Quantum Field Theory

Winter Term 2015/2016

Prof. Dr. Nora Brambilla Solution to Exercise Sheet No. 9
Assigned on December 18th 2015
Discussion January 12th and 13th 2015
Dr. J. Tarrus-Castella, Dr. J. Segovia, Dr. J. Komijani, and S. Biondini

Exercise 9.1: One Loop Renormalization of QED

a)
Plugging in the definitions of the renormalized quantities we obtain
1 1
− F0,µν F0µν = − (∂µ A0,ν − ∂ν A0,µ )(∂ µ Aν0 − ∂ ν Aµ0 )
4 4
1 1 1
= − ZA Fµν F µν = − Fµν F µν − (ZA − 1) Fµν F µν , (1)
4 4 4

1 1 1 1
− (∂ µ A0,µ )2 = − ZA Zξ−1 (∂ µ Aµ )2 = − (∂ µ Aµ )2 − (ZA Zξ−1 − 1) (∂ µ Aµ )2 , (2)
2ξ0 2ξ 2ξ 2ξ

1/2
ψ̄0 (i∂/ − Qe0 A/0 )ψ0 = Zψ ψ̄i∂/ψ − Zψ ZA Ze ψ̄(µε eQA)ψ
/
/ + (Zψ − 1)ψ̄i∂/ψ − (Zψ ZA1/2 Ze − 1)ψ̄(µε eQA)ψ,
= ψ̄(i∂/ − µε eQA)ψ / (3)

−m0 ψ̄0 ψ0 = −Zψ Zm mψ̄ψ = −mψ̄ψ − (Zψ Zm − 1)mψ̄ψ. (4)

Putting everything together we obtain the desired result
LQED = LR,QED + LCT , (5)
1 1
LR,QED = − Fµν F µν − (∂ µ Aµ )2 + ψ̄ (i∂/ − µε eQA)
 
/ − m ψ, (6)
4 2ξ
1 1
LCT = −(ZA − 1) Fµν F µν − (ZA Zξ−1 − 1)(∂ µ Aµ )2
4 2ξ
1/2
+ (Zψ − 1)ψ̄i∂/ψ − (Zψ Zm − 1)mψ̄ψ − (Zψ ZA Ze − 1)µε Qeψ̄γ µ ψAµ . (7)
Now we want to derive the Feynman rules for the counterterms. First of all we state, that
the counterterm for the gauge fixing term doesn’t introduce any new vertices, just as the
gauge fixing itself. So let us start with the counterterm
1 1 1
− (ZA − 1)Fµν F µν = − (ZA − 1)∂µ Aν (∂ µ Aν − ∂ ν Aµ ) → (ZA − 1)Aν (∂µ ∂ µ Aν − ∂µ ∂ ν Aµ )
4 2 2
1 1
= (ZA − 1)Aν (2g µν Aµ − ∂ µ ∂ ν Aµ ) = (ZA − 1)Aν (2g µν − ∂ µ ∂ ν )Aµ ,
2 2
(8)

1
where we integrated by parts to shift the partial derivative. According to the given instruc-
tion for the derivation of the Feynman rules for the vertex, we multiply this expression by
i and replace ∂ µ by −ik µ , where k µ is the momentum of the photon. Then there is also
a symmetry factor involved. Imagine sandwiching this expression between two momentum
eigenstates of the photon field. Then you want to contract the eigenstates with the fields
and for that you have exactly two possibilites. Hence, you need to multiply the vertex by a
factor of 2. Finally, discard all fields. The resulting expression is
1
i2 (ZA − 1)((−i)2 k 2 g µν − (−i)2 k µ k ν ) = −i(ZA − 1)(k 2 g µν − k µ k ν ). (9)
2
Let us proceed with
(Zψ − 1)ψ̄i∂/ψ − (Zψ Zm − 1)mψ̄ψ. (10)
We repeat the same procedure and remark that in this case there is no symmetry factor
involved. If we sandwich the above expression between two electron momentum eigenstates,
then there is only one possibility to contract the eigenstates with the fields: ψ must be
contracted with the electron momentum eigenket, while ψ̄ is contracted with the electron
momentum eigenbra. All other contractions vanish. We denote the momentum of the elec-
tron pµ and obtain


i(Zψ − 1)i(−i)p/ − i(Zψ Zm − 1)m = i (Zψ − 1)p/ − (Zψ Zm − 1)m . (11)
Finally we are left with the counterterm
1/2
−(Zψ ZA Ze − 1)µε Qeψ̄γ µ ψAµ . (12)
No symmetry factors arise and since there are no derivatives with respect the fields, this
vertex will not depend on the field momenta. All we need to do is just multiply the expression
by i and discard the fields. This yields
1/2
−i(Zψ ZA Ze − 1)µε Qeγ µ . (13)

b)
The superficial degree of divergence for QED diagrams was determined in the exercise 7.1
and is given by
3
d = 4 − NA − Nψ . (14)
2
Let us now enumerate the six superficially divergent diagrams in QED.
• NA = 1, Nψ = 0 ⇒ Γ̃(0,1) with d = 3
This corresponds to a photon tadpole diagram. The diagram carries one Lorentz index
(because of one external photon) but it doesn’t depend on the momentum of the photon
q µ , so expanding it around q = 0 we obtain

q
µ = tµ (0). (15)

2
 To cancel such a divergence, the corresponding counterterm must be proportional to
a single photon field Aµ , which would violate gauge invariance. Therefore, if QED is
renormalizable, then the tadpole diagram must be finite, since otherwise we cannot
remove the corresponding divergence.

