynamics
of the meson fields after a symmetry-breaking shift along the -direction has the
form (5.3). It is completed by the quark–meson Yukawa interaction:

g p
LY D qN i .x/ @ı
/ ij C p .x/ıij C i5 2Nf Tij .x/ qj .x/; i; j D 1; : : : ; Nf :
a a
N
(5.149)
The notation T a , a D 1; : : : ; Nf2 1, is introduced for the generators of the SUV .Nf /
group. The resulting model is a generalization of the Nf D 2 pion-nucleon theory of
Gell-Mann and Lévy, except that the proton and the neutron are replaced by quarks.
One has to emphasize that this hybrid construction violates both the O.N/ and the
original UL .Nf / UR .Nf / symmetry of the model, which will be reflected in the
counterterm structure
p generated in the renormalization process.
The factor 1= N is introduced into (5.149) in such a way that the quark mass
generated by the symmetry-breaking shift (5.2) remains finite .O.N 0 //:
mq D gv: (5.150)
In phenomenological applications, this mass is identified with the constituent mass

of the additive quark model, nearly equal to one-third of the nucleon mass. Since the
value of the mass parameter of the light quarks in the QCD Lagrangian is 5 MeV,
one can convincingly argue for an approximation in which the constituent quark
mass is fully generated by the chiral symmetry-breaking. This would not be the
case for the pion–nucleon effective model.
The integration over the quark fields can be performed, and the following
additional term is generated to the mesonic action from the quantum fluctuations
of the quarks:
h p i
SM D Nc Tr log @/ C mq ıij C g .x/ıij C i5 2Nf Tija a .x/ : (5.151)
Here the factor Nc reflects the color degeneracy of the quarks and the “color
blindness” of the quark–meson interaction.
With one functional derivation with respect to .x/, one obtains the additional
correction to the equation of state (5.18):

p h gNc Nf
0D Nv m2 C .v 2 C G .x; x// p Tr./@ C mq /1 : (5.152)
6 v N
The factor Nf in the numerator appears from the trace over the quark flavor p
indices. The fermion correction is O.N 0 /. It is suppressed by a factor 1= N
relative to the LO mesonic part, but it dominates over the next order in the
mesonic contribution. Therefore, the combination of the LO mesonic and quark
contributions is a consistent treatment of the large-N hierarchy of the quark–meson
model.
The fermionic tadpole appearing here can be expressed through the tadpole
integral introduced above for the bosonic fields:
.0;1/ .0;2/
Tr./@ C mq /1 D 4mq T .mq / D 4mq .T F .mq / C Id .m2q M02 /Id /;
(5.153)
which requires the following additional pieces in the counterterms:
.1=2/
1 ı 4
ıLM D ım21=2 v 2 C v v 4
2 24
4g2 Nf Nc .0;1/ .0;2/ .1=2/ 24g4Nc Nf .0;2/
ım21=2 D .Id C M02 Id /; ıv4 D Id : (5.154)
N N
The index 1=2 refers explicitly to the order of the large-N expansion in which these
terms are included.
For the pion propagator, a nontrivial wave function renormalization has to be
assumed. The propagator equation completed with the quark-loop contribution
looks like
Z
2 g2 Nc Nf
G1 2
.k/ D Z k C m C
2
v C T .M / trD .DF .q/5 DF .q C k/5 / :
6 N q
(5.155)
In order to arrive at the present form, one performs the trace on the unitary indices
with the help of the normalization of the SU.Nf / generators, trT a T b D ı ab =2, while
the Dirac trace trD is still to be done. For this one uses the explicit representation
im qm C mq
DF .q/ D : (5.156)
q2 C m2q
5.7 The Quark–Meson Theory at Large Nf 137
After working out the Dirac trace, one has

2 4g2 Nc Nf k2
G1 2 2
.k/ D Z k C m C v C T .M / T .mq / B.mq ; k/ ;
6 N 2
(5.157)
where T .mq / and B.mq ; k/ are bosonic tadpole and bubble integrals evaluated with
the quark mass mq . The comparison of the equation of state (5.152) with G .k D 0/
demonstrates the validity of Goldstone’s theorem also at this order:
h
G1 2
.k D 0/ M D : (5.158)
v
This leads to the finite pion-propagator equation

2g2 Nc Nf F
G1
.k/ D k 2
1 C B .k; m q / C M2 ; (5.159)
N
where the finite self-consistent equation determining M follows from (5.157)

evaluated at k2 D 0:
2 4g2 Nc Nf F
M2 D m2R C v C T F .M / T .mq /: (5.160)
6 N
The leading large-N behavior of the -propagator was found in the auxiliary
variable formulation of the model. Now after renormalization, the matrix
element of the inverse propagator matrix of the
sector receives an additional
contribution:

4g2 Nf Nc k2 F
G1
.k/ D D 1
.k/ T .m q / B .k; m q / 2m 2 F
B .k; m q /
N 2 q
8g4 Nc Nf 2 F
D G1
.k/ C v B .k; mq /: (5.161)
N
The counterterm couplings of the Lagrangian ensuring the finiteness of the inverse
propagators are given as
2g2 Nc Nf .0;2/
ıZ;1=2 D ıZ;1=2 D Id ; ıv2 D ıv4 ; ıv2 D 3ıv4 :
N
(5.162)
The spectra of the excitations in the

sector are determined by the determinant
of the 22 matrix of the inverse propagators, evaluated with the modified expression
of G1
given above:

3 1 1 8g4 Nc Nf 2 F
0D C BF .k/ G .k/ C v B .k; mq / C v 2 : (5.163)
2 N
Phenomenological finite temperature and finite density applications of this model

will be discussed in Chap. 7.
References
1. S.-K. Ma, Modern Theory of Critical Phenomena (Benjamin, Reading, 1976)

2. J. Zinn-Justin, Quantum Field Theory and Critical Phenomena, 4th edn. (Clarendon Press,
Oxford, 2004)
3. M. Moshe, J. Zinn-Justin, Quantum field theory in the large N limit: a review. Phys. Rep. 385,
69 (2003)
4. J. Goldstone, Nuovo Cimento 19, 154 (1961)
5. J.-P. Blaizot, E. Iancu, U. Reinosa, Nucl. Phys. A 736, 149 (2004)
6. M. Lüscher, P. Weisz, Nucl. Phys. B 318, 705 (1989)
7. G. Aarts, D. Ahrensmeier, R. Baier, J. Berges, J. Serreau, Phys. Rev. D 66, 045008 (2002)
8. G. Fejős, A. Patkós, Zs. Szép, Phys. Rev. D 80, 025015 (2010)
9. G. Fejős, A. Patkós, Zs. Szép, Phys. Rev. D 90, 016014 (2014)
10. D. Dominici, U.M.B. Marconi, Phys. Lett. B 319, 171 (1993)
11. J. Berges, Sz. Borsányi, U. Reinosa, J. Serreau, Ann. Phys. 320, 344 (2005)
12. G. Fejős, A. Patkós, Phys. Rev. D 82, 045011 (2010)
13. H. Meyer-Ortmanns, B.J. Schaefer, Phys. Rev. D 53, 6586 (1996)
14. G. Fejős, Phys. Rev. D 87, 056006 (2013)
15. D.U. Jungnickel, C. Wetterich, Phys. Rev. D 53, 5142 (1996)
Chapter 6
Dimensional Reduction and Infrared Improved
Treatment of Finite-Temperature Phase
Transitions
The thermal energy kB T introduces a new energy scale into field theories. In
thermal field theories, a specific series of Feynman diagrams of increasing loop
order can be found that reveals the presence of increasing powers of the ratio
kB T=mc2 in the self-energy, dangerous at high temperature in the infrared domain
(T
m). Resummation is invoked to avoid the emerging singularity from this class.
Application of the method of dimensional reduction is motivated by the observation
that it automatically realizes the resummation leading to the replacement of the
renormalized mass by an infrared safe thermal mass of T.
The main content of this chapter is a detailed presentation of the procedure
and features of dimensional reduction. We begin here with a short reminder of the
so-called daisy diagram resummation in thermal perturbation theory so that in the
bulk of this chapter we can compare to it the results of dimensional reduction. The
presentation as it appears in the classical paper of Dolan and Jackiw [1] is followed.
The lowest (one-loop tadpole) contribution to the self-energy of the one-
component self-interacting massive ˚ 4 theory is given by
Z
.1/ X 1
˙T D T ; !n D 2Tn; !p D .m2 C p2 /1=2 ; (6.1)
2 n p !n2 C !p2
where m is the renormalized mass of the field, and is the strength of the self-
coupling (precise notation will be introduced in the next subsection; the units kB D
c D „ D 1 are used). After normalizing this expression, one finds the following
expression for it in terms of an expansion in m=T:

.1R/ T 2 mT
˙T D C O.m2 ln.m2 =T 2 // : (6.2)
2 12 4

DOI 10.1007/978-3-319-22620-0_6
140 6 Dimensional Reduction and Infrared Improved Treatment of...
The two-loop (double tadpole or “double-scoop”) correction contains the same

quantity multiplied by the vertex strength =2 and an integral over the product
of two Euclidean propagators:
Z XZ
.2/ X 1 1
˙T D T 2 2
T 2
2 n
p !n C !p 2 2 2
p .!n0 C !p /
n0
Z !
X 1 d XZ 1
D T T 2
2 n p !n2 C !p2 2 dm2 p !n0 C !p
2
n0

.1/ T
D ˙T : (6.3)
2 8m
.1/
The large ratio multiplying ˙T challenges the validity of the thermal perturbation
theory.
One can continue to put an increasing number of tadpoles along the basic
loop (this class of diagrams is called “daisy”) and find increasingly dangerous
contributions, containing increasingly high powers of T=m. The contribution of the
daisy diagram with l petals is written as
Z !l
.l/ X 1 XZ 1
˙T D T T 2
2 2 n p !n C !p2
2
p .!n0 C !p2 /lC1
n0
Z !l Z
X 1 1 dl X 1
D T T 2
: (6.4)
2 6 n p !n C !p2
2 2
lŠ d.m / l
0 p !n0 C !p2
n
However, the sum of this series represents a translation in the m2 variable:

" Z # Z !
X 1 d X 1
˙T D exp T T : (6.5)
2 n p !n C !p dm2
2 2 2 0n p0 !n0 C !p20
2
In the exponent of the “translation” operator, one can keep the leading contribution
at high T in (6.2), while in the expression on which the m2 -translation acts, both
of the first two terms are retained. In this way, one obtains for the temperature-
dependent resummed mass an expression free of any infrared singular behavior:
1=2
2 T T 2 T
m2 C ˙T m2 C T m2 C m2T mT : (6.6)
24 8 24 8
This expression is the same

p as the one-loop expression (6.2), with mT replacing m.
At large temperature, T defines the infrared-safe effective mass scale. In this
way, the resummation proves essential for avoiding accumulation of infrared-
sensitive self-energy contributions. The “daisy” summation will be shown below
6 Dimensional Reduction and Infrared Improved Treatment of... 141
to arise automatically from the infrared safe integration over the nonzero Matsubara
frequency modes and will serve as the mass parameter of the effective theory of the
static Matsubara fields.
The Matsubara decomposition of the imaginary temporal (“thermal time”)
dependence of finite-temperature quantum fields defined in d spatial dimensions
naturally leads to a separation of the infrared-sensitive light modes of E p
and heavy modes E kB T at sufficiently high temperatures (p T). The
Wilsonian interpretation of the infrared behavior of field theories suggests a strategy
of gradual functional integration for the evaluation of the partition function defining
the thermodynamics of the theory [2]. The primary integration over the heavy modes
in the background of an arbitrary light configuration results in an effective theory
of these modes. The most important feature of this effective theory is a thermal
mass generation for most of the light modes, essentially improving the infrared
convergence features.
At high temperatures, where kB T is much larger than the original characteristic
masses of the theory, this effective theory containing in principle an infinite number
of operators can be constructed perturbatively, for instance in a loop expansion
of increasing loop number. The coupling strength in front of most operators will
be inversely proportional to powers of kB T, as dictated by simple dimensional
considerations. For practical purposes, one keeps those contributions to the coupling
strengths that do not vanish with infinitely growing T (these coincide with the
“perturbatively renormalizable” operators). In a second stage, one applies various
techniques to performing the remaining functional integrations.
In the first part of this chapter, the example of the one-component ˚ 4 theory will
be used for learning the techniques of the determination of the effective reduced
theory [3]. The integration over the heavy modes will be presented with two-loop
accuracy. The main emphasis will be put on presenting how the renormalization
program applied to the cutoff regularized .d C 1/-dimensional theory generates the
cutoff-dependent counterterms for the d-dimensional effective theory. A discussion
of the finite-temperature phase transition occurring in this model will be based on
solving the effective dimensionally reduced model. This treatment goes beyond the
conventional Landau-type mean field approximation.
In the second part, another technique is presented for the determination of the
effective reduced theory. The so-called matching strategy is based on determining
the relationship among the couplings of the original and the effective theories by
requiring coinciding results for a set of n-point functions as obtained from the
.d C 1/-dimensional (full) and the d-dimensional (reduced) theories [4].
In the third part, we shall apply the technique of integrating over the hard
Matsubara modes to the thermodynamics of the soft bosonic sector of the U.1/-
invariant scalar electrodynamics with one-loop accuracy [7]. The corresponding
discussion of the cosmological electroweak phase transition based on the SU.2/
local symmetry group will be presented in Chap. 8. We close the chapter with a
short discussion of the order of the finite-temperature phase transition in a generic
dimensionally reduced theory.
6.1 The ˚ 4 Theory at Finite Temperature
The thermodynamics of the ˚ 4 theory is defined by the Euclidean path integral

Z
ˇF
Ze D D˚.x; /eSE Œ˚ ; (6.7)
˚per
where the field ˚.x; / is defined in the direct product space of a three-dimensional
spatial cube of volume V and the thermal dimension of extension ˇ D .kB T/1 .
Only fields that are periodic along the thermal extension
˚.x; / D ˚.x; C ˇ/ (6.8)
are included in the path integral. The Euclidean action is given as

Z Z " 2 #
ˇ
1 @˚ 1 2 1 C ı
SE Œ˚ D d d3 x C .r˚/ C .m2 C ım2 /˚ 2 C ˚4 :
0 2 @ 2 2 24
(6.9)
Here the bare mass and self-interaction parameters are written as the sum of the
renormalized couplings and the countercouplings. The wave function renormaliza-
tion is neglected for simplicity. The Fourier expansion along the compact thermal
dimension is called Matsubara expansion:
1
X 2
˚.x; / D ˚n .x/ei!n ; !n D n: (6.10)
nD1
ˇ
It will be clear soon that these are the static (!0 D 0) modes, which are most
sensitive to the infrared fluctuations. Therefore, it is convenient to divide the
functional integration into two stages using the notation
Z ˇ p X
˚.x; / D ˚0 .x/ C '.x; /; d'.x; / D 0; '.x; / D ˇ 'n .x/ei!n :
0
n¤0
(6.11)
Here a special convention is used for the Fourier expansion in the variable . With
this, one can rewrite (6.7) as
Z
Z
.2/ Œ'S Œ˚ ;'
ZD D˚0 .x/eS0 Œ˚0 D'.x; /eS I 0 0
: (6.12)
6.1 The ˚ 4 Theory at Finite Temperature 143
The original Euclidean action splits into a sum of the three terms indicated above:
Z
1 2 1 C ı 4
S0 Œ˚0 D ˇ d3 x
.m C ım2 /˚02 C .r˚0 /2 C ˚0 ; (6.13)
2 2 24
X Z
.2/ 1 2 3 2 2 2
S Œ' D ˇ d x'n .x/ 4 C m C ˚0 .x/ C !n 'n .x/ (6.14)
2 2
n¤0
and
Z
1 2 ı 2
SI Œ˚0 ; ' D ˇ d 3 x ım2 C ˚0 .x/ 'n .x/'n .x/
2 24
Z X
C ı
C ˇ3 d3 x ı.n1 C n2 C n3 C n4 /'n1 .x/'n2 .x/'n3 .x/'n4 .x/
24
n1 ;n2 ;n3 ;n4 ¤0
Z X
C ı
C ˇ 5=2 d 3 x˚0 .x/ ı.n1 C n2 C n3 /'n1 .x/'n2 .x/'n3 .x/: (6.15)
6
n1 ;n2 ;n3 ¤0
In this representation of the canonical partition function, the thermal dimension

“disappears,” and in its place, infinitely many three-dimensional fields .'n .x//
emerge. Those with index n ¤ 0 have in the kernel of the quadratic part of the
action the term !n2 , which dominates m2 at high temperatures. This “squared thermal
mass” contribution is missing for ˚0 , which makes these static configurations the
most vulnerable in the infrared. For this reason, one performs the path integration
first in the curly brackets of (6.12) with an arbitrary background ˚0 .x/.
For the perturbative integration, one needs the propagator of the nonstatic modes,
which is the functional inverse of the kernel defining S.2/:
1
2 2 2 2
Dn .x y/ D ˇ 4x C m C !n C ˚0 .x/ ı.x y/ ; (6.16)
2
which will be expanded into a power series of ˇ 2 .m2 C =2˚02 / ˇ 2 M 2 .x/:

h 2 2
Dn .x y/ D D.0/
n .x z 1 / ı.z1 y/ D.0/n .z1 y/ ˇ .m C =2˚0 .y//
2
2 2
CD.0/
n .z1 z2 / ˇ .m C =2˚0 .z2 /
2
2 2 i
D.0/
n .z 2 y/ ˇ .m C =2˚ 2
0 .y/ C : : : ;
Z
1
D.0/ .x y/ D eip.xy/ D.0/ .0/
n .p/; Dn .p/ D 2 : (6.17)
n
p ˇ .!n C p2 /
2
6.2 Two-Loop Integration Over the Nonstatic Fields
The perturbative evaluation of the functional integral over the nonstatic modes is
done by expanding exp.SI / into a Taylor series and performing the integrals over
each term weighted by the Gaussian exp.S.2/ /. The exponent of the partition
function, e.g., the free energy of the system, is perturbatively determined by the
contributions from the one-particle-irreducible diagrams (see Sect. 2.7!). To two-
loop order, one has
Z
1X
ˇF D ln D˚eSE D S0 Trx log D1
n .x; x/
2
n¤0
Z X
1 ı 2
ˇ2 d3 x h'n .x/'n .x/i ım2 C ˚0 .x/
2 24
n¤0
Z X

ˇ 3 d3 x ı.n1 C n2 C n3 C n4 /h'n1 .x/'n2 .x/ih'n3 .x/'n4 .x/i
8
n1 ;n2 ;n3 ;n4 ¤0
2 Z Z XX
5 3
Cˇ d x d3 y˚0 .x/˚0 .y/ ı.n1 C n2 C n3 /ı.m1 C m2 C m3 /
12
ni ¤0 mi ¤0
h'n1 .x/'m1 .y/ih'n2 .x/'m2 .y/ih'n3 .x/'m3 .y/i: (6.18)
Here the propagators defined with the help of S.2/ are defined as
Z
1 .2/
h'n .x/'m .y/i D eS 'n .x/'m .y/ D Dn .x y/ınm : (6.19)
Z .2/
Recognizing that the propagators are diagonal in the Matsubara index, one obtains
the following simplified expression:
1 X
F D ˇ 1 S0 C Trx log D1
n .x; x/
2ˇ
n¤0
Z X
1 3 2 ı 2
C ˇ d x Dn .x; x/ ım C ˚ .x/
2 24 0
n¤0
Z X

C ˇ2 d3 x Dn1 .x; x/Dn2 .x; x/
8
n1 ;n2 ¤0
Z Z
2
ˇ4 d 3 x d3 y˚0 .x/˚0 .y/
12
X
ı.n1 C n2 C n3 /Dn1 .x y/Dn2 .x y/Dn3 .x y/: (6.20)
n1 ;n2 ;n3 ¤0
6.2 Two-Loop Integration Over the Nonstatic Fields 145
One can reproduce the conventional perturbation theory in a ˚0 .x/ background by

applying in (6.20) the expansion (6.17). Then the “tracelog” is rewritten in an infinite
series of momentum integrals:
( Z
1 X 1 TX
Trx log Dn .x; x/ D V log.ˇ 2 !n2 C ˇ 2 p2 /
2ˇ 2 p
n¤0 n¤0
Z " #) ˇ
X1
.1/lC1 3
ˇ 2 Q2
.p / ˇ
Vd p j M j p jC1 ˇ
‘C ˘jD1
l
ˇ ;
l 3
.2/ ˇ .!n C pj /
2 2 2 ˇ
lD1 plC1 Dp1
Z
Q 2 1
M .p/ D d 3 xeipx M 2 .x/: (6.21)
V
R
The l D 1 term is simply proportional to d3 x˚02 .x/. Therefore, it contributes to the
mass term of the reduced theory. Below, we shall discuss the interest of the nonlocal
features of the l D 2 term, but for the moment, this expression plus the other terms
appearing in F will be discussed assuming a constant-˚0 background. Then the
free energy accounts exclusively for the potential energy density of constant fields.
When ˚0 is a constant, the contributions of (6.20) can be expressed in terms of
the following fundamental sums and integrals:
XZ
I.M/ D T log ˇ 2 .!n2 C p2 C M 2 /
p
n¤0
Z X
K.M; M; M/ D ˇ 4 d ı.n1 C n2 C n3 /Dn1 ./Dn2 ./Dn3 ./
n1 ;n2 ;n3 ¤0
X
D T2 ı.n1 C n2 C n3 /
n1 ;n2 ;n3 ¤0
Z Z Z
.2/3 ı .3/ .p1 C p2 C p3 /
: (6.22)
p1 p2 p3 .!n21 C p21 C M 2 /.!n22 C p22 C M 2 /.!n23 C p23 C M 2 /
With these abbreviations, (6.20) is expressed as

h1
2 C ı 4 1 2 1 ı 2 0
F D V .m C ım C /˚02
˚0 C I.M/ C 2
ım C ˚ I .M/
2 24 2 2 24 0
0 2 2 i
C I .M/ ˚02 K.M; M; M/ ; (6.23)
8 12
(I 0 .M/ D dI=dM 2 ). The integrals can be expanded in powers of M 2 (which is

equivalent to an expansion in ˚02 ), since the heavy modes make the Matsubara
summation finite. For the analysis of the effective potential of the reduced theory
up to terms ˚04 below, we need the expansion of I.M/ up to O.M 6 /:
I.M/ D I0 C I1 M 2 C I2 M 4 C I3 M 6 C : : : ;
.2/ .1/ .0/ .ln/ .0/
I1 D I1 2 C I1 T C I1 T 2 ; I2 D I2 ln C I2 ;
T
.0/
I3
I3 D ; (6.24)
T2
with the coefficients explicitly given in Eq. (C.2) of Appendix C. The coefficient I0 is
an unimportant constant. Therefore, one can get rid of it by redefining the zero point
of the effective potential of the reduced theory appropriately. A similar expansion of
K.M; M; M/ is needed to O.M 2 /, and one has by dimensional analysis the following
terms:
K.M; M; M/ D K0 C K1 M 2 C : : : ;
.2/ .1/ .ln/ .0/
K0 D K0 2 C K0 T C K0 T 2 ln C K0 T 2 ;
T

