Basics of Thermal Field Theory (By Mikko Laine)

Basics of Thermal Field Theory1
Mikko Laine
February 6, 2008
Notes based on lectures delivered at the University of Bielefeld, Germany, during the summer semester
2004 and during the winter semester 2007-2008.
Contents
1
Quantum Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.1
Basic structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2
Canonical partition function . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.3
Path integral for the partition function . . . . . . . . . . . . . . . . . . . . .
1.4
Evaluation of the path integral for harmonic oscillator . . . . . . . . . . . .
1.5
Exercise 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Free scalar fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
12
2.1
12
2.2
Fourier representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
13
2.3
Evaluation of thermal sums . . . . . . . . . . . . . . . . . . . . . . . . . . .
15
2.4
Exercise 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
18
2.5
Low-temperature expansion . . . . . . . . . . . . . . . . . . . . . . . . . . .
19
2.6
High-temperature expansion . . . . . . . . . . . . . . . . . . . . . . . . . . .
22
2.7
Properties of the Euler gamma and Riemann zeta functions . . . . . . . . .
25
2.8
Exercise 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
27
Interacting scalar fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
28
3.1
Weak-coupling expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . .
29
3.2
Wick theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
30
3.3
Propagator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
32
3.4
Naive free energy density to O(): ultraviolet divergences . . . . . . . . . .
35
3.5
Naive free energy density to O(2 ): infrared divergences . . . . . . . . . . .
35
3.6
Exercise 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
36
3.7
Proper free energy density to O(): ultraviolet renormalization . . . . . . .
39
3.8
Proper free energy density to O( 2 ): infrared resummation . . . . . . . . .
42
3.9
4
Exercise 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
44
Fermions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
46
4.1
Path integral for the partition function of a fermionic harmonic oscillator .
46
4.2
The Dirac field at finite temperature . . . . . . . . . . . . . . . . . . . . . .
49
4.3
Fermionic thermal sums . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
52
4.4
Exercise 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
53
Gauge fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
55
5.1
56
5.2
Gauge fixing and ghosts . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
60
5.3
Feynman rules for Euclidean continuum QCD . . . . . . . . . . . . . . . . .
61
5.4
Exercise 7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
63
5.5
Thermal gluon mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
65
5.6
Free energy density to O(g 3 ) . . . . . . . . . . . . . . . . . . . . . . . . . .
71
5.7
Exercise 8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
74
Low-energy effective field theories . . . . . . . . . . . . . . . . . . . . . . . . . . . .
75
6.1
The infrared problem of thermal field theory . . . . . . . . . . . . . . . . .
75
6.2
A simple example of an effective field theory . . . . . . . . . . . . . . . . .
76
6.3
Dimensionally reduced effective field theory for hot QCD . . . . . . . . . .
80
6.4
Exercise 9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
83
Finite density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
85
7.1
Complex scalar field with a finite chemical potential . . . . . . . . . . . . .
86
7.2
Effective potential and Bose-Einstein condensation . . . . . . . . . . . . . .
88
7.3
Dirac fermion with a finite chemical potential . . . . . . . . . . . . . . . . .
90
7.4
How about chemical potentials for gauge symmetries? . . . . . . . . . . . .
91
7.5
Exercise 10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
92
Real-time observables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
94
8.1
Different Greens functions . . . . . . . . . . . . . . . . . . . . . . . . . . .
94
8.2
From Euclidean correlator to spectral function . . . . . . . . . . . . . . . . 104
8.3
Exercise 11 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
8.4
Hard Thermal Loop effective theory . . . . . . . . . . . . . . . . . . . . . . 110
8.5
Relation to classical kinetic theory . . . . . . . . . . . . . . . . . . . . . . . 118
8.6
Exercise 12 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
10
Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
9.1
Thermal phase transitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
9.2
Bubble nucleation rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
9.3
Exercise 13 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
9.4
Particle production rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
9.5
Dark matter abundance in cosmology . . . . . . . . . . . . . . . . . . . . . 145
9.6
Appendix: relativistic Boltzmann equation . . . . . . . . . . . . . . . . . . 151
9.7
Transport coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
Further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
1.
1.1.
Quantum Mechanics
Basic structure
The properties of the system can be described by a Hamiltonian, which for non-relativistic spin-0
particles in one dimension takes the form
2
= p + V (
H
x) ,
2m
(1.1)
where m is the particle mass. The dynamics is governed by the Schr

odinger equation,
i~
|i = H|i
.
t
(1.2)
Formally, the time evolution can be solved in terms of the time-evolution operator:
t0 )|(t0 )i ,
|(t)i = U(t;
(1.3)
where, for a time-independent Hamiltonian,
0)
(t; t0 ) = e ~i H(tt
.
U
(1.4)
For the state vectors |i, various bases can be chosen. In the |xi-basis,
hx|
x|x i = xhx|x i = x (x x ) ,
In the energy basis,
hx|
p|x i = i~x hx|x i = i~x (x x ) .
H|ni
= En |ni .
(1.5)
(1.6)
In the classical limit, the system of Eq. (1.1) can be described by the Lagrangian
LM =
1
mx 2 V (x) .
2
(1.7)
A Legendre transform leads to the classical Hamiltonian:

p=
LM
,
x
H = xp
LM =
p2
+ V (x) .
2m
(1.8)
A most important example of a quantum mechanical system is provided by a harmonic oscillator:

V (
x)
1
m 2 x
2 .
2
In this case the energy eigenstates can be found explicitly:

1
En = ~ n +
, n = 1, 2, 3, . . . .
2
(1.9)
(1.10)
All states are non-degenerate.

It will turn out to be useful to view (quantum) mechanics formally as (0+1)-dimensional (quantum) field theory: the operator x
can be viewed as the field operator at a certain point,
x
(0)
.
(1.11)
In quantum field theory operators are usually represented in the Heisenberg picture rather than in
the Schrodinger picture; then
xH (t) H (t, 0) .
(1.12)
In the following we use an implicit notation whereby showing the time coordinate t as an argument
implies automatically the Heisenberg picture, and the corresponding subscript is left out.
1
1.2.
Canonical partition function
Taking now our quantum mechanical system to a finite temperature T , the basic quantity to
compute is the partition function Z. We employ the canonical ensemble, whereby Z is a function
of T . Introducing units where kB = 1 (i.e., There kB TSI-units ), the partition function is defined
by
1
.
(1.13)
Z(T ) Tr [e H ] ,
T
From the partition function, other observables are obtained, for instance the free energy F , the
entropy S, and the average energy E:
F
T ln Z ,
F
1
H ] = F + E ,
= ln Z +
Tr [He
T
TZ
T
T
1
Tr [He
].
Z
(1.14)
(1.15)
(1.16)
Let us now compute these quantities for the harmonic oscillator. This can be trivially done in
the energy basis:
Z
hn|e H |ni =
n=0
e~( 2 +n) =
n=0
1
e~/2
.
=
~
~
1e
2 sinh 2T
(1.17)
Consequently,
F
=
T ~
T ~
=
T ~
T ~
=
T ~
T ~

~
~
~
~
2T
2T
=
T ln e e
+ T ln 1 e
2
~
2
T
,
T ln

~
~
1
~
+
ln 1 e
~
T e
1
~ ~
e T
T
T
,
1 + ln
~

1
1
F + T S = ~
+ ~
2
e
1
~
2
T .
(1.18)
(1.19)
(1.20)
(1.21)
(1.22)
(1.23)
(1.24)
(1.25)
(1.26)
Here we have also shown the behaviours of the various functions at low temperatures T ~ and
at high temperatures T ~. Note how in most cases one can identify the contribution of the
ground state, and of the thermal states, with their Bose-Einstein distribution function.
Note also that the average energy rises linearly with T at high temperatures, with a coefficient
counting the number of degrees of freedom (i.e. the degeneracy).
1.3.
Path integral for the partition function
In the case of the harmonic oscillator, energy eigenvalues are known, and Z can easily be evaluated.
In many other cases, however, En are difficult to compute. A more useful representation of Z is
obtained by writing it as a path integral.
In order to get started, let us recall some basic relations. First of all,
hx|
p|pi = phx|pi = i~x hx|pi hx|pi = Ae
ipx
~
(1.27)
where A is some constant. Second, we will need completeness relations, which we write as
Z
Z
dp
|pihp| = 1 ,
dx |xihx| = 1 ,
(1.28)
B
where B is another constant. The choices of A and B are not independent. Indeed,
Z
Z
Z
Z
Z
Z
i(p p)x
dp dp
dp dp
1 =
dx
|pihp|xihx|p ihp | = dx
|pi|A|2 e ~ hp |
B
B
B
B
Z
Z
2 Z
dp dp
dp
2~|A|
2~|A|2
1.
|pi|A|2 2~(p p)hp | =
|pihp| =
=
B
B
B
B
B
(1.29)
(1.30)
Thereby B = 2~|A|2 ; we choose A 1, so that B = 2~.

We then move to the evaluation of the partition function. We do this in the x-basis, whereby
Z
Z
H
H
Z = Tr [e H ] = dx hx|e H |xi = dx hx|e ~ e ~ |xi .

(1.31)
Here we have split e H into a product of N 1 different pieces, and defined ~/N .
The trick now is to insert
1=
dpi
|pi ihpi | ,
2~
i = 1, . . . , N
(1.32)
on the left side of each exponential, with i increasing from right to left; and
Z
1 = dxi |xi ihxi | , i = 1, . . . , N
(1.33)
on the right side of each exponential, with again i increasing from right to left.
Thereby we are left to consider matrix elements of the type
x)
hxi+1 |pi ihpi |e ~ H(p,
|xi i =
ipi xi+1
~
hpi |e ~ H(pi ,xi )+O( ) |xi i

xi+1 xi
p2i
ipi
+ V (xi ) + O()
.
exp
~ 2m
(1.34)
Moreover we need to note that on the very right, we have

hx1 |xi = (x1 x) ,
(1.35)
which allows to carry out the integral over x; and that on the very left, the role of hxi+1 | is played
by the state hx| = hx1 |. Finally, we remark that the O()-correction in Eq. (1.34) can be eliminated
by sending N .
In total, then, we can write the partition function as
Z = lim

Z Y
N
dxi dpi
i=1
2~

N

1 X p2j
xj+1 xj
exp
ipj
+ V (xj )
~ j=1 2m
.
xN +1 x1 ,~/N
(1.36)
Oftentimes this is symbolically written as a continuum path integral, as

Z
Z
1 ~
DxDp
[p( )]2
exp
ip( )x(
) + V (x( ))
.
d
Z=
~ 0
2m
x(~)=x(0) 2~
(1.37)
Note that in this form, the integration measure is well-defined (as a limit of that in Eq. (1.36)).
The integral over the momenta pi is gaussian, and can be carried out explicitly:

r

Z
xi+1 xi
p2i
m(xi+1 xi )2
dpi
m
=
.
exp
ipi
exp
~ 2m
2~
2~
2~
(1.38)
Thereby Eq. (1.36) becomes

Z = lim

2

N

1 X m xj+1 xj
p
+ V (xj )
exp
~ j=1 2
2~/m

i=1
Z Y
N
dxi
.
xN +1 x1 ,~/N
(1.39)
We may also try to write this in a continuum form, like in Eq. (1.37). Of course, the measure
contains then a factor which appears quite divergent at large N ,
N/2

m
N
mN
.
(1.40)
C
= exp
ln
2~
2
2~2
This factor is, however, completely independent of the properties of the potential V (xj ). Thereby
it contains no dynamical information, and we actually do not need to worry too much about the
apparent divergence in any case, we will return to C from another angle in the next section.
For the moment, then, we can again write a continuum form for the functional integral,
2

Z
Z
m dx( )
1 ~
+ V (x( ))
.
(1.41)
d
Z=C
Dx exp
~ 0
2
d
x(~)=x(0)
Let us end by giving an interpretation to the result in Eq. (1.41). We recall that the usual
path integral at zero temperature has the exponential
Z
2

i
m dx
exp
V (x) .
(1.42)
dt LM , LM =
~
2 dt
We note that Eq. (1.41) could be obtained with the following recipe:
(i) Carry out a Wick rotation, denoting it.
(ii) Introduce
LE LM ( = it) =
2
m dx
+ V (x) .
2 d
(1.43)
(iii) Restrict to the interval 0...~.

(iv) Require periodicity over .
With these steps, the exponential becomes

Z
1
1 ~
exp SE exp
d LE .
~
~ 0
(1.44)
where the subscript E stands for Euclidean, in contrast to Minkowskian. It will turn out that
this recipe works, almost without modifications, also in field theory, and even for spin-1/2 and
spin-1 particles, although the derivation of the path integral itself looks quite different in those
cases.
4
1.4.
Evaluation of the path integral for harmonic oscillator
As a crosscheck, we would now like to evaluate the path integral in Eq. (1.41) for the case
of a harmonic oscillator, and check that we get the correct result in Eq. (1.17). To make the
exercise more interesting, we carry out the evaluation in Fourier space (with respect to the time
coordinate ) rather than in configuration space. Moreover, we would like to see if we can deduce
the information contained in the divergent constant C without making use of its actual value in
Eq. (1.40).
Let us start by representing an arbitrary function x( ), 0 < < ~, with the property x(~) =
x(0), as a Fourier sum:
X
x( ) T
xn ein ,
(1.45)
n=
where the factor T is a convention. Periodicity requires

ein ~ = 1 ,
i.e. n ~ = 2n ,
nZ.
(1.46)
The values n = 2T n/~ are called Matsubara frequencies.

Apart from periodicity, we also impose reality on x( ):
x( ) R x ( ) = x( ) xn = xn .
(1.47)
If we write xn = an + ibn , it follows that

xn
= an ibn = xn = an + ibn
an = an
bn = bn
In particular, b0 = 0, and xn xn = a2n + b2n . Thereby, we now have the representation

X
in
in
.
(an + ibn )e
+ (an ibn )e
x( ) = T a0 +
(1.48)
(1.49)
n=1
Here, a0 is called (the amplitude of) the Matsubara zero-mode.

With the representation of Eq. (1.49), general quadratic structures in configuration space can be
written as
Z
Z
X
1 ~
1 ~
2
d x( )y( ) = T
d ei(n +m )
xn ym
~ 0
~
0
m,n
X
X
1
2
xn yn .
(1.50)
= T
xn ym n,m = T
T
n
m,n
In particular, the argument of the exponential in Eq. (1.41) becomes

Z
1 ~ m dx( ) dx( )
+ 2 x( )x( )
d
~ 0
2
d
d
i
h
Eq. (1.50)
mT X
=
xn in in + 2 xn
2 n=
n =n
mT X
( 2 + 2 )(a2n + b2n )
2 n= n
X
mT 2 2
a0 mT
(n2 + 2 )(a2n + b2n ) .
2
n=1
(1.51)
Next, we need to consider the integration measure. Let us make a change of variables from x( ),
(0, ~), to the Fourier components an , bn . As we have seen, the independent variables are then
a0 and {an , bn }, n 1.
This change of variables introduces a determinant,

i
hY

x( )
.
da
db
da
Dx( ) = det
n
n
0
xn
(1.52)
n1
The change of bases is purely kinematical, however, and independent of the potential V (x). Thus
we can define

x( )
C C det
,
(1.53)
xn
and consider now C as an unknown coefficient.

p
R
Making use of the gaussian integral dx exp(cx2 ) = /c, the expression in Eq. (1.41) now
becomes

Z
Z hY
i
X
1
2 2
2
2
2
2
Z = C
da0
dan dbn exp mT a0 mT
(n + )(an + bn )
2
n1
n1
r
2T n
2 Y
, n =
.
(1.54)
= C
2
2
2
mT n=1 mT (n + )
~
It remains to determine C . How to do this?
Since C is independent of , we can determine it in the limit = 0, whereby the system
simplifies.
The integral over the zero-mode a0 in Eq. (1.54) is, however, divergent for 0. We call
such a divergence an infrared divergence: the zero-mode is the lowest-energy mode.
But we can still take 0 if we momentarily regulate the integration over the zero-mode
in some other way. We note from Eq. (1.49) that
1
~
d x( ) = T a0 ,
(1.55)
so that T a0 represents the average value of x( ). In terms of Eq. (1.31), we can identify the
average value with the boundary condition x, over which we integrate.
Let us then simply regulate the system by putting it in a box, i.e. by restricting the values
of x to some (asymptotially wide by finite) interval x, and those of a0 to the interval x/T .
With this setup, we can proceed to match for C .
Side A: effective theory computation. In the presence of the regulator, Eq. (1.54) becomes

Z
Z hY
i
X
lim Zregulated = C
da0
n2 (a2n + b2n )
dan dbn exp mT
0
x/T
n1
x Y
,
= C
T n=1 mT n2
n =
n1
2T n
.
~
(1.56)
Side B: full theory computation. In the presence of the regulator, and in the absence of
V (x), Eq. (1.31) can be computed in a very simple way:
Z
p
2
lim Zregulated =
dx hx|e 2mT |xi
0
Z
Zx
p
2
dp
dx
=
hx|e 2mT |pihp|xi
2~
x
Z
Z
dp p2
e 2mT hx|pihp|xi
dx
=
x
2~
1
= x
2mT .
(1.57)
2~
Matching the two sides. Equating Eqs. (1.56) and (1.57), the regulator x drops out, and we
find
Y
mT n2
T
2mT
C =
.
(1.58)
2~
n=1
Since the infrared regulator has dropped out, we may called C an ultraviolet coefficient.
Now we can continue with the full Eq. (1.54). Inserting C from Eq. (1.58), we get
Z
T Y
1
~ n=1 1 + 22
(1.59)
=
Making use of
1
T
i .
h
~ Q 1 + (~/2T )2
2
n=1
n

Y
sinh x
x2
1+ 2 ,
=
x
n
n=1
(1.60)
(1.61)
then yields directly Eq. (1.17): the result is correct!
Thus, we have indeed managed to reproduce the correct result from the path integral, without
ever making recourse to Eq. (1.40) or, for that matter, to the discretization that was present in
Eqs. (1.36), (1.39).
Let us end with a couple of final remarks. First of all, in Quantum Mechanics, the partition function Z, and all other observables, are certainly finite and well-defined functions of the parameters
T, m, and , if computed properly. We saw that with path integrals this is not always obvious at
every intermediate step, but at the end does work out. In Quantum Field Theory, on the contrary,
divergences may remain, even if we compute everything correctly. These are then taken care of
by renormalization. It is important to realise, however, that the ambiguity of the functional
integration measure (through C ) is not in itself the source of these divergences, as our quantum
mechanical example has demonstrated!
As the second remark, it is appropriate to stress that in many physically relevant observables,
the coefficient C drops out completely, and the procedure is thereby simpler. An example of such
a quantity is discussed in Exercise 1.
As a final remark, it should be noted that many of the concepts and techniques that were
introduced with this simple example zero-modes, infrared divergences, their regulation, matching
computations, etc will also play a role in much less trivial quantum field theoretic examples
later on, so it is important to master them as early as possible.
1.5.
Exercise 1
Defining
x( ) e
H
~
x
e
H
~
0 < < ~ ,
(1.62)
we define a 2-point Greens function as

G( )
The corresponding path integral reads
R
G( ) =
i
1 h H
Tr e
x
( )
x(0) .
Z
x(~)=x(0) Dx x( )x(0) exp[SE /~]
x(~)=x(0) Dx
exp[SE /~]
(1.63)
(1.64)
whereby the coefficient C has dropped out. The task is to compute explicitly G( ) for the harmonic
oscillator, by making use of
x
(a) the canonical formalism [expressing H,
in terms of a
, a
].
(b) the path integral formalism, in Fourier space.
Solution to Exercise 1
(a) In terms of a
, a
, we can write

1
= ~ a
H
a
+
,
2
x
=
~
(
a+a
) ,
2m
[
a, a
] = 1 .
(1.65)
In order to construct x
( ), we make use of the expansion
A
[A,
B]]
+ 1 [A,
[A,
[A,
B]]]
+ ... .
+ [A,
B]
+ 1 [A,
eA Be
=B
2!
3!
(1.66)
In particular,
a
[H,
] = ~[
a a
, a
] = ~
a,
2
[H, [H, a
]] = (~) a
,
a
[H,
] = ~[
a a
, a
] = ~
a ,
[H,
a
[H,
]] = (~)2 a
,
(1.67)
and so forth, so that we can write

r

H
H
1
1
~
a
1 + ( )2 + ... + a
1 + + ( )2 + ...
e ~ xe ~ =
2m
2!
2!
r

~
(1.68)
a
e + a
e .
=
2m
Inserting Z from Eq. (1.17), we then get

~ X

1
~
+a
|ni .
a
e
+a
e a
hn|e~(n+ 2 )
G( ) = 2 sinh
2
2m
n=0
We now use a
|ni =
(1.69)
n + 1|n + 1i and a
|ni = n|n 1i to identify the non-zero matrix elements,
hn|
aa
|ni = n + 1 ,
hn|
a a
|ni = n .
(1.70)
Thereby
G( ) =
i
~ X
h

~
~
exp
sinh
e~n e + n e + e .
m
2
2
n=0
(1.71)
The sums are simple,
e~n
1
,
1 e~
n=0
n=0
ne~n
1
1 d
e~
=
.
~
~ d 1 e
(1 e~ )2
(1.72)
In total, then,
G( )
=
=
=

1
e~
~
e
1e
+ e
+e
2m
1 e~
(1 e~ )2
i
h
~
1
e + e( ~)
2m 1 e~
h
i
~
cosh
2
~
i
h
.
2m
sinh ~
2
(1.73)
(b) Integration measure:

Z
da0
hY
n1
i
dan dbn .
(1.74)
Exponential:

X
X
1
1
2 2
2
2
2
2
2
2
2
exp mT a0 mT
m(n + )|xn | .
(n + )(an + bn ) = exp T
2
2 n=
(1.75)
n1
Fourier representation:
x( )
x(0)

X
ik
ik
,
= T a0 +
(ak + ibk )e
+ (ak ibk )e

= T a0 +
k=1
2al
l=1
Observable:
G( ) = hx( )x(0)i
(1.76)
(1.77)
RQ
da0
n1 dan dbn x( ) x(0) exp[ ]
RQ
R
.
da0
n1 dan dbn exp[ ]
(1.78)
Since the exponential is quadratic in a0 , an , bn R, we have

ha0 ak i = ha0 bk i = hak bl i = 0 ,
hak al i = hbk bl i kl .
(1.79)
Thereby
G( ) = T
a20
Here
ha20 i
=
=
=
ha2k i =
=
2a2k
k=1
ik
+e
ik
+

da0 a20 exp 12 mT 2 a20

R
da0 exp 21 mT 2 a20

Z

r

2 d
1
2 d
2
2 2
=
ln da0 exp mT a0
ln
m 2 dT
2
m 2 dT
m 2 T
1
,
m 2 T
R

dak a2k exp mT (k2 + 2 )a2k
R
dak exp [mT (k2 + 2 )a2k ]
1
.
2
2m(k + 2 )T
(1.80)
(1.81)
(1.82)
Inserting into Eq. (1.80), we get

G( ) =

X
T X eik
eik + eik
T 1
=
+
,
m 2
k2 + 2
m
k2 + 2
k=1
(1.83)
k=
where k = 2kT /~.

There are various ways to evaluate the sum in Eq. (1.83). We will encounter one generic method
in the later sections, so let us here present a different approach. We start by noting that

T X ik
~
d2
(1.84)
e
= ( mod ~) ,
2 + 2 G( ) =
d
m
m
k=
10
where we made use of a standard summation formula.1

We now first solve Eq. (1.84) for 0 < < ~:

d2
2 + 2 G( ) = 0
d
G( ) = Ae + Be ,
(1.85)
where A, B are unknown constants. The definition, Eq. (1.83), indicates that G(~ ) = G( ),
which allows to related A and B:
i
h
(1.86)
G( ) = A e + e(~ ) .
The remaining unknown A can be obtained by approaching the limit 0+ . Then, from (1.83),

G(0) = A 1 + e~
X
T X
T ~2
1
1
=
=

~ 2
2
m
k2 + 2
m (2T )2
k=
k= k + 2T

~
~ cosh 2T
,
=
(1.87)
~
2m sinh 2T
P
2
2
where we made use of
k= 1/(k + x ) = /x tanh(x). Solving for A, and inserting into
Eq. (1.86), yields the important final result:
h
i
~
cosh
2
~
h
i
G( ) =
.
(1.88)
2m
sinh ~
2
P
ikx = C(x mod 2), where C is some constant. In order to determine C, let us
Clearly
k= e
integrate both sides of this equation from 0 to 2 + 0 . On the left-hand side we get 2k,0 , on the right-hand
side
C = 2. Replacing now x by 2T /~, and using (ax) = (x)/|a| on the right-hand side, yields
P C. Thus,
ik = ~ ( mod ~).
k= e
1 Proof:
11
2.
2.1.
Free scalar fields

We start by deriving a path integral representation for scalar field theory, by making use of the
result obtained for the quantum mechanical harmonic oscillator (HO) in the previous section.
In quantum field theory, the form of the theory is most economically defined in terms of the
(for instance, Lorentz
corresponding classical Lagrangian LM , rather than the Hamiltonian H
symmetry is explicit only in LM ). Let us therefore start from Eq. (1.7), and re-interpret x as an
internal degree of freedom , situated at the origin 0 of d-dimensional space, like in Eq. (1.11):
Z
HO
SM
=
dt LHO
(2.1)
M ,
LHO
M

2
m (t, 0)
V ((t, 0)) .
2
t
(2.2)
Let us compare this with the usual action of scalar field theory (SFT) in d-dimensional space:
Z
Z
SFT
SM
=
dt dd x LSFT
,
(2.3)
M
LSFT
M
1
1
1
V () = (t )2 (i )(i ) V () ,
2
2
2
(2.4)
where we assume that repeated indices are summed over (irrespective of whether they are up and
down, or both at the same altitude), and the metric is (+).
Comparing Eq. (2.2) with Eq. (2.4) we see that formally, scalar field theory is nothing but a
collection (sum) of almost independent harmonic oscillators with m = 1, one at every x. These
oscillators only interact via the derivative term (i )(i ) which, in the language of statistical
physics, couples nearest neighbours:
i
(t, x + i) (t, x)
,
(2.5)
where i is a unit vector in the direction i.

We then realise, however, that such a coupling does not change the derivation of the path
integral (for the partition function) in Sec. 1.3 in any essential way: it was only important that
the Hamiltonian was quadratic in the canonical momenta, p = mx t . In other words, the
derivation of the path integral is only concerned with objects having to do with the time coordinate
(or time derivatives), and these appear identically in Eqs. (2.2) and (2.4). Therefore, we can directly
take over the result from Eqs. (1.41)(1.44):
Z SFT (T ) =
LSFT
E

Z
Z
Y
1 ~
[ C D(, x) ] exp
,
d dd x LSFT
E
~ 0
(~,x)=(0,x) x
2 X

2
d
1
1
+
+ V () .
LSFT
(t
i
)
=
M
2
2 xi
i=1
Z
(2.6)
(2.7)
We will drop out the superscript SFT in the following and also, for brevity, mostly write LE in
the form LE = 21 + V ().
12
2.2.
Fourier representation
We now parallel the strategy in Sec. 1.4, and rewrite the path integral in Fourier representation. In
order to simplify the notation somewhat, we measure time in units where ~ = 1.05 1034 Js = 1.
Then the dependence of can be expressed as
(, x) = T
n , x)ein ,
(
n = 2T n , n Z .
(2.8)
n=
For the space coordinates, it is useful to momentarily take each direction finite, of extent Li , and
impose periodic boundary conditions, just like for the time direction. Then the dependence on xi
can be represented as
f (xi ) =
1
Li
f(ni )eiki xi ,
ki =
ni =
2ni
,
Li
ni Z ,
(2.9)
where 1/Li plays the same role as T in the time direction. In the infinite volume limit, the sum in
Eq. (2.9) goes over to the usual Fourier integral,
Z
1 X
1 X
dki
Li
=
,
(2.10)
ki
Li n
2 n
2
i
so that the finite volume is really just an intermediate regulator. The whole function in Eq. (2.8)
now becomes
X 1 X
n , k)ein +ikx , V L1 L2 ...Ld .
(
(2.11)
(, x) = T
V
Like in Sec. 1.4, the reality of (, x) implies that the Fourier modes satisfy
h
i
n , k) = (
(
n , k) .
(2.12)
Thereby only half of the Fourier-modes are independent; we can choose, for instance,
n , k) ,
(
k) ,
(0,
n1;
k1 > 0 ;
0, k2 , ...) ,
(0,
k2 > 0 ; . . . ;
0) (2.13)
and (0,
as the integration variables. Note again the presence of a zero-mode.

Quadratic forms can be written as
Z
dd x 1 (, x)2 (, x) = T
In particular, in the free case, i.e. for V ()

exp(SE )
X 1 X
1 (n , k)2 (n , k) .
V
1
2
m2 2 , the exponent in Eq. (2.6) becomes

Z Z
= exp
d dd x LE
0

1 X 1 X 2
n , k)|2
= exp T
(n + k2 + m2 )|(
2 V
k
n

X
Y
T
n , k)|2
.
(n2 + k2 + m2 )|(
=
exp
2V
k
(2.14)
13
(2.15)
The exponential here is precisely the same as the one in Eq. (1.75), with the replacements
m(HO)
1
,
V
2(HO) k2 + m2 ,
(HO)
|x2n |
n , k)|2 .
|(
(2.16)
The result thus factorises into a product of harmonic oscillator partition functions, for which we
know the answer already. In fact, rewriting Eqs. (1.54), (1.59), (1.18) for the case ~ = 1, the
harmonic oscillator partition function can be represented as

Z Y
mT X
Z HO = C
dxn exp
(2.17)
(n2 + 2 )|xn |2
2 n=
n0
= T
n=1
n2
+ n2
(2.18)
1
(n2 + 2 ) 2
n=
(n2 ) 2
(2.19)
n =

1
,
+ T ln 1 e
= exp
T 2
(2.20)
where n means that the zero-mode n = 0 is omitted. Note in particular that all dependence on
m(HO) has dropped out.
Combining Eq. (2.15) with Eqs. (2.17)(2.20), we obtain two useful representations for Z SFT .
First of all, denoting
p
Ek k2 + m2 ,
(2.21)
Eq. (2.19) yields

Z
SFT

F SFT
=
= exp
T

Y
Y Y
2 12
2
2 21
(n )
T
(n + Ek )
n
(2.22)
X

1 X
1 X
exp
ln(n2 + Ek2 )
. (2.23)
ln n2
ln T +
2
2 n
Taking then the infinite-volume limit, the free-energy density, F/V , can be written as

Z d X
X1
1
d k
1
F SFT
2
2
2
2
T
=
ln(
+
E
)
T
ln(
)
T
ln(T
)
.
lim
n
k
n
V
V
(2)d
2
2
2
(2.24)
Second, making directly use of Eq. (2.20), we get

Y
F SFT
1 Ek
Z SFT = exp
+ T ln 1 eEk
=
,
exp
T
T 2
k
Z d

d k Ek
F SFT
Ek
.
=
+
T
ln
1
e
lim
V
V
(2)d 2
We will return to the evaluation of the momentum integration in Secs. 2.5 and 2.6.
14
(2.25)
(2.26)
2.3.
Evaluation of thermal sums
Due to the previously established equality between Eqs. (2.19), (2.20), we have arrived at two
different representations for the free energy density of free scalar field theory, Eqs. (2.24), (2.26).
The purpose of this section is to take the step from Eq. (2.24) to (2.26) directly, and learn on the
way how to carry out thermal sums such as those in Eq. (2.24) also in more general cases.
In fact, the very sum in Eq. (2.24) is a somewhat special case: it includes a physical term,
the first one, which depends on the energy (and thus on the mass of the scalar particle); as well
as unphysical subtractions, the second and third terms, which are independent of the energy,
but are needed in order to make the sum convergent. Only the result from energy-dependent term
survives in Eq. (2.26). In order not to lose focus on the peculiarities of this special case, we will
mostly concentrate on another, convergent sum:
X
1 dj(E)
1
,
=T
2 + E2
E dE
i(E)
(2.27)
from which the first term appearing in Eq. (2.24),

j(E) T
X1
n
ln(n2 + E 2 ) ,
(2.28)
can be obtained by integration with respect to E, apart from an integration constant, which should
be chosen according to the 2nd and 3rd terms in Eq. (2.24).
To begin with, we want to be completely general. Let f (p) be a function which is analytic in the
complex plane, and regular on the real axis. We then consider the sum
X
f (n ) .
(2.29)
ST
n
Consider now the auxiliary function

i nB (ip)
i
.
exp(ip) 1
(2.30)
This function has poles at ip = 2n, n Z, i.e. p = n . Expanding in Laurent series around any
pole we get
i nB (i[n + z]) =
i
exp(i[n + z]) 1
i
T
+ O(1) .
exp(iz) 1
z
(2.31)
Therefore, the residue at any pole is T . This means that we can replace the sum in Eq. (2.29) by
a complex integral:
S
dp
f (p) inB (ip)
2i
+i0+
i0+
dp
f (p) nB (ip) +
2
+i0+
++i0+
dp
f (p) nB (ip) , (2.32)
2
where, as indicated, the integration contour runs clockwise around the real axis of the complex
p-plane.
In the latter term in Eq. (2.32), we can furthermore substitute integration variables as p p,
and note that
nB (ip) =
1
exp(ip) 1 + 1
=
= 1 nB (ip) .
exp(ip) 1
1 exp(ip)
15
(2.33)
We thereby get the final formula

S
=
=
+i0+
o
dp n
f (p) + [f (p) + f (p)]nB (ip)
i0+ 2
Z +
Z +i0+
dp
dp
f (p) +
[f (p) + f (p)]nB (ip) ,
2
2
+
i0
(2.34)
where in the first term we were able to return to the real axis (because there are no singularities
there), and substitute once again p p. Thus we have managed to convert the sum in Eq. (2.29)
to a complex integral.
The first term in Eq. (2.34) is temperature-independent: it gives the zero-temperature vacuum
contribution. The latter term determines how thermal effects change the result.
As a technical point let us note, furthermore, that in the lower half plane,

yT y yx |p|
1
p=xiy

e
|nB (ip)| = ix y
e
.
e e 1
(2.35)
Therefore, if the function f (p) grows slower than e|p| at large |p| (in particular, polynomially), the
integration contour can be closed in the lower half plane, and the result is determined by the pole
locations and residues of the function f (p) + f (p). Physically, we therefore say that the thermal
contribution to S is related to on-shell particles.
Let us now apply the general formula in Eq. (2.34) to the particular example of Eq. (2.27). In
fact, without any additional cost, we can consider a slight generalization,
i(E; c) T
X
n
1
,
(n + c)2 + E 2
cC.
(2.36)
In terms of Eq. (2.29), we then have

f (p) =
f (p) + f (p) =

1
i
1
1
=
(p + c)2 + E 2
2E p + c + iE
p + c iE

i
1
1
1
1
.
+
2E p + c + iE
p c + iE
p + c iE
p c iE
(2.37)
(2.38)
For Eq. (2.34), we need the poles in the lower half plane; for | Im c| < E, these are at p = c iE.
According to Eq. (2.38), the residue at each pole is i/2E. The vacuum term in Eq. (2.34) then
produces
i
1
1
(2i)
=
,
(2.39)
2
2E
2E
while the matter part yields

1
i
1
1
(2i)
+
.
(2.40)
2
2E e(Eic) 1 e(E+ic) 1
In total, then,
i
1 h
1 + nB (E ic) + nB (E + ic) .
(2.41)
2E
We note, first of all, that the result is periodic in c c + 2T n, n Z, as it must be according to
Eq. (2.36); and that the appearance of ic resembles that of a chemical potential. Indeed, as we will
see in Exercise 2, setting ic corresponds to a situation where we have averaged over a particle
(chemical potential ) and an antiparticle (chemical potential .)
i(E; c) =
16
To conclude this section, let us integrate the result of Eq. (2.41) with respect to E, in order to
obtain a generalization of the function in Eq. (2.28),
j(E; c) T
X1
n
ln[(n + c)2 + E 2 ] .
(2.42)
The relation given in Eq. (2.27) continues to hold in the presence of c. Noting that

d
ex
1
x
,
1
e
=
=
ex 1
1 ex
dx
(2.43)
h
i
i h
T
E
(E+ic)
(Eic)
,
+ 1e
ln 1 e
j(E; c) = const. + +
2
2
(2.44)
we immediately get
where the constant can depend on T and c.

Setting now c = 0, and comparing with Eqs. (2.24), (2.26), we notice that we have been lucky:
the extra terms in Eq. (2.24) are such that they precisely cancel against the integration constant
in Eq. (2.44). So, the full physical result containing j(E; 0) can be deduced directly from i(E; 0).
That the same is true even for ic 6= 0, if we interpret j(E; c) as a free energy density averaged
over a particle and an antiparticle (in a harmonic oscillator potential, i.e. a free field), is shown in
Exercise 2.
17
2.4.
Exercise 2
Compute the partition function of the harmonic oscillator in the presence of a chemical potential,
i
h
(2.45)
eF (T,) Z(T, ) Tr e(HN ) ,
and show that the expression
i
1h
F (T, ic) + F (T, ic)
2
agrees with the E-dependent part of Eq. (2.44).
(2.46)
Trivially,

N
)|ni = ~ n + 1 n = (~ )n + ~ .
hn|(H
2
2
Evaluating the partition function in the energy basis yields

~

exp
X
2T
~
~
.

Z HO =
exp
n =
2T
T
1 exp ~
n=0
(2.47)
(2.48)
Putting ~ 1, E, ic, we can rewrite the result as

Eic
1 E
Z HO = exp
.
(2.49)
+ T ln 1 e T
T 2
h
i
Reading from here F (T, ) according to Eq. (2.45), and computing 12 F (T, ic) + F (T, ic) , yields
directly the E-dependent part of Eq. (2.44).
18
2.5.
Low-temperature expansion
Our next goal is to carry out the momentum integration in Eqs. (2.24), (2.26). We denote
Z d

d k Ek
Ek
(2.50)
+ T ln 1 e
J(m, T )
(2)d 2

X Z dd k 1
2
2
= T
ln(
+
E
)
const.
,
(2.51)
n
k
(2)d 2
I(m, T )
=
=
1 d
J(m, T )
m dm
Z d
i
d k 1 h
1 + 2nB (Ek )
d
(2) 2Ek
X Z dd k
1
T
,
d 2 + E 2
(2)
n
k
(2.52)
(2.53)
(2.54)
where d is the space dimension, Ek = k2 + m2 , and we made use of the fact that m1 d/dm =
Ek1 d/dEk . In order to simplify the notation in the following, we will denote
Z
P
X Z dd k
,
(2)d
Z
P
X Z dd k
,
(2)d
dd k
,
(2)d
(2.55)
(n , k), and a prime denotes that the zero-mode is omitted. The tilde in K
is a
where K
reminder for Euclidean metric.
At low temperatures, T m, we expect that the results resemble those of the zero-temperture
theory. Therefore we can write
J(m, T ) = J0 (m) + JT (m) , I(m, T ) = I0 (m) + IT (m) ,
where J0 is the temperature-independent vacuum energy density,
Z
Ek
J0 (m)
,
k 2
and JT is the thermal part of the free energy density,
Z

JT (m)
T ln 1 eEk .
(2.56)
(2.57)
(2.58)
The sum-integral I(m, T ) can be divided in a similar way. It is clear that J0 (m) is ultraviolet
divergent, and can only be evaluated in the presence of a regularization; as indicated by Eq. (2.55),
we will mostly be employing dimensional regularization here. In contrast, the integrand in JT is
exponentially small for |k| T , and therefore the integral is well convergent.
Let us start by evaluating J0 (m). Writing the mass dependence explicitly, the task is to evaluate
J0 (m) =
1
dd k 1 2
(k + m2 ) 2 ,
(2)d 2
(2.59)
and subsequently insert d = 32. For generality and future reference, we will in fact first compute
Z d
1
d k
,
(2.60)
F (m, d, A)
(2)d (k2 + m2 )A
and obtain then J0 as J0 (m) =
1
2
F (m, d, 21 ).
19
Since the integrand only depends on |k|, angular integrations can be carried out, and the integration measure obtains the known form2
dd k =
d2
d/2
(k2 ) 2 d(k2 ) ,
(d/2)
(2.61)
where (s) is the Euler gamma function (we reiterate some of its basic properties in Sec. 2.7).
Substituting k2 z m2 t in Eq. (2.60), we get
Z
d2
d/2
1
F (m, d, A) =
dz z 2 (z + m2 )A
d
(d/2) (2) 0
Z
md2A
=
dt td/21 (1 + t)A .
(2.62)
(4)d/2 (d/2) 0
A further substitution t 1/s 1, dt ds/s2 yields
F (m, d, A) =
md2A
(4)d/2 (d/2)
ds sAd/21 (1 s)d/21 .
(2.63)
We now recognize a standard integral that can be expressed in terms of the gamma function:
F (m, d, A) =
(A d/2)
dd k
1
1
1
.
=
(2)d (k2 + m2 )A
(A)
(4)d/2
(m2 )Ad/2
(2.64)
We now return to J0 (m) in Eq. (2.59), i.e. set A = 21 , d = 3 2, A d2 = 2 + in Eq. (2.64),

and multiply by 12 . The basic property (s) = s1 (s + 1) allows to transport the argument of the
gamma function to the vicinity of 1/2 or 1, where a Taylor expansion is easily carried out (basic
formulae can be found in Sec. 2.7):
(2 + ) =
=
1
( ) =
2
1
(1 + )
(2 + )(1 + )

1
1+
1 + (1 E ) + O() ,
2
2
1
2( ) = 2 .
2
(2.65)
(2.66)
(2.67)
The other parts of Eq. (2.64) are written as
i
3
2 h
(4) 2 + =
1
+
ln(4)
+ O(2 ) ,
(4)2
2

2
4 2
(m2 )2 = m4 2
=
m
+ O(2 ) ,
1
+
ln
m2
m2
(2.68)
(2.69)
where is an arbitrary scale parameter, introduced as 1 = 2 2 .

Collecting now everything together, we get

3
2
m4 2 1
+
ln(4)
+
+
ln
+
O()
.
J0 (m) =
E
64 2
m2
2
(2.70)
It is convenient at this point to introduce the MS scheme scale parameter

, through
ln
2 ln 2 + ln(4) E .
2 On
c(d)
R
0
dk k d1
(2.71)
dd k exp(tk 2 ) = [ dk1 exp(tk12 )]d = (/t)d/2 . On the other, dd k exp(tk 2 ) =

R
2
exp(tk 2 ) = c(d)td/2 0 dx xd1 ex = c(d)(d/2)/2td/2 . Thereby c(d) = 2 d/2 /(d/2).
one hand,
20
Thereby

3
2
m4 2 1
+ ln 2 + + O() .
J0 (m) =
64 2
m
2
(2.72)
For I0 (m), we obtain from Eqs. (2.52), (2.72) the expression

Z
m2 2 1
2
1
=
+
1
+
O()
.
(2.73)
+
ln
I0 (m) =
16 2
m2
k 2Ek
R
Incidentally, note that dk0 /(2) 1/(k02 + Ek2 ) = 1/(2Ek ), so that I0 (m) can also be written
as
Z d+1
1
d k
.
(2.74)
I0 (m) =
(2)d+1 k2 + m2
We then move to the finite-temperature parts, JT (m) and IT (m). As already mentioned, the corresponding integrals are finite. Therefore, we can normally set d = 3 to begin with.3 Substituting
|k| T x in Eq. (2.58), and taking the derivative in Eq. (2.52), we find
Z

T4
2
x2 +y 2
dx
x
ln
1
e
JT (m) =
,
(2.75)
2 2 0
ym/T

Z

dx x2
1
T2

p
IT (m) =
.
(2.76)

2 2 0
x2 + y 2 e x2 +y2 1
ym/T
Unfortunately, these integrals cannot be expressed in terms of elementary functions. In retrospective, though, this may even be understandable: as we will see, they contain so much important
physics that it would be unrealistic to find it in any simple well-behaved analytic function! At the
same time, Eqs. (2.75), (2.76) can of course be numerically evaluated without problems.
Even though Eq. (2.75) cannot be evaluated exactly, we can still find approximate expression
valid in various limit. In this section we are interested in low temperatures, i.e. y = m/T 1. Let
us thus evaluate the leading term of Eq. (2.75) in an expansion in exp(y) and 1/y. We can write
Z
Z
2 2

2
x2 +y 2
dx x2 e x +y + O(e2y )
=
dx x ln 1 e
0
0
Z
p
w x2 +y 2
=
dw w w2 y 2 ew + O(e2y )
y
Z
p
vwy
= ey
dv(v + y) 2vy + v 2 ev + O(e2y )
0

3 y Z
1
v
v 12 v
e + O(e2y )
1+
dv v 2 1 +
= 2y 2 e
y
2y
0
h
1

i
3 3
= 2( )y 2 ey 1 + O
+ O ey ,
(2.77)
2
y
where (3/2) = /2.

