Solutions to Peskin and Schroder – Andrzej Pokraka

Problem 9.1: Scalar QED

This problem concerns the theory of a complex scalar field φ interacting with the electromagnetic
field Aµ . The Lagrangian is
1 2
L = − Fµν + (Dµ φ)∗ (Dµ φ) − m2φ φ∗ φ, (1)
4
where Dµ = ∂µ + ieAµ is the usual gauge-covariant derivative.
(a) Use the functional method of Section 9.2 to show that the propagator of the complex scale
field is the same as that of a real field:

i
= (2)
p p2 − m2φ + i

Also derive the Feynman rules for the interactions between photons and scalar particles; you
should find µ ν

p0
µ = −ie(p + p0 )µ = 2ie2 g µν (3)
p

(b) Compute, to lowest order, the differential cross section for e+ e− → φφ∗ . Ignore the electron
mass (but not the scalar particle’s mass), and average over the electron and positorn polariza-
tions. Find the asymptotic angular dependence and total cross section. Compare your results
to the corresponding formulae for e+ e− → µ+ µ− .
(c) Compute the contribution of the charged scalar to the photon vacuum polarization, using
dimensional regularization. Note that there are two diagrams. To put the answer into the
expected form,
Πµν (q 2 ) = (g µν q 2 − q µ q ν )Π(q 2 ),
it is useful to add the two diagrams at the beginning, putting both terms over a common
denominator before introduction a Feynman parameter. Show that, for −q  m2 , the charged
boson contribution to Π(q 2 ) is exactly 1/4 that of a virtual electron-positron pair.

9.1 (a)
Let us start with the action for the theory
Z  
1 2
S = d4 x − Fµν + (Dµ φ)∗ (Dµ φ) − m2φ φ∗ φ
4
Z  
1 2
= d4 x − Fµν + ∂µ φ∗ ∂ µ φ − ieAµ [φ∗ (∂µ φ) − (∂µ φ∗ )φ] + e2 Aµ Aµ φ∗ φ − m2φ φ∗ φ
4
Z  
4 1 2 ∗ 2 2 µ ∗ ∗ 2 2 2
IBP = d x − F − φ (∂ + m )φ − ieA [φ (∂µ φ) − (∂µ φ )φ] + e A |φ|
4 µν φ µ

≡ S0γ + S0φ + Sint . (4)

1
 Solutions to Peskin and Schroder – Andrzej Pokraka

The free theory generating functional is given by

Z Z Z
γ R 4 µ φ R 4 ∗ ∗
Z0 [Jem , Js , Js∗ ] = DA Dφ Dφ∗ eiS0 +i d x Aµ Jem +iS0 +i d x {Js φ+φ Js }
d4 x d4 y µ
( 12 Jem ν
(y)+Js∗ (x)D(x−y)Js (y))
R R
= Z0 [0, 0, 0] e−i (x)Dµν (x−y)Jem
. (5)
Here, we assume that the Faddeev and Popov procedure has been preformed to restrict the path
integral over A to physically unique field configurations. We have also ignored the overall infinite
constant factor that this procedure generates (recall that we are usually interested in ratios).
Expanding the exponential with the interaction term we obtain the full generating functional as
a series in e
Z Z Z  Z 
Z [Jem , Js , Js∗ ] = DA Dφ Dφ∗ Z0 [Jem , Js , Js∗ ] 1 + ie2 d4 x A2µ (x)|φ(x)|2 + O(e4 )
 Z 
× 1 + e d4 z Aµ (z) [φ∗ (z)(∂µ φ(z)) − (∂µ φ∗ (z))φ(z)] + O(e2 ) . (6)

We can now read off the propagator and interaction vertices. The scalar field propagator is
d4 k
Z
i
h0|Tφ(x1 )φ∗ (x2 )|0i = D(x1 − x2 ) = eik·(x1 −x2 ) (7)
(2π)4 p2 − m2φ + i
while the photon propagator is
d4 k
Z  
i kµ kν
h0|TAµ (x1 )Aν (x2 )|0i = Dµν (x1 − x2 ) = 4 2
gµν − (1 − ξ) 2 . (8)
(2π) k − i k
Fourier transforming the fields in the term,
Z
e d4 z Aµ (k) [φ∗ (z)(∂µ φ(z)) − (∂µ φ∗ (z))φ(z)]

d4 k d4 p d4 p0 µ
Z Z Z Z
0
= (−ie) d z 4
à (k)φ̃∗ (p0 )φ̃(p)(pµ + p0µ )e−iz·(p−p −k)
(2π)4 (2π)4 (2π)4
d4 p d4 p0 µ
Z Z
= Ã (p − p0 )φ̃∗ (p0 )φ̃(p)(−ie)(pµ + p0µ ) (9)
(2π)4 (2π)4
yields the vertex

p0
µ = −ie(p + p0 )µ (10)
p

while the term, Z

ie2 d4 x A2µ (x)|φ(x)|2 ,

yields the vertex

µ ν

= 2ie2 g µν . (11)

where the 2 comes from symmetry in the two A fields.

