Prepared for submission to JHEP BI-TP 2014/06

Three-loop HTLpt thermodynamics at finite

temperature and chemical potential
arXiv:1402.6907v2 [hep-ph] 12 May 2014

Najmul Haque,a Aritra Bandyopadhyay,a Jens O. Andersen,b Munshi G. Mustafa,a

Michael Strickland,c and Nan Sud
a
Theory Division, Saha Institute of Nuclear Physics, 1/AF Bidhannagar, Kolkata-700107, India
b
Department of Physics, Norwegian University of Science and Technology, N-7491 Trondheim,
Norway
c
Department of Physics, Kent State University, Kent, Ohio 44242, United States
d
Faculty of Physics, University of Bielefeld, D-33615 Bielefeld, Germany

Abstract: We calculate the three-loop thermodynamic potential of QCD at finite

temperature and chemical potential(s) using the hard-thermal-loop perturbation theory
(HTLpt) reorganization of finite temperature and density QCD. The resulting analytic
thermodynamic potential allows us to compute the pressure, energy density, and entropy
density of the quark-gluon plasma. Using these we calculate the trace anomaly, speed of
sound, and second-, fourth-, and sixth-order quark number susceptibilities. For all observ-
ables considered we find good agreement between our three-loop HTLpt calculations and
available lattice data for temperatures above approximately 300 MeV.
Contents

1 Introduction 2

2 Hard-thermal-loop perturbation theory 4

3 Contributions to the HTLpt thermodynamic potential through NNLO 5

4 NNLO HTLpt thermodynamic potential 7

4.1 NNLO result for equal chemical potentials 7
4.2 NNLO result – General case 11

5 Mass prescription 12

6 Thermodynamic functions 13
6.1 Running coupling 13
6.2 Scales 13
6.3 Pressure 14
6.4 Energy density 15
6.5 Entropy density 16
6.6 Trace anomaly 17
6.7 Speed of sound 17

7 Quark number susceptibilities 18

7.1 Baryon number susceptibilities 19
7.2 Single quark number susceptibilities 21

8 Conclusions and outlook 23

A Expansion in mass parameters 25

A.1 One-loop sum-integrals 25
A.2 Two-loop sum-integrals 27
A.3 Three-loop sum-integrals 30

B Sum-Integrals 35
B.1 One-loop sum-integrals 35
B.2 Two-loop sum-integrals 37
B.3 Three-loop sum-integrals 37

C Three-dimensional integrals 38
C.1 One-loop integrals 38
C.2 Two-loop integrals 39

–1–
D Properties of the ℵ functions 39

1 Introduction

Quantum chromodynamics (QCD) describes the propagation and interaction of quarks

and gluons which are believed to be the fundamental constituents of all hadronic matter.
Based solely on the QCD Lagrangian it is possible to calculate the finite temperature and
chemical potential partition function of QCD which results in the so-called equation of state
(EoS). The determination of the QCD EoS is extremely important to the phenomenology
of the quark-gluon plasma (QGP). At this time, the most reliable method to calculate
the QCD thermodynamic functions at finite temperature and zero chemical potential is
lattice gauge theory (see e.g.[1–15]). Importantly, lattice QCD can be used to probe the
behavior of QCD matter near the transition temperature where QCD matter undergoes a
phase transition from the hadronic phase to the deconfined QGP phase. Near the phase
transition, the running coupling is large and non-perturbative methods like lattice QCD
must be used. Finite temperature lattice QCD calculations are now quite sound; however,
due to the sign problem, it is not straightforward to extend such calculations to finite
baryon chemical potential. In practice, it is possible to obtain information about the
behavior of the thermodynamic functions at small baryon chemical potential by making
a Taylor expansion of the partition function around µB = 0 and extrapolating the result.
This requires the calculation of various quark-number susceptibilities evaluated at zero
chemical potential.
Since extrapolations based on a finite number of Taylor coefficients can only be trusted
within the radius of convergence of the expansion, it would be nice to have an alternative
framework for calculating the finite temperature and chemical potential QCD thermody-
namic potential and associated quantities. This is important in light of the ongoing beam
energy scan at the Relativistic Heavy Ion Collider (RHIC) and the forthcoming experi-
ments at the Facility for Antiproton and Ion Research (FAIR). As an alternative to lattice
QCD calculations, one natural option is to compute the thermodynamic potential using
perturbation theory. In principle, this should work since, at sufficiently high temperature,
the value of the strong coupling constant is small; however, one does not know a priori how
large the temperature should be for this method to result in a good approximation to real-
ity. The calculation of the thermodynamic potential using the weak-coupling expansion in
the strong coupling constant, g, has a long history [16–24] and the perturbative expansion
of the pressure of QCD at both zero [25] and non-zero chemical potential [26–28] are now
known through order g6 ln g.
Unfortunately, it turns out that a strict expansion in the coupling constant converges
only for temperatures many orders of magnitude higher than those relevant for heavy-ion
collision experiments. The source of the poor convergence comes from contributions from
soft momenta, p ∼ gT . This suggests that one needs a way of reorganizing the perturbative
series which treats the soft sector more carefully. There are various ways of reorganizing

–2–
the finite temperature/chemical potential perturbative series. For scalar field theories one
can use “screened perturbation theory” (SPT) [29–33] which was inspired in part by vari-
ational perturbation theory (VPT) [34–39]. For gauge theories, however, it is not possible
to use a scalar gluon mass. As a result, a gauge-invariant generalization of SPT called
hard-thermal-loop perturbation theory (HTLpt) was developed. HTLpt has been used to
calculate thermodynamic functions at one loop HTLpt [40–44], at two loops [45–48], and
at three loops at zero chemical potential [49–54] as well as at finite chemical potential [55].
Application of some hard-thermal-loop motivated approaches can be found in [56–67]. In
addition to lattice QCD and HTLpt, there are also various model calculations on the mar-
ket. For example, the Nambu-Jona-Lasinio (NJL) [68, 69], Polyakov-loop extended Nambu-
Jona-Lasinio (PNJL) [70–80], quasi-particle models [81–85], Polyakov-loop extended quark
meson (PQM) model [86–88] have been used to calculate various thermodynamic functions.
There have also been works which apply Pade and Borel-Pade methods to the perturbative
QCD pressure [90–92]. Finally, we note that some results from holographic QCD for the
quark number susceptibilities can be found in refs. [93, 94].
In this paper we calculate the thermodynamic potential at finite temperature and
chemical potential to three-loop order in HTLpt. The result for equal quark chemical
potentials was first presented in ref. [55]. Herein, we present the details of this calculation
and extend our results to the case that the quarks can possess flavor-dependent chemical
potentials. The resulting three-loop thermodynamic potential is renormalized using only
known vacuum, mass, and coupling constant counterterms and the final result is completely
analytic and gauge independent. The resulting analytic thermodynamic potential is then
used to obtain expressions for the pressure, energy density, entropy density, trace anomaly,
speed of sound, and various quark number susceptibilities. We find that there is good
agreement between our NNLO HTLpt results and lattice data down to temperatures on
the order of 300 MeV.
The paper is organized as follows. In section 2 we specify the HTLpt calculational
framework and the necessary counterterms to renormalize HTLpt. In section 3 we discuss
the diagrams that contribute to the HTLpt thermodynamic potential through NNLO. In
section 4 we present our final results for the NNLO thermodynamic potential. In section 5
we discuss the mass prescription for the in-medium masses mD and mq . We present our
results for the thermodynamic functions and compare them with results from lattice gauge
simulations in section 6. In section 7 we present our results for the second-, fourth-, and
sixth-order baryon and quark number susceptibilities. We also compare our results for
these quantities with available lattice data. In section 8 we summarize and conclude. In
appendix A the necessary diagrams are reduced to scalar sum-integrals and expanded in
powers of mD /T and mq /T . We list the necessary non-trivial sum-integrals and integrals
in appendices B and C. Finally, in appendix D we list some properties of the ℵ functions
which appear repeatedly in finite density calculations.

–3–
2 Hard-thermal-loop perturbation theory

The QCD Lagrangian density in Minkowski space can be written as

1
LQCD = − Tr[Gµν Gµν ] + iψ̄γ µ Dµ ψ + Lgh + Lgf + ∆LQCD , (2.1)
2
where the field strength is Gµν = ∂ µ Aν − ∂ ν Aµ − ig[Aµ , Aν ] and the covariant derivative
is D µ = ∂ µ − igAµ . The term ∆LQCD contains the counterterms necessary to cancel
ultraviolet divergences in perturbative calculations. The ghost term Lgh depends on the
form of the gauge-fixing
h iterm Lgf . In this paper we work in general covariant gauge where
Lgf = −ξ −1 Tr (∂µ Aµ )2 with ξ being the gauge-fixing parameter.
Hard-thermal-loop perturbation theory is a reorganization of in-medium perturbation
theory for QCD. The HTLpt Lagrangian density can be written as

L = (LQCD + LHTL )|g→√δg + ∆LHTL , (2.2)

where the HTL improvement term is [95]

    !
yµ 1 y α yβ
LHTL = (1 − δ)im2q ψ̄γ µ ψ − (1 − δ)m2D Tr Gµα Gµβ , (2.3)
y ·D ŷ 2 (y ·D)2 ŷ

where y µ = (1, ŷ) is a light-like four-vector with ŷ being a three-dimensional unit vector
and the angular bracket indicates an average over the direction of ŷ. The two parameters
mD and mq can be identified with the Debye screening mass and the thermal quark mass,
respectively, and account for screening effects. HTLpt is defined by treating δ as a formal
expansion parameter. By coupling the HTL improvement term (2.3) to the QCD La-
grangian (2.1), HTLpt systematically shifts the perturbative expansion from being around
an ideal gas of massless particles to being around a gas of massive quasiparticles which are
the appropriate physical degrees of freedom at high temperature and/or chemical potential.
The HTLpt Lagrangian (2.2) reduces to the QCD Lagrangian (2.1) if we set δ = 1.
Physical observables are calculated in HTLpt by expanding in powers of δ, truncating
at some specified order, and then setting δ = 1. This defines a reorganization of the
perturbative series in which the effects of m2D and m2q terms in (2.3) are included to leading
order but then systematically subtracted out at higher orders in perturbation theory by
the δm2D and δm2q terms in (2.3). To obtain leading order (LO), next-to-leading order
(NLO), and next-to-next-leading order (NNLO) results, one expands to orders δ0 , δ1 , δ2 ,
respectively. Note that HTLpt is gauge invariant order-by-order in the δ expansion and,
consequently, the results obtained are independent of the gauge-fixing parameter ξ.
If the expansion in δ could be calculated to all orders, the final result would not de-
pend on mD and mq when we set δ = 1. However, any truncation of the expansion in δ
produces results that depend on mD and mq . As a consequence, a prescription is required
to determine mD and mq as a function of T , µ and αs . Several prescriptions had been
discussed in [53] at zero chemical potential. The HTLpt expansion generates additional
ultraviolet divergences. In QCD perturbation theory, renormalizability constrains the ul-
traviolet divergences to have a form that can be cancelled by the counterterm Lagrangian

–4–
∆LQCD . We will demonstrate that the renormalization of HTLpt can be implemented by
including a counterterm Lagrangian ∆LHTL among the interaction terms in (2.3). There
is no all-order proof that the HTL perturbation expansion is renormalizable, so the gen-
eral structure of the ultraviolet divergences is unknown. However, as shown previously in
refs. [45–47, 53], it is possible to renormalize the NNLO HTLpt thermodynamic potential
using only a vacuum counterterm, a Debye mass counterterm, a fermion mass counterterm,
and a coupling constant counterterm. The necessary counterterms for renormalization of
the NNLO thermodynamic potential are
dA
∆E0 = (1 − δ)2 m4D , (2.4)
128π 2 ǫ
11cA − 4sF
∆m2D = αs δ(1 − δ)m2D , (2.5)
12πǫ
3 dA
∆m2q = αs δ(1 − δ)m2q , (2.6)
8πǫ cA
11cA − 4sF 2 2
δ∆αs = − αs δ , (2.7)
12πǫ
where, with the standard normalization, the QCD Casimir numbers are cA = Nc , dA =
Nc2 − 1, sF = Nf /2, dF = Nc Nf , and s2F = CF sf with CF = (Nc2 − 1)/2Nc . Note that the
coupling constant counterterm (2.7) is consistent with one-loop running of αs .
In practice, in addition to the δ expansion, it is also necessary to make a Taylor
expansion in the mass parameters scaled by the temperature, mD /T and mq /T , in order
to obtain analytically tractable sum-integrals. An added benefit of this procedure is that
the final result obtained at NNLO is completely analytic. In order to truncate the series
in mD /T and mq /T one treats these quantities as being O(g) at leading order, keeping all
terms that naively contribute to the thermodynamic potential through O(g5 ). In practice,
such an truncated expansion works well [33, 44] and the radius of convergence of the scaled
mass expansion seems to be quite large, giving us confidence in this approximate treatment
of the necessary sum-integrals.
In addition to calculations of the thermodynamic potential, hard-thermal-loop pertur-
bation theory has been used to calculate various physical quantities which are relevant to
the deconfined state of matter. Quantities such as the dilepton production rate [96, 97],
photon production rate [98], single quark and quark anti-quark potentials [99–107], fermion
damping rate [108–110], photon damping rate [111], gluon damping rate [112, 113], jet en-
ergy loss [114–125], plasma instabilities [126–132], thermal axion production [133], and
lepton asymmetry during leptogenesis [134, 135] have also been calculated using HTLpt.
We note, however, that most of the papers above have only worked at what we would call
leading order in HTLpt.

3 Contributions to the HTLpt thermodynamic potential through NNLO

The diagrams needed for the computation of the HTLpt thermodynamic potential through
NNLO can be found in figures 2 and 3 of ref. [53]. In ref. [53] the authors computed the
NNLO thermodynamic potential at zero chemical potential. Here we extend the NNLO

–5–
calculation to finite chemical potential.1 For this purpose, one needs to only consider
diagrams which contain at least one quark propagator; however, for completeness we also
list the purely gluonic contributions below. In the results we will express thermodynamic
quantities in terms of two dimensionless variables: m̂D = mD /(2πT ) and µ̂ = µ/(2πT ).
The complete NNLO HTLpt thermodynamic potential can be expressed in terms of
these diagrams as
 g h i
g g g  f f f
ΩNNLO = dA F1a + F1b + F2d + F3m + dF F1b + F2d + F3i
h i
g g g g g g g g
+dA cA F2a + F2b + F2c + F3h + F3i + F3j + F3k + F3l
h i
f f f f f f f f
+dA sF F2a + F2b + F3d + F3e + F3f + F3g + F3k + F3l
h i h i
g g g g g g g f f
+dA c2A F3a + F3b + F3c + F3d + F3e + F3f + F3g + dA s2F F3a + F3b
h 1 i h i
f f f f f f
+dA cA sF − F3a + F3m + F3n + F3o + dA s2F F3c + F3j
2
∂ ∂
+∆0 E0 + ∆1 E0 + ∆2 E0 + ∆1 m2D 2 ΩLO + ∆1 m2q ΩLO
∂mD ∂m2q
∂ ∂ ∂ ∂
+∆2 m2D 2 ΩLO + ∆2 m2q 2
ΩLO + ∆1 m2D 2 ΩNLO + ∆1 m2q ΩNLO
∂mD ∂mq ∂mD ∂m2q
   
1 ∂2 2 2

1 ∂2
2
+ 2 2
ΩLO ∆1 mD + 2 2
ΩLO ∆1 m2q
2 (∂mD ) 2 (∂mq )
" g f
#
cA F2a+2b+2c + sF F2a+2b
+dA ∆1 αs , (3.1)
αs

where the necessary counterterms at any order in δ can be calculated using eqs. (2.4)-(2.7).
The expressions for the one- and two-loop diagrams above can be found in refs. [45, 46].
g g
The expressions for the three-loop bosonic diagrams F3a –F3m are presented in section 3
f f
of ref. [50], and the three-loop diagrams with fermions F3a –F3i can be found in section
3 of ref. [51]. The three-loop diagrams specific to QCD, i.e., the non-Abelian diagrams
involving quarks, are given by
Z h i
f 1P
F3m = Tr Γα (R − P, R, P )S(P )Γβ (P − Q, P, Q)S(Q)Γγ (Q − R, Q, R)S(R)
6 {P QR}
×Γµνδ (P − R, Q − P, R − Q)∆αµ (P − R)∆βν (Q − P )∆γδ (R − Q) , (3.2)
Z
f P µν
F3n = − Π̄g (P )∆να (P )Π̄αβ βµ
f (P )∆ (P ) , (3.3)
P Z
1 P h i
f
F3o = − g2 Tr Γαβ (P, −P, Q, Q)S(Q) ∆αµ (P )∆βν (P )Π̄µν
g (P ) , (3.4)
2 P {Q}
1
Some additional details concerning the LO and NLO finite chemical potential calculations can be found
in refs. [43, 44] and [47].