.2ln/ 2 .ln/ .0/
K1 D K1 ln C K1 ln C K1 : (6.25)
T T
The values of the expansion coefficients are given in Eq. (C.6) of Appendix C. In
order to separate the dependence of the potential on the cutoff and the temperature,
one introduces the renormalization scale . The renormalized potential is obtained
by fixing its value at the origin to zero (simply throwing away all ˚0 -independent
terms) and fixing the first two nonzero derivatives with respect to ˚0 :
d2 F ˇˇ d4 F ˇˇ
ˇ D Vm2 ; ˇ D V: (6.26)
d˚02 TD˚0 D0 d˚04 TD˚0 D0
These conditions determine the countercouplings, which might contain specific

finite terms beyond the cutoff-dependent parts. The finite pieces that correspond to
the choice of a specific renormalization scheme are denoted below by m2.l/ ; .l/ ,
where the index l refers to the loop order at which the corresponding quantity
is defined. Their explicit expressions will not be given, since they do not give
any further insight. The countercouplings ensuring the ultraviolet finiteness of the
effective potential at T D 0 are found with one- and two-loop accurate reduction
and are written here separately:

1 .2/ 2 .ln/
ım2.1/ D m2.1/ I1 C m2 I2 ln ;
2

1 .2/ 1 .2/ .2/ .ln/
ım2.2/ D m2.2/ 2 .1/ I1 K0 2 I1 I2 2 ln
2 6
6.2 Two-Loop Integration Over the Nonstatic Fields 147

2 .ln/ .ln/ 2 .ln/ .0/ 2 2 1 .ln/ 2 2
m.1/ I2 C .1/ I2 m 2I2 I2 m K1 m ln
6
2
.ln/
C 2.I2 /2 2 m2 ln ; (6.27)

.ln/
ı.1/ D .1/ 32 I2 ln ;

2
.ln/ 2 3
ı.2/ D .2/ C 9.I2 /

.ln/ .ln/ .0/ 3 .ln/ 3
6.1/ I2 12I2 I2 K1 ln : (6.28)

The resulting temperature-dependent pieces still contain cutoff-dependent terms
whose coefficients vanish for T D 0. These are the induced counterterms of the
reduced three-dimensional effective theory. In order to simplify the expression, it
is convenient to choose the normalization scale as D T. This choice, however,
does not mean that the counterterm operators will depend on the temperature (see
the discussion in Sect. 3.5.1). This leads to
1 1
F D ˚ 4
V 24 0

1 1 .0/ .0/ .0/ 1 .0/ .0/ .ln/ 1 .ln/
C ˚02 m2 C T 2 I1 2 I1 I2 C K0 C I1 I2 C K0 ln
2 2 6 6 T

1 .1/ .1/ .0/ 1 .1/ .1/ .ln/
C T I1 2 I1 I2 C K0 C I1 I2 ln : (6.29)
2 6 T
The superrenormalizable nature of the effective theory in the present approximation
is manifested by the fact that counterterms were induced only to the ˚02 term of the
potential.
Let us return to the l D 2 term of (6.21), which defines a momentum-dependent
contribution the Lagrangian density of the reduced theory containing four powers
of ˚0 . The nonlocal part of this term is absent in the original formulation of the
three-dimensional theory, which follows the same pattern as the four-dimensional
one. We shall discuss in the next section to what extent it is nonnegligible for the
consistency of the reduced effective model.
The full quartic contribution is written with the help of the nonstatic part of the
bubble integral B in terms of the Fourier transform of ˚02 .x/ as
Z XZ
2 1
˚Q 02 .q/˚Q 02 .q/Bns .q/; Bns .q/ D T 2
:
16 q p .!n C p2 /.!n2 C .p C q/2 /
n¤0
(6.30)
The q D 0 finite part of B is already included in the local reduced quartic term
.0/
as 2I2 . For dimensional reasons, the remaining q-dependent part depends only on
the ratios D q=T, x D q= (q D jqj). When performing (partially) the sum-
integrations, one arrives, for the kernel of the strictly nonlocal (q-dependent) part,
through (C.3), at
.0/
˝.q/ Bns .q/ 2I2 : (6.31)
6.3 The Effective Three-Dimensional Theory and Its

Two-Loop-Accurate Effective Potential
The effective theory of the static ˚0 .x/ modes can be formulated in terms of the
appropriately rescaled fields and couplings:
p 1
˚3 .x/ D ˇ˚.x/; 3 D T; ˝3 .q/ D ˝.q/: (6.32)
T
After combining (6.29) and (6.31) and adding to it a three-dimensional kinetic
density, one has for the effective three-dimensional action,
Z
1 3
S3D D d 3 x .r˚3 /2 C .m2T C ı3 m2 /˚32 C ˚34
2 24
Z
1
23 ˚Q 32 .q/˝3 .q/˚Q 32 .q/; (6.33)
16 q
where the thermal mass and the mass counterterm of the three-dimensional theory
are given as

1 .0/ .0/ .0/ 1 .0/
m2T D m2 C T 2 I1 2 I1 I2 C K0 ;
2 6

2 2 .0/ .ln/ 1 .ln/
ı3 m D T ln I I C K0
T 1 2 6

1 .1/ .1/ .0/ 1 .1/ .1/ .ln/
C T I1 2 I1 I2 C K0 C I1 I2 ln : (6.34)
2 6 T
A computation of the effective potential U3D .˚30 / for a constant three-
dimensional background ˚30 that fully contains all two-loop contributions of
the four-dimensional theory has three types of contributions (involving, however,
infrared safe propagators for the static modes). The “classical” potential
1 2 2 3 4
U3D .˚30 /class D mT ˚30 C ˚30 (6.35)
2 24
6.3 The Effective Three-Dimensional Theory and Its Two-Loop-Accurate. . . 149
includes the effect of one- and two-loop diagrams in which all lines correspond to
nonstatic (heavy) modes. The (resummed) two-loop-accurate contributions originat-
ing from diagrams containing exclusively static propagators have the same topology
as their 4D counterparts (6.23), except that here, the couplings and the phase space
of the momentum integrations are three-dimensional:
1 3 0 2 2
U3D .˚30 /static D I3 .mT / C I3 .mT / 3 ˚02 K3 .mT ; mT ; mT /: (6.36)
2 8 12
Finally, the contribution from the nonlocal part of L3D should be taken into
account only at the one-loop level in order to account for the mixed static–nonstatic
two-loop diagrams of the four-dimensional model. In order to find its one-loop
contribution, one has to introduce the appropriate splitting of ˚3 .x/ D ˚30 C '3 .x/
and keep only the quadratic part in '3 .x/. One has to calculate the tracelog of the
quadratic part of the action, which looks like
Z
.2/ 1 3 2 2 ˚ 2
S3D Œ'3 D '3 .q/ q2 C m2T C ˚30 3 30 ˝3 .q/ '3 .q/: (6.37)
2 q 2 2
The resulting tracelog can be expanded in powers of the nonlocal part:
Z
1 2 2 3 2
U3D .˚30 /staticnonstatic D ln q C mT C ˚30
2 q 2
1 Z !l
X .1/lC1 23 ˚30
2
˝3 .q/
C : (6.38)
lD1
2lC1 l q q2 C m2T C 2 ˚30
2
The first term is already included in the static one-loop contribution (D I3 .mT /=2),
and therefore, only the infinite sum should be added to the sum of the three
different types of contributions in order to avoid double counting. In the nonlocal
contribution, all terms are finite but the first one, which contributes to the three-
dimensional mass divergence:
2 Z 2 2
23 ˚30 ˝3 .q/ 23 ˚30 2 mT
2
D 4
ln C ln C C f :
4 2 2
q q C mT C 2 ˚30 64 T T T T T2
(6.39)
The constant C as well as the finite function f .m2T =T 2 / can be determined numeri-
cally.
The integrals figuring in (6.36) can be evaluated analytically:

.1/ 2 3 2 .0/ 2 3 2 3=2
I3 .mT / D 2J mT C ˚30 C 2J mT C ˚30 ;
2 2
2 9.m2T C 3 ˚30
2
=2/
K3 .mT ; mT ; mT / D L.ln/ ln C L.0/
L.ln/
ln ; (6.40)
2 2
with the coefficients L.i/ ; J .i/ taken from Sect. C.2.

The sum of the three-dimensional divergences emerging from the static and the
nonlocal contributions exactly cancel the divergences induced by the integration
over the nonstatic modes. This shows that it is compulsory to include also the
nonlocal piece in the effective three-dimensional action at the two-loop level. One
has again the freedom to conveniently choose also a finite counterterm 3 m2 ,
thereby simplifying the local contributions to the potential. One can arrange the
following form:
1 2 2 3 4
U3D .˚30 / D mT ˚30 C ˚30
2 24

.0/ 2 3 2 3=2 23 2 .ln/ 9.m2T C 3 ˚30
2
=2/
CJ mT C ˚30 C ˚30 L ln 2
2 12
C U3D .˚30 /nonlocal;ren : (6.41)
The last term emerges from (6.38) after subtraction of the divergent terms plus the
first term on the right-hand side, which is already present in the first term of the
second line.
Before developing further the formalism, one can relate the one-loop part of the
potential to the “daisy” resummation of the self-energy reviewed in the introduction
of this chapter. One expands the potential near ˚30 D 0 to quadratic order to find
the mass of the static field. Scaling back to four-dimensional quantities and using
J .0/ D .12/1 , one obtains for U3D D ˇU4D

1 2 2 .0/ 2 3 2 3=2 1 2 T
m ˚ CJ mT C ˚30 const. C ˇ mT mT ˚02 C : : : ;
2 T 30 2 2 8
(6.42)
which exactly reproduces the resummed mass (6.6). Clearly, the higher loop
corrections in the nonstatic integration go beyond the simple “daisy” approximation.
The natural question is whether one could choose the couplings of a local 3D
theory to reproduce the result obtained by including the nonlocal piece. The method
to find an answer to this question is called the matching procedure, and its two
versions will be outlined in the next section.
6.4 Local Coarse-Grained Effective Theory via Matching
Matching of two theories means to require coincidence of a number of n-point

functions for fixed values of their variables when computed simultaneously in both
theories (recall Sect. 3.5). The conditions should be imposed in a region where one
expects both theories to be valid. Since the scale of nonlocality is T, one puts
into the matching conditions momentum values much below this scale. The number
of matching conditions is chosen to provide the necessary number of equations
6.4 Local Coarse-Grained Effective Theory via Matching 151
for the determination of the relations between the two coupling sets. The local
three-dimensional theory that will be matched to the nonlocal theory (6.33) has the
following Lagrangian density:
1 3loc 4
3D D
Lloc .r'3 /2 C .m23loc C ı3loc m2 /'32 C ' : (6.43)
2 24 3
Here m23loc ; 3loc are couplings to be determined in terms of m2T ; 3 , and also there is
a finite rescaling relating the field variables:
˚3 .x/ D 1=2 '3 .x/: (6.44)
The mass counterterm of the local superrenormalizable theory at the two-loop level
can be extracted from the divergences of the local two-loop contributions to U3D
computed in the previous point, just replacing the couplings by those of the local
theory:
23loc .ln/ 2loc

ı3loc m2 D J .1/ 3loc loc C L ln 2 : (6.45)
6
The cutoff loc has to be chosen below the scale T.
The most obvious choice for the matching conditions is to require the coinci-
dence of the 2- and 4-point functions of the two theories at vanishing momentum.
One also imposes a condition relating the kinetic terms:
2nonloc .mT ; 3 ; T; p D 0/ D 2loc .m3loc ; 3loc ; T; p D 0/;

4nonloc .mT ; 3 ; T; p D 0/ D 2 4loc .m3loc ; 3loc ; T; p D 0/;
@ nonloc @
2
2 .mT ; 3 ; T; p D 0/ D 2 2loc .m3loc ; 3loc ; T; p D 0/: (6.46)
@p @p
Since the nonlocality of (6.33) is additive, one can decompose the n-point functions
of the nonlocal theory into the sum of a local and an intrinsically nonlocal
contribution:
2nonloc .mT ; 3 ; T; p D 0/ D 2loc .mT ; 3 ; T; p D 0/ C 2nonloc .mT ; 3 ; T; 0/;

4nonloc .mT ; 3 ; T; p D 0/ D 4loc .mT ; 3 ; T; p D 0/ C 4nonloc .mT ; 3 ; T; 0/;
@ nonloc @
2
2 .mT ; 3 ; T; p D 0/ D 2 2loc .mT ; 3 ; T; p D 0/
@p @p
@
C 2nonloc .mT ; 3 ; T; 0/; (6.47)
@p2
where nonloc denotes the one-loop contribution involving one nonlocal vertex to
the corresponding n-point function. The obtained expressions should be substituted
into the left-hand sides of the previous set of conditions.
This additivity suggests that one looks for linearized relations
D 1 C 23 ; m23loc D m2T C 23 m ; 3loc D 3 C 33 : (6.48)
After substituting these ansatzes on the right-hand side of (6.46), one expands to
linear order in i and solves the resulting set of linear equations.
It is suggestive to apply the matching strategy directly between the original four-
dimensional model and the dimensionally reduced local model [4]. Naturally, one
then applies on the left-hand side of the matching conditions (6.46) derivatives
of .4d/ . The two-point function computed perturbatively in the four-dimensional
theory to two-loop order can be written in a form whereby the contribution of the
purely light (static) modes is separated explicitly:
.4d/
2 .m; ; T; p/ D p2 C m2 C ˙T .p2 / D p2 C m2 C ˙Tnonstatic C ˙Tstatic : (6.49)
In the nonstatic self-energy contribution one includes also mixed diagrams contain-
ing at least two nonstatic propagators. Since these contributions are infrared safe,
one is free to expand the nonstatic self-energy in powers of k2 :
d nonstatic 2 ˇˇ
˙ nonstatic .k2 / D ˙ nonstatic .0/ C ˙ .k /ˇ 2 k2 C O.k4 =T 2 /: (6.50)
dk2 k D0
The renormalized self-energy is calculated with two-loop (O.2 /) accuracy, its

derivatives
p with one loop. The validity of the reduced model is restricted to momenta
k < T. Then the order of magnitude of the omitted terms of the expansion is
estimated as 3 T 2 . The effective coupling of the static model is T. Therefore,
the two-loop-accurate static self-energy is O.2 T 2 /. Taking all these order-of-
magnitude estimates into account, one can approximately rearrange the two-point
function as

.4d/ d
2 .m; ; T; p/ D 1 C 2 ˙ nonstatic .0/
dp

2 2 d nonstatic static 2
p C .m C ˙ nonstatic
.0// 1 2 ˙ .0/ C ˙ .p / : (6.51)
dp
The expansion leading to this form assumes d˙ nonstatic =dp2 O.2 /; ˙ nonstatic
O./ and it allows the following 4d $ 3d identifications:

d nonstatic
˚32
D ˇ 1 C 2˙ .0/ ˚42 ;
dp

2 2 d nonstatic
mT D .m C ˙ nonstatic
.0// 1 2 ˙ .0/ : (6.52)
dp
6.5 Dimensional Reduction of a Gauge Theory at the One-Loop Level 153
A similar calculation of the four-point function at the origin in both models fixes the
self-coupling 3 :
2
.4d/ d nonstatic
4 .m; ; T; p D 0/ D ˇ 1 C 2 ˙ .0/ 4loc .mT ; 3 ; T; p D 0/:
dp
(6.53)
The explicit expressions for i in the case of the (3D nonlocal) $ (3D
local) matching were calculated for the ˚ 4 theory with cutoff regularization in
an optimally chosen renormalization scheme [3]. The rules of the (4D local) $
(3D local) matching were computed for a rather general class of models using
dimensional regularization and MS renormalization [4].
The general constructional rule of the nonlocal three-dimensional theory is to
include in the reduced theory all nonlocal vertex operators that are necessary
to reproduce, within the renormalized perturbation theory of the reduced model,
the contribution of all Feynman diagrams containing mixed (static and nonstatic)
propagators in the full theory. The induced counterterms automatically ensure
the finiteness of the result obtained in the reduced perturbation theory. Clearly,
the number of necessary nonlocal vertices is increasing with the order of the
perturbation theory without any limit.
The matching of two theories both containing a fixed number of local pertur-
batively renormalizable operators can be done by just imposing equality of some
lower n-point functions (and their derivatives) at as many appropriately chosen
values of their argument(s) as there are couplings that appear in the reduced model.
The renormalization procedure is independently implemented in both theories, the
matching results in relating the renormalized couplings. This procedure is easier
to realize and its outcome is more transparent, but there is no guarantee of the
coincidence of any of the n-point functions in an extended range of the arguments
even in the lowest orders of the perturbation theory. Parallel applications to specific
physical problems of the full and the reduced theories give one a chance to evaluate
the accuracy of the matching for physically relevant observables. A nice research
program of outstanding physical interest offering the opportunity for this kind of
comparative study was the investigation of the electroweak symmetry-restoring
phase transition in the 1990s (see Chap. 8).
6.5 Dimensional Reduction of a Gauge Theory

at the One-Loop Level
The origin of the matter–antimatter asymmetry of our cosmic neighborhood is one

of the most intriguing questions in the focus of both high-energy and astrophysical
research. The famous Sakharov conditions [5] for the dynamical generation of
this asymmetry include the necessity of far-from-equilibrium evolutionary period(s)
during the history of the universe. Since the standard model possesses the other two
features (existence of baryon-number-changing processes and CP-violation), it was

an important research direction in the mid-1990s to clarify whether the evolution
based on a joint solution of the equations of the gravitational and quantum field
theories could lead to substantial deviations from thermal equilibrium in the era of
the hot universe.
The most common phenomenon in which thermal equilibrium is temporarily
suppressed is a discontinuous (first-order) phase transition. For this reason, it was
natural to ask whether the Brout–Englert–Higgs symmetry-breaking phenomenon
occurred in the early universe via first-order phase transition. A most economical
way to answer this question was the Monte Carlo investigation of the change in the
ground state of the standard model in the framework of the dimensionally reduced
finite-temperature theory, as initiated by K. Kajantie and collaborators [6].
As preparation for the discussion of the electroweak phase transition in Chap. 8,
we shall present the steps and the peculiarities requiring the special attention of
dimensional reduction in the case of gauge theories. The example to be used here
will be the simplest gauge theory, namely the scalar electrodynamics or the U.1/
symmetric Higgs model.
The model is defined with the Euclidean action
Z ˇ Z
1 1
SU.1/ D d d3 x
Fmn Fmn C j.@m C igAm /˚.x; /j2
0 4 2

1
C m2 j˚j2 C j˚j4 C SU.1/;ct : (6.54)
2 24
The Euclidean continuation of the field-strength tensor Fmn is defined as the

antisymmetric derivative of the vector potential: Fmn D @m An @n Am . The electric
charge of the complex field ˚.x; / is g. All couplings are renormalized and Sc:t:
stands for the counterterms necessary for the cancellation of the four-dimensional
UV-divergences of the renormalized perturbation theory.
The one-loop integration over the nonstatic gauge and scalar modes will be done
in a static real ˚0 .x/ and A0 .x/ background. For the integration, the most convenient
approach is to use the so-called thermal axial gauge, whereby the nonstatic part of
the m D 0 component of the vector potential is set to zero. Therefore, denoting the
nonstatic modes by lowercase symbols, one has the following decomposition of the
fields:
˚.x; / D ˚0 .x/ C '.x; /; Ai .x; / D ai .x; /; A0 .x; / D A0 .x/;