• NA = 2, Nψ = 0 ⇒ Γ̃(0,2) with d = 2
This is the familiar vacuum polarization diagram. The diagram carries two Lorentz
indices (due to two external photons) and can be naively expanded around q = 0 as

q −q
µ ν

= aµν (0) + qµ bν (0) + qν bµ (0) + gµν q 2 c(0) + qµ qν d(0) + finite terms. (16)

Since the diagram is invariant under q → −q, the linear terms must vanish. Then, to
cancel the possibly divergent aµν (0) contribution, we need a counterterm proportional
to Aµ Aν which would again violate gauge invariance. Hence, that contribution must
be finite. So we are left with gµν q 2 c(0) + qµ qν d(0) which is expected to be logarithmic
divergent and needs a counterterm of the form ∂ 2 A2 . Indeed, such a counterterm
arises from the renormalization of the photon field and is absorbed by ZA and Zξ .
In fact, if we remember that the photon self-energy is constrained to be of the form
(gµν q 2 −qµ qν )Π(q 2 ) by the Ward identity, then we immediately can say that c(0) = d(0).
The Ward identity allows us to choose ZA = Zξ , such that we can eliminate Zξ .

• NA = 3, Nψ = 0 ⇒ Γ̃(0,3) with d = 1
This is a triangle diagram with three external photons. The diagram carries three
Lorentz indices and can be expanded around q1 = 0, q2 = 0 as
µ
q1
ρ = aµνρ + [((q1 )µ bνρ (0) + permutations in µ, ν, ρ]
q2
q 1 + q2
ν
+ [((q2 )µ cνρ (0) + permutations in µ, ν, ρ] + finite terms.
(17)

To absorb the possible linear divergence of aµνρ , we need a counterterm proportional

to Aµ Aν Aρ . For the presumably logarithmic divergence of (q1 )µ bνρ and (q2 )µ cνρ , coun-
terterms proportional to ∂ µ Aν Aρ are required. Both would break gauge invariance,
such that this diagram must be finite.

• NA = 4, Nψ = 0 ⇒ Γ̃(0,4) with d = 0
This corresponds to a box diagram with four external photons. The diagram can be
at most logarithmic divergent, hence

µ ρ
= aµνρσ (0) + finite terms.
ν σ
(18)

3
 Here, to cancel a possible divergence of aµνρσ (0) we would require a counterterm pro-
portional to Aµ Aν Aρ Aσ which is not gauge invariant. Therefore, also this diagram
should be finite.

• NA = 0, Nψ = 2 ⇒ Γ̃(2,0) with d = 1
This corresponds to the familiar electron self-energy diagram. The diagram is at most
linear divergent with the expansion

p −p
= a(0) + p/b(0) + finite terms. (19)

The linear divergence of a(0) is canceled by a counterterm proportional to ψ̄ψ, while

the logarithmic divergence of b(0) requires something like ψ̄∂µ ψ. Both counterterms are
admittable and are in fact present in the counterterm Lagrangian as (Zψ Zm − 1)mψ̄ψ
and (Zψ − 1)ψ̄i∂/ψ.

• NA = 1, Nψ = 2 ⇒ Γ̃(2,1) with d = 0
Our last superficially divergent diagram is the one-loop vertex correction. The diver-
gence is at most logarithmic, hence we can expand as

p −p

q = aµ (0) + finite terms

µ (20)

The counterterm must contain one photon field and two fermion fields, i.e. it must
be proportional to Aµ ψ̄ψ. Such a term is not gauge invariant by itself, but together
with ψ̄i∂µ ψ gauge invariance is preserved. In the counterterm Lagrangian this term
1/2
manifests itself as −(Zψ ZA Ze − 1)ψ̄(µε eQA)ψ.
/

So we have seen, that out of the six possibly divergent vertex functions, only three are
allowed to be divergent, if QED is assumed to be renormalizable. The three divergent
functions have four divergent contributions

• c(0) from the vacuum polarization diagram (logarithmic divergence)

• a(0) from the electron self-energy diagram (linear divergence)

• b(0) from the electron self-energy diagram (logarithmic divergence)

• aµ (0) from the vertex correction diagram (logarithmic divergence)

for which we naively need four counterterms. However, by the virtue of the Ward identity
that allows us to relate two of the four renormalization constants, only three counterterms
are needed.
Looking at the diagrams that were supposed to be divergent but were shown to be finite,
it turns out that Γ̃(0,1) and Γ̃(0,3) are not only finite, but even zero. The reason for this is the

4
charge conjugation symmetry C, which is, as we know, an exact symmetry of QED. Under
the charge conjugation the photon field transforms as

CAµ (x)C −1 = −Aµ (x). (21)

Furthermore, we have

C|0i = |0i, (22)

since the vacuum is invariant under change conjugation. Now consider a 2n+1 point function
for the photons, i.e. a Green function with and odd number of photons

h0|Aµ1 Aµ2 . . . Aµ2n+1 |0i = h0|CAµ1 C −1 CAµ2 C −1 C . . . Aµ2n+1 C|0i (23)

= (−1)2n+1 h0|Aµ1 Aµ2 . . . Aµ2n+1 |0i (24)
= −h0|Aµ1 Aµ2 . . . Aµ2n+1 |0i = 0. (25)

This result is known as Furry’s theorem (after W.H. Furry), which forbids interactions of an
odd number of photons.

c)
 