–6–
where
Z
1 2 P µν,αβ
Π̄µν
g (P ) = g Γ (P, −P, Q, −Q)∆αβ (Q)
2 Q
Z
1 P
+ g2 Γµαβ (P, Q, −P − Q)∆αβ (Q)Γνγδ (P, Q, −P − Q)∆γδ (−P − Q)
2 Q
Z
P Qµ (P + Q)ν
+g2 2 2
, (3.5)
Q Q (P + Q)
Z
µν 2P
Π̄f (P ) = −g Tr [Γµ (P, Q, Q − P )S(Q)Γν (P, Q, Q − P )S(Q − P )] . (3.6)
{Q}

Thus Π̄µν (P ) is the one-loop gluon self-energy with HTL-resummed propagators and ver-
tices as in ref. [53]:
µν
Π̄µν (P ) = cA Π̄µν
g (P ) + sF Π̄f (P ) . (3.7)

4 NNLO HTLpt thermodynamic potential

One can evaluate the sum-integrals necessary analytically by expanding in the ratios mD /T
and mq /T . For details concerning this expansion and intermediate results, we refer the
reader to appendix A. We consider first the case that all quarks have the same chemical
potential µf = µ = µB /3 where f is a flavor index and µf ∈ {µu , µd , µs , · · · , µNf }. Af-
ter presenting the steps needed for this case, we present the general result with separate
chemical potentials for each quark flavor.

4.1 NNLO result for equal chemical potentials

When all quarks have the same chemical potential µf = µ = µB /3 we can straightforwardly
combine the results for the various sum-integrals. In this case, the unrenormalized three-
loop HTLpt thermodynamic potential is
  
Ω3loop 7 dF 120 2 240 4 sF αs 5


= 1+ µ̂ + µ̂ + − 1 + 12µ̂2 5 + 12µ̂2
Ω0 4 dA 7 7 π 8
  
15 2

15 1 Λ̂ 3 2
+ 1 + 12µ̂ m̂D + − 1 − ℵ(z) + 4 ln − 2 ln m̂D m̂D − 90m̂q m̂D
2 2 ǫ 2

 α 2 15  ′
s ζ (−1) ζ ′ (−1) 2
+ s2F 35 − 32 + 472µ̂2 + 384 µ̂ + 1328µ̂4
π 64 ζ(−1) ζ(−1)
 
2 2
+64 − 36iµ̂ℵ(2, z) + 6(1 + 8µ̂ )ℵ(1, z) + 3iµ̂(1 + 4µ̂ )ℵ(0, z)

45 2


− m̂D 1 + 12µ̂
2
(4.1)

–7–
 "
 s α 2 5
2
m̂2q
F s
+ 1 + 12µ̂2 + 30 1 + 12µ̂2
π 4m̂D m̂D
(  !
25 72 2 144 4 1 Λ̂ 31 6 68 ζ ′ (−3)
+ 1 + µ̂ + µ̂ + 6 ln + + γE −
24 5 5 ǫ 2 10 5 25 ζ(−3)
12 24 8 ζ ′ (−1)
+ (25 + 12γE )µ̂2 + (61 + 36γE )µ̂4 − (1 + 12µ̂2 )
5 5 5 ζ(−1)
144 h
− 8ℵ(3, z) + 3ℵ(3, 2z) + 12iµ̂ (ℵ(2, z) + ℵ(2, 2z)) − 12µ̂2 ℵ(1, 2z)
5 )
i
− iµ̂(1 + 12µ̂2 ) ℵ(0, z) − (3 + 20µ̂2 )ℵ(1, z)
( !
15 2

1 Λ̂
− 1 + 12µ̂ + 4 ln − 2 ln m̂D
2 ǫ 2
  ) #
4
+(1 + 12µ̂2 ) − ℵ(z) + 24ℵ(1, z) m̂D
3
" (  !
 c α   s α  15
235 792 2 1584 4 1 Λ̂
A s F s 2
+ 1 + 12µ̂ − 1+ µ̂ + µ̂ + 6 ln
3π π 2m̂D 32 47 47 ǫ 2
 
1809 8600 2 28720 4 48γE
32
ζ ′ (−1)
+ 1+ µ̂ + µ̂ − 1 + 12µ̂2 − 1 + 6µ̂2
470 603 603 47 47 ζ(−1)
′
464 ζ (−3) 288
288 h
− − 1 + 12µ̂2 ln m̂D − 2iµ̂ℵ(0, z)
235 ζ(−3) 47 47
)

i
2
− 3 + 68µ̂ ℵ(1, z) + 72iµ̂ℵ(2, z) + 26ℵ(3, z)
(  !
315 132 2 1 Λ̂ 88 440 2 22

+ 1+ µ̂ + 6 ln − 2 ln m̂D + + µ̂ + 1 + 12µ̂2 γE
8 7 ǫ 2 21 7 7
) #
8 ζ ′ (−1) 4 264 m̂2q Ω3loop,YM
− + ℵ(z) + ℵ(1, z) m̂D + 90 + , (4.2)
7 ζ(−1) 7 7 m̂D Ω0

Ω
where Ω0 = −dA π 2 T 4 /45 and 3loop,YM
Ω0 is the pure Yang Mills unrenormalized three-loop
thermodynamic potential [53]. Above, ℵ(z) = Ψ(z) + Ψ(z ∗ ) with z = 1/2 − iµ̂ and Ψ being
the digamma function (see app. D for more details and useful properties of ℵ(z)).

–8–
 The sum of all counterterms through order δ2 is
∆Ω ∆Ω1 + ∆Ω2
=
Ω0 Ω0
" ! #
sF αs 15 1 Λ̂ 3
= − + 2 ln − 2 ln m̂D m̂D
π 2 ǫ 2
" (  !
 c α   s α  235 792 2 1584 4 1 Λ̂ 56 ζ ′ (−1)
A s F s
+ 1+ µ̂ + µ̂ + 4 ln +
3π π 32 47 47 ǫ 2 47 ζ(−1)
  )
149 2376 2 4752 4 1584
1056 ζ ′ (−1)
+ 1+ µ̂ + µ̂ + 1 + 4µ̂2 ℵ(1, z) + µ̂2
47 149 149 47 47 ζ(−1)
(  !
315 132 2 1 Λ̂ 8 ζ ′ (−1) 61
− 1+ µ̂ + 4 ln − 2 ln m̂D − + + 44µ̂2
8 7 ǫ 2 7 ζ(−1) 21
) # " (  !
264  s α 2 25 72 2 144 4 1 Λ̂
F s
+ ℵ(1, z) m̂D + − 1 + µ̂ + µ̂ + 4 ln + 3
7 π 24 5 5 ǫ 2
)
144
8
ζ ′ (−1)
+ 1 + 4µ̂2 ℵ(1, z) + 1 + 12µ̂2
5 5 ζ(−1)
( ! ) #
15
1 Λ̂ 7 ∆ΩYM
+ 1 + 12µ̂2 + 4 ln − 2 ln m̂D + + 24ℵ(1, z) m̂D + ,
2 ǫ 2 3 Ω0
(4.3)

where ∆ΩYM is the pure-glue three-loop HTLpt counterterm [53]

!
∆ΩYM 45 4 495  cA αs  1 Λ̂g
= m̂ + + 2 ln − 2 ln m̂D m̂3D
Ω0 8ǫ D 8 3π ǫ 2
" !
 c α 2 165 1 Λ̂g ζ ′ (−1)
A s
+ + 4 ln +2+4
3π 16 ǫ 2 ζ(−1)
! #
1485 1 Λ̂g 4 ζ ′ (−1)
− + 4 ln − 2 ln m̂D + + 2 m̂D . (4.4)
8 ǫ 2 3 ζ(−1)

Adding the total three-loop HTLpt counterterm (4.3) to the unrenormalized three-loop
HTLpt thermodynamic potential (4.2) we obtain our final result for the NNLO HTLpt

–9–
thermodynamic potential in the case that all quarks have the same chemical potential
  
ΩNNLO 7 dF 120 2 240 4 sF αs 5


= 1+ µ̂ + µ̂ − 1 + 12µ̂2 5 + 12µ̂2
Ω0 4 dA 7 7 π 8
  
15 2

15 Λ̂ 3 2
− 1 + 12µ̂ m̂D − 2 ln − 1 − ℵ(z) m̂D + 90m̂q m̂D
2 2 2

 α 2 15 
s
ζ ′ (−1)
+ s2F 35 − 32 1 − 12µ̂2 + 472µ̂2 + 1328µ̂4
π 64 ζ(−1)
 
2 2
+64 − 36iµ̂ℵ(2, z) + 6(1 + 8µ̂ )ℵ(1, z) + 3iµ̂(1 + 4µ̂ )ℵ(0, z)
  "
45 2

sF αs 2 5
2
m̂2q
− m̂D 1 + 12µ̂ + 1 + 12µ̂2 + 30 1 + 12µ̂2
2 π 4m̂D m̂D
( 
25 72 144 4 Λ̂ 1
3
2
+ 1 + µ̂2 + µ̂ ln + 1 + 168µ̂2 + 2064µ̂4 + 1 + 12µ̂2 γE
12 5 5 2 20 5
8 ζ ′ (−1) 34 ζ ′ (−3) 72 h
− (1 + 12µ̂2 ) − − 8ℵ(3, z) + 3ℵ(3, 2z) − 12µ̂2 ℵ(1, 2z)
5 ζ(−1) 25 ζ(−3) 5
)
i
+12iµ̂ (ℵ(2, z) + ℵ(2, 2z)) − iµ̂(1 + 12µ̂2 ) ℵ(0, z) − 2(1 + 8µ̂2 )ℵ(1, z)
( !) #
15 2

Λ̂
− 1 + 12µ̂ 2 ln − 1 − ℵ(z) m̂D
2 2
" ( 
 c α   s α  15
235 792 2 1584 4 Λ̂
A s F s 2
+ 1 + 12µ̂ − 1+ µ̂ + µ̂ ln
3π π 2m̂D 16 47 47 2
 
144
319 2040 2 38640 4 24γE

− 1 + 12µ̂2 ln m̂D + 1+ µ̂ + µ̂ − 1 + 12µ2
47 940 319 319 47
  ′ ′ h
44 156 2 ζ (−1) 268 ζ (−3) 72
− 1+ µ̂ − − 4iµ̂ℵ(0, z)
47 11 ζ(−1) 235 ζ(−3) 47
)

i m̂2q
+ 5 − 92µ̂2 ℵ(1, z) + 144iµ̂ℵ(2, z) + 52ℵ(3, z) + 90
m̂D
(   
315 132 2 Λ̂ 11 2

9 132 2
+ 1+ µ̂ ln + 1 + 12µ̂ γE + 1+ µ̂
4 7 2 7 14 9
) #
2 ΩYM
+ ℵ(z) m̂D + NNLO . (4.5)
7 Ω0

– 10 –
where ΩYM
NNLO is the NNLO pure-glue thermodynamic potential [50]
2

" ! #
ΩYM
NNLO 15 3 cA αs 15 45 135 2 495 Λ̂g 5 3
= 1 − m̂D + − + m̂D − m̂ − ln + + γE m̂D
Ω0 4 3π 4 2 2 D 4 2 22
" !
 c α 2 45 165 Λ̂g 72 84 6 74 ζ ′ (−1) 19 ζ ′ (−3)
A s
+ − ln − ln m̂D − − γE − +
3π 4m̂D 8 2 11 55 11 11 ζ(−1) 11 ζ(−3)
! #
1485 Λ̂g 79 π2
+ ln − + γE + ln 2 − m̂D , (4.6)
4 2 44 11

The result contained in eq. (4.5) was first presented in ref. [55]. Note that the full thermo-
dynamic potential (4.5) reduces to thermodynamic potential of ref. [53] in the limit µ → 0.
In addition, the above thermodynamic potential produces the correct O(g5 ) perturbative
result when expanded in a strict power series in g [26, 27].3

4.2 NNLO result – General case

It is relatively straightforward to generalize the previously obtained result (4.5) to the case
that each quark has a separate chemical potential µf . The final result is
  
ΩNNLO 7 dF 1 X 120 2 240 4 sF αs 1 X 5


= 1+ µ̂f + µ̂f − 1 + 12µ̂2f 5 + 12µ̂2f
Ω0 4 dA Nf 7 7 π Nf 8
f f
  
15 2

15 Λ̂ 3 2
− 1 + 12µ̂f m̂D − 2 ln − 1 − ℵ(zf ) m̂D + 90m̂q m̂D
2 2 2
 
s2F  αs 2 X 15
ζ ′ (−1)
+ 35 − 32 1 − 12µ̂2f + 472µ̂2f + 1328µ̂4f
Nf π 64 ζ(−1)
f
 
2 2
+64 − 36iµ̂f ℵ(2, zf ) + 6(1 + 8µ̂f )ℵ(1, zf ) + 3iµ̂f (1 + 4µ̂f )ℵ(0, zf )

45 2


− m̂D 1 + 12µ̂f
2
"
 s α 2 1 X 5
m̂2q 4

Λ̂
F s
+ 96 1 + 12µ̂2f + 1 + 12µ̂2f 5 + 12µ̂2f ln
π Nf 16 m̂D 3 2
f
1 64 ζ ′ (−3) 32 ζ ′ (−1)
+ + 4γE + 8(7 + 12γE )µ̂2f + 112µ4f − − (1 + 12µ̂2f )
3 15 ζ(−3) 3 ζ(−1)
#
n o
2
−96 8ℵ(3, zf ) + 12iµ̂f ℵ(2, zf ) − 2(1 + 2µ̂f )ℵ(1, zf ) − iµ̂f ℵ(0, zf )

2
Note that chemical potential dependence also appears in pure-glue diagrams from the internal quark
loop in effective gluon propagators and effective vertices. This chemical potential dependence enters through
the chemical potential dependence of the Debye mass.
3
There is a mismatch in one term proportional to s2F α2s compared to result published in refs. [26, 27].
We found that the second term proportional to s2F α2s is 32(1 − 12µ̂2 )ζ ′ (−1)/ζ(−1), whereas in refs. [26, 27]
it was listed as 32(1 − 4µ̂2 )ζ ′ (−1)/ζ(−1). The author of refs. [26, 27] has agreed that this was a typo in his
article.