X X
'.x; / D 'n .x/ei!n ; ai .x; / D ain .x/ei!n : (6.55)
n¤0 n¤0
In this gauge and with the present background gauge field, the three-dimensional
U.1/ local symmetry is preserved. Therefore, the gradient piece of the reduced
theory will contain the three-dimensional covariant derivative. Therefore, it is
sufficient to find the renormalized kinetic piece for the scalar field. Also, it is
unnecessary to introduce a nonzero Ai .x/, since every nonzero mass or higher-power

term is forbidden by the reduced gauge invariance.
One-loop integration requires that one finds the part of the action quadratic in
the nonstatic fields. The relevant pieces of the Lagrangian density for the one-loop
computation are L.0/C L.2/ , which are easily found:
1 1 1 1
Fmn Fmn D .@i A0 .x//2 C .@i a0 .x//2 C .@i aj /2 .@i ai /2 ; (6.56)
4 2 2 2
1
Dm ˚ Dm ˚ !
2
1
! .@i ˚0 .x//2 C g2 A0 .x/2 ˚02 C .@m '1 /2 C .@m '2 /2 C g2 A20 .'12 C '22 /
2
1
C 2gA0 .'2 'P1 '1 'P 2 / 2gai .˚0 @i '2 ˚0 @'1 / C g2 a2i ˚02 ;
2
1 2
m ˚ ˚ C .˚ ˚/2 !
2 24
1 1
! m2 ˚02 C ˚04 C m2 .'12 C '22 / C ˚02 3'12 C '22 : (6.57)
2 24 2 12
Here the dot on the top of a symbol denotes the -derivative.
On a strictly homogeneous background, one finds the contributions to the poten-
tial energy density, though no information is obtained on any change (rescaling) of
the kinetic term. In this case, the quadratic piece in Fourier space has the expression
"
.2/ 1 2
L D '˛n .k/ !n2 C k2 C g2 A20 C m2 C ˚0 ı˛ˇ
2 6
#
2
2˛ˇ i!n gA0 C ˚0 ı˛1 ıˇ1 'ˇn .k/
3
1 2 2 2 2
j
C ai
n .!n C k /ıij ki kj C g ˚0 ıij an .k/
2
n .k/'2 .k/:
Cig0 ki ai (6.58)
It is obvious that the two scalar components ('˛ , ˛ D 1; 2) are coupled to the lon-
gitudinal gauge mode @i ain . The two transversal gauge modes are eigenmodes of
the harmonic expansion around the background. Using the identity “Trlog = logDet”
a little algebra allows us to find the contribution of the .3 3/-dimensional coupled
sector to the action of the effective potential of the three-dimensional action:
1 2 1
U3D D g A0 .x/2 ˚02 C m2 ˚02 C ˚04
2 2 24
Z "
1X 3
d k h
C 3
ln .!n2 C MA2 / .K 2 C g2 A20 C MH2 /.K 2 C g2 A20 C MG2 /
2 .2/
n¤0
#
i
4!n2 g2 A20 k 2
MA2 .K 2 C g2 A20 C MH2 / 2
C 2 ln.K C MA2 / : (6.59)
The following quantities are abbreviated in this formula:
2 2
MA D g˚0 ; MG2 D m2 C ˚ ; MH2 D m2 C ˚ ; K 2 D !n2 C k2 (6.60)
6 0 2 0
The Taylor expansion of the logarithms in powers of the masses can be applied
to the above expression without any risk, since in the denominators of the different
terms of the expansion, expressions of the form .!n2 /l1 .K 2 /l2 , n ¤ 0 will figure.
After performing the corresponding Matsubara sum-integrals, one proceeds to the
determination of the leading-order countercouplings ım21 ; ı1 ; ıg1 . The divergences
occurring in the coefficients of the pure ˚0 powers are of the same character as
in the ˚ 4 theory, except there will also be contributions proportional to g2 ; g.
Since in this calculation the field renormalization is neglected, one cannot determine
correctly the charge countercoupling ıg1 from the divergent coefficient of A20 ˚02 . At
this approximation, therefore, ıg1 is also kept at zero. Finally, the coefficients of the
pure A0 powers should turn out free of four-dimensional divergences, since there are
no counterterms to these operators.
Let us see in some detail the emerging thermal mass of A0 . Collecting all terms
proportional to g2 A20 , one obtains
XZ 1 !n2

g2 A20 2 4 : (6.61)
q q2 C !n2 .q2 C !n2 /2
n¤0
For the relevant Matsubara sums we have the following results:
X
1 1 2 1 ˇq
D 1 C 2y ; yD
ˇ 2 .q2 C !n2 / 4y e 1 4y2 2
n¤0
X
4ˇ 2 !n2 1 1 1
4 2 2 2
D 1 C 2 2y
C 2e2y 2y : (6.62)
ˇ .q C !n / 2y e 1 .e 1/2
n¤0
The momentum integral cutoff at jqjmax D over the sum can be performed
exactly:

1 g2 T 2 g2 T
ˇV A20 : (6.63)
2 3 2 2
With similar calculation, one obtains, for the effective action up to operators of
fourth power (with classical kinetic terms and choosing a renormalization scheme
where the quartic self-coupling of the reduced model is just T),
1 2 1 T 2 g4 ˇ 1
m3 ' ' C m2D AN 20 C .' '/ C 3 2 AN 40 C gN 2 AN 20 ' 2 ;
pot
L3D D
2 2 24 24 2
1N N 1 1 N 2 N
3D D Fij Fij C Di ŒA' Di ŒA' C .@i A0 / ; Di ŒA' D .@i C ig3 Ai /';
Lkin
4 2 2

1 2 2 1 N2 2
Lcounter
3D D ' ' 3g 3 C C A g : (6.64)
2 2 2 3 2 0 3
The following notation is used:

1 2 1 2
m2D D g T; m23 D C 3g23 T; (6.65)
3 3 24 3
and
p p p
'D ˇ˚; AN i D ˇAi ; AN 0 D ˇA0 ; g23 D g2 T: (6.66)
The expression of the coupling m23 is given in terms of the renormalized quantities
of the four-dimensional theory. Its explicit form depends on the renormalization
conditions imposed on U3D .˚0 ; A0 / [7].
A one-loop-accurate solution of the effective models offers some insight into the
mechanism of the finite-temperature symmetry restoration. The most convenient
choice for the evaluation of U3D with one-loop accuracy on a constant background
'0 is to choose the so-called ’t Hooft gauge, which suppresses the AL -'2 coupling
O in the quadratic part of the action (the last line below):
.ALi D kO i kA/
Zh
.2/ 1
S3D D Ai .k/.k2 ıij ki kj /Aj .k/ C g23 '02 jAi .k/j2
k 2
1
C A0 .k/.k2 C m2D C g23 '02 /A0 .k/
2

1
C '˛ .k/ k2 C m23 C '02 ı˛ˇ C ı˛1 ıˇ1 '02 'ˇ .k/
2 6 3
i i
C g3 '0 kAL .k/'2 .k/ '2 .k/kAL .k/ : (6.67)
2
The following gauge-fixing functional is chosen:
F D rA.x/ g3 '0 '2 .x/; (6.68)

R
which gives for the gauge-fixed action, after adding to it .1=2/ F 2 ,
Zh
.2/0 tHooft 1 1
S3D D Ai .k/ k2 ıij ki kj 1 C g23 '02 Aj .k/
k 2
1
C A0 .k/.k2 C m2D C g23 '02 /A0 .k/
2

1
C '1 .k/ k2 C m23 C C g23 '02 '1 .k/
2 6
i
2 2 2
C'2 .k/ k C m3 C '0 '2 .k/ : (6.69)
3
The quadratic form is now diagonal; each degree of freedom independently

contributes to the effective potential. One also has to take into account the opposite-
sign contribution from the Fadeev–Popov determinant:
2 2
R
ln.k2 Cg23 '02 /
Det.4 g23 '02 / D eTrlog.4g3 '0 / D eV k : (6.70)
For definiteness, we choose D 1 when the masses of the ghost and the transversal
gauge boson become equal. These are the degrees of freedom that do not receive any
thermal (Debye) mass. One checks explicitly that in the effective potential, the linear
divergences ( ) are exactly canceled by the induced counterterm contributions.
Finally, the finite contribution is rewritten in terms of four-dimensional quantities:

1 2 1 2 2 2
ˇU4D .˚0 / D m C T 3g C ˚02 C ˚04
2 24 3 24
"
1 1 2
T g3 ˚03 C m2 C T 2 3g2 C
12 24 3
3=2 3=2
1 2
C g2 C ˚02 C m2 C T 2 3g2 C C ˚02
6 24 3 2
3=2 #
1 2 2
C g T C g2 ˚02 : (6.71)
3
The presence of the cubic terms is a clear signal for some sort of infinite
resummation, since it is not analytic in the field amplitude. The potential is of “van
der Waals type,” since the coefficient of the sum of cubic contributions is negative.
Were one to estimate the temperature-dependence of the effective potential without
References 159
the 3D contributions, the usual mean field argument would signal a second-order
phase transition at Tc , where the thermal mass cancels the wrong sign m2 < 0.
The question of interest is, how intensive should the cubic piece be for turning this
conclusion into one predicting a first-order phase transition?
This question is discussed as a last point of this section using the example
of a simpler function, still reflecting the main features of the variation of the
potential under the effect of changing temperature. Consider a generic potential
parameterized as
3=2
U.˚/ D A.T/˚ 2 C B.T/˚ 4 C C.T/ ˚ 2 C K 2 .T/ ; C.T/ < 0: (6.72)
Calculating the second derivative of U at the origin, one easily obtains the
equation that determines the temperature below which the symmetric point becomes
absolutely unstable:
V 00 .0/ D 2A.T2 / C 3C.T2 /K.T2 / D 0: (6.73)
For T > T2 , the first derivative vanishes at three points. One root is the origin; the
other two are given as
2 1 h 2 2 2 2
3=2 i
˚˙ .T/ D 9C 16AB ˙ 3jCj 9C C 32.2B K AB/ : (6.74)
32B2
The temperature where
˚C .T1 / D ˚ .T1 / (6.75)
is the temperature where a metastable minimum appears first on decreasing the

temperature. If T1 > T2 , one sees how the nontrivial (metastable) minimum gets
deeper until eventually it touches the ˚-axis. This temperature T2 < Tc < T1 defines
the first-order transition temperature.
In Chap. 8, we shall analyze, along the lines sketched above, the electroweak
transition. After discussing results from a gauge-invariant version of the optimized
perturbation theory, we shall briefly describe the results of the nonperturbative
(lattice) investigations.
References
1. L. Dolan, R. Jackiw, Phys. Rev. D 9, 3320 (1974)

2. P. Ginsparg, Nucl. Phys. B 170, 388 (1980)
3. A. Jakovác, Phys. Rev. D 53, 4538 (1996)
4. K. Kajantie, M. Laine, K. Rummukainen, M. Shaposhnikov, Nucl. Phys. B 458, 90 (1996)
5. A. Sakharov, J. Exp. Theor. Phys. 5, 24 (1967)
6. K. Kajantie, K. Rummukainen, M. Shaposhnikov, Nucl. Phys. B 407, 356 (1993)
7. A. Jakovác, K. Kajantie, A. Patkós, Phys. Rev. D 49, 6810 (1994)
Chapter 7
Thermodynamics of the Strong Matter
The focus of this chapter is on the changing nature of the thermodynamic degrees
of freedom characterizing the strong matter. The theories employed are variants
of the effective meson–quark theories theoretically discussed in Chaps. 4 and 5
of these notes. Here we shall mostly rely on the renormalized equations derived
previously. The discussion begins with the thermodynamics of the system
completed at a more advanced stage by quark degrees of freedom. In this family
of models, the excitation spectra at zero temperature and baryon density are found
with the help of a large-N solution, which allows one to determine the actual values
of the couplings. The nature of the finite-temperature phase transformations and
the behavior of the excitations in a neighborhood of the transition temperature
follows, and finally, we turn to the effect of the finite baryo-chemical potential
(B ) on the transition. The second model family (the three-flavor meson–quark
model) is more complete, since it includes both the scalar and the pseudoscalar
nonets. The process of zero-temperature parameterization was described in the OPT
framework in Chap. 4. Here the change in the spectra at finite T and B will be
discussed. The scope of the investigations is expressed best in the form of a phase
diagram (see Fig. 7.6), which exhibits the nature of the finite-temperature transition
as established for different values of the baryo-chemical potential on arbitrarily
chosen values of the mass of the Goldstone bosons (the pion and the kaon). In the
second part of this chapter, we shall concentrate on the equation of state of the strong
matter. The functional dependence of the pressure on the energy density plays a
central role in any theoretical approach to high-energy heavy-ion experiments or the
internal structure of compact astrophysical objects, such as neutron stars. Here the
style of the discussion is closer to field-theory-motivated phenomenological model
investigations than to searching for solutions using the formal rules of the relevant
quantum field theories.

DOI 10.1007/978-3-319-22620-0_7
162 7 Thermodynamics of the Strong Matter
7.1 The Thermodynamics of the Quark System
7.1.1 The T D 0 Excitation Spectra
Nontrivial phenomenology can be built on the LO solution of the Gell-Mann–Lévy

linear sigma model discussed in Chap. 5. The input parameters used for fixing the
renormalized couplings are the T D 0 mass of the pion mG0 140 MeV and the
decay constant f 90 MeV characterizing the weak decay of the charged pions.
Also, one sets N D 4. Then applying the PCAC theorem to the pion decay and using
Eqs. (5.50), (5.60), and (5.61), one arrives at the relations

f h 2 m2G0 e
v0 D p ; m2G0 D ; m2 D f C m2G0 1 ln ;
N v0 6N 96 2 M02
(7.1)
which determine the (renormalized) parameters v0 ; h; m2 , using the explicit form
of the finite M0 -dependent part of the tadpole integral. The zero of the inverse -
propagator (5.64) continued to Minkowski metrics determines the mass of the sigma
particle (jpj D 0):
R 2 1
iG1 2 2
.p0 / D p0 mG0 f D 0: (7.2)
3N 1 R
6 B .p0 ; mG0 /
F
Here one needs the finite part of the bubble integral:

" p #
1 m 2 p 1 x 1 m2G0
B F .p0 ; mG0 / D ln G0 1 x ln p ; xD :
16 2 M02 1xCx p20
(7.3)
It is notable that it has a finite limiting expression also in the chiral limit. This set of
equations obeys the renormalization group invariance of the full solution. Namely,
one can find a functional form of .M02 / leading to the same m2 for a fixed input
f ; mG0 , independent of M0 .
The model predicts the -mass as a function of . The physically interesting
root of (7.2) is in the lower half-plane of the second complex Riemann sheet
parameterized as m i . In the chiral case, the easiest approach is to search
for it in the form
p0 D M0 ei'0 ; m D M0 cos '0 ; D M0 sin '0 : (7.4)
If the pion mass is finite, one shifts the parameterization to the location of the !
cut:
N 0 ei'N0 :
p0 D 2mG0 C M (7.5)
7.1 The Thermodynamics of the Quark System 163
3.5
2.5
1.5
1 h≠0 Mσ/fπ
h=0 Mσ/fπ
0.5 h=0 -Γ/(2fπ)
h≠0 -Γ/(2fπ)
ln(ML(T)/Mσ)
0
0 100 200 300 400 500 600 700
λR
Fig. 7.1 The real and imaginary parts of the complex pole of the -propagator for T D 0.
Results are presented both for the chiral limit and for a value of the explicit symmetry-breaking
corresponding to realistic pion mass. Also is shown the purely imaginary Landau pole, which
clearly sets a natural upper bound on (from [1])
In Fig. 7.1, the -dependence of the mass m and the width are displayed both
for h D 0 and h ¤ 0. In the figure, a bunch of descending curves appears,
which corresponds to the logarithm of the purely imaginary pole iML calculated
in proportion of m for several temperatures. It reflects the Landau-pole singularity
of the scalar self-coupling discussed in Sect. 5.3. The theory makes sense only if
the absolute scale of the Landau ghost lies much higher than the physical excitation
spectra. This requirement clearly puts an upper limit on the allowed range of the
renormalized self-couplings (max 300, since above this value ML =M <
exp.1:5/ 4:5). The same mechanism is at work in the standard model, where
one has to include also the effect of the top quark in a realistic computation of the
scale of the electroweak Landau singularity.
It is curious that the numerical solution of (7.2) displays a maximum. Apparently,
the value of the -mass cannot exceed 4f . This value is on the lower edge of the
mass range, quoted by the Particle Data Group. It is below the value obtained from
analyzing the phase shift in the isoscalar–scalar channel of the scattering in
the framework of the chiral perturbation theory [2]. This remark raises a number of
interesting questions: Does this upper bound becomes looser at NLO? Does there
exist such an upper bound in extended versions of the linear -model containing
more meson/quark fields?
The LO solution of the extended quark–meson model, discussed in Sect. 5.7, led
to the -propagator (5.161). After Minkowskian continuation, one has to find the
complex root of the following modified equation:
2 1 8g4 Nc Nf 2 F
iG1 1
.q/ D iG .q/ v
C v B .q; mQ /: (7.6)
3 1 6 B F .q; MG / N
One can exploit the freedom to introduce into the fermionic renormalized bubble
integral the renormalization scale M0F , which is somewhat different from that used
in the bosonic part (M0B ), although they should be chosen to be of same order of
magnitude. One can attempt to tune the ratio M0B =M0F exp. / to minimize the
renormalization-scale-dependence of the characteristics of the thermal transition.
By the analyticity of the propagator, the spectral function defined with zero
spatial momentum q D 0 as
1

.!/ D lim Im ŒiG .! C i; 0/ (7.7)
!0
is sufficient to characterize fully the propagator through a dispersive representation.

In Fig. 7.2, the ideal resonance-like shape is largely distorted, but the location of the
maxima moves toward the phenomenologically more acceptable range due to the
fermionic contribution.
0.035
M0B[MeV]=886
λ=400 850
0.03 680
0.025
η=-.25
ρσ(p0)fπ2
0.02
0.015
0.01
0.005
0 2 4 6 8 10
p0/fπ
Fig. 7.2 The spectral function of the excitation at T D 0 in the quark–meson theory ( D
ln.M0B =M0F / D 1=4) for three somewhat different values of M0B (from [3])
7.1 The Thermodynamics of the Quark System 165
7.1.2 Change of the Ground State at Finite Temperature

and Finite Density
In the quark system at finite temperature, the pion tadpole receives a thermal
Bose–Einstein contribution, and similar corrections appear in the quark tadpole
characterized by the Fermi–Dirac distribution. The two will modify the equation
of state (5.160) to
R 2 R M2 .T/e g2 Nc 2 MQ2 .T/e

M2 .T/ D m2R C v .T/ C M 2
.T/ ln M .T/ ln
6 96 2 M0B 2 8 2 Q M0F 2
Z 1
R T 2
C dy.ey 1/1 .y2 M2 .T/=T 2 /1=2
12 2 M .T/=T
Z 1
g2 Nc T 2
C dy.ey C 1/1 .y2 MQ2 .T/=T 2 /1=2 : (7.8)
2 MQ .T/=T
First, one can analyze the chiral limit when M D 0. At the critical temperature Tc ,
the vacuum expectation value of vanishes (MQ .Tc / D gv.Tc / D 0). Both integrals
can be calculated, and one arrives at the equation

R Tc2
0 D m2R C C g2 Nc : (7.9)
6 12
The case of nonzero baryo-chemical potential is obtained from (7.8) by replacing the
Fermi–Dirac factor .ey C1/1 in the last integral by .zey C1/1 C.ey =zC1/1 , where
z D e=T is called fugacity. Then the integral over the Fermi–Dirac distribution can
be performed term by term using an expansion of the integrand in powers of z.
The result can be expressed in terms of a specific polylogarithmic function Li2 .z/.
Making use of the identity (B.10), one arrives at the simple equation