The one loop correction to the vertex function Γ̃(2,1) (p, −p0 , q) corresponds to the follow-
µ
ing Feynman diagram without the external lines
k

p p+k p0 + k p0
q (26)
µ
and is given by

− iµε QeΛµ (p, p0 )

dD k i(k/ + p/0 + m) i(k/ + p/ + m) −ig αβ
Z
ε 3
= (−iµ Qe) γβ γµ γ α
(2π)D (k + p0 )2 − m2 + iη (k + p)2 − m2 + iη k 2 + iη
dD k γα (k/ + p/0 + m)γµ (k/ + p/ + m)γ α
 Z 
ε 2ε 2
= (−iµ Qe) −iµ (Qe) .
(2π)D ((k + p0 )2 − m2 + iη)((k + p)2 − m2 + iη)(k 2 + iη)
| {z }
Λµ (p,p0 )
(27)

c) (i)
We begin by evaluating the numerator, where the terms involving p/, p/0 and m are neglected

γα (k/ + p/0 + m)γµ (k/ + p/ + m)γ α → γα k/γµ k/γ α . (28)

5
In D dimensions the anticommutation relation for the gamma matrices

{γ µ , γ ν } = 2g µν (29)

remains unchanged, but for the trace of the metrics and for the products of gamma matrices
we now have

Tr(g) = g µν gµν = D, (30)

γ µ γµ = D. (31)

From these three relations it is straightforward to derive the following relations for the
products of gamma matrices in D dimensions.

γ µ γ ν γµ = −(D − 2)γ ν = −2γ ν − (D − 4)γ ν , (32)

γ µ γ ν γ ρ γµ = 4g νρ − (4 − D)γ ν γ ρ , (33)
γ α γ ν γ µ γ σ γα = −2γ σ γ µ γ ν + (4 − D)γ ν γ µ γ σ . (34)

Hence, for D = 4 − 2ε we obtain

γα k/γµ k/γ α = −2k/γµ k/ + 2εk/γµ k/

= 2k 2 γµ − 4kµ k/ + 2εk/γµ k/
= 2k 2 γµ − 4kµ k/ + O(ε)
= (2g αβ γµ − 4δµβ γ α )kα kβ + O(ε), (35)

where we used the anticommutation relation for the gamma matrices and that k/k/ = k 2 .
As far as the denominator is concerned, we now want to apply the Feynman parametriza-
tion. However, instead of doing the whole integral from scratch we can use the integral that
we have already computed in the exercise 6.2 e). Recall that there we had

dD k Πsj=1 k νj
Z
ν1 ...νs 2
Ir,s =µ , (36)
(2π)D Πrl=1 [(k + pl )2 − m2l + iη]
(−1)r−1 i h g νρ Γ(r − 3 + ) Γ(r − 2 + )
Z i
νρ 2  ν ρ
Ir,2 = (4πµ ) dX − E E .
(4π)2 2 (E 2 − F − iη)r−3+ (E 2 − F − iη)r−2+
(37)

and the integral we are now interested in is

Λµ (p, p0 ) = (−i(Qe)2 (2g αβ γµ − 4δµβ γ α ))

dD k
Z
2ε kα kβ
×µ 0
(2π) ((k + p ) − m + iη)((k + p)2 − m2 + iη)(k 2 + iη)
D 2 2

= (−i(Qe)2 (2g αβ γµ − 4δµβ γ α )I3,2,αβ . (38)

From Eq. (37) we obtain

Z
i 2 
I3,2,αβ = (4πµ ) dx1 dx2 dx3 δ(1 − x1 − x2 − x3 )
(4π)2
hg Γ() Γ(1 + ) i
αβ
× − E E
α β (39)
2 (E 2 − F − iη) (E 2 − F − iη)1+

6
To deduce what Eα and F are, we look at the denominator of Eq. (38) in terms of Feynman
parameters
x1 (k + p0 )2 − x1 m2 + x2 (k + p)2 − x2 m2 + x3 k 2 + iη
= k 2 + 2x1 k · p0 + 2x2 k · p + x1 p02 + x2 p2 − x1 m2 − x2 m2 + iη
= (k + x1 p0 + x2 p)2 − (x1 p0 + x2 p)2 + x1 (p02 − m2 ) + x2 (p2 − m2 ) + iη (40)
and compare this to
l2 − E 2 + F + iη, (41)
such that
Eα = (x1 p0 + x2 p)α , (42)
F = x1 (p02 − m2 ) + x2 (p2 − m2 ). (43)
Observe that the second piece of I3,2,αβ is finite for ε → 0 and can therefore be discarded,
since we are interested only in the singular contribution. We obtain
Z 1 Z 1−x1
i 2  gαβ Γ()
I3,2,αβ = (4πµ ) dx 1 dx2 + finite terms
(4π)2 | {z } 0 0 2 (E 2 − F − iη)
1+O(ε) | {z }
1
ε
+ finite terms

i gαβ 1 1−x1
Z Z
1 i gαβ 1
= dx1 dx2 = + finite terms. (44)
(4π)2 2 0 0 ε (4π)2 4 ε
Putting everything together we get
i gαβ 1
Λµ (p, p0 ) = −i(Qe)2 (2g αβ γµ − 4δµβ γ α )


+ finite terms (45)
(4π)2 4 ε
and since
(2g αβ γµ − 4δµβ γ α )gαβ = 2Dγµ − 4γµ = (8 − 4ε)γµ − 4γµ = 4γµ + O(ε), (46)
the final result is
e2 1 α 2 1
Λµ (p, p0 ) = Q 2
γµ + finite terms = Q γµ + finite terms, (47)
(4π)2 ε 4π ε
where we used that α = e2 /4π.
Now we demand that the one-loop vertex function
 

(2,1)