2
 Solutions to Peskin and Schroder – Andrzej Pokraka

9.1 (b)
To lowest order e+ e− → φφ∗ is given by the diagram:
e+ φ∗
0
p
k0
p = iM(e+ e− → φφ∗ ). (12)
q
k
e− φ
Applying the Feynman Rules we obtain
igµν
iM(e+ e− → φφ∗ ) = v̄(p0 )(−ieγ µ )u(p) (ie)(−k + k 0 )ν
(p + p0 )2
v̄(p0 )(−k
+ k
0 )u(p)
= ie2 . (13)
(p + p0 )2
The spin averaged matrix element is
2
e2

1X 1
|M(e+ e− → φφ∗ )|2 =
 0 
0 2
Tr (p − me )(me − k
)(p + me )(me − k
)
4 4 (p + p )


spin

e4

= 2(p · k) [p0 · k − p0 · k 0 ]
(p + p0 )4
+ 2(p · k 0 ) [p0 · k 0 − p0 · k]
+ (p · p0 ) [2k · k 0 − k · k − k 0 · k 0 ]

+ m2e [k · k + k 0 · k 0 − 2k · k 0 ] . (14)

In the centre of mass frame p = −p0 =⇒ p + p0 = (2Ep , 0) and k = −k0 =⇒ k + k 0 = (2Ek , 0).
This means that Ep = Ek ≡ E. We also assume that |p|  me so that E ≈ |p|. We need the dot
products
p · k = p0 · k 0 = Ep Ek − |p||k| cos θ ≈ E (E − |k| cos θ) (15)
0 0
p · k = p · k = Ep Ek + |p||k| cos θ ≈ E (E + |k| cos θ) (16)
0
p·p = Ep2 2
+ |p| ≈ 2E 2
(17)
0
k·k = Ek2 2
+ |k| = E + |k| 2 2
(18)
With these relations (and taking me = 0), the spin averaged matrix element squared becomes
!
1X + − ∗ 2 e4 |k|2 |k|2 2
m2φ
|M(e e → φφ )| ≈ − 2 2 cos θ − 2
4 4 E2 E E
spin

= 8π 2 α2 β 2 sin2 θ, (19)
q
where β = 1 − m2φ /E 2 is the velocity of the particles and α = 4πe2 is the fine structure constant.
Since this is a two particle reaction we can use the simplified cross section formula (4.84)
dσ(e+ e− → φφ∗ ) 1 |k| 1 X
= |M(e+ e− → φφ∗ )|2
dΩ 2E 16π 2 E 4
2
spin
2 3
α β
≈ sin2 θ. (20)
4E 2

3
 Solutions to Peskin and Schroder – Andrzej Pokraka

Integrating over the polar coordinate yields the total cross section

2πα2 β 3
σ(e+ e− → φφ∗ ) = . (21)
3E 2
Comparing our results to those of the e+ e− → µ− µ+
" ! ! #
dσ(e+ e− → µ− µ+ ) α2 β m2µ m2µ 2
= 1 + 2 + 1 − 2 cos θ ,
dΩ 4E E E
!
4πα2 β m2µ
σ(e+ e− → µ− µ+ ) = 1 + , (22)
3E 2 2E 2

we see that the angular dependence of the differential cross sections are very different. The scalar
particles have a are more likely to move perpendicular to the electron beam axis while the muons
are more likely to move along the electron beam axis.