R Tc ./2 g2 Nc 2
m2R C C g2 Nc C D 0: (7.10)
6 12 4 2
The dependence of the critical temperature on the baryo-chemical potential can be

deduced from comparing this equation with the previous one, (7.9), where Tc D
Tc . D 0/ is understood.
The line of second-order transitions changes into a first-order line at a tricritical
point (TCP) where also the coefficient of the quartic ( v 4 ) term in the effective
potential vanishes. This term is obtained by setting the derivative of the equation of
state (7.8) with respect to v 2 to zero. This leads to the following analytic expression:
ˇ
R g4 Nc @ ˇ const: TTCP
C .Lin .z/ C Lin .1=z// ˇ ln D 0:
6 4 2 @n nD0 M0 B
(7.11)
140
120 λ=400
100
T [MeV]
80
TCP
60
40
nd
2 order line
20 1st order line
spinodal
0
0 100 200 300 400 500 600 700 800 900 1000
μB [MeV]
Fig. 7.3 The phase diagram in the T -plane (M0B D M0F D 886 MeV). On the left from the
tricritical point (TCP) the symmetric and the broken symmetry phases are separated by a second-
order transition (dashed) line. On the two sides of the first-order line on the right of TCP, the
(dotted) lines enclose the region of spinodal instability (from [3])
The root of this equation when combined with (7.10) determines the coordinates
TCP ; TTCP of the tricritical point. The solution with realistically chosen parameters
is displayed in Fig. 7.3. In the case of explicit breaking of the chiral symmetry,
the second-order line will change into a crossover region, and in place of TCP, a
critical endpoint (CEP) of the first-order line appears. The location of the CEP for
realistic pion mass is close to that of the TCP displayed in the figure. The existence
of a CEP is largely debated in numerical simulations of QCD. The existing Monte
Carlo simulations suffer from unknown systematic errors because of the negative
weights associated with certain gauge configurations. The first numerical estimates
of the CEP location [4] indicate lower CEP and higher TCEP values than what is
obtained with the large-N approximation presented here. There are studies that have
led to opposing the existence of the CEP [5]. The result presented here, however, is
fully consistent with most the estimates derived to date from various versions of the
effective chiral models [6–11].
7.2 The U.3/ U.3/ Meson Model
An approximate solution of this broader model at T D 0, described in Chap. 4,

relies on the optimized perturbation theory at the one-loop level. The renormalized
couplings were determined exclusively with data from the pseudoscalar sector,
avoiding the use of the scalar meson spectra, where the experimental knowledge
is much less reliable.
7.2 The U.3/ U.3/ Meson Model 167
Particularly important is the role of the optimization conditions (4.44) imposed

on the and K propagators. These equations are extended to finite T by evaluating
the contributing one-loop self-energy diagrams as
.TD0/ .T/
˙M .p2 D m2 ; mi ; / D ˙M .p2 D m2 ; mi ; / C ˙M .p2 D m2 ; mi ; /;
(7.12)
where M D ; K and one separates the T D 0 and T ¤ 0 contributions. In principle,
one should evaluate the self-energy self-consistently on the respective mass shells
of the pion and the kaon. In practice, the computation is much simpler when one
assumes that the momentum-dependence of the self-energies is negligible. Then the
temperature-dependent part of the self-energies is computed at p2 D 0. At this point,
the bubble integrals can be expressed with help of tadpoles:
T F .m1 ; T/ T F .m2 ; T/
B F .p D 0; m1 ; m2 / D : (7.13)
m22 m21
As a consequence, the temperature-dependent corrections to the consistency condi-

tions will consist of a linear combination of the T-dependent parts of the tadpole
contributions evaluated with appropriate meson masses:
m2 D 2 C 2.2f1 C f2 /x2 C 4f1 y2 C 2gy

X
CZ ˙.1loop;TD0/ .m2 ; mi .m /; / C c˛ T F .mi .m /; T ¤ 0/;
˛Di ;i
p
m2K 2 2 2
D C 2.2f1 C f2 /x C 4f1 y C 2gy 2x.2f2 y g/
.1loop;TD0/
CZK ˙K .m2K ; mi .mK /; /
X
C cK˛ T F .mi .mK /; T ¤ 0/: (7.14)
˛Di ;i
The notation mi .m / or mi .mK / emphasizes that in the equations, all meson masses
are expressed with the help of the appropriately chosen Goldstone boson (for the
field, the 1-loop pole mass; for all others the tree-level masses were used in [12]).
When these equations are solved together with the Ward identities (4.48),
one obtains the temperature-dependent quantities x.T/; y.T/; m .T/; mK .T/. Based
on the PCAC relations (4.49), one determines also the temperature-dependence
of f ; fK . The tree-level mass relations transfer the temperature-dependence of
the Goldstone bosons on the masses of all other fields. The transition is well
characterized by the variation of these quantities displayed in Fig. 7.4. The left-hand
figure displays the variation of the nonstrange (x) and strange (y) order parameters
in proportion to f .T D 0/ for four different choices of the renormalization scale
(denoted in the figure by l) It is obvious that the crossover happens around T D
200 MeV and is dominated by the nonstrange order parameter. The variation of the
renormalization scale only mildly influences the transition. In the right-hand figure,
0.9 1350
0.8 1200
0.7 1050
0.6 900
0.5 750
MeV
0.4 y/fπ: l=1300 MeV 600
1200
0.3 1100 450
mf
1000 mκ0
x/fπ: l=1300 MeV ma
0.2 mη’0 300
1200
1100 mσ
0.1 mη 150
1000 mK
mπ
0 0
50 100 150 200 250 300 50 100 150 200 250 300 350
T [MeV] T [MeV]
Fig. 7.4 Left: the T-variation of the order parameters x and y and the sensitivity to the choice of
the renormalization scale. Right: the T-dependence of the masses in the pseudoscalar (; K; ; 0 )
and scalar (; a0 ; ; f0 ) multiplets (from [12])
the variation of the meson masses is displayed. Here one notices the masses of the
parity partners from the scalar and pseudoscalar nonet approaching each other. The
approximate degeneracy is realized well above the transition region (T > 250 MeV).
7.2.1 Phase Transition in the Three-Flavor Quark–Meson

Model
For the determination of the temperature-dependence of the nonstrange (x.T/) and

strange (y.T/) condensates, it is sufficient to consider in addition to the equations
of state, the equation of the pion mass solved self-consistently. For convenience,
Eqs. (4.54), (4.36) will be given here again, in a slightly more explicit form:
X
x D x.2 C 2gy C 2.2f1 C f2 /x2 C 4f1 y2 / C tix T F .mi .m /
i
4gF Nc Mu T F .Mu /;
X
y D y.2 C 4f1 xy C 4.f1 C f2 /y2 C 4f1 x2 / C gx2 C
y
ti T F .mi .m //
i
p
2 2gF Nc Ms T F .Ms /;
m2 D 2 C 2.2f1 C f2 /x2 C 4f1 y2 C 2gy C ˙ .p D 0; mi .m /; Mu /: (7.15)
Here also the fermionic tadpoles are expressed through bosonic tadpoles of mass
MQ , and in the evaluation of the meson contribution, the mass eigenbasis of the
fields is used. All masses are expressed (as was explained in Chap. 4) through m .
The tadpole integrals are split into the sum of a T D 0 and a temperature-
dependent part, where one has to choose carefully, where it is appropriate, between
7.2 The U.3/ U.3/ Meson Model 169
the Bose–Einstein and the Fermi–Dirac distributions:

Z 1 q
1
T .mi / D T .mi ; T D 0/ C
F F
dE E2 m2i nBE .E=T/;
2 2 mi
T F .MQ / D T F .MQ ; T D 0/
Z 1 q
1
2 dE E2 MQ2 .nFD ..E C B /=T/ C nFD ..E B /=T// :
4 MQ
(7.16)
A discontinuous jump in x.T/ signals first-order transitions for large enough

baryo-chemical potential B . With diminishing B , the gap diminishes, and at the
CEP it vanishes. The line of first-order transitions exists for > CEP , as appears
in Fig. 7.5 (dashed line). Left from the CEP one observes finite maxima of the
order parameter susceptibility dx=dx . The location of these maxima in the
B T-plane draws a pseudocritical line (dashed line), which can be associated
with the interval of rapid changes in the thermodynamic quantities. This is the
crossover region. The pseudocritical line found with this approximate solution hits
the B D 0-axis at Tcross .B D 0/ 155 MeV, which agrees rather well with the
crossover temperature region identified in Monte Carlo simulations of the full QCD
[14, 15].
160 lattice CEP cross-over line

1st order line
spinodal
140 freeze-out curve
120
TIsing
100
T [MeV]
Scaling region of the CEP

105
χ/χfree
100
80 2
CEP
95 1.6
1.2
60 90 0.8
85 0.4
40 80
75 CEP
20
70
780 800 820 840 860 880 900 920
0
0 100 200 300 400 500 600 700 800 900 1000 1100
μB [MeV]
Fig. 7.5 Phase diagram of the three-flavor quark–meson model in the T D -plane. In the insert,
the scaling behavior of the baryo-chemical susceptibility is displayed. By its scaling exponent, the
CEP of this model belongs to the universality class of the Ising model (from [13])
A scaling behavior of can be searched in the form of the following general

parameterization:
D .jB CEPj cos ˛ C jT TCEP j sin ˛/ : (7.17)
The angle ˛ determines the orientation of the main scaling axes in the plane. In
fact, one of the main axes lies along the tangential to the pseudocritical curve at
the CEP. The value of the exponent is close to unity, which gives some evidence
of the universality class of the second-order transition at the CEP belonging to the
Ising class. In the inset of Fig. 7.5, the susceptibility is shown in proportion to the
susceptibility of a free massless fermion gas. The darker region corresponds to the
larger ratios. This graphical representation nicely depicts the main scaling direction.
7.2.2 Dependence of the Nature of the Transition

on the Masses of the Goldstone Fields
Lattice simulations of the full QCD thermodynamics can be performed on a wide

range of quark masses, which correspond to varying the masses of the light
pseudoscalar Goldstone bosons. In the chiral limit (zero pion and kaon mass) and
in a certain (unknown) mass region close to it, one expects a first-order transition
between the phase of broken and restored chiral symmetry. It has been firmly
established that the physical point in the m mK -plane lies outside this regime.
It is located in the crossover region. Some effort has been invested in localizing
the boundary of the first-order regime in numerical simulations, at least along the
diagonal m D mK [16–18]. At present, no firmly accepted boundary mass is known
even along this direction of higher mass degeneracy.
The existence and the location of the second-order phase boundary was first
intensively studied in the framework of effective models at B D 0. The first
computations [19] used just the one-loop-level equations (described above) and
scanned through the .m ; mK /-plane using for the other couplings those values
that arise in substituting into their expressions the actual pion and K-masses. It
was realized that in the framework of the three-flavor nonlinear sigma model, the
Goldstone masses influence the other masses and the weak decay constants rather
subtly. In [20], the parameters of the linear model are fitted to the exact relations
derived in the chiral perturbation theory. There is no space in the present notes to
enter into a discussion of the details of matching the couplings in the two models.
Only results obtained in the linear sigma model with the correctly tuned parameters
will be discussed briefly below.
Figure 7.6 displays the boundary found with different parameterizations. The
condition used for recognizing the second-order nature of the transition was the
infinite response of the order parameter x to the variation of T at a fixed value of
B . The phase boundary could be constructed for mK > m . The m D mK line
7.3 Equation of State of the Strong Matter 171
Fig. 7.6 The surface of μB,CEP

second-order phase
transitions in B m mK
space. In the region to the left μB [MeV]
and above the surface, the
restoration of chiral 1000
900
symmetry proceeds via 800
first-order transitions below 700
the surface via crossover. The 600
500 physical point 500
physical points that might be 400
reached in our actual world 300 400
]
eV
lie along the vertical line 200 300
[M
100 diagonal
(from [13]) 200
K
0
m
60 80 100 120 140 100
160 180
mπ [MeV]
cannot be reached, since there, the parameterization based on having separate self-
consistent gap equations for the pion and the kaon becomes singular. It also stays
away from the m D 0 axis; in fact, for mK > 400 MeV, the surface curves back
and does not cross it at all.
The critical surface begins orthogonal to the m mK -plane. Then it bends
rightward, and the equations definitely exclude an eventual bending backward.
By this monotonic change, the existence of the CEP (as suggested by the figure)
depends critically on the starting tendency of the bending near mB D 0. The
corresponding derivatives can be estimated in numerical simulations, avoiding the
problem of negative weights.
7.3 Equation of State of the Strong Matter
In the previous sections, we discussed the spectrum and the description of the phase
transition with the help of the effective action. For the bulk thermodynamics, the
most relevant relation is the dependence of the free energy on the temperature and/or
chemical potential (called the equation of state, EoS). In homogeneous systems,
the complete thermodynamics is derived from this relation. In the remaining part
of the chapter, we discuss the thermodynamics at zero chemical potential with the
aim of achieving a physically fully adequate, analytic though phenomenological
characterization.
The strong matter can be described in terms of qualitatively different degrees
of freedom in the low- and high-temperature (energy-density) regimes. Although
there is no strict phase transition between the two regimes, we shall speak about
low- and high-temperature phases. Because of the continuous (crossover) nature of
the transition, in principle the two regimes can be analytically continued from each
phase to the other; but if one tries to use the degrees of freedom appropriate for
6
lattice continuum limit SB
5
4
p/T4
3
2
HTL NNLO
1
HRG
0
200 300 400 500
T[MeV]
Fig. 7.7 The pressure of quark–hadron matter in 2 + 1 flavor QCD by the numerical simulation
of the BMW group [21]. The HRG approximation (dashed line) and the normalization-scale-
dependent range covered by the perturbative results (the dashed-dotted lines corresponding to
renormalization scales T; 2T; 4T, respectively) are also shown here
a given phase to describe the other one, one typically gets very badly converging
series.
No direct experimental measurements can be constructed for the EoS of QCD,
but one can perform lattice MC simulations. After decade-long efforts, the most
influential groups in the field of finite-temperature MC simulations provided results
that are in reasonable agreement with each other [21, 22]; see Fig. 7.7. In view
of the confidence in QCD as the exact theory of strong interactions, the results
of numerical simulations can be accepted as the true behavior of the pressure of
strongly interacting matter. The measured equation of state predicts a continuous
transformation of the phases. This means that there is no discontinuity in any
derivatives of the free energy, but only peaky but rounded structures can be observed
in the temperature-dependence of the different thermodynamic susceptibilities. The
position of the peak depends on the chosen observable. From the susceptibility of
the chiral condensate, the best known result for the pseudocritical temperature is
Tc D 155 MeV with a few MeV uncertainty. This value is referred to as the “critical”
Tc in the sequel.
In the high-temperature phase, the degrees of freedom of the strong interaction
are those that define QCD: quarks and gluons. At very high temperatures, the
coupling constant of QCD is vanishingly small (asymptotic freedom), and so one
expects that the free gas approximation will be valid. But this is true only for very
high temperatures T 1000Tc. In the region where the temperature is just a few
times Tc , the pressure reaches only about 80%–85% of the Stefan–Boltzmann limit
corresponding to free quarks and gluons. This means that in the moderately, but not
asymptotically, high temperature region, the interaction plays a crucial role in the
quantitative description of the QCD pressure.
...
Fig. 7.8 A ladder diagram with arbitrary number of gluon ladders gives the same O.g6 /
contribution to the self-energy: by adding an extra rung, the additional coupling factors from the
two vertices are canceled by the inverse magnetic mass from the additional gluon propagator
The first theoretical tool to account for this behavior is perturbation theory.
Straightforward perturbation theory with thermal masses but without any other
optimization, however, shows very bad convergence features near Tc : the series of
the computed quantities in different orders no longer converge. One needs better
methods to obtain reliable results. In QCD, the most successfully used methods
are dimensional reduction (DR) discussed in Chap. 6, and hard thermal loop (HTL)
resummation [23], not discussed in these notes. Presently, the HTL perturbative
results are three-loop accurate, while the DR results have reached the maximal g5
accuracy. For higher precision, one necessarily encounters nonperturbative correc-
tions present at each higher order. A nonperturbative resummation is needed to treat
them, since starting from order g6 , the diagrams with increasing number of loops
contribute at the same level (see Fig. 7.8). This is the so-called Linde problem [24].
A similar phenomenon appears also in the electroweak case (cf. Sect. 8.2). There
one could take into account the resulting net effect by introducing an appropriate
magnetic mass. In QCD, one can fix the g6 terms by numerical simulations [25].
As a result, one gets stable results from both approximation procedures. As is
customary in perturbation theory, these approximations still have some uncertainty,
arising, for example, from the choice of the renormalization scale. The knowledge
of some experimentally measured data allows one to select the procedure that best
reproduces the data; cf. Fig. 7.9. The results very closely follow the lattice data;
nevertheless, they do not show any signal for flattening toward low temperatures.
Shortly after the pressure calculated in perturbation theory deviates from the MC
data (in the lower left corner of the figure), it crosses zero, i.e., it apparently
describes an unstable state of matter.
At low temperatures, we are in the hadronic phase, where the relevant degrees of
freedom are the color neutral hadrons. From the point of view of thermodynamics,
one starts with a free gas of hadrons. One faces here, however, the problem that
most hadrons are not stable. It is not perfectly clear what should happen with a
degree of freedom that decays before it can really be part of the thermal medium.
We return to this problem below. The first approximation is certainly the assumption
that all hadronic resonances can be taken into account as stable particles: this is the
hadron resonance gas (HRG) limit. There, all hadronic resonances give a partial
contribution to the total pressure in the form of an ideal quantum gas
Z 1
T X
PD 2
dp p2 ./ ln 1 eE.p;mn /=T ; (7.18)
2 n
0
1.0
0 MeV 1 loop ; 176 MeV
MS
0.8
0.6
ideal
0.4
0.2 NNL OHTLpt
Wuppertal Budapest
0.0
200 400 600 800 1000
T MeV
Fig. 7.9 The best perturbative results for the QCD pressure at high temperature from [23]. The
shaded region is the variation of the perturbation theory under the change of the renormalization
scale by a factor of 2
where the ˙ sign applies to bosonic/fermionic modes, and E2 .p; m/ D p2 C m2 .

Within a nonrelativistic approximation (all hadrons are much heavier than the
equivalent energy density corresponding to the actual low temperature), we can
forget about the difference of statistics between bosons and fermions and have
T2 X 2
PD m K2 .ˇmn /; (7.19)
2 2 n n
where K2 is the modified Bessel function. The hadron masses are taken from
experiments [26].
Although this approach seems to be oversimplified, it works surprisingly well, at
least below a certain temperature that depends on the observable; for the pressure, it
describes well the MC data up to .1 1:2/Tc [21]. We should remark here, however,
that the HRG pressure is a continuously increasing function that finally reaches the
Stefan–Boltzmann (SB) limit belonging to the gas of hadrons, instead of the SB limit
of the QCD degrees of freedom. Assuming the existence of infinitely many stable
hadrons, one observes eventually a singular diverging behavior in the hadronic
pressure. The reason is [27] that toward large masses, the density of hadronic
resonances increases exponentially: %H .m/ em=TH , where TH is the Hagedorn
temperature. This tendency can also be extracted from the experimental data [28].
There are several fits for the experimentally observed hadronic resonance density,
for example those with powerlike prefactors in
H , but already a simple exponential
does the expected job. Then the complete pressure behaves as
Z1 Z1
m=T
P dm %H .m/e D dm em.1=T1=TH / ; (7.20)
0 0
which is divergent when T > TH . This singularity, called a Hagedorn singularity,

signals the inapplicability of the hadronic description for temperatures higher than
TH .
Therefore, the validity range of the two descriptions, approaching the critical
temperature from below and from above, are more or less overlapping, but not
analytically related. If there were a first- or second-order phase transition between
the two regimes, then we could simply stick those regimes together, changing
description abruptly at Tc . From a purely phenomenological point of view, this
can still be satisfactory, since it provides a good estimate for the pressure at any
temperature.
However, for understanding the physics of the transition regime, and in particular
to design a predictive strategy for applications in other physical situations, this
dual approach is unsatisfactory. The phase transformation is continuous, and so
we are pushed unavoidably to the conclusion that during the crossover regime, the
excitations of the other phase appear continuously. The MC data display in dp=dT
an inflection. But as we have emphasized, the pressure of the HRG is increasing
monotonically, while the pressure of the HTL calculations decreases monotonically.
These two facts suggest that within these approximation schemes, the continuous
crossover to the other phase cannot be described. The precondition for the correct
physical interpretation is an appropriate understanding of how the various degrees
of freedom appear/disappear in the transient state of the system.
7.3.1 Characterization of the Excitations
Let us first discuss what kind of excitations or particles we want to describe. The
notion of particles in several cases means free particles (for example in an ideal gas),
or, as in scattering theory, the asymptotic states. But as we have stated before, the
HRG approach takes into account all hadronic resonances, even those that live for
a very short time. A particle with a finite lifetime cannot be an energy eigenstate;
it is a collective excitation instead: we call them quasiparticles. The physical origin
of the appearance of a quasiparticle is that the original free particle modes are
mixed with the multiparticle modes of their environment, as was first demonstrated
by the Weisskopf–Wigner approximation to the coupled atom + radiation system.
The collective excitations maintain to some extent a nearly energy-eigenstate-like
behavior to the degree that the real-time evolution of their response function is
e t ei!t . The quasiparticle properties, the damping rate , and the energy (mass)
! depend on how the quasiparticle was formed. Consequently, they usually depend
on the temperature/chemical potential.
The identification of quasiparticles therefore requires the observation of the
characteristic asymptotic behavior of the response function Gret .t/. Since the
response function is perfectly described by the spectral function %, the quasiparticle
is best characterized as a peak in % that behaves near the peak as a Lorentzian:
A
%.k0 !/ D : (7.21)
.k0 !/2 C 2
Before proceeding further, we have to clarify here an interesting question of a

general nature. A finite system always contains discrete energy levels En . Then the
spectral function reads
X
%.k0 / D rn ı.k0 En / (7.22)
n
with some spectral weights rn . Now if the time evolution starts from an energy
eigenstate, then it remains there jn; ti D eiEn t jni. There is no damping or decay
whatsoever, and we get a stable system as in quantum mechanics. How, then, can
we observe decays of the real excitations at all?
The answer stems from understanding the way in which the final state of the
system is created. Consider a system having just two energy levels E1 and E2 , and
the initial state is composed of them coherently: j i D c1 j1i C c2 j2i. Then the
expectation value of a Hermitian observable O O at time t reads
D E D E
O .t/ D ; tjOj
O O ; t D jc1 j2 O11
C jc2 j2 O22 C .c1 c2 ei.E1 E2 /t C c2 c1 ei.E2 E1 /t /O12 ; (7.23)
D E
where Oij D ijOjjO . If we want to determine whether just one of the energy levels
was initially excited, or both, but we do not know the matrix elements exactly,
then we should rely on the time evolution. If both energy levels are excited, we
should observe a beat with frequency E D jE1 E2 j. The point is that we should
monitor the evolution for a time duration minimally lasting t 1=E to be able
to recognize that there is a time-dependence in at least one of the expectation values.
The creation of a state is just the reverse of the measurement process. A control
process that forces the system to some initial state actually requires a preparation
time t 1=E.
The particle creation process, on the other hand, is rather fast. Therefore,
although a quasiparticle peak consists of a large number of Dirac deltas (in a finite
system), the dynamics is not able to pick up any of them to excite it separately. We
can excite only an interval of the eigenstates located under the peak on an equal
footing. This leads to an intermediate “final” state