0 ε α 21 1/2
Γ̃ (p, −p , q) = −µ Qeγµ 1 + Q + finite terms + (Zψ ZA Ze − 1) (48)
µ 4π ε
is finite, which means that the 1ε -divergence must be absorbed by the counterterm. In the
M S scheme we only subtract the divergence and hence obtain at one loop

1/2 α 1
(ZψM S ZAM S Ze − 1) = 12 δAM S + δeM S + δψM S + O(α2 ) = − Q2 . (49)
4π ε
Here we used that at O(α) we have
q
MS M S 1/2 MS 2
Ze = (1 + δψ + O(α )) 1 + δAM S + O(α2 )(1 + δeM S + O(α2 ))


Zψ ZA
1
= (1 + δψM S + O(α2 ))(1 + δAM S + O(α2 ))(1 + δeM S + O(α2 ))
2
= 1 + 2 δA + δe + δψ + O(α2 ).
1 MS MS MS
(50)
1 α
In the M S scheme, ε
in − 4π Q2 1ε must be replaced by 1
ε
− γE + log(4π).

7
c) (ii)
First of all we want to bring Λµ (p, p) to the requested form by using the given identity. We
have
dD k γα (k/ + p/ + m)γµ (k/ + p/ + m)γ α
Z
2ε 2
Λµ (p, p) = −iµ (Qe)
(2π)D ((k + p)2 − m2 + iη)((k + p)2 − m2 + iη)(k 2 + iη)
dD k
Z
1 1 1
= −iµ2ε (Qe)2 D
γα γµ γα 2
(2π) k/ + p/ − m + iη k/ + p/ − m + iη k + iη
| {z }
− ∂p∂µ 1
/−m
/+p
k

dD k
Z
∂ 1 1
= µ iµ2ε (Qe)2 γα γ α
∂p (2π)D k/ + p/ − m k 2 + iη
| {z }
≡Σ(p,m)
∂
= Σ(p, m). (51)
∂pµ
Then, we plug our vertex function at one loop into the renormalization condition
 
!
ū(p) Γ̃(2,1) (p, −p0 , 0)u(p) = −µε Qeū(p)γµ u(p) for p2 = m2 . (52)
µ

to obtain
h i
1/2 !
ū(p) Λµ (p, p) + (Zψ ZA Ze − 1)γµ u(p) = 0 for p2 = m2 . (53)
or
 
∂ 1/2 !
ū(p) Σ(p, m) + (Zψ ZA Ze − 1)γµ u(p) = 0 for p2 = m2 . (54)
∂pµ
From this we conclude that


1/2 OS ∂ 
(ZψOS ZAOS 1 OS
δeOS δψOS 2


Ze − 1)γµ = δ
2 A
+ + γµ + O(α ) = − µ Σ(p, m) . (55)
∂p p2 =m2

The identity that we used to rewrite Λµ (p, p) can be easily shown by direct computation.
∂ 1 ∂ k/ + p/ + m
− = −
∂pµ k/ + p/ − m ∂pµ (k + p)2 − m2
2(k + p)µ (k/ + p/ + m) γµ
= 2 2 2
−
((k + p) − m ) (k + p)2 − m2
2(k + p)µ (k/ + p/ + m) − γµ ((k + p)2 − m2 )
= . (56)
((k + p)2 − m2 )2
Compare this to
1 1 (k/ + p/ + m)γµ (k/ + p/ + m)
γµ =
k/ + p/ − m + iη k/ + p/ − m + iη ((k + p)2 − m2 )2
 
−γµ (k/ + p/ − m) + 2(k + p)µ (k/ + p/ + m)
=
((k + p)2 − m2 )2
2(k + p)µ (k/ + p/ + m) − γµ ((k + p)2 − m2 )
= (57)
((k + p)2 − m2 )2
to convince yourself that the identity indeed holds.

8
d)
The corresponding Feynman diagram is

p p+k p
(58)

and the electron self-energy function at one loop is this diagram without the external legs,
i.e.
dD k i(k/ + p/ + m)
Z
Ψ 2ε 1
iΣ (p, m) = −iµ D
(−ieQγ µ ) 2 2
(−ieQγµ ) 2 (59)
(2π) (k + p) − m + iη k + iη
dD k γ µ (k/ + p/ + m)γµ
Z
2 2ε 2
= i µ (eQ) . (60)
(2π)D ((k + p)2 − m2 + iη)(k 2 + iη)
Using

γ µ k/γµ = −Dk/ + 2k/ = −2(1 − ε)k/, (61)

such that

γ µ (k/ + p/ + m)γµ = −2(1 − ε)(k/ + p/) + Dm (62)

we can rewrite the self-energy function as (using integrals from the exercise sheet 6)

ΣΨ (p, m) = i(Qe)2 −2(1 − ε) γα I2,1 α

  
+ p/I2,0 + Dm I2,0 , (63)

with
Z
i 2  Γ()
I2,0 = 2
(4πµ ) dx dy δ(1 − x − y) 2 , (64)
(4π) (E − F − iη)
−i
Z
α 2  Γ()
I2,1 = 2
(4πµ ) dx dy δ(1 − x − y)E α 2 . (65)
(4π) (E − F − iη)
and

E α = xpα , (66)
F = −x(m2 − p2 ) (67)

for both integrals. The integrals evaluate to

Z 1
i 2 ε Γ(ε)
I2,0 = 2
(4πµ ) dx 2 2 (68)
(4π) 0 (x p + x(m2 − p2 ) − iη)ε
Z 1
α −i 2 ε α Γ(ε)
I2,1 = (4πµ ) dx xp . (69)
(4π)2 0 (x2 p2 + x(m2 − p2 ) − iη)ε

9
Hence,
 Z 1 
Ψ α 2 Γ(ε)
Σ (p, m) = p/ Q 2(1 − ε)(4πµ2 )ε dx(1 − x) 2 2
4π 0 (x p + x(m2 − p2 ) − iη)ε
| {z }
Σψ 2 2
V (p ,m )
 Z 1 
α 2 2 ε Γ(ε)
+ m − Q D(4πµ ) dx 2 2
4π 0 (x p + x(m2 − p2 ) − iη)ε
| {z }
Σψ 2 2
S (p ,m )

= p/ΣψV (p2 , m2 ) + mΣψS (p2 , m2 ). (70)

d) (i)
In the minimal subtraction schemes we demand that our self-energy function at one-loop i.e.