– 11 –
 " (
 s α 2 1 X 5


F s
+ 2
1 + 12µ̂f 1 + 12µ̂g + 90 2 (1 + γE ) µ̂2f µ̂2g
2
π Nf2 4m̂D
f,g
n  
− ℵ(3, zf + zg ) + ℵ(3, zf + zg∗ ) + 4iµ̂f ℵ(2, zf + zg ) + ℵ(2, zf + zg∗ ) − 4µ̂2g ℵ(1, zf )
)
o
−(µ̂f + µ̂g )2 ℵ(1, zf + zg ) − (µ̂f − µ̂g )2 ℵ(1, zf + zg∗ ) − 4iµ̂f µ̂2g ℵ(0, zf )
! #
15 2

Λ̂
− 1 + 12µ̂f 2 ln − 1 − ℵ(zg ) m̂D
2 2
" (
 c α   s α  X 15
235 792 2 1584 4

Λ̂
A s F s 2
+ 1 + 12µ̂f − 1+ µ̂f + µ̂f ln
3π πNf 2m̂D 16 47 47 2
f
 
144
319 2040 2 38640 4 24γE

− 1 + 12µ̂2f ln m̂D + 1+ µ̂f + µ̂f − 1 + 12µ̂2f
47 940 319 319 47
  ′ ′ h
44 156 2 ζ (−1) 268 ζ (−3) 72

− 1+ µ̂f − − 4iµ̂f ℵ(0, zf ) + 5 − 92µ̂2f ℵ(1, zf )
47 11 ζ(−1) 235 ζ(−3) 47
) ( 
i 2
m̂q 315 132 2 Λ̂
+144iµ̂f ℵ(2, zf ) + 52ℵ(3, zf ) + 90 + 1+ µ̂f ln
m̂D 4 7 2
  ) #
11
9 132 2 2 ΩYM
+ 1 + 12µ̂2f γE + 1+ µ̂f + ℵ(zf ) m̂D + NNLO , (4.7)
7 14 9 7 Ω0

where the sums over f and g include all quark flavors, zf = 1/2 − iµ̂f , and ΩYM NNLO is the
pure-glue contribution as before. The result contained in eq. (4.7) is new compared to what
was reported in ref. [55] since it includes separate chemical potentials for all quark flavors.

5 Mass prescription

As discussed in ref. [53], the two-loop perturbative electric gluon mass, first introduced
by Braaten and Nieto in [23, 24] is the most suitable for three-loop HTLpt calculations.
We use the Braaten-Nieto (BN) mass prescription for mD in the remainder of the paper.
Originally, the two-loop perturbative mass was calculated in refs. [23, 24] for zero chemical
potential, however, Vuorinen has generalized it to finite chemical potential. The resulting
expression for m2D is [26, 27]
( ! "
2 αs c2A αs Λ̂g 1 X

m̂D = cA + 5 + 22γE + 22 ln + sF 1 + 12µ̂2f
3π 12π 2 Nf
f
!
cA sF αs 2

2

2

Λ̂
+ 9 + 132µ̂f + 22 1 + 12µ̂f γE + 2 7 + 132µ̂f ln + 4ℵ(zf )
12π 2
! #)
s2 αs
Λ̂ 3 s2F αs

+ F 1 + 12µ̂2f 1 − 2 ln + ℵ(zf ) − 1 + 12µ̂2f . (5.1)
3π 2 2 π

The effect of the in-medium quark mass parameter mq in thermodynamic functions is

small and following ref. [53] we take mq = 0 which is the three-loop variational solu-

– 12 –
tion. The maximal effect on the susceptibilities comparing the perturbative quark mass,
m̂2q = αs (1 + 4µ̂2 )/6π, with the variational solution, mq = 0, is approximately 0.2% at
T = 200 MeV. At higher temperatures, the effect is much smaller, e.g. 0.02% at T = 1
GeV.

6 Thermodynamic functions

In this section we present our final results for the NNLO HTLpt pressure, energy density,
entropy density, trace anomaly, and speed of sound.

6.1 Running coupling

Below we will generally use the self-consistent one-loop running coupling implied by eq. (2.7),
however, in some places we will try to gauge the sensitivity of the result to the order of
the running coupling by comparing the impact of using one- or three-loop running. The
three-loop running coupling can be expressed approximately as [136, 137] 4

1 b1 ln t b21 (ln2 t − ln t − 1) + b0 b2
αs (Λ) = 1− 2 +
b0 t b0 t b40 t2

#
b31 ln3 t − 25 ln2 t − 2 ln t + 12 + 3b0 b1 b2 ln t
− , (6.1)
b60 t3

with t = ln(Λ2 /Λ2MS ) and

11cA − 2Nf
b0 = , (6.2)
12π
17c2A − 5cA Nf − 2CF Nf
b1 = , (6.3)
24π 2

2857c3A + 54CF2 − 615CF cA − 1415c2A Nf + (66CF + 79cA ) Nf2
b2 = . (6.4)
3456π 3
For one-loop running, we take b1 = b2 = 0. For both one- and three-loop running we
fix the scale ΛMS by requiring that αs (1.5 GeV) = 0.326 which is obtained from lattice
measurements [138]. For one-loop running, this procedure gives ΛMS = 176 MeV, and for
three-loop running, one obtains ΛMS = 316 MeV.

6.2 Scales
For the renormalization scale we use separate scales, Λg and Λq , for purely-gluonic and
fermionic graphs, respectively. We take the central values of these renormalization scales
p
to be Λg = 2πT and Λ = Λq = 2π T 2 + µ2 /π 2 . In all plots the thick lines indicate the
result obtained using these central values and the light-blue band indicates the variation of
the result under variation of both of these scales by a factor of two, e.g. πT ≤ Λg ≤ 4πT .
For all numerical results below we use cA = Nc = 3 and Nf = 3.

– 13 –
 1.0 1.0
Μ B = 0 MeV 1 loop Αs ; L =176 MeV Μ B = 400 MeV 1 loop Αs ; L =176 MeV
MS MS

ææ
æ æ æ
0.8 æ
æ æ 0.8
æ
æ æ
æ
æ æ ææ
æ
æ
0.6 0.6 æ
P P ideal

P P ideal
æ
æ
ææ æ
æ æ
æ æ
æ
æ
0.4 æ
æ
0.4
æ
æ æ
æ æ
NNLO HTLpt
æ
0.2 æ
0.2 æ
NNLO HTLpt
ææ Wuppertal - Budapest
æ æ
æ
æ æ
æ
æ æ HotQCD æ
æ
æ Wuppertal - Budapest
æ
ææ æ
0.0 æææ 0.0
200 400 600 800 1000 200 400 600 800 1000
T @MeVD T @MeVD

Figure 1. Comparison of the Nf = 2 + 1, µB = 0 (left) and µB = 400 MeV (right) NNLO HTLpt
pressure with lattice data from Borsanyi et al. [1, 4] and Bazavov et al. [13]. For the HTLpt results
a one-loop running coupling constant was used.

1.0 1.0
Μ B = 0 MeV 3 loop Αs ; L = 316 MeV Μ B = 400 MeV 3 loop Αs ; L = 316 MeV
MS MS

ææ
æ æ æ
0.8 æ
æ æ 0.8
æ
æ æ
æ
æ æ ææ
æ
æ
0.6 0.6 æ
P P ideal

P P ideal

æ
æ
ææ æ
æ æ
æ æ
æ
æ
0.4 æ
æ
0.4
æ
æ æ
æ æ
NNLO HTLpt
æ
0.2 æ
0.2 æ
NNLO HTLpt
ææ Wuppertal - Budapest
æ æ
æ æ
æ
æ
æ æ HotQCD æ
æ
æ Wuppertal - Budapest
æ
ææ æ
0.0 æææ 0.0
200 400 600 800 1000 200 400 600 800 1000
T @MeVD T @MeVD

Figure 2. Same as fig. 1 except with a three-loop running coupling constant.

6.3 Pressure
The QGP pressure can be obtained directly from the thermodynamic potential (4.5)

P(T, Λ, µ) = −ΩNNLO (T, Λ, µ) , (6.5)

where Λ above is understood to include both scales Λg and Λq . In figures 1 and 2 we

compare the scaled NNLO HTLpt pressure for µB = 0 (left) and µB = 400 MeV (right)
with lattice data from refs. [1, 3, 13]. In order to gauge the sensitivity of the results to
the order of the running coupling, in fig. 1 we show the results obtained using a one-loop
running and in fig. 2 the results obtained using a three-loop running. As can be seen by
comparing these two sets, the sensitivity of the results to the order of the running coupling
4
We have checked that for the scale range of interest, this is a very good approximation to the exactly
integrated QCD three-loop β-function.

– 14 –
 0.7
1 loop Αs ; Μ B = 400 MeV
L =176 MeV Free
0.6 MS
HTLpt
0.5 æææ æ Wuppertal - Budapest
æ
Μ B = 300 MeV
æ

0.4 Free

DP T 4
æ

æ
HTLpt
0.3 æææ æ Wuppertal - Budapest
æ æ æ
æ
æ
æ
0.2 æ
æ
æ
æ æ
æ æ
æ
æ
0.1 æ
æ æ

0.0
200 300 400 500 600
T @MeVD

Figure 3. Comparison of the Stefan-Boltzmann limit (dashed lines) and NNLO HTLpt (solid lines)
results for the scaled pressure difference with lattice data from Borsanyi et al. [4].

is small for T & 250 MeV. As a result, unless the order of the running coupling turns out
to have a significant effect on a given observable (see e.g. the fourth-order baryon number
susceptibility), we will show the results obtained using a one-loop running coupling consis-
tent with the counterterms necessary to renormalize the NNLO thermodynamic potential
(2.7).
For an additional comparison we can compute the change in the pressure

∆P = P(T, Λ, µ) − P(T, Λ, 0) . (6.6)

In figure 3 we plot ∆P as a function of the temperature for µB = 300 MeV and µB = 400
MeV. The solid lines are the NNLO HTLpt result and the dashed lines are the result
obtained in the Stefan-Boltzmann limit. We note that in fig. 3 the lattice data from
the Wuppertal-Budapest group [3] is computed up to O(µ2B ), whereas the HTLpt result
includes all orders in µB . As can be seen from this figure, the NNLO HTLpt result is quite
close to the result obtained in the Stefan-Boltzmann limit. Note that the small correction
in going from the Stefan-Boltzmann limit to NNLO HTLpt indicates that the fermionic
sector is, to good approximation, weakly coupled for T & 300 MeV.

6.4 Energy density

Once the pressure is known, it is straightforward to compute other thermodynamic func-
tions such as the energy density by computing derivatives of the pressure with respect to
the temperature and chemical potential. The energy density can be obtained via
∂P ∂P
E =T +µ −P. (6.7)
∂T ∂µ
In figure 4 we plot the scaled NNLO HTLpt energy density for µB = 0 (left) and µB = 400
MeV (right) together with µ = 0 lattice data from ref. [1]. As we can see from this figure,

– 15 –
 1.0 1.0
Μ B = 0 MeV 1 loop Αs ; L =176 MeV Μ B = 400 MeV 1 loop Αs ; L =176 MeV
MS MS

0.8 ææ ææ ææ 0.8
æææ
ææ
æ
æ
æ
æ
æ
0.6 æ 0.6
E E ideal

E E ideal
æ
æ
æ
æ
0.4 æ
æ 0.4
æ
æ
æ
æ NNLO HTLpt
0.2
æ
æ
0.2
æ
Wuppertal - Budapest NNLO HTLpt

0.0 0.0
200 400 600 800 1000 200 400 600 800 1000
T @MeVD T @MeVD

Figure 4. Comparison of the Nf = 2 + 1, µB = 0 (left) and µB = 400 MeV (right) NNLO HTLpt
energy density with lattice data. The µB = 0 lattice data shown in the left panel are from ref. [1].
For the HTLpt results a one-loop running coupling constant was used.

1.0 1.0
Μ B = 0 MeV 1 loop Αs ; L =176 MeV Μ B = 400 MeV 1 loop Αs ; L =176 MeV
MS MS

0.8 ææææææ
ææ ææ 0.8
æææ
æææ
æææ
æ
ææ
ææ
0.6 æ
0.6
S S ideal

S S ideal

æ
æ
æ
æ
æ
æ
æ
æ
æ
æ
æ
0.4 æ
æ
0.4
æ
æ
æ
æ
æ NNLO HTLpt
0.2æ
æ 0.2
æ
æ
Wuppertal - Budapest NNLO HTLpt

0.0 0.0
200 400 600 800 1000 200 400 600 800 1000
T @MeVD T @MeVD

Figure 5. Comparison of the Nf = 2 + 1, µB = 0 (left) and µB = 400 MeV (right) NNLO HTLpt
entropy density with lattice data. The µB = 0 lattice data shown in the left panel are from ref. [1].
For the HTLpt results a one-loop running coupling constant was used.

there is reasonable agreement between the NNLO HTLpt energy density and the lattice
data when the central value of the scale is used.

6.5 Entropy density

Similarly, we can compute the entropy density
∂P
S(T, µ) = . (6.8)
∂T
We note that in the ideal gas limit, the entropy density becomes
  
4dA π 2 T 3 7 dF 60 2
Sideal (T, µ) = 1+ 1 + µ̂ . (6.9)
45 4 dA 7

– 16 –
 5 1 loop Αs ; L =176 MeV 1 loop Αs ; L =176 MeV
Μ B = 0 MeV Μ B = 400 MeV
MS
5 æææ MS
æ
æ
4 æ
æ
æ æ
æ 4
æ

H E -3PL T 4
H E -3PL T 4

æ NNLO HTLp t æ NNLO HTLp t

3 æ æ
3 æ
æ æ WB æ WB
æ
2 æ æ æ
æ 2
æ æ
æ
1 æ
æ
æ 1
æ æ
æ æ
0 0
200 400 600 800 1000 200 400 600 800 1000
T @ MeVD T @ MeVD

Figure 6. Comparison of the Nf = 2 + 1, µB = 0 (left) and µB = 400 MeV (right) NNLO HTLpt
trace anomaly with lattice data. The µB = 0 lattice data are from [1] and the µB = 400 MeV
lattice data are from [4]. For the HTLpt results a one-loop running coupling constant was used.

In figure 5 we plot the scaled NNLO HTLpt entropy density for µB = 0 (left) and µB = 400
MeV (right) together with µ = 0 lattice data from ref. [1]. As we can see from this figure,
there is quite good agreement between the NNLO HTLpt entropy density and the lattice
data when the central value of the scale is used.

6.6 Trace anomaly

Since it is typically the trace anomaly itself which is computed on the lattice and then
integrated to obtain the other thermodynamic functions, it is interesting to compare di-
rectly with lattice data for the trace anomaly. The trace anomaly is simply I = E − 3P.
In the ideal gas limit, the trace anomaly goes to zero since E = 3P. When interactions are
included, however, the trace anomaly (interaction measure) becomes non-zero. In figure
6 we plot the NNLO HTLpt trace anomaly scaled by T 4 for µB = 0 (left) and µB = 400
MeV (right) together with lattice data from refs. [1] and [4]. As we can see from this figure,
there is quite good agreement between the NNLO HTLpt trace anomaly and the lattice
data for T & 220 MeV when the central value of the scale is used.

6.7 Speed of sound

Another quantity which is phenomenologically interesting is the speed of sound. The speed
of sound squared is defined as
∂P
c2s = . (6.10)
∂E
In figure 7 we plot the NNLO HTLpt speed of sound for µB = 0 (left) and µB = 400 MeV
(right) together with lattice data from refs. [1] and [4]. As we can see from this figure,
there is quite good agreement between the NNLO HTLpt speed of sound and the lattice
data when the central value of the scale is used.