Inserting into Eq. (2.58), we thus obtain

3

m 2 m
T
4
m/T
T
JT (m) = T
e
1+O
.
+O e
2T
m
The derivative in Eq. (2.52) subsequently yields

3

T
T3 m 2 m
m/T
e T 1+O
.
+O e
IT (m) =
m 2T
m
(2.78)
(2.79)
3 In multiloop computations, J (m) or I (m) could get multiplied by a divergent term, 1/, in which case
T
T
contributions of O() would be needed as well. They could then be obtained by noting from Eq. (2.61) that
2 dd k/(2)d = {1 + [ln(
2 /4k2 ) + 2] + O(2 )}d3 k/(2)3 , for d = 3 2 and an integrand only depending on k2 .
21
Thereby we have arrived at the main conclusion of this section: at low temperatures, T m,
finite-temperature effects in a theory with a mass gap are exponentially suppressed by the Boltzmann factor, exp(m/T ). In fact the results agree with those in non-relativistic classical statistical
mechanics. Consequently, the results for the functions J(m, T ), I(m, T ) can be well approximated
by their zero-temperature limits, J0 (m), I0 (m), given in Eqs. (2.72), (2.73), respectively.
2.6.
High-temperature expansion
We now move to the opposite limit than in the previous section, i.e. T m or, in terms of
Eq. (2.75), y = m/T 1. It appears obvious that the procedure then should be a Taylor
expansion in y 2 , around y 2 = 0. The zeroth order term, for instance, yields
Z

T4
2 T 4
2
JT (0) = 2
,
(2.80)
dx x2 ln 1 e x =
2 0
90
which is nothing but the free-energy density (minus the pressure) of black-body radiation with one
massless degree of freedom. A term of order O(y 2 ) can also be computed exactly.
However, that is as far as it works: trying to go to the second order, O(y 4 ), one finds that
the integral determining the coefficient of y 4 is power-divergent for small x! In other words, the
function JT (m) is not analytic around the origin in the variable m2 .
Nevertheless, Eq. (2.75) can still be expanded in a generalized sense, as we will see. The result
will in fact read

E
8
3
m6 (3)
me
m4
2 T 4 m2 T 2 m3 T
m
+
ln
+O() , (2.81)
+O
+
JT (m) =
2
4
2
90
24
12 2(4)
4T
4
3(4) T
T4
where m (m2 )1/2 . It is the cubic term in Eq. (2.81) which first indicates that JT (m) is not
analytic in m2 around the origin (because there is a branch cut); this term plays a very important
role in certain physical contexts, as we will see later on.
Our goal now is to derive the expansion in Eq. (2.81). A classic derivation, starting directly
from the definition in Eq. (2.75), can be found in a paper by Dolan and Jackiw, Phys. Rev. D 9
(1974) 3320. It will be easier, and ultimately more useful, to carry out another type of derivation,
however:
we start from Eq.P
(2.51) rather than Eq. (2.50), and now carry out first the integration
R
,
and
only
then
the
sum
n .
k
Of course, Eq. (2.51) contains inconvenient additional constant terms, which appears problematic. Fortunately we already know that mass-independent correct value of J(0, T ): it is given in
Eq. (2.80). Therefore it is enough to study I(m, T ), in which case the starting point, Eq. (2.54),
is simple enough, and subsequently integrate for J(m, T ) as
Z m
J(m, T ) =
dm m I(m , T ) + J(0, T ) .
(2.82)
0
Proceeding now with I(m, T ) in Eq. (2.54), the essential insight is to split the sum into the
contribution of the zero-mode, n = 0, and that of the non-zero modes, n 6= 0. Using the
notation of Eq. (2.55), we thus write
Z
Z
Z
P
P
.
(2.83)
+T
=
Let us first compute the contribution of the last term, which will be denoted by I (n=0) .
22
To start with, let us return to the infrared divergences alluded to above. Trying naively a Taylor
expansion in m2 , we would get

Z d
Z
d k 1
m2
m4
1
=
T
+
+
.
.
.
.
(2.84)
I (n=0) = T
2
2
(2)d k2
k4
k6
k k +m
For d = 3 2, the first term is ultraviolet divergent, i.e. grows at large |k|; while the second
and subsequent terms are infrared divergent, i.e. explode at small |k| too fast to be integrable.
(Of course, in dimensional regularization, every term in Eq. (2.84) appears strictly speaking to be
zero. The total result is non-zero, however, as we will see: thus the Taylor expansion in Eq. (2.84)
is really not justified, whichever way one looks at it.)
We now compute the integral in Eq. (2.84) properly. The result can be directly read from
Eq. (2.64), by just setting d = 3 2, A = 1:
I (n=0) = T F (m, 32, 1) = T
( 21 + )
1
1
3/2
(1)
(4)
(m2 )1/2+
( 21 )=2
Tm
+O() . (2.85)
4
This result is quite remarkable: a linearly divergent integral over a manifestly positive function is
finite and negative in dimensional regularization! According to Eq. (2.52), the corresponding term
in J (n=0) reads
T m3
J (n=0) =
+ O() .
(2.86)
12
Given the importance of the result, and its somewhat counter-intuitive appearance, it is worthwhile to demonstrate that Eq. (2.85) is not an artifact of dimensional regularization. Indeed, let
us compute it with cutoff regularization, by restricting |k| to be smaller than an upper bound, :
I
(n=0)
=
=
=
Z
dk k2
4
T
(2)3 0 k2 + m2

Z

T
T
dk
2
m
=
m
arctan
2
2
2 2
2 2
m
0 k +m

2
m
m
.
T
+O
2 2
4
(2.87)
We now observe that, due to the first term, Eq. (2.87) is indeed positive. This term is unphysical,
however: it must be cancelled by similar terms emerging from the non-zero modes, since the
temperature-dependent part of Eq. (2.53) is manifestly finite. Representing a power divergence, it
does not appear in dimensional regularization at all. The second term in Eq. (2.87) is the physical
one; it indeed agrees with Eq. (2.85). The remaining terms in Eq. (2.87) vanish when we take the
cutoff to infinity, and are the analogy of the O()-terms of Eq. (2.85).
We next turn to the contribution of the non-zero Matsubara modes, which will be denoted by
I (m, T ). It is important to realise that in this case, a Taylor expansion in m2 can be carried out:
the integrals will be of the type
Z
(m2 )n
, n 6= 0 ,
(2.88)
2
2 n+1
k (n + k )
and thus the integrand remains finite for small |k|, i.e., there are no infrared divergences. (There
could be ultraviolet divergences at small n, but these are taken care of by the regularization.)
More explicitly,
I (m, T )
X Z dd k
1
d
2
(2) n + k2 + m2
23
Taylor
=
Z
X
m2l
dd k X
l
(1)
2T
(2)d
[(2nT )2 + k2 ]l+1
n=1
l=0
X
X
Eq. (2.64)
(1)l m2l
=
2T
(l + 1 d2 )
1
1
d/2
(l + 1) (2nT )2l+2d
(4)
n=1 l=0
l

X
(l + 1 d2 )
1
m2
2T
(2l + 2 d) , (2.89)
(2T )2
(l + 1)
(4)d/2 (2T )2d
l=0
where in the last step we have interchanged

Pthe orders of the two summations, and identified the nsum as a Riemann zeta function, (s) n=1 ns . Some of the properties of (s) are summarised
in Sec. 2.7.
Let us work out the orders l = 0, 1, 2 explicitly. For d = 32, the order l = 0 requires evaluating
( 21 + ) and (1 + 2); l = 1 requires evaluating ( 21 + ) and (1 + 2); and l = 2 requires
evaluating ( 32 + ) and (3 + 2). We give some more details in Sec. 2.7 and in Exercise 3. In any
case, a straightforward computation yields
6
E

m
e
T2
2m2 2 1
2m4 (3)
I (m, T ) =
+
O
+
ln
+
+ O() .
(2.90)
12
(4)2
2
4T
(4)4 T 2
T4
Adding the zero-mode contribution, Eq. (2.85), we get

6
E
T 2 mT
2m4 (3)
2m2 2 1
m
e
I(m, T ) =
+
+ O() .
+
O
+
ln
12
4
(4)2
2
4T
(4)4 T 2
T4
Subtracting Eq. (2.73), finally yields

E
6
1
2m4 (3)
me
mT
2m2
m
T2
+
ln
+ O() .
+
O
IT (m) =
12
4
(4)2
4T
2
(4)4 T 2
T4
(2.91)
(2.92)
Note how the divergences and

have cancelled from IT (m), as must be the case.
We can now transport these results to various versions of the function J, by making use of
Eqs. (2.80) and (2.82). From Eq. (2.90), we get

8
E
m6 (3)
1
m2 T 2
m4
m
e
2 T 4
+
+ O() . (2.93)
+O
+
+ ln
J (m, T ) =
90
24
2(4)2 2
4T
3(4)4 T 2
T4
Adding the zero-mode contribution from Eq. (2.86) leads to

8
E
2 T 4
m6 (3)
1
m2 T 2 m3 T
m4
m
e
J(m, T ) =
+
+ O() .
+O
+
+ ln
90
24
12
2(4)2 2
4T
3(4)4 T 2
T4
(2.94)
Subtracting the zero-temperature part, J0 (m) in Eq. (2.72), leads finally to the expansion for
JT (m), given in Eq. (2.81). Note again the cancellation of 1/ and
from JT (m). The numerical
convergence of the high-temperature expansion is inspected in Exercise 3.
24
2.7.
Properties of the Euler gamma and Riemann zeta functions
(s)
The function (s) is to be viewed as a complex function, with a complex argument s. For
Re(s) > 0, it can be defined as
Z
(s) =
dx xs1 ex ,
(2.95)
0
while for Re(s) 0, the values can be obtained through iterative use of the relation
(s) =
(s + 1)
.
s
(2.96)
On the real axis, Im(s) = 0, (s) is regular at s = 1; as a consequence of Eq. (2.96), it then has
first order poles at s = 0, 1, 2, ... .
In practical applications, the argument s is typically either close to an integer, or close to a
half integer. In the former case, we can use Eq. (2.96) to relate the desired value to the values of
(s) and its derivatives around s = 1; these can then be worked out from the convergent integral
representation in Eq. (2.95). In particular,
(1) = E ,
(1) = 1 ,
(2.97)
where E is the Euler constant, E = 0.577215664901... . In the latter case, we can use Eq. (2.96)
to relate the desired value to the value of (s) and its derivatives around s = 21 ; these can then
be worked out from the convergent integral representation in Eq. (2.95). In particular,
1
2
1
2
(E 2 ln 2) .
(2.98)
The values required for Eq. (2.90) thus become

1
=
+
2

1
+
=
2
3

+
=
2
2 + O() ,
i
h
1 (E + 2 ln 2) + O(2 ) ,
+ O() .
2
(2.99)
(2.100)
(2.101)
We have gone one order deeper in the middle one, because the result turns out to be multiplied
by 1/ (cf. Eq. (2.111)).
(s)
The function (s) is to be viewed as a complex function, with a complex argument s. For
Re(s) > 1, it can be defined as
(s) =
n=1
1
=
(s)
dx xs1
.
ex 1
(2.102)
The equivalence
of the two forms in Eq. (2.102) can be seen by writing 1/(ex 1) = ex /(1
P nx
x
, and using the definition of the gamma function in Eq. (2.95). The remarkable
e ) = n=1 e
properties of (s) follow from the fact that by writing

1
1
1
1
1
,
(2.103)
=
=
ex 1
2 ex/2 1 ex/2 + 1
(ex/2 1)(ex/2 + 1)
25
and then substituting integration variables, x 2x, we find an alternative integral representation
for (s),
Z
dx xs1
1
.
(2.104)
(s) =
(1 21s )(s) 0 ex + 1
The integral here is defined for Re(s) > 0. Moreover, even though it diverges at s 0, the function
(s) also diverges at the same point, and consequently (0) will be finite and regular around s = 0:
(0)
(0) =
1
2
n=1
1
ln(2) .
2
Finally, for Re(s) 0, an analytic continuation is obtained through

s
(s) = 2s s1 sin
(1 s)(1 s) .
2
(2.105)
(2.106)
(2.107)
On the real axis, Im(s) = 0, (s) has a pole only at s = 1. Its values at even arguments are
easy; in fact, at negative even integers, Eq. (2.107) implies that
(2n) = 0 ,
n = 1, 2, 3, ... ,
(2.108)
while at positive even integers the values can be related to the Bernoulli numbers,
(2) =
2
,
6
(4) =
4
, ... .
90
(2.109)
Negative odd integers can be related to the positive even ones through Eq. (2.107), which equation
also allows to determine the behaviour around the pole at s = 1. In contrast, odd positive integers
larger that unity, i.e. s = 3, 5, ..., yield new transcendental numbers.
The values required for Eq. (2.90) become
1
1
(2)(2) + O() = + O() ,
(2.110)
2 2
12

i
h
1
+ cos
(1 2)(2)
(1 + 2) = 21+2 2 sin
2
2
2

1
1
= 2(1 + 2 ln 2)(1 + 2 ln )(1)
(1 + 2E )
(1 2 ln 2) + O()
2
2
1
=
+ E + O() ,
(2.111)
2
(1 + 2) =
(3 + 2) = (3) + O() 1.2020569031... + O() ,
(2.112)
where in the first two cases we made use of Eq. (2.107), and in the second also of Eqs. (2.105),
(2.106).
26
2.8.
Exercise 3
(a) Complete the derivation leading to Eq. (2.90).

(b) Inspecting JT (m), Eq. (2.81), sketch the regimes where the low-temperature and hightemperature expansions are numerically accurate.
Solution
(a) For the term l = 0 in Eq. (2.89), we make use of the values in Eqs. (2.99), (2.110):

1
T2
1
2
+ O() =
+ O() .
I (m, T )|l=0 = 2T
(2T
)
3/2
1
12
12
(4)
(2.113)
For the term l = 1 in Eq. (2.89), we make use of the values in Eqs. (2.100), (2.111):

i1
m2 h
(4)
12
1 (E + 2 ln 2)
(1 + 2E ) + O()
(2T
)
I (m, T )|l=1 = 2T
2
3/2
(2T )
2
(4)

2
2
1=2 2 m
2 1
=
+
ln(4)
+
2[
ln(4)]
+ O()
+
ln
E
E
(4)2
T2

eE
m2 2 1
2
=
+ ln 2 + 2 ln
+ O() ,
(2.114)
(4)2
T
4
where in the last step we introduced the MS scheme scale parameter through Eq. (2.71). For the
term l = 2 in Eq. (2.89), we make use of the values in Eqs. (2.101), (2.112):
2m4 (3)
1
m4 21
+ O() .
(2.115)
(3) + O() =
I (m, T )|l=2 = 2T
(2T )
4
3/2
(2T )
2
(4)4 T 2
(4)
(b) We again denote y m/T , and inspect then the function
Z

1
JT (m)
2
x2 +y 2
.
=
J (y)
dx
x
ln
1
e
T4
2 2 0
(2.116)
Apart from evaluating this expression numerically, we also consider the low-temperature result
from Eq. (2.78),
23
y1
y
J (y)
ey + O() ,
(2.117)
2
as well as the high-temperature result from Eq. (2.81),

E
y1
3
y 6 (3)
ye
y2
y3
y4
2
+
ln
+ O() .
+
J (y) =
2
90 24 12 2(4)
4
4
3(4)4
(2.118)
The results of the comparison are shown in Fig. 1. We observe that if we keep terms up to y 6 in
the high-temperature expansion, its numerical convergence is reasonable for y <
3. On the other
hand, the low-temperature expansion converges reasonably well as soon as y >
6. In between, a
numerical evaluation is needed.
27
Figure 1: The exact result from Eq. (2.116); the low-temperature approximation from Eq. (2.117);
and five orders of the high-temperature approximation from Eq. (2.118).
3.
Interacting scalar fields
In order to move from a free to an interacting theory, we now choose

V ()
1 2 2 1 4
m +
2
4
(3.1)
in Eq. (2.4), where > 0 is a dimensionless coupling constant. Thereby the Minkowskian and
Euclidean Lagrangians become
LM
LE
1
1
1
m2 2 4 ,
2
2
4
1
1
1
+ m2 2 + 4 ,
2
2
4
(3.2)
(3.3)
where repeated indices are again summed over, irrespective of whether they are up and down, or
both at the same altitude; and the case with all indices down implies the use of Euclidean metric
(i.e. no minus signs), like in Eq. (2.7).
Now, in the presence of > 0, it is no longer possible to exactly determine the partition function
of the system. We therefore need to develop approximation methods, which could in principle be
either analytic or numerical. In the following we restrict our attention to the simplest analytic
procedure which, as we will see, already teaches us quite a lot about the nature of the system.
28
3.1.
Weak-coupling expansion
In the weak-coupling expansion the theory is solved by assuming that 1, and by expressing
the result for the physical observable in question as a (generalized) Taylor series in .
The physical observable we are interested in, is the partition function in Eq. (2.6). Defining the
free and interacting parts of the Euclidean action as

Z Z
1
1
d3 x + m2 2 ,
S0
d
(3.4)
2
2
V
0

Z Z
1 4
3
d x ,
SI
d
(3.5)
4
V
0
the partition function can be written as
Z
Z SFT (T ) = C D eS0 SI

Z
1 2 1 3
S0
= C D e
1 SI + SI SI + . . .
2
6

1 2
1 3
SFT
= Z(0) 1 hSI i0 + hSI i0 hSI i0 + . . . ,
2
6
where
SFT
=C
Z(0)
D eS0
(3.6)
(3.7)
is the free partition function that we determined in Sec. 2, and the expectation value h i0 is
defined as
R
C D [ ] exp(S0 )
R
.
(3.8)
h i0
C D exp(S0 )
The free energy density then reads
F SFT (T, V )
V
=
=
=
T
ln Z SFT
V

SFT
F(0)
T
1 2
1 3
ln 1 hSI i0 + hSI i0 hSI i0 + . . .
V
V
2
6

SFT
F(0)
2
1
2
T
SI 0 +
SI 0 SI 0
V
V
2

2
1
3
SI 0 3 SI 0 SI 0 + 2 SI 0 + . . . ,
(3.9)
(3.10)
where we have Taylor-expanded the logarithm, ln(1 x) = x x2 /2 x3 /3 + ... . The first term,
SFT
F(0)
/V , is given in Eq. (2.26), while the subsequent terms correspond to corrections of orders
O(), O(2 ), and O(3 ). As we will see, the combinations that appear within the square brackets
in Eq. (3.10) have a very specific significance: Eq. (3.10) is actually simpler than Eq. (3.9)!
Inserting Eq. (3.5) into the various terms in Eq. (3.10), we are lead to evaluate expectation values
of the type
h(x1 )(x2 ) . . . (xn )i0 .
(3.11)
These can be reduced to products of free two-point correlators, h(xk )(xl )i0 , through the Wick
theorem, as we now recall.
29
3.2.
Wick theorem
The Wick theorem states that free (Gaussian) expectation values of any number of integration
variables can be reduced to products of two-point correlators, according to
X
h(x1 )(x2 ) . . . (xn1 )(xn )i0 =
h(x1 )(x2 )i0 h(xn1 )(xn )i0 .
(3.12)
all combinations
Before applying this to the terms in Eq. (3.10), we briefly recall, for completeness, how the theorem
can be derived with (path) integration techniques.
Let us assume that we discretise the space-time such that the coordinates x only take a finite
number of values (provided that the volume is finite as well). Then we can collect the values
(x), x, into a single vector v. The free action can be written as S0 = 12 v T Av, where A is a
matrix; we assume that A1 exists and that A is symmetric, AT = A.
The trick allowing to evaluate various integrals weighted by exp(S0 ) is to introduce a source
vector b, and to take derivatives with respect to its components. Specifically, we introduce
Z
h
i
h 1
i
exp W (b)
dv exp vi Aij vj + bi vi
2
iZ
i
h
h 1
vi vi +A1
b
1
ij j
(3.13)
=
exp bi A1
b
dv exp vi Aij vj ,
ij j
2
2
where we made use of a substitution of integration variables. We then obtain
h
i
R
dv (vk vl ...vn ) exp 12 vi Aij vj
i
h
hvk vl ...vn i0 =
R
dv exp 12 vi Aij vj
n

o
d d
d
...
exp
W
(b)
dbk dbl
dbn
b=0

=
exp W (0)
io
h1
n d d
d
...
exp bi A1
=
ij bj
db db dbn
2
b=0
k l
h
i
d d
1
d
1 1 2
1
1
=
1 + bi A1
b
A
b
b
A
b
+
.
.
.
...
b
+
.
i ij j r rs s
ij j
dbk dbl dbn
2
2 2
b=0
(3.14)
Taking the derivatives in Eq. (3.14), it is clear that:

h1i0 = 1.
If there is an odd number of components of v in the expectation value, the result is zero.
hvk vl i0 = A1
kl .
1 1
1 1
1
hvk vl vm vn i0 = A1
kl Amn + Akm Aln + Akn Alm
= hvk vl i0 hvm vn i0 + hvk vm i0 hvl vn i0 + hvk vn i0 hvl vm i0 .
And that, in general, we are lead to a discretized version of Eq. (3.12).

Since all of the operations carried out are purely combinatorial in nature, it is clear that removing
the discretization does not modify the result, and that thereby Eq. (3.12) indeed also holds in the
continuum limit.
30
Let us now apply Eq. (3.12) to Eq. (3.10). We will denote

F (T, V )
,
V
V
f (T ) lim
(3.15)
where we have dropped out the superscript SFT for simplicity. From Eqs. (2.26), (2.51), (3.10),
the leading order result is the familiar one,
f(0) (T ) = J(m, T ) .
(3.16)
At the first order, we get

T
T
f(1) (T ) = lim
hSI i0 = lim
V V
V V
d3 x h(x)(x)(x)(x)i0 .
4
V
(3.17)
Here we can use the Wick theorem, Eq. (3.12). Noting furthermore that h(x)(y)i0 can only
depend on x y, due to translational invariance, the space-time integral is trivial, and we get
3
h(0)(0)i0 h(0)(0)i0 .
4
f(1) (T ) =
(3.18)
Finally, at the second order, we get

i
T h 2
f(2) (T ) = lim
hSI i0 hSI i20
V
2V
Z

2
T
h(x)(x)(x)(x) (y)(y)(y)(y)i0
= lim
V
2V x,y 4

Z
Z
, (3.19)
h(x)(x)(x)(x)i0
h(y)(y)(y)(y)i0
x 4
y 4
where we have denoted
Z
d3 x .
(3.20)
Carrying out contractions according to Eq. (3.12), the role of the subtraction term, the second
one in Eq. (3.19), now becomes clear: it cancels all disconnected contractions, i.e. contractions
of the type where all fields at point x are contracted with other fields at the same point. In
other words, the combination in Eq. (3.19) amounts to taking into account only the connected
contractions. This miraculous combinatorial fact is brought about by the logarithm in Eq. (3.10),
i.e., by going from the partition function to the free energy!
As far as the connected contractions are concerned, the Wick theorem tells that
h(x)(x)(x)(x) (y)(y)(y)(y)i0,c
= 4 h(x)(y)i0 h(x)(x)(x) (y)(y)(y)i0,c +

+ 3 h(x)(x)i0 h(x)(x) (y)(y)(y)(y)i0,c
= 4 3 h(x)(y)i0 h(x)(y)i0 h(x)(x) (y)(y)i0,c +

+ 4 2 h(x)(y)i0 h(x)(x)i0 h(x) (y)(y)(y)i0,c +
+ 3 4 h(x)(x)i0 h(x)(y)i0 h(x) (y)(y)(y)i0,c

= 4 3 2 h(x)(y)i0 h(x)(y)i0 h(x)(y)i0 h(x)(y)i0 +
+ (4 3 + 4 2 3 + 3 4 3)h(x)(x)i0 h(x)(y)i0 h(x)(y)i0 h(y)(y)i0 ,
(3.21)
where the subscript (...)c refers to connected.
31
Inspecting the two-point correlators in Eq. (3.21), we note that they either depend on x y, or
on neither x nor y (in the cases where fields at the same point are contracted). Thereby one of
the spacetime integrals is again trivial (just substitute x x + y, and note that h(x + y)(y)i0 =
h(x)(0)i0 ), and cancels against the factor T /V = 1/(V ) in Eq. (3.19). In total, then

Z
2 Z
(3.22)
12 h(x)(0)i40 + 36h(0)(0)i20 h(x)(0)i20 .
f(2) (T ) =
4
x
x
One could go on with the third-order terms in Eq. (3.10): again, it could be verified that the
subtraction terms cancel all disconnected contractions, so that only the connected ones contribute
to f (T ); and that one spacetime integral cancels against the explicit factor T /V . These facts are,
in fact, of general nature, and hold at any order in the weak-coupling expansion.
To summarise, the Wick theorem has allowed us to reduce the terms in Eq. (3.10) to various
structures made of the two-point correlator h(x)(0)i0 . We now turn to its properties.
3.3.
Propagator
The two-point correlator h(x)(y)i0 is usually called the free propagator. Denoting
Z
ei(P +Q)x
= pn +qn ,0 (2)d (d) (p + q) ,
(P + Q)
(3.23)
where P (
pn , p), and pn are bosonic Matsubara frequencies; and employing the Fourier-space
presentation
Z
P iP x
(P )e
,
(3.24)
(x)
P
we recall from basic quantum field theory that the propagator can be written as
P )(
Q)i
0
h(
h(x)(y)i0
= (P + Q)
,
2
P + m2
Z
P iP (xy)
1
e
=
.
2
P + m2
P
(3.25)
(3.26)
Before inserting these expressions into Eqs. (3.18), (3.22), we briefly review their derivation, as
well as some basic properties of h(x)(y)i0 .
In order to carry out the derivation, we return for a moment to a finite volume V , and proceed
as in Sec. 2.2. Let us start by inserting Eq. (3.24) into the definition of the propagator,
Z
P
P )(
Q)i
0.
eiP x+iQy h(
(3.27)
h(x)(y)i0 =
P ,Q
In order to compute this expectation value, we insert Eq. (3.24) also into the free action, S0 , finding
Z
Z
1 P 2
P )|2 .
P ) = 1 P (P 2 + m2 )|(
(3.28)
(P )(P + m2 )(
S0 =
2
2
P
P ) = a(P ) + i b(P ), with a(P ) = a(P ), b(P ) = b(P ); only half of the Fourier
We write (
components are independent, and we could choose the ones specified in Eq. (2.13).
Restricting the sum to independent components, and making use of the symmetry properties of
a(P ) and b(P ), Eq. (3.28) becomes
T X 2
S0 =
(P + m2 )[a2 (P ) + b2 (P )] .
(3.29)
V
Pindep.
32
The Gaussian integral,

R
dx x2 exp(cx2 )
1
R
=
,
dx exp(cx2 )
2c
(3.30)
and the symmetries of a(P ) and b(P ), thus imply that

0
ha(P ) b(Q)i
= 0,
0
ha(P ) a(Q)i
1
V
,
2T P 2 + m2
V
1
,
= (P ,Q P ,Q )
2
2T P + m2
0
hb(P ) b(Q)i
(3.31)
(3.32)
= (P ,Q + P ,Q )
(3.33)
where the delta-functions are of Kronecker-type in finite volume. Thereby the propagator becomes
P )(
Q)i
0
h(
+ i a(P ) b(Q)
+ i b(P ) a(Q)
b(P ) b(Q)i
0
= ha(P ) a(Q)
V
1
1
= P ,Q
= pn +qn ,0 V p+q,0
.
2
2
2
T P +m
P + m2
(3.34)
In the infinite-volume limit,

Z d
d p
1 X
,
V p
(2)d
V p,0 (2)d (d) (p) ,
(3.35)
and we recover Eq. (3.25) which, in combination with Eq. (3.27), in turn leads to Eq. (3.26).
We would now like to learn something more about the behaviour of the propagator h(x)(y)i0 ,
at short and large separations x y. We note, first of all, that as shown by Eqs. (1.84), (1.73),
h
i
X eipn
1 cosh 2 E
h i
.
(3.36)
T
=
p2n + E 2
2E
sinh E
p
2
This equation is valid for 0 ; as is obvious from the left-hand side, we can extend the
validity to by replacing by | |. Thereby the propagator from Eq. (3.26) now
becomes
h
i
Z d
cosh
|x
y
|
Ep
0
0
2
d p ip(xy) 1

i
h
G0 (x y) h(x)(y)i0 =
e
.
(3.37)

E
(2)d
2Ep

2 2
sinh 2 p
Ep
p +m
Since G0 only depends on the separation x y, we set y = 0 in the following.
Consider first short distances, |x|, |x0 | 1/T, 1/m. We may expect the dominant contribution in
the Fourier transform in Eq. (3.37) to come from the regime |p||x| 1, so let us assume |p| T, m.
Then Ep |p|, and Ep |p|/T 1. Consequently,
h
i
i
h
exp 2 |x0 | Ep
cosh 2 |x0 | Ep
i
i
h
h
e|x0 ||p| .
(3.38)
E
E
sinh 2 p
exp 2 p
We note that
1 |x0 ||p|
e
=
2|p|
whereby
G0 (x)
dp0 eip0 x0
,
2
2
2 p0 + p
dd+1 P eiP x
,
(2)d+1 P 2
33
(3.39)
(3.40)
with P (p0 , p).

At this point we can make use of rotational symmetry, in order to choose x in the direction of
the component p0 . Then
Z d Z
Z d+1
d p
d P eiP x
dp0 eip0 |x|
=
d+1
2
d
(2)
P
(2) 2 p20 + p2
Z d
d p e|x||p|
=
(2)d 2|p|
Z
d/2
1
=
d|p||p|d2 e|x||p|
(2)d (d/2) 0
(d 1)
=
,
(3.41)
d/2
(4) (d/2)|x|d1
where we made use of Eq. (2.61). Inserting d = 3 and (3/2) = /2, we find
G0 (x)
1
,
4 2 |x|2
|x|
1 1
,
.
T m
(3.42)
We note that this behaviour is independent of T and m: at short enough distances (in the ultraviolet regime), temperature and masses do not play a role, and the propagator diverges.
Consider, on the other hand, large distances, |x| 1/T (the temporal coordinate x0 remains by
construction always small, i.e. is at most 1/T ). We expect the Fourier-transform in Eq. (3.37)
to now be dominated by small momenta, |p| T . If we simplify the situation further by assuming
that we are also at very high temperatures, T m, then Ep 1 in the relevant regime, and we
can expand the hyperbolic functions in Taylor series, cosh() 1, sinh() . Then
G0 (x) T
dd p eipx
.
(2)d p2 + m2
(3.43)
Restricting for simplicity to d = 3, we can write4

G0 (x)
T
(2)2
T
(2)2
+1
dz
d|p| |p|2
ei|p||x|z
|p|2 + m2
d|p| |p|2 ei|p||x| ei|p||x|

2
2
i|p||x|
0 |p| + m
Z
ip|x|
dp p e
T
(2)2 i|x| p2 + m2
= T
em|x|
,
4|x|
|x|
1
,
T
(3.44)
where the last integral was carried out by closing the contour in the upper half plane.
We note from Eq. (3.44) that at large distances (in the infrared regime), thermal effects modify
the behaviour of the propagator in an essential way. In particular, if we were to set the mass to
zero, then Eq. (3.42) would be the exact behaviour at zero temperature, both at small and at
large distances, while Eq. (3.42) shows that a temperature would slow down the long-distance
decay to T /(4|x|). In other words, we can say that at a non-zero temperature the theory is more
sensitive to infrared physics than at zero temperature.
4 For
a general d,
dd p eipx
(2)d p2 +m2
1
( m )d/21 Kd/21 (m|x|),
(2)d/2 |x|
34
where K is a modified Bessel function.
3.4.
Naive free energy density to O(): ultraviolet divergences
We now return to the free energy density of scalar field theory, given by Eqs. (3.16), (3.18), (3.22).
Noting from Eqs. (2.54) and (3.26) that G0 (0) = I(m, T ), we obtain to O() that
f (T ) = J(m, T ) +
3
[I(m, T )]2 .
4
(3.45)
According to Eqs. (2.72), (2.73),

J(m, T ) =
I(m, T ) =

3
2
m4 2 1
+ ln 2 + + O() + JT (m) ,
64 2
m
2

2
2
m
2 1
+ ln 2 + 1 + O() + IT (m) ,
16 2
(3.46)
(3.47)
where the functions JT (m) and IT (m) are finite, and were evaluated in various limits in Eqs. (2.78),
(2.79), (2.81), (2.92).
Inserting Eqs. (3.46), (3.47) into Eq. (3.45), we note that the result is, in general, ultraviolet
divergent. For instance, restricting for simplicity to very high temperatures, T m, and making
use of Eq. (2.92),
T2
mT
IT (m)
+ O(m2 ) ,
(3.48)
12
4
the dominant term at 0 reads

1 2 2
3
2
4
3
4
2
m + T m
T m + O(m ) + O( ) + O(1) .
(3.49)
f (T )
64 2
2
2
This result is obviously non-sensical: the divergences appear even to depend on the temperature.
A proper procedure requires renormalization; we return to this in Sec. 3.7, but identify first also
another problem with the naive result.
3.5.
Naive free energy density to O(2 ): infrared divergences
Let us consider the second order contribution to Eq. (3.45), given in Eq. (3.22). With the notation
of Eqs. (3.20), (3.37), we can write it as
Z
Z
9 2
3 2
2
4
[G0 (x)] [I(m, T )]
[G0 (x)]2 .
(3.50)
f(2) (T ) =
4
4
x
x
The question we would like to answer is, what happens if we take the limit that the particle mass
m is very small, m T . As Eqs. (3.45), (2.81), (3.48) show, at O() this limit is perfectly welldefined. Consider then the first term in Eq. (3.50). We know from Eq. (3.42) that the behaviour
of G0 (x) is independent of m at small x; thus nothing particular happens for |x| T 1 . On the
other hand, for large |x|, G0 (x) is given by Eq. (3.42). We can estimate the contribution as
m|x| 4
Z Z
Z
Te
.
(3.51)
[G0 (x)]4
d3 x
d
4|x|
1
1
>
>
|x| T
|x| T
0
This integral is convergent even for m 0; so in the first term of Eq. (3.50), taking m 0 does
not cause any divergences.
Consider then the second term in Eq. (3.50). Repeating the argument of Eq. (3.51), we get
m|x| 2
Z Z
Z
Te
3
2
.
(3.52)
d x
[G0 (x)]
d
4|x|
1
1
|x| >
|x| >
0
T
T
35
If we now attempt to take m 0, the integral is linearly divergent! Because the problem emerges
from large distances, we call this an infrared divergence.
In fact, it is easy to be more precise about the form of the divergence. We can write
Z
[G0 (x)]
eiP x P eiQx
2 + m2
P 2 + m2 Q Q
P
Z Z
P
Z
P
Z
P
(P + Q)
P ,Q
1
2
2
2 + m2 )
(P + m )(Q
1
2
[P + m2 ]2
P
d
I(m, T ) .
dm2
(3.53)
Inserting Eq. (3.48), we get

Z
1 d
T
[G0 (x)]2 =
I(m, T ) =
+ O(1) .
2m
dm
8m
x
(3.54)
Therefore, for m T , Eq. (3.50) evaluates to

9
T4 T
f(2) (T ) = 2
+ O(m0 ) ,
4 144 8m
(3.55)
and indeed diverges for m 0.

It is clear that, like the ultraviolet divergence in Eq. (3.49), the infrared divergence in Eq. (3.55)
must also be an artifact of some sort: a gas of weakly interacting massless scalar particles should
certainly have a finite pressure and other thermodynamic properties, just like a gas of massless
photons has. We return to the resolution of this problem in Sec. 3.8.
3.6.
Exercise 4
(a) Let us consider a scalar field theory where the interaction term
Derive the expression for f (T ) up to O( 3 ).
1
4
4 is replaced with
1
3
3 .
(b) What kind of infrared divergences does this expression have, if we take the limit m 0?
(a) According to Eq. (3.10), the radiative corrections to the free energy density can be compactly
represented by the formula
f(1) (T ) =
=
E
1 D
exp(SI )
V
0,connected
E
D
1 2
SI SI + . . .
R .
2
0,connected,drop overall x
(3.56)
The term of O() obviously vanishes, since it contains an odd number of fields. The term of
O( 2 ) reads
Z
2
h(x)(x)(x)(0)(0)(0)i0,c .
(3.57)
f(2) (T ) =
18 x
36
Here, according to the Wick theorem,

h(x)(x)(x) (0)(0)(0)i0,c
and we get
3 h(x)(0)i0 h(x)(x) (0)(0)i0,c +
+
=
2 h(x)(x)i0 h(x) (0)(0)(0)i0,c

3 2 h(x)(0)i0 h(x)(0)i0 h(x)(0)i0 +
+
+
3 1 h(x)(0)i0 h(x)(x)i0 h(0)(0)i0 +

2 3 h(x)(x)i0 h(x)(0)i0 h(0)(0)i0
=
+
6 h(x)(0)i0 h(x)(0)i0 h(x)(0)i0 +

9 h(x)(x)i0 h(x)(0)i0 h(0)(0)i0 ,
1
f(2) (T ) = 2
3
[G0 (x)]3
1 2
[G0 (0)]2
2
G0 (x) .
(3.58)
(3.59)
It turns out, however, that the latter term in Eq. (3.59) should actually be neglected. The
reason is that once we add the interaction 13 3 , then the ground state of the theory is no
longer at hi = 0, as we have (implicitly) assumed; to see this, just compute hi to O()! On
the other hand, the problem can be rectified by also adding another term to the potential
V (), of the type . Then there are additional contributions both to hi and to f (T ):
Z
1
h(0)i h(0) [(x) (x)(x)(x)]i0
3
x
Z
(3.60)
= [ + G0 (0)] G0 (x) ,
x
Z D
E
1
1
f (T )
2 (x) (0) + [(x)(x)(x) (0) + (x) (0)(0)(0)]
2 x
3
0
Z
1 2
(3.61)
= [ + 2G0 (0)] G0 (x) .
2
x
Summing together Eq. (3.59) and (3.61), we get
Z
Z
1
1 2
2
3
[G0 (x)] [ + G0 (0)]
G0 (x) .
f(2) (T ) =
3
2
x
x
(3.62)
Hence, we note that if we choose the coefficient such as to make the expression in Eq. (3.60)
vanish, as it should, we simultaneously need to remove part of the terms in f(2) (T ).
To summarise, a physically meaningful expression is given by
Z
1
f (T ) = J(m, T ) 2 [G0 (x)]3 + O( 4 ) .
3
x
(3.63)
(b) Inspecting the large-|x| behaviour like in Eqs. (3.51), (3.52), we obtain
Z
1
|x| >
T
[G0 (x)]3
d
0
m|x| 3
Te
d3 x
.
4|x|
1
>
|x| T
The radial part of the integration yields

Z
Z

d
dx 3mx
dx
e
=
ln x e3mx
dx
x0
x0 x
Z
i
h
3mx
= lnx e
dx ln x e3mx
+ 3m
x0
x0
37
(3.64)
=
=
=
Z
dx lnx e3mx + O(x0 m)
ln x0 e3mx0 + 3m
0
Z
x
x
ln x0 +
d
x ln
e + O(x0 m)
3m
0
Z
Z
x
ln x0 ln(3m)
d
xe +
d
x ln
x ex + O(x0 m)
0
1
E + O(x0 m) ,
ln
3x0 m
(3.65)
where we denoted x |x|; inserted a lower bound x0 to the integration, assuming x0 1/m;
and set x0 0 whenever that does not lead to an ultraviolet divergence.
We now see that Eq. (3.65) diverges logarithmically if we set m 0.
38
3.7.
Proper free energy density to O(): ultraviolet renormalization
In Sec. 3.4 we attempted to compute the free energy density f (T ) of scalar field theory up to
O(), but found a result which appeared divergent. Let us now show that, as must be the case
in a renormalizable theory, the divergences disappear order-by-order in perturbation theory, if we
re-express f (T ) in terms of finite renormalized parameters.
In order to proceed properly, we need to first change the notation somewhat. The zero-temperature
parameters we employed before, m2 , , will now be re-interpreted as bare parameters, m2B , B .
(The temperature T , in contrast, is a physical property of the system, and is not subject to any
modifications.) The expansion in Eq. (3.45) can then be written as
f (T ) = (0) (m2B , T ) + B (1) (m2B , T ) + O(2B ) .
(3.66)
As a second step, we introduce some renormalized parameters, m2R , R . These could either be
directly physical quantities (say, the mass of the scalar particle, and the scattering amplitude with
some particular kinematics), or quantities which are not yet directly physical, but are related
to physical quantities by finite equations (say, so-called MS scheme parameters). In any case,
it is natural to choose the renormalized parameters such that in the limit of an extremely weak
interaction, R 1, they formally agree with the bare parameters. In other words,
m2B
m2R + R f (1) (m2R ) + O(2R ) ,
(3.67)
R + 2R g (1) (m2R ) + O(3R ) .
(3.68)
Note that renormalized parameters are defined at zero temperature, so no T can appear in these
relations. The functions f (i) and g (i) are, in general, divergent in the limit that regularization is
removed; for instance, in dimensional regularization, they contain poles like 1/.
The idea now is simply to convert the expansion in Eq. (3.66) into an expansion in R , by
inserting the expressions from Eqs. (3.67), (3.68), and Taylor-expanding in R :

(0) (m2R , T ) (1) 2
f (T ) = (0) (m2R , T ) + R (1) (m2R , T ) +
f
(m
)
+ O(2R ) .
(3.69)
R
m2R
We note that to O(2R ), only the mass parameter needs to be renormalized.
To carry out the renormalization in practice, we need to choose a scheme. We will here choose the
so-called pole mass scheme, where m2R is taken to be the physical mass squared of the -particle,
denoted by m2phys . In Minkowskian spacetime, this appears as the exponential time-evolution
factor,
eiE0 t eimphys t ,
(3.70)
in the propagator of a particle at rest, p = 0. In Euclidean spacetime, this corresponds to the
exponential fall-off, exp(mphys ), of the propagator. Therefore, in order to determine m2phys to
O(R ), we need to compute the full propagator, G(x), to O(R ) at zero temperature.
The full propagator can be defined as the generalization of Eq. (3.37) to the interacting case:
G(x)
h(x)(0) exp(SI )i0

hexp(SI )i0
h(x)(0)i0 h(x)(0)SI i0 + O(2B )
=
1 hSI i0 + O(2B )
i
h
= h(x)(0)i0 h(x)(0)SI i0 h(x)(0)i0 hSI i0 + O(2B ) .
(3.71)
We may recall from Quantum Field Theory that the second term inside the square brackets serves
to cancel disconnected contractions, just like the subtractions in Eq. (3.10) did for the free energy
39
density. Therefore, we will drop the second term in the following, and replace the expectation
value in the first term by h...i0,c , like we already did in Eq. (3.21).
Now, let us start by inspecting the leading (zeroth order) term in Eq. (3.71), in order learn how
mphys can be conveniently extracted from the propagator. Introducing the notation
Z d+1
Z
Z
P
d P
=
lim
,
(3.72)
T 0
(2)d+1
P
the free propagator reads (cf. Eq. (3.26))

G0 (x) = h(x)(0)i0 =
eiP x
.
P 2 + m2
(3.73)
For Eq. (3.70), we need to project to zero spatial momentum, p = 0; evidently, this can be achieved
by taking a spatial average of G0 (x):
Z
Z
dp0 eip0
dd x h(, x)(0)i0 =
.
(3.74)
2 p20 + m2
We see that we get an integral which can be evaluated with the help of the Cauchy theorem and,
in particular, that the exponential fall-off of the correlation function is determined by the pole
position of the momentum-space propagator:
Z
em
1
2i
, 0.
(3.75)
dd x h(, x)(0)i0 =
2
2im
Hence,

m2phys =0 = m2 ,
(3.76)
and, more generally, the physical mass can be extracted by determining the pole position of the full
propagator in momentum space.
We then proceed to the second term in Eq. (3.71):
Z
B
h(x)(0)SI i0,c =
h(x)(0) (z)(z)(z)(z)i0,c
4 z
Z
B
4 3 h(x)(z)i0 h(z)(0)i0 h(z)(z)i0
=
4 z
Z
= 3B G0 (0) G0 (z)G0 (x z)
z
Z Z
Z
1
1
1
eiQz eiR(xz)
= 3B
2
2
2
2
2
Q + mB R + m2B
P P + mB z Q,R
Z
eiRx
= 3B I0 (mB )
.
(3.77)
2 + m2 )2
(R
R
B
Summing together with Eq. (3.73), the full propagator reads

Z
1
1
iP x
2
e
G(x) =
3B I0 (mB )
+ O(B )
P 2 + m2B
(P 2 + m2B )2
P
Z
eiP x
+ O(2B ) ,
=
2 + m2 + 3B I0 (mB )
P P
B
(3.78)
where we have effectively resummed a series of higher order corrections.

