Resummation and Renormalization in Eff Ective Theories of Particle Physics
Resummation and Renormalization in Eff Ective Theories of Particle Physics
Resummation and Renormalization in Eff Ective Theories of Particle Physics
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]
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
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.
vii
viii Abbreviations
ix
x Contents
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
(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
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].
4 1 Effective Theories from Nuclear to Particle Physics
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
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)
1 Effective Theories from Nuclear to Particle Physics 5
(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
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
6 1 Effective Theories from Nuclear to Particle Physics
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:
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
1 Effective Theories from Nuclear to Particle Physics 7
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
8 1 Effective Theories from Nuclear to Particle Physics
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
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.
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
(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 /
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.
14 2 Finite Temperature Field Theories: Review
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
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.
2.2 Quantization and Path Integral 15
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)
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)
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)
2.2 Quantization and Path Integral 17
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:
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.
18 2 Finite Temperature Field Theories: Review
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)
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
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
20 2 Finite Temperature Field Theories: Review
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
t i−iβ
2.4 Propagators 21
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)
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:
.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
.12/
iGAB .!/ D ˛eˇ.!Cq/ iG.21/ .!/: (2.49)
.21/ .12/
%AB .x/ D iGAB .x/ iGAB .x/: (2.50)
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
.ra/ .ar/
iGAB .x/ D .t/%AB .x/; iGAB .x/ D .t/%AB .x/: (2.55)
24 2 Finite Temperature Field Theories: Review
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"
.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
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
2nT bosons
!n D (2.62)
.2 C 1/nT fermions:
Zˇ Z
.33/ i!n .33/ d! %AB .!; k/
GAB .!n ; k/ D d e GAB .; k/ D : (2.63)
2 ! i!n
0
.33/
GAB .i!n ! k0 C i"; k/ D Gra
AB .k/: (2.64)
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
which implies that the spectral function % D % also satisfies the equation
K .i@/%.x/ D 0: (2.68)
26 2 Finite Temperature Field Theories: Review
In case of general relativistic theories with a single mass shell, there exists a
differential operator D.i@/ (Klein–Gordon divisor [4]), for which
where %0 .k/ is the spectral function of the one-component bosonic field (Klein–
Gordon field)
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
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
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
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)
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
28 2 Finite Temperature Field Theories: Review
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
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
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,
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
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
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.
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:
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.
32 2 Finite Temperature Field Theories: Review
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
i ˛ˇV D R E
WŒJ D Jx Gxy Jy Tr ln KE C e JA eiSint ŒA 1 : (2.99)
2 2 conn
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
@i
D Jx : (2.103)
@AN x
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)
Together with (2.104), this formula further supports the interpretation of as the
quantum effective action.
34 2 Finite Temperature Field Theories: Review
@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
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
which means
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
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:
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,
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"
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.
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)
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
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)
@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
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
38 2 Finite Temperature Field Theories: Review
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
40 2 Finite Temperature Field Theories: Review
In this subsection we will perform some changes in the integration variable of the
path integral
Z
ZŒJ D DAeiSŒACJA : (2.139)
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
@ @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.
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
Therefore,
ˇ X
@S ˇˇ ˝ ˛
n
Aa1 : : : Aan D i.1/ iC1
ıiak Aa1 : : : Aak1 AakC1 : : : Aan : (2.152)
@A ˇADA.i/ kD1
@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
X
n
˝ ˛
0D Aa1 : : : Aak1 Aak AakC1 : : : Aan : (2.156)
kD1
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)
44 2 Finite Temperature Field Theories: Review
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)
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)
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
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
@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
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
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
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.
52 3 Divergences in 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.
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.
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
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:
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
56 3 Divergences in Perturbation Theory
and
Moreover,
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.
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.
58 3 Divergences in Perturbation Theory
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,
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
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).
60 3 Divergences in Perturbation Theory
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
@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
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 Scale-Dependence and the Callan–Symanzik Equations 61
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
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
62 3 Divergences in Perturbation Theory
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.
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.
3.5 Scale-Dependence and the Callan–Symanzik Equations 63
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.
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
64 3 Divergences in Perturbation Theory
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.