– 17 –
 0.35 Μ B = 0 MeV 1 loop Αs ; L =176 MeV 0.35 Μ B = 400 MeV 1 loop Αs ; L =176 MeV
MS MS

æ æ æ æ æ
æ
0.30 æ
0.30 æ
æ

ææ
æ æ
æ
0.25 0.25 æ
æ
c 2s

c 2s
æ æ

0.20 0.20 æ

æ æ

æ NNLO HTLpt NNLO HTLpt

0.15 æ Wuppetral - Budapest 0.15 æ Wuppetral - Budapest
æ
ææ Stefan - Boltzman limit Stefan - Boltzman limit
0.10 0.10
200 400 600 800 1000 200 400 600 800 1000
T @ MeVD T @ MeVD

Figure 7. Comparison of the Nf = 2 + 1, µB = 0 (left) and µB = 400 MeV (right) NNLO HTLpt
speed of sound squared with lattice data. The µB = 0 lattice data are from [1] and the µB = 400
MeV lattice data are from [4]. For the HTLpt results a one-loop running coupling constant was
used.

7 Quark number susceptibilities

Using the full thermodynamic potential as a function of chemical potential(s) and tem-
perature we can compute the quark number susceptibilities. In general, one can intro-
duce a separate chemical potential for each quark flavor giving a Nf -dimensional vector
µ ≡ (µu , µd , ..., µNf ). By taking derivatives of the pressure with respect to chemical po-
tentials in this set, we obtain the quark number susceptibilities 5

∂ i+j+k+ ··· P (T, µ)

χijk ··· (T ) ≡ . (7.1)
∂µiu ∂µjd ∂µks · · · µ=0

Below we will use a shorthand notation for the susceptibilities by specifying derivatives by
a string of quark flavors in superscript form, e.g. χuu ds
2 = χ200 , χ2 = χ011 , χ4
uudd = χ
220 ,
etc.
When computing the derivatives with respect to the chemical potentials we treat Λq
as being a constant and only put the chemical potential dependence of the Λq in after the
derivatives are taken. We have done this in order to more closely match the procedure
used to compute the susceptibilities using resummed dimensional reduction [44].6

– 18 –
 1 loop Αs ; L =176 MeV 3 loop Αs ; L = 316 MeV
MS MS
1.0 1.0
œœ
œçóó œœ
ó á
óó
õ œæ óœ
õ œ
õ ó œáõóœ óœ
ó
õ õ õ ó
õ ó ó õ
ó õ
õ õ á
ó õ
ó
œœ
œçóó œœ
ó á
óó
õ œæ óœ
õ œ
õ ó œáõóœ óœ
ó
õ õ õ ó
õ ó ó õ
õ õ á
ó õ
ó õ ó
œ
ó óó
çœç¢
õææ
çç
ç¢æ
æõ
çæ
õ
çç
œ
ó óó
çœç¢
õææ
çç
ç¢æ
æõ
çæ
õ
çç
ç æ õæ ç æ õæ
0.8 çççæõ¢æõ@ @ 0.8 çççæõ¢æõ@ @
ç¢ ç¢
çæ çæ
 Χ f2

 Χ f2
çæ@ çæ@
çæ çæ
0.6çæ NNLO HTLpt 0.6çæ NNLO HTLpt
@ @
BH u L

BH u L
æ æ
WBH BL , stout WBHBL , stout
æ@ æ@
BNL - BIH BL , N Τ = 8H HISQL BNL - BIHBL , N Τ = 8HHISQL
Χ2

Χ2
æ æ
0.4@ WBH u L , stout 0.4@ WBHu L , stout
æ æ
@ BNL - BIH u L , HISQ @ BNL - BIHu L , HISQ
BNL - BIH u L , p4 BNL - BIHu L , p4
@ @
0.2@ TIFRH u L , N Τ = 8H SQL 0.2@ TIFRHu L , N Τ = 8HSQL
@ BNL - BIH sL , HISQ @ BNL - BIHsL , HISQ
MILCH sL , N Τ = 8H asqtad L MILCHsL , N Τ = 8Hasqtad L
0.0 0.0
200 400 600 800 1000 200 400 600 800 1000
T @MeVD T @MeVD

Figure 8. The scaled second order baryon number susceptibility compared with various lattice
data using one-loop running (left) and three-loop running (right). The lattice data labeled WB,
BNL-BI(B), BNL-BI(u,s), MILC, and TIFR come from refs. [2], [9], [10], [12], and [139], respectively.

1.1 ò
1.1 ò
1 loop Αs ; L =176 MeV 3 loop Αs ; L = 316 MeV
à MS à MS
ò ò
ò ò
1.0 ò 1.0 ò
òà ò òà ò
òò òò
B,f

B,f

0.9 0.9
Χ4B  Χ 4

Χ4B  Χ 4

0.8 0.8
NNLO HTLpt NNLO HTLpt
0.7 WBHBL , stout 0.7 WBHBL , stout

BNL- BIHBL , N Τ = 8HHISQL BNL- BIHBL , N Τ = 8HHISQL

0.6 0.6
200 400 600 800 1000 200 400 600 800 1000
T @MeVD T @MeVD

Figure 9. The scaled fourth order baryon number susceptibility compared with various lattice
data using one-loop running (left) and three-loop running (right). The lattice data labeled WB,
BNL-BI(B), BNL-BI(u,s), MILC, and TIFR come from refs. [2], [9], [10], [12], and [139], respectively.

7.1 Baryon number susceptibilities

We begin by considering the baryon number susceptibilities. The nth -order baryon number
susceptibility is defined as

∂nP
χnB (T ) ≡ . (7.2)
∂µnB µB =0
5
We have specified that the derivatives should be evaluated at µ = 0. In general, one could define the
susceptibilities at µ = µ0 .
6
One could instead put the chemical potential dependence of the Λq in prior to taking the derivatives
with respect to the chemical potentials. If this is done, the central lines obtained are very close to the ones
obtained using the fixed-Λq prescription, however, the scale variation typically increases in this case.

– 19 –
 ò
à
1.0 ò
ò
àò
ò
0.8 ò

N f2 T 2 Χ 4B  Χ 2B
àò
ò
òàò ò
òò
0.6

0.4
NNLO HTLpt
WBHBL , stout
0.2 BNL- BIHBL , N Τ = 8HHISQL
Stefan - Boltzmann limit
0.0
200 400 600 800 1000
T @ MeVD

Figure 10. Comparison of the Nf = 2 + 1 NNLO HTLpt ratio of the fourth to second order
baryon susceptibility with lattice data. For the HTLpt results a one-loop running coupling constant
was used. The data labeled WB and BNL-BI(B) come from refs. [5, 6] and [9], respectively.

For a three flavor system consisting of (u, d, s), the baryon number susceptibilities can be
related to the quark number susceptibilities [15]
1 h uu i
χB2 = χ2 + χdd ss ud ds
2 + χ2 + 2χ2 + 2χ2 + 2χ2
us
, (7.3)
9
and
1 h uuuu
χB
4 = χ + χdddd + χssss + 4χuuud + 4χuuus + 4χdddu + 4χddds + 4χsssu
81 4 4 4 4 4 4 4 4
i
+ 4χsssd
4 + 6χuudd
4 + 6χddss
4 + 6χuuss
4 + 12χuuds
4 + 12χddus
4 + 12χssud
4 . (7.4)

If we treat all quarks as having the same chemical potential µu = µd = µs = µ = 31 µB ,

eqs. (7.3) and (7.4) reduce to χB uu B
2 = χ2 and χ4 = χ4
uuuu . This allows us to straightfor-

wardly compute the baryon number susceptibility by computing derivatives of (4.5) with
respect to µ.
In figure 8 we compare the NNLO HTLpt result for the second order baryon number
susceptibility with lattice data from various groups. In the left panel of this figure we used
the one-loop running and on the right we used the three-loop running. As one can see,
for this quantity, the size of the light-blue band becomes larger if one uses the three-loop
running, however, the central value obtained is very close in both cases.
Comparing to the lattice data we see that the NNLO HTLpt prediction is approxi-
mately 10% higher than the lattice data at T = 250 MeV and approximately 2% higher at
T = 800 MeV. We note in this context that recently the four-loop second-order baryon num-
ber susceptibility has been computed in ref. [44] using the resummed dimensional reduction
method. The result from this approach lies within the NNLO HTLpt scale variation band
and is even closer to the lattice data with the error at T = 250 MeV being approximately
2% and . 1% at T = 800 MeV. Our result, taken together with the resummed dimensional
reduction results seem to indicate that the quark sector of the QGP can be quite accurately

– 20 –
 0.0

- 0.1

- 0.2

Χ B6
- 0.3

2
N f6 T
- 0.4

- 0.5

- 0.6

- 0.7
200 400 600 800 1000
T @ MeVD

Figure 11. The Nf = 2 + 1 NNLO HTLpt scaled sixth-order baryon susceptibility as a function
of temperature.

described using resummed perturbation theory for temperatures above approximately 300
MeV.
In figure 9 we compare the NNLO HTLpt result for the fourth order baryon number
susceptibility with lattice data. Once again we show in the left and right panels, the result
obtain using the one-loop running coupling and three-loop running coupling, respectively.
Both the one- and three-loop running results are consistent with the lattice data shown;
however, the lattice error bars on this quantity are somewhat large and the data are
restricted to temperatures below 400 MeV, making it difficult to draw firm conclusions from
this comparison. That being said, HTLpt makes a clear prediction for the temperature
dependence of the fourth order baryon number susceptibility. It will be very interesting to
see if future lattice data agree with this prediction.
In figure 10 we plot the scaled ratio of the fourth and second order baryon number
susceptibilities as a function of temperature along with lattice data for this ratio. As we
can see from this figure, this ratio very rapidly approaches the Stefan-Boltzmann limit if
one considers the central NNLO HTLpt line. Comparing with the lattice data we see that
the NNLO HTLpt result is below the lattice data for temperatures less than approximately
300 MeV. Without lattice data at higher temperatures, it is hard to draw a firm conclusion
regarding the temperature at which HTLpt provides a good description of this quantity.
In figure 11 we show the NNLO HTLpt prediction for the sixth order baryon number
susceptibility. To the best of our knowledge there is currently no publicly available lattice
data for this quantity. It will be very interesting to see if these NNLO HTLpt predictions
agree with lattice data as they becomes available.

7.2 Single quark number susceptibilities

We now consider the single quark number susceptibilities (7.1). For these we use the general
expression for the NNLO thermodynamic potential with different quark chemical potentials
(4.7). The resulting susceptibilities can either be diagonal (same flavor on all derivatives)
or off-diagonal (different flavor on some or all indices). In HTLpt there are off-diagonal

– 21 –
 0.10
0.6 ô
æ

æ
NNLO HTLpt
ô
æì æ BNL - BIH uudd L
0.5 æ
æ æ
æ ìõ ì
0.08 æ

õ
õæ æõæõ õìõæõæ BNL - BIH uussL
õõ æ
0.4
0.06
Χ uuuu

Χ uudd
æ
ô
4

0.3

4
NNLO HTLpt 0.04 æ
ô
WB, stout
0.2 BNL - BIHu L , N Τ = 8Hp4L æ
ô
BNL - BIHsL , N Τ =12Hp4L
BNL - BIHu L , N Τ = 8HHISQL 0.02 æ
ô
0.1 TIFRHu L , N Τ = 8HSQL
Stefan - Boltzmann limit
0.0 0.00
200 400 600 800 1000 200 400 600 800 1000
T @MeVD T @MeVD

Figure 12. Comparison of the Nf = 2 + 1 NNLO HTLpt ratio of the fourth order diagonal
single quark number susceptibility (left) and the only non-vanishing fourth order off-diagonal quark
number susceptibility (right) with lattice data. In the left figure the dashed blue line indicates
the Stefan-Boltzmann limit for this quantity. For the HTLpt results a one-loop running coupling
constant was used. The data labeled BNL-BI(uudd), BNL-BI(u,s), BNL-BI(uuss), and TIFR come
from refs. [9], [10], [11], and [139], respectively.

0.6
õ
0.5 õ õ õõõ õ õ õ
T 2 Χ4 u  Χ2 u

0.4

0.3

0.2
NNLO HTLpt
0.1 WB, stout

0.0
200 400 600 800 1000
T @MeVD

Figure 13. Comparison of the Nf = 2 + 1 NNLO HTLpt ratio of the fourth to second order single
quark susceptibility with lattice data. For the HTLpt results a one-loop running coupling constant
was used. The data labeled WB come from refs. [5, 6].

f f
susceptibilities emerging explicitly from graphs F3c and F3j ; however, the latter vanishes
when we use the variational mass prescription for the quark mass (mq = 0), so we need
f
only consider the F3c graph. Additionally, there are potential off-diagonal contributions
coming from all HTL terms since the Debye mass receives contributions from all quark
flavors. In practice, however, because we evaluate derivatives with respect to the various
chemical potentials and then take µf → 0, one finds that all off-diagonal second order

– 22 –
susceptibilities vanish in HTLpt. Therefore, for the three-flavor case one has

χud ds su
2 = χ2 = χ2 = 0 , (7.5)

and, as a result, the single quark second order susceptibility is proportional to the baryon
number susceptibility
1 B
χuu
2 = χ2 . (7.6)
3
For the fourth order susceptibility, there is only one non-zero off-diagonal susceptibility,
namely χuudd
4 = χuuss
4 = χddss
4 , which is related to the diagonal susceptibility, e.g. χuuuu
4 =
dddd
χ4 ssss
= χ4 , as

χuuuu
4 = 27χB uudd
4 − 6χ4 . (7.7)

As a consequence, one can compute χuuuu 4 directly from (4.7) or by computing χB4 using
(4.5) and χ4 uudd using (4.7) and applying the above relation. In our final plots we compute
χuuuu
4 directly from (4.7), however, we have checked that we obtain the same result if we
use (7.7) instead.
In figure 12 (left) we plot our result for the fourth order single quark susceptibility
χ4uuuu compared to lattice data from refs. [10], [9], [11], and [139]. As we can see from
this figure, for the fourth order susceptibility there is very good agreement with available
lattice data. In addition, the scale variation of the HTLpt result is quite small for this
particular quantity. In figure 12 (right) we plot our result for the fourth order off-diagonal
single quark susceptibility χuudd
4 compared to lattice data. From this right panel we also
see reasonably good agreement between the NNLO HTLpt result and the available lattice
data.
In figure 13 we plot the scaled ratio of the fourth- and second-order single quark
susceptibilities. Once again we see good agreement between the NNLO HTLpt result and
lattice data. Once again, for both figures 12 and 13, the lattice data are confined to
relatively low temperatures. It will be interesting to compare higher temperature lattice
data with the NNLO HTLpt prediction as they become available.
Finally, in figure 14 we plot the diagonal and off-diagonal sixth-order quark number
susceptibilities χ600 (left), χ420 (middle), and χ222 (right). In the left panel of figure 14
we show lattice data available from the RBC-Bielefeld collaboration [140]. At this point in
time the lattice sizes are small and the errors bars for χ600 are large, so it is hard to draw
a firm conclusion from this comparison. Regarding the off-diagonal six-order single quark
susceptibilities (center and right panels), we are unaware of any lattice data for these. As
before, it will be very interesting to see if these NNLO HTLpt predictions agree with lattice
data as it becomes available.