9.1 (c)
The scalar contribution to the photon vacuum polarization is given by the following diagrams
k−q
µ ν = iMµν (23)
q q 1

and k

µ ν = iMµν
q q 2 . (24)

Now applying the Feynman rules (in d-dimensions) we obtain

dd k (−ie)(2k − q)µ (i)(−ie)(2k − q)ν (i)

Z
iMµν
1 = ,
(2π)d (k 2 − m2φ − i)((k − q)2 − m2φ + i)
dd k (2k − q)µ (2k − q)ν
Z
= e2 , (25)
(2π)d (k 2 − m2φ − i)((k − q)2 − m2φ + i)
dd k (2ie2 g µν )(i)
Z
iMµν
2 = ,
(2π)d k 2 − m2φ − i
dd k −2g µν
Z
2
=e . (26)
(2π)d k 2 − m2φ − i

4
 Solutions to Peskin and Schroder – Andrzej Pokraka

Following the hint we add these diagrams together before introducing Feynman parameters

iΠµν = iMµν1 + iM2

µν
!
dd k (−ie)(2k − q)µ (i)(−ie)(2k − q)ν (i) (2ie2 g µν )(i)
Z
= +
(2π)d (k 2 − m2φ − i)((k − q)2 − m2φ + i) k 2 − m2φ − i
!
d µ ν µν 2 2
(2k − q) (2k − q) − 2g ((k − q) − m )
Z
d k φ
= e2
(2π)d (k 2 − m2φ − i)((k − q)2 − m2φ + i)
!
d4 k 4k µ k ν − 2k µ q ν − 2k ν q ν + q µ q ν − 2g µν (k 2 − 2k · q + q 2 − m2φ )
Z
2
=e . (27)
(2π)d (k 2 − m2φ − i)((k − q)2 − m2φ + i)

Introducing the Feynman parameters x, y we have

1 1
dd k
Z Z Z
iΠµν = e2 dx dy δ(x + y − 1)
0 0 (2π)d
 
µ ν µ ν ν ν µ ν µν 2 2
 4k k − 2k q − 2k q + q q − 2g (k − 2k · q + q − m2φ ) 
× h i2 
(x + y)k 2 − 2yk · q + yq 2 − (x + y)m2φ − i)
 
Z 1 d µ ν µ ν ν ν µ ν µν 2 2 2
d k  4k k − 2k q − 2k q + q q − 2g (k − 2k · q + q − mφ ) 
Z
= e2 dx d  i2  . (28)
(2π)
h
0 k 2 − 2xk · q + xq 2 − m2φ − i

By letting ` = k − xq we complete the square in the denominator

dd `
Z 1 Z 
1
iΠµν = e2 dx 4(` + xq)µ (` + xq)ν − 2(` + xq)µ q ν − 2(` + xq)ν q µ
0 (2π)d [`2 − ∆]2

+ q µ q ν − 2g µν ((` + xq)2 − 2(` + xq) · q + q 2 − m2φ ) ,
1
dd `
Z Z 
2 1
=e dx 4(`µ `ν + x`[µ q ν] + x2 q µ q ν ) − 2(`µ q ν + xq µ q ν ) − 2(`ν q µ + xq ν q µ )
0 (2π)d [`2 − ∆]2

+ q µ q ν − 2g µν (`2 + 2xq · ` + x2 q 2 − 2(` · q + xq 2 ) + q 2 − m2φ ) ,
1
dd `
Z Z 
2 1
=e dx 4(`µ `ν + x2 q µ q ν ) − 2(xq µ q ν ) − 2(xq ν q µ ) + q µ q ν
0 (2π)d [`2 − ∆]2

µν 2 2 2 2
− 2g (` + (x − 1) q − mφ ) ,
1
dd ` 4`µ `ν − 2g µν `2 + (2x − 1)2 q µ q ν − 2g µν (x − 1)2 q 2 + 2g µν m2φ
Z Z
= e2 dx 2 ,
0 (2π)d [`2 − ∆]
4
1
dd ` −(2 − d )g µν `2 + (2x − 1)2 q µ q ν − 2g µν (x − 1)2 q 2 + 2g µν m2φ
Z Z
= e2 dx 2 , (29)
0 (2π)d [`2 − ∆]

5
 Solutions to Peskin and Schroder – Andrzej Pokraka

where we have ignored terms linear in ` and ∆ = −x(1 − x)q 2 + m2φ . Integrating over ` yields
Z 1
   
4
iΠµν = e2 g µν I1 (∆) + (2x − 1)2 q µ q ν − 2g µν (x − 1)2 q 2 + 2g µν m2φ I0 (∆) ,


dx − 2 −
0 d
Z 1    
4
= e2 g µν I1 (∆) + 2g µν m2φ I0 (∆) + (2x − 1)2 q µ q ν − 2(x − 1)2 g µν q 2 I0 (∆) ,


dx − 2 −
0 d
(30)

where

dd ` Γ(2 − d/2)
Z
1 i
I0 (∆) = = , (31)
(2π)d (`2 − ∆)2 (4π)d/2 ∆2−d/2 Γ(2)
dd ` `2 −i d Γ(1 − d/2)
Z
d
I1 (∆) = = =− ∆I0 (∆). (32)
(2π)d (`2 − ∆)2 (4π)d/2 2 ∆1−d/2 Γ(2) 2−d