X
j iQP cn jni; (7.24)
n
where the cn coefficients express the degree of sensitivity of the creation process
to the energy level jni. Its modulus jcn j2 is a spectral weight, describing the
ability to create or measure an energy D level.
E If we assign a quantum field to the
O
creation/measurement, then cn D nj˚j0 . Then the overlap between the state
j ; tiQP and the initial state (i.e., the propagator) reads as
X Z
d!
h ; 0j ; tiQP D jcn j2 eiEn t %.!/ei!t D %.t/; (7.25)
n
2
according to the definition of the spectral function (A.5). This formula, with the help
of (7.21), yields the exponentially damped behavior in time.
Why is this argument relevant in the discussion of the changing nature of the
degrees of freedom? Because it makes clear that the quasiparticles have nothing to
do with single separate energy levels of the system. They are characterized only
by their weighted density, i.e., the spectral function. Therefore, if the “degrees of
freedom” change in a system, the appropriate quantity for the characterization of
the change is provided by the spectral function.
All that has been said above applies only if the members of a set of energy
eigenstates have the same quantum numbers. Otherwise, one easily chooses an
operator OO that is sensitive to a distinct value of a conserved quantity. Then in (7.23),
O11 ¤ 0, while the other matrix elements are zero, and so we easily can determine
whether c1 D 0. The same applies to particle creation: one usually calls a particle
that carries well-defined values of the conserved quantities, i.e., it has definite
quantum numbers, a quasiparticle.
Therefore, we can use different spectral functions for different quantum channels
(superselection classes), i.e., for different sets of quantum numbers. But it makes no
sense to assign more than one spectral function to the same quantum channel; more
precisely, it would introduce tacitly a new quantum number, i.e., it would change the
symmetry of the quantum system. The result is then a physically different system.
The corollary therefore is that to each superselection class we can assign one, but
only one, spectral function.
7.3.2 Spectral Functions and Thermodynamics
As we have seen, the excitations of the system should be characterized by the

spectral functions. The next step is to examine how to extract the thermodynamics
from a given set of spectral functions.
The complete partition function of the system is written as

X
ZD eˇEn;q ; (7.26)
n;q
where q labels the distinct quantum channels and n is the index of the energy
eigenstates in channel q. Usually, we do not know the exact spectral functions,
except for quadratic (free) theories. In a free theory, the partition function can
be evaluated, and we need only to know the dispersion relation of the one-
particle channel: it determines all the other channels as well as the complete
thermodynamics.
In our discussion, we assume that the interacting theory, after we have taken into
account all sorts of interactions, finally behaves like a free theory with a modified
particle spectrum. For low enough energy, in the strong interactions this assumption
is certainly a good approximation to the thermodynamics, as the success of the HRG
demonstrates. We will assume that this approach can also be consistently applied in
the crossover regime, with much more distorted excitation spectra than in the low-
energy regime.
For each system with a single degree of freedom, therefore, we will assume
that there is a fundamental channel (similar to the original free particles of the
system) with spectral function %, and this determines the thermodynamics as well
as the density of states of all the other channels. If there are bound states in the
multiparticle channels, then these channels also should be treated as fundamental,
where we must take care to avoid double counting.
Mathematically, these basic principles are implemented using a quadratic theory
with a generic kernel K , containing tunable parameters. Let us discuss a theory of
scalar fields to avoid complications with the internal degrees of freedom. Then the
action reads (compare to (2.65))
Z
1
SŒ' D d 4 x '.x/K .i@/ '.x/: (7.27)
2
In Fourier space, we can write

Z
1 d4 p
SŒ' D ' .p/K .p/'.p/: (7.28)
2 .2/4
The ' field is real, which in Fourier space means that '.p/ D ' .p/. Since S is
also real, the kernel must satisfy the relations K .p/ D K .p/ D K .p/. From
the generic rules, we have for the retarded propagator and the spectral function
G.ra/ .p/ D K 1
.p0 C i"; p/; %.p/ D Disc iG.ra/ .p/: (7.29)
p0
For the inverse relation, we can use the Kramers–Kronig relation (2.56), and then
the relation between the kernel and the retarded Green’s function. Finally, we obtain
Z 1
d! %.!; p/
K .p/ D lim Re : (7.30)
"!0 2 p0 ! C i"
One should ask whether the field theory defined with help of
.!; p/ is
consistent. For an arbitrary kernel it is not, but it can be proven [29] that if the
spectral function satisfies the generic properties discussed in Appendix A (namely
that %.p0 > 0, p/ > 0, %.p0 ; p/ D %.p0 ; p/) and it is normalized, then the field
theory is consistent. If in addition, the spectral function is Lorentz-invariant, then the
theory will be relativistically invariant, too. Therefore, the spectral function should
be chosen as the input, and we should determine the kernel using the dispersion
relation above.
Now we have a model theory that is capable of reproducing any spectra we want,
at least in the fundamental channel. What is the thermodynamics it describes? It is
best to start with the computation of the expectation value of the energy density.
This quantity is the T00 component of the energy–momentum tensor, which is the
Noether charge density generated by the time translation-invariance of the system.
It is well defined microscopically. With the standard techniques, we arrive at the
energy momentum tensor (cf. [29])
1
T .x/ D '.x/ D K .i@/ '.x/; (7.31)
2
where
" ˇ #
@K .p/ ˇˇ
D K .i@/ D i@ g K .i@/; (7.32)
@p ˇp!i@
sym
and the symmetrized derivative is defined as
1 X
n
f .x/Œ.i@/n sym g.x/ D Œ.i@/a f .x/Œ.i@/na g.x/: (7.33)
n C 1 aD0
Although this expression is somewhat complicated, when we take its expectation,

it becomes much simpler. For the expectation, we can use the KMS relation (2.49).
Finally, we obtain
Z
1 d4 p @K 1
" D hT00 i D 4
p0 C n.p0 / %.p/; (7.34)
2 .2/ @p0 2
where we also exploited the equation of motion K .p/%.p/ D 0 (cf. (2.68)). If we

use the expression of the free spectral function %.p/ D .2/ sgn p0 ı.p2 m2 / and the
free kernel K .p/ D p2 m2 , then we get back from this formula the ideal gas energy
density (2.88). The integrand is symmetric, so it is enough to take into account the
positive frequency part of K and omit the zero-point 1=2 factor. This formula is
not sensitive to the normalization of the spectral function (the normalization of K
cancels it).
This result already suffices to recover the complete thermodynamics. In particu-
lar, the pressure reads
Z
1 d 4 p @K
pD %.p/ ln.1 eˇp0 /: (7.35)
2 .2/4 @p0
We remark that an analogous formula can be derived using the generalized Beth–
Uhlenbeck approach [30], where the imaginary time formula of the thermodynamic
potential is expressed through the phase shifts.
7.3.3 (In)distinguishability of Particles and the Gibbs Paradox
Formula (7.34) has remarkable properties. First of all, if in a quantum channel there
are Dirac-delta excitations with dispersion relation p0 D Ei .p/, then the formula can
be evaluated. In this case, the spectral function and the propagator are both additive:
X X Zi
%.p/jp0 >0 D 2Zi ı.p0 Ei .p//; G.ra/ .p/ D ;
i i
p0 C i" Ei .p/
(7.36)
where Zi are the individual wave-function renormalization constants. The kernel is
not additive, however, since it is K D 1=G. Its derivative reads
@K .p0 / 1 @G 1 X Zi
D 2 D !2 : (7.37)
@p0 G @p0 X Zi i
.p0 Ei /2
i
p 0 Ei
Because the spectral function is concentrated on the energy levels, in (7.34) we have
to evaluate p0 K 0 .p0 / near these points. As can be easily seen,
@K .p0 / 1
lim D : (7.38)
p0 !Ei @p0 Zi
Therefore, the Zi factors drop out from the formula of the energy density. As a result,
we obtain a sum of free energy densities
X .n/
"D "0 ; (7.39)
n
.n/
where "0 is the contribution to the free energy density from the nth particle.
This means in particular that a gas of N particles with the same quantum numbers,
but with different masses, behaves as a mixture of N distinguishable particle species,
despite the fact that all these particles are just peaks of the same spectral function.
In this case, therefore, the number of degrees of freedom was N, although the
number of fundamental quantum channel was just one. The two numbers, therefore,
can be different. For the characterization of the number of degrees of freedom of the
system, we can introduce a somewhat heuristic measure:
Z1
@K
Ndof D dp0 %.p0 /: (7.40)
@p0
0
This is a dimensionless quantity, and it gives N for the above example of the
spectrum consisting of N Dirac-delta peaks.
As we see, the number of degrees of freedom as well as the complete energy
density is a functional of the spectral function. This makes it possible to experience
a dynamical change in the degrees of freedom. In this way, we can account also
for the Gibbs paradox. In the original version of the paradox, Gibbs argued that the
entropy of mixing two gases depends on whether the molecules of the gases are
distinguishable. In the above example, we physically had a mixture of N species of
Bose gases. We have seen that the number of species is explicitly measurable here
(in contrast to the original “mixture entropy” problem), since, for example, at high
temperature, the Stefan–Boltzmann limit of the total energy density is N 2 T 4 =30.
Two species are distinguishable in the present formalism if in the spectrum we can
find two distinct energy levels, if i.e., their masses are different. If the particles
are free, then no matter what the mass difference (provided it is not zero) and
the relative wave function renormalization constants may be, the two species are
always distinguishable, yielding a factor of 2 in the Stefan–Boltzmann limit. The
formulas (7.34) and (7.40) provide the analytic expressions reflecting the essence of
the paradox: when two Dirac deltas merge into one, the resulting energy density or
Ndof changes nonanalytically.
In reality, the quasiparticles always have a finite width, and this smoothens the
nonanalytic behavior of " and Ndof . An analytically calculable example is the case
of two Lorentzians with the retarded propagator
Z Z
G.ra/ .!/ D C : (7.41)
! C i ! E C i
Fig. 7.10 Two nearby 0.8

quasiparticle peaks with γ =0.01
various widths, D =E. 0.7 γ =0.05
γ =0.1
The peaks are scaled with
0.6 γ =0.2
appropriately chosen Z to
ensure optimal visibility
0.5
0.4
0.3
0.2
0.1
0
-0.5 0 0.5 1 1.5 2
Physically, this describes two quasiparticles with mass difference E, and if they are
independent, with lifetime 1 . For simplicity, we take here the same width and
same wave-function renormalization. The corresponding spectral function can be
seen in Fig. 7.10.
What do we expect in this system? Assume that ¤ 0, but that it is small enough
that a single peak can be considered a sufficiently long-lived quasiparticle. Now if
E
, then the two peaks are far from each other, and the system must behave as
if it consisted of two Dirac-delta peaks as analyzed above. But when E , then
in the spectrum, we cannot disentangle the two peaks. Since the physics is encoded
in the spectrum, we expect that we have a system with a single degree of freedom.
Therefore, we should see the continuous disappearance of a degree of freedom as
the mass difference of the particles vanishes.
Exactly this picture arises when we calculate Ndof . One can perform the integral
in (7.40), resulting in
E2 16 2
Ndof D 1 C E2 : (7.42)
.E2 C 16 2 /2
This function is plotted in Fig. 7.11. If E ! 0, then Ndof ! 1, and if E ! 1, then

Ndof ! 2. Numerically, E= O.10/ is the regime in which the effective number
of degrees of freedom changes the fastest. We can also see a dip where Ndof < 1;
this is a peculiarity of the simple quantity we defined.
7.3.4 Melting of the Particles and Phenomenology

of the Crossover Regime in QCD
In a quantum field theory, the spectra consists not just of discrete energy levels,
but also in each channel, a multiparticle continuum is present, the continuum of
Fig. 7.11 The effective 2

number of degrees of
freedom for two peaks as a
function of their energy
1.5
difference scaled by the
width. For large separation,
we see two degrees of
N eff
freedom; for zero separation, 1
just one. This is the smooth
version of the Gibbs paradox
0.5
0
0 5 10 15 20 25 30 35 40
E/Γ
the scattering states. Then the relevant physics scenario induced by an increase of
the width of a particle is not the merging of two different particle species, but the
merging of the quasiparticle into the continuum. Since the continuum contains many
more states, the result of this process is that the particle peak becomes unidentifiable
in the spectrum. Therefore, all the physical effects associated with the particle peak
should also be suppressed, meaning that the particle has literally vanished from the
ensemble. This process is called particle melting.
If we want to illustrate this process quantitatively, then we may use the trial
spectral functions (cf. [31])
%.p/ D Z.%Lor .p/ C %cont .p//; (7.43)
where
4h 2 p2
%Lor .p/ D ;
.p2 m2 /2 C 4 2 p2
q
p
%cont .p/ D 2 Im m2th C iS p2 : (7.44)
p C m2th
In this parameterization, h is the height of the peak, is the peak width, Z is the
wave-function renormalization constant. The continuum is modeled by a 2-particle
spectral function with complex threshold mass (m2th C iS) and a correction prefactor
is also introduced to ensure that the spectral function vanishes at p D 0. The overall
normalization factor is not important, since it does not influence the physics. We
depicted this spectral function with different values in Fig. 7.12, normalizing the
quasiparticle peak height to unity for better visibility. We have chosen here m D 1,
mth D 2, with different values shown in the figure, and S that is consistent with
the requirement that the quasiparticle width be the height of the spectral function
at m.
Fig. 7.12 Trial spectral 200 γ=0.005

functions containing a γ=0.015
quasiparticle peak with γ=0.045
different widths and a 150 γ=0.137
continuum. The quasiparticle
peaks (at m D 1) are scaled to
the same height for better
ρ
100
visibility
50
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
ω
Fig. 7.13 The pressure γ=0 γ=0.046

calculated from trial spectral 0.14
γ=0.018 γ=0.066
functions 0.12 γ=0.032 γ=0.095
0.1 SB
p/T4
0.08
0.06
0.04
0.02
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
T
Now we can use (7.35) to calculate the pressure of the system. We get the
series of curves plotted in Fig. 7.13. What we see here is that the pressure in the
melting process becomes smaller and smaller. In the limit ! 1, we get zero
pressure, and the particle has indeed become nonexistent from the point of view of
thermodynamics. We can also observe that these curves can scale to each other by a
multiplication
p.T/ p0 .T/Neff ; (7.45)
where p0 .T/ is the pressure of the free particle gas with D 0, and Neff is a
more-or-less temperature-independent constant; cf. Fig 7.14. The coefficient Neff
can be interpreted as the effective number of thermodynamical degrees of freedom.
It is the function(al) of the spectral parameters and does not depend directly on the
temperature. In case of rather complicated trial spectral functions [32], a stretched
˛
exponential Neff D eA was always a satisfactory fit.
Fig. 7.14 The change in the data

effective number of degrees 1 fit
of freedom as a function of
the quasiparticle peak width. 0.8
The width of the band shows
the temperature-dependence
0.6
Neff
0.4
0.2
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
γ
The framework obtained through the analysis presented in the last three sub-
sections can be used to draw a qualitative picture about the fate of the hadronic
excitations in the crossover regime of QCD itself. According to the commonly
accepted interpretation of the MC simulations [21, 22], all thermodynamic quan-
tities are very well described by assuming a pure, phenomenologically composed
hadron gas up to T Tc . This would suggest that there are no QGP degrees of
freedom present in this temperature range, at least that their contribution is not very
important. On the other hand, above Tc , hadronic states are definitely observed
[33–37], although these are not quasiparticle-like excitations. Combining these
observations, one is tempted to suggest that in the spectral function of the partons of
the QGP phase, there are no peaky structures below Tc (although other, scattering-
like, states might still exist and even might dominate the continuum part of the
spectral function in some channels). In the hadronic channels, one observes well
identifiable peaks below Tc , and these do not disappear abruptly (discontinuously)
at any well-defined temperature.
Analyzing these possibilities, one can propose a scenario whereby the hadrons
vanish from the spectra by melting. The simplest approach is to assume that all
hadrons melt in the same way. Therefore, one can adapt a common Neff ;hadron for
them. One has to treat the degrees of freedom of the QGP phase, too, again with
a common Neff ;QGP . A natural expectation is that the width of the QGP excitations
˛
depends on the available hadronic resonances, for example QGP Neff ;hadron , where
˛ is a conveniently chosen power of order 1. Even with this simple scenario, one can
obtain a smoothly varying spectral content through the crossover region instead of
producing the Hagedorn singularity. This smooth variation can be nicely fitted to
the MC observations. Figure 7.15 shows a scenario in which the hadrons survive for
a very long time: they are present up to T 3Tc in the plasma, and they even give
half of the pressure at T 2Tc [32].
Fig. 7.15 A possible 5

hadrons+partons
scenario for QCD crossover hadrons
with melting hadrons, plotted partons
4
with the MC measurements. lattice MC data
Hadrons can survive in this
scenario up to 3Tc 3
4
p/T
2
0
0 100 200 300 400 500 600 700 800
T (MeV)
Correspondingly, the QGP degrees of freedom appear only continuously, and

they become the dominant degree of freedom of the plasma only beyond 2Tc . In
any case, the regime Tc < T < 3Tc is a regime in which the standard quasiparticle
picture cannot be applied; the interplay between the quasiparticle peak(s) and the
continuum affects the physics crucially.
References
1. A. Patkós, Zs. Szép, P. Szépfalusy, Phys. Lett. B 537, 77 (2002)

2. I. Caprini, G. Colangelo, H. Leutwyler, Phys. Rev. Lett. 96, 132001 (2006)
3. A. Jakovác, A. Patkós, Zs. Szép, P. Szépfalusy, Phys. Lett. B 582, 179 (2004)
4. Z. Fodor, S.D. Katz, J. High Energy Phys. 0404, 050 (2004)
5. P. de Forcrand, O. Philipsen, J. High Energy Phys. 0701, 077 (2007)
6. M. Asakawa, K. Yazaki, Nucl. Phys. A 504, 668 (1989)
7. A. Barducci, R. Casalbuoni, G. Pettini, R. Gatto, Phys. Rev. D 49, 426 (1994)
8. M.A. Halasz, A.D. Jackson, R.E. Shrock, M.A. Stephanov, J.J.M. Verbaarschot, Phys. Rev. D
58, 096007 (1998)
9. O. Scavenius, A. Mocsy, I.N. Mishustin, D.H. Rischke, Phys. Rev. C 64, 045202 (2001)
10. Y. Hatta, T. Ikeda, Phys. Rev. D 67, 014028 (2003)
11. A. Barducci, R. Casalbuoni, G. Pettini, L. Ravagli, Phys. Rev. D 72, 056002 (2005)
12. T. Herpay, Zs. Szép, Phys. Rev. D 74, 025008 (2006)
13. P. Kovács, Zs. Szép, Phys. Rev. D 75, 025015 (2007)
14. Y. Aoki, G. Endrődi, Z. Fodor, S.D. Katz, K.K. Szabó, Nature 443, 675–678 (2006)
15. M. Cheng et al., [HotQCD Collaboration] Phys. Rev. D 74, 054507 (2006)
16. F. Karsch, E. Laermann, C. Schmidt, Phys. Lett. B 520, 41 (2001)
17. N.H. Christ, X. Liao, Nucl. Phys. Proc. Suppl. 119, 514 (2003)
18. P. de Forcrand, O. Philipsen, Nucl. Phys. B 673, 170 (2003)
19. H. Meyer-Ortmanns, B.-J. Schaefer, Phys. Rev. D 53, 6586 (1996)
20. T. Herpay, A. Patkós, Zs. Szép, P. Szépfalusy, Phys. Rev. D 71, 125017 (2005)
21. S. Borsanyi, Z. Fodor, C. Hoelbling, S.D. Katz, S. Krieg, K.K. Szabo, Phys. Lett. B 730, 99
(2014)
22. A. Bazavov et al., [HotQCD Collaboration] Phys. Rev. D 90(9), 094503 (2014)
References 187
23. J.O. Andersen, N. Haque, M.G. Mustafa, M. Strickland, N. Su, arXiv:1411.1253 [hep-ph]
(2014)
24. A.D. Linde, Phys. Lett. B 96, 289 (1980)
25. K. Kajantie, M. Laine, K. Rummukainen, Y. Schroder, J. High Energy Phys. 0304, 036 (2003)
26. K.A. Olive et al. (Particle Data Group), Chin. Phys. C, 38, 090001 (2014)
27. R. Hagedorn, Nuovo Cimento Suppl. 3, 147 (1965)
28. W. Broniowski, W. Florkowski, L.Y. Glozman, Phys. Rev. D 70, 117503 (2004)
29. A. Jakovac, Phys. Rev. D 86, 085007 (2012)
30. D. Blaschke, D. Zablocki, M. Buballa, A. Dubinin, G. Roepke, Ann. Phys. 348, 228 (2014)
31. A. Jakovac, Phys. Rev. D 88, 065012 (2013)
32. T.S. Biro, A. Jakovac, Phys. Rev. D 90, 094029 (2014)
33. S. Datta, F. Karsch, P. Petreczky, I. Wetzorke, Phys. Rev. D 69, 094507 (2004)
34. T. Umeda, K. Nomura, H. Matsufuru, Eur. Phys. J. C 39S1, 9 (2005)
35. M. Asakawa, T. Hatsuda, Phys. Rev. Lett. 92, 012001 (2004)
36. A. Jakovac, P. Petreczky, K. Petrov, A. Velytsky, Phys. Rev. D 75, 014506 (2007)
37. P. Petreczky, J. Phys. Conf. Ser. 402, 012036 (2012)
Chapter 8
Finite-Temperature Restoration
of the Brout–Englert–Higgs Effect
The origin of the matter–antimatter asymmetry of our cosmic neighborhood is one

of the outstanding challenges faced at present by cosmology and particle physics.
This asymmetry plays a central role in the successful modeling of the primordial
nucleosynthesis (for a general introduction to particle-physics aspects of cosmology,
see [1]). The observational estimation of the concentration of light elements of
primordial origin can be interpreted with the help of the single asymmetry parameter
nbaryon nantibaryon
baryon D 6 1010 ; (8.1)
n
where n is the photon number density and nbaryon ; nantibaryon are the respective
densities of matter and antimatter. The estimate for this number was further
sharpened with the help of the cosmic microwave background radiation data.
The common standpoint is that this feature of extreme importance for our
very existence has a dynamic origin. It should have been generated from a
symmetric initial state not later than when the spontaneous symmetry-breaking of
the electroweak interactions occurred in the cooling universe.
There are three distinct necessary features that every quantum field theory has to
possess for the generation of this asymmetry [2]:
• violation of the quark number (B) conservation;
• violation of the combined charge-conjugation–parity (CP) transformation sym-
metry;
• extended time interval during the cosmic evolution when quantum fields were far
from thermal equilibrium.