8 Conclusions and outlook

In this paper, we presented the results of a NNLO (three-loop) HTLpt calculation of the
thermodynamic potential of QCD at finite temperature and chemical potential(s). Our

– 23 –
 0.6 0.00
æ
0.000
0.4 1 loop Α s ; L MS =176 MeV -0.01
-0.001
T 2 Χ 600

T 2 Χ 420

T 2 Χ 222
0.2 -0.02
ææ -0.002
0.0 æ æ
æ
-0.03
æ
æ
æ
æ
-0.04 1 loop Α s ; L MS =176 MeV -0.003
-0.2 æ
NNLO HTLpt
RBC - BI , N Τ = 4 1 loop Α s ; L MS =176 MeV
RBC - BI , N Τ = 6
-0.05 -0.004
200 400 600 800 1000 200 400 600 800 1000 200 400 600 800 1000
T @MeVD T @MeVD T @MeVD

Figure 14. The Nf = 2 + 1 NNLO HTLpt scaled sixth-order diagonal and off-diagonal single
quark susceptibilities χ600 (left), χ420 (middle), and χ222 (right) as a function of temperature. In
the left panel we show lattice data available from the RBC-Bielefeld collaboration for Nτ = 4 and
Nτ = 6 lattices [140].

final result (4.7) is completely analytic, gauge invariant, and should be valid in the region
of the phase diagram for which µf . 2πT . Based on the resulting thermodynamic potential
we proceeded to calculate the pressure, energy density, entropy density, trace anomaly, and
speed of sound of the QGP. In all cases we found very good agreement between the results
obtained using the central values of the renormalization scales and available lattice data.
Additionally, we have made predictions for the diagonal and off-diagonal sixth-order baryon
number and single quark susceptibilities.
Looking to the future there are still many avenues for improvement in the HTLpt
approach: (1) inclusion of the effects of finite quark masses (2) extension of results to µf &
2πT and eventually to T = 0, and (3) to potentially resum logarithms in order to reduce the
scale variation of the final results (light-blue bands in all figures). Of these three, the second
task is the most straightforward; however, in order to make more definitive and constrained
statements it now seems necessary to start moving in directions (1) and (3) as well. In
closing, we emphasize that HTLpt provides a gauge invariant reorganization of perturbation
theory for calculating static quantities in thermal field theory. Since the NNLO HTLpt
results are in good agreement with lattice data for various thermodynamic quantities down
to temperatures that are relevant for LHC, it would therefore be interesting and challenging
to apply HTLpt to the calculation of dynamic quantities, especially transport coefficients,
at these temperatures.

Acknowledgments

We thank S. Borsanyi, S. Datta, F. Karsch, S. Gupta, S. Mogliacci, P. Petreczky, and A.

Vuorinen for useful discussions. N.H., A.B., and M.G.M. were supported by the Indian De-
partment of Atomic Energy. M.S. was supported in part by DOE Grant No. DE-SC0004104.
N.S. was supported by the Bielefeld Young Researchers’ Fund.

– 24 –
A Expansion in mass parameters

In refs. [45, 46] the NLO HTLpt thermodynamic potential was reduced to scalar sum-
integrals. Evaluating these scalar sum-integrals exactly seems intractable, however, the
sum-integrals can be calculated approximately by expanding them in powers of mD /T and
mq /T following the method developed in ref. [33]. We will adopt the same strategy in
this paper and include all terms through order g 5 assuming that mD and mq are O(g)
at leading order. At each loop order, the contributions can be divided into those coming
from hard and soft momenta, which are the momenta proportional to the scales T and
gT , respectively. In the one-loop diagrams, the contributions are either hard (h) or soft
(s), while at the two-loop level, there are hard-hard (hh), hard-soft (hs), and soft-soft
(ss) contributions. At three loops there are hard-hard-hard (hhh), hard-hard-soft (hhs),
hard-soft-soft (hss), and soft-soft-soft (sss) contributions.

A.1 One-loop sum-integrals

We now review the mass expansion of the necessary one-loop sum-integrals considering
separately the contributions from hard and soft momenta. We list the purely gluonic
contributions when they are necessary for simpler exposition of the final result. Note that
in order to simplify the results, when possible, it is best to add the corresponding iterated
polarization and self-energy insertions that appear at higher order in δ, e.g. below we will
g f g f
also include F2d , F2d , F3m , and F3i as “one-loop” contributions.

Hard contributions
g g
For one-loop gluon (F1a ) and one-loop ghost (F1b ) diagrams, we need to expand in order
m2D :
Z
X Z
g(h) 1 2 1 2 X1
F1a+1b = (d − 1) ln P + mD
2 2 P2
P P
Z 
X 
1 4 1 2 2d 2 d 2
− m − − 4 TP + 2 2 TP + 4 TP . (A.1)
4(d − 1) D P 4 p2 P 2 p p P p
P

g
The one-loop graph with a gluon self-energy insertion (F2d ) has an explicit factor of m2D
and, therefore, we only need to expand the sum-integral to first order in m2D :
Z Z  
g(h) 1 2 X1 1 4
X 1 2 2d 2 d 2
F2d =− mD + m − − 4 TP + 2 2 TP + 4 TP . (A.2)
2 P 2 2(d − 1) D P 4 p2 P 2 p p P p
P P

g
The one-loop graph with two gluon self-energy insertions (F3m ) must be expanded to zeroth
2
order in mD
Z 
X 
g(h) 1 1 2 2d 2 d 2
F3m = − m4D − − T P + T P + T . (A.3)
4(d − 1) P 4 p2 P 2 p4 p2 P 2 p4 P
P

– 25 –
The sum of eqs. (A.1)-(A.3) is very simple
Z
X
g(h) 1
π2
F1a+1b+2d+3m = (d − 1) ln P 2 = − T 4 . (A.4)
2 45
P

f
The one-loop fermionic graph F1b needs to expanded to second order in m2q
Z
X Z
X Z 
X 
f (h) 2 2 1 4 2 1 2 1 2
F1b = −2 log P − 4mq + 2mq − + TP − 2 2 TP . (A.5)
P2 P 4 p2 P 2 p2 P 2 p P0
{P } {P } {P }

f
The one-loop fermion loop with a fermion self-energy insertion F2d must be expanded to
2
first order in mq ,
Z
X Z 
X 
f (h) 2 1 4 2 1 2 1 2
F2d = 4mq − 4mq − + TP − 2 2 TP . (A.6)
P2 P 4 p2 P 2 p2 P 2 p P0
{P } {P }

f
The one-loop fermion loop with two self-energy insertions F3i must be expanded to zeroth
order in m2q :
Z 
X 
f (h) 4 2 1 2 1 2
F3i = 2mq − + TP − 2 2 TP . (A.7)
P 4 p2 P 2 p2 P 2 p P0
{P }

The sum of eqs. (A.5)-(A.7) is particularly simple

Z
X
f (h)
F1b+2d+3i = −2 ln P 2
{P }
 
7π 2 4 120 2 240 4
=− T 1+ µ̂ + µ̂ . (A.8)
180 7 7

This is the free energy of an ideal gas consisting of a single massless fermion.

Soft contributions
g g g
The soft contributions in the diagrams F1a+1b , F2d , and F3m arise from the P0 = 0 term in
the sum-integral. At soft momentum P = (0, p), the HTL self-energy functions reduce to
ΠT (P ) = 0 and ΠL (P ) = m2D . The transverse term vanishes in dimensional regularization
because there is no momentum scale in the integral over p. Thus the soft contributions
come from the longitudinal term only and read
Z    
g(s) 1 2 2

m3D T Λg 2ǫ 8
F1a+1b = T ln p + mD = − 1+ ǫ , (A.9)
2 12π 2mD 3
p

Z  2ǫ
g(s) 1 1 m3D T Λg
F2d = − m2D T = [1 + 2ǫ] , (A.10)
2 p2 + m2D 8π 2mD
p

– 26 –
 Z
g(s) 1 1 m3D T
F3m = − m4D T = − . (A.11)
4 (p2 + m2D )2 32π
p

The total soft contribution from eqs. (A.9)-(A.11) is

 2ǫ  
g(s) m3D T Λg 8
F1a+1b+2d+3m = − 1+ ǫ . (A.12)
96π 2mD 3

There are no soft contributions from the leading-order fermion diagrams or HTL countert-
erms (polarization and self-energy insertions).

A.2 Two-loop sum-integrals

For hard momenta, the self-energies are suppressed by mD /T and mq /T relative to the
inverse free propagators, so we can expand in powers of ΠT , ΠL , and Σ. As was the case for
the one-loop contributions, we once again treat the polarization and self-energy insertion
NNLO diagrams as two-loop graphs in order to simplify the resulting expressions.

(hh) contribution
We first consider the contribution from fermionic diagrams. The (hh) contribution from
f f
F2a and F2b reads

 
Z
X Z
X
f (hh) 2 1 2 
F2a+2b = (d − 1)g  − 
P 2 Q2 P 2 Q2
{P Q} P {Q}
Z 
X 
1 1 d−2 1
+2m2D g2 TP + 4 2 −
p 2 P 2 Q2 P Q d − 1 p 2 P 2 Q2
P {Q}
Z 
X 
d+1 1 4d q2 2d P ·Q
+m2D g2 − − TR
d − 1 P 2 Q2 r 2 d − 1 P 2 Q2 r 4 d − 1 P 2 Q2 r 4
{P Q}
Z 
X 3−d 1 2d P ·Q d+2 1 4d q2
+m2D g2 + − +
d − 1 P 2 Q2 R 2 d − 1 P 2 Q2 r 4 d − 1 P 2 Q2 r 2 d − 1 P 2 Q2 r 4
{P Q}
 Z 
X 
4 q2 2 2 d+3 1 2 r 2 − p2
− + 2mq g (d − 1) − +
d − 1 P 2 Q2 r 2 R 2 d − 1 P 2 Q2 R 2 P 2 Q4 q 2 P 2 Q2 R 2
{P Q}
Z 
X 
1 p2 − r 2
+2m2q g2 (d − 1) + TQ
P 2 Q20 Q2 P 2 q 2 Q20 R2
{P Q}
Z 
X 
2 1
+2m2q g2 (d − 1) − TQ . (A.13)
P 2 Q4 P 2 Q20 Q2
P {Q}

f f
We consider next the (hh) contributions from F3d and F3f . The easiest way to calculate
f f
these term, is to expand the two-loop diagrams F2a and F2b to first order in m2D . This

– 27 –
yields
Z 
X 
f (hh) 1 1 d−2 1
F3d+3f = −2m2D g2 2 2 2
TP + 4 2 −
p P Q P Q d − 1 p P 2 Q2
2
P {Q}
Z 
X 
d+1 1 4d q2 2d P ·Q
−m2D g2 − − TR
d − 1 P 2 Q2 r 2 d − 1 P 2 Q2 r 4 d − 1 P 2 Q2 r 4
{P Q}
Z 
X 3−d 1 2d P ·Q d+2 1
−m2D g2 2 2 2
+ 2 2 4
−
d−1P Q R d−1P Q r d − 1 P Q2 r 2
2
{P Q}

4d q2 4 q2
+ − . (A.14)
d − 1 P 2 Q2 r 4 d − 1 P 2 Q2 r 2 R 2
f f f f
Next we consider the (hh) contribution from the diagrams F3e , F3g , F3k and F3l

Z  
f (hh)  X 1 p2 − r 2
F3e+3g+3k+3l = −2m2q g2 (d − 1)  + TQ
P 2 Q20 Q2 P 2 q 2 Q20 R2
{P Q}
Z 
X 
2 1
+ 2 4
+ 2 2 2 TQ
P Q P Q0 Q
P {Q}

Z 
X d+3 
1 2 r 2 − p2 
+ − + . (A.15)
d − 1 P 2 Q2 R 2 P 2 Q4 q 2 P 2 Q2 R 2
{P Q}

The sum of eqs.(A.13)-(A.15) is

 
Z
X Z
X
f (hh) 2 1 2 
F2a+2b+3d+3e+3f+3g+3k+3l = (d − 1)g  − 
P 2 Q2 P 2 Q2
{P Q} P {Q}

π2 αs 4


= T 1 + 12µ̂2 5 + 12µ̂2 . (A.16)
72 π
For completeness, the hard-hard contribution coming from two-loop pure-glue diagrams is
[50]
Z
X
g(hh) 1 2 2 1
F2a+2b+2c+3h+3i+3j+3k+3l = (d − 1) g
4 P Q2
2
PQ
π2 αs 4
= T (A.17)
36 π
(hs) contribution
In the (hs) region, one gluon momentum is soft but the fermionic momentum is always
f f
hard. The terms that contribute through order g2 m3D T and g2 m2q mD T from F2a and F2b

– 28 –
were calculated in ref. [45–47] and read
Z Z 
X 
f (hs) 2 1 1 2q 2
F2a+2b = 2g T − 4
p2 + m2D Q2 Q
p {Q}
Z Z 
X 
1 1 2 q2 8 q4
+2m2D g2 T − (3 + d) 6 +
p2 + m2D Q4 d Q d Q8
p {Q}
Z Z "
X  #
1 3 4q 2 4 2 1
−4m2q g2 T − 6 − 4 TQ − 2 . (A.18)
p2 + m2D Q4 Q Q Q (Q·Y )2 ŷ
p {Q}

f f
The (hs) contribution from diagrams F3d and F3f can again be calculated from the diagrams
f f
F2a and F2b by Taylor expanding their contribution to first order in m2D . This yields
Z Z 
X 
f (hs) 1 1 2q 2
F3d+3f = 2m2D g 2 T − 4
(p2 + m2D )2 Q2 Q
p {Q}
Z Z 
X 
p2 1 2 q2 8 q4
−2m2D g 2 T − (3 + d) 6 +
(p2 + m2D )2 Q4 d Q d Q8
p {Q}
Z Z "
X  #
1 3 4q 2 4 2 1
−4m2D m2q g 2 T − 6 − 4 TQ − 2 .(A.19)
(p2 + m2D )2 Q4 Q Q Q (Q·Y )2 ŷ
p {Q}

f f f f
We also need the (hs) contributions from the diagrams F3e , F3g , F3k and F3l . Again we
f f
calculate these contributions by expanding the two-loop diagrams F2a and F2b to first order
in m2q . This yields
Z Z "
X  #
f (hs) 1 3 4q 2 4 2 1
F3e+3g+3k+3l = 4m2q g 2 T − 6 − 4 TQ − 2 . (A.20)
p p2 + m2D Q4 Q Q Q (Q·Y )2 ŷ
{Q}

– 29 –
The sum of eqs. (A.18)-(A.20) is
f (hs)
F2a+2b+3d+3e+3f +3g+3k+3l
"Z Z #Z  
1 1 X 1 2q 2
2 2
= 2g T 2 + mD
2 − 4
2
p p + mD p p 2 + m2 Q2 Q
D {Q}
"Z Z #Z  
1 p 2 X 1 2 q2 8 q4
2 2
+ 2g mD T 2 2 −
2 − (3 + d) 6 +
Q4 d Q d Q8
p p + mD p p 2 + m2 D {Q}
Z Z "
X  #
2 2 1 3 4q 2 4 2 1
− 4mq mD T
2 − 6 − 4 TQ − 2
p p 2 + m2 Q4 Q Q Q (Q·Y )2 ŷ
D {Q}
"Z Z #Z
1 1 X 1
2 2
= −2(d − 1)g T 2 + mD
2
2
p p + mD p p 2 + m2 Q2
D {Q}
Z Z
X Z Z
X
d−1 2 4 1 1 d−3 2 2 2 1 1
− g mD T
2 4
−8 g mq mD T
2 ,
3 p p2 + m2D Q d−1 p p 2 + m2 Q4
{Q} D {Q}
1
αs
=− αs 1 + 12µ̂2 mD T 3 − 2 m2q m2D T
12 4π
    2ǫ
αs 1 Λ 2ǫ Λ
− − 1 − ℵ(z) . (A.21)
48π 2 ǫ 4πT 2mD

A.3 Three-loop sum-integrals

We now list the mass-expanded sum-integrals necessary at three loops. As before we orga-
nize the contributions according to whether the momentum flowing in a given propagator
is hard or soft.