Γ̃(2,0) (p, −p)

= p/(1 + ΣψV ) + m(ΣψS − 1) + (Zψ − 1)p/ − (Zψ Zm − 1)m
= p/(Zψ + ΣψV ) + m(ΣψS − Zψ Zm )
= p/(1 + δψ + O(α2 ) + ΣψV ) − m(1 + δψ + δm + O(α2 ) − ΣψS ) (71)

is finite. Note that here we inserted Eq. (70) into the definition of Γ̃(2,0) (p, −p) from the ex-
ercise sheet. Now to find the divergent part of the self-energy function we expand ΣψV (p2 , m2 )
and ΣψS (p2 , m2 ) in ε . This yields

α 2
ΣψV (p2 , m2 ) = Q 2(1 − ε)(1 + ε log(4πµ2 ) + O(ε2 )) 1


ε
− γE + O(ε)
4π
Z 1
× dx(1 − x)(1 − ε log((x2 p2 + x(m2 − p2 ) − iη) + O(ε2 )
0
Z 1
4πµ2
   
α 2 1
= Q 2 dx(1 − x) − γE + log − 1 + O(ε)
4π 0 ε x2 p2 + x(m2 − p2 ) − iη
Z 1
µ2
  
α 2
= Q ∆−1+2 dx(1 − x) log + O(ε)
4π 0 x2 p2 + x(m2 − p2 ) − iη
(72)

and
α
ΣψS (p2 , m2 ) = − Q2 4(1 − 12 ε)(1 + ε log(4πµ2 ) + O(ε2 )) 1ε − γE + O(ε)


Z 4π
1
× dx(1 − ε log((x2 p2 + x(m2 − p2 ) − iη) + O(ε2 ))
0
Z 1 
4πµ2
  
α 2 1 1
=− Q 4 dx − γE + log − + O(ε)
4π 0 ε x2 p2 + x(m2 − p2 ) − iη 2
Z 1
µ2
  
α 2
= − Q 4∆ − 2 + 4 dx log + O(ε), (73)
4π 0 x2 p2 + x(m2 − p2 ) − iη

10
with ∆ = 1ε − γE + log(4π). Plugging these results back into Eq. (71) and demanding that
the infinities are to be absorbed by the renormalization constants δψ and δm , we can read
off the δi and Zi values. In the M S scheme we have
α 2 α 2
δψM S = − Q ∆, ZψM S = 1 − Q ∆, (74)
4π 4π
MS α α 2
δm = − Q2 3∆, Zm MS
=1− Q 3∆. (75)
4π 4π
and in the M S scheme we just have to perform the replacement ∆ → 1ε .

d) (ii)
In the on-shell scheme we formulate the renormalization condition in terms of the renormal-
ized self-energy

ΣψR = Σψ (p, m) + (Zψ − 1)p/ − (Zψ Zm − 1)m

= (ΣψV + Zψ − 1)(p/ − m) + (ΣψS + ΣψV + Zψ (1 − Zm ))m
= (ΣψV + δψ )(p/ − m) + (ΣψS + ΣψV − δm + O(α2 ))m, (76)

where we plugged Eq. (70) into the definition of the renormalized self-energy from the
exercise sheet. Now we start expanding the defining relation
 
ψ 1
lim ΣR (p, m) u(p)
p2 →m2 p/ − m
 
ψ 2 2 m ψ 2 2 ψ 2 2 !
= 2lim 2 ΣV (p , m ) + δψ + (ΣS (p , m ) + ΣV (p , m ) − δm ) u(p) = 0 (77)
p →m p/ − m

around p2 = m2
 
ψ 2 2 m ψ 2 2 ψ 2 2
ΣV (p , m ) + δψ + (Σ (p , m ) + ΣV (p , m ) − δm ) u(p)
p/ − m S
 
ψ m 
2 2
= ΣV (m , m ) + δψ + ΣψS + ΣψV (m2 , m2 ) − δm
p/ − m
 0  
2 2 ψ ψ 2 2 2 2
+ (p − m ) ΣS + ΣV (m , m ) + O(p − m ) u(p)

m  ψ  
= ΣS + ΣψV (m2 , m2 ) − δm + ΣψV (m2 , m2 )
p/ − m
 0 
ψ ψ 2 2 2 2
+ δψ + m(p/ + m) ΣS + ΣV (m , m ) + O(p − m ) u(p) 2 → 2 0. (78)
p →m

The requirement that the above expression vanishes in the limit p2 → m2 leads to two
conditions
 
δm = ΣψS + ΣψV (m2 , m2 ), (79)
"  #
ψ 2 2 ∂  ψ ψ

2

2 
lim ΣV (m , m ) + δψ + m(p
/ + m) ΣS + Σ V (p , m ) 2 2 u(p) = 0 (80)
p2 →m2 ∂p2 p =m