Thus,
1   
2d − 4
Z
µν 2 2 µν 2 µ ν 2 µν 2


iΠ =e dx ∆ + 2mφ g + (2x − 1) q q − 2(x − 1) g q I0 (∆),
0 2−d
Z 1
i Γ()
= e2 dx 2x(1 − x)q 2 g µν + (2x − 1)2 q µ q ν − 2(x − 1)2 g µν q 2
 
2−
,
0 (4π) ∆ Γ(2)
Z 1  
 i 4π
= e2 dx −(2x − 1)2 q 2 g µν − q µ q ν + (2x − 1)g µν q 2


2
Γ(),
0 16π ∆
iα 1 4πe−γE
Z   
 1
dx −(2x − 1)2 q 2 g µν − q µ q ν + (2x − 1)g µν q 2


= + log . (33)
4π 0  ∆

For the last term in the first brackets we change variables y = x − 12 . Then (1 − 2x) → 2y while
∆ → −(y 2 − 14 )q 2 − m2φ . Since ∆ is even in y while the last term in the first brackets is linear the
integral is odd and vanishes. Thus,

1 4πe−γE
Z   
µν iα 2 µν µ ν 2 1
iΠ = q g −q q dx (1 − 2x) + log ,
4π 0  ∆

1 4πe−γE
Z   
iα 2 µν 1
= q g − qµ qν dx (1 − 2x)2 + log ,
4π 0  ∆
= q 2 g µν − q µ q ν iΠ(q 2 ),


(34)

where
" !#
1
4πe−γE
Z
2 α 2 1
iΠ(q ) = dx (1 − 2x) + log . (35)
4π 0  m2φ − x(1 − x)q 2

The physically relevant part, iΠ̂, is

iΠ̂(q 2 ) = iΠ(q 2 ) − iΠ(0)

!
m2φ
Z 1
α
= dx (1 − 2x)2 log . (36)
4π 0 m2φ − x(1 − x)q 2

6
 Solutions to Peskin and Schroder – Andrzej Pokraka

Problem 9.2: Quantum statistical mechanics

(a) evaluate the quantum statistical partition function

Z = Tr[e−βH ]

using the strategy of section 9.1 for evaluating the matrix elements of e−iHt in terms of
functional integrals. Show that once again one finds a functional integral, over functions
defined on a domain that is of length β and periodically connected in the time direction. Note
that the Euclidean form of the Lagrangian appears in the weight.
(b) Evaluate this integral for a simple harmonic oscillator,
1 2 1 2 2
LE = ẋ + ω x ,
2 2
by introducing a Fourier decomposition of x(t):
X 1
x(t) = xn √ e2πint/β .
n
β

The dependence of the result on β is a bit subtle to obtain explicitly, since the measure of the
integral over x(t) depends on β in any discretization. However, the dependence on ω should be
unambiguous. Show that, up to a (possibly divergent and β-dependent) constant the integral
reproduces exactly the familiar expression for the quantum partition function of an oscillator.
[You many find the identity
∞ 
z2
Y 
sinh z = z 1+
n=1
(nπ)2

useful.]
(c) Generalize this construction to field theory. Show that the quantum statistical partition func-
tion for a free scalar field can be written in terms of a functional integral. The value of this
integral is given formally by
−1/2
det(−∂ 2 + m2 )


where the operator acts on functions of Euclidean space that are periodic in the time direction
with periodicity β. As before, the β dependence of this expression is difficult to compute
directly. However, the dependence on m2φ is unambiguous. (More generally, one can usually
evaluate the variation of a functional determinant with respect to any explicit parameter in the
Lagrangian.) Show that the determinant indeed reproduces the partition function of relativistic
scalar particles.
(d) Now let ψ(t), ψ̄(t) be two Grassmann-valued coordinates, and define a fermionic oscillator by
writing the Lagrangian
LE = ψ̄ ψ̇ + ω ψ̄ψ.
This Lagrangian corresponds to the Hamiltonian

H = ω ψ̄ψ, with {ψ̄, ψ} = 1;

that is , to a simple two-level system. Evaluate the functional integral, assuming that the
fermions obey anti-periodic boundary conditions: ψ(t + β) = −ψ(t). (Why is this reasonable?)
Show that the result reproduces the partition function of a quantum-mechanical two-level
system, that is, of a quantum state with Fermi statistics.