The same steps that lead us from Eq. (3.74) to (3.76) now produce
m2phys = m2B + 3B I0 (mB ) + O(2B ) .
40
(3.79)
Recalling from Eq. (3.68) that m2B = m2R + O(R ), B = R + O(2R ), this relation can be inverted,
to give
m2B = m2phys 3R I0 (mphys ) + O(2R ) .
(3.80)
This corresponds precisely to Eq. (3.67). The function I0 , given in Eq. (2.73), diverges in the limit
0,

m2phys 2 1
+
ln
+
1
+
O()
,
(3.81)
I0 (mphys ) =
16 2
m2phys
and we may hope that the divergence cancels the unphysical ones that we found in f (T ).
Indeed, let us take the step from Eq. (3.66) to Eq. (3.69), employing the explicit expression from
Eq. (3.45),
3
f (T ) = J(mB , T ) + B [I(mB , T )]2 + O(2B ) .
(3.82)
4
Recalling from Eq. (2.52) that
I(m, T ) =
d
1 d
J(m, T ) ,
J(m, T ) = 2
m dm
dm2
(3.83)
we can expand the two terms in Eq. (3.82) as a Taylor series around m2phys ,
J(mB , T ) =
=
B [I(mB , T )]2
J(mphys , T ) + (m2B m2phys )
J(mphys , T )
+ O(2R )
m2phys
3
R I0 (mphys )I(mphys , T ) + O(2R ) ,
2
R [I(mphys , T )]2 + O(2R ) ,
J(mphys , T )
(3.84)
(3.85)
where we inserted Eq. (3.80). Eq. (3.82) then becomes

h
i
3
f (T ) = J(mphys , T ) + R I 2 (mphys , T ) 2I0 (mphys ) I(mphys , T ) + O(2R )
4

3
3
=
J0 (mphys ) R I02 (mphys ) + JT (mphys ) + R IT2 (mphys ) +O(2R ) , (3.86)
4
4
|
{z
} |
{z
}
T = 0 part
T 6= 0 part
where we have inserted the definitions from Eq. (2.56), J(m, T ) = J0 (m) + JT (m), I(m, T ) =
I0 (m) + IT (m).
Recalling Eqs. (2.72), (2.73), the first term in Eq. (3.86), the T = 0 part, is divergent. However, this term is completely independent of the temperature. Thus it does not play a role in
thermodynamics, but rather corresponds to a vacuum energy density: it only plays a physical role
in connection with gravity. If we included gravity, however, we should also include a bare cosmological constant, B , to the bare Lagrangian; it would contribute additively to Eq. (3.86), and we
can simply define
3
(3.87)
phys B + J0 (mphys ) R I02 (mphys ) + O(2R ) .
4
The divergences are now eaten up by B , and phys is finite.
In contrast, the second term in Eq. (3.86), the T 6= 0 part, is finite: it contains the functions
JT , IT for which we have determined analytically various limiting values in Eqs. (2.78), (2.79),
(2.81), (2.92), and general integral expressions in Eqs. (2.75), (2.76). Therefore all thermodynamic
quantities obtained from derivatives of f (T ), such as the entropy density, specific heat, etc, are
manifestly finite. In other words, the temperature-dependent ultraviolet divergences that we found
in Sec. 3.4 have indeed disappeared through zero-temperature renormalization, as must be the case.
41
3.8.
Proper free energy density to O( 2 ): infrared resummation
We now consider the limit mphys 0, as would be the case (in perturbation theory) for, say,
gluons in QCD. According to Eq. (3.80), this corresponds to mB 0, since I0 (0) = 0 according
to Eq. (2.73). Then we are faced with the infrared problem introduced in Sec. 3.5.
In the limit of a small mass, we can employ high-temperature expansions for the functions
J(m, T ), I(m, T ), given in Eqs. (2.91), (2.94). Employing Eqs. (3.45), (3.55), we write the leading
terms in the small-mB expansion as
O(0B ) :
f(0) (T ) =
O(1B ) :
f(1) (T ) =
=
=
O(2B ) :
f(2) (T ) =
J(mB , T ) =
2 T 4
m2 T 2
m3 T
+ B
B + O(m4B ) ,
90
24
12
3
B [I(mB , T )]2
4
2
2
T
mB T
3
B
+ O(m2B )
4
12
4
4

3
T
mB T 3
B
+ O(m2B T 2 ) ,
4
144
24
4
9
T
T
2B
+ O(m0B ) .
4
144 8mB
(3.88)
(3.89)
(3.90)
Let us inspect, in particular, odd powers of mB . As is obvious from Eqs. (3.88)(3.90), they are
becoming increasingly important as we go further in the expansion. We remember from Sec. 2.6
that odd powers of mB are necessarily associated with contributions from the Matsubara zeromodes. In fact, the odd power in Eq. (3.88) is directly the zero-mode contribution from Eq. (2.86),
odd f(0) = J (n=0) =
m3B T
.
12
(3.91)
The odd power in Eq. (3.89) comes from a cross-term between the zero-mode contribution and the
leading non-zero mode contribution:
odd f(1) =
B mB T 3
3
B I (0, T ) I (n=0) =
.
2
32
(3.92)
Finally, the divergence in Eq. (3.90) comes from a product of two non-zero mode contributions
with a particularly infrared sensitive zero-mode contribution:
odd f(2) =
dI (n=0)
2 T 5
9 2
.
B [I (0, T )]2
= 3B
2
4
dmB
8 mB
(3.93)
Comparing these structures, we immediately see that the expansion parameter related to the odd
powers is
odd f(1)
odd f(2)
B T 2
.
(3.94)
odd f(0)
odd f(1)
8m2B
Thus, if we try to take m2B 0 (or even just m2B B T 2 ), the loop expansion breaks down.
In order to cure this problem, our only hope is to identify and sum the corresponding divergent
terms to all orders. We may then expect that the complete sum has a form where we can take m2B
0, without meeting any more divergences. This procedure is often referred to as resummation.
Fortunately, it is indeed possible to identify the problematic terms. Eqs. (3.91)(3.93) already
show that at order N in B , they are associated with terms containing N non-zero mode contributions I (0, T ), and one zero-mode contribution. Graphically, this corresponds to a loop with one
zero-mode line, dressed with N non-zero mode bubbles. Such graphs are usually called ring
diagrams or, adopting botanistic terminology, daisy diagrams.
42
To be more quantitative, we consider Eq. (3.56) at order N

B:
D
(1)N +1 N E
1
SI
f (T ) =
SI SI2 + . . . +
R
2
N!
(1)N +1
N!
B
4
N *
(1)N +1
N!
B
4
N
2(N 1)
2(N 2)
(3.95)
6
0,...
2 N Z d
N
T
d p
1
.
6N [2(N 1)][2(N 2)]...[2]
T
{z
} |{z}
|
12
(2)d p2 + m2B
{z
}
|
2N 1 (N 1)!
I (0, T )
zero-mode ring
Let us compute the zero-mode part (for simplicity we omit terms of O()):

Z 32
m3B
1
mB
d
d
p
,
=
=
N =1:
(2)32 p2 + m2B
4
dm2B
6

Z 32
d
p
mB
m3B
d
1
d
d
N =2:
,
=
(2)32 (p2 + m2B )2

dm2B
4
dm2B dm2B
6
Z 32
Z 32
d
p
d
p
d
1
1
1
generally :
=
2
2
32
2
N
32
2
(2)
(p + mB )
N 1 dmB (2)
(p + m2B )N 1
N 1 Z 32

d
p
1
1
d
1
1
=
N 1
N 2
1
dm2B
(2)32 p2 + m2B
N 3

mB
d
(1)N
.
(3.96)
=
2
(N 1)! dmB
6
Combining Eqs. (3.95), (3.96), we get
N 3

N

2 N
(1)N +1 3B
mB
(1)N
d
T
N 1
odd f(N ) =
2
(N 1)!
T
2
N!
2
12
(N 1)! dmB
6
N 3

2 N
mB
d
T 1 B T
.
(3.97)
=
2 N!
4
dm2B
6
As a crosscheck, it can easily be verified that this expression reproduces the explicit results for
N = 0, 1, 2 shown in Eqs. (3.91)(3.93).
Now, thanks to the fact that Eq. (3.97) has precisely the right structure to correspond to a
Taylor-expansion, we can sum the contributions in Eq. (3.97) to all orders:
N

N

3
X
d
m3B T
T
B T 2 2
1 B T 2
2
m
+
.
(3.98)
B
N!
4
dm2B
12
12
4
N =0
A miracle has happened: from Eq. (3.98), the limit m2B 0 can be taken without divergences! But
3/2
there is a surprise: we get a contribution of O(B ), rather than O(2B ) as we naively expected
in Sec. 3.5. In other words, infrared divergences in finite-temperature field theory modify even
qualitatively the structure of the weak-coupling expansion.
Setting finally m2B 0, and collecting all the finite terms from Eqs. (3.88)(3.90), we find the
correct expansion of f (T ) in the massless limit:

3/2
2 T 4
B T 2
B T 4
T
f (T ) =
+
+ O(2B T 4 )
(3.99)
90
12
4
4 48

3

15 R 2
15 R
2 T 4
2
+
1
=
+ O(R ) .
(3.100)
90
32 2
16 2
43
Here we have inserted B = R + O(2R ). Eq. (3.100) is our final result.

It is perhaps appropriate to add that despite the complications that we have found, higher
order terms can also be added to Eq. (3.100). In fact, as of today (fall 2007), the coefficients
5/2
5/2
of the five subsequent terms, of orders O(2R ), O(R ln R ), O(R ), O(3R ln R ), and O(3R ),
are also known.5 This progress is possible due to the fact that the resummation of higher order
contributions that we have carried out explicitly in this section, can be implemented much more
elegantly and systematically with so-called effective field theory methods. We return to the general
procedure in later sections, but some flavour of the idea can perhaps be obtained by organizing
the computation in yet another way, outlined in Exercise 5.
3.9.
Exercise 5
(a) Following the zero-temperature computation of m2phys in Eq. (3.79), repeat the determination
of the pole mass at T 6= 0, mB 0. The result can be called the thermal mass term, m2eff .
(b) Argue that in the weak-coupling limit (R 1), the thermal mass term is important only
for the Matsubara zero mode.
(c) Let us write the Lagrangian for m2B = 0 as
LE
1
1
1
1
+ m2eff 2n=0 + B 4 m2eff 2n=0 .
2
2
4
2
|
{z
} |
{z
}
L0
LI
(3.101)
Treating L0 as the free theory, and LI as an interaction of order R , write down the contributions to f(0) and f(1) . Check that once the computation is reorganized this way, the series
is well-behaved, and the result agrees with what we got in Eq. (3.99).
(a) RThe computation
proceeds precisely like the one leading to Eq. (3.79), with the replacement
R
.
Consequently,
P
P
h
i
R T 2
m2B + 3B I(mB , T ) = 3B I(0, T ) =
+ O(2R ) .
4
mB 0
m2eff = lim
2
(3.102)
(b) For the non-zero Matsubara modes, with n 6= 0, we note that R T 2 /4 n2 , so that
the thermal mass term plays a subdominant role in the propagator. In contrast, for the
Matsubara zero-mode, m2eff modifies the propagator significantly for p2 m2eff , removing
any infrared divergences.
(c) The free propagators are now different for the Matsubara zero-modes and non-zero modes:
0
h (P ) (Q)i
0
hn=0 (P )n=0 (Q)i
1
,
n2 + p2
1
= (P + Q)
.
2
p + m2eff
= (P + Q)
(3.103)
(3.104)
Eq. (3.16) gets replaced with

f(0) (T ) =
Z
Z
P1
1
ln(P 2 ) + T
ln(p2 + m2eff ) const.
2
2
p
5 A.
Gynther, M. Laine, Y. Schr

oder, C. Torrero and A. Vuorinen, Four-loop pressure of massless O(N) scalar
field theory, JHEP 04 (2007) 094 [arXiv:hep-ph/0703307].
44
=
=
J (0, T ) + J (n=0) (meff , T )

2 T 4 m3eff T
90
12
(3.105)
Eq. (3.17) now contains the two terms of the LI in Eq. (3.101), and Eq. (3.18) becomes
f(1) (T ) =
=
=
3
1
B h(0)(0)i0 h(0)(0)i0 m2eff hn=0 (0)n=0 (0)i0
4
2
h
i2 1
3
B I (0, T ) + I (n=0) (meff , T ) m2eff I (n=0) (meff , T )
4
2
4

3
1 2 meff T
T
meff T 3
m2eff T 2
+ meff
B
+
.
4
144
24
16 2
2
4
(3.106)
Inserting Eq. (3.102) into the last term, we see that this contribution precisely cancels the
large linear term within the square brackets. As we recall from Eq. (3.92), the linear term
was part of the problematic series that needed to be resummed; hence the problematic series
does not get generated at all with the present reorganization,
Combining Eqs. (3.105), (3.106), we then get
f (T ) =
T4
2 T 4
3
m3 T
+ R
eff + O(2R ) .
90
4
144
12
(3.107)
which exactly agrees with Eq. (3.99).

The cancellation that took place in Eq. (3.106) can also be verified at higher orders. In
particular, proceeding to O(2R ), it can be seen that the structure in Eq. (3.93) gets cancelled
as well: the resummation of infrared divergences that we carried out explicitly in Eq. (3.98)
can indeed be fully captured by the reorganization in Eq. (3.101).
45
4.
4.1.
Fermions
Path integral for the partition function of a fermionic harmonic oscillator
As in the bosonic case, the structure of the path integral can most easily be derived by considering
a non-interacting field living in a zero-dimensional space (d = 0). We refer to this system as the
harmonic oscillator.
In order to introduce the fermionic harmonic oscillator, let us start by briefly summarizing the
main formulae for the bosonic case. In the operator description, the basic commutation relations,
the Hamiltonian, the energy eigenstates, and the completeness relations, can be expressed as
[
a, a
] = 0 , [
a , a
] = 0 , [
a, a
] = 1 ;

~
1
= ~ a
=
(
a a
+a
a
) ;
H
a
+
2
2
a
|ni = n + 1|n + 1i , a
|ni = n|n 1i ,
Z
Z
X
dp
1=
|nihn| = dx |xihx| =
|pihp| .
2~
n
(4.1)
(4.2)
n = 0, 1, 2, ... ;
(4.3)
(4.4)
and the various path integral represenThe observable we are interested in is Z = Tr [exp( H)],
tations we obtained for this are (cf. Eqs. (1.37), (1.41), (1.44))
Z

Z
1 ~
[p( )]2
DxDp
exp
ip( )x(
) + V (x( ))
d
=
~ 0
2m
x(~)=x(0) 2~
2

Z
Z
1 ~
m dx( )
+ V (x( ))
Dx exp
= C
d
~ 0
2
d
x(~)=x(0)

Z
Z ~
1
= C
Dx exp
d LE , LE = LM (t i ) ,
~ 0
x(~)=x(0)
Z
(4.5)
(4.6)
(4.7)
where C is a constant.
In the fermionic case, we replace the algebra of Eq. (4.1) through
{
a, a
} = 0 ,
{
a , a
} = 0 ,
{
a, a
} = 1 .
(4.8)
Consider next the Hilbert space, the analogue of Eq. (4.3). We define a vacuum state |0i by
a
|0i 0 .
(4.9)
|1i a
|0i .
(4.10)
Then we define a one-particle state |1i by
It is now easy to see that the Hilbert space does not contain any other states: operating on |1i
with a
or a
gives either the already known state |0i, or nothing:
a
|1i = a
a
|0i = [1 a
a
]|0i = |0i ,

a
|1i = a
a
|0i = 0 .
(4.11)
(4.12)
The Hamiltonian, the analogue of Eq. (4.2), is an operator acting in this space, and can be
defined as

1 ~
~ a
=
(
a a
a
a
) .
(4.13)
H
a
2
2
46
The observable remains the partition function,

i
h
Z = Tr e H = h0|e H |0i + h1|e H |1i .
(4.14)
Due to the simple structure of the Hilbert space, the partition function can be evaluated almost
trivially:
Z
X
~
(~)n
n
h0|0i +
h1|(
a a
) |1i e 2
{z
}
|
n!
n=0
1
i ~
~
h
1 + e~ e 2 = 2 cosh
,
2T
(4.15)
to be compared with Eq. (1.17). However, like in the bosonic case, it will ultimately be more useful
to write a path integral representation for the partition function, rather than to remain with the
operator formulation, and this will be our goal in the following.
As we recall from the bosonic case, the essential ingredient in the derivation of the path integral
is a repeated use of the completeness relations in Eq. (4.4) (cf. Sec. 1.3). Therefore, we now
need to find some analogues of Eqs. (4.4) for the fermionic system. This can be achieved with
the help of Grassmann variables. In short, the answer will be that while in the bosonic case the
system of Eqs. (4.1) leads to commuting classical fields, x( ), p( ), = 0...~, in the fermionic
case the system of Eqs. (4.8) leads to anti-commuting Grassmann fields, c( ), c ( ), = 0...~.
Furthermore, while x( ) is periodic (there is no periodicity constraint for p( )), the fields c( ), c ( )
will both be anti-periodic over the compact -interval.
We define the Grassmann variables c, c (and, more generally, the Grassmann fields c( ), c ( ))
through the following axioms:
c, c are treated as independent variables, like x, p.
c2 = (c )2 0, cc = c c.
R
R
R
R
integration is defined through dc = dc 0, dc c = dc c 1.
the integration is also Grassmann-like, in the sense that {c, dc} = {c, dc } = {c , dc} =
{c , dc } = 0, {dc, dc} = {dc, dc } = {dc , dc } = 0.
R
by convention we write the integration measure over both c and c in the order dc dc.
a field c( ) is a collection of independent Grassmann variables, one at each point (0, ~).
c, c are defined to anticommute with a

, a
as well, so that products like c
a act as regular

bosonic operators, i.e. [c
a , c ] = 0.
We now define a ket-state, |ci, and a bra-state, hc|, which are eigenstates of a
(from the left)
and a
(from the right), respectively:
|ci
hc|
eca |0i = (1 c
a )|0i ;
h0|eac = h0|(1 a
c ) ;
a
|ci = c|0i = c|ci ,
hc|
a = h0|c = hc|c .
(4.16)
(4.17)
Such states possess the transition amplitude

hc |ci = h0|(1 a
c )(1 c
a )|0i = 1 + h0|
ac c
a |0i = 1 + c c = ec
47
(4.18)
With these states, we note that we can define the objects we need:
Z
Z
c c
dc dc e
|cihc| =
dc dc (1 c c)(1 c
a )|0ih0|(1 a
c )
Z
= |0ih0| + dc dc c a
|0ih0|
a c
=
Z
dc dc ec c hc|A|ci
=
=
=
=
|0ih0| + |1ih1| = 1 ,
Z
c
dc dc (1 c c)h0|(1 + a
c )A(1
a )|0i
Z
dc dc h0|
h0|A|0i
a c A c a
|0i
Z
h0|A|0i h1| dc dc c c A|1i
(4.19)
+ h1|A|1i
= Tr [A]
,
h0|A|0i
(4.20)
where we assumed A to be a bosonic operator, for instance the Hamiltonian.

Representing now the trace in Eq. (4.14) as in Eq. (4.20), and splitting up the exponential into
a product of N small terms like in Eq. (1.31), we can write
Z
H
~
H
.
(4.21)
Z = dc dc ec c hc|e ~ e ~ |ci ,
N
R
Then we insert Eq. (4.19) in between the exponentials, as 1 = dci dci eci ci |ci ihci |. Thereby we
are faced with objects like
i

h

(4.16), (4.17)
eci+1 ci+1 hci+1 |e ~ H(a ,a) |ci i
=
exp ci+1 ci+1 hci+1 |ci i exp H(ci+1 , ci )
~
h
i
(4.18)
=
exp ci+1 ci+1 + ci+1 ci H(ci+1 , ci )
~

c
i+1
i
+ H(ci+1 , ci )
. (4.22)
=
exp ~ci+1
~
Finally, some more attention needs to be paid to the right-most and left-most exponentials in
Eq. (4.21). We may define c1 c, which clarifies the fate of the right-most exponential, but the
left-most needs to be inspected in detail:
Z
Z

dc1 dc1 ec1 c1 hc1 |e ~ H(a ,a) | dcN dcN |cN i
Z
Z
i
h
=
dc1 dc1 dcN dcN exp c1 c1 c1 cN H(c1 , cN )
~

Z
Z
c1 + cN
=
dc1 dc1 dcN dcN exp ~c1
+ H(c1 , cN )
~

Z
Z
c1 cN
=
dc1 dc1 dcN dcN exp ~c1
+ H(c1 , cN )
.
(4.23)
~
The total becomes

Z
SE

1
=
exp SE ,
~

N
X

ci+1 ci
+ H(ci+1 , ci )
=
~ci+1

i=1
Z
dcN dcN
dc1 dc1
cN +1 c1 ,c
N +1 c1
48
(4.24)
.
(4.25)
Finally, taking the formal limit N , 0, with ~ = N kept fixed, we arrive at

Z
Z
dc( )
1 ~
+ H(c ( ), c( ))
.
d ~c ( )
Z = c (~)=c (0) Dc ( )Dc( ) exp
~ 0
d
(4.26)
c(~)=c(0)
To summarize, the fermionic path integral resembles the bosonic one in Eq. (4.5), but the Grassmann fields obey antiperiodic boundary conditions, unlike in the bosonic case.
In fact, the analogy between the bosonic and the fermionic path integrals can be pushed even
further, but for that we need to specify precisely the form of the fermionic Hamiltonian, rather
than just use the ad hoc definition in Eq. (4.13). To do this, we need to turn to Dirac fields.
4.2.
The Dirac field at finite temperature
In order to make use of the result of the previous section, we need to construct the Hamiltonian
for the Dirac field, and identify the objects that play the roles of the operators a
, a
. Our starting
point will be the classical Minkowskian Lagrangian,
m) ,
LM = (i
where
0 ,
{ , } 2 ,
(4.27)
m m 144 .
( ) = 0 0 ,
(4.28)
The conjugate momentum is defined by

=
LM
0 = i ,
= i
(0 )
(4.29)
and the Hamiltonian density becomes

i
H = 0 LM = [i
i + m] .
(4.30)
If we now go to the operator language and recall the canonical commutation relations,
{ (x0 , x), (x0 , y)} = { (x0 , x), (x0 , y)} = 0 ,
{ (x , x),
(x , y)} = { (x
0
, x), i (x0 , y)}
= i
(4.31)
(d)
(x y) ,
(4.32)
where the subscripts refer to the Dirac indices, , (1, ..., 4), we note that , play precisely
the same roles as a
, a
in Eq. (4.8), and that from the operator point of view the Hamiltonian has
indeed the structure of Eq. (4.13),
Z
0 , x) .
H = dd x (x0 , x)[i 0 i i + m 0 ](x

(4.33)
Therefore, rephrasing Eq. (4.26) [we denote c , c , and set again ~ = 1], the object
within square brackets in Eq. (4.26), which we define to be the Euclidean Lagrangian, reads now
0 i i i + m] .
LE + [i 0 i i + m 0 ] = [
(4.34)
Most remarkably, a comparison of Eqs. (4.27) and (4.34) shows that our old recipe from Eq. (1.43),
LE = LM ( = it), again works! (Note that i0 = it .)
It is conventional and convenient to simplify the appearance of Eq. (4.34) by introducing so-called
Euclidean Dirac matrices through
0 0 ,
i i i .
49
(4.35)
According to Eq. (4.28), these satisfy the algebra

= .
{
, } = 2 ,
We also define
0 ,
(4.36)
i i .
(4.37)
Thereby Eq. (4.34) can be written in the simple form

+ m] ,
LE = [
(4.38)
and the partition function becomes

Z=

Z Z
d
D(, x)D(, x) exp

d d x LE ,
(,x)=
(0,x)
(4.39)
(,x)=(0,x)
and set again ~ 1. It is important

where we have substituted integration variables from to ,
to keep in mind that in the path integral formulation, and are independent integration variables.
In order to evaluate Z, it is useful to go to Fourier space, like with the scalar fields. We write
Z
Z
P iP x
P iP x
(P ) .
(4.40)
e
e
(P ) , (x)
(x)
P
The anti-periodicity in Eq. (4.39) requires that P be of the form

P = (nf , p) ,
ein = 1 ,
(4.41)
whereby the fermionic Matsubara frequencies become

1
nf = 2T n +
,
2
nZ,
(4.42)
i.e. nf = T, 3T, ... . Note that anti-periodicity removes the Matsubara zero-mode from the
spectrum. Correspondingly, recalling the discussion in Sec. 3.8, it is clear that there are no infrared
problems associated with fermions!
In the Fourier representation, the exponent in Eq. (4.39) becomes
SE
d dd x (x)[
+ m](x)
0
Z Z Z
P P i(P Q)x
Q)[i
P )
P + m](
(
e
=
f
f
Z
P
P ) ,
=
(P )[iP/ + m](
(4.43)
Pf
where we made use of Eq. (3.23), and defined P/ P . In contrast to real scalar fields, all Fourier
modes are independent. Up to a constant, we can then also change the integration variables in
Eq. (4.39) to be the Fourier modes. Furthermore, we note that
Z
Z
c ac
dc dc e
= dc dc [c ac] = a ,
(4.44)
R
R
dc dc c c
dc dc c c ec ac
1
R
R
=
= ,
(4.45)
a
dc dc ec ac
dc dc [c ac]
50
and recall that the generalizations to a multicomponent case read

Z nY

o

dci dci exp ci Mij cj = det(M ) ,
(4.46)

c
c
exp
c
M
c
dc
dc
k
ij
j
i
i
i
l
i
= (M 1 )kl .

R Q
i dci dci exp ci Mij cj
R Q
(4.47)
Armoured with this knowledge, we can now work out the partition function Z, as well as the
propagator, which is needed for computing perturbative corrections to the partition function. From
Eqs. (4.39), (4.43), (4.46),
Y
det[iP/ + m]
Z = C
Pf
Y
det[iP/ + m]
Y
Pf
Pf
21
det[iP/ + m]
,
(4.48)
where C is some constant, and have we replicated the determinant and compensated for that by
taking the square root. The reason for the replication is that
[iP/ + m][iP/ + m] = P/ P/ + m2 = (P 2 + m2 )144 ,
(4.49)
where we made use of Eq. (4.36). Thereby

Z = C
Y
Pf
21
Y
det[(P 2 + m2 )144 ]
= C (P 2 + m2 )2 ,
(4.50)
Pf
and the free energy density f (T ) becomes

f (T ) =
=
=

F
T
= lim ln Z
V V
V
V
X
T
ln(P 2 + m2 ) + const.
lim
2
V V
Pf
Z
P 1
ln(P 2 + m2 ) + const. ,
4
2
lim
(4.51)
Pf
where we identified the sum-integration measure from Eqs. (2.9), (2.55).

The following remarks are in order:
the sum-integral appearing in Eq. (4.51) is similar to the bosonic one in Eq. (2.51), but
is preceded by a minus-sign, and contains fermionic Matsubara frequencies. These are the
characteristic properties of fermions.
the factor 4 in Eq. (4.51) corresponds to the independent spin degrees of freedom contained
in a Dirac spinor.
like for the scalar field theory in Eq. (2.51) (or Eq. (2.24)), there is a constant part in the
sum, independent of the energy (or mass). We will not specify it explicitly here; rather, there
will be an implicit specification in Sec. 4.3, where we relate generic fermionic thermal sums
to the already known bosonic ones.
51
Finally, from Eqs. (4.43), (4.47), we find the propagator:
0 = (P Q)[i
P/ + m1]1 = (P Q)
[iP/ + m1] .
h (P ) (Q)i
P 2 + m2
(4.52)
Once interactions are added, they can be reduced to a product of propagators with the Wick
theorem in the same way as in Sec. 3.2 (remembering, though, that the Grassmann nature of the
Dirac fields produces a minus-sign in every commutation). Finally, we reiterate that taking m2 0
in Eq. (4.52) does not lead to any divergences, because the denominator of a fermion propagator
can never vanish at finite temperature (cf. Eq. (4.42)), unlike that of a boson.
4.3.
Fermionic thermal sums
Let us now consider the same problem as in Sec. 2.3, but with fermionic Matsubara frequencies:
X
f (n ) .
(4.53)
Sf T
f
n
For clarity we also denote the sum in Eq. (2.29) by Sb from now on. We can write:
Sf (T ) = T [... + f (3T ) + f (T ) + f (T ) + ...]
= T [... + f (3T ) + f (2T ) + f (T ) + f (0) + f (T ) + f (2T ) + ...]
T [... + f (2T ) + f (0) + f (2T ) + ...]

T
T
i

Th
T
T
T
= 2
... + f 6
+ f 4
+ f 2
+ f 0 + f 2
+ f 4
+ ...
2
2
2
2
2
2
T [... + f (2T ) + f (0) + f (2T ) + ...]
T
Sb (T ) .
(4.54)
= 2Sb
2
Thereby all fermionic sums follow from the known bosonic ones!
To give an example, consider Eq. (2.34),
Sb (T ) =
dp
f (p) +
2
+i0+
i0+
dp
[f (p) + f (p)]nB (ip) .
2
(4.55)
Eq. (4.54) now implies

Sf (T ) =
dp
f (p) +
2
+i0+
i0+
h
i
dp
(T /2)
(T )
[f (p) + f (p)] 2nB (ip) nB (ip) .
2
(4.56)
Thereby the zero-temperature part remains unchanged, while the finite-temperature part has the
new weight
(T /2)
2nB
(T )
(ip) nB (ip) =
=
=
2
1
exp(2ip) 1 exp(ip) 1

2
1 exp(ip)
1
1 =
exp(ip) 1 exp(ip) + 1
[exp(ip) 1][exp(ip) + 1]
(T )
nF (ip) ,
(4.57)
where nF (p) 1/[exp(p) + 1] is the Fermi-Dirac distribution function. In total, then, fermionic
sums can be converted to integrals according to
Sf (T ) =
dp
f (p)
2
+i0+
i0+
52
dp
[f (p) + f (p)]nF (ip) .
2
(4.58)
4.4.
Exercise 6
(a) Defining
Z
i
1 P h 2
ln(P + m2 ) const. ,
2
P
Z f
P
1
,
2
P + m2
P
Jf (m, T ) =
If (m, T ) =
(4.59)
(4.60)
and writing
Jf (m, T ) = J0 (m) + JTf (m) ,
find the general expressions for
JTf (m),
If (m, T ) = I0 (m) + ITf (m) ,
(4.61)
ITf (m).
(b) Derive the low-temperature and the high-temperature expansions for JTf (m), ITf (m). Note
the absence of odd powers of m in the high-temperature expansions.
(c) Consider the fermionic version of Eq. (1.83),
Gf ( ) T
X ein
,
n2 + 2
0< <.
(4.62)
What is the explicit expression for Gf ( )?

(a) We proceed according to Eq. (4.54). From Eq. (2.50),
Z

T ln 1 eEk ,
JTb (m) =
(4.63)
so that
JTf (m)
=
=
Zk
k
i

h
T ln 1 e2Ek ln 1 eEk

T ln 1 + eEk .
(4.64)
(4.65)
From Eq. (2.53),

ITb (m) =
1
nB (Ek ) ,
Ek
and then the same steps as in Eq. (4.57) lead to

Z
1
nF (Ek ) .
ITf (m) =
k Ek
(4.66)
(b) From Eq. (2.78), the low-temperature expansion for JTb reads
JTb (m)
m
2T
23
e T .
(4.67)
In Eq. (4.54), the first term is exponentially suppressed, and thus

JTf (m) T 4
53
m
2T
32
e T .
(4.68)
From Eq. (2.79), the low-temperature expansion for ITb reads

3
T3 m 2 m
b
IT (m)
e T .
m 2T
Again the first term in Eq. (4.54) is exponentially suppressed, so that

3
T3 m 2 m
e T .
ITf (m)
m 2T
From Eq. (2.81), the high-temperature expansion for JTb reads

E
3
m6 (3)
me
m3 T
m4
2 T 4 m2 T 2
+
ln
+ ... .
+
JTb (m) =
2
90
24
12
2(4)
4T
4
3(4)4 T 2
(4.69)
(4.70)
(4.71)
According to Eq. (4.54), we then get

E

1 2 T 4 1 m2 T 2
3
me
m3 T
m4
8m6 (3)
ln
+
2
+
ln
2
+
2
8 90
2 24
12
2(4)
4T
4
3(4)4 T 2

3
m6 (3)
meE
m2 T 2
m3 T
m4
2 T 4
ln
...
+
+
2
90
24
12
2(4)
4T
4
3(4)4 T 2

E
7 2 T 4 m2 T 2
3
7m6 (3)
me
m4
+
ln
+ ... .
(4.72)
2
8 90
48
2(4)
T
4
3(4)4 T 2
JTf (m) =
+
=
We note that the cubic term indeed disappears from the difference. Finally, from Eq. (2.92),
the high-temperature expansion for ITb reads

E
T 2 mT
1
2m4 (3)
me
2m2
ITb (m) =
+
ln
+ ... ,
(4.73)
2
12
4
(4)
4T
2
(4)4 T 2
and Eq. (4.54) then yields
ITf (m) =
E

1
1 T2
me
mT
4m2
16m4 (3)
ln
+
ln
2
+
2
2 12
4
(4)
4T
2
(4)4 T 2

1
2m4 (3)
T 2 mT
meE
2m2
ln
+ ...
+
+
2
12
4
(4)
4T
2
(4)4 T 2

E
T2
1
14m4 (3)
me
2m2
+
ln
+ ... .
2
24
(4)
T
2
(4)4 T 2
(4.74)
Again, the term odd in m has disappeared from the result.

(c) According to Eq. (1.88),
Gb ( )
h
i
cosh
X ein
2
1
h i
T
=
2
2
n +
2
sinh
b
n
( )
1 e
+e
(4.75)
2
e 1
(4.54), we get

i
i
h
1
1 h ( )
2
(2 )

e
+e
e
+e
2 e2 1
e 1
io
n
h
1
1
2e(2 ) + 2e (e + 1) e( ) + e
2 (e 1)(e + 1) |
{z
}
i
h
(e 1) e( ) e
i
h
1
(4.76)
nF () e( ) e .
2
=
Employing Eq.
Gf ( )
=
=
i
h
1
=
nB () e( ) + e .
2
54
5.
Gauge fields
As with fermions in Sec. 4.2, our starting point with new kinds of fields is their classical Lagrangian
in Minkowskian spacetime, which for non-Abelian gauge fields reads
1
a
LM = F a F
,
4
a
F
= Aa Aa + gf abc Ab Ac ,
(5.1)
where g is the (bare) gauge coupling. Introducing a covariant derivative in the adjoint representation,
Dac ac + gf abc Ab ,
(5.2)
a
we note for later convenience that F
can be expressed as
a
F
= Aa Dac Ac
(= Dac Ac Aa ) .
(5.3)
We can naturally also supplement Eq. (5.1) with matter fields: for instance, letting be a fermion
in the fundamental representation, a scalar in the fundamental representation, and a scalar in
the adjoint representation, we could add
D m) + (D ) D + Dac c Dbd d V ( , a a ) ,
LM = (i
(5.4)
where D = igAa T a is a covariant derivative in the fundamental representation, and T a are

the Hermitean generators of SU(Nc ), satisfying the algebra [T a , T b ] = if abc T c and by convention
normalised as Tr [T a T b ] = ab /2.
The construction principle behind Eqs. (5.1), (5.4) is gauge invariance. With U exp[iga (x)T a ],
the Lagrangian is invariant in the transformations A A , , , , with
i
= U A U 1 + U U 1 = A + iga [T a , A ] + T a a + O(2 )
g
= Aa + Dac c + O(2 ) ,
(5.6)
= U = (1 + iga T a ) + O(2 ) ,
(5.7)
= U = (1 + iga T a ) + O(2 ) ,
(5.8)
= U U 1 = + iga [T a , ] + O(2 )
(5.9)
a
A Aa
T
Aa
a T a
a
= a + gf abc b c + O(2 ) .
(5.5)
(5.10)
We would now like to quantise the theory in Eqs. (5.1), (5.4). At this point, the role of gauge
invariance becomes rather convoluted, however. We may remember that:
The classical theory is constructed by insisting on gauge invariance.
Canonical quantization and the derivation of the Euclidean path integral necessitate an explicit breaking of gauge invariance.
The final Euclidean path integral itself again displays gauge invariance.
Yet carrying out perturbation theory with the Euclidean path integral necessitates once again
an explicit breaking of gauge invariance.
Nevertheless, only gauge invariant observables are considered to be physical.
So, conceptually, it is not a completely transparent setting!
55
In fact, as far as canonical quantization and the derivation of the Euclidean path integral are
concerned, there are two procedures followed in the literature. The idea of the most common one is
to carry out a complete gauge fixing (going to the axial gauge Aa3 = 0), identifying physical degrees
of freedom (Aa1 , Aa2 and the corresponding canonical momenta; Aa0 is expressed in terms of these
by imposing a further constraint, the Gauss law); and then following the quantization procedure
of scalar field theory.
We will here follow another approach, where the idea is to do as little gauge fixing as possible;
the price to pay is that then one has to be careful about the states over which the physical Hilbert
space is constructed6 . The advantage of the approach is that the role of gauge invariance remains
less compromised during quantization; if the evaluation of the resulting Euclidean path integral
were also to be carried out non-perturbatively (within lattice regularization, for instance), then it
would perhaps become clearer why only gauge invariant observables are physical.
5.1.
For simplicity, let us restrict to Eq. (5.1) in the following. For canonical quantization, the first
step is to construct the Hamiltonian. We will do this after setting
Aa0 0 ,
(5.11)
which fixes the gauge only partially (according to Eq. (5.6), time-independent gauge transformations are still allowed, since Aa0 remains zero in them). In some sense, the philosophy is to break
gauge invariance only to the same soft degree that Lorentz invariance is also broken in the
canonical formulation, through the special role that is given to the time coordinate.
The spatial components Aai are treated as the canonical coordinates. According to Eq. (5.3),
= 0 Aai , and Eq. (5.1) becomes
a
F0i
LM =
1
1
0 Aai 0 Aai Fija Fija .
2
4
(5.12)
The canonical momenta corresponding to Aai , to be denoted by Eia , are

Eia
LM
= 0 Aai ,
(0 Aai )
(5.13)
and the Hamiltonian density reads

H = Eia 0 Aai LM =
1 a a 1 a a
E E + Fij Fij .
2 i i
4
(5.14)
We also note that the multiplier of Aa0 (before gauge fixing) reads, according to Eq. (5.3),

Z
1
SM
b
b
bc c
b
,
(5.15)
=
(0 Ai Di A0 )F0i = Diab F0i
Aa0
Aa0 x 2
where we made use of
(x)Dab g b (x)
g a (x)Dab f b (x) .
(5.16)
The object in Eq. (5.15), the left-hand side of the Gauss law, will play an important role later on.
The theory is now canonically quantized by making Aai and Eia into operators, and by imposing
the standard bosonic equal-time commutation relations between them,
jb (t, y)] = i ab ij (x y) .
[Aai (t, x), E
(5.17)
6 This approach dates back to C.W. Bernard, Feynman Rules for Gauge Theories at Finite Temperature, Phys.
Rev. D 9 (1974) 3312, and particularly D.J. Gross, R.D. Pisarski and L.G. Yaffe, QCD and Instantons at Finite
Temperature, Rev. Mod. Phys. 53 (1981) 43.
56
The Hamiltonian becomes

=
H

1 a a 1 a a
E E + Fij Fij .
d x
2 i i
4
d
(5.18)
A very important role in the quantization will be played by what we call the Gauss law operators.
b
Combining Eq. (5.15) and F0i
= 0 Abi = Eib , we write these as
a = D
ab E
b ,
G
i
i
a = 1, . . . , Nc2 1 .
(5.19)
Furthermore we define an operator parametrized by time-independent gauge transformations,

n Z
o
exp i dd x a (x)G
a (x) .
U
(5.20)
generates gauge transformations. Let us prove this to leading non-trivial
We now claim that U
a
order in . First of all,
Z
Abj (y)U 1 = Abj (y) i dd x a (x)[G
a (x), Abj (y)] + O(2 )
U
Z
n
o
b
ia (x), Abj (y)] + gf acd Aci (x)[Eid (x), Abj (y)] + O(2 )
= Aj (y) i dd x a (x) ix [E
Z
o
n
= Abj (y) dd x a (x) ix ab ij (x y) + gf acd Aci (x) db ij (x y) + O(2 )
= Abj (y) + j b (y) + gf bca Aci (y)a (y) + O(2 )
jba a (y) + O(2 ) = Ab

= Abj (y) + D
j (y) ,
(5.21)
where in the last step we used Eq. (5.6). Similarly,

Z
E
jb (y)U 1 = E
jb (y) i dd x a (x)[G
a (x), Ejb (y)] + O(2 )
U
Z
n
o
jb (y) i dd x a (x) +gf acd[Aci (x), Ejb (y)]Eid (x) + O(2 )
= E
b (y) +
= E
j
dd x a (x)gf acd cb ij (x y)Eid (x) + O(2 )
jb (y) + gf bda E
jd (y)a (y) = E
jb (y) + O(2 ) ,
= E
(5.22)
where we have identified the transformation law of an adjoint scalar according to Eq. (5.10).
a commute with the
One important consequence of Eqs. (5.21), (5.22) is that the operators G
Hamiltonian H. This follows from the fact that the Hamiltonian in Eq. (5.18) is gauge-invariant
a transforms as an adjoint scalar):
in time-independent gauge transformations (if E
i
H
U
1 = H
[G
a , H]
=0.
U
(5.23)
transforms eigenstates as well: if Aa |Aa i = Aa |Aa i, then

Another important consequence is that U
i
i
i
i
a
a
a 1
1 |Aai i = U
1 Aa
1 Aa
Aai U
|Aai i ,
i |Ai i = U
i |Ai i = Ai U
(5.24)
where we made use of Eq. (5.21). Consequently,

1 |Aai i = |Aa
U
i i.
(5.25)
1 |physi = |physi.
Let us now define a physical state, |physi, to be one which is gauge-invariant: U
a
Expanding to first order in , we see that physical states must satisfy
a |physi = 0 .
G
57
(5.26)
This is an operator manifestation of the statement that physical states must obey the Gauss law.
a , we can choose the basis vectors of the
To summarise: since the Hamiltonian commutes with G
and G
a . Among all of these states, only the
Hilbert space to be simultaneous eigenstates of H
a
are physical. It is then only these states which are to be used in
ones with zero eigenvalue of G
for instance.
the evaluation of Z = Tr [exp( H)],
After all these preparations, we are finally in a position to derive a path integral expression for
and with a
Z. In terms of basic quantum mechanics, we have a system with a Hamiltonian H
(whose role is played by G
a ). One could consider the grand canonical
commuting quantity, Q
Q)]}
partition function Z(T, ) = Tr {exp[(H

but, according to what we have said, we are
only interested in the contribution to Z from the states with zero charge, Q|physi
= 0. Assuming
are integers, we label the states with the eigenvalues
for concreteness that the eigenvalues q of Q
q , qi = Eq |Eq , qi, Q|E
q , qi = q|Eq , qi. We can now write the relevant partition
Eq , q, so that H|E
function by taking a trace over all states, but inserting a Kronecker delta function inside the trace:
i
h
X
X
H
Zphys
.
(5.27)
hE0 , 0|eE0 |E0 , 0i =
hEq , q|q,0 eEq |Eq , qi = Tr Q,

0e
E0
Eq ,q

Since Q,

0 = Q,
0 Q,
0 and [H, Q] = 0, we can finally write
i
h
H
H
H
e
e
.
.
.
,
Zphys = Tr Q,
0
Q,0
Q,0
|
{z
}
(5.28)
N parts
where = /N , N as before. We may represent

Q,

0 =
di ii Q
=
e
2
dci ici Q
,
e
1
/ 2
(5.29)
and insert unity operators as in Eq. (1.34), but now the momentum state representation is placed
between Q,

0 and exp(H). The typical building block then reads
x)
|pi ihpi |eH(p,
|xi i
hxi+1 |eici Q(x,p)

xi+1 xi
p2
+ V (xi ) + O()
.
= exp ici Q(xi+1 , pi ) + i ipi
2m
(5.30)
It remains to: (i) take the limit 0, whereby xi , pi , ci become functions, x( ), p( ), c( ); (ii)
go from d = 0 to a general dimension; (iii) replace x( ) Aai (x), p( ) Eia (x), c( ) Aa0 (x),
Q Diab Eib , m 1. Then the integral over the square brackets in Eq. (5.30) becomes
Z

Z

1
1
1 a a
1
Ei Ei iEia Aai Diab Ab0 + Fija Fija ,
iAa0 Diab Eib + Eia Eia iEia Aai + Fija Fija =
2
4
4
x 2
x
(5.31)
where we made use of Eq. (5.16).
At this point we recognize inside the round brackets in Eq. (5.31) an expression of the form in
Eq. (5.3). Of course, the field Aa0 is not the original Aa0 -field, which was set to zero. Rather, it is
a new field, which we are however free to rename to be Aa0 . Indeed, in the following we leave out
the tilde from Aa0 , and redefine a Euclidean field strength tensor according to
a
F0i
Aai Diab Ab0 .
(5.32)
1
1 a a
1 a a
a
a 2
Ei Ei iEia F0i
= (Eia iF0i
) + F0i
F0i ,
2
2
2
(5.33)
Noting furthermore that
58
we can carry out the Gaussian intergral over Eia , and end up with the desired path integral
expression:

Z Z
Z
Z
1 a a
Zphys = C DAa0
DAai exp
d
dd x LE , LE = F
F .
(5.34)
a
a
4
Ai (,x)=Ai (0,x)
0
V
Two remarks are in order on the final result in Eq. (5.34):
(i) The field Aa0 was introduce in order to impose the Gauss law at every value of ; therefore,
the integrations at each are independent of each other. In other words, it is not obvious
from the derivation of the path integral whether the field Aa0 should satisfy periodic boundary
conditions like the spatial components Aai do.
Now, any field defined on a compact interval (0, ) can be expressed as a sum of a
periodic and an anti-periodic function: Aa0 ( ) = 12 [Aa0 ( )+Aa0 ( )]+ 21 [Aa0 ( )Aa0 ( )].
The periodic part is what we would expect, and the question then is, what happens with
the antiperiodic part? Note that this part could be Fourier-decomposed with fermionic
Matsubara frequencies.
For the moment the answer does not seem obvious, but we will see in Exercise 7 that indeed
only a periodic Aa0 ( ) leads to physical results.
(ii) For scalar field theory and fermions, Eqs. (2.7), (4.34), we found after a careful derivation of
the Euclidean path integral that the result could be interpreted in terms of a simple recipe:
LE = LM (t i ). We may now ask whether the same is true for gauge fields as well?
A comparison of Eqs. (5.1), (5.34) shows that, indeed, the recipe again works; the only
complication is that the Minkowskian Aa0 needs to be replaced with iAa0 (of which we have
normally left out the tilde), just like t gets replaced with i . This, of course, should be
expected from gauge invariance, since covariant derivatives need to change as Dt iD .
59
5.2.
Gauge fixing and ghosts
Eq. (5.34) is gauge invariant and could be evaluated as such in lattice regularization, for instance.
As before, we will here restrict to perturbation theory, however. Then gauge invariance needs once
again to be broken, because the quadratic part of LE does otherwise not contain an invertible
matrix M , so that no propagators can be defined. For completeness, let us briefly recall the main
steps of this procedure.
Let Ga be some function(al)s (no longer the Gauss law; the notation has changed!) of the path
integration variables in Eq. (5.34), for instance Ga (x) Aa3 (x) or Ga (x) = Aa (x). The idea is
to insert the object
h Ga (x) i
Y
(Ga ) det
(5.35)
b (y)
x,y,a,b
as a multiplier in front of the exponential in Eq. (5.34), in order to remove the infinities from the
integrations over the gauge orbits. Indeed, it is easy to see that this insertion does not change
the value of gauge invariant expectation values (of course, in Z it induces an overall constant,
analogous to C). First of all, since LE is gauge invariant, its value does not depend (within each
set of gauge-equivalent configurations) of the particular value selected by the constraint Ga = 0.
Second, let us inspect the integration measure. We can imagine dividing the integration into one
over gauge non-equivalent fields, A , and gauge transformations thereof, parametrized by . Then
Z
n Z
o
h Ga i
exp
L
(A
)
=
DA (Ga ) det
E
b
x
Z
Z
h Ga i
o
n Z
b
a
)
=
DA D (G ) det
L
(
A
exp
E
b
x
Z
Z
o
n Z
a
a
=
DA DG (G ) exp LE (A )
x
Z
o
n Z
(5.36)
=
DA exp LE (A ) .
x
In other words, the dependence on the particular choice of the functions Ga disappears.
It is perhaps appropriate to point out that the somewhat formal manipulations in Eq. (5.36) can
be given a precise meaning in lattice regularization, where the integration measure is defined as
the gauge invariant Haar measure on SU(Nc ). Nevertheless, the message remains the same as with
our simplistic argument.
Given that the outcome is independent of Ga , it is conventional and convenient to furthermore
replace (Ga ) by (Ga f a ), where f a is some Aa -independent function, and then to take an
average over the f a s with a Gaussian weight:

Z
Z
1
a a
a
a
a
a
f f
(G )
Df (G f ) exp
2 x

Z
1
= exp
(5.37)
Ga Ga .
2 x
Here an arbitrary parameter, , has been inserted, in order to allow for a check later on that the
results indeed are independent of its value.
Finally, the other structure in Eq. (5.35), the determinant, is conventionally written in terms of
Faddeev-Popov ghosts, by making use of Eq. (4.46):
Z

det(M ) = D
c Dc exp
cM c .
(5.38)
60
It should be noted, however, that since the matrix Ga /b is purely bosonic, the ghost fields have
to obey the same boundary conditions as the gauge fields, i.e. periodic, in spite of their Grassmann
nature.
In total, then, we can write the gauge-fixed version of Eq. (5.34), adding now also Dirac fermions
to complete the theory into QCD, as
Z
Z
Z
DAa
D
ca Dca
Zphys = C
D D
periodic
Z
exp
periodic
anti-periodic
Ga
1 a a
1
D
+ m)
dd x
d
cb + (
F F + Ga Ga + ca
4
2
b
V

(5.39)
where m is the diagonal quark mass matrix.
A particularly convenient gauge choice is that of covariant gauges. Then
Ga
Aa ,
1
Aa Aa ,
2
a
i
h

+ b = ab + gf acb Ac ,
1 a a
G G
2
Ga
b
Ga
cb
ca
b
=
=
ca ca + gf abc ca Ab cc ,
(5.40)
(5.41)
(5.42)
(5.43)
where we made use of Eqs. (5.2), (5.6).