11
Since the second condition is to be evaluated on the mass shell, we can use the Dirac equation
to rewrite
 
(p/ + m)u(p)p2 =m2 = 2mu(p)p2 =m2 (81)

such that this condition becomes

 
2 ∂

ψ ψ ψ
δψ = − ΣV + 2m ΣS + ΣV  2 2.
 (82)
∂p2 p =m

Now we evaluate both expression explicitly using our previous results for ΣψS and ΣψV .
 Z 1  2 
α 2 µ
δm = Q −3∆ + 1 + dx(−2 − 2x) log
4π 0 x m2
2
 Z 1   2 
α 2 µ
= Q −3∆ + 1 − 2 dx(1 + x) log − 2 log(x)
4π 0 m2
   2 
α 2 µ
= Q −3∆ + 1 − 3 log +5
4π m2
  2 
α 2 4 µ
= − Q 3 ∆ + + log . (83)
4π 3 m2

To evaluate δψ we need to expand ΣψV (m2 , m2 ) for small ε. As usual, to regulate UV singu-
larities we assume that ε > 0.
 Z 1  2 
ψ 2 2 α 2 µ
ΣV (m , m ) = Q ∆−1+2 dx(1 − x) log
4π 0 x m2
2
 Z 1   2 
α 2 µ
= Q ∆−1+2 dx(1 − x) log − 2 log(x)
4π 0 m2
| {z  i }
µ2
h 
1
log +3
2 m2
2
   
α 2 µ
= Q ∆ + log +2 . (84)
4π m2
 
However, when we try do the same for 2m2 ∂
∂p2
ΣψS + ΣψV p2 =m2 , this fails, since we obtain

2 ∂  ψ 
ψ 
2m ΣS + ΣV p2 =m2 (85)
∂p2
 Z 1
µ2
 
α 2 2 ∂

= Q 2m −2 dx(1 + x) log  (86)
4π ∂p2 0 x2 p2 + x(m2 − p2 ) p2 =m2
 Z 1
µ2
 
2 ∂

∼ 2m −2 dx log  (87)
∂p2 0 x2 p2 + x(m2 − p2 ) p2 =m2
Z 1  
2 ∂ 2 2 2 2

= 4m dx log(x p + x(m − p ))  (88)
0 ∂p2 2 2
p =m
Z 1 2 1
− −
Z
x x 1 x
= 4m2 dx 2 2 = −4 dx (89)
0 xm 0 x

12
which is divergent at x → 0 and indicates an IR-divergence. We also see that apart from the
IR-divergence, the expression is UV-finite. The IR-divergence can be regularized by taking
ε < 0 and denoting it as εIR in the end of the computation.
Therefore, we now want to compute  the
 above quantity in a proper way, by considering
2 ∂ ψ ψ 
the full expression for 2m ∂p2 ΣS + ΣV p2 =m2 and expanding it for small ε with ε < 0
only after we take the derivative.

2 ∂

ψ ψ 
2m ΣS + Σ V 
∂p2 p2 =m2
 Z 1
2 α 2 2 ε ∂ Γ(ε)
= 4m Q (4πµ ) 2
(1 − ε) dx(1 − x) 2 2
4π ∂p (x p + x(m2 − p2 ))ε
Z 1 0
D Γ(ε) 
− dx 2 2 
2 0 (x p + x(m2 − p2 ))ε p2 =m2
 Z 1
2 α 2 2 ε −εΓ(ε)
= 4m Q (4πµ ) (1 − ε) dx(1 − x)(x2 − x) 2 2 1+ε
4π (x m )
Z 1 0
D −εΓ(ε)
− dx(x2 − x) 2 2 1+ε
2 0 (x m )
Z 1
4πµ2
   
α 2
= 4 Q 1 + ε log (1 − εγE ) (1 − ε) dx(1 − x)2 x−1−2ε
4π m2 0
ε 1
 Z 
−1−2ε
−2 1− dx(1 − x)x
2 0
4πµ2
          
α 2 1 1 3
= Q 2 1 + ε log − γE − +2 + + 2 + O(ε)
4π m2 ε ε 2
4πµ2
   
α 2 2
= Q − 2γE + 2 log +2 , (90)
ε→εIR 4π εIR m2
where we used that
Z 1  
−1−2ε 1 1 1 1
dx(1 − x)x =− − =− + 2 + O(ε), (91)
0 2ε 1 − 2ε 2 ε
Z 1  
2 −1−2ε 1 2 1 1 1
dx(1 − x) x =− − + =− + 3 + O(ε). (92)
0 2ε 1 − 2ε 2 − 2ε 2 ε
Our final result for δψ is thus
4πµ2
   
α 1 2
δψ = − Q2 + − 3γE + 3 log +4 . (93)
4π ε εIR m2

e)
In the M S scheme we found that
1 MS α 21
δ
2 A
+ δeM S + δψM S + O(α2 ) = − Q = δψM S . (94)
4π ε
Hence,
δeM S = − 12 δAM S (95)

13
The same trivially holds also in the M S scheme.
In the on-shell scheme we found that
 
1 OS OS OS

2 ∂  ∂

δ + δe + δψ γµ + O(α ) = − µ Σ(p, m)
2 A
= − µ p/ΣV + mΣS 
∂p p2 =m2 ∂p p2 =m2
 
∂ ∂ 
= − γµ ΣV + p/ µ ΣV + m µ ΣS 
∂p ∂p p2 =m2
 
∂ ∂ 
= − γµ ΣV + p/2pµ 2 ΣV + 2mpµ 2 ΣS  , (96)
∂p ∂p p2 =m2

where we applied the chain rule

∂ ∂p2 ∂ ∂
µ
f (p) = µ 2
f (p) = 2pµ 2 f (p). (97)
∂p ∂p ∂p ∂p
Contracting our expression with pµ and then left multiplying it by p/−1 we obtain
 