7
 Solutions to Peskin and Schroder – Andrzej Pokraka

(e) Define the partition function for the photon field as the gauge-invariant functional integral
Z  Z 
1
Z = DA exp − d4 xE (Fµν )2
4

over vector fields Aµ that are periodic in the time direction with period β. Apply the gauge-
fixing procedure discussed in Section 9.4 (working, for example, in Feynman gauge). Evaluate
the functional determinants using the result of part (c) and show that the functional integral
does give the correct quantum statistical result (including the correct counting of polarization
states).

9.2 (a)
Let the system be described by the generalized coordinates {qi } and momenta {pi } where i =
1, 2, ..., n. The quantum mechanical partition function is given by

Z = Tr e−βH
 
Z
= dn q0 hq0 |e−βH |q0 i (37)

To evaluate (37) we split the temperature interval, β, into N equal slices of size  (we will eventually
take the  → 0 or equivalently the N → ∞ limit). The partition function becomes
Z
Z = dn q0 hq0 | e|−H .{z
. . e−H} |q0 i
N times
Z Z Z
= n
d q0 d q1 . . . dn qN −1 hq0 |e−H |q1 ihq1 | . . . |qN −1 ihqN −1 |e−H |q0 i
n

Z Z
= dn q1 . . . dn qN −1 hq0 |1 − H|q1 ihq0 | . . . |qN −1 ihqN −1 |1 − H|q0 i + O(2 ) (38)

We insert unity in terms of a complete set of momentum states,

Z
1 = dp|pihp|, (39)

N times to get
Z Z Z Z
Z = dn q0 . . . dn qN −1 dn p0 . . . dn pN −1

× hq0 |p0 ihp0 |(1 − H)|q1 ihq1 | . . . |qN −1 ihqN −1 |pN −1 ihpN −1 |(1 − H)|q0 i + O(2 ),
PN −1
ei n=0 pn ·qn
Z Z Z Z
n n n n
= d q0 . . . d qN −1 d p0 . . . d pN −1
(2π)(n/2)(N −1)
× hp0 |(1 − H)|q1 ihq1 | . . . |qN −1 ihpN −1 |(1 − H)|q0 i + O(2 ), (40)

At this point the derivation has been general. Let us specify to the standard form of the Hamiltonian,
1
Ĥ(p̂, q̂) = p̂ · p̂ + V (q̂) . (41)
2m

8
 Solutions to Peskin and Schroder – Andrzej Pokraka

Insertion into (41) yields

PN −1
ei n=0 pn ·qn
Z Z Z Z
n n n n
Z= d q0 . . . d qN −1 d p0 . . . d pN −1
(2π)(n/2)(N −1)
× hp0 |q1 i(1 − H(p0 , q1 ))hq1 | . . . |qN −1 ihpN −1 |q0 i(1 − H(pN −1 , q0 )) + O(2 ),
Z Z Z Z
= dn q0 . . . dn qN −1 dn p0 . . . dn pN −1

eip0 ·(q0 −q1 ) eipN −1 ·(q0 −q1 ) −H(p0 ,q1 )

× n
... e . . . e−H(pN −1 ,q0 ) + O(2 ),
(2π) (2π)n
Z Z Z Z
= dn q0 . . . dn qN −1 dn p0 . . . dn pN −1
2 2
eip0 ·(q0 −q1 )−(p0 /2m+V (q1 )) eipN −1 ·(q0 −q1 )−(pN −1 /2m+V (q0 ))
× . . . + O(2 ),
(2π)n (2π)n
Z Z Z Z
n n n
= d q0 . . . d qN −1 d p0 . . . dn pN −1
2 2
eip0 ·(q0 −q1 )−p0 /2m eipN −1 ·(q0 −q1 )−pN −1 /2m − PN −1 V (qk )
× . . . e k=0 + O(2 ). (42)
(2π)n (2π)n
The momentum integrals can be preformed,
Z
dpk i(pk ·(qk −qk+1 )+i|pk |2 /2m)  m n/2 − m |qk −qk+1 |2
e = e 2 , (43)
(2π)N 2π
to yield
|qk −qk+1 |2
 
 m n(N −1)/2 Z Z PN −1
−m −V (qk )
dn q0 . . . dn qN −1 e + O(2 ),
k=0 2 
Z=
2π
PN −1 m |qk −qk+1 |2
 
 m n(N −1)/2 Z Z
− +V (q k )
dn q0 . . . dn qN −1 e
2
+ O(2 ).
k=0 2 
= (44)
2π
This is of course the discretized form of the path integral with a Euclidean action,
I
U (q0 , qN , iβ) = Dq(iβ)e−SE [q] , (45)

where
β
|p|2
Z  
SE [q] = dτ + V (q) , (46)
0 2m
H
and represents the fact that the path integral is restricted to paths which start and end in the
same place. Since the initial and final paths differ by a ”time”, β, we require that the paths in the
trace be periodic in β (i.e., q(0) = q(β)).