DOI 10.1007/978-3-319-22620-0_8
190 8 Finite-Temperature Restoration of the Brout–Englert–Higgs Effect
The standard model possesses the first two features. Anomalous processes trans-
forming baryons into leptons were demonstrated to exist by ’t Hooft [3]. A specific
saddle-point solution of the field equations, called sphaleron, was constructed [4, 5],
tunneling through which changes the baryon number of the ground state. The sus-
tained presence of the nonzero baryon density generated by such processes requires
the violation of the charge conjugation + parity symmetries, which is realized
in the three-family structure of the standard model as predicted by Kobayashi
and Maskawa [6]. The irreversibility of the processes generating the asymmetry
could be due to a far-from-equilibrium state of the universe. The proposal that
the electroweak phase transition, the onset of the Brout–Englert–Higgs (BEH)
effect on cosmic scales. could be of a first-order nature and produce the necessary
nonequilibrium stage was made in 1985 by Kuzmin et al. [7].
Early investigations based on summing the infrared-sensitive daisy-diagram
contributions to the effective Higgs potential computed at the one-loop level
[8, 9] have predicted persistently first-order transitions of gradually diminishing
discontinuity in the Higgs vacuum condensate with increasing self-coupling of the
Higgs field. Two-loop corrections turned out to be extremely large [10], raising
doubts as to the reliability of the perturbative approach.
In this chapter, we describe first the results of the application of the optimized
perturbation theory (OPT) to the electroweak phase transition as treated in the
framework of the dimensionally reduced standard model. An optimized one-loop
calculation led to the conclusion that the first-order nature of the transition has
an endpoint when the quartic self-coupling of the Higgs sector exceeds the value
corresponding to MHiggs 100 GeV/c2 [11]. Since this conclusion hints at the
necessity to extend the presently known elementary interactions beyond the standard
model (BSM), it proved to be of great importance to obtain similar information from
the exact solution of the reduced model relying on Monte Carlo simulations [12, 13].
The basis of the numerical investigations and the evidence confirming the existence
of a critical endpoint of the first-order transition line will be summarized in the
second part of the chapter.
8.1 The Reduced SU.2/ Symmetric Higgs + Gauge Model
The dimensionally reduced three-dimensional Higgs + gauge model with SU.2/

gauge symmetry has the following action:
Z
1 2
S3D ŒW; ; D d3 x Tr Wij Wij C .Di ˚/ Di ˚ C m23 ˚ ˚ C 2BP
3 ˚ ˚ ;
2
(8.2)
8.1 The Reduced SU.2/ Symmetric Higgs + Gauge Model 191
where the trace is understood over the SU.2/ indices. For the fields, the following
singlet–triplet representations are used:
1 1 E E i /˚:
˚D C iE E ; Wi D E Wi ; Di ˚ D .@i ig3 E W (8.3)
2 2
(The conventions of [11] are followed, for which reason, the index BP, with
reference to the initials of its authors, appears on the Higgs self-coupling.) The real
fields ; are related nonlinearly to the four components of the complex doublet ˚c
used in the standard model:
˚c ˚c D 2Tr.˚ ˚/;

Eji D i g3 ˚c E @i ˚c .@i ˚c / E ˚c D g3 @i E @i E .@i /
E E : (8.4)
2 2
The first relation is the condition for the invariance of the local scalar potential; the
second expresses the unchanged current of the scalar fields in the interaction with
the gauge fields.
The lowest approximation to the dimensional reduction that leads to this form
of the reduced action is arrived when after the nonstatic bosonic and fermionic
modes, one integrates also over the static W 0 fields. This process, realized at the
one-loop level, determines the effective couplings g3 ; m23 ; BP
3 and also the effective
counterterms needed for finding the renormalized physical data of the model.
The integration over the nonstatic fields can be done in full conformity with
the steps followed for the abelian Higgs + gauge model presented in Sect. 6.5. For
the case of SU.2/ symmetry, the expression of the reduced action of the three-
dimensional fields is the following [14, 15]:
Z h
SŒW i ; W 0 ; ˚ D N 23 ˚ ˚ C 2N 3 .˚ ˚/2
d3 xTr Wij Wij C .Di ˚/ Di ˚ C m
i
C .Di A0 /2 C m2D W02 C 2W W04 C 4h3 W02 .˚ ˚/ (8.5)
(Di D @i iNg3 Wi ). Here the induced three-dimensional counterterms are not written
out explicitly, and (without fermions)
1 E 5 2 17g4.T /T 1
W0 D E W0 ; m2D D g .T /T 2 ; W D 2
; h3 D gN 23 ;
2 6 48 4

3 1
gN 23 D g2 .T /T; N 3 D .T /T; N 23 D m2 C
m g2 .T / C .T / T 2 ;
16 2
(8.6)
and T denotes a special choice of the renormalization scale, which ensures the
suppression of all logarithmic one-loop contributions to the expressions of gN 3 ; N 3 ; m
N 23
(this is again a sort of optimization to avoid large contributions from perturbation

theory; see Sect. 3.5.1).
For the second step of the one-loop integration, one separates the quadratic part
of the W0 -dependent three-dimensional SU.2/ symmetric action on constant ˚0 ; W N ia
background [15]:
Z
.2/ 1
SW0 D W0a .k/ (8.7)
2 k
2
k C m2D C h3 ˚02 C gN 23 .W N ic gN 23 W
N ic /2 ı ab 2iNg3 acb ki W N ia W
N ib W0b .k/:
The electric potential receives a Debye screening mass mD of O.gT/, which is softer
than the effective mass of the nonstatic Matsubara modes. Still, on reduction, this
integral is infrared safe, and there is no need for gauge fixing. Below, evidence will
be presented that the characteristic scale of the three-dimensional (magnetic) Wi -
fluctuations is O.g2 T/.
The contribution of the W0 -integration to (8.2) is obtained by expanding the result
of the integration up to fourth power in the 3D field variables with the following
result:
3=2 EN W EN /2
N ia D 1 1 gN 4 .W
m2D C gN 23 ˚02 C 3 C O.AN i /:
i j 6
W0 UŒ˚0 ; W
4 4 96 m2 C gN 2 ˚ 2 =4 1=2
D 3 0
(8.8)
The second term on the right-hand side is readily completed to a gauge-invariant
correction of the three-dimensional pure gauge action:
1 a a 1 a a 1 1 gN 23
Fij Fij ! F F ; D1C 2 : (8.9)
4 4 ij ij 24 m C gN 2 ˚ 2 =4 1=2
D 3 0
The leading (˚0 -independent) piece of this induced magnetic susceptibility can be
absorbed into a field rescaling, but it is then reflected in the final relation of g23 to the
four-dimensional gauge coupling. Also, the first term of the correction is expanded
up to ˚04 and modifies both m N 23 and N 3 .
Here we give the finite parts of the relations of the couplings of the three-
dimensional Higgs + gauge system derived in [14, 15] for the case of pure SU.2/
Higgs + gauge theory, taking into account also the effects of W0 :
3Ng23 mD
m23 D m
N 23 C O.g4 ; 2 /;
16

N 3Ng43 gN 23
BP
3 D 3 C O.g4 ; 2 /; g23 D gN 23 1 : (8.10)
128mD 24mD
8.2 Optimized Perturbation Theory for the Electroweak Phase Transition 193
The zero of m23 determines the temperature where the effective squared mass of
the Higgs field changes sign (it contains the negative renormalized squared mass
parameter m2 ), which in the mean field approximation would give the critical
temperature of a continuous (second-order) transition. In a better approximation,
one decides about the nature and the location of the transition after solving the
effective three-dimensional theory with static field fluctuations taken into account.
8.2 Optimized Perturbation Theory for the Electroweak

Phase Transition
A gauge-independent formulation of the OPT to the effective Higgs + gauge theory

has been put forward in [11]. The Lagrangian is first expanded around the scalar
background as D v C 0 and supplemented with the gauge-fixing term
1 g 2
LGF D @i Wia C v a : (8.11)
2 2
The full Lagrangian (the 3D counterterms not written out explicitly) includes also
the contribution of the anticommuting ghost triplet ca representing the Fadeev–
Popov determinant of the reduced theory:
2 2
1E E 1 E i /2 C g 3 v 2 W
E i2 C g3 v 0 W
E i2 C Eji W
Ei
LD Wij Wij C .@i W
4 2 8 4
1 1 g2
C .@i 0 /2 C BP 2 02
3 v C .@i /E 2 C 3 v 2 E 2
2 2 8
1
C .m23 C BP 2 02
3 v /. C E 2 C 2v 0 /
2
g23 E 2 02
C W . C E 2 / C BP 0 02
E 2/
3 v . C
8 i
BP
3 1 1
C . 02 C E 2 /2 C m23 v 2 C BP v2
4 2 4 3
g23 2 E i cE/
C@i cE @i c C v cE Ec C g3 @i cE .W
4
g2
C 3 v 0 cE Ec C cE .E Ec/ : (8.12)
4
For a one-loop computation in the 3D perturbation theory, one obtains for the gauge
field the following propagators, some of them explicitly depending on the gauge-
fixing parameter :
pi pj pi pj
ij .p/ D ıab DT .p/Pij C DL .p/Pij ;
Dab PTij D ıij ; PLij D ;
T L
p3 p2

g23 2 g23 2
D1 2
T D p C v ; D1 D 1
p 2
C v : (8.13)
4 L
4
The Higgs fields and the ghost propagators are partly degenerate:

g23 2
.1
/ ab
D . 1 ab
/ D ı ab p 2
C v ; 1 2 2 BP 2
D p C m3 C 33 v :
c
4
(8.14)
Here it is appropriate to recall the idea of OPT: one replaces in the propagators
the tree-level masses by the optimal mass parameter (attention: this parameter does
not necessarily coincide with the true mass of the exact solution):
g23 2 2 2
v D MW ıMW ; m23 C 3BP 2 2 2
3 v D MH ıMH : (8.15)
4
This replacement is done also in the masses proportional to the gauge-fixing
parameter . The optimization of the mass and of the transversal vector mass
parameters is expressed then by requiring the on-shell suppression of the one-loop
self-energy correction to the chosen masses:
2
ıMW C ˘T .p2 D MW
2
; MH ; MW ; / D 0;
ıMH2 C ˙.p2 D MH2 ; MW ; MH ; / D 0; (8.16)
where ˘T .p2 / is the coefficient function of the transversal part of the polarization
tensor of the W i field, and ˙.p2 / is the self-energy of 0 . The linear divergences
occurring in the one-loop calculation of these functions are automatically canceled
by the induced counterterms.
However, the equations resulting from simply copying the OPT scenario worked
out for scalar theories in Chap. 4 would depend explicitly on the gauge-fixing
parameter. The clever idea suggested in [11] was to achieve gauge independence
by appropriate resummations of the three- and four-point couplings of (8.12). The
resulting gauge-invariant piece chosen for the resummed one-loop computation
emerges by replacing m23 and BP3 through the “exact” mass parameters:

1 1 g2 M 2
Lresum D Tr Wij Wij C .Di ˚/ Di ˚ C MH2 ˚ ˚ C 3 2H .˚ ˚/2 ; (8.17)
2 2 4MW
supplemented with the gauge-fixing term
1 2
LGF;resum D @i Wia C MW a : (8.18)
2
Also, the shift of the -field is parameterized through the optimal parameters:
2MW
D 0 C : (8.19)
g3
The terms completing Lresum to the original L are shifted to the interaction piece,
which should be included only at the next (two-loop) order. In this way, one
explicitly checks in the one-loop gap equations the cancellation of the -dependent
contributions to the Higgs and gauge self-energies on the respective mass shells:
"
2 3v 3
˙.p D MH2 / D g23 .m23 C BP 2
3 v /C 2
2
.4MW C MH2 /T3 .MW
2
/
2g3 MW 4MW
3 4 2

2
C 8MW 4MW MH2 C MH4 B3 .MH2 ; MW 2
; MW2
/
8MW
#
3MH2 9M 4
C 2 T3 .MH2 / C H
2
B3 .MH2 ; MH2 ; MH2 / : (8.20)
4MW 8MW
" 2
2 2 2 MW 2 BP 2 3MW MH2 1 2
˘T .p D MW / D g3 v.m3 C 3 v / C T3 .MW /
g3 MH2 MH2 8MW 2 2
2
MH 1 63 2
C 2
C T3 .MH2 / MW B3 .MW 2
; MW 2
; MW2
/
8MW 2 64
#
2 1 2 MH4 2 2 2
C MW MH C 2
B3 .MW ; MW ; MH / : (8.21)
2 8MW
Here T3 .2 / is the tadpole integral in the three-dimensional theory computed with
propagator masses 2 D MH2 ; MW 2
. Its linearly divergent piece is taken care of by the
induced counterterms; only its finite part is to be included into the gap equations.
The first argument of the three-dimensional bubble diagram B3 gives the value of
2
the momentum where it is evaluated (p2 D MW ; MH2 ). The second and third
arguments give the propagator masses appearing on the two lines of the bubble.
The missing input into the two equations comes from the equation of state,
which determines the physical value of the background v. This equation arises from
averaging the fluctuating part of the -field to zero: h 0 i D 0. This equation turns
out to be mildly -dependent:

3 MH2
v.m23 C BP
3 v 2
/ D g M
3 W 4T3 .M 2
/ C 2
T3 .M 2
/ C T3 .M 2
/ :
4 W
MW W H
(8.22)
The integrals T3 and B3 can be calculated easily in analytic form. Taking into
account the fact that only the one-particle-irreducible diagrams contribute, one
obtains, after the cancellation of the linear divergences, the following explicit form
for the gap equations plus the equation of state:
2 g23 2
MW D v C MW g23 f .z/; MH2 D m23 C 3BP 2 2
3 v C MH g3 F.z/;
4
p
3 M3
v.m23 C BP 2
3 v /D
2
g3 4MW C MH2 C H ; (8.23)
16 MW
where the explicit form of the two functions f .z/; F.z/ depending on z D MW =MH
can be found in the original publication [11]. Note that F.z/ becomes complex for
z < 1=2, signaling the decay of into two W’s. In the physically interesting region,
each term in each of the equations is real, and one has to search for real solutions
v; MW ; MH .
The numerical solution of the set of three equations was found by fixing values
2 2 2
of x D BP3 =g3 . The quantities v=g3 ; MH =g3 ; MW =g3 were evaluated as functions of
2 4
y m3 =g3 . This variation corresponds to that induced by the temperature, since the
reduced-mass parameter m23 changes sign at some T D Tbarrier . Negative values of
x correspond to the broken symmetry phase, large positive values to the phase of
restored symmetry.
The dependence on the gauge-fixing parameter is negligible for small values
2
of BP
3 =g3 . Therefore, in this region one safely can choose the Landau gauge D 0.
The variation of the scaled order parameter v=g3 is displayed in Fig. 8.1, where the
2
continuous line gives the result for BP3 =g3 D 1=128, which (estimating by the tree-
level relations) corresponds to the T D 0 mass relation MH0 MW0 =4 (here and
below, the T D 0 masses receive an extra lower “0” index in order to distinguish
them from the temperature-dependent masses). The order parameter has large values
for y < 0, which monotonically decrease as y approaches zero. This branch of the
solution is stopped at some maximal value in the region y > 0. In the symmetric
phase, the value of the order parameter is nonzero but takes very small values almost
independently of y. The two overlapping branches of the order parameter solutions
near y D 0 demonstrate the coexistence of two minima in the free energy density.
The computation is optimized for the masses and not for the corresponding effective
potential. Therefore, it is doubtful whether one can decide which solution is the
stable one at a given y.
M2H0 / M2W0 < (M2H0 / M2W0) CEP

M2H0 / M2W0 > (M2H0 / M2W0) CEP
v/g
0 2 4
m3 / g3
Fig. 8.1 Variation of the order parameter v=g with m23 =g43 for two values of 8BP =g2 MH0
2 2
=MW0
(after [11])
M2H0 / M2W0 < (M2H0 / M2W0) CEP M2H0 / M2W0 < (M2H0 / M2W0) CEP
M2H0 / M2W0 > (M2H0 / M2W0) CEP M2H0 / M2W0 > (M2H0 / M2W0) CEP
2
MW/g3
MH/g32
0 0 2 4
m23 / g43 m3 / g3
Fig. 8.2 Variation of the scaled Higgs mass MH =g23 (left) and the scaled vector mass MW =g23 (right)
with m23 =g43 for two values of 8BP =g2 MH0
2 2
=MW0 (after [11])
One can perform also an ordinary one-loop renormalized perturbative calculation

that gives vanishing v in the symmetric phase, while it has a nonzero solution in the
broken symmetry phase approximating quite well the solution of the gap equations.
This calculation corresponds to the pioneering investigation of Kirzhnits and Linde
[16], who concluded that for small Higgs masses, the electroweak phase transition
is of first-order nature.
2
When repeating the analysis for BP 3 =g3 D 1=8 (or MH0 MW0 ), one finds
(dashed line) a continuously varying scaled order parameter over the whole range
of m23 =g43 (the perturbative calculation makes no sense for y > 0). This is a clear
message from a continuous crossover transition.
The crossover nature of the transition in the second case is further supported
by the variation of the masses in the two cases. In the left-hand figure of Fig. 8.2
one recognizes the abrupt jump downward of the Higgs mass in passing through yc
(continuous line), which clearly hints at a first-order transition (MH0 MW0 =4).
The dashed line corresponds to nearly equal Higgs and vector masses, and in the
broken-symmetry phase, one observes the Higgs mass passing through a nonzero
minimum, signaling the presence of a pseudocritical temperature.
The branch of the nonperturbatively computed vector mass (the right-hand

figure) for MH0 MW0 =4 begins close to the perturbative value in the broken phase
and ends discontinuously at some y > 0 near zero. Another branch grows for y > 0
from zero, which at large y approaches from below a small, but definitely nonzero,
constant value as opposed to the expectation based on perturbation theory. When
MH0 MW0 (dashed line), the two sections appear to form a unique continuous
curve, displaying the same nonzero asymptotic value for large y. This effect is
related to the nonperturbative dynamics of the gauge sector and defines a high-
temperature scale of O.g23 / O.g2 T/. The constant-high-temperature asymptotics
hint at the existence of magnetic screening in nonabelian pure gauge theories
like QCD. The existence of the magnetic mass has been long debated [17–19],
but one should not forget that an OPT-based evidence in favor of finite-range
magnetic screening could be an artifact arising from the restricted set of optimizable
parameters.
One can also compare the results with the simple perturbative one-loop mass
calculations. The close, almost quantitative agreement of the perturbatively calcu-
lated Higgs and vector masses with the result of the gap equations is obvious in the
broken-symmetry phase for smaller Higgs mass. The two results are also not too
far from each other in the high-temperature phase concerning the scalar mass. For
nearly equal vector and scalar masses, the quality of the agreement deteriorates even
in the broken-symmetry phase when one approaches y D 0. In the symmetric phase,
only the Higgs mass is nonzero in the perturbative framework, but again its value
diverges from the solution of the gap equation for asymptotically large temperatures.
2
c =g ) from the first-order phase transition to
The location of the transition (BP
2
the crossover-type variation lies between the two values of BP 3 =g3 analyzed here.
Since the results for MW0 MH0 become increasingly -sensitive, especially in
the region y 0, one cannot draw reliable quantitative conclusions about the
2
accuracy of the one-loop location of the transition (BP c =g 0:0406). Taking
into account all the uncertainties of the approximate solution, one can, however,
safely claim that the first-order nature of the electroweak transition changes into a
smooth crossover somewhere for MH0 MW0 . Since the mass of the Higgs particle
discovered at CERN is outside this region, one has to abandon the idea of the
electroweak generation of the matter–antimatter asymmetry, under the condition that
this conclusion becomes supported by more accurate mappings of the electroweak
phase structure.
The accurate localization of the endpoint of the first-order line of electroweak
phase transitions was found with the help of Monte Carlo simulations. It will be
reviewed with emphasis on the results obtained in the framework of resummed
(dimensionally reduced) effective models in the next section.
8.3 Results from the Numerical Simulation of the Electroweak Model Reduced. . . 199
8.3 Results from the Numerical Simulation

of the Electroweak Model Reduced with One-Loop
Accuracy
In the lattice regularization of (8.2), the gauge field Wj .x/ is replaced by the link
variable Ux;j D exp.ig3 Wj .x/a/, pointing from the lattice site x to x C aej (a is
the lattice constant; ej is the unit vector along the j.D 1; 2; 3/-axis). The Wilson
action of the gauge field is proportional to the SU.2/-trace of the plaquette variable