2 ∂

1 OS OS OS ψ ψ ψ OS


δ + δe + δψ = − ΣV + 2m
2 A
ΣS + Σ V  2 2 = δψ ,
 (98)
∂p2 p =m

where the second equality follows from Eq. (82). Hence, we again verify that
δeOS = − 21 δAOS . (99)

f)
The corresponding Feynman diagram is

k
q q
µ ν (100)
q+k
The photon self-energy is given by


dD k Tr γν (k/ + /q + m)γµ (k/ + m)
Z
2 2ε
−iΠµν (q) = −(Qe) µ . (101)
(2π)D ((k + q)2 − m2 + iη)(k 2 − m2 + iη)
The trace in the numerator evaluates to
Tr γν (k/ + /q + m)γµ (k/ + m) = Tr γν (k/ + /q)γµ k/ + γν γµ m2

= (k + q)ρ k σ 4(gνρ gµσ − gνµ gρσ + gνσ gρµ + 4m2 gνµ )

= 4[2kµ kν + kµ qν + qµ kν − gµν (k 2 + k · q − m2 )]
= 4[2kµ kν + kµ qν + qµ kν − 12 gµν ((k 2 − m2 ) + ((k + q)2 − m2 ) − q 2 )] (102)
where we used that
Tr(γ µ γ ν ) = 4g µν , (103)
Tr(γ µ γ ν γ ρ γ σ ) = 4(g µν g ρσ − g µρ g νσ + g µσ g νρ ), (104)
Tr(any odd # of γ 0 s) = 0 (105)

14
and

q · k = 12 [((k + q)2 − m2 ) − (k 2 − m2 ) − q 2 )]. (106)

Written in terms of the given tensor and scalar integrals, the self-energy becomes
Q2 α
Πµν (q) = [2Bµν (q; m, m) + qµ Bν (q; m, m) + qν Bµ (q; m, m)
π
+ 21 gµν q 2 B0 (q; m, m) − gµν A0 (m) ,

(107)

where we used the shifting property of loop integrals

dD k dD k
Z Z
1 1
D 2 2
= . (108)
(2π) ((k + q) − m + iη) (2π) (k − m2 + iη)
D 2

Now we want to determine the coefficients for the given expansions of the tensor integrals

B µ = q µ B1 , (109)
B µν = g µν B00 + q µ q ν B11 . (110)

To determine B1 we simply multiply the first equation by qµ , such that

qµ µ
B1 = B (111)
q2
We can deal with Eq. (110) in the same manner (i.e. multiply by gµν and qµ qν and
 then solve 
for B00 and B11 ), but a smarter way to do this is to introduce the projectors gµν − qµq2qν
 
qµ qν
and −gµν + D q2 with the following trivial properties
   
qµ q ν µν qµ qν
gµν − 2 g = D − 1, −gµν + D 2 g µν = 0, (112)
q q
  µ ν   µ ν
q µ qν q q qµ q ν q q
gµν − 2 2
= 0, −gµν + D 2 =D−1 (113)
q q q q2

If we multiply Eq. (110) by each of these two projectors, we immediately obtain

 
1 q µ qν
B00 = gµν − 2 B µν , (114)
D−1 q
 
1 qµ qν
B11 = 2
−gµν + D 2 B µν . (115)
(D − 1)q q
Let us now determine the coefficients explicitly in terms of the scalar integrals A0 and B0

qµ µ (4π)2 1 2ε dD k
Z
2kq
B1 = 2 B = µ
q i 2q 2 (2π)D ((k + q)2 − m2 + iη)(k 2 − m2 + iη)
(4π)2 1 2ε dD k ((k + q)2 − m2 ) − (k 2 − m2 ) − q 2
Z
= µ
i 2q 2 (2π)D ((k + q)2 − m2 + iη)(k 2 − m2 + iη)
1 1
= 2 A0 (m) − A0 (m) − q 2 B0 (q; m, m) = − B0 .


(116)
2q 2

15
 (4π)2 1
Z
dD k k 2 − q12 (q · k)2
B00 = µ2ε
i D−1 (2π)D ((k + q)2 − m2 + iη)(k 2 − m2 + iη)
2 2 2 (q·k) 2 2 2 2 2
(4π)2 1 dD k k − m + m − 2q2 [((k + q) − m ) − (k − m ) − q )]
Z
2ε
= µ
i D−1 (2π)D ((k + q)2 − m2 + iη)(k 2 − m2 + iη)

1
= A0 (m) + m2 B0 (q; m, m)
D−1
(4π)2 1 dD k dD k
 Z 
q·k q·k
Z
2ε 2ε
− µ −µ
i 2q 2 (2π)D k 2 − m2 + iη (2π)D (k + q)2 − m2 + iη
| {z }
1
A (m)
2 0

(4π)2 1 2ε dD k

q·k
Z
+ µ
i 2 (2π)D ((k + q)2 − m2 + iη)(k 2 − m2 + iη)
| {z }
q2
2
B1 (q;m,m)

1
A0 (m) + 2m2 B0 (q; m, m) + q 2 B1 (q; m, m)
 
=
2(D − 1)
q2
   
1 2
= A0 (m) + 2m − B0 (q; m, m) . (117)
2(D − 1) 2

Here we used that

dD k q·k
Z
=0 (by antisymmetry under k → −k), (118)
(2π) k − m2 + iη
D 2

dD k q·k dD k q · k − q 2 dD k
Z Z Z
2 1
D 2 2
= D 2 2
= −q .
(2π) (k + q) − m + iη k→k−q (2π) k − m + iη (2π) k − m2 + iη
D 2