5.3.
Feynman rules for Euclidean continuum QCD
For completeness, we now collect together the Feynman rules that apply to computations with the
theory in Eq. (5.39), when the gauge is fixed according to Eq. (5.40).
Consider first the free (quadratic) part of the Euclidean action. Expressing everything in Fourier
representation, this becomes

Z
i
P
1 a h a
1 a a
a
SE =
iP A (P ) iQ A (Q) iQ A (Q) + iP A (P )iQ A (Q) +
(P + Q)
2
2
Pb ,Q
b
Z
Z
i
h
P
+ m]A (Q)
ca (Q)
+ P (P + Q)
A (P )[i
iPca (P )iQ
Q
+
(P + Q)
P ,Q
P ,Q
b
=
+
f f

i

h

1
a
a
2
P 1 1 P P
+
A (P )A (Q)
(P + Q)
2
P ,Q
Z
Z b b
i P
h
P
a a 2
.
A (P )[iP/ + m]A (Q)
c (Q)P +
(P + Q)
(P + Q) c (P )
Z b
P
(5.44)
Pf ,Q
f
Pb ,Q
b
For the quarks, the index A is assumed to comprise both colour and flavour indices, while in the
Dirac space and are treated as vectors.
The propagators are obtained by inverting the matrices in Eq. (5.44):
D
Aa (P )Ab (Q)
D
E
b
ca (P )c (Q)
=
=

P P
P P
P 2
P 2
,
+
(P + Q)
P 2
P 2
1 ,
ab (P Q)
P 2
ab
61
(5.45)
(5.46)
A (P )B (Q)
iP/ + m .
AB (P Q)
P 2 + m2
(5.47)
Finally, we list the interactions. It is convenient to symmetrize these through changes of integration and summation variables as much as possible. Thereby the three-gluon vertex becomes
Z
1
(AAA)
SI
=
( Aa Aa )gf abc Ab Ac
x 2
Z
P
1 a b c
+ R)

=
A (P )A (Q)A (R) (P + Q
3!
Pb ,Qb ,Rb
h
i
) + (R
Q
) + (Q
P ) . (5.48)
igf abc (P R
The four-gluon vertex reads
Z
1 4 abc ade b c d e
(AAAA)
SI
=
g f f A A A A
x 4
Z
P
1 a b c d
+R
+ S)

A (P )A (Q)A (R)A (S) (P + Q
=
4!
,R
,S
Pb ,Q
h b b b
i
eab ecd
2
g f f ( ) + f eac f ebd ( ) + f ead f ebc ( ) .
(5.49)
The ghost interaction can be written as

Z
(
cAc)
SI
=
ca gf abc Ab cc
x
Z

P
cc (R)
(P + Q
+ R)
igf abcP .
ca (P )Ab (Q)
=
(5.50)
,R
Pb ,Q
b b
Finally, the fermion interaction is contained in

Z

(A)
a
Aa B
A igTAB
=
SI
x
Z

P
a
B (R)
(P + Q
+ R)
igTAB
.
=
A (P )
Aa (Q)
,R
Pf ,Q
b f
62
(5.51)
5.4.
Exercise 7
(a) Compute the free energy density f (T ) for free gluons (Nc colours) and massless quarks (Nf
flavours), starting from Eq. (5.44).
(b) Deduce from here the result for usual electromagnetic blackbody radiation.
Note that ghosts play a role in both cases!
(a) We remember from Eqs. (2.51), (2.80), (4.59), (4.72) that
Z
i
1 P h 2
2 T 4
Jb (0, T ) =
,
ln(P ) const. =
2
90
Pb
Z
i 7 2 T 4
1 P h 2
ln(P ) const. =
Jf (0, T ) =
.
2
8 90
(5.52)
(5.53)
Pf
We just need to figure out the prefactors of these terms. There will be contributions from
gluons, ghosts and quarks, and we inspect them one at a time.
In the gluonic case, we are faced with the matrix

1
P P .
M = P 2 1
(5.54)
Let us introduce two further matrices,

T
P

P P
,
P 2
L
P
P P
.
P 2
(5.55)
As matrices, these satisfy P T P T = P T , P L P L = P L , P T P L = 0, P T + P L = 1. Thereby

P T , P L are projection operators, which implies that their eigenvalues are either zero or unity.
The numbers of the unit eigenvalues can be found by taking the traces from the matrices:
Tr [P T ] = 1 = d, Tr [P L ] = 1.
We can now write
1
L
T
.
M = P 2 P
+ P 2 P
(5.56)
Thereby M has d eigenvalues P 2 and one P 2 /. Also, there are a = 1, . . . , Nc2 1 copies of
this structure. In total, then,

Z
Z h
i 1P
i
1
1 P h 2
ln(P ) const. +
ln( P 2 ) const.
f (T )|gluons = (Nc2 1) d
2
2
Pb
P

b
Z
1P
= (Nc2 1)
ln() + (d + 1)Jb (0, T ) .
(5.57)
2
Pb
Furthermore, the first term in Eq. (5.57) vanishes in dimensional regularization, because it
does not contain any scales.
For ghosts, the Gaussian integral yields (cf. Eq. (4.46))
Z Y
io
n h
Y
1
a a
a
dc d
c exp(c P 2 ca ) =
P 2 = exp 2(Nc2 1) ln(P 2 ) .
2
a
a
63
(5.58)
Furthermore we have to remember that ghosts obey periodic boundary conditions. Thereby
f (T )|ghosts = 2(Nc2 1)Jb (0, T ) .
(5.59)
Finally, quarks work out as in Eq. (4.51), except that they now come in Nc colours and Nf
flavours:
f (T )|quarks = 4NcNf Jf (0, T ) .
(5.60)
Summing together Eqs. (5.57), (5.59), (5.60), inserting the values from Eqs. (5.52), (5.53),
and setting d = 3, we get
f (T )|QCD =
i
2 T 4 h
7
2(Nc2 1) + Nf Nc .
90
2
(5.61)
This result is often referred to as the (QCD-version of the) Stefan-Boltzmann law.

It is important to realize (i) that the contribution from the ghosts is essential: according to
Eq. (5.59), it cancels half of the result in Eq. (5.57), thereby yielding the correct number of
physical degrees of freedom as a multiplier in Eq. (5.61); (ii) that the assumption that Aa0 has
only periodic modes also played a role; had it also had antiperiodic ones, Eq. (5.61) would
have obtained a further unphysical term (to be more precise, this statement assumes that
the ghosts remain periodic; it might be possible to compensate for an antiperiodic part of
Aa0 through an antiperiodic part in the ghost determinant, but the setup would then become
rather complicated).
(b) The case of QED is obtained by setting Nc 1, and Nc2 1 1 [since the group is U(1)
rather than SU(1)]:
7 i
2 T 4 h
2 + Nf .
(5.62)
f (T )|QED =
90
2
The factor 2 corresponds to the two photon polarizations; the factor 4, multiplying 87 Nf ,
corresponds to a spin- 21 particle and a spin- 21 antiparticle. If neutrinos were included, they
would only contribute with a factor 2 78 Nf .
64
5.5.
Thermal gluon mass
Consider the propagator of the Matsubara zero modes of gauge fields. We wish to see whether an
effective thermal mass meff is generated for them, like for the scalar fields (cf. Eq. (3.102)). (Note
that we do not need to consider non-zero Matsubara modes since, like in Eq. (3.101), the thermal
mass corrections are subdominant in this case, g 2 T 2 (2T )2 . For the same reason, we do not
need to consider thermal mass correction for fermions at the present order.) The observable to
consider is the full propagator, the analogue of Eq. (3.71).
In order to simplify the task somewhat, we choose to carry out the computation in the so-called
Feynman gauge, 1. Then the free propagator from Eq. (5.45) becomes
D
E
Aa (P )Ab (Q)

P P
P P
P 2
P 2
+
(P + Q)
P 2
P 2
ab
=1
.
ab (P + Q)
P 2
(5.63)
At 1-loop level, in QCD, there are corrections to Eq. (5.63) from two graphs involving gauge
loops (via a quartic vertex in SI and via cubic vertices in +SI2 /2); from one graph involving
ghosts (via cubic vertices in +SI2 /2); and from one graph involving fermions (via cubic vertices
in +SI2 /2). (If the theory contains scalar fields, there are two graphs like from gauge loops; for
simplicity, we omit this contribution in the following.)
For future reference, we will treat P as a general Euclidean four-momentum for the moment,
even though for the Matsubara zero modes only the spatial part is non-zero.
Let us start by considering the gauge loop via a quartic vertex. Denoting the structure in
Eq. (5.49) by
cdef
C
f gcd f gef ( ) + f gce f gdf ( ) + f gcf f gde ( ) ,
(5.64)
we get
D
E
)
Aa (P )Ab (Q)(S
I
0,c
Z
E
2D
g a b P
Ad (S)
Ae (T)Af (U
) (R
+ S + T + U)
C cdef
=
Ac (R)
A (P )A (Q)
24
0,c
,S

R
b b ,Tb ,Ub
Z
g2 P
+ S + T + U)h
Aa (P )Ac (R)i
0 hAb (Q)
Ad (S)i
0 hAe (T)Af (U
)i0 C cdef ,
=
(R
2
Rb ,Sb ,Tb ,Ub
(5.65)
cdef
where we made use of the complete symmetry of C
. Inserting Eq. (5.63), we get
Aa (P )Ab (Q)(S
I)
0,c
Z
g2 P
+ S + T + U
) (P + R)
(Q
+ S)
(T + U)

(R
,T ,U
Rb ,S
b b b
1
cdef
ac bd ef C
.
2
P Q2 T2
(5.66)
S,
U
are trivially carried out. Moreover, we note that
The sum-integrals over R,
h
i
cdef
ac bd ef C
= ef f gae f gbf ( ) + f gaf f gbe ( )
=
2 d f age f bge ,
(5.67)
where we made use of the antisymmetry of the structure constants, and the fact that = d + 1
(which in dimensional regularization equals 4 2). The structure constants satisfy f age f bge =
65
Nc ab , and in total we then get

D
E
Aa (P )Ab (Q)(S
I)
0,c
= g 2 Nc d ITb (0) ab (P + Q)
.
(P 2 )2
(5.68)
Note that the structure in the colour and spacetime indices is the same as in Eq. (5.63).
We then move on to the gluon loop originating from the cubic interactions. Denoting the combination in Eq. (5.48) by
S,
T) (R
T ) + (T S ) + (S R
) ,
D (R,
(5.69)
we get
E
D

1 S2
Aa (P )Ab (Q)
I
2
0,c
Z
Z
g 2 D a b P c d e P g h i E
=
A (U )A (V )A (X)
A (P )A (Q) A (R)A (S)A (T )

72
0,c

Ub ,Vb ,Xb
Rb ,Sb ,Tb
f
=
cde ghi
+ S + T) (U
+ V + X)
D (R,
S,
T) D (U
, V , X)
(R
Z
2P
0 hAb (Q)
Ag (U
)i0 hAd (S)
Ah (V )i0 hAe (T)Ai (X)i
0
hAa (P )Ac (R)i
g
2
,S

R
b b ,Tb ,Ub ,Vb ,Xb
cde ghi

+ V + X)
D (R,
S,
T) D (U
, V , X)
,
(R + S + T) (U
(5.70)
S,
T) in simultaneous interchanges
where we made use of the complete symmetry of f cdeD (R,
d, , S).
of all indices labelling a particular gauge field (for instance c, , R

Inserting Eq. (5.63), let us inspect in turn colour indices, spacetime indices, and momenta. The
colour contractions result in a factor
ac bg dh ei f cde f ghi = f ade f bde = Nc ab .
(5.71)
The spacetime indices can be transported to the D-functions: , , , . The

momentum dependence can be written as
(Q
+ U)
(S + V ) (T + X)
(P + R)
2 S2 T2
P 2 Q
+ S + T) (U
+ V + X)
D (R,
S,
T) D (U
, V , X)
(R
(Q
+ U)
(S + V ) (T + X)
(P + R)
=
2
2
2
2
P Q S T
S T) D (P , S,
T) D (Q,
S,
T)
(P + S + T) (Q
(Q
+ U)
(S + V ) (T + X)
(P + R)
=
2 S2 T2
P 2 Q
D (P , S,
T) D (P , S,
T)
(P + S + T) (P + Q)
U
, V , X
and T. Thereby
We can now integrate over R,

D
E
1 SI2
Aa (P )Ab (Q)
2
0,c
Z
2
ab
P
g Nc (P + Q)
1
P S)
D (P , S,
P + S)
.
=
D (P , S,
2
2
2
2
S
(
P
S)
(P )
(5.72)
(5.73)
Finally, we are faced with the tedious task of inserting Eq. (5.69) and carrying out the contractions:
P S)
D (P , S,
P + S)
D (P , S,
66
= [ (2P + S ) + (P 2S ) + (P + S )]
[ (2P + S ) + (P 2S ) + (P + S )]
= (4P 2 4P S + S2 + P 2 + 2P S + S2 ) (d + 1)(P P 2P S 2P S + 4S S )
[(2P + S )(P 2S ) + (2P + S )(P + S ) + (P 2S )(P + S ) + ( )]
2 + S2 ] (d 5)P P + (2d 1)(P S + P S ) (4d 2)S S .
= [4P 2 + (P S)
(5.74)
Inserting Eq. (5.74) into Eq. (5.73), we observe that the result depends in a non-trivial way on the
external momentum P . This is an important fact which will play a role later on. Nevertheless,
for the moment we may note that since the tree-level gluon propagator, Eq. (5.63), is massless, the
pole position lies at P 2 = 0. This pole position may get shifted by the correction in Eq. (5.73), like
happened in scalar field theory (cf. Eq. (3.102)), but since the correction is small (suppressed by
the coupling), we may in fact insert P 2 = 0 inside the already small loop correction in Eq. (5.74),
thereby only making an error of O(g 4 ). Proceeding this way, we get
Z
D

i
E
P
1 h 2
g 2 Nc ab (P + Q)
1 SI2
Aa (P )Ab (Q)
(5.75)
2S + (4d 2)S S .
2 2
2
2
0,c
(P 2 )2
(S )
S
b
Now, symmetries tell that the latter integral in Eq. (5.75) must be proportional to , just like
the first one. However, its value could still be different for = = 0 and = = i. This is
because the heat bath constitutes a preferred frame, and thus breaks Lorentz invariance. In fact,
we can write
Z
Z
Z
P S02
P Si2
P S S
= 0 0
+ i i
2 2
2 2
2 2
(S )
(S )
(S )
S
S
S
b
b
b
Z
Z
P S02
1 P s2
+ i i
= 0 0
2 2
d (S2 )2
(S )
Sb
S
b
Z
Z 2 2
P S02
P
S S0
1
= 0 0
+ i i
.
(5.76)
2 )2
d
(
S
(S2 )2
S
S
b
At this point, let us inspect the sum-integral

Z
Z
X
P 1
T2
1
=T
=
+ O() .
ITb (0) =
2
2
2
(2nT ) + s
12
S
n= s
(5.77)
Sb
Taking the derivative T 2 d/dT 2 = T /2 d/dT on both sides, we find

Z
Z
X
X
T2
(2nT )2
1
1
T
=
T
+ O() ,
2 n= s (2nT )2 + s2
[(2nT )2 + s2 ]2
12
n= s
which can be used to solve for the unknown sum-integral,
Z
P S02
1
= ITb (0) + O() .
2
2
2
(S )
Sb
Inserting Eqs. (5.76), (5.79) into Eq. (5.75) finally yields

E
D

g 2 Nc ab (P + Q)
1
1 S2
(4d
2)
+
Aa (P )Ab (Q)
0 0
2 I 0,c
2
2
(P 2 )2
67
(5.78)
(5.79)

4d 2 3
ITb (0)
+i i 2 +
d
2
2

T
g 2 Nc ab (P + Q)
3 0 0 + 7 i i
,
2
2
2
12
(P )
(5.80)
where we inserted d = 3 in the last step, like already in Eq. (5.79).

Consider next the ghost loop; the vertex is in Eq. (5.50). Proceeding as in Eq. (5.70), we obtain
D
E

1 S2
Aa (P )Ab (Q)
2 I 0,c
Z
Z
2D
g a b P c d e P g h i E
c (U )A (V )
c (X)
c (T )
A (P )A (Q) c (R)A (S)
2
0,c

Ub ,Vb ,Xb
Rb ,Sb ,Tb
f
Z
P
2
cde ghi
+ S + T) (U
+ V + X)
R
U
(R
g
0 hAb (Q)
Ah (V )i0 h
cc (R)i
0
hAa (P )Ad (S)i
ce (T)c (U
)i0 h
ci (X)
,S

R
b b ,Tb ,Ub ,Vb ,Xb
cde ghi
,
+ V + X)
R
U
(R + S + T) (U
(5.81)
where the Grassmann nature of the ghosts induced a minus sign.

Inserting the gluon propagator from Eq. (5.63) and the ghost propagator from Eq. (5.46), let us
inspect in turn colour indices, spacetime indices, and momenta. The colour contractions result in
a factor
ad bh eg ic f cdef ghi = f cae f ebc = Nc ab .
(5.82)
. The momentum
= R
U
U
The spacetime indices can be transported to the momenta: R
dependence can be written as
(Q
+ V ) (T U)
(X
R)
(P + S)
+ S + T) (U
+ V + X)
R
U
(R
2
2 T2 X
P 2 Q
(Q
+ V ) (T U)
(X
R)
(P + S)
P + T) (T Q
+ X)
X
T
(X
2
2
2
2
P Q T X
(Q
+ V ) (T U)
(X
R)
(P + S)
P + T) (P + Q)
(T P )T . (5.83)
=
(X
2 T 2 (T P )2
P 2 Q
V , U
, R
and X.
Thereby
We can now integrate over S,
Z
D
1 E
P
ab (P + Q)
1
2
2
a b
(S P )S ,
= g Nc
S
A (P )A (Q)
2
2
2
2 I 0,c
S
(
P
S)
(P )
(5.84)
where we renamed T S.
The structure in Eq. (5.84) is identical to some of the terms in Eq. (5.74). In particular, if we
again set the external momentum to zero, we return back to the integral in Eq. (5.76). Putting
furthermore d 3, we arrive at
2
D

E
1
ab (P + Q)
T
1
2
1 S2
Aa (P )Ab (Q)
g
N
.
(5.85)
c
0
0
i
i
I
2
2
2
2
2
12
0,c
(P )
Finally we consider the fermion loop; the vertex is in Eq. (5.51). Proceeding as in Eq. (5.70), we
obtain
E
D

1 SI2
Aa (P )Ab (Q)
2
0,c
68
Z
Z
E
P
g 2 D a b P
B (T) C (U
A (P )A (Q) A (R)
Ac (S)
Ad (V )D (X)
2
0,c

Rf ,Sb ,Tf
Uf ,Vb ,Xf
+ S + T) (U
+ V + X)T
c Td
(R
AB CD
Z
i
h
a
c
b
d
2P
)i0
0 hB (T) (U
0 hA (Q)
A (V )i0 Tr hD (X)
(R)i
hA (P )A (S)i
g
C
A
,S

R
f b ,Tf ,Uf ,Vb ,Xf
+ S + T) (U
+ V + X)T
c Td ,
(R
AB CD
(5.86)
where the Grassmann nature of the fermions induced a minus sign. The capital indices include
both colour and flavour.
Inserting the gluon propagator from Eq. (5.63) and the fermion propagator from Eq. (5.47), let
us inspect in turn colour + flavour indices, the Lorentz indices, and momenta. The colour + flavour
contractions result in a factor
c
d
TCD
= Tr [T a T b ] =
ac bd DA BC TAB
Nf
.
2
(5.87)
(For simplicity, we assumed that the flavours are degenerate in mass.) The spacetime indices yield
( + ) + m2 ]
U
Tr [(iR
/ + m)
(iU
/ + m)
] = 4[R
] .
R
U
U
+ m2 ) R
U
(5.88)
= 4[ (R
The momentum dependence can be written as
(Q
+ V ) (X
R)
(T U)
(P + S)
+ S + T) (U
+ V + X)
f (R,
U
)
(R
2 (X
2 + m2 )(T 2 + m2 )
P 2 Q
(Q
+ V ) (X
R)
(T U)
(P + S)
P + T) (T Q
+ X)
f (X,
T)
=
(X
2 (X
2 + m2 )(T 2 + m2 )
P 2 Q
(Q
+ V ) (X
R)
(T U)
(P + S)
P + T) (P + Q)
f (T P , T) . (5.89)
(X
2
2
2
2
2
2
P Q [T + m ][(T P ) + m ]
V , R,
U
and X.
Thereby
We can now integrate over S,
E
D

1 SI2
Aa (P )Ab (Q)
2
0,c
Z
ab
(P + Q) P (S2 P S + m2 ) 2S S + P S + P S
= 2g 2 Nf
,
2 + m2 ]
(P 2 )2
[S2 + m2 ][(P S)
(5.90)
The structure in Eq. (5.90) is again identical to some of the terms in Eq. (5.74), except that
the Matsubara frequencies are fermionic. In particular, if we again set the external momentum to
zero, and also consider the limit T m, so that quark masses can be ignored, we get
Z
P S2 2S S
(S2 )2
S
f
Z
Z
P S2 2S02
P S2 2Si2
+ i i
(S2 )2
(S2 )2
S
S
f
f
Z
Z
P S2 2S02
P S2 (2/d)s2
= 0 0
+ i i
(S2 )2
(S2 )2
S
S
f
f
Z
Z
P (1 2/d)S2 + (2/d)S02
P S2 2S02
+ i i
.
= 0 0
(S2 )2
(S2 )2
S
S
= 0 0
69
(5.91)
The relation in Eq. (5.79) continues to hold in the fermionic case; setting d 3, we thereby get
Z
1
P S2 2S S
1 f
I (0) .
= 0 0 2ITf (0) + i i
2 )2
3
3 T
(
S
(5.92)
The breaking of Lorentz invariance by the finite temperature is quite explicit here. Inserting finally
ITf (0) = T 2 /24 (cf. Eq. (4.74)), we arrive at
D
2
E

T
ab (P + Q)
2
1 S2
g
N
+
0
.
Aa (P )Ab (Q)
f
0 0
i i
I
2
2
2
6
0,c
(P )
(5.93)
Summing the contributions from Eqs. (5.68), (5.80), (5.85), (5.93), we get
D
E

SI + 1 SI2
Aa (P )Ab (Q)
2
0,c
2

ab
T
7
3
1
1
(P + Q)
Nc + 2Nf 0 0 + 3 +
Nc i i
3+
g 2
2
2
2
2
12
(P 2 )2

ab
Nc
Nf
(P + Q)
.
(5.94)
0 0 g 2 T 2
+
=
2
2
3
6
(P )
It is very important to realise that all corrections have cancelled from the spatial part.
The result obtained has a direct physical meaning. Indeed, we recall from the discussion of scalar
field theory, Eq. (3.78), that Eqs. (5.63), (5.94) can be interpreted as a (resummed) full propagator,
D
E
ab
(P + Q) ,
Aa (P )Ab (Q)
P 2 + 0 0 m2
(5.95)
where
m2E
g T
Nf
Nc
+
3
6
(5.96)
is called the Debye mass parameter. Its existence corresponds to the fact the colour-electric fields
( A0 ) get screened in a thermal plasma. In contrast, colour-magnetic fields ( Ai ) do not get
screened at least not at this order! We will return to the physical significance of these effects
later on.
70
5.6.
Free energy density to O(g 3 )
As an immediate application of the results of the previous section, we compute the free energy
density of QCD to O(g 3 ), parallelling the method introduced for scalar field theory in Exercise 5.
We recall that the essential insight is to correct the quadratic part of the Lagrangian of the
Matsubara zero modes by the effective thermal mass computed from the full propagator, and to
treat minus the same term as part of the interaction Lagrangian. The non-interacting free energy
density computed with the corrected propagator then yields the result for the ring sum, while the
bilinear interaction term cancels the corresponding contributions order by order from multiloop
graphs. In the present case, given the result in Eq. (5.95), only the zero-components of the gauge
fields need to be corrected.
The correction of O(g 3 ) to the tree-level result7 in Eq. (5.61) can therefore be immediately
written down, by employing Eq. (2.86) and taking into account that there are Nc2 1 gauge field
components:

T m3E
f(3/2) (T )QCD = (Nc2 1)
12

3
1
Nc
Nf 2
2
4 3
= (Nc 1)T g
+
12
3
6
3

2 4
T
Nc
Nf 2
2
3 15
=
(Nc 1)g
+
90
2 3 3
6
3
2 32
2 4
T
Nc
Nf 2
g
2
=
,
(5.97)
60(Nc 1)
+
90
4 2
3
6
where the effective mass mE was taken from Eq. (5.96).
Next, we need to consider the contributions of O(g 2 ). In analogy with Eq. (3.106), these terms8
come from the non-zero mode contributions to 2-loop graphs:
E
D

1
(5.98)
f(1) (T )QCD = SI SI2 + . . .
R .
2
At this point, it is useful to compare Eq. (5.98) with the computation of the full propagator in the
previous section, Eq. (5.94). We note that, apart from an overall minus sign, the two computations
are quite similar; the main difference is in a combinatorial factor. In fact, the claim is that we only
need to close the gluon line in the results of the previous section, and simultaneously divide the
graph by 1/2n, where n is the number of gluon lines in the vacuum graph. Let us prove this claim
by direct inspection.
Consider first the contribution from the 4-gluon vertex. In vacuum graphs, this leads to the
combinatorial factor
hA A A Ai0,c = 3hA Ai0 hA Ai0 ,
(5.99)
while in the propagator we got
hA A A A A Ai0,c = 4 3hA Ai0 hA Ai0 hA Ai0 ,
(5.100)
so there is indeed a difference of 4 = 2 n, with n = 2 being the number of contractions in

Eq. (5.99). Similarly, with the contribution from two 3-gluon vertices, the vacuum graphs lead to
the combinatorial factor
hA A A A A Ai0,c = 3 2hA Ai0 hA Ai0 hA Ai0 ,
7 J.I.
(5.101)
Kapusta, Quantum Chromodynamics at High Temperature, Nucl. Phys. B 148 (1979) 461.
Shuryak, Theory of Hadronic Plasma, Sov. Phys. JETP 47 (1978) 212; S.A. Chin, Transition to Hot Quark
Matter in Relativistic Heavy Ion Collision, Phys. Lett. B 78 (1978) 552.
8 E.V.
71

hA A A A A A A Ai0,c = 6 3 2hA Ai0 hA Ai0 hA Ai0 hA Ai0 ,
(5.102)
so there is indeed a difference of 6 = 2n, with n = 3 the number of contractions in Eq. (5.101).
Finally, the ghost and fermion contributions to the vacuum graphs lead to the combinatorial factor
h
c A c c A ci0,c = hA Ai0 hc ci0 hc ci0 ,
(5.103)

hA A c A c c A ci0,c = 2hA Ai0 hA Ai0 hc ci0 hc ci0 ,
(5.104)
so there is indeed a difference of 2 = 2n, with n = 1 the number of contractions in Eq. (5.103).
The contribution of the 4-gluon vertex can hence be extracted from Eq. (5.68):

Z
P aa
1
=
g 2 Nc d ITb (0)
4
P 2
P
b
g2
Nc (Nc2 1)d(d + 1)[ITb (0)]2 .
4
(5.105)
The contribution of the 3-gluon vertices can be extracted from Eq. (5.73); noting from Eq. (5.74)
that
P S)
D (P , S,
P + S)
D (P , S,
2
2
2
+ S ] (d 5)P 2 + 2(2d 1)P S (4d 2)S2 .
= (d + 1)[4P + (P S)
= {P 2 [5d + 5 + d 5] + P S[2d
2 4d + 2] + S2 [2d + 2 + 4d 2]}
2 + S2 } ,
= 3d{P 2 + (P S)
(5.106)
we get

Z
Z
2 + S2
3g 2 Nc P aa P P 2 + (P S)
d
2
2
P 2 S
S2 (P S)
P
g2
Nc (Nc2 1)d 3[ITb (0)]2 .
4
(5.107)
Note that here it is important to keep the full P -dependence, unlike in Eq. (5.75), because all
values of P are integrated over.
Similarly, the contribution of the ghost loop can be extracted from Eq. (5.84):

Z
Z
P aa P S2 P S
g 2 Nc
2
P 2 S S2 (P S)
Pb
b
Z
2 P 2
P
g2
S2 + (P S)
Nc (Nc2 1)
2
2
2
4
P S (P S)
1
2
Pb ,S
b
g
Nc (Nc2 1)[ITb (0)]2 .
4
(5.108)
Finally, the contribution of the fermion loop can be extracted from Eq. (5.90):
=
1
2
2g 2 Nf

Z
Z
P aa P (d + 1)(S2 P S + m2 ) 2S2 + 2P S
.
2 + m2 ]
[S2 + m2 ][(P S)
P 2
Pb
S
f
72
(5.109)
We again simplify by setting m/T 0, and then get

=
Z
P
g 2 Nf (Nc2 1)
Pb ,S
f
g 2 Nf (Nc2 1)
(d 1)(S2 P S)
2
P 2 S2 (P S)
Z
d 1P
2
Pb ,S
f
2 P 2
S2 + (P S)
2
P 2 S2 (P S)
g
Nf (Nc2 1)(d 1){2ITb (0)ITf (0) [ITf (0)]2 } .
2
(5.110)
In the last step, careful attention needs to be paid to the nature of Matsubara frequencies appearing
in the propagators.
Adding together the terms from Eqs. (5.105), (5.107), (5.108), (5.110); setting d = 3 (there are
no divergences); and using ITb (0) = T 2 /12, ITf (0) = T 2 /24, we get

9
1
1
1
T4
3 +
Nc 2
Nf
f(1) (T )QCD = g 2 (Nc2 1)
144
4
4
2
4

T4
5
= g 2 (Nc2 1)
Nc + Nf
144
4

2 4
T
5
5 g2
=
.
(5.111)
N
+
(Nc2 1)
N
c
f
90
2 4 2
4
Combining finally the effects from Eqs. (5.61), (5.111), and (5.97), we get

2 T 4 2
5
s
7 Nf Nc
5
f (T )|QCD =
Nc + Nf
(Nc 1) 1 +
+
45
4 Nc2 1 4
4

3 3
Nc
Nf 2 s 2
+ O(2s ) ,
+30
+
3
6
(5.112)
were we have denoted s g 2 /4.

A few remarks are in order:
The result in Eq. (5.112) can be compared with that for scalar field theory in Eq. (3.100).
The general structure is seen to be identical. In particular, in both cases, the first relative
correction is negative. This means that the interactions between the particles in a plasma
tend to decrease the pressure that the plasma exerts on the walls of an imaginary container.
The second correction is positive, however. Such an alternating structure, even though in
principle convergent (in the usual asymptotic sense) for small enough renormalized coupling,
indicates that it may be difficult to quantitatively estimate the magnitude of radiative corrections to the non-interacting result. [But recall: 1 21 + 13 14 ... = ln 2 = 0.693...,
1 + 21 + 31 + 41 ... = ; maybe an alternating structure is good after all!]
As of today (fall 2007), the coefficients of the four subsequent terms, of orders O(2s ln s ),
5/2
O(2s ), O(s ), and O(3s ln s ), are also known.9 Like for scalar field theory, this progress
is possible via the use of effective field theory methods that we discuss in the next chapter.
9 T. Toimela, The next term in the thermodynamic potential of QCD, Phys. Lett. B 124 (1983) 407; P. Arnold and
C. Zhai, The three loop free energy for pure gauge QCD, Phys. Rev. D 50 (1994) 7603; C. Zhai and B. Kastening, The
free energy of hot gauge theories with fermions through g 5 , Phys. Rev. D 52 (1995) 7232; K. Kajantie, M. Laine,
K. Rummukainen and Y. Schr
oder, The pressure of hot QCD up to g 6 ln(1/g), Phys. Rev. D 67 (2003) 105008
[hep-ph/0211321].
73
5.7.
Exercise 8
In the computation in Sec. 5.6, we assumed that only the Matsubara zero modes of the fields Aa0
need to be resummed (i.e., get an effective thermal mass). We do know that fermions do not need
to be resummed in any case, but how about ghots? Show that ghosts do not get any thermal mass,
and thus behave like the spatial components of the gauge fields.
The tree-level ghost propagator is in Eq. (5.46), and we now consider corrections to this expressions.
The relevant vertex is the one in Eq. (5.50). We obtain
D
1 E
b
S2
ca (P )c (Q)
2 I 0,c
Z
Z
g 2 D a b P c d e P g h i E
c (U )A (V )
c (X)
c (T )
c (P )c (Q) c (R)A (S)

=
2
0,c

Ub ,Vb ,Xb
Rb ,Sb ,Tb
f
Z
2P
cde ghi
+ S + T) (U
+ V + X)
R
U
(R
c
g
cb (Q)i
0 hAd (S)
Ah (V )i0
h
ca (P )c (R)i
ce (T)c (U
)i0 h
ci (X)
0 h
,S

R
b b ,Tb ,Ub ,Vb ,Xb
cde ghi

.
+ V + X)
R
U
(R + S + T) (U
(5.113)
Inserting the gluon propagator from Eq. (5.63) and the free ghost propagator from Eq. (5.46), let
us inspect in turn colour indices, Lorentz indices, and momenta. The colour contractions result in
a factor
ac eg ib dh f cde f ghi = f ade f edb = Nc ab .
(5.114)
The spacetime indices yield simply . The momentum dependence can be written as
(T U
) (X
Q)
(S + V )
(P R)
+ S + T) (U
+ V + X)
R
U
(R
2 S2
P 2 T 2 Q
(T U
) (X
Q)
(S + V )
(P R)
P T
(P + S + T) (T S + Q)
2
2
2
2
P T Q S
(T U
) (Q
X)
(S + V )
(P R)
P T .
(5.115)
=
(P + S + T) (P + Q)
2 T 2 (P T)2
P 2 Q
U
, X,
V and S.
Thereby
We can now integrate over R,
D
Z
1 E
P
P S
ab (P Q)
b
SI2
= g 2 Nc
ca (P )c (Q)
,
2
2
2
2
2
0,c
(P )
S (P S)
S
(5.116)
We note that the result here is proportional to the external momentum P . Therefore, the
result in Eq. (5.116) does not represent an effective mass correction; it is rather a wave function
(re)normalization contribution.
74
6.
6.1.
Low-energy effective field theories

The infrared problem of thermal field theory
Let us consider the types of integrals that appear in perturbation theory. According to Eqs. (2.34),
(4.58), each new loop order (corresponding to an additional loop momentum) produces one of
Z
P
f (nb , p) =
Z
P
f (nf , p) =
Pb
Pf
Z
p
1
2
Z
1
2
+i0+
i0+
+i0+
i0+

d
[f (, p) + f (, p)][1 + 2nB (i)] ,
2
(6.1)

d
[f (, p) + f (, p)][1 2nF (i)] ,
2
(6.2)
depending on whether the new line is bosonic or fermionic. The functions f here contain propagators, and possibly structures emerging
from the vertices; in the simplest case, f (, p) 1/( 2 +E 2 ),
p
2
where we again denote E p + m2 .
Now, the structures which are most important (i.e. yield the biggest contributions) are those
where the functions f are largest. Let us inspect this in terms of both the left-hand and the
right-hand sides of Eqs. (6.1), (6.2).
For bosons, the largest contribution on the left-hand side of Eq. (6.1) is associated with the
Matsubara zero-mode, nb = 0; for f (, p) 1/( 2 + E 2 ), this gives the term

n X o
T
2 .
f
T
(6.3)
E

b
b =0
n
In terms of the right-hand side, we have to close the contour in the lower half-plane, and the large
term is associated with Bose-Einstein enhancement around the pole = iE:

i
1 1
1
1 2i 2 h
1 + 2nB (E) =
+ E/T
{. . .}
2 2 2iE
E 2
e
1

1 1
T
1
1
.
(6.4)
+
.
.
.
=
+
O
+
E 2
E/T + E 2 /2T 2
E2
T
For fermions, there is no Matsubara zero-mode on the left-hand side of Eq. (6.2), so that the
largest terms have the magnitude

n X o
T
1
f
T
.
(6.5)
2
(T
)
T

f
n
f =T
n
Similarly, in terms of the right-hand side of Eq. (6.2), we can estimate

i
1 1
1
1 2i 2 h
1 2nF (E) =
E/T
{. . .}
2 2 2iE
E 2
e
+1

1
1 1
1
+ ... = O
E 2
2 + E/T
T
(6.6)
Given these estimates, let us try to construct a dimensionless expansion parameter associated
with the loop expansion. Apart from an additional propagator, each loop order also brings in an
additional vertex (or vertices); let us denote the corresponding coupling by g 2 , as would be the
75
case in gauge theory. Morover, the Matsubara summation involves a factor T , so we can assume
that the expansion parameter contains the combination g 2 T . We now have to use the other scales
in the problem to transform this into a dimensionless numbers. For the Matsubara zero-modes,
Eq. (6.3) tells that we are allowed to use inverse powers of E or, after integration over the spatial
momenta, inverse powers of m. Therefore, we can assume that for large temperatures, T m,
the largest possible expansion parameter is
b
g2T
.
m
(6.7)
For fermions, in contrast, Eq. (6.5) indicates that inverse powers of E or, after integration over
spatial momenta, m, cannot appear; we are lead to the estimate
f
g2T
g2 .
T
(6.8)
Obviously, we expect also some typical loop factor like 1/(4)2 to appear, but this has been omitted
for simplicity.
Assuming that we work in the weak-coupling limit, g 2 1, we can thus conclude the following:
Fermions appear to be purely perturbative in these computations.
Bosonic Matsubara zero-modes appear to be purely non-perturbative for m 0.
The resummations we saw in Exercise 5 for scalar field theory and in Sec. 5.5 for QCD produce
an effective thermal mass, m2eff g 2 T 2 . Then we may expect the expansion parameter in
Eq. (6.7) to become g 2 T /gT = g. In other words, a small expansion parameter exists in
principle, but the structure of the weak-coupling series is peculariar, with odd powers of g,
as we have seen before.
As we found in Eq. (5.95), colour-magnetic fields do not develop a thermal mass at O(g 2 T 2 ).
2
This might still happen at higher orders, so we can say that meff <
g T . Thereby the expan2
2
sion parameter in Eq. (6.7) reads b >
g T /g T = 1. In other words, colour-magnetic fields
cannot be treated perturbatively; this is known as the infrared problem (or Linde problem) of
thermal gauge field theories10 .
The situation we observe here, namely that serious infrared problems exist, but that they are
related to quite particular degrees of freedom, is common in field theory. Correspondingly there is
also a generic tool, called the effective field theory approach, which allows to isolate the infrared
problems to a simple Lagrangian, and treat them in this simplified setting. In fact, the concept
of effective field theories is not restricted to finite-temperature physics, but applies even at zero
temperature, if the system possesses a scale hierarchy (in the thermal context, the hierarchy is
often expressed as g 2 T / gT T , where the first scale refers to the non-perturbative one
generated for the colour-magnetic fields). Given the generic nature of effective field theories, let
us indeed start by discussing the basic idea in a simple zero temperature setting.
6.2.
A simple example of an effective field theory
Let us consider a zero-temperature Lagrangian containing two different scalar fields, and H, with
widely different masses, m and M , respectively11 :
Lfull
10
1
1
1
1
1
1
+ m2 2 + H H + M 2 H 2 + g 2 2 H 2 + 4 + H 4 .
2
2
2
2
4
4
(6.9)
A.D. Linde, Infrared problem in thermodynamics of the Yang-Mills gas, Phys. Lett. B 96 (1980) 289.
example and the related discussion follow closely those in the book Renormalization, by J.C. Collins.
11 This
76
We assume that there exists the scale hierarchy m M (or, to be more precise, mR MR ; we
leave out the subscripts in the following). The effective theory now has the form
Leff =
1 1 2 2 1 4
+ m
+ + . . . ,
2
2
4
(6.10)
where an infinite set of higher dimensional operators have been dropped out. Note that in principle,
if we also wanted to describe gravity with these theories, we could add a fundamental cosmological
in Leff .
constant in Lfull , and an effective cosmological constant
The statement concerning the effective description goes roughly as follows: Let us assume that
m<
gM , and consider external momenta p <
gM . In addition, we assume that all the couplings
are parametrically of the same order of magnitude, g 2 . Then the one-particle-irreducible
n , computed within the effective theory, reproduce those of the full theory, n ,
Greens functions
with a relative error
n n |
n
|
k
<
(6.11)
O(g ) , k > 0 ,
n
n
are tuned suitably. (We assume the normalizations of the fields to be so chosen
if m
2 and
that the kinetic term in Eq. (6.10) has its canonical form.) The number k may depend on the
dimensionality of spacetime; it may also depend on n although a universal lower bound should exist;
the lower bound could be arbitrarily increased by adding suitable higher dimensional operators to
Leff , and in the limit of infinitely many such operators the effective description becomes exact.
A weaker form of the effective theory statement, although already sufficiently strong and simultaneously one which may be easier to make precise (for instance, it is effectively implemented in
the so-called non-perturbative Symanzik improvement program of lattice QCD12 ), is that Greens
functions are matched only on-shell, rather than for arbitrary external momenta.
The general effective theory statement is, as sometimes said, at the same time almost trivial, and
yet extremely difficult to prove. We will not prove the statement here either, but let us try to get
an impression on how it arises, operationally, by inspecting explicitly 2-point Greens functions.
To be more precise, let us inspect the inverse propagator of the light field . In the full theory,
at 1-loop level, this reads
G1
+
(1)
(1)
= p2 + m2 + l (0; m2 ) + h (0; M 2 ) ,
(6.12)
where the dashed line represent the light field; the solid line the heavy field; and the subscripts
(1)
(1)
l, h stand for light and heavy, respectively. The first argument of the functions l , h is the
external momentum.
Within the effective theory, the same computation yields
1
G
=
=
+
(1) (0; m
p2 + m
2 +
2) .
l
(6.13)
The equivalence of all Greens functions at the on-shell point should also imply the equivalence
of pole masses, i.e. the locations of the on-shell points. By matching Eqs. (6.12) and (6.13), we see
that this can indeed be achieved provided that
(1)
m
2 = m2 + h (0; M 2 ) + O(g 4 ) .
12 K.
(6.14)
Jansen et al., Non-perturbative renormalization of lattice QCD at all scales, Phys. Lett. B 372 (1996) 275.
77
Note that within perturbative theory the matching needs to be carried out order-by-order:
m
(1) (0; m
2 ) is already of 1-loop order, so inside it ,

2 can be replaced by , m2 , respectively,
l
given that the difference between and as well as m

2 and m2 is itself of 1-loop order, whereby
(1)
(1)
2
2
4
l (0; m
) = l (0; m ) + O(g ).
The situation becomes considerably more complicated once we go to the 2-loop level. Let us
analyse various types of graphs that exist in the full theory, and try to understand how they could
be matched with the simpler contributions existing within the effective theory.
First of all, there are graphs involving only the light fields,
.
(6.15)
These can be directly matched with the corresponding graphs within the effective theory computation; as above, the fact that different parameters appear in the propagators (and vertices) is a
higher order effect.
Second, there are graphs which clearly account for the insignificant higher order effects that
we omitted in the 1-loop matching, but would play a role once we go to the 2-loop level:
(1)
(m
2 m2 )
l (0; m2 )
,
m2
(6.16)
(1)
)
(
l (0; m2 )
.
(6.17)
(1) (0; m
These two combine to reproduce (a part of) the 1-loop effective theory expression
2 ) with
l
2-loop full theory accuracy.
Third, there are graphs only involving heavy fields in the loops:
.
(6.18)
Obviously we can account for their effects by a 2-loop correction to m

2.
Finally, there remain the by far most complicated graphs: structures involving both heavy and
light fields, in a way that the momenta flowing through the two sets of lines do not factorise:
=
(6.19)
Naively, the representation on the right-hand side might suggest that this graph is simply part of
)(1) (0; m2 )/, just like the graph in Eq. (6.17); this, however, is not the
the correction (
l
case, because the substructure appearing,
,
(6.20)
is momentum-dependent, unlike the effective vertex .