(hhh) contribution
f f
The (hhh) contributions from diagrams F3a and F3b are
" Z Z Z
X 1 X 1 X 1
f (hhh) 4
F3a = g (d − 1) 4 2
−2 2
P P Q R (Q + R)2
2 2
P {P } {QR}
Z
X
1 1
− (d − 7)
2 P 2 Q2 R2 (P+ Q + R)2
{P QR}
Z Z #
X 1 X (P − Q)2
+(d − 3) +2
P 2 Q2 (P − R)2 (Q − R)2 P 2 Q2 R2 (P − R)2 (Q − R)2
{P QR} {P Q}R
n   1 o
f0,0 + 2N1,−1 .
= g4 (d − 1) 4 I10 − 2Ĩ10 τ̃ − (d − 7) N0,0 + (d − 3)M (A.22)
2

– 30 –
 " Z Z Z Z
X 1 X 1 2 X 1 X 1
f (hhh) 4 2
F3b = −g (d − 1) 2
− 2 4
− 2
P P P P Q R (Q + R)2
2 2 2
P {P } {P } {P QR}
Z Z #
X 1 X (P − Q)2
+ +
P Q (P − R)2 (Q − R)2
2 2 P 2 Q2 R2 (P − R)2 (Q − R)2
{P QR} {P QR}
n 2 o
f0,0 + M
= −g4 (d − 1)2 I10 − Ĩ10 Ĩ20 − 2Ĩ10 τ̃ + M f1,−1 . (A.23)

f f f
Expressions for F3a , F3b , and F3c can be found in ref. [26, 27]. The results are
f (hhh) f (hhh)
F3a + F3b
"
α2s T 4 ζ ′ (−1) ζ ′ (−1) 2
= 35 − 32 + 472µ̂2 + 384 µ̂ + 1328µ̂4
192 ζ(−1) ζ(−1)
#
 
+64 − 36iµ̂ℵ(2, z) + 6(1 + 8µ̂2 )ℵ(1, z) + 3iµ̂(1 + 4µ̂2 )ℵ(0, z) . (A.24)

 6ǫ "  
f (hhh) 5α2 Λ 72 2 144 4 1 31 6 68 ζ ′ (−3)
F3c = − s T4 1 + µ̂ + µ̂ + + γE −
216 4πT 5 5 ǫ 10 5 25 ζ(−3)
12 8 ζ ′ (−1)
+ (25 + 12γE )µ̂2 + 120µ̂4 − (1 + 12µ̂2 )
5 5 ζ(−1)
144 h
− 8ℵ(3, z) + 3ℵ(3, 2z) + 12iµ̂ (ℵ(2, z) + ℵ(2, 2z)) − (3 + 20µ̂2 )ℵ(1, z)
5 #
i
− iµ̂(1 + 12µ̂2 ) ℵ(0, z) − 12µ̂2 ℵ(1, 2z) , (A.25)

– 31 –
where ℵ(n, z) is defined in appendix D. The (hhh) contribution proportional to cA sF is
1 f (hhh) f (hhh)
− F + F3m+3n
2 3a ( Z
X Z
X
2 1 1 1
= g (d − 1) 2(d − 5) 4 2 2
+ (d − 3)
P Q R 2 P Q (P − Q)2 (Q − R)2
2 2
P Q{R} {P QR}
Z
X Z
X
1 1 (P − Q)2
− (d − 7) +
4 P 2 Q2 (P − R)2 (Q − R)2 P 2 Q2 R2 (P − R)2 (Q − R)2
{P Q}R {P Q}R
Z
X Z
X
1 1
−(d − 3) −2
P 2 Q2 R2 (Q − R)2 P 2 Q2 R2 (Q − R)2
P {QR} {P QR}
Z )
X R4
+2
P 2 Q2 (P − Q)4 (Q − R)4
{P QR}
" #
2 1 1
f00 − (d − 7)N00 + N1,−1 − (d − 3)I1 τ̃ − 2Ĩ1 τ̃
= g (d − 1) 2(d − 5)I20 I10 Ĩ10
+ (d − 3)M 0 0
2 4
"  !  
25α2s T 4 72 2 1584 4 1 Λ̂ 369 2840 2 28720 4
=− 1 + µ̂ − µ̂ + 6 ln − 1+ µ̂ + µ̂
864 25 25 ǫ 2 250 123 123
48
536 ζ ′ (−3) 32
ζ ′ (−1) 288 h
+ 1 + 12µ̂2 γE + + 1 + 6µ̂2 + 26ℵ(3, z)
25 125 ζ(−3) 25 ζ(−1) 25
#

i
2
+ 3 − 68µ̂ ℵ(1, z) + 72iµ̂ℵ(2, z) + 2iµ̂ℵ(0, z) , (A.26)

where the integrals appearing above are evaluated in app. B.3. Finally, we note that there
f
is no (hhh) contribution from F3o since this is a purely HTL diagram.

(hhs) contribution
The (hhs) contribution to the F3a , F3b , and F3c+3j are
Z " Z
f (hhs) 1 X 4Q0 R0
4
F3a = 2(d − 1)g T 2
2
p + mD Q R4 (Q + R)2
2
p Q{R}
Z Z #
X 1 X 2Q0 R0
− + , (A.27)
Q2 R2 (Q − R)2 Q4 R 4
{QR} {QR}


Z Z Z
f (hhs) 4 1  X 4Q0 R0 X 1
F3b = 2(d − 1)g T 2  − 2 4 2
+
2
p + mD Q R (Q + R) Q R (Q − R)2
2 2
p Q{R} {QR}
 
Z
X 1 X 1Z Z
X 1 
+ (d − 3) 4  2
−  , (A.28)
Q R R2
{Q} R {R}

– 32 –
  2
Z Z  
f (hhs) 1 X 1 2q 2 
F3c+3j = −4g4 T 2  − 
(p2 + mD )2 Q2 Q4
p {Q}
Z Z 
X Z  
p2 1 2q 2 X 1 2 r2 8 r4
+8g4 T − 4 − (3 + d) 6 +
(p2 + m2D )2 Q2 Q R4 d R d R8
p {Q} {R}
Z Z 
X Z "  #
1 1 2q 2 X 3 4r 2 4 2 1
−16m2q g4 T − 4 − 6 − 4 TR − 2
(p2 + m2D )2 Q2 Q R4 R R R (R·Y )2 ŷ
p {Q} {R}
Z Z
X Z Z
X
4 2 1 1 4 p2 1
= −4g T (d − 1) + g4 T (d − 1)2
(p + m2D )2
2 2
Q R 2 3 2 2
(p + mD )2 Q R4
2
p {QR} p {QR}
Z Z
X
1 1
+32m2q g4 T (d − 3) . (A.29)
(p2 + m2D )2 Q2 R 4
p {QR}

Computing the necessary sum-integrals one finds

f (hhs) α2s mD T 3
F3a+3b = (1 + 12µ̂2 ) , (A.30)
4π
and
   
f (hhs) α2s mD T 3 1 + 12µ̂2 2 4
F3c+3j = + (1 + 12µ̂ ) − ℵ(z) + 24ℵ(1, z)
12π ǫ 3
πα2s T 5
2 α2s m2q T 3

− 1 + 12µ̂2 − 1 + 12µ̂2 . (A.31)
18mD 3πmD

– 33 –
Similarly, one obtains
1 f (hhs) f (hhs)
− F3a + F3m+3n+3o
2 ( Z Z
X
2 2 1 1
= g T (d − 1) 2(d − 1) 2
2
(p + mD )2 Q R2
2
p Q{R}
Z Z
X
1 1 1
+ (d − 3)
2 (p2 + m2D ) Q2 R2 (Q − R)2
p {QR}
Z Z
X Z Z
X
1 2Q0 R0 1 2
p2 1
− 2 4 4
− d − 11d + 46 2
2
(p + mD ) Q R 3 2
(p + mD )2 Q R2
4
p {QR} p Q{R}
Z Z
X
1 p2 1
− (d − 1)2
3 (p2 + m2D )2 Q2 R 4
p Q{R}
Z Z
X Z "   #)
1 1 X 3 4r 2 4 2 1
+4m2q (d − 1) − 6 − 4 TR − 2 ,
(p2 + m2D )2 Q2 R4 R R R (R.Y )2 ŷ
p Q {R}
( Z Z
1 X 1
2 2
= g T (d − 1) 2(d − 1) 2
2
(p + mD )2 Q R2
2
p Q{R}
Z Z
X
1 2
p2 1
− d − 11d + 46
3 (p2 + m2D ) Q4 R 2
p Q{R}
Z Z Z Z )
1 p2 X 1 1 X 1
− (d − 1)2 + 8m2q (d − 3)
3 (p2 + m2D )2 Q2 R 4 (p2 + m2D )2 Q2 R 4
p Q{R} p Q{R}
 2ǫ  4ǫ "
αs mD T 3 Λ 1 Λ
88

=− 7 + 132µ̂2 + + 440µ̂2 + 22 1 + 12µ̂2 γE
48π 2mD ǫ 4πT 3
#
ζ ′ (−1) πα2s T 5
α2s
−8 + 4ℵ(z) + 264ℵ(1, z) − 1 + 12µ̂2 − m2 T 3 . (A.32)
ζ(−1) 9mD 3πmD q

– 34 –
(hss) contribution
f
The only three-loop diagram involving a fermionic line that has a (hss) contribution is F3n
which can be written as
Z "
f (hss) 4 2
F3n = g4 T 2 2

2

2 −

2
2 2
p + mD q + mD (p + q) p + mD q 2 + m2D
2 2
pq
#Z  
8m2D X 1 2r 2
−

2 − 4
(p + q)2 p2 + m2 q 2 + m2 R2 R
D D {R}
Z "
4 2
= −g 4 T 2 (d − 1)

2 −

2
p2 + m2D q2 2
+ mD (p + q) p2 + m2D q 2 + m2D
pq
#Z
8m2D X 1
−

2
(p + q)2 p2 + m2 q 2 + m2 R2
D D {R}
  4ǫ  2ǫ
α2s T 4 1+ 12µ̂2
Λ Λ
= + 2 1 + 12µ̂2 + 12ℵ(1, z) (A.33)
12 ǫ 2mD 4πT

B Sum-Integrals

We can define a set of “master” sum-integrals as in [26, 27]

Z
X
m P0m
In = , (B.1)
P 2n
P
Z
X
e m P0m
In = , (B.2)
P 2n
{P }
Z
X 1
τ̃ = , (B.3)
P 2 Q2 (P + Q)2
{P Q}
Z
X 1
Mm,n = , (B.4)
P 2 Q2 (R2 )m [(P − Q)2 ]n (P − R)2 (Q − R)2
P QR
Z
X
fm,n = 1
M , (B.5)
P 2 Q2 (R2 )m [(P − Q)2 ]n (P − R)2 (Q − R)2
{P QR}
Z
X 1
Nm,n = . (B.6)
P 2 Q2 (R2 )m [(P − Q)2 ]n (P − R)2 (Q − R)2
{P Q}R

B.1 One-loop sum-integrals

The specific bosonic sun-integrals needed are
Z     
X 1 T2 Λg 2ǫ ζ ′ (−1)
I10 = = 1 + 2ǫ 1 + , (B.7)
P2 12 4πT ζ(−1)
P

– 35 –
 Z    
X 1 1 Λg 2ǫ 1
I20 = = + 2γE . (B.8)
P4 (4π)2 4πT ǫ
P
The specific fermionic sun-integrals needed are
Z  
e 0
X 1 T2 Λ 2ǫ 

I1 = 2
=− 1 + 12µ̂2 + 2ǫ 1 + 12µ̂2 + 12ℵ(1, z) , (B.9)
P 24 4πT
{P }

and
Z    
X 1 1 Λ 2ǫ 1
Ie20 = = − ℵ(z) . (B.10)
P4 (4π)2 4πT ǫ
{P }

Using the two basic one-loop sum-integrals above, we can construct other one-loop
sum-integrals that will be necessary here as follows:
XZ Z
1 2 X p2
= (B.11)
P2 d P4
{P } {P }

Z
X Z Z
1 4 X p2 24 X p4
= = ,
P4 d P6 d(d + 2) P8
{P } {P } {P }
Z
X Z  
1 2 X 1 1
= (d − 1) TP = − . (B.12)
P4 d−1 P 2 (P · Y )2 ŷ
{P } {P }

This allows us to compute the following sum-integrals

Z    
X p2 T2 Λ 2ǫ 2 4 2


=− 1 + 12µ̂ + ǫ 1 + 12µ̂ + 18ℵ(1, z) , (B.13)
P4 16 4πT 3
{P }

Z    
X 1 1 1 Λ 2ǫ 1
TP = + 1 − ℵ(z) , (B.14)
P4 2 (4π)2 4πT ǫ
{P }

Z
X  2ǫ  
p2 3 1 Λ 1 2
= − − ℵ(z) , (B.15)
P6 4 (4π)2 4πT ǫ 3
{P }

Z
X  2ǫ  
p4 5 1 Λ 1 16
= − − ℵ(z) , (B.16)
P 8 8 (4π)2 4πT ǫ 15
{P }

Z      
X 1 1 1 Λ 2ǫ 1
=− − 1 − ℵ(z) , (B.17)
P 2 (P.Y )2 ŷ (4π)2 4πT ǫ
{P }

Z
X  2ǫ
P0 1 Λ
4
= [iµ̂ + ℵ(0, z)ǫ] . (B.18)
P (4π) 4πT
{P }

– 36 –
B.2 Two-loop sum-integrals
For the purposes of this paper we only need one new two-loop sum-integral
Z
X  4ǫ  
1 T2 Λ µ̂2
τ̃ = = − + 2µ̂2 − 2iµ̂ℵ[0, z] . (B.19)
P 2 Q2 (P + Q)2 (4π)2 4πT ǫ
{P Q}

B.3 Three-loop sum-integrals

The three-loop sum-integrals necessary are
Z
X 1
M00 =
P Q R (P + Q + R)2
2 2 2
P QR
 2  6ǫ  
1 T2 Λg 6 182 ζ ′ (−3) ζ ′ (−1)
= + − 12 + 48 . (B.20)
(4π)2 12 4πT ǫ 5 ζ(−3) ζ(−1)

Z
X 1
N00 =
P 2 Q2 R2 (P + Q + R)2
{P QR}
 2 2   "
1 T Λ 6ǫ 3

2 2 173 2 4 24 ζ ′ (−3)
= 1 + 12µ̂ + + 210µ̂ + 1284µ̂ −
(4π)2 12 4πT 2ǫ 20 5 ζ(−3)
#
h i
2 2
+ 144 (1 + 8µ̂ )ℵ(1, z) + 4µ̂ ℵ(1, 2z) − 4iµ̂ [ℵ(2, z) + ℵ(2, 2z)] − 2ℵ(3, z) − ℵ(3, 2z) .