9.2 (b)
We want to evaluate the integral (46) for the simple harmonic oscillator. Inserting the Fourier
decomposition (know exists because from part (a) we showed that the paths in the trace are periodic)
into the Euclidean Lagrangian we obtain,
 
1 XX 2πin 2πim
LE = xn xm + ω 2 e2πiτ (n+m)/β . (47)
2β n m β β

9
 Solutions to Peskin and Schroder – Andrzej Pokraka

Lets first evaluate the action

Z β
SE [x(τ )] = dτ LE
0
Z β ∞  
1 X 2πin 2πim
= dτ xn xm + ω e2πiτ (n+m)/β
2
0 2β
n,m=−∞
β β
∞
X  2πin 2πim 
1
= + ω 2 (βδn,−m )
2β n,m=−∞ β β
∞
4π 2 n2
 
1 X 2
= xn x−n +ω
2 n=−∞ β2
∞  2 2 
1 X 4π n
= |xn |2 + ω 2
.
2 n=−∞ β2
∞ 
ω 2 2 X 4π 2 n2

2
= x0 + 2
+ ω |xn |2 . (48)
2 n=1
β

We have used that fact that the expansion of x implies: x−n = x∗n (reality condition of x). Thus,
the path integral, Z, is a Gaussian
∞ 
" #
ω 2 2 X 4π 2 n2
Z 
Z = Dx(τ ) exp − x0 − 2
+ ω 2 |xn |2 . (49)
2 n=1
β

Here, the path integral is understood to be over the Fourier coefficients of x,

Z Z Z Z Z
Z ∝ dx0 dRex1 . . . dRex∞ dImx1 . . . dImx∞
∞ 
" !#
ω 2 2 X 4π 2 n2

2 2 2


× exp − x + +ω (Rexn ) + (Imxn ) ,
2 0 n=1 β2
∞ Z !
ω2 2
Z   Y   2 2 
4π n
= dx0 exp − x0 dRexn exp − 2
+ ω 2 (Rexn )2
2 n=1
β
∞ Z   2 2  !
Y 4π n 2 2
× dImxn exp − + ω (Imxn ) ,
n=1
β2
 2
r ∞ s
2π  Y π
=  ,
ω 2 n=1 4πβ22n2 + ω 2
∞
r !
2π Y β 2 π
= ,
ω 2 n=1 4π 2 n2 1 + β 22ω22
4π n
√ ∞
! ∞
!
2π Y β 2 βω 2 Y 1
= 2 ,
ω n=1
4πn2 2 βω n=1 1 + (βω/2)
π 2 n2
∞
!
√ Y β2 1
= 2πβ 2
,
n=1
4πn 2 sinh (βω/2)
1
= N (β) . (50)
2 sinh (βω/2)

10
 Solutions to Peskin and Schroder – Andrzej Pokraka

Now the overall constant N (β) is not well defined, however, neither was the integration measure.
Dividing by β in each integral will get rid of the extra powers of β in N (β).

Part (c)
From part (a) the field theoretic generalization follows readily:
Z
−βH
Z = hφ0 (x)|e |φ0 (x)i = Dφ e−SE [φ] (51)

where the φ are periodic, φ(x)|x0 =0 = φ(x)x0 =β . The Euclidean action is obtained from the
Minkowski action by Wick rotating the time component of x, x0 → −ix0 . For the free real scalar
field, the Euclidean action is
Z Z   Z  
1 1 1 2 2
SE = i d(−ix0 ) d3 x − (∂E µ φ)2 − mφ2 = d4 x (∂E µ φ)2 + mφ φ . (52)
2 2 2

To evaluate the statistical path integral we expand the field, φ, in its Fourier modes, as we did
for the Harmonic oscillator of part (b),
X e2πint/β 1 X
φ(x) = √ √ e−ik·x φ(kn0 , k)
n
β V k
X e2πint/β 1 X
≡ √ √ e−ik·x φn,k (53)
n
β V k

where we have taken the limit that space is a finite volume so that the k are discretized. The Fourier
coefficients, φ0,k may be complex. However, since φ(x) is real, they must satisfy the reality condition
φ−n,−k = φ∗n,k .
Inserting the mode expansion of φ into the action, we obtain
Z
1 µ
d4 x ∂E µ φ∂E φ + m2φ φ2