Nevertheless, it should be possible to split up Eq. (6.19) in two parts:
+
(2)
mixed (p2 ; m2 , M 2 )
(2)
2
2
2
mixed (p ; m , M )
78
(2)
2
2
2
+
mixed (p ; m , M ) .
(6.21)
(6.22)
(2) is, by definition, characterised by the fact that it depends non-analytically on

The first part
the mass parameter m2 of the light field; therefore the internal field is soft in this part, i.e.
gets a contribution only from momenta p m. In this situation, the momentum dependence of
Eq. (6.20) is of subleading importance; indeed it can only be significant if the momentum flowing
through the loop is of order p M . In other words, this part of the graph does contribute simply
)(1) (0; m2 )/, as we naively expected.
to (
l
(2) is, by definition, analytic in the mass parameter m2 . We associate this
The second part
with a situation where the internal is hard: even though its mass is small, it can have a large
momentum p M , transmitted to it through interactions with the heavy modes. In this situation,
the momentum dependence of Eq. (6.20) plays an essential role. At the same time, the fact that
all internal momenta are hard, allows for a Taylor-expansion in the small external momentum:
(2) (p2 ; m2 , M 2 ) =
(2) (0; m2 , M 2 ) + p2
(2) (0; m2 , M 2 )
mixed
mixed
p2 mixed
1
2 (2)
+ (p2 )2
(0; m2 , M 2 ) + . . . .
(6.23)
2
(p2 )2 mixed
The first term here represents now a 2-loop correction to m
2 , just like the graph in Eq. (6.18).
Finally,
The second term can be compensated for by a change of the normalization of the field .
the further terms have the appearance of higher order (derivative) operators, truncated from the
structure shown explicitly in Eq. (6.10). Compared with the leading kinetic term, the magnitude
of the third term is very small, however:
2 2
g 4 (pM )2
< g6 ,
(6.24)
p2
for p <
gM . Therefore a truncation is indeed justified unless we want to go beyond a certain relative
accuracy. The structures in Eq. (6.23) are collectively denoted by a 2-point blob in Eq. (6.21).
To summarise, we see that the explicit construction of the effective field theory becomes quite
subtle at higher loop orders. Another illuminating example of the difficulties met with mixed
graphs is given in Exercise 9.
Nevertheless, the following conjecture/philosophy/recipe for an effective field theory description
can be formulated, and is assumed to hold in general (in weakly-coupled theories):
(1) Identify the light or soft degrees of freedom, i.e. the ones that are IR-sensitive.
(2) Write down the most general Lagrangian for them, respecting all the symmetries of the
system, and including local operators of arbitrary order.
(3) The parameters of this Lagrangian can be determined by matching:
compute the same observable in the full and in the effective theory, with the same UVregularization and IR-cutoff.
subtract the results.
the IR-cutoff should disappear; the result of the subtraction should be analytic in p2 , and
allow for a matching of the parameters and field normalizations of the effective theory.
if not, the degrees of freedom, or the form of the effective theory, have not been correctly
identified!
(4) Truncate the effective theory by dropping higher dimensional operators suppressed by 1/M k ,
which can only give a relative contribution of order
m k
gk .
(6.25)
M
79
6.3.
Dimensionally reduced effective field theory for hot QCD
We now apply the recipe of the previous section to the problem outlined in Sec. 6.1.
(1) Identification of soft degrees of freedom. As discussed in Sec. 6.1, the soft degrees of
freedom are the bosonic Matsubara zero-modes. Since they do not depend on the coordinate ,
they live in d = 3 2 spatial dimensions; thus the construction of the effective theory is in this
context called high-temperature dimensional reduction13 .
(2) Symmetries. First of all, since the heat bath breaks Lorentz invariance, the time direction
and the space directions are not symmetric. Indeed, the effective theory needs only to be invariant
in spatial rotations and translations.
Second, the full theory possesses discrete symmetries; QCD is invariant in C, P and T separately.
The effective theory inherits some reflection of these symmetries; in turns out, for instance, that
Leff is symmetric in A0 A0 (unless, say, C of QCD is broken by giving a chemical potential to
the quarks).
Third, consider gauge symmetries, Eqs. (5.5), (5.6):
A
i
U A U 1 + U U 1 .
g
(6.26)
Since we now restrict to static (i.e. -independent) fields, U (or a ) should not depend on either.
Thus, the effective theory should be invariant under
Ai
A0
i
U Ai U 1 + U i U 1 ,
g
U A0 U 1 .
(6.27)
(6.28)
In other words, the spatial components Ai remain gauge fields, while the temporal components A0
have turned into scalar fields in the adjoint representation (cf. Eq. (5.9)).
With these ingredients, we can write down the general form of the effective Lagrangian. It is
illuminating to start by simply rewriting the full Lagrangian, Eq. (5.34), in terms of the soft degrees
of freedom. Noting from Eq. (5.32),
a
F0i
Aai Diab Ab0 ,
(6.29)
a
that in the static case, Fi0
= Diab Ab0 , we get
LE =
1 a a 1 ab b
F F + (D A )(Diac Ac0 ) .
4 ij ij 2 i 0
(6.30)
It is often convenient to note that

T a Diab Ab0 = i A0 + gf acb T a Aci Ab0 = i A0 ig[Ai , A0 ] = [Di , A0 ] ,
(6.31)
where Di = i igAi is the covariant derivative in the fundamental representation. Thereby,

(0)
Leff =
1 a a
F F + Tr {[Di , A0 ][Di , A0 ]} .
4 ij ij
(6.32)
13 P. Ginsparg, First and second order phase transitions in gauge theories at finite temperature, Nucl. Phys. B 170
(1980) 388; T. Appelquist and R.D. Pisarski, High-temperature Yang-Mills theories and three-dimensional Quantum
Chromodynamics, Phys. Rev. D 23 (1981) 2305.
80
Next, we complete the tree-level structure by adding all operators allowed by the symmetries. It
is useful to proceed in order of increasing dimensionality. The following structures can be written
down:
Tr [A20 ] ;
Tr [A40 ] ,
dim = 2 :
(6.33)
(Tr [A20 ])2 ;

Tr {[Di , Fij ][Dk , Fkj ]} , . . . .
dim = 4 :
dim = 6 :
(6.34)
(6.35)
In the last case we have only shown one example; in total there is quite a large number of sixdimensional operators14.
Combining Eqs. (6.32)(6.34), we can write the effective action as

Z
1
1 a a
2
2
(1)
2 2
(2)
4
d
Seff =
F F + Tr ([Di , A0 ][Di , A0 ]) + m
Tr [A0 ] + (Tr [A0 ]) + Tr [A0 ] + . . . .
d x
T
4 ij ij
(6.36)
R
Here the prefactor 1/T comes from the integration 0 d ; since none of the soft fields depend on ,
we just get 1/T , like in classical statistical physics. Sometimes this theory is referred to as EQCD,
for Electrostatic QCD15 .
(3) Matching. If we restrict to 1-loop order, then the matching for the parameters in Eq. (6.36)
is rather simple, as explained in Eq. (6.14): we simply need to compute Greens functions with
soft fields (with vanishing external momenta) on the external legs, and heavy modes in the loop.
For the parameter m
2 , this is precisely the computation that we carried out in Sec. 5.5. Therefore,
the result can be read directly from Eq. (5.96):

Nc
Nf
m
2 = g2T 2
+ O(g 4 T 2 ) .
(6.37)
+
3
6
(1) ,
(1) can, in turn, be obtained by considering 4-point functions with soft
The parameters
modes of A0 on the external legs, and non-zero Matsubara modes in the loop:
+
(6.38)
Clearly the effect is of O(g 4 ) and, using the same notation as in Eq. (5.96), the actual values are16
4
(1) = g + O(g 6 ) ,
4 2
4
(2) = g (Nc Nf ) + O(g 6 ) .
12 2
(6.39)
(4) Truncation of higher dimensional operators. In some sense the most non-trivial and
critical part of any effective field theory construction is the analysis of the accuracy that can be
reached with the effective theory, once higher dimensional operators are dropped. In other words,
the challenge is to determine the constant k in Eq. (6.11). Let us illustrate the procedure by
considering the error that we make by dropping the operator in Eq. (6.35).
We need to know, first of all, the parametric magnitude of the coefficient with which the operator would enter Leff , if it were kept. This operator could be generated through the momentum
dependence of graphs like
n6=0
g2
(i Fija )2 ,
(6.40)
T2
14 S.
(1994)
15 E.
16 S.
(1988)
498.
Chapman, A new dimensionally reduced effective action for QCD at high temperature, Phys. Rev. D 50
5308.
Braaten and A. Nieto, Free energy of QCD at high temperature, Phys. Rev. D 53 (1996) 3421.
Nadkarni, Dimensional reduction in finite temperature Quantum Chromodynamics. 2, Phys. Rev. D 38
3287; N.P. Landsman, Limitations to dimensional reduction at high temperature, Nucl. Phys. B 322 (1989)
81
where the dashed lines stand for the spatial components Ai . If we now drop this term, the
corresponding Greens function will not be computed correctly; however, it still has some value,
namely that which would be obtained within the effective theory:
0
A
Z
g2T
1
3 (i Fija )2 .
(6.41)
g 2 (i Fija )2 T
2
2
3
)
m
p (p + m
Here we noted that there are two propagators, but to account for the dependence on the external
momentum (represented by the derivative i in front of Fija ), we need to Taylor-expand in p to
first non-trivial order.
We note from Eq. (6.41) that the Greens function is also non-zero within the effective theory;
in fact, the contribution in Eq. (6.41) is larger than that in Eq. (6.40)! Therefore, the error made
through the omission of Eq. (6.40) is small:
m
g2 m
3
3
2 2
g3 .
T g T
T
(6.42)
In other words, for the dimensionally reduced effective theory of hot QCD, we get k = 3.17
Once the effective theory of Eq. (6.36) is there, we can take a further step: the field A0 is massive,
and can be integrated out. Thereby we arrive at an even simpler effective theory:

Z
1
1 a a
Seff
=
(6.43)
dd x
F ij F ij + . . . .
T
4
Sometimes this theory is referred to as MQCD, for Magnetostatic QCD15 . It is important to
realise that this theory, the three-dimensional Yang-Mills theory (up to higher order operators such
a
as the one in Eq. (6.35)), only has one parameter, the gauge coupling. Furthermore, if the fields A
i
a T 1/2 , then the coefficient 1/T in Eq. (6.43)
a A
are rescaled by an appropriate power of T 1/2 , A
i
2
disappears. The coupling constant squared that appears afterwards is g T , and this is the only
scale in the system. Therefore all dimensionfull quantities (correlation lengths, string tension, free
2
energy density, ...) must be proportional to an appropriate power of g T , with a non-perturbative
coefficient. This is the essence of the non-perturbative physics pointed out by Linde10 .
The implication of the complete setup for the weak-coupling expansion is the following. Consider
a generic observable O, with an expectation value
hOi g m T n [1 + #g p + ...] .
(6.44)
It could now happen that: (i) p is even; # is determined by the heavy scale T , and is purely
perturbative; (ii) p is even or odd; # is determined by the intermediate scale gT , and is purely
perturbative; (iii) m + p is even; # is determined by the soft scale g 2 T , and is non-perturbative;
(iv) p > k; # can only be determined correctly by adding higher dimensional operators to the
effective theory.
A final remark, relevant for effective field theories quite in general, is in order. Indeed, we have
seen that the omission of higher order operators usually leads to a small error, since the same
Greens function is produced with a larger coefficient within the effective field theory. It could
happen, however, that there is some approximate symmetry in the full theory, which becomes
exact within the effective theory, if we truncate to some order. For instance, many Grand Unified
Theories induce violation of baryon minus lepton number (B L), while in the Standard Model this
is an exact symmetry. It is then only broken by some higher dimensional operator18. Therefore, if
we consider B L violation with the Standard Model, we make an infinitely large relative error.
17 K. Kajantie, M. Laine, K. Rummukainen and M.E. Shaposhnikov, High temperature dimensional reduction and
parity violation, Phys. Lett. B 423 (1998) 137.
18 S. Weinberg, Baryon and Lepton Nonconserving Processes, Phys. Rev. Lett. 43 (1979) 1566; F. Wilczek and
A. Zee, Operator Analysis of Nucleon Decay, Phys. Rev. Lett. 43 (1979) 1571.
82
6.4.
Exercise 9
Let us consider the full theory

Lfull
1
1
1
1
1
+ m2 2 + H H + M 2 H 2 + H3 .
2
2
2
2
6
(6.45)
For simplicity, we assume that the dimensionality of spacetime is 3 (i.e. d = 2 2 in our standard
notation). Moreover we work at zero temperature, like in Sec. 6.2.
(a) Integrating out H in order to construct an effective theory amounts to the computation of
the graph
.
(6.46)
After Taylor-expanding in external momenta, write down all the corresponding operators.
at vanishing external momenta. What kind of
(b) Let us consider the 4-point function of s
contributions do the operators computed in part (a) give to this Greens function?
(c) Let us then consider directly the graph
(6.47)
at vanishing external momenta. How does the result compare with the Taylor-expanded
result of part (b)? What is the lesson?
(a) The construction of the effective theory proceeds essentially as in Eq. (3.56), except that only
the H-fields are integrated out. We get
Seff
=
=
=
=
=
D 1 E
SI2
2
H,connected
2 Z
3 (x)3 (y)hH(x)H(y)i0
72 x,y
Z
Z ip(xy)
e
2
3
3
(x) (y)
2 + M2
72 x,y
p
p
X

Z
Z
2
(1)n (p2 )n
3 (x)3 (y) eip(xy)

72 x,y
(M 2 )n+1
p
n=0

X
Z
(2x )n
2
3
3
(x) (y)
(x y)
72 x,y
(M 2 )n+1
n=0
Z X
2
(2x )n 3
(x) .
3 (x)
72 x n=0
(M 2 )n+1
(b)
D
Seff
(0)
(0)
(0)e
(0)
2 D
(0)(0)(0)(0)
72
EX
[(P4 + P5 + P6 )2 ]n
1 ) . . . (P
6)
(i Pi )(P
(M 2 )n+1
P1 ,...,P6
n=0
83
(6.48)
2
6 (2 3 2 + 3 4 2)
72
(i Pi )
P1 ,...,P6
X
[(P4 + P5 + P6 )2 ]n
(M 2 )n+1
n=0
(P
1 )ih(0)
(P
2 )ih(0)
(P
5 )ih(0)
(P
6 )ih(P
3 )(P
4 )i
h(0)
Z
2
n
X (P )
1
(0)
3
3 2 2 4
.
2
2
(m
) P3 P3 + m
n=0 (M 2 )n+1
(6.49)
The integrals in Eq. (6.49) can all be carried out in dimensional regularization; for instance,
the two leading terms read

Z
m
1
1
1
,
(6.50)
= 2
n=0:
M 2 P3 P32 + m
2
M
4
Z
Z
1
3
m
2
1 m
P32
1
n=1:
4
=
=
,
(6.51)
M P3 P32 + m
2
M 4 P3 P32 + m
2
M 4 4
where we made use of Eq. (2.85). We note that, indeed, the terms get smaller with increasing
n, apparently justifying a posteriori the Taylor-expansion that we carried out in part (a).
(c) Let us, on the other hand, carry out the integral corresponding to Eq. (6.47), without a
Taylor expansion. The contractions remain as in part (b), and we simply need to replace the
integral in Eq. (6.49) by

Z
Z
1
1
1
1
1
=
2
2
2
2 P32 + M 2
2 P32 + m
2
P3 + M 2
P3 M m
P3 P3 + m

1
1
=
(m
M)
2
2
M m
4
1
=
4(M + m)

m
m
2
1
1
(6.52)
+ 2 + ... .
=
4M
M
M
Comparing now Eqs. (6.50), (6.51) with Eq. (6.52), we note that by carrying out the Taylorexpansion, i.e. the naive matching to various effective parameters, too early, we missed the
leading contribution, the dominant term in Eq. (6.52)! The largest term we found, Eq. (6.50),
is only next-to-leading order in Eq. (6.52).
The reason for this problem is the same as in Eq. (6.21): it has to be taken into account
that the light fields can also have large momenta P3 , P3 M , in which situation a Taylorexpansion of 1/(P32 + M 2 ) is not allowed. Rather, we have to view Eq. (6.47) in analogy with
Eq. (6.21):
=
(6.53)
The first term here corresponds to a naive replacement of Eq. (6.46) by a momentumindependent 6-point vertex, times a dynamical effect from soft fields; indeed the result,
Eq. (6.50), is non-analytic in the parameter m
2.
The second term, on the other hand, corresponds to a contribution from hard -modes to
the effective 4-point vertex. The result, the leading term in Eq. (6.52), is indeed analytic in
the parameter m
2.
84
7.
Finite density
Let us consider a system which possesses some conserved charge, Q. We assume the conserved
charge to be additive, i.e. the charge can in principle have any (integer) value. Physical examples
of possible Qs include:
the baryon number B and the lepton number L. (In fact, within the Standard Model, the
combination B + L is not conserved because of an anomaly19 , so that strictly speaking only
the linear combination B L is conserved; however, in practice the rate of violation of B + L
is exponentially small, so that we can treat both B and L as conserved quantities.)
if weak interactions are switched off (i.e., if we inspect phenomena at a time scale well below
1010 s, or distances well below 1 cm within the collision region of a particle experiment),
then quantities like the strangeness S and the isospin I are also conserved.
in some supersymmetric theories, there is a quantity called the R-charge which is conserved.
in non-relativistic field theories, the particle number N is conserved.
The case of a conserved Q turns out the be analogous to the case of gauge fields, treated on
p. 58; indeed, as we will see, the introduction of a chemical potential, , as a conjugate variable to
Q, is closely related to the introduction of the gauge fields, A0 , that were needed for imposing the
Gauss law, Q = 0, in the gauge field case. However, in contrast to that situation, we will work
in a grand canonical ensemble in the following, so that the quantum mechanical partition function
is in general of the type
i
h
(7.1)
Z(T, ) Tr e(HQ) .
On p. 58 the projection operator Q,

0 was effectively imposed as
Z
dc icQ
Q,
=
e
,

0
2
(7.2)
with c A0 . Comparing Eqs. (7.1), (7.2), we see that a chemical potential corresponds to something like a constant purely imaginary gauge field A0 .
Now, let us go back to classical field theory for a moment, and recall that if the system possesses
a global U(1) symmetry, then there exists, according to Noethers theorem, a conserved global
current, J . The integral of the zero-component of the current, i.e. charge density, over the spatial
volume, defines the conserved charge,
Z
Q dd x J0 (t, x) .
(7.3)
Conversely, we can assume that in order to describe a system which does have a conserved global
charge, then there should exist a global U(1) symmetry in its field-theoretic description. In general
this indeed is the case, and we will restrict to these situations in the following. (One notable
exception is free field theory where, due to lack of interactions, particle number is conserved even
without a global symmetry; another is that a discrete symmetry, , may also lead to the
concept of a generalized parity, which acts as a multiplicative quantum number, with possible
values 1; however, in this case no non-trivial charge density = hQi/V

can be defined in the
theormodynamic limit).
An immediate consequence of the inverse use of the Noether theorem is that a real scalar
field cannot describe a system with an additive conserved charge. As the simplest example, let us
therefore consider a system of a complex scalar field.
19 G.
t Hooft, Symmetry breaking through Bell-Jackiw anomalies, Phys. Rev. Lett. 37 (1976) 8.
85
7.1.
Complex scalar field with a finite chemical potential
The classical Lagrangian of a complex scalar field reads

LM = V () ,
(7.4)
V () m2 + ( )2 .
(7.5)
where the potential has the form
The system is symmetric in the global (position-independent) phase transformation

ei ,
ei ,
(7.6)
where R. The corresponding Noether current is defined as

J
LM
LM
+
( ) ( )
= i i
= i[ ] = 2 Im[ ] .
(7.7)
Let us note that the overall sign (i.e., what we call particles and what antiparticles) is a matter of
convention; we could equally well have defined the global symmetry through ei ,
ei , and then J would have the opposite sign.
The first task now, as always, is to write down a path integral expression for the partition
function in Eq. (7.1). Subsequently, we may try to evaluate the partition function, in order to see
what kind of phenomena take place in this system.
In order to write down the path integral, we start from the known expression of Z of a real scalar
field without chemical potential, i.e. the generalization to field theory of Eq. (1.37):

Z Z
Z
Z
1
1 2
2
1 i1 1 + (i 1 ) + V (1 )
,
(7.8)
D1 D1 exp
Z
d
2
periodic
x 2
0
where 1 = 1 /t. Here the combination
Hamiltonian density, H(1 , 1 ).
1
2
12 +
1
2
(i 1 )2 + V (1 ) is nothing but the classical
In order to
be able to make use of Eq. (7.8), let us rewrite the complex scalar field as =
(1 + i2 )/ 2, i R. Then
=
1
1
1 1 + 2 2 ,
2
2
1 2
( + 22 ) ,
2 1
and the classical Hamiltonian density reads

i 1
1h 2
1 + 22 + (i 1 )2 + (i 2 )2 + m2 21 + m2 22 + (21 + 22 )2 .
H=
2
4
(7.9)
(7.10)
In order to go to the grand-canonical ensemble, we need to add from Eqs. (7.3), (7.7) the classical
to the Hamiltonian, cf. Eq. (7.1):
version of Q
Z
h
i
Q = Im (1 i2 )(t 1 + it 2 )
Zx
Z
= (1 2 2 1 ) =
(1 2 2 1 ) .
(7.11)
x
Since the charge can be expressed in terms of the canonical variables, nothing changes in the
derivation of the path integral, and we can simply replace the Hamiltonian of Eq. (7.8) by the sum
of Eqs. (7.10), (7.11).
86
Finally, we again carry out the Gaussing integrals over 1 , 2 :

2
Z
1 2
1
1 1
+ 2
+ i2
d1 exp 1 + 1 i
= const. exp
, (7.12)
2
2

2

Z
2
1 2
1
= const. exp
. (7.13)
1
i1
d2 exp 22 + 2 i
2
2
Afterwards, we can go back to the complex notation:

2

2
1 1
1 2
+ i2 +
i1
2
2

2
2

1
1
1
2
2
1
+ i 2
+
=
1
2 (21 + 22 )
2
2
|
{z
}

[( ) ][( + )] .
(7.14)
In total, then, the path integral representation for the grand canonical partition function reads

Z Z
Z
.
Z(T, ) = C
D exp
d
( ) ( + ) + i i + m2 + ( )2
periodic
(7.15)
We observe that, as anticipated, appears in a way reminiscent of an imaginary gauge field A0 .
Let us finally work out the properties of the free theory in the presence of . Going to momentum
space, the quadratic part of the Euclidean action becomes
Z
i
P h
P )
(P ) (in )(in + ) + p2 + m2 (
SE =
P
Z b
i
P h
(P )(P ) (n i)2 + p2 + m2 .
(7.16)
Pb
We observe that the chemical potential corresponds simply to a shift of the Matsubara freqencies by
a constant imaginary term (this is the reason for considering a corresponding sum in Eq. (2.36)).
In particular, the propagator reads
P ) (Q)i
0 = (P Q)
h(
(n
1
,
+ p2 + m2
i)2
(7.17)
while the free energy density is obtained from Eqs. (2.44), (2.49), after replacing c i and
noting that for a complex scalar field, all Fourier modes are independent, whereby Eqs. (2.44),
(2.49) are to be multiplied by a factor 2:

Z d

d p
E+
E

T
T
(7.18)
E
+
T
ln
1
e
+
ln
1
e
f (T, ) =
2 2 .
d
(2)
E= p +m
We may wonder how the existence of 6= 0 affects the infrared problem of finite-temperature
field theory. In Sec. 2.6 we found that the high-temperature expansion (T m) of Eq. (7.18) has
a peculiar structure, because Eq. (7.18) has a branch cut starting at m2 = 0. From the last term in
Eq. (7.18), we note that this problem has become worse in the presence of > 0: now the integral
is not defined even at a finite m if > m, because then exp((E )/T ) > 1 at small |p|. Of
course, in an interacting theory, thermal corrections generate an effective mass m2eff T 2 , which
postpones the problem to larger . Nevertheless, at large enough it still exists.
It turns out that there is a significant physics consequence from this infrared problem: the
existence of Bose-Einstein condensation, to which we now turn.
87
7.2.
Effective potential and Bose-Einstein condensation
In order to treat properly complex scalar field theory with a chemical potential, two things need
to be realized:
(i) In contrast to gauge field theory, the infrared problem exists now even in the non-interacting
limit. Therefore it cannot be cured by a non-perturbatively generated confinement scale.
Rather, it corresponds to a strong dependence of the properties of the system on the volume,
so we should indeed keep the volume finite.
(ii) The chemical potential is a most useful quantity in theoretical computations, but it is
somewhat abstract from a practical point of view; the physical properties of the system
are typically best characterised not by but by the variable conjugate to , i.e. the number
density of the conserved charge. Therefore, rather than trying to give some specific value,
we should in the first place fix the number density.
So, let us put the system in a periodic box, V = L1 L2 L3 . The spatial momenta get discretized
like in Eq. (2.9),
n n n
1
2
3
p = 2
,
,
,
(7.19)
L1 L2 L3
with ni Z. The mode with b = 0, p = 0 will be called the condensate, and denoted by .
n
We now rewrite the partition function in Eq. (7.15) as

Z

Z
SE [=+
]
d
Z(T, ) =
D e
periodic,P 6=0

Z
V
.
d exp Veff ()
T
(7.20)
Here contains all modes with P 6= 0, and Veff is called the (constrained) effective potential. The
R R
factor V /T is the trivial spacetime integral, 0 d V dd x.
Let us write down the effective potential explicitly for the free theory, = 0. From Eqs. (7.15),
(7.16), we get
Z
h
i
P
V
2
2
P ) (n i)2 + p2 + m2 ,
SE [ = + ] = (m ) +
(P )(
(7.21)
T
Pb 6=0
where we made use of the fact that the crossterm between and vanishes, given that by definition
has no zero-momentum mode:
Z Z
dd x = 0 .
(7.22)
d
0
Note that the latter term in Eq. (7.21) does not correspond to only non-zero Matsubara modes:
Matsubara zero-modes (nb = 0) with non-zero spatial momentum p 6= 0 are also included. In
the limit of a large volume, mL 1, the omission of a single mode does not matter (its effect is
(T /V ) ln(m2 2 )). The path integral over the latter term yields then Eq. (7.18), and in total
we get
= (m2 2 ) + f (T, ) .
(7.23)
Veff ()
Physically, the first term corresponds to the contribution to the free-energy density from particles
that have formed a condensate, while the latter term represents free particles in the plasma.
Now, if we go toward zero temperature, T m, and assume furthermore that <

m, then the
latter term in Eq. (7.18) vanishes. The vacuum contribution left over is independent of , and can
be omitted as well. Therefore
(m2 2 ) .
(7.24)
Veff ()
88
The remaining task is to carry our the final regular integral over in Eq. (7.20). At this point we
+
need to make contact with the particle number density. From the definition Z = Tr [exp( H
Q)], we obtain
hQi
T ln Z
=
V
V

R
d 2 exp V Veff ()
T
(7.25)

R
V
d exp T Veff ()
(7.26)
Let us now consider a situation where we decrease the temperature towards low values, T m,
and attempt simultaneously to keep the particle number density, the left-hand side of Eq. (7.26),
fixed. How should we choose in this situation? There are three possibilities:
(i) if || > m, the integrals are not defined (in the free theory). This simply means that such
choices of are not physically meaningful.
(ii) if || < m, the integrals can be carried out: in fact their result corresponds to the propaga
tor of :
T
= 2
.
(7.27)
2
V (m 2 )
We note that if T 0, then 0. This conflicts with our assumption that the particle
number density is kept constant; therefore this range of is again not physically relevant for
our situation.
(iii) According to the preceding points, the only possible choice is || = m. More specifically, if
> 0, we need to choose = +m. According to Eq. (7.27), this situation needs in fact to
be approached by a careful tuning of m as we put T 0. However, assuming that
this is done, and that we furthermore add an infinitesimal interaction > 0 to the theory so
that the integrals in Eq. (7.26) are defined, we obtain the relation
>0.
= 2mh i
(7.28)
Eq. (7.28) manifests the phenomenon of Bose-Einstein condensation (at zero temperature in the
free limit): the conserved particle number is converted to a non-zero scalar condensate.
Obviously, it would also be easy to include the effects of a finite temperature, by starting from
Eq. (7.23), and the effects of interactions, by keeping > 0. These very interesting developments go
beyond the scope of the present lectures, however. On the other hand, the concepts of a condensate
and an effective potential will be met again in later chapters.
89
7.3.
Dirac fermion with a finite chemical potential
The Lagrangian of a Dirac fermion,

LM = A (iD
/ AB m AB )B ,
possesses a global symmetry,
A ei A ,
a
D
/ AB = (AB igAa TAB
),
(7.29)
A ei A ,
(7.30)
in addition to the usual non-Abelian (local) gauge symmetry. Therefore there is an associated
conserved quantity, and we can again consider the behaviour of the system in the presence of a
chemical potential.
The conserved Noether current reads
J
=
=
LM A
( A )
A i iA = A A .
(7.31)
The corresponding charge is Q = x J0 , and as an operator it commutes with the Hamiltonian,

Q]
= 0. Therefore, like with the scalar field theory, we can treat the combination H
Q
as
[H,
an effective Hamiltonian, and directly write down the corresponding path integral, by simply
adding
Z
Q = dd x A 0 A
(7.32)
to the Euclidean action. The path integral thereby reads

Z Z
Z
d
Z(T, ) =
DD exp
d d x [
D + 0 + m] .
antiperiodic
(7.33)
For perturbation theory, let us consider the quadratic part of the Euclidean action. Going to
momentum space, we get
Z
P
P ) .
SE =
(P )[i
0 n + i
i pi + 0 + m](
(7.34)
Pf
Therefore, just like in the previous section, the existence of a chemical potential corresponds simply
to the shift n n i of the Matsubara frequencies. Everything else propagators, Feynman
rules, etc remains unchanged.
In particular, let us write down the free energy density of a single free Dirac fermion. Compared
with a complex scalar field, there is an overall factor 2 (rather than 4 in Eq. (4.51), where we
compared with a real scalar field). Otherwise, the chemical
potential appears in identical ways in
Eqs. (7.16) and Eq. (7.34), so Eq. (4.54), Sf (T ) = 2Sb T2 Sb (T ), continues to apply. Employing
it with Eq. (7.18) we get

Z d
d p
2(E)
2(E+)
+
ln
1
e
f (T, ) = 2
E
+
T
ln
1
e
(2)d

ln 1 e(E+) ln 1 e(E)

E=
p2 +m2

dd p
E+
E

T
T
E
+
T
ln
1
+
e
+
ln
1
+
e
2 2 . (7.35)
d
(2)
E= p +m
This integral is well-defined for any ; thus free elementary fermions do not suffer from infrared
problems even with 6= 0, and do not undergo any type of condensation.
90
7.4.
How about chemical potentials for gauge symmetries?
We already saw, after Eq. (7.2), that a chemical potential has some relation to a gauge field A0 .
However, in cases like QCD, a chemical potential has no colour structure (i.e. it is an identity
matrix in colour space), while A0 is a traceless matrix in colour space. On the other hand, in
QED, A0 is not traceless. In fact, in QED, the gauge symmetry is nothing but a local version of
that in Eq. (7.30). We may therefore ask whether we can associate a chemical potential with the
electric charge of QED, and what is the precise relation of A0 and in this case.
Let us first recall what happens in such a situation physically. A non-zero chemical potential
corresponds to the a system which is charged. Moreover, if we want to describe it perturbatively
with the QED Lagrangian, we had better choose a system where the charge carriers (particles) are
essentially free; such a system could be a metal or a plasma. In this situation, though, the free
charge carriers interact repulsively with a long-range force, and hence all the net charge resides on
the surface. In other words, the homogeneous bulk of the medium is neutral (i.e. has no free
charge). On the other hand, the charged body as a whole has a non-zero electric potential, V0 ,
with respect to the neutral ground.
Let us now try to understand how to reproduce this behaviour directly from the partition function, Eq. (7.33), adapted to QED:
Z Z

Z

1 2
d
Z(T, ) = DA DD exp
d d x
F + 0 ( ieA0 + ) + i Di + m
.
4
0
(7.36)
The usual boundary conditions over the time direction are assumed, even if not shown explicitly.
The basic claim is, according to the physical picture above, that if we by construction assume the
system to be homogeneous, i.e. consider the bulk situation, then the partition function should
not depend on , in order to ensure the neutrality that we expect:
f
=0.
(7.37)
How does this arise?

The key observation is that we should again think of the system in terms of an effective potential,
.
like in Eq. (7.20). The role of the condensate is now given to the field A0 ; let us denote it by A
0
The last integral to be carried out is

Z
V
(7.38)
dA0 exp Veff (A0 ) .
Z(T, ) =
T
+ , so
Now, we can deduce from Eq. (7.36) that can only appear in the combination ieA
0
that Veff (A0 ) = f (A0 + i/e). Moreover, we know from Eq. (6.36) that in a large volume and large
temperature,
) 1 m2 (A
+ i/e)2 + O(A
+ i/e)4 ,
Veff (A
(7.39)
0
0
0
2 E
where m2E e2 T 2 . (The exact effective potential, which indeed only has a quadratic and a quartic
term, can be deduced from the results of Exercise 10.) In the infinite-volume limit, the integral
in Eq. (7.38) can be carried out by making use of the saddle point approximation. The saddle
= i/e. The

point is located in the complex plane at the position where Veff
(A0 ) = 0, i.e. at A
0
value of the potential at the saddle point, as well as the Gaussian integral around it, are clearly
independent of . This, then, leads to Eq. (7.37).
It is interesting to note that the saddle point is purely imaginary. Recalling the relation of
Minkowskian and Euclidean A0 from page 59, this corresponds to a real Minkowskian A0 . Thus
there indeed is a real electric potential V0 , just as we expected physically.
91
7.5.
Exercise 10
The free energy density of a single Dirac fermion,

Z
P
ln[(n i)2 + E 2 ] ,
f (T, ) = 2
(7.40)
Pf
can be computed explicitly for the case m = 0 (i.e. E = |p|). Show that, subtracting the vacuum
part, the result is
2 4

7 T
2 T 2
4
f (T, ) =
.
(7.41)
+
+
180
6
12 2
We start from the expression in Eq. (7.35), subtracting the T -independent vacuum term and setting
m = 0, d = 3:

Z 3
d p
|p|
|p|+
T
T
+
ln
1
+
e
f (T, ) = 2T
ln
1
+
e
(2)3

Z
T4
x+y
xy
2
= 2
,
(7.42)
+ ln 1 + e
dx x ln 1 + e
0
where we set x |p|/T , y /T on the latter row, and carried out the angular integral.
A possible trick now is to expand the logarithms in Taylor series,
ln(1 + z) =
(1)n+1
n=1
zn
,
n
|z| < 1 .
(7.43)
Assuming y > 0, this is indeed possible with the first term of Eq. (7.42), while in the second term a
direct application is not possible. However, if ex+y > 1, we can write 1+ex+y = ex+y (1+exy ),
where exy < 1; thereby the Taylor expansion can be written as
h

X
X
(1)n+1 xn yn i
(1)n+1 xn yn
x+y
. (7.44)
e
e + (y x) y x +
e e
= (x y)
ln 1 + e
n
n
n=1
n=1
Inserting this into Eq. (7.42), we get
Z y

X
T4
(1)n+1 2 xn yn
xn yn
2
3
f (T, ) = 2
+e
e
x e e
dx yx x +
n
0
n=1
Z X

(1)n+1 2 xn yn
xn yn
+
dx
x e
e +e
e
n
y
n=1
Z y

X
(1)n+1 2 xn yn
T4
exn eyn
x e e
dx yx2 x3 +
= 2
n
0
n=1
Z X

(1)n+1 2 xn yn
+
dx
e + exn eyn .
x e
n
0
n=1
All the x-integrals can be carried out:

Z y
1
1 4
1
2
3
dx (yx x ) =
y4 =
y ,
3
4
12
0

Z y
2
2
2y
y2
2 x
y
,
2+
dx x e
= 3 +e
Z 0
2
.
dx x2 exn =
3
n
0
92
(7.45)
(7.46)
(7.47)
(7.48)
Inserting these into Eq. (7.45) we get

f (T, ) =

2
2
y 4 X (1)n+1 yn
2y y 2
yn
3 +e
e
2+
+
12 n=1
n
n
n3
n
n

2
2
2
2y
y2
2
eyn 3 + eyn 3 2
+ eyn 3 + eyn 3
n
n
n
n
n
n

2y 2
T 4 y 4 X (1)n+1 4
,
+
+
2
12 n=1
n
n3
n
T4
2
(7.49)
where a remarkable cancellation took place. The sums can be carried out:
(2)
X
(1)n+1
1
1
1
1
= 2 2 + 2 2 +
2
n
1
2
3
4
n=1
2
1
2
(2)
=
(2)
=
,
22
2
12
X (1)n+1
1
1
1
1
= 4 4 + 4 4 +
4
n
1
2
3
4
n=1
= (2)
(4)
= (4)
2
7
7 4
(4) = (4) =
.
4
2
8
8 90
(7.50)
(7.51)
Inserting into Eq. (7.49), we end up with

T4
f (T, ) = 2
y4
2 2 7 4
+
y +
12
6
180
which after the substitution y = /T reproduces Eq. (7.41).
93
(7.52)
8.
Real-time observables
We now move to a new class of observables, those including both a Minkowskian time t and a
temperature T . Examples are production rates from a thermal plasma of various types of weaklyinteracting particles; rates of oscillation and damping of waves travelling in the plasma; as well as
transport coefficients such as electric and thermal conductivity and bulk and shear viscosity. We
start by developing some aspects of the general formalism needed for treating these observables,
and return later on to specific applications.
8.1.
Different Greens functions
Practically all the observables of interest in the following can be reduced to two-point correlation
functions of elementary or composite operators. Let us therefore list some common definitions and
relations that apply to such correlation functions20 .
We denote Minkowskian space-time coordinates by x = (t, xi ) and momenta by Q = (q 0 , q i ),
= (
while their Euclidean counterparts are denoted by x
= (, xi ), Q
q0 , qi ). Wick rotation is
0
carried out by it, q0 iq . Scalar products are defined as Q x = q 0 t + q i xi = q 0 t q x,
x
Q
= q0 + qi xi = q0 q x. Arguments of operators denote implicitely whether we are in
Minkowskian or Euclidean space-time. In particular, Heisenberg-operators are defined as

x) eiHt
O(t,
O(0, x)eiHt ,
x) eH
O(,
O(0, x)eH .
(8.1)
The thermal ensemble is defined by the density matrix = Z 1 exp( H).
Bosonic case
We first consider operators that are bosonic in nature, i.e. commuting (modulo possible contact
terms). We denote the operators which appear in the two-point functions by (x), (x).
We can define various classes of correlation functions. The physical correlators are defined as
Z
E
D
>
(8.2)
(Q)
dt d3 x eiQx (x) (0) ,
Z
D
E
(8.3)
dt d3 x eiQx (0) (x) ,
<
(Q)
Z
D1 h
iE
(Q)
dt d3 x eiQx
(x), (0) ,
(8.4)
2
Z
oE
D1 n
(x), (0)
,
(8.5)
(Q)
dt d3 x eiQx
2
where is called the spectral function, while the retarded/advanced correlators can be
defined as
Z
Dh
i
E
R
(Q)
i
dt d3 x eiQx (x), (0) (t) ,
(8.6)
Z
D h
i
E
(8.7)
A
dt d3 x eiQx (x), (0) (t) .
(Q) i
20 A.L. Fetter and J.D. Walecka, Quantum Theory of Many-Particle Systems (McGraw-Hill, New York, 1971);
S. Doniach and E.H. Sondheimer, Greens Functions for Solid State Physicists (Benjamin, Reading, 1974);
J.W. Negele and H. Orland, Quantum Many Particle Systems (Addison-Wesley, Redwood City, 1988).
94
On the other hand, from the computational point of view one is often faced with time-ordered
correlation functions,
Z
D
E
T
(8.8)
(Q)
dt d3 x eiQx (x) (0)(t) + (0) (x)(t) ,
which appear in time-dependent perturbation theory, or with the Euclidean correlator
E
(Q)
D
E
d3 x eiQx (
x) (0) ,
(8.9)
which appears in non-perturbative formulations. Note that the Euclidean correlator is also timeordered by definition, and can be computed with standard imaginary-time functional integrals in
the Matsubara formalism; q0 is a bosonic Matsubara frequency
It follows from Eq. (8.1) that
i
1 h
h (, x) (0, 0)i = Tr e H e H (0, x)e H (0, 0) = h (0, 0) (0, x)i .

Z
(8.10)
This is a version of the so-called Kubo-Martin-Schwinger (KMS) relation, which in general relates
<
>
and to each other.
More generally, all of the correlation functions defined above can be related to each other. In
particular, all correlators can be expressed in terms of the spectral function, which in turn can be
determined as a certain analytic continuation of the Euclidean correlator. In order to do this, we
<
may first insert sets of energy eigenstates into the definitions of >
and :
>
(Q)
=
=
<
(Q) =
1
Z
1
Z
XZ
1
Z
eiqx
1
Z
dt d3 x eiQx Tr e H
1
Z
XZ
1
Z
dt d3 x eiQx Tr e H+iHt
=
=
(0, x)eiHt
|mi hm|
1 (0, 0)
|{z}
|ni hn|
dt d3 x eiQx e(+it)Em eitEn hm| (0, x)|ni hn| (0, 0)|mi
m,n
eEm 2 (q 0 + Em En )hm| (0, x)|ni hn| (0, 0)|mi , (8.11)
m,n
1 (0, 0)eiHt
|{z}
1
|{z}
|ni hn|
1
|{z}
(0, x)eiHt
|mi hm|
dt d3 x eiQx e(it)En eitEm hn| (0, 0)|mi hm| (0, x)|ni
(8.12)
m,n
eiqx
x
0
eEn 2 (q 0 + Em En )hm| (0, x)|ni hn| (0, 0)|mi

{z
}
|
m,n
En =Em +q0
eq >
(Q) .
(8.13)
This is the Fourier-space version of the KMS relation. Consequently

(Q) =
1 >
1 q0
[ (Q) <
(e
1)<
(Q)] =
(Q)
2
2
(8.14)
and, conversely,
0
<
(Q) = 2nB (q ) (Q) ,
>
(Q) = 2
(8.15)
q0
e
(Q) = 2[1 + nB (q 0 )] (Q) ,
eq0 1
95
(8.16)
where nB (x) 1/[exp(x) 1]. Moreover,

(Q) =
1 >
0
[ (Q) + <
(Q)] = [1 + 2nB (q )] (Q) .
2
(8.17)
Note that 1 + 2nB (q 0 ) = [1 + 2nB (q 0 )], so that if is odd in Q Q, then is even.