(B.21)

Z
X
f00 = 1
M
+ Q + R)2
P 2 Q2 R2 (P
P Q{R}
 2 2   "
1 T Λ 6ǫ 3 2 4

179
=− 1 + 24µ̂ − 48µ̂ + + 111µ̂2 − 210µ̂4
(4π)2 12 4πT 4ǫ 40
#
ζ ′ (−1) 2 24 ζ ′ (−3) h i
+ 48 µ̂ + + 72 (1 − 8µ̂2 )ℵ(1, z) + 6ℵ(3, z) + 12iµ̂ℵ(2, z) .
ζ(−1) 5 ζ(−3)
(B.22)

 2  6ǫ 
1 T2 Λ 3

173
N1,−1 = − 2 1 + 12µ̂2 1 − 4µ̂2 + + + 114µ̂2
2 (4π) 12 4πT2ǫ 20
12 ζ ′ (−3) ζ ′ (−1) h
+ 132µ̂4 − − 96µ̂2 − 144 2ℵ(3, z) + 2ℵ(3, 2z) − 4iµ̂ ℵ(2, z)
5 ζ(−3) ζ(−1)

1
i
2 2 2
+ 8iµ̂ ℵ(2, 2z) − 1 − 4µ̂ ℵ(1, z) − 8µ̂ ℵ(1, 2z) − iµ̂ 1 + 12µ̂ ℵ(0, z) .
3
(B.23)

– 37 –
 Z
X Q·R
H3 =
P 2 Q2 R2 (P + Q)2 (P + R)2
{P }QR
 2 2   "
1 T Λ 6ǫ 3

2 2 361 3 ζ ′ (−3)
= 1 + 12µ̂ + −
(4π)2 12 4πT 8ǫ 160 5 ζ(−3)
( 
141 2 501 4 1

+ µ̂ + µ̂ − 9 2 4
+ µ̂ + 2µ̂ ℵ(z) + 2iµ̂ 1 + 4µ̂2 ℵ(0, z)
4 2 8
)#
2


+ 2 1 − 12µ̂ ℵ(1, z) + 24iµ̂ℵ(2, z) + 16ℵ(3, z) . (B.24)

Z
X
f−2,2 = R4
M
P 2 Q2 (P − Q)4 (Q − R)2 (R − P )2
{P QR}
 2  6ǫ "
1 T2 Λ 1
89
=− 29 + 288µ̂2 − 144µ̂4 +
(4π)2 12 4πT12ǫ 12
  ′
10 84 ζ (−1) 8 ζ ′ (−3)
+4γE + 2(43 + 24γE )µ̂2 − 68µ̂4 + 1 + µ̂2 +
3 5 ζ(−1) 3 ζ(−3)
#
 2

+24 10ℵ(3, z) + 18iµ̂ℵ(2, z) + 2(2 − 5µ̂ )ℵ(1, z) + iµ̂ℵ(0, z) . (B.25)

C Three-dimensional integrals

Dimensional regularization can be used to regularize both the ultraviolet divergences and
infrared divergences in 3-dimensional integrals over momenta. The spatial dimension is
generalized to d = 3 − 2ǫ dimensions. Integrals are evaluated at a value of d for which they
converge and then analytically continued to d = 3. We use the integration measure
Z  γE 2 ǫ Z
e Λ d3−2ǫ p
≡ . (C.1)
4π (2π)3−2ǫ
p

C.1 One-loop integrals

The general one-loop integral is given by
Z
1
In ≡
(p + m2 )n
2
p

1 γE 2 ǫ Γ(n − 32 + ǫ) 3−2n−2ǫ
= (e Λ ) m . (C.2)
8π Γ( 12 )Γ(n)

– 38 –
Specifically, we need
Z
I0′ ≡ log(p2 + m2 )
p
  
m3 Λ 2ǫ 8 2


=− 1+ ǫ+O ǫ , (C.3)
6π2m 3
 
m Λ 2ǫ 

I1 = − 1 + 2ǫ + O ǫ2 , (C.4)
4π 2m
 
1 Λ 2ǫ
I2 = [1 + O (ǫ)] . (C.5)
8πm 2m

C.2 Two-loop integrals

We also need a few two-loop integrals on the form
Z
1 1 1
Jn = . (C.6)
p + m (q + m ) (p + q)2
2 2 2 2 n
pq

Specifically, we need J1 and J2 which were calculated in refs. [50]:

 4ǫ  
1 Λ 1
J1 = + 2 + O(ǫ) , (C.7)
4(4π)2 2m ǫ
 4ǫ
1 Λ
J2 = [1 + O(ǫ)] . (C.8)
4(4π)2 m2 2m

D Properties of the ℵ functions

For some frequently occurring combinations of special functions we will apply the following
abbreviations

ζ ′ (x, y) ≡ ∂x ζ(x, y) , (D.1)

n+1 ′
ℵ(n, z) ≡ ζ ′ (−n, z) + (−1) ζ (−n, z ∗ ) , (D.2)
∗
ℵ(z) ≡ Ψ(z) + Ψ(z ) , (D.3)

where n is assumed to be a non-negative integer and z is a general complex number given

here by z = 1/2 − iµ̂. Above ζ denotes the Riemann zeta function, and Ψ is the digamma
function
Γ′ (z)
Ψ(z) ≡ . (D.4)
Γ(z)

Below we list Taylor expansions of the function ℵ(z) and ℵ(n, z) for values of n necessary
for calculation of the susceptibilities presented in the main text. For general application

– 39 –
we evaluate the ℵ functions exactly using Mathematica.

ℵ(z) = −2γE − 4 ln 2 + 14ζ(3)µ̂2 − 62ζ(5)µ̂4 + 254ζ(7)µ̂6 + O(µ̂8 ) , (D.5)

14 62
ℵ(0, z) = 2 (2 ln 2 + γE ) iµ̂ − ζ(3)iµ̂3 + ζ(5)iµ̂5 + O(µ̂7 ) , (D.6)
3 5
 
1 ζ ′ (−1) 7
ℵ(1, z) = − ln 2 − − (1 − 2 ln 2 − γE ) µ̂2 − ζ(3)µ̂4
12 ζ(−1) 6
31
+ ζ(5)µ̂6 + O(µ̂8 ) , (D.7)
15

1 ζ ′ (−1)
2 ζ(3)
4
ℵ(1, z + z ′ ) = − − (1 − γE ) µ̂ + µ̂′ − µ̂ + µ̂′
6 ζ(−1) 6
ζ(5)
6
+ µ̂ + µ̂′ + O(µ̂8 , µ̂′8 ) , (D.8)
15 
1 ζ ′ (−1) 1
ℵ(2, z) = 1 + 2 ln 2 − 2 iµ̂ + (3 − 2γE − 4 ln 2) iµ̂3
12 ζ(−1) 3
7
+ ζ(3)iµ̂5 + O(µ̂7 ) , (D.9)
15
1 ζ ′ (−1) 
1
3
ℵ(2, z + z ′ ) = − 1 − 2 i µ̂ + µ̂′ + (3 − 2γE ) i µ̂ + µ̂′
6 ζ(−1) 3
ζ(3)
5
+ i µ̂ + µ̂′ + O(µ̂7 , µ̂′7 ), (D.10)
15
   
1 ζ ′ (−3) 1 ζ ′ (−1)
ℵ(3, z) = ln 2 − 7 + 5 + 6 ln 2 − 6 µ̂2
480 ζ(−3) 24 ζ(−1)
1 7
+ (11 − 6γE − 12 ln 2) µ̂4 + ζ(3)µ̂6 + O(µ̂8 ) , (D.11)
12 30

1 ζ ′ (−3) 1 ζ ′ (−1) 
2
ℵ(3, z + z ′ ) = − 5−6 µ̂ + µ̂′
60 ζ(−3) 12 ζ(−1)
1
4 ζ(3)
6
+ (11 − 6γE ) µ̂ + µ̂′ + µ̂ + µ̂′ + O(µ̂8 , µ̂′8 ) . (D.12)
12 30

References
[1] S. Borsanyi, G. Endrodi, Z. Fodor, A. Jakovac, S. D. Katz, S. Krieg, C. Ratti and
K. K. Szabo, The QCD equation of state with dynamical quarks, JHEP 1011, 077 (2010).
[2] S. Borsanyi, Z. Fodor, S. D. Katz, S. Krieg, C. Ratti and K. Szabo, Fluctuations of
conserved charges at finite temperature from lattice QCD, JHEP 1201, 138 (2012)
[3] S. Borsanyi, S. Durr, Z. Fodor, C. Hoelbling, S. D. Katz, S. Krieg, D. Nogradi and
K. K. Szabo et al., QCD thermodynamics with continuum extrapolated Wilson fermions I,
JHEP 1208, 126 (2012).
[4] Sz. Borsányi, G. Endrődi, Z. Fodor, S.D. Katz, S. Krieg, C. Ratti and K.K. Szabó, QCD
equation of state at nonzero chemical potential: continuum results with physical quark masses
at order µ2 , JHEP 08, 053 (2012).

– 40 –
 [5] S. Borsanyi, Thermodynamics of the QCD transition from lattice, Nucl. Phys. A904-905
2013, 270c (2013).
[6] S. Borsanyi, Z. Fodor, S. D. Katz, S. Krieg, C. Ratti and K. K. Szabo, Freeze-out
parameters: lattice meets experiment, Phys. Rev. Lett. 111, 062005 (2013).
[7] S. Sharma, QCD Thermodynamics on the Lattice, Adv. High Energy Phys. 2013, 452978
(2013).
[8] F. Karsch, B. -J. Schaefer, M. Wagner and J. Wambach, Towards finite density QCD with
Taylor expansions, Phys. Lett. B 698, 256 (2011).
[9] A. Bazavov, H. -T. Ding, P. Hegde, O. Kaczmarek, F. Karsch, E. Laermann, Y. Maezawa
and S. Mukherjee et al., Strangeness at high temperatures: from hadrons to quarks, Phys.
Rev. Lett. 111, 082301 (2013).
[10] A. Bazavov, H. -T. Ding, P. Hegde, F. Karsch, C. Miao, S. Mukherjee, P. Petreczky and
C. Schmidt et al., Quark number susceptibilities at high temperatures, arXiv:1309.2317
[hep-lat].
[11] A. Bazavov, H. T. Ding, P. Hegde, O. Kaczmarek, F. Karsch, E. Laermann, S. Mukherjee
and P. Petreczky et al., Freeze-out Conditions in Heavy Ion Collisions from QCD
Thermodynamics, Phys. Rev. Lett. 109, 192302 (2012).
[12] C. Bernard et al., QCD thermodynamics with three flavors of improved staggered quarks,
Phys. Rev. D 71, 034504 (2005).
[13] A. Bazavov, T. Bhattacharya, M. Cheng, N. H. Christ, C. DeTar, S. Ejiri, S. Gottlieb and
R. Gupta et al., Equation of state and QCD transition at finite temperature, Phys. Rev. D
80, 014504 (2009).
[14] A. Bazavov et al., Fluctuations and Correlations of net baryon number, electric charge, and
strangeness: A comparison of lattice QCD results with the hadron resonance gas model,
Phys. Rev. D 86, 034509 (2012).
[15] P. Petreczky, Lattice QCD at non-zero temperature, J. Phys. G 39, 093002 (2012).
[16] E. V. Shuryak, Theory of Hadronic Plasma, Sov. Phys. JETP 47, 212 (1978).
[17] S.A. Chin, Transition to Hot Quark Matter in Relativistic Heavy Ion Collision, Phys. Lett. B
78, 552 (1978).
[18] J. I. Kapusta, Quantum Chromodynamics at High Temperature, Nucl. Phys. B 148, 461
(1979).
[19] T. Toimela, The Next Term in the Thermodynamic Potential of QCD, Phys. Lett. B 124,
407 (1983).
[20] P. Arnold, C. X. Zhai, The Three loop free energy for pure gauge QCD, Phys. Rev. D 50,
7603 (1994).
[21] P. Arnold, C. X. Zhai, The Three loop free energy for high temperature QED and QCD with
fermions, Phys. Rev. D 51, 1906 (1995).
[22] C. X. Zhai and B. M. Kastening, The Free energy of hot gauge theories with fermions
through g 5 , Phys. Rev. D 52, 7232 (1995).
[23] E. Braaten and A. Nieto, Effective field theory approach to high temperature
thermodynamics, Phys. Rev. D 51, 6990 (1995).

– 41 –
[24] E. Braaten and A. Nieto, Free energy of QCD at high temperature, Phys. Rev. D 53, 3421
(1996).
[25] K. Kajantie, M. Laine, K. Rummukainen and Y. Schroder, The Pressure of hot QCD up to
g 6 ln 1/g , Phys. Rev. D 67, 105008 (2003).
[26] A. Vuorinen, Quark number susceptibilities of hot QCD up to g 6 ln g, Phys. Rev. D 67,
074032 (2003).
[27] A. Vuorinen, The Pressure of QCD at finite temperatures and chemical potentials, Phys. Rev.
D 68, 054017 (2003).
[28] A. Ipp, K. Kajantie, A. Rebhan, and A. Vuorinen, The Pressure of deconfined QCD for all
temperatures and quark chemical potentials, Phys. Rev. D 74, 045016 (2006).
[29] F. Karsch, A. Patkos and P. Petreczky, Screened perturbation theory, Phys. Lett. B 401, 69
(1997).
[30] S. Chiku and T. Hatsuda, Optimized perturbation theory at finite temperature, Phys. Rev. D
58, 076001 (1998).
[31] J.O. Andersen, E. Braaten and M. Strickland, Screened perturbation theory to three loops,
Phys. Rev. D 63, 105008 (2001).
[32] J.O. Andersen and L. Kyllingstad, Four-loop screened perturbation theory, Phys. Rev. D 78,
076008 (2008).
[33] J.O. Andersen and M. Strickland, Mass expansions of screened perturbation theory, Phys.
Rev. D 64, 105012 (2001) .
[34] V.I. Yukalov, Remarks on quasiaverages, Teor. Mat. Fiz. 26 403, (1976).
[35] P.M. Stevenson, Optimized perturbation theory, Phys. Rev. D 23, 2916 (1981).
[36] A. Duncan and M. Moshe, Nonperturbative physics from interpolating actions, Phys. Lett. B
215, 352 (1988).
[37] A. Duncan and H.F. Jones, Convergence proof for optimized δ expansion: the anharmonic
oscillator, Phys. Rev. D 47, 2560 (1993).
[38] A.N. Sisakian, I.L. Solovtsov, and O. Shevchenko, Variational perturbation theory, Int. J.
Mod. Phys. A 9, 1929 (1994).
[39] W. Janke and H. Kleinert, Convergent strong-coupling expansions from divergent
weak-coupling perturbation theory, Phys. Rev. Lett. 75, 2787 (1995).
[40] J. O. Andersen, E. Braaten, and M. Strickland, Hard thermal loop resummation of the free
energy of a hot gluon plasma, Phys. Rev. Lett. 83, 2139 (1999).
[41] J. O. Andersen, E. Braaten, and M. Strickland, Hard thermal loop resummation of the
thermodynamics of a hot gluon plasma, Phys. Rev. D 61, 014017 (1999).
[42] J. O. Andersen, E. Braaten, and M. Strickland, Hard thermal loop resummation of the free
energy of a hot quark - gluon plasma, Phys. Rev. D 61, 074016 (2000).
[43] J. O. Andersen, S. Mogliacci, N. Su and A. Vuorinen, Quark number susceptibilities from
resummed perturbation theory Phys. Rev. D 87, 074003 (2013);
[44] S. Mogliacci, J. O. Andersen, M. Strickland, N. Su and A. Vuorinen, Equation of State of hot
and dense QCD: Resummed perturbation theory confronts lattice data JHEP 1312, 055
(2013).