UP.jl/ .x/ D Ux;l UxCael ;j UxCaej ;l Ux;j . The scalar multiplet is parameterized in the same
way as in (8.3); the Higgs self-coupling will be redefined in this section as 3 D
6BP
3 . The lattice regularized action in this notation is given as [13, 20]
1X h X X
Slat D Tr ˇ UP.jl/ C 4 ˚ .x/Ux;j ˚.x C aej /
2 x jl j
2 23L 2 i
˚ .x/˚.x/ ˚ .x/˚.x/ : (8.24)
3
The dimensional couplings of the effective theory given at one-loop accuracy in the
previous section are related to the dimensionless couplings of (8.24) as
4
aT; ˇD ; 3L D 3 a;
g23 a
r !
1 3 2 15g3 5
D m23 a2 C 6 ˙.L/ g C ;
2 32 6
˙.1/ D 0:252731: (8.25)
Note that in the coefficient of the so-called hopping term ( 1 ), also a size-
dependent ( ˙.L/) piece is present beyond the additive contribution (C6) arising
from the discretization of the kinetic term. This contribution to the squared mass or
. 2 a2 /1 is linearly divergent near the continuum ( 1=a) and replaces the previous
induced counterterm, generated now by evaluating the three-dimensional tadpole
integral on a cubic lattice with lattice constant a.
There exist detailed introductory textbooks helping the reader to gain deeper
insight into the technology of lattice field theory [21, 22]. Here we proceed by
presenting the findings of [13] without entering into any technical details. The
correspondence of the lattice couplings to the four-dimensional couplings was fixed
by setting D 1, that is, a D 1=T, and choosing ˇ D 9. This choice determines
g23 , and by the one-loop relation, the four-dimensional renormalized gauge coupling
g. The simulation has been realized for a number of 3 values. By its one-loop
expression, one obtains, in the knowledge of g, the corresponding values of the
Higgs self-coupling in four dimensions.
Fig. 8.3 Variation of the 7

maxima of the field-length
susceptibility (8.26) with the
lattice size for L D 8 to
L D 48 for five different 6
values of 3 (from [13])
ln χΦ2,max
5
2 2.5 3 3.5 4
ln L
Using the physical value of the vacuum condensate v D 246 GeV, one can use the
corresponding tree-level expressions to estimate the W and Higgs masses. The set of
variables corresponds to fixing the W-mass near its physical value and investigating
the nature of the symmetry-restoring transition for various Higgs masses. For each
set of couplings, one tunes , which corresponds to tuning jm2 j=T 2 . The goal is
to find the maximum of certain thermodynamic susceptibilities of the system. The
corresponding value of , called max , characterizes the temperature (in units of the
renormalized mass) where the system turns from the broken-symmetry regime into
the symmetric phase.
A widely used class of signatures allowing distinction between first-order
transitions and crossovers is the finite-size behavior of various susceptibilities. In
Fig. 8.3, the size-dependence of
˚ 2 D Volume h.˚ ˚ h˚ ˚i/2 i (8.26)
is plotted on a log-log diagram. For a first-order transition, ˚ 2 should increase with

the size of the lattice system as V D L3 , while for a crossover, it should tend to a
constant. The figure clearly reveals that for 3 D 0:170190 (the upper left group
of points), with increasing volume the variation of the susceptibilities approaches
a linear behavior in V. The lower three sequences of points obviously tend to
constants. These measurements were done for 3 D 0:313860; 0:401087; 0:498579,
respectively. A power law behavior with nontrivial power (critical exponent) is
observed for 3 D 0:291275, which is the accepted signal for a second-order
transition. The corresponding Higgs self-coupling is the point separating the first-
order and crossover regimes.
Another method for finding the critical endpoint is to study the behavior of the
partition function when is continued to complex values [23]. The partition function
of the complexified theory displays a number of zeros (0 ). With increasing size L,
8.3 Results from the Numerical Simulation of the Electroweak Model Reduced. . . 201
5e-05
4e-05 λ3,crit.=0.2853(48)
3e-05
R 2e-05
1e-05
0.3 0.4 0.5

λ3
Fig. 8.4 Variation of the constant R of (8.27) with 3 extrapolated linearly to R D 0. The
horizontal interval gives the error of the extrapolation but does not reflect the systematic error
from the one-loop mapping (from [13])
the imaginary part of 0;min , the zero lying the closest to the real -axis, exhibits the
scaling behavior
Im.0;min / D CL.1= / C R.3 /: (8.27)
The range of 3 values where R ¤ 0 corresponds to the crossover-type transitions

where the partition function does not contain any nonanalytic behavior. The critical
power exponent characterizes the vanishing of the Higgs mass in the critical
endpoint (R.3c / D 0) as a function of the temperature. In Fig. 8.4, the variation of
the constant R is plotted for four larger values of 3 and is extrapolated to the point
below which it vanishes. The critical endpoint was located as 3c D 0:2853.48/
quite close to the value found from analyzing the field susceptibility. Using the one-
loop relations for the couplings and the corresponding expression of the Higgs mass,
the critical Higgs mass value was estimated as mHc D 75:4.6/.
The intriguing question of the existence of a magnetic screening mass in the
symmetric (high-T) phase has been also investigated. For this study, the asymptotic
behavior of the W W correlator was extracted from a lattice simulation of the
model with Landau-gauge fixing. The variation with of the dimensionless W mass
(MW a), extracted as the inverse characteristic length of the exponential decrease of
the correlator, is seen in Fig. 8.5. For < c , i.e., in the high-T phase, the existence
of a constant screening mass has been clearly demonstrated.
0.9
0.8
0.7
0.6
0.5
mw(κ)
0.4
0.3
0.2
0.1
0 ||
0 0.17 0.172 0.174 κc 0.176 0.178 0.18
κ
Fig. 8.5 Variation of the magnetic mass with the temperature (low values correspond to high
temperatures (from [13]))
8.4 On the High-Accuracy Determination of the Critical

Higgs Mass
The field theory defined in (8.2) is superrenormalizable. This means that only the
operator m23 requires a counterterm, and the couplings 3 ; g23 are renormalization
group invariants within the three-dimensional model. Moreover, only one- and two-
loop Feynman diagrams provide divergent contributions. Therefore, the counterterm
can be determined exactly by analyzing a finite number of diagrams. Its structure is
the following:
ım23 D f1m ˙lin C f2m ˙log : (8.28)
The first term shows up as an induced counterterm in the coefficient of ˚02 when one
expands the one-loop nonstatic + W0 contribution to the effective potential computed
on constant-˚0 background. If one solves the reduced model also perturbatively,
this induced counterterm ensures the finiteness of the potential or of any other
observable. But when one treats the reduced model with the method of lattice field
theory, one finds the necessary counterterm by analyzing the lattice version of the
tadpole integral supplemented with the requirement that it should provide the same
finite part as with momentum cutoff regularization. This gives the last term in the
expression of 1 in (8.25).
The two-loop divergence comes from the so-called setting-sun diagrams, which
are logarithmically divergent. The divergent piece has the following expression
8.4 On the High-Accuracy Determination of the Critical Higgs Mass 203
when calculated on a lattice (compare to Appendix C.2):

1 6
˙log D log C 0:09 : (8.29)
16 2 a3
The finite piece again stems from requiring the same finite parts as calculated with
momentum cutoff regularization. Collecting all the diagrams, for the coefficient of
˙log one obtains
51 4 2
f2m D g C 9BP
3 g3 123 :
BP2
(8.30)
16 3
The renormalization of the logarithmic divergences necessarily introduces a new
scale (independent of the renormalization scale generated in the 4D ! 3D reduction
step), denoted by 3 . One has an exact renormalization group equation
@m23 .3 / 1 1 m
D 2
f2m ! m23 .3 / D 2
f2m log ; (8.31)
@ log 3 16 16 3
where m is a renormalization group invariant constant of integration. This constant

is found by comparing the two-loop effective potential calculated with couplings
2
BP
3 ; m3 in three dimensions and the two-loop finite-temperature potential calculated
directly in four dimensions. This comparison implies that at high temperatures, m
will be predominantly proportional to T. Then, similarly to the optimal choice T
considerably simplifying the expression of mN 23 , one chooses the scale 3 optimally.
There is an additional source of “counterterms” that comes from the difference
of finite two-loop contributions of double-scoop type, calculated respectively in
the continuum and the lattice schemes. These terms for the SU.2/ group give an
additional counterterm to (8.25) [24]:
1 4 2

f2m ˙log C 5:0g 3 C 5:2 3 g 3 : (8.32)
16 2
A high-precision Monte Carlo simulation of the three-dimensional Higgs + gauge
2
model was performed [12] for several values of the ratio x D BP 3 =g3 . The zero-
temperature W-mass was chosen near its physical value by setting v D 246 GeV
and g D 2=3. The variation of the temperature was realized by varying y D m23 =g43 .
These two dimensionless variables have been expressed with the tree-level zero
temperature Higgs mass m2H D v 2 =3 and the ratio m2H =T 2 . Including also the
counterterm contributions just discussed above, the following relations were used
for analyzing the data:
x D 0:00550 C 0:12622h2;
m2
y D 0:39818 C 0:15545h2 0:00190h4 2:58088 T H2 ; (8.33)
where h D mH =.80:6 GeV/.

Fig. 8.6 Variation of the 4

field-length susceptibility
10
with the lattice size for L D 8 βG = 8
to L D 64 for seven different βG = 12 35 GeV 60 GeV
values of 3 (from [12]) βG = 20
3
10
70 GeV
80 GeV
2
χmax
10
1/2
V
95 GeV
1
10
120 GeV
180 GeV
0
10
2 3 4 5
10 10 10 10
2 -3
volume/(g3 )
The simulations covered the range mH 2 35–180 GeV. Three different

lattice spacings were used to control possible discretization errors. The volume-
dependence of the susceptibility (8.26) was studied for lattice sizes L D
12; 16; 24; 32; 48; 64. In Fig. 8.6, the variation of the maximal values max with the
lattice volume is displayed. It is clear that the curves in the mH D 35; 60; 70 GeV
cases signal a first-order transition, while those of 95; 120; 180 GeV correspond to a
crossover. The slope of the mH D 80 GeV data seems to follow a nontrivial scaling
exponent. Therefore, the conclusion was (in agreement with [13]) that the endpoint
is located below mH 80 GeV.
The investigation of the nature of the electroweak symmetry restoration was
performed also in [25] using the reduced three-dimensional effective model. These
authors obtained mHc D 67:0.8/ GeV for the critical Higgs mass. All these
investigations done in the dimensionally reduced effective model are compatible
among each other and also with the result of Csikor et al. obtained by simulating the
original finite-temperature four-dimensional model [26].
The conclusion of the intensive study of the SU.2/ symmetric Higgs + gauge
model in the middle of the 1990s can be best summarized through the phase diagram
of electroweak theory in the Tc MH -plane as it appears in Fig. 8.7.
References 205
Fig. 8.7 The phase diagram

of the SU.2/ symmetric
Higgs + Gauge theory
(RHW D mH =mW , from [26])
References
1. E.W. Kolb, M.S. Turner, The Early Universe (Addison-Wesley, New York, 1990)
2. A. Sakharov, J. Exp. Theor. Phys. 5, 24 (1967)
3. G. t’Hooft, Phys. Rev. D 14, 3432 (1976)
4. N. Manton, Phys. Rev. D 28, 2019 (1983)
5. F. Klinkhamer, N. Manton, Phys. Rev. D 30, 2212 (1984)
6. M. Kobayashi, T. Maskawa, Prog. Theor. Phys. 49, 652 (1973)
7. V.A. Kuzmin, V.A. Rubakov, M.E. Shaposhnikov, Phys. Lett. B 155, 36 (1985)
8. G.W. Anderson, L.J. Hall, Phys. Rev. D 45, 2685 (1992)
9. M. Carrington, Phys. Rev. D 45, 2933 (1992)
10. Z. Fodor, A. Hebecker, Nucl. Phys. B 432, 127 (1994)
11. W. Buchmuller, O. Philipsen, Nucl. Phys. B 443, 47 (1995)
12. K. Kajantie, M. Laine, K. Rummukainen, M. Shaposhnikov, Phys. Rev. Lett. 77, 2887 (1996)
13. F. Karsch, T. Neuhaus, A. Patkós, J. Rank, Proceedings of Lattice 96 international conference.
Nucl. Phys. Proc. Suppl. 53, 623 (1997)
14. K. Farakos, K. Kajantie, K. Rummukainen, M. Shaposhnikov, Nucl. Phys. B 425, 67 (1994)
15. A. Jakovác, K. Kajantie, A. Patkós, Phys. Rev. D 49, 6810 (1994)
16. D.A. Kirzhnits, A.D. Linde, Phys. Lett. B 72, 471 (1972)
17. G. Alexanian, V.P. Nair, Phys. Lett. B 352, 435 (1995)
18. A. Patkós, P. Petreczky, Zs. Szép, Eur. Phys. J. C5, 337 (1998)
19. F. Eberlein, Phys. Lett. B 439, 130 (1998)
20. F. Karsch, T. Neuhaus, A. Patkós, J. Rank, Nucl. Phys. B 474, 217 (1996)
21. I. Montvay, G. Münster, Quantum Fields on Lattice (Cambridge University Press, Cambridge,
1994)
22. H.J. Rothe, Lattice Gauge Theories: An Introduction, 4th edn. (World Scientific, Singapore,
2012)
23. C.N. Yang, T.D. Lee, Phys. Rev. 87, 404 (1952)
24. M. Laine, Nucl. Phys. B 451, 484 (1995)
25. M- Gürtler, E.-M. Ilgenfritz, A. Schiller, Nucl. Phys. Proc. Suppl. 63, 566 (1998)
26. F. Csikor, Z. Fodor, J. Heitger, Phys. Rev. Lett. 82, 21 (1999)
Appendix A
The Spectral Function
We briefly mention some properties of the spectral function.
A.1 Sum Rules
Expressed with operators, we have

1
%AB .x/ D Tr eˇH ŒA.x/; B.0/˛ ; (A.1)
Z0
i.e., this is the expectation of the (anti)commutator of the two operators. For
fundamental fields, the (anti)commutator is just a Dirac delta (cf. (2.8)), which
implies
Z
dk0
%A˘ .t D 0; x/ D iı.x/; %A˘ .k0 ; k/ D iI (A.2)
2
this is the basic form of sum rules. In general, the (anti)commutator of a local
operator at equal times must be proportional to ı.x/, but the proportionality constant
can be complicated. Thus in the generic sum rule, the right-hand side is not just i,
but in addition, a (temperature-dependent) constant can appear.
A.2 Positivity
Using the time translation-invariance of the equilibrium and the properties of the
(anti)commutators, it is easy to show that %AB .k/ D %B A .k/, which means that
%AA .k/ D %AA .k/, i.e., the spectral function is real for B D A . Moreover,

DOI 10.1007/978-3-319-22620-0
208 A The Spectral Function
%AB .k/ D ˛%BA .k/, which means that the spectral function is antisymmetric for
B D A.
Continuing with the real spectral functions %AA , we can insert into the definition
a complete system with states hPnj, where P is the four-vector of the energy and
momentum, and n is the eigenvalue of N, the operator of a conserved charge. We
obtain
X˝ ˇ ˇ ˛
Tr eˇ.HN/ ŒA.x/; A .0/˛ D Pn ˇeˇ.HN/ ŒA.x/; A .0/˛ ˇ Pn
Pn
X ˝ ˛˝ ˇ ˇ ˛
eˇ.En/ ˛eˇ.E n / Pn jA.x/j P0 n0 P0 n0 ˇA .0/ˇ Pn
0 0
D
PP0 nn0
X 0 0
0 ˝ ˛
D eˇ.En/ ˛eˇ.E n / ei.PP /x j Pn jA.0/j P0 n0 j2 ; (A.3)
PP0 nn0
where we have used A.x/ D eiPx A.0/eiPx . If A has a definite charge q, then n0 Cq D
n must be true. In Fourier space, we have
X ˝ ˛
%AA .k/ D eˇ.En/ 1 ˛eˇ.k0 Cq/ .2/4 ı.k C P P0 /j Pn jA.0/j P0 n0 j2 :
PP0 nn0
(A.4)
This form shows that for k0 C q > 0, the spectral function is strictly positive.
The same formula also demonstrates that the spectral function is a sum of Dirac
deltas, at values of k, where two energy eigenstates satisfy the equality P0 D P C k.
At zero temperature and chemical potential, only the E D 0 ground state remains,
where
X
%AA .k0 > 0/jTDD0 D Ck .2/4 ı.k P/; (A.5)
P
where Ck D j hkq jA.0/j 0i j2 . This is proportional to the density of states.

Appendix B
Computation of the Basic Diagrams
Here we summarize the results of the computation of the tadpole and the bubble
diagrams in a one-component bosonic or fermionic theory.
B.1 Tadpole Integral
The contribution of the tadpole diagram (see the first diagram of Fig. 2.2) is
momentum-independent, and so it yields the same result in every formalism. In
real space, the diagram is proportional to G.33/ .x D 0/. This will be considered the
tadpole integral with the definition
T D ˛G.33/ .x D 0/: (B.1)
We can write it in the imaginary time formalism as

Z
X d3 p 1
T D ˛T ; (B.2)
.2/ !n C !p2
3 2
where !n D 2n or .2n C 1/ for the bosonic or fermionic case, respectively. We
do not perform the sum, instead we use the line of argument of Eq. (2.82): using the
definition in (2.42), we obtain
Z
d4 p ˛
T D C n .p
˛ 0 / %0 .p/
.2/4 2
Z
d3 p
D 3
˛ C n˛ .!p / C n˛ .!p C / : (B.3)
.2/ 2!p

DOI 10.1007/978-3-319-22620-0
210 B Computation of the Basic Diagrams
It naturally splits into a zero- and finite-temperature/chemical potential part:
1 ˇ
T D ˛T0 C TT; C TT; ; T0 D T ˇDTD0 : (B.4)
2
Here T0 is UV divergent. We need to regularize the integral in order to be able to
handle it. With momentum cutoff, we have
Z Z
1 d3 p 1 1 dpp2 2 m2 42
T0 D 3
p D p D ln :
2 .2/ 2
p Cm 2 4 2 p 2 C m2 8 2 16 2 em2
0
(B.5)
We can also apply dimensional regularization (cf. Appendix B.3) to compute its
value:1
Z
2" d32" p 2
2 1=2 m2 1 m2
p C m D C E 1 C ln C O."/ ;
2 .2/32" 16 2 " 42
(B.6)
where E is the Euler constant, E D 0:577216.
The quantity TT; is UV finite, due to the exponential suppression of the
distribution functions. For the evaluation, we change the integration variable p !
x D ˇp, and obtain
Z1 p
T2 x2 1
TT; D dx ; D x2 C .ˇm/2 : (B.7)
2 2 e ˇ ˛
0
The only analytic problem with this integral is a pole for the bosonic (˛ D 1)
case for > m. Therefore, the bosonic expression T is defined only in the range
m < < m, just as for the thermodynamics itself.
In the convergent regimes, one can study different limits. The small-temperature
regime is common for the bosonic and fermionic cases:
Z1 p
T2 mT
TT;m D dx exCˇ x2 .ˇm/2 D K1 .ˇm/eˇ
2 2 2 2
ˇm
3=2
1 mT
eˇ.m/ : (B.8)
m 2
1
Usually, it is computed in 4 2" dimensions, using Wick rotation; but the result is the same.
B.1 Tadpole Integral 211
For the leading order of the small-mass expansion, we also can get analytic
expressions by formally substituting m D 0:
Z1
T2 x ˛T 2
TT; .m D 0/ D dx D Li2 .˛eˇ /; (B.9)
2 2 eˇ ex ˛ 2 2
0
where Li2 is the dilogarithm (Spence’s) function. In the bosonic case, the pole is
regularized by principal value integration. Using the identity [1]
2 1 2
Li2 .z/ C Li2 .1=z/ D ln .z/; (B.10)
6 2
one obtains
8 2
ˆ T 2
1 < for ˛ D 1
2
TT; .m D 0/ C TT; .m D 0/ D 122 82 (B.11)
2 :̂ T C for ˛ D 1:
24 8 2
Note that for the ˛ D 1 bosonic case, only jj < m is the physical domain.
The next terms of the small-mass expansion come from the expansion of the
Bose–Einstein or Fermi–Dirac distribution. We will compute the function [2]
Z1
dx e"x 1
I˛" .z/ D ˇ
; 2 D x2 C z2 : (B.12)
e e ˛
0
It is connected to the tadpole as
dTT; 1
2
D 2 I˛ .ˇm/: (B.13)
dm 4
We work out only the D 0 case.
In the bosonic case, the expansion of the distribution function reads
X1
1 1 1