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)
66 3 Divergences in Perturbation Theory
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
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
where the infrared regulator function Rk .q/ obeys the limiting behaviors
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:
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
References
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
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)
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
and we assume that the difference between the original and the arbitrary new
parameters is of higher loop order:
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
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)
74 4 Optimized Perturbation Theory
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
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
76 4 Optimized Perturbation Theory
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.
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
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
78 4 Optimized Perturbation Theory
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 .
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
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)
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-
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.
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.
82 4 Optimized Perturbation Theory
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])
where m2 and m2K are the tree-level masses computed with the optimally chosen
mass parameter m2 :
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)
4.5 Optimized Perturbation Theory for the SU.3/L SU.3/R Symmetric. . . 83
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
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:
These equations are written more explicitly by separating the true one-loop contri-
butions and the result of the optimizing mass shift:
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
84 4 Optimized Perturbation Theory
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 ; /
m2 xy D 2gx; (4.47)
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.
86 4 Optimized Perturbation Theory
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
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:
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)
88 4 Optimized Perturbation Theory
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
This parameterization is used in Chap. 7 for studying the phase diagram of the
quark–meson model at finite temperature and finite baryonic density.
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.
1 ren 4 1 ı 4
L D ' K .i@/' ' C 'ıK .i@/' ' ; (4.63)
2 24 2 24
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
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
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
and ˙div .K ; p/ D ıZp2 ım2 , which is, in fact, just the definition of ıZ and ım2 .
4.7 The 2PI Formalism as Resummation and Its Renormalization 91
˙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 :
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.
4.7 The 2PI Formalism as Resummation and Its Renormalization 93
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/ :
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
where
Z
1
V D C GVG: (4.76)
2
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
@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
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
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
100 5 The Large-N Expansion
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
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/
102 5 The Large-N Expansion
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)
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
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
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
104 5 The Large-N Expansion
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
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
106 5 The Large-N Expansion
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
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
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:
resulting in
R 1
Q.M/
s .qM / D : (5.43)
3 1 R q2M
96 2
ln M02
R 2 1
iG1.M/
D q2M v D 0; (5.45)
3 1 R q2M
96 2
ln M02
108 5 The Large-N Expansion
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.
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
m2 C
D M 2 : (5.50)
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
110 5 The Large-N Expansion
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
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/
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
.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].
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
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
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-
5.5 Renormalization of the O.N/-Model in the Auxiliary Formulation 113
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
ŒG
D I G
.k/: (5.63)
12 d
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
114 5 The Large-N Expansion
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 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)
5.5 Renormalization of the O.N/-Model in the Auxiliary Formulation 115
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.
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
116 5 The Large-N Expansion
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.
5.5 Renormalization of the O.N/-Model in the Auxiliary Formulation 117
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 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/:
Direct analytic integration allows one to determine the difference B0F .p/ BF .p/,
which one needs here with O.1=p2/ accuracy:
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)
)
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
5.5 Renormalization of the O.N/-Model in the Auxiliary Formulation 119
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
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
.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
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
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
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
122 5 The Large-N Expansion
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
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
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
.
124 5 The Large-N Expansion
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.
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
A theory invariant under the global chiral transformation defined with the unitary
transformations UL=R ,
can be constructed with the help of the independent local chiral invariants
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
126 5 The Large-N Expansion
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
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
5.6 Large-n Approximation in U.n/ Symmetric Models 127
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)
128 5 The Large-N Expansion
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
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)
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//
5.6 Large-n Approximation in U.n/ Symmetric Models 129
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)
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)
G1 1 2
.x y/ D Gs .x y/ D . C M /ı.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)
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
5.6 Large-n Approximation in U.n/ Symmetric Models 131
In order to find the divergent contributions explicitly, one has to expand B.k/
BM .k/ B0 .k/ with k4 accuracy:
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
5.6 Large-n Approximation in U.n/ Symmetric Models 133
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
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)
134 5 The Large-N Expansion
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
g 2 x D g 1 y0 : (5.148)
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].
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)
Here the factor Nc reflects the color degeneracy of the quarks and the “color
blindness” of the quark–meson interaction.
136 5 The Large-N Expansion
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)
.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
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
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)
138 5 The Large-N Expansion
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
References
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
.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
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
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.
142 6 Dimensional Reduction and Infrared Improved Treatment of...