Inserting the representation
(t) = i
d eit
+
2 + i0
(8.18)
into the definitions of R , A , we obtain

Z
Z 4
d P iP x
3
iQx
R
(Q)
=
i
dt
d
x
e
2(t)
e
(P )
(2)4
Z
Z
Z
0
0
d dp0 ei(q p )t
(p0 , q)
= 2 dt
2
2 + i0+
Z
Z
d dp0 2(q 0 p0 )
(p0 , q)
= 2
2
2
+ i0+
Z 0
dp (p0 , q)
=
,
0
0
+
p q i0
and similarly
A
(Q)
Making use of
(8.19)
dp0 (p0 , q)
.
0
0
+
p q + i0
(8.20)
1
1
=
P
i() ,
i0+
(8.21)
we find
Im R
(Q) = (Q) ,
Im A
(Q) = (Q) .
(8.22)
A
Furthermore, the real parts of R and A agree, so that i[R
(Q) (Q)] = 2 (Q).
Moving on to T , and making use of the inverse transforms of Eqs. (8.15), (8.16), we find
Z
Z 4
i
0
d P iP x h
(t)2ep nB (p0 ) + (t)2nB (p0 ) (P )
e
T (Q) =
dt d3 x eiQx
4
(2)

Z
Z
Z
0
0
0
0
d dp0 ei(q p )t p0 ei(q p +)t
= 2i dt
nB (p0 ) (p0 , q)
e
+
2
2
+ i0+
+ i0+

Z
Z
d dp0 2(q 0 p0 ) p0 2(q 0 p0 + )
nB (p0 ) (p0 , q)
e
+
= 2i
2
2
+ i0+
+ i0+

Z
0
dp0
ep
1
= i
nB (p0 ) (p0 , q)
q 0 p0 + i0+ q 0 p0 i0+
Z 0
dp i (p0 , q)
=
+ 2 (q 0 , q)nB (q 0 )
0 p0 + i0+
<
= iR
(Q) + (Q) ,
(8.23)
0
where in the penultimate step we inserted Eq. (8.21) as well as the identity nB (p0 )ep = 1+nB (p0 ).
Finally, we note that the sums in Eq. (8.11) are exponentially convergent for 0 < it < .
Therefore we can relate the two functions
D
E
D
E
(x) (0)
and
(
x) (0)
(8.24)
96
by a direct analytic continuation t i , ot it , with 0 < < . Thereby

Z 4

Z Z
d P iP x > 0
x
E
3
iQ
e
(p , p)
(Q) =
d d x e
(2)4
0
it
Z
Z
0
dp p0 > 0
d eiq0
=
e
(p , q)
0
2
Z
Z
0
dp0 p0 2ep
(p0 , q)
e
=
d eiq0
ep0 1
0
2

Z
0
dp0 (p0 , q) e(iq0 p )
=
p0
i
q0 p0 0
1 e
Z
0
dp0 (p0 , q) ep 1
=
p0 i
q0 p0
1 e
Z 0
dp (p0 , q)
,
=
p0 i
q0

(8.25)
where we inserted Eq. (8.16) for > (Q), and changed orders of integration. This relation is called
the spectral representation of the Euclidean correlator.
It is useful to note that Eq. (8.25) implies the existence of a simple sum rule:
Z
Z 0
dq (q 0 , q)
=
d E
(, q) .
0
q
0
(8.26)
Here we set q0 = 0 and used the definition in Eq. (8.9) on the left-hand side of Eq. (8.25). The
usefulness of the sum rule is that it relates directly (integrals over) Minkowskian and Euclidean
correlators to each other.
Finally, the spectral representation can formally be inverted by making use of Eq. (8.21),
(q 0 , q)
1
Disc E
q0 iq 0 , q)
(
2i
i
1h E
0
+
(i[q 0 + i0+ ], q) E
(i[q i0 ], q) .
2i
(8.27)
(8.28)
Furthermore, a comparison of Eqs. (8.19) and (8.25) shows that

E
R
q0 i[q 0 + i0+ ], q) .
(Q) = (
(8.29)
In the context of the spectral representation, Eq. (8.25), it will often be useful to note from
Eq. (4.75), viz.
i
X eib
nB () h ( )
(8.30)
e
+ e ,
=
T
2
2
b +
2
that, for 0 < < ,

T
X
b
1
eib
0
p ib
X ib + p0
ib
2 + (p0 )2 e
( + p0 )T
X
b
b2
eib
+ (p0 )2
i
nB (p ) h
0
0 ( )p0
0
0 p0
(p
+
p
)e
+
(p
+
p
)e
2p0
nB (p0 )e p .
97
(8.31)
This relation turns out to be valid both for both p0 < 0 and p0 > 0 (to show this, substitute
b b and use Eq. (8.32)). We also note that, again for 0 < < ,
T
X
b
p0
1
eib
ib
= T
X
b
0
1
eib ( ) = nB (p0 )e( )p .
p0 ib
(8.32)
P
In particular, taking the inverse Fourier transform (T q0 eiq0 ) from the left-hand side of
Eq. (8.25), and employing Eq. (8.32), we get the relation
Z
D
E
d3 x eiqx (, x) (0, 0)
Z 0
0
dp
(p0 , q)nB (p0 )e( )p
=

h
h
i
i
0
Z 0
cosh
p0
0
0 sinh
0
0
2
2
(p ) + (p )
dp
(p ) (p )

=
,
+

2
2
0
sinh p0
sinh p0
2
(8.33)
where we symmetrised and anti-symmetrised the kernel nB (p0 )e( )p with respect to p0 , and
for brevity left out the argument q from on the last line. Normally (when and are
identical) the spectral function is antisymmetric in p0 p0 , and then only the second term on
the last line of Eq. (8.33) contributes. Thereby we obtain a potentially very powerful identity: if
the left-hand side of Eq. (8.33) can be measured non-perturbatively on a Euclidean lattice with
Monte Carlo simulations as a function of , then an inversion of Eq. (8.33) could lead to a
non-perturbative estimate of the Minkowskian spectral function.
Example
Let us illustrate the use of some of the relations obtained above with the example of a free propagator in scalar field theory:
E (Q) =
=
1
q02 + Eq2

1
1
1
,
+
2Eq i
q0 + Eq
i
q0 + Eq
p
m2 + q2 . According to Eqs. (8.29), (8.21)

1
1
1
R
+
(Q) =
2Eq q 0 + Eq + i0+
q 0 + Eq i0+

h
i
1
1
1
0
0
=
P 0
P 0
+ i (q Eq ) (q + Eq )
2Eq
q + Eq
q Eq

i
i h 0
1
+
= P
(q Eq ) (q 0 + Eq ) ,
0
2
2
(q ) Eq
2Eq
(8.34)
where Eq =
(8.35)
and according to Eq. (8.22),

(Q) =
i
h 0
(q Eq ) (q 0 + Eq ) .
2Eq
(8.36)
Finally, according to Eq. (8.23),

o

n 0
i
0
0
0
+
(q
E
)[1
+
2n
(q
)]
(q
+
E
)[1
+
2n
(q
)]
T (Q) = P
q
B
q
B
(q 0 )2 Eq2
2Eq
98
=
=
=

i
h 0
i
0
+
(q
E
)
+
(q
+
E
)
[1 + 2nB (|q 0 |)]
q
q
(q 0 )2 Eq2
2Eq

i
0 2
2
0
P
+
(q
)
E
q [1 + 2nB (|q |)]
(q 0 )2 Eq2

i
0 2
2
0
+
2
(q
)
E
q nB (|q |) ,
(q 0 )2 Eq2 + i0+
P
(8.37)
where in the second step we made use of the identity 1 + 2nB (Eq ) = [1 + 2nB (Eq )].
It is useful to note that Eq. (8.37) is, in some sense, closely related to the relation in Eq. (2.34).
However, Eq. (2.34) is true in general, while Eq. (8.37) was derived for a special case; thus it is
not always true that the thermal effects can simply be obtained by replacing the zero-temperature
time-ordered propagator by Eq. (8.37), even if surprisingly often such a simple recipy does indeed
function.
Fermionic case
Let us next consider two-point correlation functions built out of fermionic operators20. In contrast
to the bosonic case, however, we take for generality the density matrix to be of the form =
Q)].
Z 1 exp[(H
We denote the operators which appear in the two-point functions by j (x), j (x). They could be
elementary field operators, in which case the indices , label Dirac and/or flavour components,
but they could also be composite operators consisting of a product of elementary field operators.
Nevertheless, we assume the validity of the relation
= j (0, x) .
[j (0, x), Q]
(8.38)
To motivate this, note that for j , j = , the canonical commutation relations of Eq. (4.32),
{ (x0 , x), (x0 , y)} = (d) (x y) ,
and the expression for the conserved charge in Eq. (7.32),
Z
Z
d
Q = d x 0 = dd x ,
(8.39)
(8.40)
B
C]
= AB
C
B
C A = AB
C+
B
AC
B
AC
B
C A = {A,
B}
C
B{
A,
C},
as well as the identity [A,

imply that in this case Eq. (8.38) is indeed satisfied. Eq. (8.38) implies that
X
X
1
1
n j (0, x) =
+ 1)n = j (0, x)eQ e ,
()n (Q)
()n j (0, x)(Q
n!
n!
n=0
n=0
(8.41)
and consequently that
D
E
i
1 h (H
Q)
j (, x)j (0, 0)
=
Tr e
e j (0, x)e H j (0, 0)
Z
i
1 h
=
Tr j (0, x)e e(HQ) j (0, 0)
Z
i
1 h
=
e Tr j (0, x)e(HQ) j (0, 0)
Z D
E
(8.42)
= e j (0, 0)j (0, x) .
eQj (0, x) =
This is a fermionic version of the KMS relation.

99
With this setting, we can again define various classes of correlation functions, like in the bosonic
case. The physical correlators are now set up as
Z
D
E
3
iQx
j (0) ,
>
(Q)
dt
d
x
e
j
(x)
(8.43)
Z
D
E
(8.44)
dt d3 x eiQx j (0)j (x) ,
<
(Q)
Z
D1 n
oE
j (x), j (0)
(Q)
dt d3 x eiQx
,
(8.45)
2
Z
D1 h
iE
j (x), j (0) ,
(Q)
dt d3 x eiQx
(8.46)
2
where is the spectral function, while retarded and advanced correlators can be defined as
Z
Dn
o
E
R
(Q) i dt d3 x eiQx j (x), j (0) (t) ,
(8.47)
Z
D n
o
E
A
dt d3 x eiQx j (x), j (0) (t) .
(8.48)
(Q) i
On the other hand, the time-ordered correlation function reads
Z
D
E
T (Q)
dt d3 x eiQx j (x)j (0)(t) j (0)j (x)(t) ,
(8.49)
while the Euclidean correlator is
E
(Q)
D
E
d3 x e(iq0 ) iqx j (
x)j (0) .
(8.50)
Note again that the Euclidean correlator is time-ordered by definition, and can be computed
with standard imaginary-time functional integrals in the Matsubara formalism. The Matsubara
frequencies q0 in Eq. (8.50) are fermionic, and the additional factor in the exponential in Eq. (8.50)
is chosen in order to cancel the extra multiplicative factor in Eq. (8.42).
We can establish relations between the different Greens functions just like in the bosonic case:
Z
i
h
1
1
1 j (0, 0)
eQ j (0, x)eiHt |{z}
dt d3 x eiQx Tr e H+iHt |{z}
>
(Q)
=
Z
P
P
m |mi hm|
n |ni hn|
Z
1 X
=
dt d3 x eiQx e(+it)Em eitEn e hm|j (0, x)eQ |ni hn|j (0, 0)|mi
Z m,n
Z
X
1
=
eiqx
e(Em ) 2 (q 0 + Em En )hm|j (0, x)eQ |ni hn|j (0, 0)|mi ,
Z x
m,n
(8.51)
<
(Q) =
1
Z
dt d3 x eiQx Tr e H eQ
1
|{z}
P
n
XZ
j (0, 0)eiHt
|ni hn|
1
|{z}
j (0, x)eiHt
|mi hm|
dt d3 x eiQx e(it)En eitEm hn|j (0, 0)|mi hm|j (0, x)eQ |ni
m,n
eiqx
e(q
+)
eEn 2 (q 0 + Em En )hm|j (0, x)eQ |ni hn|j (0, 0)|mi

{z
}
|
m,n
En =Em +q0
>
(Q) .
(8.52)
<
Consequently, (Q) = [>
(Q) (Q)]/2 and, conversely,
0
>
(Q) = 2[1 nF (q + )] (Q) ,
100
0
<
(Q) = 2nF (q + ) (Q) ,
(8.53)
where nF (x) 1/[exp(x) + 1]. Moreover, (Q) = [1 2nF (q 0 + )] (Q), in which relation
the combination [1 2nF (q 0 + )] has a specific symmetry property: [1 2nF (q 0 )] = [1
2nF (q 0 + )].
The relation of R , A and T to the spectral function can now be derived in complete analogy
with Eqs. (8.18)(8.23). For brevity we only cite the final results:
Z 0
Z 0
dp (p0 , q)
dp (p0 , q)
A
,
(Q)
=
,
(8.54)
R
(Q)
=
0
0
+
0
0
+
p q + i0
p q i0
Z 0
dp i (p0 , q)
T (Q) =
2nF (q 0 + ) (q 0 , q)
0 p0 + i0+
<
iR
(Q) + (Q) .
(8.55)
Note that when written in a generic form, where no distribution functions are visible, the end
results are identical with the bosonic ones.
Finally, writing the argument inside the -integration in Eq. (8.50) as a Wick rotation of the
inverse Fourier transform of the left-hand side of Eq. (8.43), inserting Eq. (8.53), and changing
orders of integration, we get
Z
Z
dp0 p0 > 0
(i
q0 )
=
d
e
E
(
Q)
e
(p , q)
0
2
Z
Z
0
dp0 p0 2e(p +)
(i
q0 )
e
(p0 , q)
=
d e
e(p0 +) + 1
0
2
Z
Z
0
0
dp0 e(p +)
0
d e(iq0 p )
(p
,
q)
=
(p0 +) + 1
e
0
(iq0 p0 )
Z
0
dp0 e(p +)
e
=
(p0 , q)
(p0 +) + 1
i
q0 p0 0
e
Z
0
0
dp0 e(p +)
e(p +) 1
0
=
(p , q)
(p0 +) + 1
i
q0 p0
e
Z 0
0
(p , q)
dp
=
.
(8.56)
0 i[
p
q0 + i]
Like in the bosonic case, this relation can be inverted by making use of Eq. (8.21),
1
1
Disc E
q0 + i i[q 0 + i0+ ], q) =
Disc R
(
(Q) ,
2i
2i
where the discontinuity is defined like in Eq. (8.28).
(q 0 , q) =
(8.57)
We also note that the fermionic Matsubara sum over the structure in Eq. (8.56) can be carried
out explicitly. This could be verified by making use of Eq. (4.76), in analogy with the bosonic
analysis in Eqs. (8.31), (8.32), but let us for a change proceed in another way. We may recall, first
of all, that
X
T
eib = ( mod ) .
(8.58)
b

According to Eq. (4.54), viz. Sf (T ) = 2Sb T2 Sb (T ), we can write
X
T
eif = 2( mod 2) ( mod ) .
(8.59)
We assume for a moment that p0 + > 0. Employing the representation

Z
1
=
ds e(+i)s , > 0 ,
+ i
0
101
(8.60)
and inserting subsequently Eq. (8.59), we get

Z
X
X
0
1
if
T
e
=
ds T
eif p ss+if s
0
p + if
0
f
f
Z
h
i
0
=
ds e(p +)s 2( + s mod 2) ( + s mod )
0
(p0 +)( +2n)
n=1
0
e(p
+)( +n)
n=1

X
X
(p0 +)n
2(p0 +)n
+)
e
2
e
n=1
n=1
e(p
e(p
+)
{z
0
e2(p +)
0
1e2(p +)
{z
{z
0
e(p +)
0
1e(p +)
2
1
(p0 +)
0
0
(e(p +) 1)(e(p +) +1)
e
1
nF (p0 + ) ,
}
}
(8.61)
where we assumed 0 < < . As an immediate consequence,

T
X
f
p0
X
0
1
1
eif = T
eif ( ) = e( )(p +) nF (p0 + ) .
0 + i
+ if
p
f
(8.62)
Furthermore, it is not too difficult to show (by substituting f f ) that these relations continue
to hold also for p0 + < 0.
As a consequence of Eq. (8.61), let us note that
T
X
f
ei(f i)
(f i)2 + 2
1
( if )( + if + )
f

X
1
1
1
+
e T
eif
2 + + if
if
f
i
e h (+)
e
nF ( ) e() nF ( )
2
i
e h ( )(+)
e
nF ( + ) e () nF ( )
2
i
1 h
nF ( + )e( )+ nF ( )e ,
2
= e T
=
=
=
=
eif
(8.63)
which constitutes a generalization of Eq. (4.76) to the case of a finite chemical potential.
Example
Let us illustrate some of the relations obtained by considering the structure of the free fermion
propagator in the presence of a chemical potential. With fermions, in the presence of 6= 0,
one unfortunately has to be extremely careful with definitions. Suppressing spatial momenta and
indices, Eq. (5.47) and the presence of a chemical potential a
` la Eq. (7.34) imply that the free
propagator can be written in the form
)(0)i
=T
h(
ei(p0 i)
p0f
102
iA(
p0 i) + B
,
(
p0 i)2 + E 2
(8.64)
where an additional exponential has been inserted into the Fourier transform, in order to respect
the property in Eq. (8.42). The correlator in Eq. (8.50) then becomes
Z
X
iA(
p0 i) + B
E
ei(p0 i)
(
q0 ) =
d e(iq0 ) T
(
p0 i)2 + E 2
0
p
0f
iA(
q0 + i) + B
.
(
q0 + i)2 + E 2
(8.65)
According to Eq. (8.57),

1 A(q 0 + i0+ ) + B
A(q 0 i0+ ) + B
0
(q ) =
2i (q 0 + i0+ )2 + E 2
(q 0 i0+ )2 + E 2

0
+
1 1 A(q + i0 ) + B
A(q 0 + i0+ ) + B
A(q 0 i0+ ) + B
A(q 0 i0+ ) + B
=
+
2i 2E
q 0 + i0+ + E
q 0 + i0+ E
q 0 i0+ + E
q 0 i0+ E

1 1
AE + B
AE + B
AE + B
AE + B
=
0
0
+ 0
2i 2E q 0 + i0+ + E
q + i0+ E
q i0+ + E
q i0+ E

0
0
(q E) AE + B + AE + B + (q + E) AE B + AE B
=
4E
h
i
=
(q 0 E) (q 0 + E) (Aq 0 + B) .
(8.66)
2E
Note that the tree-level spectral function is independent of the temperature and of the chemical
potential! The retarded propagator reads
R (q 0 ) =
A(q 0 + i0+ ) + B
,
(q 0 + i0+ )2 + E 2
(8.67)
and, from Eqs. (8.53), (8.55), the time-ordered propagator can be determined after a few steps:

1
1
i
T 0
0
+
(q ) = (Aq + B)
2E E q 0 i0+ E + q 0 + i0+
h
i
2
0
0
0
nF (q + ) (q E) (q + E)
2E

1
1
Aq 0 + B
iP
iP
=
2E
E q0
E + q0
h
i
h
i
0
0
0
0
+(q E) 1 2nF (q + ) (q + E) 1 2nF (q + )

Aq 0 + B
2E
=
iP
2E
E 2 (q 0 )2
h
i
h
i
0
0
0
0
+(q E) 1 2nF (q + ) + (q + E) 1 2nF (q )

2E
Aq 0 + B
0 2
2
iP
+
2E
(q
)
E
=
2E
E 2 (q 0 )2
h
i
0
0
0
0
2 (q E)nF (q + )] + (q + E)nF (q )

i
0 2
2
0
0
0
2 (q ) E nF |q | + sign(q )
= (Aq + B)
. (8.68)
(q 0 )2 E 2 + i0+
All temperature and density effects are again seen to reside in an on-shell part and, to some
extent, one could hope to account for finite temperature effects simply by replacing free zerotemperature propagators through Eq. (8.68); the proper procedure, however, is to carry out the
analytic continuation for the complete observable considered, and this may not always amount to
the simple replacement through Eq. (8.68) of all the free propagators appearing in the graph.
103
8.2.
From Euclidean correlator to spectral function
As an application of the relations given in Sec. 8.1, let us carry out an explicit computation
illustrating the steps. This computation will also find a practical application in Sec. 9.5.
) interacting
To motivate the correlator, consider a model describing a right-handed neutrino (N
with the usual left-handed leptons (L , = 1, 2, 3) and the Higgs field ( = i2 ) through Yukawa
interactions, with dimensionless but possibly complex coupling constants h . The Minkowskian
Lagrangian can be written as
LM =
1

N
h N
aR N
h L
1 MN
Ni /
N
aL L ,
2
2
(8.69)
where repeated indices are summed over, and aL (15 )/2, aR (1+5 )/2 are chiral projectors.
Let us now consider a correlator that would play the role of a self-energy correction for the
right-handed neutrino,
Z Z
(0)(0)i

eiQx aL h (
x)L (
x) L
aR .
(8.70)
E
(
Q)
d
is fermionic. In the Standard Model, the Higgs and lepton doublets have the forms
Here Q

1
0 + i3
.
(8.71)
, L =
=
2 + i1
2
Carrying out the contractions, we can write
Z Z
3
X
x
iQ
=
E
(
Q)
e
d
aL hc (
x)c (0)i0 hc (
x)c (0)i0 aR
2 0
x
c=0
Z Z
3 Z
P
iP/ + mc
1
X

ei(Q+P +R)x aL
aR
d
=
2
2
2
2 c=0 0
P + m R + m2
x

Pf ,Rb
3 Z
X
c=0
X iP/ aR
1
,
2
2
p0 + E1 (
p0 + q0 )2 + E22
(8.72)
p0f
where we inserted the free scalar and fermion propagators, and denoted
q
q
E1 m2c + p2 , E2 m2c + (p + q)2 .
(8.73)
Moreover the left and right projectors removed the mass term from the numerator. We have been
somewhat implicit about the assignments of the masses mc , mc to the corresponding fields, but
for our purposes more details are not needed.
The essential issue in handling Eq. (8.72) is the treatment of the Matsubara sum. More generally,
let us inspect the structure
F =T
X
p0f
[
p20
f (i
p0 , p, i
q0 , q)
,
2
+ E1 ][(
p0 + q0 )2 + E22 ]
(8.74)
where we assume that the function f in the numerator depends on its arguments at most linearly.
We can write:
X X
f (i
p0 , p, i
q0 , q)
(
r0b p0f q0f ) 2
T
F = T
2 ][
2 + E2]
[
p
+
E
r
0
1
0
2
r0b
p0f
X ir
X
Z
e 0b
f (i
p0 , p, i
q0 , q)
T
,
(8.75)
eip0f
=
d eiq0f T
2
2
p0 + E1
r02 + E22
0
r0b
p0f
104
where we wrote
(
r0b p0f q0f ) =
d ei(r0b p0f q0f ) .
(8.76)
Now we can make use of Eqs. (8.30), (8.63) and time derivatives thereof:
X eir0b
r02 + E22
i
nB (E2 ) h ( )E2
+ e E2 ,
e
2E2
(8.77)
X eip0
p20 + E12
i
nF (E1 ) h ( )E1
e E1 ,
e
2E1
(8.78)
X i
p0 eip0
i
nF (E1 ) h
E1 e( )E1 + E1 e E1 .
2E1
(8.79)
r0b
p0f
p0f
p20
E12
Thereby
Z
F=
nF (E1 )nB (E2 )
4E1 E2

e( )(E1+E2 ) f (E1 , p, i
q0 , q)
d eiq0f
q0 , q)
e( )E2+ E1 f (E1 , p, i
q0 , q)
e( )E1+ E2 f (E1 , p, i
(E1 +E2 )
f (E1 , p, i
q0, q)
(8.80)
As an example, let us focus on the third structure in Eq. (8.80). The -integral can be carried
out, noting that q0f is fermionic:
Z
d eE1 e (iq0f E1 +E2 )
=
=
Thus
F |3rd =
h
i
eE1
e(E2 E1 ) 1
i
q0 E1 + E2
eE2 + eE1
i
q0 + E1 E2
h
i
1
1
n1
B (E2 ) + nF (E1 ) .
i
q0 + E1 E2
i f (E , p, i
1 h
q0 , q)
1
nF (E1 ) + nB (E2 )
.
4E1 E2
i
q0 + E1 E2
(8.81)
(8.82)
Finally we set q0 i(q 0 + i0+ ) and take the imaginary part according to Eq. (8.28). Furthermore, making use of Eq. (8.21), we note that
i
1
1
1h
= (q 0 + ) ,
2i q 0 + + i0+
q 0 + i0+
(8.83)
whereby the denominator in Eq. (8.82) simply gets replaced with () times a Dirac delta function.
Special attention needs to be paid here to the possibility that q0 could also appear in the numerator
in Eq. (8.82); however, we can then write
i
q0 = i
q + E E2 +E2 E1 ,
| 0 {z1
}
no discontinuity
105
(8.84)
so that in total
n
o

Im F(i
q0 q 0 + i0+ )
=
3rd
i
h
nF (E1 ) + nB (E2 ) (q 0 + E1 E2 )f (E1 , p, E2 E1 , q)
| {z }
4E1 E2
q0
E1
E2
e
+e
1
0
2(q 0 + E1 E2 )f (E1 , p, q 0 , q)n1
F (q ) (E2 E1 )
2 4E1 E2
[e
+ 1](eE1 + 1)(eE2 1)
eE1 [1 + e(E2 E1 ) ]
1
0
2(q 0 + E1 E2 )f (E1 , p, q 0 , q)n1
F (q ) (E2 E1 )
2 4E1 E2
[e
+ 1](eE1 + 1)(eE2 1)
1
0
=
2(q 0 + E1 E2 )f (E1 , p, q 0 , q)n1
(8.85)
F (q )nB (E2 )[1 nF (E1 )] .
2 4E1 E2
q
Moreover, we remember that E2 = m2c + (p + q)2 , so that we can write
=
g(E2 ) =
q

d3 p2
3 (3)
2 + p2 .
m
(2)
(q
+
p
p
)g
2
2
c
(2)3
(8.86)
Let us now return to Eq. (8.72). We had there the object iP/ , which now plays the role of the
function f in the analysis above, and according to Eq. (8.85) becomes
iP/ = i
p0 0 + ipj j E1 0 + ipj (i j ) = P/ ,
(8.87)
where we also made use of the definition of the Euclidean Dirac-matrices in Eq. (4.35) (Eq. (8.85)
shows that any possible iQ
/ can also be replaced by Q
/ ). Furthermore, the factors 1/2 in Eq. (8.72)
and (8.85) combine into 1/4. In total, then, the spectral function becomes
(Q) =
Z
Z
3
d3 p1
d3 p2
X 1 0
P/1 aR
nF (q )
3
4 c=0
(2) 2E1
(2)3 2E2

1
(2)4 (4) (P1 + P2 Q) nF1 nB2 + 2
Q
1
+ (2)4 (4) (P2 P1 Q) nB2 (1 nF1 ) +
Q
2
+ (2)4 (4) (P1 P2 Q) nF1 (1 + nB2 ) +

4 (4)
+ (2) (P1 + P2 + Q) (1 nF1 )(1 + nB2 ) ,
Q
1
Q
2
(8.88)
where we renamed P P1 and added the results of other channels as well. Furthermore, we
denoted nFi nF (Ei ), nBi nB (Ei ). The graphs in Eq. (8.88) illustrate the various processes
that the energy-momentum conserving delta functions correspond to, with a dashed line indicating
a scalar particle, a solid line a lepton, and a dotted line the right-handed neutrino.
Eq. (8.88) is for the moment our final result. We return, however, to the evaluation of the phase
space integrals in Sec. 9.5.
As a final remark, we note that the spectral function has the important property that, in a
CP-symmetric situation, it is even in Q:
(Q) = (Q) .
(8.89)
Let us demonstrate this explicitly with the 3rd channel in Eq. (8.88). The energy-dependent part
reads
0
0
n1
F (q )(E1 E2 q )nF1 (1 + nB2 )
QQ
0
0
n1
F (q )(E1 E2 + q )nF1 (1 + nB2 )
106
=
=
=
=
(eq + 1)eE2
(E1 E2 + q ) E1
+ 1)(eE2 1)
(e
0
e(E2 q )
E
1
+ 1)(eE2 1)
(e
eE1
0
(E1 E2 + q 0 )n1
F (q ) E1
+ 1)(eE2 1)
(e
0
(E1 E2 + q 0 )(eq + 1)
0
0
n1
F (q )(E2 E1 q )nB2 (1 nF1 ) .
(8.90)
We see that we get back the structure of the 2nd channel, which indeed is the desired result,
because that is precisely what we need from the point of view of changing q q in the delta
function conserving the spatial momentum. Similarly, it can be checked that the 4th term goes
over into the 1st term, and vica versa.
107
8.3.
Exercise 11
In the text we made use of tree-level propagators, but in general the propagators need to be
resummed, and obtain a more complicated form. In this situation it may be useful to express them
as in the spectral representation of Eq. (8.25). In particular, the scalar propagator can in general
be written as
Z 0
dq S (q 0 , q)
(P + Q)
,
(8.91)
=
(
P
+
Q)
h(P )(Q)i0 = 2
0
q0 + q2 + S (
q0 , q)
q0
q i
while the fermion propagator contains two possible structures (or, if chirality is not a symmetry,
even more)21 ,

iQ
/
0
= (P Q) 2
+
q0 + q2 + L (
q0 , q) q02 + q2 + U (
q0 , q)

Z 0
Z 0
0
0
(q
,
q)
dq
dq
L
U (q , q)
iQ
= (P Q)
/
,
+
u
/
0
0
q0
q0
q i
q i

P )(
h(
Q)i0
(8.92)
where u = (1, 0) is the plasma four-velocity. Carry out the steps from Eq. (8.72) to (8.88) in this
situation.
The structure in Eq. (8.74) now becomes
X X Z d1 Z d2 fF (i
p0 , p, u)F (1 , p)S (2 , p + q)
,
F =T
[1 i
p0 ][2 i(
p0 + q0 )]
(8.93)
p0f F =L,U
where we assume that the function fF in the numerator depends on its arguments at most linearly.
We can write:
X Z d1 d2
F (1 , p)S (2 , p + q)
F =
2
F =L,U
p0f
X
r0b
(
r0b p0f q0f )
fF (i
p0 , p, u)
.
[1 i
p0 ][2 i
r0 ]
(8.94)
Employing Eqs. (8.76), (8.31) and (8.62), as well as the time derivative of the last one,
T
X
f
i
dh
if
eif =
nF (E)e( )E = nF (E) E e( )E ,
E if
d
(8.95)
we get
F
X Z
F =L,U
d1 d2
F (1 , p)S (2 , p + q)
2
d eiq0f nF (1 )nB (2 )fF (1 , p, u)e( )1+ 2 .
(8.96)
21 H.A. Weldon, Effective Fermion Masses of O(gT ) in High Temperature Gauge Theories with Exact Chiral
Invariance, Phys. Rev. D 26 (1982) 2789.
108
The -integral can be carried out, noting that q0f is fermionic:

Z
d nF (1 )nB (2 ) e1 e (iq0f 1 +2 )
=
=
=
i
nF (1 )nB (2 )e1 h (2 1 )
e
1
i
q0 1 + 2
i
nF (1 )nB (2 ) h 2
+ e1
e
i
q0 + 1 2
i
nF (1 )nB (2 ) h 1
nF (1 ) + n1
(
)
2
B
i
q0 + 1 2
h
i
1
nF (1 ) + nB (2 ) .
(8.97)
i
q0 + 1 2
Finally we set q0 i(q 0 + i0+ ) and take the imaginary part. Making use of Eq. (8.83), the
denominator in Eq. (8.97) simply gets replaced with () times a Dirac delta function. Thus, in
total,
n
o
X Z d1 d2
Im F(i
q0 q 0 + i0+ )
=
F (1 , p)S (2 , p + q)
2
F =L,U
h
i
nF (1 ) + nB (2 ) (q 0 + 1 2 )fF (1 , p, u)
Z
d1 d2
1 X
F (1 , p)S (2 , p + q)
=
2
2
F =L,U
0
2 (q 0 + 1 2 )fL (1 , p, u)n1
F (q )nB (2 )[1 nF (1 )] , (8.98)
where we parallelled the steps in Eq. (8.85). Moreover, representing

Z 3
d p2
g(p + q) =
(2)3 (3) (q + p p2 )g(p2 ) ,
(2)3
(8.99)
and defining P1 (1 , p) (1 , p1 ), P2 (2 , p2 ), we get

Z
3
i
0 Z
d1 d3 p1
d2 d3 p2 h
X n1
F (q )
P
/
(P
)
/
u
(P
)
aR S (P2 )
(Q) =
1
L
1
U
1
4 c=0 2
(2)3
(2)3

1
4 (4)
Q
2
(2) (P2 P1 Q) nB2 (1 nF1 )
(8.100)
where nFi nF (i ), nBi nB (i ). If we insert the free spectral functions from Eqs. (8.36), (8.66),
and note that in free limit U = 0, this result goes over into Eq. (8.88).
109
8.4.
Hard Thermal Loop effective theory
In the case of static observables, we realized in Sec. 3.5 that the perturbative series suffers in
general from serious infrared divergences. However, as discussed in Sec. 6.1, these divergences can
only be associated with bosonic Matsubara zero modes. They can therefore be cleanly isolated by
constructing an effective field theory for the bosonic Matsubara zero modes, as we did in Sec. 6.3.
The situation is somewhat more complicated in the case of real-time observables. Indeed, as
Eq. (8.27) shows, the dependence on all the Matsubara modes is needed in order to carry out
the analytic continuation leading to the spectral function, even if we were only interested in its
behaviour at small frequencies |q 0 | T . (The same holds also in the opposite direction: as the
sum rule in Eq. (8.26) shows, the information contained in the Matsubara zero mode is spread
out to all q 0 s in the Minkowskian formulation.) Therefore, it is not necessarily easy to isolate the
soft/light degrees of freedom for which to write down the most general effective Lagrangian.
Nevertheless, it turns out that the Dimensionally Reduced effective field theory of Sec. 6.3 can
to some extent be generalized to apply to real-time observables as well. In the case of QCD, the
generalization is known as the Hard Thermal Loop effective theory. The effective theory dictates
what kind of resummed propagators should be used for instance in the computation of Sec. 8.2
in order to generate a systematic weak-coupling series. Since the issue is technically relatively
complicated we will, however, restrict ourselves to gluons here, the goal being then essentially to
generalize the analysis of Sec. 5.5 to a real-time situation.
More precisely, Hard Thermal Loops (HTL) can operationally be defined as follows22 :
Consider soft external frequencies and momenta, |q 0 |, |q| <
gT .
Inside the loops, sum over all Matsubara frequencies n .
Subsequently, integrate over hard spatial loop momenta, |s| >

T , Taylor-expanding to leading non-trivial order in |q 0 |/|s|, |q|/|s|.
In order to illustrate the procedure, let us compute the gluon self-energy in this situation. The
computation is much like that in Sec. 5.5, except that now we need to keep the external momentum
non-zero while carrying out the Matsubara sum, because the full dependence on q0 is needed
(Q)
in order to be able to carry out the analytic continuation; we can take q 0 , q to be soft only after
the analytic continuation.
For simplicity, we will focus on the quark loop for now. The contributions of the gluon and ghost
loops can be analysed in precisely the same way, and in the end they manifest themselves through
an additional prefactor precisely like in the parameter m2E = g 2 T 2 ( N3c + N6f ) in Eq. (5.96).
As a starting point, we then take Eq. (5.90), interpreted as a self-energy contribution, (Q).
Furthermore, we restrict to the spatial part, ij , for the moment. Taking also the high-temperature
limit T m, Eq. (5.90) can be rewritten as
=
ij (Q)
Z
j Si
i Sj + Q
S)
2Si Sj + Q
P ij (S2 Q
2g Nf
S)
2
S2 (Q
S
f

Z ij 1 (Q
j Si
i Sj + Q
S)
2 + 1 S2 1 Q
2 2Si Sj + Q
P
2
2
2
2g 2 Nf
S)
2
S2 (Q
2
S
f
22 R.D. Pisarski, Scattering Amplitudes in Hot Gauge Theories, Phys. Rev. Lett. 63 (1989) 1129; J. Frenkel and
J.C. Taylor, High Temperature Limit of Thermal QCD, Nucl. Phys. B 334 (1990) 199; E. Braaten and R.D. Pisarski,
Soft Amplitudes in Hot Gauge Theories: a General Analysis, Nucl. Phys. B 337 (1990) 569; J.C. Taylor and
S.M.H. Wong, The Effective Action of Hard Thermal Loops in QCD, Nucl. Phys. B 346 (1990) 115.
110
2g 2 Nf
X ij
s0f
S2

ij 2
1
.(8.101)
2si sj + qi sj + qj si
Q
S)
2
2
S2 (Q
Here, for generality, we can assume that s0f = f i.

The Matsubara sums can now be carried out. Denoting
E1 |s| ,
E2 |s q| ,
(8.102)
we can read from Eq. (8.63) that

X
1
(f i)2 + E12
=
=
i
1 h
nF (E1 + )e(E1 +) nF (E1 )
2E1
i
1 h
1 nF (E1 + ) nF (E1 ) .
2E1
(8.103)
It is somewhat more tedious to carry out the other sum. We proceed in analogy with the analysis
of Eq. (8.74). Denoting the result by G, we get
G
= T
X
s0f
1
[
s20f + E12 ][(
q0b s0f )2 + E22 ]
1
2 + E2]
[
s20f + E12 ][
r0f
2
r0f
s0f
X is X ir
Z
e 0f
e 0f
T
,
=
d eiq0b T
2
2
s0f + E1
r0f + E22
0
r
s
= T
(
r0f + q0b s0f )
(8.104)
0f
0f
where we proceeded like in Eq. (8.76). The sums can be carried out by making use of Eq. (8.63):
T
X eir0f
2 + E2
r0f
2
X eis0f
s20f + E12
= e T
r0f
s0f
i
1 h
nF (E2 + )e( )E2+ nF (E2 )e E2 ,
2E2
(8.105)
X eis0f ( )
s0f
s20f + E12
i
1 h
nF (E1 )e( )E1 nF (E1 + )e E1 ,
2E1
(8.106)
where in the latter equation attention needed to be paid to the fact that Eq. (8.63) only applies
for 0 and that there is a shift due to the chemical potential in s0f .
Inserting into Eq. (8.104) and carrying out the integral over , we get
G
d e
0
i
q0b

1
nF (E1 )nF (E2 + )e( )(E1 +E2 )
4E1 E2
nF (E1 )nF (E2 )e (E2 E1 )+(E1 )
nF (E1 + )nF (E2 + )e (E1 E2 )+(E2 +)

(E1 +E2 )
+nF (E1 + )nF (E2 )e

h
i
1
1
nF (E1 )nF (E2 + )
1 e(E1 +E2 )
4E1 E2
i
q0b E1 E2
i
h
1
e(E2 ) e(E1 )
nF (E1 )nF (E2 )
i
q0b + E2 E1
111
i
h
1
e(E1 +) e(E2 +)
i
q0b + E1 E2
h
i
1
(E1 +E2 )
e
1
+nF (E1 + )nF (E2 )
i
q0b + E1 + E2

h
i
1
1
nF (E1 ) + nF (E2 + ) 1
4E1 E2 i
q0b E1 E2
h
i
1
nF (E2 ) nF (E1 )
+
i
q0b + E2 E1
h
i
1
nF (E1 + ) nF (E2 + )
+
i
q0b + E1 E2
h
i
1
+
.
1 nF (E1 + ) nF (E2 )
i
q0b + E1 + E2
nF (E1 + )nF (E2 + )
(8.107)
At this point we could carry out the analytic continuation i

q0b q 0 +i0+ , but it will be convenient
to postpone it for a moment; we just have to keep in mind that after the analytic continution, i
q0b
becomes a soft quantity.
The next step is a Taylor-expansion to leading order in q 0 , q. In fact, we will be satisfied with
the leading term, of O(1). We can then write
E1 = s |s| ,
E2 = |s q| s qi
|s| = s qi vi ,
si
(8.108)
where
si
, i = 1, 2, 3 ,
s
are referred to as the velocities of the hard particles.
vi
(8.109)
The non-trivial terms in the Taylor-expansion are the 2nd and the 3rd ones in Eq. (8.107).
Furthermore, it has to be realized that a Taylor-expansion is sensible only in terms where there is
a scale guaranteeing that s is hard, s T . This can only happen in the presence of the thermal
distribution functions. Hence we cannot carry out a Taylor-expansion in the vacuum part; it
could be separately verified that this part vanishes for small q 0 , q (there is no gluon mass at zero
temperature!), and here we simply omit the zero-temperature part. With these approximations,
h
i
1
1
G
n
(s
n
(s
+
)
F
F
4s2 2s
1
+
(q v)nF (s )
i
q0b q v
1
(+q v)nF (s + )
i
q0b + q v
i
1h
+
nF (s + ) nF (s ) + O(q 0 , q) .
2s
+
(8.110)
Now we can insert Eqs. (8.103), (8.110) into Eq. (8.101). Through the substitution s s
(whereby v v), the 3rd row in Eq. (8.110) can be put in the same form as the 2nd row.
Furthermore, terms containing q0 or q in the numerator in Eq. (8.101) are seen to be of higher
order. Thereby
Z h
i
ij
2
nF (s ) nF (s + )
ij (Q) 2g Nf
s 2s
i
2si sj 1 h
nF (s ) nF (s + )
2
4s s
i
q0b q v i
q0b h
2si sj i
nF (s ) + nF (s + )
2
4s
i
q0b q v
112
2g 2 Nf
Z
i
ij h
nF (s ) + nF (s + )
2s
s
i
vi vj h
+
nF (s ) + nF (s + )
2s
i
vi vj h
nF (s ) + nF (s + )
2
h
i
vi vj i
q0
+
nF (s ) + nF (s + )
.
2(i
q0 q v)
(8.111)
The integrals appearing here are finite, and can be carried out with d = 3. We can write the
measure as
Z Z
Z
Z
dv
d3 s
=
,
(8.112)
=
3
(2)
4
s
s
where the radial integration reads
Z
(2)3
ds s2 ,
while the angular integration goes over the directions of v, and is normalized to unity:
Z
dv
=1.
4
Then, the following identities can be verified
Z h
i
nF (s ) + nF (s + )
s
Z h
i
1
nF (s ) + nF (s + )
s s
Z
dv
vi vj
4
The integration
Z
(8.113)
(Exercise 12):
Z h
i
1
= 2
nF (s ) + nF (s + ) ,
s
s2

1 T
2
=
+ 2 ,
4 3
ij
=
.
3
dv
vi vj
4 i
q0 q v
(8.114)
(8.115)
(8.116)
(8.117)
(8.118)
can also be carried out (Exercise 12) but we do not need its value for the moment.
With these ingredients, Eq. (8.111) becomes
Z h
i
1
2g 2 Nf
nF (s ) + nF (s + )
ij (Q)
s
s
Z

1
dv vi vj i
q0
1
1
ij + +
2
6
3
4 i
q0 q v
2

Z
2
T
v
v
i
d
i j q0
v
= g 2 Nf
+ 2
.
6
2
4 i
q0 q v
Including also gluons and ghosts, the complete result reads
Z
q0
= m2E dv vi vj i
ij (Q)
+ O(
q0 , q) ,
4 i
q0 q v
(8.119)
(8.120)
where mE is the generalization of Eq. (5.96) to the case of a fermionic chemical potential,
2

T
2
T2
.
(8.121)
+ Nf
+ 2
m2E g 2 Nc
3
6
2
113
Eq. (8.120), known for the case of QED since a long time23 , is quite a remarkable expression.
Even though it is of O(1) is we count i
q0 and q as quantities of the same order, it depends very
non-trivially on the ratio i
q0 /|q|. In particular, for q 0 = i
q0 0, i.e. in the static limit, ij
obviously vanishes. This corresponds to the result that we obtained in Eq. (5.94). On the other
hand, for 0 < |q 0 | < |q|, it contains both a real and an imaginary part (cf. Eqs. (8.200), (8.204)).
The imaginary part is related to the physics of Landau damping: it means that nearly static gauge
fields can lose energy to the hard particle degrees of freedom in the plasma.
So far, we were only concerned with the spatial self-energy correction ij . An interesting question now is to generalize the computation to the full . Fortunately, it turns out that all the
information needed can be extracted from Eq. (8.120), as we now show.
Indeed, the self-energy , obtained by integrating out the hard modes, must produce a structure which is gauge-invariant in soft gauge transformations; therefore it must be transverse.
However, what we may mean by transversality changes from the case of zero temperature, because
the heat bath introduces a preferred frame, and thus breaks Lorentz invariance. More precisely,
we can now introduce two different projection operators,
T
T
T
(Q) = P0i
(Q) = Pi0
(Q) 0 ,
P00
ij qi qj ,
PijT (Q)
2
q
(8.122)
q q
T
E
P
(Q) ,
P
(Q)
2
Q
which both are transverse:
(8.123)
T
E
P
Q = P
Q = 0 .
E
T
P
P
Also, the two projectors are transverse to each other,

= 0; this can
by choosing a frame where q = q e3 ; then
0 0 0 0
q2
0 0 q q0
1
0
1
0
0
0
0
0
0
E
T
P =
0 0 1 0 , P = Q
2 0
0 0
0
0 0 0 0
q q0 0 0
q02
(8.124)
be seen for instance
With these projectors, we can write

Z
q0
= m2E dv vi vj i
T (Q)
+ PijE (Q)
E (Q)
.
ij (Q)
PijT (Q)
4 i
q0 q v
Contracting Eq. (8.126) with ij and with qi qj leads to the equations

q02
q2
2
=
2
+
E ,
mE i
q0 L = 2T + 1 2
E
T
2
q0 + q 2
q0 + q 2

Z
dv (q v)2 i
q02 q 2
q0
(q 2 )2
m2E
=
E ,
= 0T + q 2 2
E
4 i
q0 q v
q0 + q 2
q02 + q 2
(8.125)
(8.126)
(8.127)
(8.128)
where L denotes the integral in Eq. (8.192). The integral on the left-hand side of Eq. (8.128) can
be written as
Z
Z
q0
q0 i
q0 )(q v)i
q0
dv (q v + i
dv (q v)2 i
=
4 i
q0 q v
4
i
q0 q v
Z
qv
dv
= (i
q0 )2
4 i
q0 q v
h
i
q0 L .
(8.129)
= (i
q0 )2 1 + i
23 V.P. Silin, On the electromagnetic properties of a relativistic plasma, Sov. Phys. JETP 11 (1960) 1136 [Zh. Eksp.
Teor. Fiz. 38 (1960) 1577]; V.V. Klimov, Collective Excitations in a Hot Quark Gluon Plasma, Sov. Phys. JETP
55 (1982) 199 [Zh. Eksp. Teor. Fiz. 82 (1982) 336]; H.A. Weldon, Covariant Calculations at Finite Temperature:
the Relativistic Plasma, Phys. Rev. D 26 (1982) 1394.
114
Inserting the expression from Eq. (8.192), we thus get

m2E (i
(i
q0 )2
i
q0 + |q|
q0 )2
i
q0
T (Q) =
1
,
ln
+
2
q2
2|q|
q2
i
q0 |q|

q0 )2
i
q0 + |q|
i
q0
= m2E 1 (i
E (Q)
ln
.
1
q2
2|q| i
q0 |q|
(8.130)
(8.131)
Eqs. (8.130), (8.131) have a number of interesting limiting values. In the limit i
q0 0 but with
|q| =
6 0, T 0, E m2E . This corresponds to the physics of Debye screening, familiar to us
from Eq. (5.95). On the contrary, if we consider homogeneous but time-dependent waves, i.e. take
|q| 0 with i
q0 6= 0, it can be seen that T , E m2E /3. This genuinely Minkowskian structure
in the resummed self-energy corresponds to plasma oscillations, or plasmons.
We can now also write down a general version of a resummed gluon propagator: in a general
gauge, where the tree-level propagator has the form in Eq. (5.45) and the static resummed propagator the form in Eq. (5.95), we get
x)Ab (
y )i = ab
hAa (
Z
P
eiQ(xy)
T
P
(Q)
2 + T (Q)
E
P
(Q)
2 + E (Q)

q q
,
2 )2
(Q
(8.132)
where is the gauge parameter.