D C 2 : (B.14)
e 1 2 nD1
C .2n/2
2
We note that the first term is the contribution of the zeroth (static) Matsubara mode.
We have
Z1 1
!
x" x" X x"
"
IC .z/ D dx C 2 : (B.15)
2 2 nD1
x2 C z2 C .2n/2
0
In the first term, we can take " D 0, which yields =.2z/. The second term is similar
to the zero-temperature result
Z1
1 z" " 1C" 1 1 z
dx x" .x2 C z2 /1=2 D p . / . /D C ln : (B.16)
2 4 2 2 2" 2 2
0
In the last term, we can take z D 0 and obtain
1 Z
X
1 1
X " .2n/1C"
x"
2 dx 2 D
nD1 0
x C .2n/2 nD1
cos "=2
.2/" .1 "/ 1 E 1

D D C ln.2/; (B.17)
2 cos "=2 2" 2 2
where is the Riemann zeta function. We then have
1 z E
IC .z/ D C ln C : (B.18)
2z 2 4 2
In the fermionic case, we write
X1
1 1

D 2 : (B.19)
e C1 2 nD1
C ..2n C 1//2
2
Here there is no 1= term, since there is no zeroth Matsubara mode. The first term
is the same as before; the second term yields
1 Z
X
1 1
X " .2n C 1/1C"
x" " .2 2" /.1 "/
2 dx 2 2
D D
nD1 0
x C ..2n C 1// nD1
cos "=2 2 cos "=2
1 E 1
D C ln : (B.20)
2" 2 2 2
Then we have
E 1 z
I .z/ D ln : (B.21)
2 2
B.2 The Bubble Diagram 213
To have the result for the tadpole integral, we integrate (B.13). We obtain for
bosons
" #
C T 2 mT m2 m2
TT D ln 2 2 1 ; (B.22)
12 4 16 2 cC T
for fermions

T2 m2 m2
TT D C ln 2 2 1 : (B.23)
24 16 2 c T
The constant terms are
ln cC WD E C ln 4 D 1:95381; cC D 7:05551

ln c WD E C ln D 0:567514; c D 1:7638: (B.24)
Therefore in the high-temperature expansion, we have
2 T 2 mT m2 2
bosons with cutoff W T D C ln 2 2
8 2 12 4 16 2 dC T
" #
2 2
T mT m 1 42
bosons with dim:reg: W T D E C ln 2 2 ;
12 4 16 2 " cC T
2 T2 m2 2
fermions with cutoff W T D 2
C C 2
ln 2 2
8 24 16 d T
2 2
T m 1 42
fermions with dim:reg: W T D C E C ln 2 2 ;
24 16 2 " c T
(B.25)
where
ln dC D E C 1 C ln 2 D 2:26066; dC D 9:58943;

ln d D E C 1 C ln D 0:874367; d D 2:39736: (B.26)
2
B.2 The Bubble Diagram
Similarly to the tadpole case, we just define the basic integral. In imaginary time,
this corresponds to the expression
IE .x/ D ˛G33 .m1 ; x/G33 .m2 ; x/; (B.27)

where ˛ D 1 if both propagating particles are fermions. This formula is true if two
different particles propagate on the two lines. In this note we discuss only the case
in which the particles are identical and there is no chemical potential in the system.
In Fourier space, it reads
Z
X d3 p 1
IE .q/ D ˛T 2 2
: (B.28)
.2/3 .!n2 C !1p /..q0 !n /2 C !2.qp/ /
One can tell immediately the result of the integral for the q D 0 case. There we have
Z
X d3 p 1 @T
IE .q/ D ˛T 3 2 2 2 2
D 2: (B.29)
.2/ .!n C p C m / @m
At zero temperature, we can use (B.5) or (B.6), and we have with cutoff

@ m2 em2 1 42
IETD0 .0/ D 2 ln D ln 1 (B.30)
@m 16 2 42 16 2 em2
and with dimensional regularization

2
@ m 1 m2
IETD0 .0/ D 2 C E 1 C ln
@m 16 2 " 42

1 1 m2
D E ln : (B.31)
16 2 " 42
The finite-temperature correction reads, using (B.13),
1
IET .0/ D I˛ .ˇm; D 0/: (B.32)
4 2
At high temperatures, we can use (B.25) and write
T 1 2
bosons with cutoff W IE .0/ D C ln 2 2
8m 16 2 dC T
" #
T 1 1 42
bosons with dim:reg: W IE .0/ D C E C ln 2 2 ;
8m 16 2 " cC T
1 2
fermions with cutoff W IE .0/ D ln
16 2 d 2 T2

1 1 42
fermions with dim:reg: W IE .0/ D E C ln :
16 2 " c2 T 2
(B.33)
To obtain the momentum-dependence, we change to real time, and write the

retarded expression
Z
d4 p .rr/
iI .ar/ .q/ D 2 4
iG .p/iG.ra/ .q p/ : (B.34)
.2/
This is a causal function, so it can be reproduced from its discontinuity. We obtain

Z
d4 p
Disc iI .ar/ .q/ D %.p/%.q p/.1 C n.p0 / C n.q0 p0 //: (B.35)
q0 .2/4
Using the free spectral function and the notation k D q p, we have
Disc iI .q/
Z
d 3 p .2/
D ı.q0 !p !k / ı.q0 C !p C !k / 1 C np C nk
.2/3 4!p !k

C ı.q0 !p C !k / ı.q0 C !p !k / nk np : (B.36)
We see that Disc iI .q0 ; q/ D Disc iI .q0 ; q/, so it is enough to work out the
q0 > 0 case. We obtain
Z
d 3 p .2/
Disc iI .q0 > 0; q/ D 3
ı.q0 !p !k / 1 C np C nk
.2/ 4!p !k

C2ı.q0 !p C !k / nk np : (B.37)
This integral depends on the angle between p and q. We change from cos ! k
with the help of the relation
k
k2 D q2 C p2 C 2qpx ) pdx D dk; jq pj < k < q C p; (B.38)
q
which means that k; p; q forms a triangle. This means that
Z1
1 pq
Disc iI .k/ D dqdp .k; p; q/ ı.k0 !p !q /.1 C np C nq /
8k !p !q
0

C 2ı.k0 !p C !q /.nq np /
Z1 Z!C
1 p
D dp d!q ı.k0 !p !q /.1 C np C nq /
8k !p
0 !

C 2ı.k0 !p C !q /.nq np /
Z1
1
D d! .!C > k0 ! > ! /.1 C n.!/ C n.k0 !//
8k
m

C2.!C > ! k0 > ! /.n.! k0 / n.!// ; (B.39)
2
where !˙ D .k ˙ p/2 C m2 . The constraints taking into account both theta functions
can be summarized as
!C > jk0 !j > !

jK 2 2k0 !j < 2kp
4K 2 ! 2 4K 2 k0 ! C K 4 4k2 m2 < 0: (B.40)
The positions of the zeros are

r !
˙ 1 4m2
˝ D k0 ˙ k 1 2 : (B.41)
2 K
This is real if K 2 > 4m2 or K 2 < 0. The solution of the inequality then reads
K 2 > 4m2 ) ˝C > ! > ˝

K2 < 0 ) ˝C < ! or ! < ˝ : (B.42)
Since ˝ < 0 if K 2 < 0, this term does not contribute in the integration range
Œm; 1. In the other cases, one can prove that for K 2 > 4m2 , we have ˝ > m, and
for K 2 < 0, we have ˝C > m. Taking into account that we still have k0 ˝C D ˝ ,
we obtain
r
1 2 2 4m2
Disc iI .k/ D .K > 4m / 1 2
8 K
Z˝C
1
C .K 2 > 4m2 / C .K 2 < 0/ d!n.!/: (B.43)
4k
j˝ j
The integral can be performed:

r
1 2 2 4m2
Disc iI .k/ D .K > 4m / 1 2 C
8 K
1 1 eˇ˝C
C .K 2 > 4m2 / C .K 2 < 0/ ln : (B.44)
4k 1 eˇj˝ j
At zero temperature, we have only the first term: this yields a branch cut starting
at 2m with a square-root threshold behavior. This corresponds to the creation of two
real particles. At finite temperature, the amplitude of the branch cut is modified,
and we also obtain a contribution below the light cone (K 2 < 0). This latter term
is allowed because the finite temperature leads to Lorentz invariance-breaking (it
singles out a rest frame). The name of this contribution is Landau damping.
We can study some limiting cases:
• For small q0 , the Landau damping reads
0 s 1
q0 @ q 4m 2
Disc iI .q0 q/ D n 1C 2 A: (B.45)
4q 2 q
• For small q, the finite-temperature correction to the zero-temperature result reads

s
1 4m2 q0
.Q2 > 4m2 / 1 2 n : (B.46)
4 Q 2
• At high temperature (T
m; q0 ; q), the second term reads
ˇ ˇ s
T ˇ q0 C q ˇ 4m2
2 2 2 ˇ
.Q > 4m / C .Q < 0/ ln ˇ ˇ; D 1 :
4q q0 q ˇ Q2
(B.47)
Once we know the discontinuity, we can recover the original function using the
Kramers–Kronig relation. The only problem is that the zero-temperature part goes
to a constant at large q0 value, and therefore, the Kramers–Kronig integral does not
converge. To overcome this difficulty, we calculate the contribution directly at zero
temperature at finite momenta. We write for the 11 part
Z
d4 p 1
iITD0 .q/ D : (B.48)
.2/ .p m C i"/..q p/2 m2 C i"/
4 2 2
Wick rotation is performed with the rule

Z Z
dp0 dp0E
f .p0 / D i f .ip0E /; (B.49)
2 2
and we apply Feynman parameterization:
Z1 Z
d 4 pE 1
ITD0 .q/ D dx : (B.50)
.2/ .pE C m C q2E x.1 x//2
4 2 2
0
We now use 4D momentum cutoff to evaluate this expression. With z D p2E we

obtain
Z1 Z2
1 z
ITD0 .q/ D dx dz
16 2 .z C m2 C q2E x.1 x//2
0 0
Z1
1 2 1 m2 C q2E x.1 x//2
D ln 1 C dx ln : (B.51)
16 2 m2 16 2 m2
0
This means that
Z1
1 m2 q2 x.1 x/
ITD0 .q/ D ITD0 .q D 0/ C dx ln : (B.52)
16 2 m2
0
This formula should be true independently of the chosen regularization. Therefore,

the full expression can be written as
ˇ
Z1 ˇ
1 m2 q2 x.1 x/ ˇˇ
I .ra/ .q/ D ITD0 .q D 0/ C dx ln ˇ
16 2 m2 ˇ
0 q0 Ci"
Z1
d! Disc iIT .!; q/
C ; (B.53)
2 q0 ! C i"
1
where in the second line, only the finite-T part of the discontinuity has to be taken
into account.
B.3 Dimensional Regularization
The basic idea of dimensional regularization is that we evaluate the integrals in

arbitrary dimension and identify the divergences by the poles of the result in going
to integer dimensions.
B.3 Dimensional Regularization 219
The basic input to dimensional regularization is that the surface of a d-

dimensional sphere Kd D 2.d=2/ is an analytic function of the dimension. Therefore,
d=2
a rotationally invariant integrand can be rewritten as

Z Z
2" dd2" p 2.42 /"
D dp pd12"
.2/d2" .4/d=2 .d=2 "/
Z
.42 /"
D dz zd=2"1 ; (B.54)
.4/d=2 .d=2 "/
where in the last term, we performed the change of integration variable z D p2 . Note
that d D 1 is also a possible choice.
We need the quantity of mass dimension to maintain the original engineering
dimension of the integral. Its numerical value is arbitrary, and its appearance is
an unavoidable consequence of the presence of divergences (in the case of cutoff
regularization, the cutoff itself provides this scale).
One of the most common integrals reads
Z1
.b a/ .a/
dz za1 .M 2 C z/b D .M 2 /ab : (B.55)
.b/
0
The properties of the gamma functions are
Z1
.z/ D dx xz1 ex
0
.z C 1/ D z .z/
p
.1/ D .2/ D 1; .1=2/ D
2
1 2
."/ D E C " E C
" 2 12

1 E2 2
.1 C "/ D C E 1 " 1 E C C : (B.56)
" 2 12
Finally, we often need the expansion
"2
X " D e" ln X D 1 C " ln X C .ln X/2 : (B.57)
2
Chapter C
Integrals Relevant for Dimensional Reduction
C.1 Nonstatic Sum-Integrals
The high-temperature expansion of the nonstatic contribution to the free-energy

density of the massive ideal gas in powers of M 2 =T 2 can be related to the mass
expansion of the nonstatic tadpole integral up to an unimportant constant term:
XZ
I.M/ D T p./ log.ˇ 2 .!n2 C p2 C M 2 // D I0 C I1 M 2 C I2 M 4 C : : : ;
n¤0
XZ 1
I 0 .M/ D T p./ D I1 C 2I2 M 2 C 3I3 M 4 C : : : : (C.1)
!n2 C p2 C M 2
n¤0
The three-dimensional integrals are computed with sharp momentum cutoff with
the result [3]
.2/ .1/ .0/ .1/ 1 .1/ 1 .0/ 1

I1 D I1 2 C I1 T C I1 T 2 ; I1 D ; I1 D ; I1 D ;
8 2 2 2 12
.ln/ .0/ .ln/ 1 .0/ 1
I2 D I2 ln C I2 ; I2 D ; I2 D Œ1 C ln.2/ E ;
T 16 2 16 2
.3/
I3 D : (C.2)
192 4 T 2

DOI 10.1007/978-3-319-22620-0
222 C Integrals Relevant for Dimensional Reduction
The nonstatic bubble integral can be reduced to a one-variable integral plus some
analytic terms:
XZ 1
B nonstatic .k/ D T p./
.!n2 C p2 /.!n2 C .p C k/2 /
n¤0
.0/
D 2I2 C B.k=T/ C F.k=/;
Z 1 ˇ ˇ
1 dy y ˇˇ C 2y ˇˇ 1 1 1
B./ D 1 ln C
4 2 0 y ˇ 2y ˇ ey C 1 2 y

2 x
F.x/ D 1 ln 1 C 1: (C.3)
x 2
The expansion coefficients of the “setting-sun” integral are introduced as
X Z ./ Z ./ Z ./

.2/3 ı.p1 C p2 C p3 /ın1 Cn2 Cn3 ;0
2
T
p1 p2 p3 .!l2 C !p21 /.!m2 C !p22 /.!/2m C !p23 /
l;n;m¤0
D K0 C K1 M 2 C K2 M 4 C : : :
(C.4)
.2/ .1/ .ln/ .0/

K0 D K0 2 C K0 T C K0 T 2 ln C K0 T 2 ;
T

.2ln/ 2 .ln/ .0/
K1 D K1 ln C K1 ln C K1 ; (C.5)
T T
and they are computed mostly numerically:
.2/ .1/ .ln/ 5

K0 D 0:0001041333; K0 D 0:0029850437; K0 D
32 2
.0/ .2ln/ 3 .ln/
K0 D 0:0152887686; K1 D ; K1 D 0:001087971: (C.6)
128 4
References 223
C.2 Three-Dimensional Integrals
The integrals that are relevant for the one-loop computation of the effective potential
in three dimensions are the following:
Z
1 1
p./ log.p2 C M 2 / D 2J1 M 2 C 2J0 M 3 ; J1 D ; J0 D
4 2 12
Z ./ Z ./ Z ./
.2/3 ı.p1 C p2 C p3 / 2
D L.ln/ log C L.0/ ;
p1 p2 p3 .p21 2 2 2 2
C M /.p2 C M /.p3 C M /2 9M 2
1
L.ln/ D ; L0 D 0:00670322: (C.7)
32 2
References
1. A. Besser, Finite and p-adic polylogarithms. Compos. Math. 130, 215–223 (2002)
2. L. Dolan, R. Jackiw, Phys. Rev. D 9, 3320 (1974)
3. A. Jakovác, Phys. Rev. D 53, 4538 (1996)

Resummation and Renormalization in Eff Ective Theories of Particle Physics

Uploaded by

Copyright:

Available Formats

Resummation and Renormalization in Eff Ective Theories of Particle Physics

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Resummation and Renormalization in Eff Ective Theories of Particle Physics

Uploaded by

Copyright:

Available Formats

What is the purpose of the Lecture Notes in Physics series?

What is the purpose of the Lecture Notes in Physics series?

What are some properties of the gamma functions that are relevant for dimensional regularization?

What are some properties of the gamma functions that are relevant for dimensional regularization?

Lecture Notes in Physics 912

More information about this series at http://www.springer.com/series/5304

ISSN 0075-8450 ISSN 1616-6361 (electronic)

Springer Cham Heidelberg New York Dordrecht London

Printed on acid-free paper

Springer International Publishing AG Switzerland is part of Springer Science+Business Media

Budapest, Hungary Antal Jakovác

1PI One-particle irreducible

OMS-scheme On-mass-shell scheme

1 Effective Theories from Nuclear to Particle Physics .. . . . . . . . . . . . . . . . . . . . 1

4 Optimized Perturbation Theory.. . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 69

7 Thermodynamics of the Strong Matter . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 161

A The Spectral Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 207

B Computation of the Basic Diagrams . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 209

C Integrals Relevant for Dimensional Reduction . . . . . .. . . . . . . . . . . . . . . . . . . . 221

© Springer International Publishing Switzerland 2016 1

transformation under the SU.2/I group:

The significance of this proposition can be appreciated by neglecting for a moment

The mean field approximation corresponds to

< .x/ >D ˙; meff D mN g ˙: (1.5)

with Vj D . Lj C Rj /=2; Aj D . Rj Lj /=2. A mass term described by

and it obviously has an orthogonal O.8/ symmetry. The generalization of the

This is the matrix version of the linear -model.

M.x/ D S.x/ei j .x/=2F  S.x/UŒ :

the ultraviolet behavior of the couplings of a reduced Higgs–top–bottom Yukawa

1. E. Fermi, Ric. Sci. 2, 2 (1933); Z. Phys. 88, 161 (1934)

21. Y. Nambu, New theories in physics, in Proceedings of the XI International Symposium on

2.1 Review of Classical Field Theory

In classical field theory, the basic object is the field A W M ! V, where V

© Springer International Publishing Switzerland 2016 11

These statements together mean that j is a conserved current when evaluated at

2.2 Quantization and Path Integral

O x/; ˘O .t; y/ ˛ D i„ı.x y/;

Œ˘O .t; x/; ˘O .t; y/ ˛ D 0; (2.8)

where ˛ D ˙1 and ŒX; Y ˛ D XY ˛YX, i.e., commutator for ˛ D 1 and

O x/ j'i D '.t; x/j'i;

Their inner product, using again (2.8), is

The corresponding Grassmann algebra generators are e and e . The complete

This means that either j 0 i D 0 itself, or O j 0 i D 0. Therefore, there exists

j1i D O  j0i; O  j1i D 0; O j1i D O O  j0i D .1 C O  O /j0i D j0i:

because of the nilpotency of . The norm of j i is unity.

Tr AO n .tn / : : : AO 1 .t1 /%;

hN j˘O j'0 i O 0i

We will also encounter the matrix elements of the density matrix

the section tf ! ti . In the contour ordering, segment 2 always follows segment 1,

is used. On segment 2, where the orientation of the integration is tf ! ti , the action

where T is anti-time-ordering, T is time-ordering. In this simplified setup, F has

Fig. 2.1 Closed time path

segment 3 is called Euclidean, or imaginary, time or the Matsubara formalism. The

In the case of nonzero chemical potential, we should use the replacement H !

The following identity is always true:

.11/ .22/ .12/ .21/

To avoid the use of nonindependent quantities, it is sometimes useful to change to

.ra/ .11/ .12/

which leads in Fourier space to the KMS relation:

with Vj D .Lj C Rj /=2; Aj D .Rj Lj /=2. A mass term described by

M.x/ D S.x/ei j .x/=2F S.x/UŒ:

These statements together mean that j is a conserved current when evaluated at

O x/; ˘O .t; y/˛ D i„ı.x y/;

Œ˘O .t; x/; ˘O .t; y/˛ D 0; (2.8)

where ˛ D ˙1 and ŒX; Y˛ D XY ˛YX, i.e., commutator for ˛ D 1 and

This means that either j 0 i D 0 itself, or O j 0 i D 0. Therefore, there exists

j1i D O j0i; O j1i D 0; O j1i D O O j0i D .1 C O O /j0i D j0i:

K .i@/ .x/ D 0; (2.67)

where E D f0 ; i g are self-adjoint 4 4 matrices.

where 0 D iK 1 J. Then, by introducing a new integration variable ! 0 ,

f D " Ts n D p: (2.87)

SŒ D S.2/ŒA C Sint ŒA; (2.91)

In the Euclidean case, we proceed by the replacement i ! E . Since WE is

The value of thus defined has many interesting properties. Diagrammatically,

D S C quantum fluctuations: (2.107)

Let us investigate, in particular, the second derivative of at the physical value.

G D G0 CG0 ˙G D G0 CG˙G0 D G0 CG0 ˙G0 CG0 ˙G0 ˙G0 C : (2.112)

ı hTC AAi D iG0 ı hTC AAiamp iG0 D iG0 ˙G0 C ; (2.113)

.n/ @n ŒA N .n/ @n WŒJ

i ŒA; N D WŒJ; R Jx AN x 1 iRxy .AN x AN y C iG

Let us consider an infinitesimal transformation Ai ! A0i D Ai C "i Ai ŒA for which