SE =
2
Z
1 1 XX µ 2πint/β −ik0 ·x
= d4 x ∂E µ e2πimt/β e−ik·x φm,k ∂E e e φn,k0
2 βV mn 0
kk
0

+m2φ e2πimt/β e−ik·x φm,k e2πint/β e−ik ·x φn,k0
4π 2 mn
Z  
1 4 1 X X 2πimt/β 2πint/β −ik·x −ik0 ·x 0 2
= d x e e e e φm,k φn,k0 − − k · k + mφ
2 βV mn 0 β2
kk
 2 2 
1 XX 4π m 2
= φm,k φ−m,−k + k · k + mφ
2 m β2
k
 2 !
1 XX 2 2πm 2
= |φm,k | + Ek
2 m β
k
 2 !
1 2
XX
2 2πm 2
= |φ0,0 | + |φm,k | + Ek . (54)
2 m>0
β
k

11
 Solutions to Peskin and Schroder – Andrzej Pokraka

Substituting the above into the statistical path integral we get

Z (  2 !)
Y 1 XX 2πm
Z= dReφn,k dImφn,k exp − |φ0,0 |2 − |φm,k |2 + Ek2
2 m>0
β
m>0,k>0 k
Y 1
= N (β)
sinh(βEk /2)
k
Y e−βEk /2
= N (β) (55)
1 − eβEk
k

which is just the relativistic partition function for Ek2 = k · k + m2φ .

To relate the partition function to a functional determinant, recall that for some operator Ô the
functional determinant is defined by
Z r
R 4
d xφÔφ 1
Dφ e = const × . (56)
detÔ
Integrating the Euclidean action by parts we get that the partition function is
Z  Z 
1 4 2 2 2

1
Z = Dφ exp d xφ −∂E φ + mφ φ ≡ const × r . (57)
2 
2 φ + m2 φ2
det −∂E φ

Part (d)
We are given the Lagrangian for a fermionic harmonic oscillator,

LE = ψ̄ ψ̇ + ω ψ̄ψ. (58)

The action is given by

I β  
SE = dτ ψ̄ ψ̇ + ω ψ̄ψ (59)
0

where the Grassman fields, ψ̄, ψ are anti-periodic ψ̄(τ + β) = −ψ̄(τ ), ψ(τ + β) = −ψ(τ ). We can
expand the anti-periodic Grassman field as
∞
X
ψ(τ ) = ψn e2πi(n−1/2)τ /β (60)
n=−∞

and the complex conjugate field as

∞
X
ψ̄(τ ) = ψ̄n e−2πi(n−1/2)τ /β (61)
n=−∞

where ψn and ψ̄ are Grassman numbers. It is easy to see that this expansion is indeed anti-periodic
∞
X
ψ(τ + β) = ψn e2πi(n−1/2)τ /β e2πin e−iπ = −ψ(τ ). (62)
n=−∞

After substituting the expansion for the Grassman fields the Euclidean action becomes
∞  
X 2πi(n − 1/2)
SE = ψ̄n ψn +ω . (63)
n=−∞
β

12
 Solutions to Peskin and Schroder – Andrzej Pokraka

The functional integral is then,

Z P∞ 2πi(n−1/2)
Z = Dψ̄Dψe− n=−∞ ψ̄n ψn [ β +ω ]

∞ Z Z
2πi(n−1/2)
dψn e−ψ̄n ψn [ ]
Y
+ω
∝ dψ̄n β

n=−∞
∞ Z   
2πi(n − 1/2)
Y Z
= dψ̄n dψn 1 − ψ̄n ψn +ω
n=−∞
β
∞ Z   
2πi(n − 1/2)
Y Z
= dψ̄n dψn 1 + ψn ψ̄n +ω
n=−∞
β
∞  
Y 2πi(n − 1/2)
= +ω
n=−∞
β
 Y ∞  Y∞  
−πi 2πi(n − 1/2) 2πi(−n + 1/2)
= +ω +ω +ω
β n=1
β n=1
β
∞
 Y " 2 #
−πi 2π(n − 1/2)
= +ω + ω2 . (64)
β n=1
β

This can be simplified by using the infinite product definition of cosh:

∞ 
x2
Y 
cosh x = 1+ . (65)
n=1
π 2 (n − 1/2)2

With this, the partition function becomes

Z ∝ cosh (βω/2) = eβω/2 + e−βω/2 . (66)

The last equation is just the partition function for a two-level system.