If the propagator of Eq. (8.132) is used in practical computations, it is often useful to express
it in terms of the spectral representation, cf. Exercise 11. The spectral function appearing in the
2 + T (E) (Q)]
plays the
spectral representation can be obtained from Eq. (8.28), where now 1/[Q
E
role of what we there denoted by . After analytic continuation, i
q0 q 0 + i0+ ,
1
1
,
2 + T (E) (
(q 0 + i0+ )2 + q2 + T (E) (i(q 0 + i0+ ), q)
)
Q
q0 , q
(8.133)
where
0
T (i(q + i0 ), q) =
E (i(q 0 + i0+ ), q) =

(q 0 )2
q 0 + i0+ + |q|
q0
m2E (q 0 )2
1 2 ln 0
,
+
2
q2
2|q|
q
q + i0+ |q|

(q 0 )2
q0
q 0 + i0+ + |q|
m2E 1 2
.
1
ln 0
q
2|q| q + i0+ |q|
(8.134)
(8.135)
For |q 0 | > |q|, T , E are real. For |q 0 | < |q|, they have an imaginary part. In particular, for
|q 0 | |q|, we get

q0
(q 0 )2
m2E
i
(8.136)
+ 2 2 + ... ,
T
2
2|q|
q

q0
(q 0 )2
2
E mE 1 + i
(8.137)
2 2 + ... .
2|q|
q
The spectral functions become

0
0 2
2
, |q 0 | > |q|
sign(q
)
(q
)
Re
0
T (q , q) =
q0
m2E
,
|q 0 | |q|
4|q|5
E (q 0 , q) =

0
0
0 2
2
sign(q ) (q ) q Re E , |q | > |q|
m2E
q0
,
+ m2E )2
2|q|(q2
115
|q 0 | |q|
(8.138)
(8.139)
In fact, introducing the notation

y
q0
,
|q|
q |q| ,
(8.140)
the complete expressions for T , E can be written as

T (q 0 , q)
T (y, q)
T (y, q)
(y 2 1)E (q 0 , q)
E (y, q)
E (y, q)
(1 y 2 )T (y, q)
= (y 2 1) sign(y)(T (y, q)) + 2
,
T (y, q) + 2T (y, q)

y + 1
m2
y

q 2 (y 2 1) E y 2 +
1 y 2 ln
,
2
2
y1

m2E
y 1 y2 ,
4
(1 y 2 )E (y, q)
= (y 2 1) sign(y)(E (y, q)) + 2
,
E (y, q) + 2E (y, q)

y y + 1
q 2 + m2E 1 ln
,
2
y1
m2E
y.
2
(8.141)
(8.142)
(8.143)
(8.144)
(8.145)
(8.146)
It can be seen that there is in each case a plasmon pole, i.e. a delta function analogous to the
delta functions of the free propagator, Eq. (8.36), but displaced by an amount m2E ; as well as a
cut at |y| < 1, representing Landau damping.
So far, we have only computed the form of the resummed gluon propagator. A very interesting
question is whether also an effective action can be written, which would then not only contain the
propagators, but also new vertices, in analogy with the Dimensionally Reduced effective theory
of Eq. (6.36) in the static case. Note that since our observables are now non-static, the effective
action should be gauge-invariant also in time-dependent gauge transformations.
Most remarkable, such an effective action can indeed be found24 . We simply cite here the result
for the gluonic case. Expressing everything in Minkowskian notation (i.e. after i
q0 q 0 and using
the Minkowskian Aa0 ), the effective Lagrangian reads

Z
dv
1
1
1
1
.
(8.147)
Tr
v F
v F
LM = Tr [F F ] + m2E
2
2
4
vD
vD
Here v (1, v) is a light-like four-velocity, and D represents the covariant derivative in the adjoint
representation.
Several remarks on Eq. (8.147) are in order:
A somewhat tedious analysis, making use of the velocity integrals listed in Eqs. (8.195)
(8.203), shows that in the static limit the second term in Eq. (8.147) indeed reduces to the
mass term in Eq. (6.36) (modulo Wick rotation and the redefinition of Aa0 ).
In the static limit, we found quarks to always be infrared-safe, but this situation changes
after the analytic continuation. Therefore a dynamical quark part should be added to
Eq. (8.147)24 .
Eq. (8.147) has the unpleasant feature that it is non-local: derivatives appear in the denominator. This is something we do not usually expect from effective theories. Indeed, if
non-local structures can appear, it is quite difficult to generally analyse what kind of higher
24 J. Frenkel and J.C. Taylor, Hard thermal QCD, forward scattering and effective actions, Nucl. Phys. B 374
(1992) 156; E. Braaten and R.D. Pisarski, Simple effective Lagrangian for hard thermal loops, Phys. Rev. D 45
(1992) 1827.
116
order operators have been truncated from the effective theory and, consequently, what the
relative accuracy of the effective description is.
In some sense, the appearance of non-local terms is a manifestation of the fact that the
proper infrared degrees of freedom have not been identified. In the next section, we discuss
a reformulation of the HTL effective theory which has additional degrees of freedom, and
consequently a local appearance, thereby perhaps curing the problems. (The reformulation
does not have the form of a field theory, however, whereby it continues to be difficult to
analyse the accuracy of the effective description.)
We arrived at Eq. (8.147) by integrating out the hard modes, with momenta s T . However,
like in the static limit, the theory still has multiple dynamical momentum scales, |q|
gT, g 2 T . It can be asked what happens if the momenta |q| gT are also integrated out.
This question has been analysed in detail in the literature, and leads indeed to a simplified
(local) effective description25 , which can be used for non-perturbatively studying observables
sensitive to ultrasoft momenta, |q| g 2 T .
Apart from observables sensitive to ultrasoft momenta, another situation where loop corrections need to be computed within the theory of Eq. (8.147) is the physics around the plasmon
poles, i.e. the -functions in Eqs. (8.141), (8.144). Indeed, it turns out that the plasmons
get a finitewidth through loop corrections, whereby the delta function gets smeared for
2
26
|q 0 mE / 3| <
g T.
25 D. B
odeker, On the effective dynamics of soft non-abelian gauge fields at finite temperature, Phys. Lett. B 426
(1998) 351; P. Arnold, D.T. Son and L.G. Yaffe, Hot B violation, color conductivity, and log(1/) effects, Phys.
Rev. D 59 (1999) 105020; D. B
odeker, Diagrammatic approach to soft non-Abelian dynamics at high temperature,
Nucl. Phys. B 566 (2000) 402; From hard thermal loops to Langevin dynamics, Nucl. Phys. B 559 (1999) 502;
P. Arnold and L.G. Yaffe, High temperature color conductivity at next-to-leading log order, Phys. Rev. D 62 (2000)
125014; D. B
odeker, Perturbative and non-perturbative aspects of the non-abelian Boltzmann-Langevin equation,
Nucl. Phys. B 647 (2002) 512.
26 E. Braaten and R.D. Pisarski, Calculation of the gluon damping rate in hot QCD, Phys. Rev. D 42 (1990) 2156.
117
8.5.
Relation to classical kinetic theory
We found in the previous section that the HTL effective theory has the unpleasant feature of
being non-local. It turns out, however, that by introducing extra degrees of freedom hard onshell particles it can be recast in a local form, as a kinetic theory, involving the mentioned
particles propagating in a gauge field background. (The canonical kinetic theory is defined by
the Boltzmann equation; when gauge fields appear as force terms in the Boltzmann equation, and
gauge field equations of motion are imposed as a further constraint, the system is often referred to
as Vlasov equations.) Such a local reformulation is very useful from the point of view of further
analyses, both analytic25 and numerical27 , and is also conceptually more satisfying than a non-local
effective theory. As already mentioned in the previous section, the drawback is that the theory
no longer has the appearance of a quantum field theory, whereby some of the familiar tools and
results are no longer available.
In order to simplify the analysis a bit, we will start by considering the case of QED, i.e. an
Abelian plasma. Then the covariant derivatives in Eq. (8.147) become partial derivatives, and the
HTL-correction is quadratic. On the other hand, as we recall from Eq. (7.39), in the Abelian case
the effective theory can also possess a term linear in the gauge field A0 , if the fermions carry a
non-zero chemical potential.
More precisely, let us assume that the coupling between fermions and gauge fields is mediated
by the covariant derivative
D = igA .
(8.148)
Letting

2
2
jE g
T + 2 ,
3
m2E
2
T2
+ 2
3
(8.149)
and introducing the shorthand notation

Z
dv
4
for the velocity integrals, the HTL effective action of Eq. (8.147) can be written as

Z
v v
m2E
1
F .
F
F F jE v A
SM =
4
4
(v )2
x,v
(8.150)
(8.151)
Here we have carried out one partial integration in the denominator (this can perhaps easiest be
understood by going to momentum space).
Since the action is quadratic in the fields, its contents can equivalently be expressed through
equations of motion. It is useful to express the equations of motion in a form following from the
identities in Eq. (8.210). We obtain

Z
v v
F (x) =
jE v m2E
F0 (x) .
(8.152)
v
v
R
By making use of the properties v v = 0 and F00 = 0 it is easy to see that the right-hand side
is divergenceless (or transverse) with respect to , as it has to be in order for the equation to
be consistent.
Let us at this point introduce some notation that will be useful in the following. We denote
+ (P 2 ) 2(p0 ) (2) (P 2 ) ,
(8.153)
27 D. B
odeker, G.D. Moore and K. Rummukainen, Chern-Simons number diffusion and hard thermal loops on
the lattice, Phys. Rev. D 61 (2000) 056003; M. Hindmarsh and A. Rajantie, Phase transition dynamics in the hot
Abelian Higgs model, Phys. Rev. D 64 (2001) 065016; A. Rebhan, P. Romatschke and M. Strickland, Dynamics of
quark-gluon plasma instabilities in discretized hard-loop approximation, JHEP 09 (2005) 041.
118
such that
+ (P 2 ) f (p0 , p) =
f (p, p)
,
p
p |p| .
(8.154)
All the results will factorise in a form where a phase space integral is left over, which can be carried
out explicitly. Let
nF (p) =
1
,
+1
ep
N (p) = nF (p + ) nF (p ) .
Then, irrespective of the relative magnitudes of T, ,

Z

2
N (p) = T 2 + 2 ,
6
p
Z
Z
Z
1
1
2
1T 2
N+ (p)
N+ (p)
=
=
+ 2 ,
N (p) =
p
2 p p
2 p
4 3
p
Z
Z 2
Z
2 Z
1
1
1 N (p)
N (p)
N (p)
=
=
=
N (p) = 2 .
2
2
2
p
p
2 p p
2 p
2
p p
p
(8.155)
(8.156)
(8.157)
(8.158)
These relations can be proven with the techniques of Exercise 12.

The claim now is that the HTL structures in Eq. (8.152) can be reproduced28 by classical kinetic
theory, or Vlasov equations. To see this, let us start with classical electrodynamics29 , and define
P =
dx
,
dt
dP
= gF P .
dt
(8.159)
Then the collisionless Boltzmann equation for hard particles in a gauge field background becomes

df (x, P )
f
f
= P
+
gF
=0.
(8.160)
dt
x
P
Let us note that we may in general assume that
f (x, P ) = + (P 2 )f(x, P ) ,
(8.161)
where + (P 2 ) is defined as in Eq. (8.153), since this form is conserved by Eq. (8.160), due to
P F
(P 2 ) = 2P P F (P 2 ) = 0 .
P
(8.162)
The derivate of (p0 ) included in + (P 2 ) can be safely ignored as well, since it would contribute
only at the point p0 = p = 0, and has no effect after integration over p.
We formally solve Eq. (8.160) in powers of gF : f = f0 + f1 + f2 + ... . This leads to the
recursion relation
fn (x, P )
P fn+1 (x, P ) = gP F (x)
.
(8.163)
P
The zeroth order gives
P f0 (x, P ) = 0 .
(8.164)
We take as a solution a space-time independent function depending, in view of Eq. (8.161), nontrivially only on p0 , parametrized by T, , and applying separately to all particle species i:
(i)
(i)
f0 = + (P 2 ) f0 (p0 ; T, i ) + (P 2 ) nF (p0 + i ) ,
(8.165)
28 J.P. Blaizot and E. Iancu, Kinetic equations for long wavelength excitations of the quark gluon plasma, Phys.
Rev. Lett. 70 (1993) 3376; P.F. Kelly, Q. Liu, C. Lucchesi and C. Manuel, Deriving the hard thermal loops of QCD
from classical transport theory, Phys. Rev. Lett. 72 (1994) 3461; F.T. Brandt, J. Frenkel and J.C. Taylor, High
temperature QCD and the classical Boltzmann equation in curved space-time, Nucl. Phys. B 437 (1995) 433.
29 For a pedagogic presentation, see R.D. Pisarski, Nonabelian Debye screening, tsunami waves, and worldline
fermions, lectures at International School of Astrophysics D. Chalonge, Erice, Italy, 1997 [hep-ph/9710370].
119
where nF (p0 ) is from Eq. (8.155). Furthermore, antiparticles are always assumed to come with the
opposite signs of g and than particles. Thus, a single Dirac fermion contributes two degrees of
freedom with +g, +, two with g, .
In addition to these equations, we need the definition of the current induced by the hard particles:
X Z
gi
P f (i) (x, P ) .
(8.166)
j (x)
P
The equations of motion are

free
SM
=
1
F F ,
4
free
S
= F = j .
A M
(8.167)
The expression for j in terms of the background gauge field thus implies a non-local effective
free
action SM = SM
+ SM for the gauge fields only, where SM is to be determined from
SM = j .
A
(8.168)
Let us now work out explicit expressions. We start by considering f = f0 . Summing over two
degrees of freedom with +g, + and two with g, , we obtain from Eqs. (8.165), (8.166),
Z
2

(8.169)
j = 2g
+ (P 2 ) P N (p0 ) = 0 g T 2 + 2 ,
3
P
where we used Eqs. (8.154), (8.155), (8.156). This indeed agrees with Eq. (8.152).
For the next term we need
f1 = g
1
f0
.
P F
P
P
(8.170)
Inserting into Eq. (8.166),
j (x) =
gi2
(i)
f0 P P
F (x) .
P P
Since f0 essentially only depends on p0 , we get

Z
h
i Z v v
j (x) = 2g 2
+ (P 2 ) nF (p0 + ) + nF (p0 ) p
F0 (x) ,
P
v v
|
{z
}

T2
3
12
(8.171)
(8.172)
+
2
which indeed agrees with the second term in Eq. (8.152). This completes our proof for the Abelian
case.
Let us then move to the non-Abelian case. The formulation will be somewhat more complicated,
so we just indicate the form of the kinetic equations, and solve them in the static limit.
The simplest way to display the non-Abelian kinetic equations is to follow Eq. (8.160), but
replace f by an Nc Nc matrix. The collisionless QCD Boltzmann/Vlasov equation for each single
fundamentally charged fermionic degree of freedom, and the corresponding gauge current induced,
can be written as
h
i gn
f o
P D, f +
P F ,
=0,
(8.173)
2
P
Z
h
i
ja = g
P Tr T a f .
(8.174)
P
120
To express this in a more down-to-earth way, we may write the matrix f in terms of a singlet
distribution function f and an adjoint (or octet) distribution function f a :
f (x, P ) =
1
f (x, P ) + 2T a f a (x, P ) .
Nc
Making use of
{T a , T b } =
1 ab
+ dabc T c ,
Nc
(8.175)
(8.176)
we can write
h
i
P D, f
=
1
P f + 2(P D)ab f b T a ,
Nc
b
1 n
f o
a f
P F ,
+ 2P F
{T a , T b }
Nc
P
P
b
f
2 a f a
2
a f
P F
+
P F
+ 2dabc T cP F
.
Nc
P
Nc
P
P
n
f o
=
P F ,
P
=
(8.177)
(8.178)
Projecting then Eq. (8.173) with Tr [...] and Tr [T a ...], the equations obtain the forms30
a
a f
P f + gP F
=0,
P
c
g
g
a f
b f
+
P F
=0,
(P D)ab f b + dabc P F
2
P
2Nc
P
Z
X
gi
P f a(i) .
ja =
(8.179)
(8.180)
(8.181)
When computing the current, each quark now comes with Nc colours, in addition to two spin
degrees of freedom, both for particles and for anti-particles.
The equations can again be solved iteratively in gF : f = f0 + f1 + f2 + ... . At the zeroth
order,
(i)
f0
+ (P 2 ) nF (p0 + i ),
(8.182)
a(i)
f0
0.
(8.183)
Iterating this, we obtain at the first order,
(P
(i)
f1
0,
(8.184)
b(i)
D)ab f1
gi a
P F0 + (P 2 ) nF (p0 + i ) .
2Nc
(8.185)
Let us now solve these equations in the static limit. In Eq. (8.185) we can then write
a
P F0
= (P DA0 )a 0 (P Aa ) = (P D)ab Ab0 ,
so that
a(i)
f1
gi a
A + (P 2 ) nF (p0 + i ) .
2Nc 0
Inserting into Eq. (8.181), we obtain

X Z
a(i)
gi
P f1
ja =
i
g 2 Aa0
0 g 2 Aa0
(8.186)
(8.187)
P + (P 2 ) N+
(p0 )
N+
(p) .
(8.188)
30 U. Heinz, Kinetic Theory For Nonabelian Plasmas, Phys. Rev. Lett. 51 (1983) 351; H. Elze and U. Heinz,
Quark - Gluon Transport Theory, Phys. Rept. 183 (1989) 81.
121
After use of Eq. (8.157), and a proper account of the signs between LE and LM as well as between Euclidean and Minkowskian A0 , this indeed agrees with the derivative of the mass term in
Eq. (6.36).
Let us count the amount of information that needed to be added to
to make it local. In full generality, the classical distribution functions
coordinates. As we have discussed, f (x, P ) = + (P 2 )f(x, P ), i.e. the
on-shell; thus the dependence can thus be reduced to, say, the spatial
one further simplification is possible in Eq. (8.185), by writing
a(i)
f1
gi
+ (P 2 ) nF (p0 + i ) W a (x, v) ,
2Nc
the HTL action, in order

f (x, P ) depend on (4+4)
hard particles can be set
components p. However,
(8.189)
so that
a
(v D)ab W b (x, v) = v F0
(x) .
(8.190)
We can also perform the sum over i and the integral over p0 in Eq. (8.181), reducing thus the
dependence only to angular variables and a total of (4+2) dimensions. Nevertheless, we needed to
introduce extra dimensions, not only extra degrees of freedom, in order to make the HTL theory
local!
As a final remark we note that the equations written down so far did not specify what kind
of initial conditions should be assumed for the gauge fields. Indeed, in order to have a proper
statistical weighting over the initial conditions for the time evolution, one should also work out a
a
Hamiltonian formulation in terms of the gauge fields Aai , Eia F0i
and W a ; afterwards one can
weigh by exp(H/T ). This issue is not altogether trivial but does turn out to possess a solution31 .
31 J.P. Blaizot and E. Iancu, Soft collective excitations in hot gauge theories, Nucl. Phys. B 417 (1994) 608;
V.P. Nair, Hard thermal loops, gauged WZNW action and the energy of hot quark - gluon plasma, Phys. Rev. D 48
(1993) 3432 Hamiltonian analysis of the effective action for hard thermal loops in QCD, Phys. Rev. D 50 (1994)
4201.
122
8.6.
Exercise 12
Compute the radial and angular integrals in Eqs. (8.115)(8.118).

Eq. (8.115) can be verified by straightforward partial integration:
Z
Z
i
h
i
ds h
s nF (s ) + nF (s + ) =
ds s nF (s ) + nF (s + )
ds
ds
0
Z0
h
i
ds s2 nF (s ) + nF (s + ) . (8.191)
0
Moving the first term on the right-hand side to the left-hand side leads directly to Eq. (8.115).
Eq. (8.116) can be verified for instance by starting from a combination of Eqs. (7.35), (7.41):

Z
s+
s
T
T
s + T ln 1 + e
f (T, ) = 2
+ ln 1 + e
s
7 2 T 4 2 T 2
4
.
+
+
180
6
12 2
(8.192)
Taking the second partial derivative with respect to , we get

Z
2
2 f (T, )
s+
s
T
T
= 2
T 2 ln 1 + e
+ ln 1 + e
s

Z
d2
s+
s
T
T
= 2
T 2 ln 1 + e
+ ln 1 + e
ds
s

Z
1
s
s+
T
T
+ ln 1 + e
ln 1 + e
= 4T
2
s s
T2
2
=
+ 2 ,
3
(8.193)
(8.194)
where in the penultimate step we carried out two partial integrations.

On the other hand, the integral in Eq. (8.193) can be rewritten as

Z
1
s
s+
T
T
+ ln 1 + e
ln 1 + e
2
s s

Z
4
ds
s
s+
T
T
=
+ ln 1 + e
ln 1 + e
ds
(2)3 0
ds

s+
s
Z
1
e T
4
e T
+
=
ds
s
s+
s
(2)3 0
T
1 + e T
1 + e T
Z h
i
1
1
nF (s ) + nF (s + ) .
=
T ss
Replacing the integral in Eq. (8.193) by that in Eq. (8.195) leads then to Eq. (8.116).
Eq. (8.117) is a trivial consequence of rotational symmetry, and the fact that v2 = 1.
As far as Eq. (8.118) goes, we start by carrying out a simpler integral:
L
dv
1
4 i
q0 q v
1
2
4
123
+1
dz
1
1
i
q0 |q|z
(8.195)
=
=
Z +1
d
1
ln(i
q0 |q|z)
dz
2|q| 1
dz
1
i
q0 + |q|
ln
.
2|q| i
q0 |q|
(8.196)
Further integrals can then be obtained by making use of rotational symmetry: for instance,
Z
vi
dv
= qi f (i
q0 , |q|) ,
(8.197)
4 i
q0 q v
where, contracting both sides with q,

Z
Z
1
qv
dv
1
1
dv
f (i
q0 , |q|) =
.
1
+
i
q
=
0
q2
4 i
q0 q v
q2
4 i
q0 q v
(8.198)
Another trick, needed for having higher powers in the denominator, is to take derivatives of
Eq. (8.196) with respect to i
q0 .
Without carrying out any further steps, we list finally the results for a number of velocity integrals
that can be obtained this way. Let us change the notation a bit at this point: we now replace
i
q0 by q 0 + i0+ , as is relevant for retarded Greens functions (i0+ is not shown explicitly), and
introduce the light-like four-velocity v (1, v). Then the integrals read (i, j = 1, 2, 3)
Z
dv
= 1,
(8.199)
4
Z
dv i
v = 0,
(8.200)
4
Z
1 ij
dv i j
vv =
,
(8.201)
4
3
Z
dv 1
= L(Q) ,
(8.202)
4 v Q
Z
i
qi h
dv v i
1 + q 0 L(Q) ,
(8.203)
=
2
4 v Q
|q|

Z
i
dv v i v j
L(Q) ij q i q j
q0 h
qiqj
0
ij
=

+
1
q
L(Q)
3
,
(8.204)
4 v Q
2
|q|2
2|q|2
|q|2
Z
dv
1
1
=
,
(8.205)
4 (v Q)2
Q2
Z
i
dv
vi
qi h q0
=
L(Q) ,
(8.206)
2
2
2
4 (v Q)
|q| Q

Z
dv v i v j
qiqj
1
(q 0 )2
qiqj
1
ij
0
ij
2q
L(Q)
+
3
,(8.207)
=
4 (v Q)2
2Q2
|q|2
2|q|2
Q2
|q|2
where v (1, v i ), Q (q 0 , q), our metric convention is (+), v Q = q 0 v q, and
L(Q)
(q 0 )3
q 0 + |q|
i
q0
1
+ ... .
ln 0
+ 2+
2|q| q |q|
2|q| |q|
3|q|4
(8.208)
The following identities (which can be derived by tedious explicit use of the relations obtained)
are sometimes very useful:
Z
Z
dv v Q
dv v

I
J
=
0 I J ,
(8.209)
2
4 (v Q)
4 v Q
Z
Z
Z
dv v v
dv v v
dv v v
Q
I
Q
J
=
2
I
Q
J
=
2
Q[ I0] J . (8.210)
[ 0]
[ ] [ ]
2
4 (v Q)
4 v Q
4 v Q
Here Q[ I] Q I Q I , and I, J are arbitrary Lorentz vectors.
124
9.
9.1.
Applications
Thermal phase transitions
As a first application of the general formalism developed, we consider the existence of finitetemperature phase transitions in models of particle physics. Prime examples are the deconfinement transition taking place in QCD, and the electroweak symmetry restoring transition taking
place in the electroweak theory32 , both of which could have significance for Early Universe cosmology. For simplicity, though, the practical analysis will be carried out within the scalar field theory
discussed in Sec. 3.
In general, a phase transition can be defined to be a line in the (T, )-plane across which the
grand canonical free energy density f (T, ) is non-analytic. In particular, if f /T or f / is
discontinuous, we speak of a first order transition. The energy density

i T2
h

f
1
(H
Q)
(9.1)
e=
=
Tr He
ln Z = f T
VZ
V T
T
T
T
is then discontinuous. This means that a closed system can proceed through the transition only
if there is some mechanism for energy transfer and dissipation; thus, first order transitions always
possess non-trivial dynamics.
It is often possible to associate an order parameter with a phase transition. In a strict sense,
the order parameter should be an elementary or composite field, the expectation value of which
vanishes in one phase and is non-zero in another. In a generalized sense, we may refer to an order
parameter even if it would not vanish in either phase, provided that (in first order transitions) it
jumps across the phase boundary. A particularly simple situation is if this role is taken by some
elementary field; in the following we consider the case where a real scalar field, , plays the role of
an order parameter. More realistically, could be for instance a neutral component of the Higgs
doublet (in some gauge) in, say, the MSSM.
The Euclidean Lagrangian of the system then reads
LE
1
1
( )2 + ()2 + V () .
2
2
(9.2)
We take the potential to be of the form V () = 21 m2 2 + 41 4 , like in the Standard Model, so

that has a non-zero expectation value at zero temperature:
V
Let us now evaluate the partition function of this system with the method of the effective poten introduced in Sec. 7.2. In other words, we again put the system in a finite volume V ,
tial, Veff (),
and denote by the condensate, i.e. the mode with p = 0, b = 0. We then write
Z Z
D eSE [=+ ]
(9.3)
d
Z(V, T, ) =
P 6= 0
32 We use here the standard characterizations of these transitions, even though in a strict sense neither concept is
appropriate.
125

V
,
d exp Veff ()
T
(9.4)
where periodic boundary conditions are assumed. We note that

R theR thermodynamic limit V
and that d d3 x = 0, given that by
is to be taken only after the evaluation of Veff (),
definition only has modes with P 6= 0.
around its absolute minimum
In order to carry out the integral in Eq. (9.4), we expand Veff ()
min , and perform the corresponding Gaussian integral:

1
Veff (min ) + Veff
(min )( min )2 + ... ,
2s
V
2T
,
e T Veff (min )
Veff (min )V
=
Veff ()
Z
d e T Veff ()
whereby (we remove reference to since there is no chemical potential in this system)

ln V
.
f (T ) = Veff (min ; T ) + O
V
(9.5)
(9.6)
(9.7)
In other words, in the thermodynamic limit V , the problem reduces to determining Veff and
finding its minimum. Note that the value of min depends on the parameters of the problem, in
particular on T .
Let us now ask under which conditions a first order phase transition could emerge. We can write

T)
T ) min
f (T )
Veff (;
Veff (;
=
+
(9.8)
T
T
T

=
min

T )
Veff (;

=
,
(9.9)

T
=
min
f
6=
where we made use of the fact that min minimizes Veff . Thereby it is clear that limT Tc+ T
f
limT Tc T only if limT Tc+ min 6= limT Tc min . In other words, a first order transition necessitates a discontinuity in min :
Veff
T > Tc
0
T < Tc
_
+
min(Tc )
_
min(Tc )
Our task therefore is to evaluate Veff . Before proceeding with the practical computation, let us
formulate the generic rules for the computation of the effective potential that can be extracted
from the above:
(1) Write = + in LE .
(2) The part only depending on is the 0th order, or tree-level, contribution to Veff .
126
R R
(3) Any terms linear in should be dropped, because d d3 x = 0.
(4) The remaining contributions to Veff are obtained like f (T ) before, except that the masses
and couplings of now depend on the shift .

(5) However, among all possible graphs, one-particle-reducible graphs (i.e. graphs with a single
-propagator, the cutting of which would split the graph into two graphs) should be dropped
(in addition to disconnected graphs), since such a -propagator would necessarily carry zero
momentum, which is excluded by definition.
Remarkably, this set of rules is identical, from an operational point of view, to the set of rules
that can be shown to follow33 from a totally different (but standard) definition of an effective
potential, namely one based on a Legendre transform of the generating functional:
Z
R
W [J]
e
D eSE x J ,
(9.10)
W [J]
,
(9.11)

J
T []
for = constant .
Veff ()
(9.12)
V
However, our procedure is actually better than this standard one, because our procedure is defined
while the Legendre transform of the standard procedure requires certain propfor any value of ,
erties from the functions in order to be defined, which leads to confusing discussions concerning
for instance whether the effective potential necessarily needs to be a convex function.
W [J] J ,
[]
Let us now proceed with the practical computation. Implementing steps (1) and (2), and indicating terms dropped in step (3) by square brackets, we get
1
1
( )2
( )2 ,
(9.13)
2
2

1
1
1 m2 2 ,
m2 2 m2 2 m2
(9.14)
2
2
2
1 4 3 3 2 2
1 4
3 + 1 4 ,

+ + +
(9.15)
4
4
2
4
Z Z
V
,
(9.16)
d
d3 x =
T
0
V
1
1
(0)
Veff ()
= m2 2 + 4 .
(9.17)
2
4
(0)
Note that, consequently, V ()
is independent of the temperature T .
eff
The dominant thermal fluctuations or radiative corrections are given by the 1-loop expression. This follows from the part quadratic in (since linear terms are dropped). Combining
Eqs. (9.14), (9.15), the effective mass of reads m2eff m2 + 32 , and the corresponding
contribution to the effective potential becomes
Z
R R 3 1
(1)
V
2
2
(9.18)
D e 0 d V d x 2 [ +meff ]
e T Veff =
Z
P
2
2
2
1
T
(9.19)
=
D e V n ,p 2 [n +p +meff ]
= C
21
(n2 + p2 + m2eff )
P 6=0
(1)
=
Veff ()
T X1
ln(n2 + p2 + m2eff ) [const.]
V V
2
lim
P 6=0
33 R.
Jackiw, Functional Evaluation of the Effective Potential, Phys. Rev. D 9 (1974) 1686.
127
(9.20)
(9.21)
In the infinite-volume limit this goes over to the function J(meff , T ), defined in Eqs. (2.50), (2.51).
We return presently to the properties of this function, but let us first specify the way to compute
higher-order corrections as well.
Indeed, higher order corrections come from the remaining terms, paying attention to the rules
(4) and (5):
V
(2)
e T Veff
()
=
SE,I []
exp SE,I []
,
1PI

Z Z
1 4
3
3
,
d x +
d
4
V
0
(9.22)
(9.23)
where the propagator to be used reads

D
E
1
= V
(P ) (Q)
P ,Q 2
2
.
T
n + p + m2eff ()
(9.24)
We now return to the evaluation of the 1-loop effective potential at V = . We obtain

Z 3 h
i

d p
(1)
/T
+
T
ln
1
e
Veff =
.
q
(2)3 2
= p2 +m2
eff
The temperature-dependent part is as in Eq. (2.58),
!#
"
p
Z 3
p2 + m2eff
d p
T ln 1 exp
JT (meff ) =
(2)3
T
Z
h
2 2i
T4
=
dx x2 ln 1 e x +y
.
2
2 0
y=meff /T
(9.25)
(9.26)
(9.27)
This function was evaluated in Exercise 3, and we recall that the shape is:
1-loop correction
Tree-level result
(0)
JT
Veff
_2 1/2
y=(c1 + c2 )
In other words, the symmetric minimum becomes more favoured (has a lower free energy) at high
temperatures!
In order to be more quantitative, let us study what happens at very high T . From Eq. (2.81),

E
2 T 4
3
me
m2 T 2 m3 T
m4
JT (m) =
+ O(m6 ) .
(9.28)
ln
+
90
24
12
2(4)2
4T
4
Keeping just the leading mass-dependent term leads to

2 2 1 4
1
(0)
(1)
2
m + T + .
Veff + Veff = [-indep.] +
2
4
4
128
(9.29)
Veff
large T
small T
Therefore, for T 2m/ , the symmetry is broken; for T 2m/ , it is restored; and in
between there is a phase transition of some kind, as illustrated above.
We may then ask a more refined question, namely, what is the order of the transition? In order
to get a first impression on this, let us include the next term from Eq. (9.28). For simplicity, we
may even put m2 0, whereby
(0)
(1)
Veff + Veff = [-indep.]

+
T
2 2
3 + 1 4 .
T
(3)3/2 ||
8
12
4
(9.30)
Thus, it appears like this could be a case of a fluctuation induced / radiatively generated first
order transition:
Veff
_4
+
_3
- ||
_2
+
0
We should not rush to conclusions, however. Indeed, it can be seen from Eq. (9.30) that the
broken minimum appears where the cubic and quartic terms are of similar orders of magnitude,
i.e.,
3
3 ||
4 ||
12 T .
(9.31)
T 2 ||
However, the expansion parameter related to higher order corrections, discussed schematically in
Sec. 6.1, then becomes
1
T
T
2 T
p
O(1) .
(9.32)
meff
||
32
In other words, the perturbative prediction is not reliable for the order of the transition.
129
On the other hand, a reliable analysis can again be carried out with effective field theory techniques, as discussed in Sec. 6.3 for QCD. In the case of the scalar field theory, the dimensionally
reduced action takes the form

Z
1
1
1
1
2
Seff =
(i 3 ) + m23 23 + 3 43 + ... ,
(9.33)
d3 x
T V
2
2
4
with effective couplings of the form
m23
= m2 [1 + O()] +
= [1 + O()] .
1
T 2 [1 + O()] ,
4
(9.34)
(9.35)
This system can be studied non-perturbatively (with lattice simulations, -expansion, etc) to show
that there is a second order transition at m23 m2 + T 2 /4 0, in the 3d Ising universality
class.
Finally, we note that if the original theory is more complicated (containing for instance two
different fields and coupling constants), then it is possible to arrange the couplings so that the first
order signature seen in perturbation theory is physical. For instance:
A theory with two real scalar fields can have a first order transition, if the couplings between
the two fields are tuned appropriately34 .
A theory with a complex scalar field and U(1) gauge symmetry, which is just the GinzburgLandau theory of superconductivity, does have a first order transition, if the quartic coupling,
, is small enough compared with the electric coupling squared, e2 .35
The standard electroweak theory, with Higgs doublet(s) + SU(2)U(1) gauge symmetry, can
also have a first order transition if the scalar self-coupling is small enough36 .
34 J.
Rudnick, First-order transition induced by cubic anisotropy, Phys. Rev. B 18 (1978) 1406.
Halperin, T.C. Lubensky and S.-K. Ma, First-Order Phase Transitions in Superconductors and Smectic-A
Liquid Crystals, Phys. Rev. Lett. 32 (1974) 292.
36 D.A. Kirzhnits and A.D. Linde, Symmetry Behavior in Gauge Theories, Annals Phys. 101 (1976) 195.
35 B.I.
130
9.2.
Bubble nucleation rate
It was already mentioned in the previous section that if a first order transition takes place, its
dynamics is non-trivial, because the latent heat releases in the transition needs to be dissipated
away. The basic mechanism for this is bubble nucleation and growth: the transition does not take
place exactly at the critical temperature, Tc , but the system first supercools to some nucleation
temperature, Tn . At this point bubbles of the stable phase form, and start to grow; the latent heat
is transported away in a hydrodynamic shock wave which precedes the expanding bubble.
The purpose of this section is to determine the probability of bubble nucleation, per unit time
and volume, at a given temperature T < Tc . Combined with the cosmological evolution equation
for the temperature T as a function of the time t, this would in principle allow us to estimate Tn .
We will not get into explicit estimates, though, but rather try to illustrate some essential aspects of
the general formalism, given that it is analogous to several other rate computations in quantum
field theory, such as the determination of the rate of baryon number violation in the standard
electroweak theory. In terms of the effective potential, the general setting can be illustrated as
follows:
Veff
T < Tc
0
?
For simplicity we consider the situation in the following that a bump already exists in the tree for radiatively generated transitions, in which a bump only appears in Veff (),
level potential V ();

some degrees of freedom need to be integrated out for this to be the case.
The starting point now is to properly define what is meant with the nucleation rate. It turns
out that this task is rather non-trivial; in fact, it is not clear whether a completely general nonperturbative definition can be given at all. Nevertheless, for practical purposes, the Langer formalism37 appears sufficient.
The general idea is the following. Consider first a system at zero temperature. Suppose we use
boundary conditions at spatial infinity, lim|x| (x) = 0, in order to define metastable energy
eigenstates. We could imagine that their time evolution looks like
|(t)i = eiEt |(0)i = ei[Re(E)+i Im(E)]t |(0)i
h(t)|(t)i = e2 Im(E) t h(0)|(0)i .
(9.36)
(9.37)
Thereby we could say that such a metastable state possesses a decay rate, (E), given by
(E) = 2 Im(E) .
(9.38)
Moving then to a thermal ensemble, we could analogously expect that

= 2 Im ,
(9.39)
37 J.S. Langer, Theory of the condensation point, Ann. Phys. 41 (1967) 108; Statistical theory of the decay of
metastable states, Ann. Phys. 54 (1969) 258.
131
where is the grand canonical free energy. It should be stressed, though, that this equation is to be
taken kind of as a definition for the moment: it would almost be a miracle if a real-time observable,
the nucleation rate, could really be determined exactly from a static Euclidean observable, the free
energy.
In any case, we can now pose the question whether could indeed develop an imaginary part?
It turns out that it can, as can be seen with the following argument38 . Consider the path integral
expression for the partition function,
Z

= T ln
D exp SE []
,
(9.40)
b.c.
where b.c. refers to the usual periodic boundary conditions. Let us assume that we can find two
different saddle points, each satisfying

SE
x) = (,
x) ,
x) = 0 .
= 0 , (0,
lim (,
(9.41)
=
|x|
One solution is a trivial one, 0, and the key assumption is that there is also a non-trivial
x).
solution which we denote by (,
Let us then consider fluctuations around the saddle point. Suppose that the fluctuation operator
around has one negative eigenmode:

2 SE
f (, x) = 2 f (, x) .
(9.42)
2 =
(For the non-negative modes, we define the eigenvalues through

2 SE
fn (, x) = 2n fn (, x) .)
2 =
(9.43)
Then, with Z0 Z[ = 0], the grand canonical potential gets contributions from both saddle
points:

Z Y
Z
dfn 1 2n fn2
df 1 2 f2
SE []
2
2
.
(9.44)
e
e
T ln Z0 + e
2
2
n0
The negative eigenmode requires a careful analysis, but essentially
s
s
Z
1
df 1 2 f2
2
1
e2
i
,
2
2
2 2
(9.45)
so that there indeed is an imaginary part. Assuming that the partition function around the regular
saddle point is much larger in absolute magnitude than that around the non-trivial one, then leads
to
1
n
o

T
2 2
det 2 SE []/
.
(9.46)
exp SE []

Z0
Such formulae are generically referred to as the semiclassical approximation. Somewhat more
precise versions of this formula will be given in Eqs. (9.61), (9.64) below.
The non-trivial saddle point is generally referred to as an instanton. By definition, an instanton is

a solution of the imaginary-time classical equations of motion, but it describes the rate of real-time
transitions.
38 S.R. Coleman, The Fate of the False Vacuum. 1. Semiclassical Theory, Phys. Rev. D 15 (1977) 2929 [Erratumibid. D 16 (1977) 1248].
132
Of course, the instanton needs to respect the boundary conditions in Eq. (9.41). Depending
on the geometric shape of the instanton solution within these constraints, we can give different
physical interpretations to the kind of tunnelling that the instanton describes. In the simplest
case, when the temperature is very small ( is very large), the Euclidean time direction is identical
with the space directions, and we can expect that the solution has 4d rotational symmetry:
Such a solution is said to describe quantum tunnelling. Indeed, had we kept ~ 6= 1, Eq. (9.46)
would have had the exponential exp(SE []/~).

On the other hand, if the temperature increases and decreases, the four-volume becomes
squeezed, and this affects the form of the solution39 . The solution can be depicted as follows:
0
0
Then we can say that quantum tunnelling and thermal tunnelling both play a role.
Eventually, however, the box becomes very squeezed, and we expect that the solution only
respects 3d rotational symmetry:
^
0
x
39 A.D.
Linde, Fate of the False Vacuum at Finite Temperature: Theory and Applications, Phys. Lett. B 100
(1981) 37.
133
In this situation, like in dimensional reduction, we can factorize the integration over the coordinate, and the instanton action becomes
Z
1
= 1 ~ d3 x LE S3d []
.
SE []
(9.47)
~
~
We say that then the transition takes place through classical thermal tunnelling
or the quantum mechanical
In typical cases, the exponential factor, be it the thermal one, S3d [],
one, SE []/~, is very large. Thereby the exponential is very small, and just how small it is, is
determined predominantly by the instanton action, rather than the fluctuation determinant which
does not have any exponential factors, and is therefore of order unity. Hence we can say that
play the dominant role in determining the
the instanton solution and its Euclidean action SE []
nucleation rate.
At the same time, from a theoretical point of view, it can be said that the real art in solving
the problem, is the computation of the fluctuation determinant around the saddle point solution40 .
In fact, the eigenmodes of the fluctuation operator can be classified into:
(1) one negative mode;
(2) a number of zero-modes;
(3) infinitely many positive modes.
We have already addressed the negative mode, which is the one responsible for the imaginary part,
so let us now look at the zero modes.
The existence and multiplicity of zero modes can be deduced from the classical equations of
motion and from the expression of the fluctuation operator. Indeed, assuming

Z Z
1
3
2
d x
SE =
d
( ) + V () ,
(9.48)
2
V
0
the classical equations of motion read
=0.
SE []/
= 0 2 + V ()
(9.49)
The fluctuation operator is given by

2
2 SE []/
,
= 2 + V ()
and differentiating Eq. (9.49) by yields the equation

h
i
= 0 .
2 + V ()
(9.50)
(9.51)
Therefore, is a zero mode. Note that a zero-mode exists (i.e. is non-trivial) only if the solution
depends on the coordinate x ; thus, the trivial saddle point = 0 does not lead to any zero
modes.
A non-trivial issue related to the zero modes is their proper normalization. In fact, the integrals
over the zero modes are only defined in a finite volume; they are proportional to the volume,
V = L3 , corresponding to translations of the critical bubble. A proper normalisation shows that
! 12
Z
SE
df0
=
L
(9.52)
2
2
40 C.G. . Callan and S.R. Coleman, The Fate of the False Vacuum. 2. First Quantum Corrections, Phys. Rev. D
16 (1977) 1762.
134
This can be shown as follows.

and L for 0 .
3 ,
2 ,
for 1 ,
Let us for simplicity consider a one-dimensional case (extent L). We normalize the eigenmodes
of the fluctation operator, defined as in Eq. (9.42), through
Z
dx fm fn = mn ,
(9.53)
and write a generic deviation from the saddle point solution as

X
X
= =
n
cn f n ,
n
(9.54)
where cn are coefficients (which we assume, for simplicity, to be real). We now need to be more
precise about the integration measure in Eq. (9.44); in fact it is sensible to define
Z
Y Z dcn
D
,
(9.55)
2
n
the point being that then the fluctuation determinant obtains a simple expression,
YZ
n

Z L
2 SE []
1
dc
n exp
dx
n
n
2
2
0
2

Z L
YZ dcn
1
2
exp
=
dx n cn
2
2
0
n
Y 1
=
n
n
1
q
.
2}
det{ 2 SE []/
(9.56)
Now, we return to the question of how to deal with the zero modes. The first task is to normalize
them according to Eq. (9.53). We note that the classical equation of motion, Eq. (9.49), implies (by
2=
multiplying with x and fixing the integration constant at infinity) a virial theorem, 12 (x )
V (). We then note that

Z
2=
dx (x )
dx
h1
2
i
2 + V ()
= SE []
SE .
(x )
(9.57)
In other words, the properly normalized zero-mode reads

1
f0 = p x .
SE
(9.58)
As the last step, we note that

1
1
c0 f0 (x) = p
.
(9.59)
x + p c0 (x)
c0 x (x)
SE
SE
p
Since the box is of size L, this means that we should restrict c0 (0, SE L), in order not to
double-count. Correspondingly,
! 12
Z
SE
dc0
=
L,
(9.60)
2
2
135
in accordance with Eq. (9.52).

We are now ready to put everything together. A more careful analysis40 shows that the factor
2 in Eq. (9.39) cancels against a factor 1/2 which we missed in Eq. (9.45). Thereby Eq. (9.46) is
correct expect for the treatment of the zero modes. Rectifying this point according to Eq. (9.52),
and expressing also Z0 in the Gaussian approximation, we arrive at
SE
2
1
! 42
det [ 2 + V ()]
2

e S E ,

det[ 2 + V (0)]
(9.61)
where det means that zero-modes have been omitted (but the negative mode is kept).
On the other hand, in the classical limit, = 0. Thereby there are only three zero-modes, and
2 Im T V
S3d
2T
1
! 32
det [2 + V ()]
2

e S3d .

det[2 + V (0)]
(9.62)
Furthermore, it turns out that the definition = 2 Im in Eq. (9.39) needs to be corrected into41
Im .
(9.63)
The result for the nucleation rate is thus

V ~=0
2
S3d
2T
1
! 23
det [2 + V ()]
S
2
3d

e T .

2
det[ + V (0)]
(9.64)
Comparing Eqs. (9.39), (9.63), we may expect Eq. (9.64) to be more accurate than Eq. (9.61) for
T>
.
2
(9.65)
It should be mentioned, though, the it is not clear whether the simplistic approach based on the
negative eigenmode really gives the correct answer42 ; rather, we should understand the analysis
in the sense that a rate exists, and the formulae as giving its rough order of magnitude.
Let us end by commenting on the analogous computation of the rate of baryon number violation.
In that case, the two vacua are actually degenerate; and the role of the field is played by the
Chern-Simons number, which is a suitable coordinate for classifying topologically distinct vacua.
However, the formalism itself is identical: in particular at low temperatures there is a saddle point
solution with 4d symmetry, which is just the usual instanton, while at high temperatures the saddle
point solution has 3d symmetry, and is referred to as sphaleron. Again there are also zero modes,
which have to be treated separately43 , as well as infinitely many positive modes.
41 I.
Affleck, Quantum Statistical Metastability, Phys. Rev. Lett. 46 (1981) 388.

Arnold, D. Son and L.G. Yaffe, The hot baryon violation rate is O(5w T 4 ), Phys. Rev. D 55 (1997) 6264.
43 P. Arnold and L.D. McLerran, Sphalerons, Small Fluctuations and Baryon Number Violation in Electroweak
Theory, Phys. Rev. D 36 (1987) 581.
42 P.
136
9.3.
Exercise 13
In scalar field theory, the determination of the saddle point solution (i.e. the critical bubble) and
its Euclidean action becomes solvable if we assume (i) the classical limit of high temperatures and
(ii) that the minima are almost degenerate. Show that then
16 3
S3d =
,
3 (p)2
where
broken
d 2V ()
(9.67)
p V (0) V (broken )
(9.68)
is the surface tension, and
(9.66)
is the pressure difference in favour of the broken phase. Note that in this limit the configuration
is a thin-wall bubble.
Eq. (9.66) implies that S3d for p 0. Therefore nucleation can only take place after
some supercooling, whereby p > 0 and S3d becomes finite.
The situation can be illustrated as follows:
Veff
~T
T~
c
_
broken
0
_
broken
|x|
In the classical limit, where there is no dependence on , the equation of motion becomes threedimensional,
=0.
(9.69)
2 + V ()
Assuming now spherical symmetry, this can be written as
d2 2 d
.
+
= V ()
dr2
r dr
The boundary conditions in Eq. (9.41) can be rephrased as
(
()
=0
,
d(0)
dr = 0
137
(9.70)
(9.71)
and the action reads

S3d = 4
2

1 d
.
+ V ()
2 dr
dr r2
(9.72)
Before proceeding, it is useful to note that Eqs. (9.70), (9.71) have a mechanical analogue. Indeed,
rewriting r t, V U , x, they correspond to a classical particle in a valley-problem
with friction. The particle starts at t = 0 from x > 0, near the top of the hill in the potential U ;
and rolls then towards the other top of the hill at the origin. The starting point has to be slightly
higher than the end point, because the second term in Eq. (9.70) acts as friction, and eats up part
of the energy. Therefore the broken minimum in V has to be lower than the symmetric one in
order for a solution to exist.
Proceeding now with the solution, we introduce the following ansatz. Suppose that at r < R,
the field is constant and has a value close to that in the broken minimum:
d
0,
dr
V (broken ) .
V ()
(9.73)
The contribution to the action from this region is then

4
S3d R3 V (broken ) .
3
(9.74)
For r > R, we assume a similar situation, but now the field is close to the origin:
d
0,
dr
V (0) 0 .
V ()
(9.75)
This region does therefore not contribute to the action. Finally, let us inspect the region at r R.
If R is very large, the term 2 /R in Eq. (9.70) is very small, and can be neglected. Thereby
d2
,
V ()
dr2
(9.76)
which can through multiplication with (r) be integrated into

1
2
2
d
.
V ()
dr
(9.77)
The contribution to the action becomes

S3d
The quantity
R+
2
d
4R
dr
dr
R
Z broken
d
4R2
d
dr
0
Z broken q
.
4R2
d 2V ()
2
R+
(9.78)
Z
2
q
broken
1 d
d 2V ()
+ V ()
dr
2 dr
0
(9.79)
represents the energy density of a planar surface, i.e. a surface tension.