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
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
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
144 6 Dimensional Reduction and Infrared Improved Treatment of...
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
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
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
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)
148 6 Dimensional Reduction and Infrared Improved Treatment of...
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)
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
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.
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:
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:
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:
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.
152 6 Dimensional Reduction and Infrared Improved Treatment of...
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 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).
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
6.5 Dimensional Reduction of a Gauge Theory at the One-Loop Level 155
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
156 6 Dimensional Reduction and Infrared Improved Treatment of...
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)
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
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
6.5 Dimensional Reduction of a Gauge Theory at the One-Loop Level 157
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
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
158 6 Dimensional Reduction and Infrared Improved Treatment of...
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:
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
References
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.
R 2 1
iG1 2 2
.p0 / D p0 mG0 f D 0: (7.2)
3N 1 R
6 B .p0 ; mG0 /
F
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?
164 7 Thermodynamics of the Strong Matter
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
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
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
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
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].
.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
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,
168 7 Thermodynamics of the Strong Matter
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).
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
120
TIsing
100
T [MeV]
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])
170 7 Thermodynamics of the Strong Matter
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.
]
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.
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
172 7 Thermodynamics of the Strong Matter
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.
7.3 Equation of State of the Strong Matter 173
...
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
174 7 Thermodynamics of the Strong Matter
1.0
0 MeV 1 loop ; 176 MeV
MS
0.8
0.6
ideal
0.4
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
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
7.3 Equation of State of the Strong Matter 175
Z1 Z1
m=T
P dm %H .m/e D dm em.1=T1=TH / ; (7.20)
0 0
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)
176 7 Thermodynamics of the Strong Matter
! 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
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
7.3 Equation of State of the Strong Matter 177
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.
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
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
7.3 Equation of State of the Strong Matter 179
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
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
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.
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
7.3 Equation of State of the Strong Matter 181
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
182 7 Thermodynamics of the Strong Matter
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
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
7.3 Equation of State of the Strong Matter 183
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])
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.
184 7 Thermodynamics of the Strong Matter
ρ
100
visibility
50
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
ω
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
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.
7.3 Equation of State of the Strong Matter 185
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].
186 7 Thermodynamics of the Strong Matter
4
p/T
2
0
0 100 200 300 400 500 600 700 800
T (MeV)
References
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
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.
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.
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:
(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
192 8 Finite-Temperature Restoration of the Brout–Englert–Higgs Effect
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.
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
194 8 Finite-Temperature Restoration of the Brout–Englert–Higgs Effect
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
8.2 Optimized Perturbation Theory for the Electroweak Phase Transition 195
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.
196 8 Finite-Temperature Restoration of the Brout–Englert–Higgs Effect
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.
8.2 Optimized Perturbation Theory for the Electroweak Phase Transition 197
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])
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.
200 8 Finite-Temperature Restoration of the Brout–Englert–Higgs Effect
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
5e-05
4e-05 λ3,crit.=0.2853(48)
3e-05
R 2e-05
1e-05
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
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]))
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:
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
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
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)
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 )
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
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,
%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
Here we summarize the results of the computation of the tadpole and the bubble
diagrams in a one-component bosonic or fermionic theory.
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
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
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
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
212 B Computation of the Basic Diagrams
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
1 Z
X
1 1
X " .2n/1C"
x"
2 dx 2 D
nD1 0
x C .2n/2 nD1
cos "=2
1 z E
IC .z/ D C ln C : (B.18)
2z 2 4 2
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
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
Similarly to the tadpole case, we just define the basic integral. In imaginary time,
this corresponds to the expression
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
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)
B.2 The Bubble Diagram 215
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
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 /
216 B Computation of the Basic Diagrams
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
This is real if K 2 > 4m2 or K 2 < 0. The solution of the inequality then reads
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
B.2 The Bubble Diagram 217
• 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
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
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
Z1
1 m2 q2 x.1 x/
ITD0 .q/ D ITD0 .q D 0/ C dx ln : (B.52)
16 2 m2
0
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.
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
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
"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
The three-dimensional integrals are computed with sharp momentum cutoff with
the result [3]
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
D K0 C K1 M 2 C K2 M 4 C : : :
(C.4)
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)