– 42 –
[45] J. O. Andersen, E. Braaten, E. Petitgirard, and M. Strickland, HTL perturbation theory to
two loops, Phys. Rev. D 66 (2002) 085016.
[46] J. O. Andersen, E. Petitgirard, and M. Strickland, Two loop HTL thermodynamics with
quarks Phys. Rev. D 70, 045001 (2004).
[47] N. Haque, M. G. Mustafa, and M. Strickland, Two-loop HTL pressure at finite temperature
and chemical potential, Phys. Rev. D 87, 105007 (2013).
[48] N. Haque, M. G. Mustafa, and M. Strickland, Quark Number Susceptibilities from Two-Loop
Hard Thermal Loop Perturbation Theory, JHEP 1307, 184 (2013).
[49] N. Su., J. O. Andersen, and M. Strickland, Gluon Thermodynamics at Intermediate
Coupling, Phys. Rev. Lett. 104, 122003 (2010).
[50] J. O. Andersen, M. Strickland, and N. Su, Three-loop HTL gluon thermodynamics at
intermediate coupling, JHEP 1008, 113 (2010).
[51] J. O. Andersen, M. Strickland and N. Su, Three-loop HTL Free Energy for QED, Phys. Rev.
D 80, 085015 (2009).
[52] J.O. Andersen, L.E. Leganger, M. Strickland and N. Su, NNLO hard-thermal-loop
thermodynamics for QCD, Phys. Lett. B 696, 468 (2011).
[53] J. O. Andersen, L. E. Leganger, M. Strickland and N. Su, Three-loop HTL QCD
thermodynamics, JHEP 1108, 053 (2011).
[54] J. O. Andersen, L. E. Leganger, M. Strickland and N. Su, The QCD trace anomaly, Phys.
Rev. D 84, 087703 (2011).
[55] N. Haque, J. O. Andersen, M. G. Mustafa, M. Strickland, and N. Su, Three-loop HTLpt
Pressure and Susceptibilities at Finite Temperature and Density, Phys. Rev. D 89, 061701
(2014).
[56] P. Chakraborty, M. G. Mustafa, and M. H. Thoma, Quark number susceptibility in hard
thermal loop approximation, Eur. Phys. J. C. 23, 591 (2002).
[57] P. Chakraborty, M. G. Mustafa, and M. H. Thoma, Chiral susceptibility in hard thermal loop
approximation, Phys. Rev. D 67, 114004 (2003).
[58] P. Chakraborty, M. G. Mustafa, and M. H. Thoma, Quark number susceptibility,
thermodynamic sum rule, and hard thermal loop approximation, Phys. Rev. D 68, 085012
(2003).
[59] N. Haque, M. G. Mustafa and M. H. Thoma, Conserved Density Fluctuation and Temporal
Correlation Function in HTL Perturbation Theory, Phys. Rev. D 84, 054009 (2011).
[60] N. Haque and M. G. Mustafa, Quark Number Susceptibility and Thermodynamics in HTL
approximation, Nucl. Phys. A 862-863, 271 (2011).
[61] N. Haque and M. G. Mustafa, A Modified Hard Thermal Loop Perturbation Theory ,
arXiv:1007.2076.
[62] J.P. Blaizot, E. Iancu, and A. Rebhan, Comparing different hard thermal loop approaches to
quark number susceptibilities Eur. Phys. J. C 27, 433 (2003).
[63] J.P. Blaizot, E. Iancu, and A. Rebhan Quark number susceptibilities from HTL resummed
thermodynamics, Phys. Lett. B 523, 143 (2001).

– 43 –
[64] J.P. Blaizot, E. Iancu, and A. Rebhan, The Thermodynamics of the quark gluon plasma:
Selfconsistent resummations versus lattice data, Nucl. Phys. A 698, 404 (2002).
[65] J.P. Blaizot, E. Iancu, and A. Rebhan, The Entropy of the QCD plasma, Phys. Rev. Lett.
83, 2906 (1999).
[66] J.P. Blaizot, E. Iancu, and A. Rebhan, Selfconsistent hard thermal loop thermodynamics for
the quark gluon plasma, Phys. Lett. B 470, 181 (1999).
[67] J.P. Blaizot, E. Iancu, and A. Rebhan, Approximately selfconsistent resummations for the
thermodynamics of the quark gluon plasma. 1. Entropy and density, Phys. Rev. D 63, 065003
(2001).
[68] T. Hatsuda and T. Kunihiro, QCD phenomenology based on a chiral effective Lagrangian,
Phys. Rep. 247, 221 (1994).
[69] T. Kunihiro, Quark number susceptibility and fluctuations in the vector channel at high
temperatures, Phys. Lett. B 271, 395 (1991).
[70] K. Fukushima, Relation between the Polyakov loop and the chiral order parameter at strong
coupling, Phys. Rev. D 68, 045004 (2003).
[71] K. Fukushima, Chiral effective model with the Polyakov loop, Phys. Lett. B 591, 277 (2004).
[72] C. Ratti, M. A. Thaller, and W. Weise, Phases of QCD: Lattice thermodynamics and a field
theoretical model, Phys. Rev. D 73, 014019 (2006).
[73] S. K. Ghosh, T. K. Mukherjee, M. G. Mustafa, and R. Ray, Susceptibilities and speed of
sound from PNJL model, Phys. Rev. D 73, 114007 (2006).
[74] S. K. Ghosh, T. K. Mukherjee, M. G. Mustafa, and R. Ray, PNJL model with a Van der
Monde term, Phys. Rev. D 77, 094024 (2008).
[75] S. Mukherjee, M. G. Mustafa, and R. Ray, Thermodynamics of the PNJL model with nonzero
baryon and isospin chemical potentials , Phys. Rev. D 75, 094015 (2007).
[76] S. Roessner, C. Ratti, and W. Weise, Polyakov loop, diquarks and the two-flavour phase
diagram, Phys. Rev. D 75, 034007 (2007).
[77] C. Sasaki, B. Friman, and K. Redlich, Susceptibilities and the Phase Structure of a Chiral
Model with Polyakov Loops, Phys. Rev. D 75 074013 (2007).
[78] A. Bhattacharyya, P. Dev, S. K. Ghosh, and R. Ray, Investigation of Phase Diagram and
Bulk Thermodynamic Properties using PNJL Model with Eight-Quark Interactions, Phys.
Rev. D 82, 014021 (2010).
[79] A. Bhattacharyya, P. Dev, A. Lahiri, and R. Ray, Susceptibilities with multi-quark
interactions in PNJL model, Phys. Rev. D 82, 114028 (2010).
[80] A. Bhattacharyya, P. Dev, A. Lahiri, and R. Ray, Correlation between conserved charges in
PNJL Model with multi-quark interactions, Phys. Rev. D 83, 014011 (2011).
[81] M. Bluhm, B. Kampfer, R. Schulze, D. Seipt and U. Heinz, A family of equations of state
based on lattice QCD: Impact on flow in ultrarelativistic heavy-ion collisions, Phys. Rev. C
76, 034901 (2007).
[82] M. Bluhm and B. Kampfer, Quasiparticle model of quark-gluon plasma at imaginary
chemical potential, Phys. Rev. D 77, 034004 (2008).

– 44 –
[83] V. M. Bannur, Quasi-particle model for QGP with nonzero densities, JHEP 0709, 046
(2007).
[84] V. M. Bannur, Self-consistent quasiparticle model for 2, 3 and (2+1) flavor QGP, Phys. Rev.
C 78, 045206 (2008).
[85] F. G. Gardim and F. M. Steffens, Thermodynamics of Quasi-Particles at Finite Chemical
Potential, Nucl. Phys. A 825, 222 (2009).
[86] B. -J. Schaefer, M. Wagner and J. Wambach, QCD thermodynamics with effective models,
PoS CPOD 2009, 017 (2009)
[87] B. -J. Schaefer, M. Wagner and J. Wambach, Thermodynamics of (2+1)-flavor QCD:
Confronting Models with Lattice Studies, Phys. Rev. D 81, 074013 (2010).
[88] V. Skokov, B. Friman and K. Redlich, Quark number fluctuations in the Polyakov
loop-extended quark-meson model at finite baryon density, Phys. Rev. C 83, 054904 (2011).
[89] S. Chatterjee and K. A. Mohan, Fluctuations and Correlations of Conserved Charges in the
(2 + 1) Polyakov Quark Meson Model, Phys. Rev. D 86, 114021 (2012).
[90] G. Cvetic and R. Kogerler, Resummations of free energy at high temperature, Phys. Rev. D
66, 105009 (2002).
[91] G. Cvetic and R. Kogerler, Resummations of the pressure of quark-gluon plasma by including
contributions of order g**6(s), Phys. Rev. D 70, 114016 (2004).
[92] G. Cvetic and R. Koegerler, Method for comparing finite temperature field theory results with
lattice data, Phys. Rev. D 75, 054016 (2007).
[93] K. -i. Kim, Y. Kim, S. Takeuchi and T. Tsukioka, Quark Number Susceptibility with Finite
Quark Mass in Holographic QCD, Prog. Theor. Phys. 126, 735 (2011).
[94] L. -X. Cui, S. Takeuchi and Y. -L. Wu, Quark Number Susceptibility and QCD Phase
Transition in the Predictive Soft-wall AdS/QCD Model with Finite Temperature, Phys. Rev.
D 84, 076004 (2011).
[95] E. Braaten and R. D. Pisarski, Simple effective Lagrangian for hard thermal loops, Phys.
Rev. D 45, 1827 (1992).
[96] E. Braaten, R. D. Pisarski and T. -C. Yuan, Production of Soft Dileptons in the Quark -
Gluon Plasma, Phys. Rev. Lett. 64, 2242 (1990).
[97] C. Greiner, N. Haque, M. G. Mustafa and M. H. Thoma, Low Mass Dilepton Rate from the
Deconfined Phase, Phys. Rev. C 83, 014908 (2011).
[98] R. Baier, S. Peigne and D. Schiff, Soft photon production rate in resummed perturbation
theory of high temperature QCD, Z. Phys. C 62, 337 (1994).
[99] M. G. Mustafa, M. H. Thoma, and P. Chakraborty, Screening of moving parton in the quark
gluon plasma, Phys.Rev. C71, 017901 (2005).
[100] M. G. Mustafa, P. Chakraborty, and M. H. Thoma, Dynamical screening in a quark gluon
plasma, J. Phys. Conf. Ser. 50, 438 (2006).
[101] P. Chakraborty, M. G. Mustafa, and M. H. Thoma, Wakes in the quark-gluon plasma, Phys.
Rev. D 74, 094002 (2006).
[102] P. Chakraborty, M. G. Mustafa, R. Ray, and M. H. Thoma, Wakes in a Collisional
Quark-Gluon Plasma, J. Phys. G 34, 2141 (2007).

– 45 –
[103] M. Laine, O. Philipsen, P. Romatschke and M. Tassler, Real-time static potential in hot
QCD, JHEP 0703, 054 (2007).
[104] A. Dumitru, Y. Guo and M. Strickland, The heavy-quark potential in an anisotropic plasma,
Phys. Lett. B 662, 37 (2008).
[105] A. Dumitru, Y. Guo, A. Mocsy and M. Strickland, Quarkonium states in an anisotropic
QCD plasma, Phys. Rev. D 79, 054019 (2009).
[106] A. Dumitru, Y. Guo and M. Strickland, The Imaginary part of the static gluon propagator
in an anisotropic (viscous) QCD plasma, Phys. Rev. D 79, 114003 (2009).
[107] L. Thakur, N. Haque, U. Kakade and B. K. Patra, Dissociation of quarkonium in an
anisotropic hot QCD medium , Phys. Rev. D 88, 054022 (2013).
[108] R. D. Pisarski, Damping rates for moving particles in hot QCD, Phys. Rev. D 47, 5589
(1993).
[109] S. Peigne, E. Pilon and D. Schiff, The Heavy fermion damping rate puzzle, Z. Phys. C 60,
455 (1993).
[110] Sreemoyee Sarkar and Abhee K Dutt-Mazumder, Heavy quark damping rate in hot viscous
QCD plasma, Phys. Rev. D 88, 054006(2013).
[111] A. Abada and N. Daira-Aifa, Photon Damping in One-Loop HTL Perturbation Theory,
JHEP 1204, 071 (2012).
[112] E. Braaten and R. D. Pisarski, Resummation and Gauge Invariance of the Gluon Damping
Rate in Hot QCD, Phys. Rev. Lett. 64, 1338 (1990).
[113] E. Braaten and R. D. Pisarski, Calculation of the gluon damping rate in hot QCD, Phys.
Rev. D 42, 2156 (1990).
[114] E. Braaten and M. H. Thoma, Energy loss of a heavy fermion in a hot plasma, Phys. Rev.
D 44, 1298 (1991).
[115] E. Braaten and M. H. Thoma, Energy loss of a heavy quark in the quark - gluon plasma,
Phys. Rev. D 44, 2625 (1991).
[116] M. H. Thoma and M. Gyulassy, Quark Damping and Energy Loss in the High Temperature
QCD, Nucl. Phys. B 351, 491 (1991).
[117] P. Romatschke and M. Strickland, Energy loss of a heavy fermion in an anisotropic QED
plasma, Phys. Rev. D 69, 065005 (2004).
[118] P. Romatschke and M. Strickland, Collisional energy loss of a heavy quark in an anisotropic
quark-gluon plasma, Phys. Rev. D 71, 125008 (2005).
[119] M. G. Mustafa, Energy loss of charm quarks in the quark-gluon plasma: Collisional versus
radiative, Phys. Rev. C 72, 014905 (2005).
[120] M. G. Mustafa and M. H. Thoma, Quenching of hadron spectra due to the collisional energy
loss of partons in the quark gluon plasma, Acta Phys. Hung. A 22, 93 (2005).
[121] M. Djordjevic, Collisional energy loss in a finite size QCD matter, Phys. Rev. C 74, 064907
(2006).
[122] P. Chakraborty, M. G. Mustafa and M. H. Thoma, Energy gain of heavy quarks by
fluctuations in the QGP, Phys. Rev. C 75, 064908 (2007).

– 46 –
[123] G. -Y. Qin, J. Ruppert, C. Gale, S. Jeon, G. D. Moore and M. G. Mustafa, Radiative and
collisional jet energy loss in the quark-gluon plasma at RHIC, Phys. Rev. Lett. 100, 072301
(2008).
[124] G. -Y. Qin, A. Majumder, H. Song and U. Heinz, Energy and momentum deposited into a
QCD medium by a jet shower, Phys. Rev. Lett. 103, 152303 (2009)
[125] G. -Y. Qin and A. Majumder, A pQCD-based description of heavy and light flavor jet
quenching, Phys. Rev. Lett. 105, 262301 (2010).
[126] S. Mrowczynski and M. H. Thoma, Hard loop approach to anisotropic systems, Phys. Rev. D
62, 036011 (2000).
[127] P. Romatschke and M. Strickland, Collective modes of an anisotropic quark gluon plasma,
Phys. Rev. D 68, 036004 (2003).
[128] P. Romatschke and M. Strickland, Collective modes of an anisotropic quark-gluon plasma II,
Phys. Rev. D 70, 116006 (2004).
[129] A. Rebhan, P. Romatschke and M. Strickland, Hard-loop dynamics of non-Abelian plasma
instabilities, Phys. Rev. Lett. 94, 102303 (2005).
[130] B. Schenke and M. Strickland, Fermionic Collective Modes of an Anisotropic Quark-Gluon
Plasma, Phys. Rev. D 74, 065004 (2006).
[131] A. Rebhan, M. Strickland and M. Attems, Instabilities of an anisotropically expanding
non-Abelian plasma: 1D+3V discretized hard-loop simulations, Phys. Rev. D 78, 045023
(2008).
[132] M. Attems, A. Rebhan and M. Strickland, Instabilities of an anisotropically expanding
non-Abelian plasma: 3D+3V discretized hard-loop simulations, Phys. Rev. D 87, 025010
(2013).
[133] P. Graf and F.D. Steffen, Thermal axion production in the primordial quark-gluon plasma,
Phys. Rev. D 83 075011 (2011).
[134] C. Kiessig and M. Plumacher, Hard-Thermal-Loop Corrections in Leptogenesis II: Solving
the Boltzmann Equations, JCAP 1209, 012 (2012).
[135] C. Kiessig and M. Plumacher, Hard-Thermal-Loop Corrections in Leptogenesis I:
CP-Asymmetries, JCAP 1207, 014 (2012).
[136] J. Beringer et al., (Particle Data Group), Phys. Rev. D 86, 010001 (2012).
[137] T. van Ritbergen, J. A. M. Vermaseren and S. A. Larin, The Four loop beta function in
quantum chromodynamics, Phys. Lett. B 400, 379 (1997).
[138] A. Bazavov, N. Brambilla, X. Garcia i Tormo, P. Petreczky, J. Soto and A. Vairo,
Determination of αs from the QCD static energy, Phys. Rev. D 86, 114031 (2012).
[139] S. Datta, R.V. Gavai and S. Gupta, http://www.ilgti.tifr.res.in/tables, to appear in
the proceedings of Lattice 2013.
[140] P. Petreczky et al. [RBC-Bielefeld Collaboration], Quark number fluctuations at high
temperatures, PoS LAT 2009, 159 (2009).

– 47 –