Part (e)
We are given the Euclidean Lagrangian for the photon field,
1
LE = − (Fµν )2
4
1
= − (∂µ Aν ∂ µ Aν − ∂µ Aν ∂ ν Aµ )
2
1 1
= − ∂µ (Aν ∂ µ Aν − Aν ∂ ν Aµ ) + Aν g µν ∂ 2 − ∂ µ ∂ ν Aµ


(67)
2 2
where A is periodic in τ with a period of β. The Euclidean action is
Z β
1
d4 xE Aν g µν ∂ 2 − ∂ µ ∂ ν Aµ


SE = (68)
2 0

and the partition function is

" #
Z Z β
1 4 µν 2 µ ν


Z= DA exp − d x E Aν g ∂ −∂ ∂ Aµ . (69)
2 0

13
 Solutions to Peskin and Schroder – Andrzej Pokraka

We must use the FP procedure to impose gauge invariance, so that we are only integrating over
unique field configurations. Let G(A) be a function of the photon field that when set to zero encodes
the gauge condition. We then insert unity into the functional integral
δG(Aα ) −SE [A]
Z Z  
Z = DA Dα δ (G(Aα )) det e . (70)
δα
1
where Aα α
µ = Aµ + e ∂µ α. We change variables from A to A to get

δG(Aα ) −SE [Aα ]

Z Z  
Z = DAα Dα δ (G(Aα )) det e . (71)
δα
Now we choose the generalized Lorenz gauge for G,
G(A) = ∂ µ Aµ (x) − ω(x) (72)
for any scalar function ω. Next note that the determinant is independent of A,
δ(∂ µ Aα
 2
µ − ω(x))
 
∂
det = det . (73)
δα e
Therefore the functional integral becomes
 2Z Z
∂
−SE [Aα ]
Z = det DAα Dα δ ∂ µ Aα µ (x) − ω(x) e
e
 2  Z Z
∂
= det Dα DA δ (∂ µ Aµ (x) − ω(x)) e−SE [A] . (74)
e
Next we integrate over Z with a Gaussian wight centred at ω = 0, to get
Z  2  Z Z Z
R 4 ω2 ∂
N (ξ) Dω e− d x 2ξ Z = N (ξ)det Dα DA Dω δ (∂ µ Aµ (x) − ω(x))
e
4 ω β 4 2 µν 2 µ ν
× e− d xE 2ξ e− 2 0 d xE Aν (g ∂ −∂ ∂ )Aµ
1
R R

 2  Z Z
∂ R 4 (∂ µ Aµ )2 Rβ 4 µν 2 µ ν
DA e− d xE 2ξ e− 2 0 d xE Aν (g ∂ −∂ ∂ )Aµ
1
= N (ξ)det Dα
e
 2  Z Z
∂ Rβ 4 µν 2 µ ν
DA e− 2 0 d xE Aν (g ∂ −(1− ξ )∂ ∂ )Aµ
1 1
= N (ξ)det Dα (75)
e
where N (ξ) is a function of that normalizes the Gaussian integral and the divergent integral over
the field α will cancel in the ratios that define correlation functions. Therefore we simply write
 2Z
∂ Rβ 4 µν 2 µ ν
DA e− 2 0 d xE Aν (g ∂ −(1− ξ )∂ ∂ )Aµ .
1 1
Z = det (76)
e
Choosing the Feynman gauge ξ = 1 the functional integral becomes
 2Z
∂ 1
Rβ 4 2 ν
Z = det DA e 2 0 d xE Aν ∂ A
e
 2Z Z
∂ 1
Rβ 4
d xE A0 ∂ 2 A0 1
Rβ 4 1 2 1
= det DA e
0 2 0 DA1 e 2 0 d xE A ∂ A
e
Z Rβ 4
Z Rβ 4
1 2 2 2 1 3 2 3
× DA2 e 2 0 d xE A ∂ A DA3 e 2 0 d xE A ∂ A
" #4
2

1
∝ det ∂ p
det (−∂ 2 )
1
∝ (77)
det (−∂ 2 )

14
 Solutions to Peskin and Schroder – Andrzej Pokraka

We can now use the results of part (c) in to evaluate the functional determinant:
1 1
= lim
det (−∂ ) m→0 det (−∂ 2 + m2 )
2

Y e−βEk /2
= lim N (β)
m→0 1 − eβEk
k>0
Y e−β|k|/2
= N (β) (78)
k
1 − eβ|k|

15