Summing up the contributions to the action, we get
4
S3d (R) 4R2 R3 p ,
3
138
(9.80)
where p > 0 was defined according to Eq. (9.68). The so far free parameter R can now be
determined by extremizing,
(9.81)
R S3d = 0 ,
leading to the radius R = 2/p. Substituting back to Eq. (9.80), the saddle point action becomes
4 2
4
16 3
8 3
S3d = 4
=
.
(p)2
3 (p)2
3 (p)2
(9.82)
Finally, if we are very close to Tc , p here can be related to basic characteristics of the first
order transition. Indeed, the energy density is
e = Ts p
(9.83)
where the entropy density reads s = dp/dT . Across the transition, the pressure is continuous but
the energy density has a discontinuity, called the latent heat:
L e = Tcs = Tc
Therefore
p(T ) p(Tc ) +
d
p .
dT

dp
T
.
(T Tc ) = L 1
dT
Tc
(9.84)
(9.85)
At the same time, the surface tension remains finite at the transition point. Thereby the nucleation
action in Eq. (9.82) diverges rapidly as T Tc .
139
9.4.
Particle production rate
Let us consider a system where some particles interact strongly enough to remain in thermal
equilibrium, while others interact so weakly that they cannot follow thermal equilibrium. We can
imagine that particles of the latter type escape from the thermal system, either concretely (if
the system is of finite size) or in an abstract sense (being still within the same volume as the
thermal particles but not interacting with them). Familiar physical examples of such settings are
(i) the decoupling of weakly interacting dark matter particles in cosmology; (ii) the production of
electromagnetic hard probes, such as photons and + -pairs, from the QCD plasma generated
in heavy ion collision experiments; (iii) the neutrino emissivity of neutron stars, constituting the
most important process by which compact objects in astrophysics may cool down.
The purpose of this section is to develop the general formalism for addressing this phenomenon44 .
To keep the discussion concrete, we choose a particular example, however: the production of righthanded neutrinos within the model of Eq. (8.69),
LM =
1
a L + L
N
1 MN
F N
aR N
F L
Ni /
N
L
MSM ,
2
2
(9.86)
where we have added the Lagrangian LMSM of the Minimal Standard Model (MSM), describing
the thermalized degrees of freedom. The goal for now is to derive a master equation relating the
production rate of N s to a certain Greens function, evaluated already in Sec. 8.2. In the next
section, we then combine all ingredients and show how the dark matter abundance can be evaluated
in practice.
the
Let be the density matrix of the full theory, incorporating all degrees of freedom, and H
45
corresponding full Hamiltonian operator. Then the equation for the density matrix is
d
(t)
(t)] .
= [H,
dt
(9.87)
=H
MSM + H
N + H
int ,
H
(9.88)
i
in the form
We now split H
MSM is the Hamiltonian of the MSM, H

N is the free Hamiltonian of right-handed neutrinos,
where H
and Hint , which is proportional to the neutrino Yukawa couplings, contains the interactions between
right-handed neutrinos and the particles of the MSM:
Z
i
h

+ F N
aR N
(9.89)
Hint = d3 x F L
aL L .
is a Majorana spinor field operator. To find the concentration of right-handed neuHere now N
trinos, one has to solve Eq. (9.87) with some initial condition. We will assume that the initial
concentration of right-handed neutrinos is zero, that is
(0) = MSM |0ih0| ,
(9.90)
1
MSM ), 1/T , is the equilibrium MSM density matrix at a
where MSM = ZMSM
exp( H
temperature T , and |0i is the vacuum state for right-handed neutrinos.
0 = H
MSM + H
N as a free Hamiltonian, and H
int as an interaction term, one
Considering now H
0 t)
0 t),
can derive an equation for the density matrix in the interaction picture, I exp(iH
exp(iH
44 We follow the presentation in T. Asaka, M. Laine and M. Shaposhnikov, On the hadronic contribution to sterile
neutrino production, JHEP 06 (2006) 053 [hep-ph/0605209].
45
|i = H|i
,
dt
d
i d |ih| = [H,
|ih|] i d (t) = [H,
(t)] .
h| = h|H
dt
dt
dt
140
in the standard way:

i
d
0
0 I + eiH 0 t [H,
(t)]eiH 0 t + I H
I (t) = H
dt
0
0 I + eiH 0 t [H
0 +H
int , (t)]eiH 0 t + I H
= H
int , (t)]eiH 0 t
= eiH0 t [H
int eiH0 t eiH0 t (t)eiH0 t eiH0 t (t)eiH0 t eiH0 t H

int eiH0 t
= eiH0 t H
I , I (t)] .
= [H
(9.91)
I = exp(iH
0 t)H
int exp(iH
0 t) is the interaction Hamiltonian in the interaction
Here, as usual, H
I can be used to compute the time evolution
picture. Now, perturbation theory with respect to H
of I ; the first two terms read
I (t) = 0 i
I (t ), 0 ] + (i)2
dt [H
dt
I (t ), [H
I (t ), 0 ]] + ... ,
dt [H
(9.92)
I may break down

where 0 (0) = I (0). Note that perturbation theory as an expansion in H
at a certain time t teq due to so-called secular terms. Physically, the reason is that after teq
right-handed neutrinos enter thermal equilibrium and their concentration needs to be computed
by other means. Here we assumed that t teq and thus perturbation theory should work.
We are interested in the distribution function of the right-handed neutrinos. It is associated with
the operator
1 X
dN
a
a
,
(9.93)
d3 x d3 q
V s=1 q,s q,s
where a
q,s is the creation operator of a right-handed neutrino with momentum q and spin state s,
normalised as
{
ap,s , a
q,t } = (3) (p q)st ,
(9.94)
and V is the volume of the system. Then the distribution function dN/d3 x d3 q (number of righthanded neutrinos per d3 x d3 q) is given by

dN
dN (x, q)
Tr 3 3 I (t) .
d3 x d3 q
d xd q
(9.95)
Inserting here Eq. (9.92), the first term leads to a time-independent result, and the second term
I is linear in a
does not contribute since H
q,s and a
q,s (cf. Eqs. (9.89), (9.97), (9.98)). Thus, we get
2
that to O(F ) the rate of right-handed neutrino production reads
X

Z t
1
dN (x, q)
= R(T, q) Tr
a
q,s a
q,s
dt [HI (t), [HI (t ), 0 ]] .
(9.96)
d4 x d3 q
V
0
s=1
I appearing in Eq. (9.96) has the form in Eq. (9.89), except that
The interaction Hamiltonian H
evolves with the
we now interpret the field operators as being in the interaction picture. Since N
free Hamiltonian HN in the interaction picture, it has the form of a free on-shell field operator,
and can hence be written as
Z
i
Xh
d3 p
iP x
iP x
(x) =
p
,
(9.97)
a
u(p,
s)e
+
a
v(p,
s)e
N
p,s
p,s
(2)3 2p0 s=1
Z
i
Xh
d3 p
iP x
iP x
(x) =
p
N
,
(9.98)
a
(p,
s)e
+
a
(p,
s)e
p,s
p,s
(2)3 2p0 s=1
141
p
where we assumed the normalization in Eq. (9.94), and p0 Ep p2 + M 2 , P (p0 , p). The
spinors u, v satisfy the completeness relations
X
X
u(p, s)
u(p, s) = P/ + M ,
v(p, s)
v (p, s) = P/ M .
(9.99)
s
Inserting the free field operators into (the interaction picture version of) Eq. (9.89), we can rewrite
I as
H

Z
Z
3
X
iP x
I = d3 x p d p
p,s (x) eiP x + J (x) a
H
,
(9.100)
a
e
p,s
p,s
p,s
(2)3 2p0 s=1
where, denoting
j (x)
j (x)
(x) ,
(x)L
(x)(x)
L
,
(9.101)
(9.102)
the operators can be written as

Jp,s (x)
Jp,s
(x)
F j (x)aR v(p, s) + F u
(p, s)aL j (x) ,
F v(p, s)a j (x) + F j (x)a u(p, s) .
L
(9.103)
(9.104)
It remains to take the following steps:

(i) We insert Eq. (9.100) into Eq. (9.96) and remove the right-handed neutrino creation and
annihilation operators, by making use of Eq. (9.94). We note first that
n
o
n
o
[C,
|0i h0|]]
C|0ih0|
+ |0ih0|C B
Tr A [B,
= Tr A B
B|0ih0|
C C|0ih0|
B
n
o
C C AB
B
AC + C B
A |0i ,
= h0| AB
(9.105)
where we denoted
X
A =
a
q,s a
q,s ,
(9.106)
s=1
B
C
=
=
d x
d3 x
Z
Z
d3 p
p
(2)3 2p0
X
a
p,m Jp,m (x) eiP x
m=1
iP x
,
+ Jp,m (x) a
p,m e

X
d3 r
iRx
r,n (x ) eiRx + J (x ) a
p
.
a
e
r,n
r,n
r,n
(2)3 2r0 n=1
(9.107)
(9.108)
A non-zero trace only arises from structures of the type h0|

aa
a
a
|0i, i.e. the second and
third terms in Eq. (9.105). Thus, Eq. (9.96) becomes
Z
Z
Z
Z
n
X Z t
d3 r
1
d3 p
p
Tr
MSM
dt d3 x d3 x p
V s,m,n=1 0
(2)3 2p0
(2)3 2r0
h
Jr,n
(x )Jp,m (x)eiP xiRx h0|
ar,n a
q,s a
q,s a
p,m |0i
io
+ Jp,m
(x)Jr,n (x )eiP x+iRx h0|
ap,m a
q,s a
q,s a
r,n |0i .
(9.109)
R(T, q) =
Both expectation values evaluate to

h0|
ar,n a
q,s a
q,s a
p,m |0i = h0|
ap,m a
q,s a
q,s a
r,n |0i = ms ns (3) (r q) (3) (p q) .
142
(9.110)
Thereby
Z
Z
X Z t
1
1
3
dt
d
x
d3 x
V (2)3 2Eq s=1 0
E
D
Jq,s
(x )Jq,s (x)eiQ(xx ) + Jq,s
(x)Jq,s (x )eiQ(x x) ,
R(T, q) =
(9.111)
where from now on the expectation value refers to that with respect to MSM .
(ii) Inserting Eqs. (9.103), (9.104), we note that correlators of the type hj (x )j (x)i and
hj (x )j (x)i must vanish, since lepton numbers are conserved within the MSM. The rate
thus becomes
Z
Z
X Z t
1
1
3
R(T, q) =
dt
d
x
d3 x F F
V (2)3 2Eq s=1 0
Dh
i
u(q, s)aL j (x) eiQ(xx )
v(q, s)aL j (x )j (x)aR v(q, s) + j (x )aR u(q, s)

E
+(x x ) .
(9.112)
(iii) The spinors u, v appear in a form where the standard completeness relations mentioned in
Eq. (9.99) can be used. (In the first term, this requires writing
i
h
(9.113)
v(q, s)aL j (x )j (x)aR v(q, s) = Tr v(q, s)
v (q, s)aL j (x )j (x)aR .)
The mass terms M that are induced this way get projected out by aL , aR . Therefore,
Z
Z
X Z t
1
1
3
R(T, q) =
dt
d x d3 x F F
V (2)3 2Eq s=1 0
i
o
Dn h
/ aLj (x) eiQ(xx )
Tr Q
/ aLj (x )j (x)aR + j (x )aR Q
E
+(x x ) .
(9.114)
If we for a moment generalize the notation such that , account for Lorentz indices as well
as flavour indices, we can in fact rewrite this as
Z
Z
X Z t
1
1
3
R(T, q) =
dt
d
x
d3 x F F
V (2)3 2Eq s=1 0
h
io
E
Dn
(aR Q
/ aL) j (x )j (x) + j (x )j (x) eiQ(xx ) + (x x ) . (9.115)
(iv) Recalling the notation in Eqs. (8.43), (8.44),
Z
D
E
>
(Q)
dt d3 x eiQ(xx ) j (x)j (x ) ,
Z
D
E
<
(Q)
dt d3 x eiQ(xx ) j (x )j (x) ,
and noting that translational invariance implies

D
E
D
E
j (x)j (x ) = f (x x ) = f (x (x)) = j (x )j (x) ,
(9.116)
(9.117)
(9.118)
we can use Eq. (9.118) in the opposite direction and invert Eq. (9.116), to write
Z 4
D
E D
E
d P iP (xx ) >
j (x )j (x) = j (x)j (x )
e
(P )
=
(2)4
Z 4
d P iP (xx ) >
P P
e
(P ) . (9.119)
=
(2)4
143
Therefore, the two-point correlator in Eq. (9.115) can be written as

Z 4
i
d P iP (xx ) h >
<
hj (x )j (x) + j (x )j (x)i =
e
(P
)
(P
)
.
(2)4
(9.120)
(v) It remains to carry out the integrals over the space and time coordinates. Taking the limit
t and summing both terms in Eq. (9.115) together, yields
=
=
=
=
i
h
dt ei(QP )(xx ) + ei(P Q)(xx )

t
0
Z t h
i
0
0
0
0
3 (3)
V (2) (p q) lim
dt ei(q p )(tt ) + ei(p q )(tt )
t 0
Z 0
i
h 0 0
3 (3)
i(p q )t
i(p0 q0 )t
V (2) (p q) lim
+e
dt e
lim
d3 x
d3 x
Z
V (2)3 (3) (p q) lim
V (2)3 (3) (p q)
t
0
t t t
dt ei(p
q )t
dt ei(p
q )t
t
0
dtei(p
q0 )t
= V (2)4 (4) (P Q) ,
t t
(9.121)
which allows to cancel 1/V from Eq. (9.115) and remove P -integration from Eq. (9.120).
As a result of all these steps, we obtain (q 0 Eq )
R(T, q) =
h
i o
n
1
<
F F Tr Q
/ aL >
(Q) (Q) aR ,
3
0
(2) 2q
(9.122)
where we have returned to the convention that , label generations, and have expressed the Dirac
part through a trace. Making use of the fact that 1 nF (q 0 ) = nF (q 0 ), Eq. (8.53) yields
>
(Q) =
<
(Q)
2[1 nF (q 0 )] (Q) = 2nF (q 0 ) (Q) ,

0
2nF (q ) (Q) .
(9.123)
(9.124)
Observing furthermore that lepton generation conservation within the MSM restricts the indices
, to be equal, we finally obtain the master relation
R(T, q) =
3
h
i o
n
2nF (q 0 ) X
2
(Q)
+
(Q)
aR .
|F
|
Tr
Q
/
a
L
(2)3 2q 0 =1
(9.125)
We stress again that this relation is valid only provided that the number density of right-handed
neutrinos created is much smaller than their equilibrium concentration.
In summary, we have obtained a relation of the particle production rate, Eq. (9.96), to a finitetemperature spectral function, computed already in Eq. (8.88).
144
9.5.
Dark matter abundance in cosmology
In this section we continue the considerations of Sec. 9.4. Denoting by n(t, q) the phase space
density of right-handed neutrinos in either polarization state,
n(t, q) =
X dN (s) (t, x, q)
,
d3 x d3 q
s=1,2
(9.126)
Eq. (9.96) is of the form

n(t, q)
= R(T, q) ,
t
(9.127)
where R is given in Eq. (9.125),

R(T, q) =
3
h
i o
n
2nF (q 0 ) X
|F |2 Tr Q
/ aL (Q) + (Q) aR .
3
0
(2) 2q =1
(9.128)
This applies in flat spacetime, at a given temperature T . Note that in thermal equilibrium, in
the free limit, we would expect n(t, q) = 2nF (q 0 )/(2)3 , wherePthe factor 2 comes
the polarization
R
sum, and the factor (2)3 from that according to Eq. (9.126), s=1,2 N (s) = d3 x d3 q n(t, q). For
cosmological applications, the first task now is to generalize Eq. (9.127) to an expanding Universe,
where the temperature furthermore is a function of time.
In order to carry out the generalization, we first have to give a meaning to our variables, the
time t and the momenta q. In the following we mean by these the physical time and momenta,
i.e. the ones defined in a local Minkowskian frame. However, as is well known, local Minkowskian
frames at different times are inequivalent in an expanding background; in particular, the physical
momenta redshift. Carrying out the derivation of the rate equation in this situation is a topic of
general relativity rather than thermal field theory, so we only quote the result here: the upshot is
that the time derivative gets replaced with /t /t Hqi /qi 46 , and Eq. (9.125) becomes

n(t, q) = R(T, q) ,
(9.129)
Hqi
t
qi
where H is the Hubble parameter, H(t) a(t)/a(t),
and qi are the spatial components of q.

We will see presently that this form is in any case consistent with the expected redshift q(t) =
q(t0 ) a(t0 )/a(t), where a(t) is the scale factor.
Eq. (9.129) can be written in a simpler form through a change of variables. We note, first of all,
that R(T, q) and consequently also n(t, q) are only functions of q |q|. Changing correspondingly
the notation to n(t, q), R(T, q), Eq. (9.129) becomes

n(t, q) = R(T, q) .
(9.130)
Hq
t
q
0)
Introducing then an ansatz n(t, q) = n(t, q(t0 ) a(t
a(t) ), and noting that

a(t0 )
a(t0 )a(t)
d
q(t0 )
= q(t0 ) 2
= Hq ,
dt
a(t)
a (t)
(9.131)
Eq. (9.130) can be written as

a(t0 )
d
a(t0 )
n t, q(t0 )
= R T, q(t0 )
.
dt
a(t)
a(t)
(9.132)
46 J. Bernstein, Kinetic Theory in the Expanding Universe (Cambridge University Press, Cambridge, 1988);
E.W. Kolb and M.S. Turner, The Early Universe (Addison-Wesley, Reading, 1990).
145
This integrates immediately to

n(t0 , q(t0 )) =
t0
0

a(t0 )
,
dt R T (t), q(t0 )
a(t)
(9.133)
where we assumed the initial condition n(0, q) = 0, i.e., that there are no right-handed neutrinos
in the beginning.
At this point we need to recall the basic cosmological relations between the time t and the
temperature T . Assuming a homogeneous and isotropic metric,
ds2 = dt2 a2 (t) dx2k ,
(9.134)
and the energy-momentum tensor of an ideal fluid,

T = diag(e, p, p, p) ,
(9.135)
where e denotes the energy density and p the pressure, the Einstein equations, G = 8GT ,
reduce to
2
a
8G
k
e,
(9.136)
+ 2 =
a
a
3
d(ea3 )
= p d(a3 ) .
(9.137)
We will assume a flat Universe in the following, k = 0, and denote

1
G,
m2Pl
(9.138)
where mPl = 1.2 1019 GeV is the Planck mass.

We now combine the Einstein equations with basic thermodynamic relations. Assuming a system
with at most very small chemical potentials, the energy and entropy densities are related by
e = Ts p ,
(9.139)
where s = dp/dT is the entropy density. The derivative of Eq. (9.139) with respect to T yields
de
ds
=T
= Tc ,
dT
dT
(9.140)
where c is the heat capacity. Moving all the terms in Eq. (9.137) to the left-hand side, we get
a3 de + 3(e + p)a2 da = 0
da
de = 3T s
a
dT de
a
= 3T s .
dt dT
a
Inserting Eqs. (9.136), (9.138) and (9.140) results finally in
p
24 s(T ) e(T )
dT
=
.
dt
mPl
c(T )
(9.141)
(9.142)
(9.143)
(9.144)
In addition, Eq. (9.137) can also be written in terms of the well-known entropy conservation law:
0 =
=
=
=
d(ea3 ) + p d(a3 )
d([e + p]a3 ) a3 dp
d(T sa3 ) a3 sdT
T d(sa3 ) ,
146
(9.145)
where we inserted Eq. (9.139) and the definition s = dp/dT . Eq. (9.145) can in turn be expressed
as
1

s(T0 ) 3
a(t)
=
,
(9.146)
a(t0 )
s(T )
where t and T are related through Eq. (9.144).
It is furthermore conventional to introduce the effective numbers of massless bosonic degrees of
freedom geff (T ) and heff (T ) via the relations
e(T )
2 T 4
2 2 T 3
geff (T ) , s(T )
heff (T ) ,
30
45
(9.147)
where the prefactors follow by applying Eq. (9.139) and the line below it to the free result p(T ) =
2 T 4 /90 from Eq. (2.80). Given the equation-of-state of the plasma [i.e. the relation between the
pressure and the temperature, p = p(T )], geff (T ) and heff (T ) can according to Eq. (9.139) be found
from
45 dp
30 d p
, heff (T ) = 2 3
.
(9.148)
geff (T ) = 2 2
T dT T
2 T dT
Furthermore, we note that s(T )/T c(T ) = p (T )/T s (T ) = p (T )/e (T ) = p/e = c2s (T ), where we
identified the standard expression for the sound speed squared.
Combining Eqs. (9.144), (9.146) and the definitions above, we can now change variables in
Eq. (9.133), arriving at (q q(t0 ))
n(t0 , q) =
T0
dT
T3
1

5
mPl
T heff (T ) 3
p
.
R T, q
4 3 c2s (T ) geff (T )
T0 heff (T0 )
(9.149)
The integral over Eq. (9.149) with the measure d3 q gives the number density of right-handed
neutrinos, denoted by n(t0 ), which can subsequently be conveniently normalized with respect to
the total entropy density, Eq. (9.147), which produces the so-called yield parameter, Y :
Y (t0 )
=
=
n(t0 )
s(t0 )
r
1

Z
Z
mPl
45 4
q heff (T ) 3
dT
5
2
p
dq q
R T, T
3 2
2 2 T03 heff (T0 ) 4 3 0
T0 heff (T0 )
T0 T cs (T ) geff (T )
Z
Z
45 5
mPl
dT
p
dz z 2
R (T, T z) ,
(9.150)
2
5/2 T0 T 3 0
cs (T )heff (T ) geff (T )
where in the last step we substituted q = zT0 [heff (T0 )/heff (T )]1/3 . Note that Eq. (9.150) obtains
a constant value at low temperatures if R 0. Therefore the yield parameter is a good (i.e.
T0 -independent) characterization of the dark matter relic density.
In order to write the result more explicitly, we need to specify the function R(T, q). The relevant
information is contained in Eqs. (8.88), (8.89), (9.128). The Dirac algebra in Eq. (9.128) can be
carried out:
h i o
n
(9.151)
Tr Q
/ aL P/ 1 aR = 2Q P1 .
Furthermore, this can be written in various ways depending on the channel:
(4) (P1 + P2 Q) 2Q P1
=
=
(4) (P2 P1 Q) 2Q P1
=
=
(4) (P1 P2 Q) 2Q P1
(4) (P1 + P2 Q)[P12 + Q2 (Q P1 )2 ]

(4) (P1 + P2 Q)[P12 + Q2 P22 ] ,
(4) (P2 P1 Q)[(Q + P1 )2 P12 Q2 ]

(4) (P2 P1 Q)[P22 P12 Q2 ] ,
(4) (P1 P2 Q)[P12 + Q2 (Q P1 )2 ]

147
(4) (P1 + P2 + Q) 2Q P1
=
=
=
(4) (P1 P2 Q)[P12 + Q2 P22 ] ,

(4) (P1 + P2 + Q)[(Q + P1 )2 P12 Q2 ]
(4) (P1 + P2 + Q)[P22 P12 Q2 ] .
(9.152)
These factors are all constants, independent of the momenta p1 , p2 . Thereby we arrive at
Z
Z
3
X
X
1
d3 p1
d3 p2
2
2
2
2
R(T, q) =
|F
|
M
)
m
(m
c
c
(2)3 2q 0 =1
(2)3 2E1
(2)3 2E2
c

1
Q
(2)4 (4) (P1 + P2 Q) nF1 nB2 + 2

1
+ (2)4 (4) (P2 P1 Q) nB2 (1 nF1 ) +
Q
2
(2)4 (4) (P1 P2 Q) nF1 (1 + nB2 ) +

4 (4)
+ (2) (P1 + P2 + Q) (1 nF1 )(1 + nB2 ) ,
where E1
Q
1
Q
2
(9.153)
q
q
m2c + p2 , E2
m2c + (p + q)2 . In passing, it is interesting to note that
Eq. (9.153) resembles very much a collision term of a Boltzmann equation. We briefly recall
the structure of the latter in Sec. 9.6.
Let us analyse Eq. (9.153) more precisely. Suppose that q 0 > 0. The first question is, when do
the different channels get realized? Since all particles are massive, we can go to the rest frame of
one of them; it is then clear that the first channel gets realized for M > mc + mc ; the second
for mc > M + mc ; the third for mc > M + mc ; and the last one never. So, assuming that the
scalar mass (the Higgs mass) is larger than those of the produced particles, mc M, mc , we
can focus on the second channel, and the integral to be considered is
Z
Z
d3 p2
d3 p1
(2)4 (4) (P2 P1 Q) nB (u P2 )[1 nF (u P1 )] .
(9.154)
I
3
(2) 2E1
(2)3 2E2
Here u is the four-velocity of the thermal bath, and we have written the integral in a frameindependent way.
Remarkably, the integral in Eq. (9.154) can be written in a very simple form, in the hightemperature limit where the masses M 2 = Q2 and m2c = P12 of the produced particles can be
neglected47 . Denoting
p |p1 | , q |q| ,
(9.155)
we get
I
=
=
=
Z
d3 p2
d3 p1
(2)3 (3) (p1 + q p2 ) (2)(p + q E2 )nB (E2 )[1 nF (p)]
(2)3 2p
(2)3 2E2
Z
q

1
d3 p1
2 + (p + q)2 n (p + q)[1 n (p)]
m
p
+
q
B
F
1
c
(4)2 p(p + q)
Z
Z +1
q

1
dp p
dz p + q m2c + p2 + q 2 + 2pqz nB (p + q)[1 nF (p)] , (9.156)
8 0 p + q 1
where in the last step radial coordinates were introduced, with q as the z-axis. The delta function
gets realized when
(9.157)
p2 + q 2 + 2pq = m2c + p2 + q 2 + 2pqz ,
47 M.
Shaposhnikov, unpublished notes.
148
i.e. z = 1 m2c /2pq. This belongs to the interval (1, 1) if p > m2c /4q, so that
1

Z

1
dp p d q 2
2
2
mc + p + q + 2pqz
I=
nB (p + q)[1 nF (p)] .

8 m2 /4q p + q dz
m2 +p2 +q2 +2pqz=p+q
c
(9.158)
The derivative is trivial,

d
pq
=
,
dz
p+q
whereby we finally arrive at

I
=
=
1
8q
T
8q
m2c /4q
(9.159)
dp nB (p + q)[1 nF (p)]

m2c i
m2c h
1 nF T x +
.
dx nB T x + q +
4q
4q
(9.160)
In the last step we substituted p = m2c /4q + T x.
We can easily work out an upper bound for this integral if we make the further approximation
that T mc . Indeed, 1 nF 1, and
q+
m2c
mc 2
2
q
+ . . . m c .
= m c +
4q
m c
2
(9.161)
Thereby the distribution function nB can to a good approximation be replaced by the Boltzmann
distribution, whereby
Z
m2
m2
c
x q+ 4q
T q+ 4qc
T
=
.
(9.162)
e
dx e
I <
8q
8q
0
Inserting Eq. (9.162) into Eq. (9.153), we get
3
X
X
1
T
R(T, q) <
e
|F |2
m2c
(2)3 2q
8q
c
=1
m2
c
q+ 4q
(9.163)
Combining now with Eq. (9.150) produces
Z
Z
m2
3
c
2
X
z+
m
45 5 1 X
dz
z
dT
2
Pl
4T z
p
Y (t0 ) <
|F |2
m2c
e
5/2 128 4
2
2z2 2
T
T
c
(T
)h
(T
)
g
(T
)
0
T
eff
eff
0
s
c
=1
Z
3
m
X
X
45 5
mPl
dT
c
2
2
p
=
,
(9.164)
|F
|
m
K
c
4
2
T
T
64 13/2 =1
cs (T )heff (T ) geff (T )
T0
c
where K1 is a Bessel function. For mc /T 1, we can finally approximate

s
m
T mc
c
,
e
K1
T
2mc
(9.165)
and if we approximate c2s , heff and geff to be slowly varying, the remaining integral can be carried
out:
s
r Z
Z
T mc T =mc x
dT
1
f (T )
e
dx x7/2 f (mc x)e1/x

3
4
T
2m
m
2 0
c
T0
c
r Z
m
1
y=1/x
c
ey
dy y 3/2 f
=
3
m c 2 0
y
r
m c
5
1
f
,
(9.166)
m3c 2
2
2
149
where we replaced y in the argument of f by the value where the integral has obtained about
50% of its total magnitude. Thereby most of the dark matter abundance is indeed generated at
temperatures T <
mc , and
Y (t0 ) <
3
X mPl
135 5 X
|F |2
11/2
m c
256 2
c
=1
where we inserted (5/2) = 3 /4.
1
c2s
m c
2
heff
m c
2
r
,
m c
geff 2
(9.167)
Eq. (9.167) displays clearly the variables on which the dark matter abundance related to a specific
mechanism depends on: the coupling constants (|F |2 ), the mass of the decaying particle (mc ),
as well as the thermal history of the Universe (the functions c2s , heff , geff ).
150
9.6.
Appendix: relativistic Boltzmann equation
We recall here briefly the structure of the collision terms of a relativistic Boltzmann equation, and
compare the result with the quantum field theoretic formula in Eq. (9.153).
To understand the logic of the Boltzmann equation, a possible starting point is Fermis Golden
Rule for a decay rate:
Z
c
1n (Q) =
d1n |M1n |2 ,
(9.168)
2Eq
where the phase space integration measure is defined as

)( n
)
Z
Z (Y
m
n
m
X
X
Y
d3 pi
d3 qa
4 (4)
Pi Q
Qa ) .
(2) (
d1+mn
(2)3 2Eqa
(2)3 2Epi
a=1
a=1
i=1
i=1
(9.169)
Moreover, c is a statistical factor and M is the scattering amplitude.

Let now f (x, q) be a particle distribution function; we assume its normalization to be so chosen
that the total number density of particles at x is given by
Z 3
d q
n(x) =
f (x, q) .
(9.170)
(2)3
In particular, in thermal equilibrium, f (x, q) nF (Eq ) (or f (x, q) nB (Eq ) for bosons) is
independent of the position x, and determined uniquely by the temperature (and by possible
chemical potentials). At the same time, in vacuum, assuming a single plane wave regularized by a
finite volume V , we would have
f (x, q) =
(2)3 (3)
(q q0 ) ;
V
(9.171)
the logic is that then n(x), as defined by Eq. (9.170), evaluates to 1/V , while f (x, q0 ) = 1, where
we made use of (3) (0) = V /(2)3 .
To convert Eq. (9.168) into a Boltzmann equation, we identify the decay rate by t f /f for
the plane wave case, and deform then the structure to be Lorentz-covariant:
Eq Eq
fN 1
fN 1
q
.
t fN
x fN
(9.172)
We also modify the right-hand side of Eq. (9.168) by allowing for 1 + m particles in the initial
state, and adding Bose enhancement and Fermi blocking factors. Thereby
Z
fN
cX
q =
dN +mn
x
2 m,n

|M|2N +mn fN fa fm (1 fi ) (1 fn )

|M|2nN +m fi fn (1 fN )(1 fa ) (1 fm ) .
(9.173)
Here + applies for bosons and for fermions.
Let us compare this with Eq. (9.153). We may observe that Eq. (9.153) corresponds to the
gain terms of Eq. (9.173) (i.e. the last row), since the right-handed neutrinos are (by assumption)
non-thermal and escape: fN 0. At the same time, to obtain a complete match, we should
work out the scattering matrix elements, |M|2 , and the statistical factors, c. The strength of
the quantum field theoretic computation leading to Eq. (9.153) is that these automatically obtain
their correct values; its weakness is the allowing for (perhaps partial) equilibration is not possible
with Eq. (9.153), but can be achieved through the non-linear dependence of Eq. (9.173) on fN .
151
9.7.
Transport coefficients
An important class of observables are so-called transport coefficients. They generically describe
the response of the system to some external perturbation. For instance, electrical conductivity
determines the current that an external electric field leads to. Other standard transport coefficients
are diffusion constants related to how an externally induced charge excess in some region flattens
out, and shear and bulk viscosities, related to how excesses in momentum flow die away.
We illustrate here some basic aspects of the formalism related to transport coefficients with the
example of a decay rate. In this case, the external perturbation has displaced the system from
equilibrium by giving it a net charge of some type. We assume, however, that the charge is not
globally conserved, but that there are processes which violate charge conservation. In this case,
the system will relax back to equilibrium, i.e. the net charge will disappear, and the decay rate
describes how fast this process takes place. The formalism for treating this case can be found in
118, 122, 124, 126 of ref.48 .
(t) be the Heisenberg operator of some physical quantity, such as the lepton number. We
Let N
will consider a situation where the system is out-of-equilibrium in the sense that the expectation
(t) differs from its equilibrium expectation value, which we assume to be zero:
value of N
(t)ieq = 0 .
hN
(9.174)
(t)inon-eq , is assumed to evolve so slowly

The non-vanishing non-equilibrium expectation value, hN
(t)inon-eq should evolve towards zero, and
that all other reactions are in equilibrium. Then hN
(t)inon-eq is small in some sense [even though it should still be larger than typical thermal
if hN
2 ieq )1/2 ], we can expect the evolution to be described by an equation of first
fluctuations, (hN
(t)inon-eq :
order in hN

d
(t)inon-eq + O hN
(t)i2
(9.175)
hN (t)inon-eq = hN
non-eq .
dt
The coefficient may be called the decay rate. The goal would be to obtain an expression for
, describing dissipation, in terms of various equilibrium expectation values, of the type h...ieq ,
describing fluctuations.
Before proceeding, let us stress that the form of Eq. (9.175) applies to other transport coefficients
as well. Consider, as an example, electrical conductivity, , defined through
hji = E ,
(9.176)
where E is an external electric field. Taking a time derivative leads to

E
hji
=
= hji ,
t
t
(9.177)
where we used the Maxwell equation B E/t = hji, and assumed that the external probe
field B has been set to zero. Clearly Eq. (9.177) has precisely the form of Eq. (9.175).
In order derive an equation of the type in Eq. (9.175), let us write a generic linear response
(t)inon-eq . We can assume that at time t = , the system was in full equilibrium,
formula for hN
but then a source term was added to the Hamiltonian,
0 (t)N
(t) ,
H(t)
=H
(9.178)
(t)inon-eq from zero. Solving the equation of motion for the density
which will slowly displace hN
matrix (t),
i
d
(t) h
= H(t), (t) ,
(9.179)
i
dt
48 L.D.
Landau and E.M. Lifshitz, Statistical Physics, Part 1, 3rd Edition (Pergamon Press, Oxford, 1993).
152
to first order in the perturbation,

(t) () i
i
h
), () + ... ,
dt H(t
(9.180)
yields
(t)inon-eq
hN
(t)]
Tr [
(t)N
Z t
nh
i
o
(t ), () N
(t)
i
dt (t )Tr N
Z
D
E
(t), N
(t )] (t t ) (t ) ,
i
dt [N
eq
(9.181)
where
h...inon-eq Tr [
(t)(...)] ,
h...ieq Tr [
()(...)] ,
(9.182)
and the leading term disappeared because of the assumption in Eq. (9.174). We also made use of
0 and () commute. In addition, we inserted (t t ) and extended afterwards
the fact that H
the upper end of the integration to infinity, to stress the retarded nature of the correlator. The
equilibrium expectation value in Eq. (9.181) is called a linear response function.
Eq. (9.181) is now in principle of a type that could be used in Eq. (9.175), expect that it contains
(t )inon-eq which would be needed
the function (t ) on the right-hand side, rather the quantity hN
for indentifying . In some cases further input may allow to establish a bridge between (t ) and
(t )inon-eq , but below we choose to follow another approach. However, as we will see, Eq. (9.181)
hN
is still quite useful, because it shows explicitly that the essential information should reside in the
(t), N
(t )]ieq (t t ).
retarded Greens function h[N
Indeed, following 118 of ref.48 , let us define the correlator
oE
D1 n
(t), N
(0)
.
N
(t)
2
eq
(9.183)
The value (0) amounts to a susceptibility and is trivial to compute perturbatively; for the
number operator of a massless chiral fermion, for instance, one gets by taking the second derivative
from Eq. (7.41), and dividing by two because we are considering Majorana rather than Dirac
fermions,
i
h
2 ieq = 1 T 2 2 Tr e(H 0 N )
(0) = hN
Z
=0
h
i
2
2
= T ln Z(T, )
=0
f (T, )V i
2
T
=0
T 2
1
V T3 ,
6
(9.184)
where V is the volume. Furthermore, it is clear that (t) = (t) (by time-translational invariance). Finally, at large time separations the fluctuations are uncorrelated, so that
lim (t) = 0 .
(9.185)
Landau and Lifshitz now argue that the way (t) approaches zero (on time scales much larger
(t)inon-eq , i.e. certainly t
than the ones needed to reach equilibrium for any given non-zero hN
1/T ) is determined precisely by the coefficient in Eq. (9.175). We recall that according to
Eq. (9.181), must in any case be related to equilibrium expectation values. Concretely, the
behaviour should be48
(t) (0) exp(|t|) .
(9.186)
153
Let us at this stage define several auxiliary objects (cf. Sec. 8.1): a Fourier-transform of (t),
Z
()
dt eit (t) ;
(9.187)
a retarded correlator,
Dh
i
E
(t), N
(0) (t)
CR (t) i N
and its Fourier-transform,
CR ()
eq
(9.188)
dt eit CR (t) .
(9.189)
Standard thermodynamic relations (Eq. (8.17)) dictate that

h
i
()
= 1 + 2nB () ()
h
i1h
i
= 1 + 2nB ()
CR ( + i0+ ) CR ( i0+ ) .
2i
(9.190)
(9.191)
Carrying out the Fourier integral on Eq. (9.186) implies that, for small frequencies,
()
Z
(0)
dt e(i+)t +
dt e(i)t
2
.
(0) 2
+ 2
(9.192)
Small frequencies certainly means T ; in fact, taking 0 on both sides of Eq. (9.192)
and using Eq. (9.190) yields
1
T
(0) = lim () .
(9.193)
0
We recall that (0) here is given by Eq. (9.184), so that Eq. (9.193) in principle allows to determine
from the slope of the spectral function at zero frequency (which in turn is related to the retarded
correlator, as we expected from Eq. (9.181)). This is an example of a Kubo formula.
Eq. (9.193) looks actually somewhat suspicious. Indeed, consider the free limit. In this case, the
decay rate should disappear. Therefore, the right-hand side should diverge. However, according
to what we saw for the free spectral functions of single particles states (cf. Eqs. (8.36), (8.66)), we
would normally expect the spectral function to be zero at small enough frequencies (|| < m), in
which case the right-hand side would vanish.
The resolution to the paradox is that in the free limit the spectral function related to a conserved
charge contains a contribution of the type
()
=
(0) ()
(9.194)
1
(0) Im
.
i0+
(9.195)
Indeed, plugging this into the sum rule in Eq. (8.33), we get
D
E Z d
( )N (0) =
N
()nB ()e( ) = (0) ,
(9.196)
which shows that Eq. (9.194) corresponds to a Euclidean correlator which is contant in . This, indeed, is precisely what we expect from the correlator of a conserved quantity! Plugging Eq. (9.194)
into the right-hand side of Eq. (9.193) we indeed get infinity, so everything is in order after all.
154
Now, the structure in Eq. (9.194) is an extreme case of what is called a transport peak in the
spectral function. Once interactions are added, the peak should get smeared, so that the limit in
Eq. (9.193) is finite; in fact, according to Eqs. (9.190), (9.192), it should obtain the form ( T )
()
(0)
+ 2
= (0) Im
1
.
i
(9.197)
(9.198)
Therefore, in order to identify the transport coefficients, we need to be able to compute the spectral
function in the range . This is typically difficult because is generated by interactions, and
therefore of the type g 2 T , where g is some coupling constant: we need to be able to compute
() for soft energies g 2 T , where resummations play an important role. In any case, if the
computation is successful, the resulting structure should be of the type
() (0)
+ O(3 ) ,
(9.199)
which replaces the vanishing result at non-zero that was obtained in the free limit, Eq. (9.194).
155
10.
Further reading
[1] J.I. Kapusta,

Finite-temperature Field Theory
(Cambridge University Press, Cambridge, 1989).
Compact pedagogical presentation, concentrating mostly on Euclidean observables and
the imaginary-time formalism. The current notes borrow mostly from this classic treatise.
[2] M. Le Bellac,
Thermal Field Theory
A standard reference on real-time observables and the real-time formalism, and a detailed
introduction to particle production rate computations.
[3] J.I. Kapusta and C. Gale,
Finite-Temperature Field Theory: Principles and Applications
An update of ref.[1], including a full account of real-time observables, and reviews on
many recent developments.
[4] P. Arnold,
Quark-Gluon Plasmas and Thermalization,
Int. J. Mod. Phys. E 16 (2007) 2555
[arXiv:0708.0812].
Lecture notes on contemporary topics, particularly related to transport coefficients and
non-equilibrium phenomena such as thermalization.
156

Basics of Thermal Field Theory (By Mikko Laine)

Enviado por

Dados do documento

Título original

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Basics of Thermal Field Theory (By Mikko Laine)

Enviado por

Direitos autorais:

Formatos disponíveis

Basics of Thermal Field Theory1

Canonical partition function . . . . . . . . . . . . . . . . . . . . . . . . . . .

Path integral for the partition function . . . . . . . . . . . . . . . . . . . . .

Evaluation of the path integral for harmonic oscillator . . . . . . . . . . . .

Free scalar fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Path integral for the partition function . . . . . . . . . . . . . . . . . . . . .

Evaluation of thermal sums . . . . . . . . . . . . . . . . . . . . . . . . . . .

Properties of the Euler gamma and Riemann zeta functions . . . . . . . . .

Interacting scalar fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Naive free energy density to O(): ultraviolet divergences . . . . . . . . . .

Naive free energy density to O(2 ): infrared divergences . . . . . . . . . . .

Proper free energy density to O(): ultraviolet renormalization . . . . . . .

Proper free energy density to O( 2 ): infrared resummation . . . . . . . . .

Path integral for the partition function of a fermionic harmonic oscillator .

The Dirac field at finite temperature . . . . . . . . . . . . . . . . . . . . . .

Fermionic thermal sums . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Path integral for the partition function . . . . . . . . . . . . . . . . . . . . .

Gauge fixing and ghosts . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Feynman rules for Euclidean continuum QCD . . . . . . . . . . . . . . . . .

Thermal gluon mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Free energy density to O(g 3 ) . . . . . . . . . . . . . . . . . . . . . . . . . .

Low-energy effective field theories . . . . . . . . . . . . . . . . . . . . . . . . . . . .

The infrared problem of thermal field theory . . . . . . . . . . . . . . . . .

A simple example of an effective field theory . . . . . . . . . . . . . . . . .

Dimensionally reduced effective field theory for hot QCD . . . . . . . . . .

Complex scalar field with a finite chemical potential . . . . . . . . . . . . .

Effective potential and Bose-Einstein condensation . . . . . . . . . . . . . .

Dirac fermion with a finite chemical potential . . . . . . . . . . . . . . . . .

How about chemical potentials for gauge symmetries? . . . . . . . . . . . .

Different Greens functions . . . . . . . . . . . . . . . . . . . . . . . . . . .

From Euclidean correlator to spectral function . . . . . . . . . . . . . . . . 104

Hard Thermal Loop effective theory . . . . . . . . . . . . . . . . . . . . . . 110

Relation to classical kinetic theory . . . . . . . . . . . . . . . . . . . . . . . 118

Thermal phase transitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

Bubble nucleation rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131

Particle production rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140

Dark matter abundance in cosmology . . . . . . . . . . . . . . . . . . . . . 145

Appendix: relativistic Boltzmann equation . . . . . . . . . . . . . . . . . . 151

Transport coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152

Further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156

where m is the particle mass. The dynamics is governed by the Schr

where, for a time-independent Hamiltonian,

A Legendre transform leads to the classical Hamiltonian:

A most important example of a quantum mechanical system is provided by a harmonic oscillator:

In this case the energy eigenstates can be found explicitly:

All states are non-degenerate.

Canonical partition function

Path integral for the partition function

Thereby B = 2~|A|2 ; we choose A 1, so that B = 2~.

Z = Tr [e H ] = dx hx|e H |xi = dx hx|e ~ e ~ |xi .

hpi |e ~ H(pi ,xi )+O( ) |xi i

Moreover we need to note that on the very right, we have

Oftentimes this is symbolically written as a continuum path integral, as

Thereby Eq. (1.36) becomes

(iii) Restrict to the interval 0...~.

Evaluation of the path integral for harmonic oscillator

where the factor T is a convention. Periodicity requires

The values n = 2T n/~ are called Matsubara frequencies.

If we write xn = an + ibn , it follows that

In particular, b0 = 0, and xn xn = a2n + b2n . Thereby, we now have the representation

Here, a0 is called (the amplitude of) the Matsubara zero-mode.

and consider now C as an unknown coefficient.

then yields directly Eq. (1.17): the result is correct!