(119)

The computation for B11 is analogous to the one for B00

1 (4π)2 2ε
Z
dD k −k 2 − D q12 (q · k)2
B11 = µ
(D − 1)q 2 i (2π)D ((k + q)2 − m2 + iη)(k 2 − m2 + iη)
q2
  
1 2

1
= − A0 (m) + m B0 (q; m, m) + D A0 (m) − B1 (q; m, m)
(D − 1)q 2 2 2
1
(D − 2)A0 (m) − 2m2 B0 (q; m, m) − Dq 2 B1 (q; m, m)
 
= 2
2(D − 1)q
q2
   
1 2
= (D − 2)A0 (m) + −2m + D B0 (q; m, m) . (120)
2(D − 1)q 2 2

Putting everything together we get

qµ 0
Bµ = − B , (121)
2
q2
   
1 2
Bµν = gµν A0 + 2m − B0
2(D − 1) 2
q2
   
qµ qν 1 2
+ 2 (D − 2)A0 + −2m + D B0 . (122)
q 2(D − 1) 2

16
At last we are able to express Πµν in terms of the scalar integrals A0 and B0

Q2 α q2
 
Πµν = 2Bµν + qµ Bν + qν Bµ + gµν B0 − gµν A0
π 2
"  2 q2   2
q 2

Q2 α 2m − 2 qµ qν (D − 2)A0 + D 2 − 2m
= gµν B0 + 2 B0
π D−1 q D−1
#
q2 A0
− qµ qν B0 + gµν B0 − gµν A0 + gµν
2 D−1
Q2 α qµ qν (2 − D)(2A0 − q 2 B0 ) + 4m2 B0
 
= gµν − 2
π q 2(D − 1)
2 2


≡ q gµν − qµ qν Π(q ), (123)

where

2 Q2 α 1 (2 − D)(2A0 − q 2 B0 ) + 4m2 B0
Π(q ) = . (124)
π q2 2(D − 1)

The expansions in small ε for A0 and B0 are already known, since those integrals are
proportional to I1,0 and I2,0

(4π)2 Γ(−1 + ε)
I1,0 E = 0, F = −m2 = −(4πµ2 )ε


A0 =
i (m2 )−1+ε
  2 
µ
= m2 ∆ + log + 1 + O(ε). (125)
m2
Z 1
(4π)2 2 2

2 ε Γ(ε)
B0 = I2,0 Eµ = xqµ , F = xq − m = (4πµ ) dx
i 0 (x(x − 1)q 2 + m2 )ε
Z 1
µ2
 
=∆+ dx log + O(ε). (126)
0 x(x − 1)q 2 + m2

f ) (i)
In the mininmal subtraction schemes we have

ΠR,µν (q 2 ) = Πµν (q 2 ) + (q 2 gµν − qµ qν )(ZA − 1) = (q 2 gµν − qµ qν ) Π(q 2 ) + δA ,

(127)

where we evalutate Π(q)2 explicitly using our results from f)

17
 Q2 α 1 (2 − D)(2A0 − q 2 B0 ) + 4m2 B0
Π(q 2 ) =
π q2 2(D − 1)
2
 
Qα1 1 2
h  2
µ
i
= −2(1 − ε) 2m ∆ + log 2 + 1
π q 2 2 · 3(1 − 23 ε) m
 Z 1     Z 1  
2 µ2 2 µ2
−q ∆+ dx log x(x−1)q2 +m2 + 4m ∆ + dx log x(x−1)q2 +m2
0 0
Z 1
Q2 α 1
   2   
m2 2 µ 1 µ2
−2 q2 ∆ + 3 + log m2 +∆− 3 + dx log x(x−1)q 2 +m2
π 3
 Z 1    0
2 µ 2
+ 2 mq2 ∆ + 32 + dx log x(x−1)q 2 +m2
0
 Z 1 Z 1 
α 4 2 1

µ2

m2

m2
= Q ∆− 3 + dx log x(x−1)q2 +m2 + 2 q2 dx log x(x−1)q2 +m2 .
4π 3 0 0
(128)

Hence,
α 4 2
δAM S = − Q ∆, (129)
4π 3
1 1
with ∆ = ε
in the M S scheme and ∆ = ε
− γE + log(4π) in the M S scheme.

f ) (ii)
Let us evaluate the on-shell condition using L’Hospital’s rule
   
1 ν 1 2 2 ν
lim ΠR,µν (q)ε (q) = lim (q gµν − qµ qν )ΠR (q )ε (q)
q 2 →0 q2 q 2 →0 q2
 2
q ΠR (q 2 )

= lim εµ (q)
q 2 →0 q2
l’Hosp. ∂
q 2 ΠR (q 2 ) εµ (q),


= lim 2
(130)
q →0 ∂q
2

where we also used the equation of motion qµ εµ (q) = 0. Now from ΠR (q 2 ) = Π(q 2 ) + δA we
can see that
 
∂ 2 2

OS
lim q Π(q ) + δA εµ (q) = 0 (131)
q 2 →0 ∂q 2
∂

⇔ δAOS = − 2 q 2 Π(q 2 ) q2 =0 . (132)
∂q

18
Using Eq. (124) for Π(q 2 ) we can compute δAOS explicitly
∂

δAOS = − q 2
Π(q 2
) 2
q =0
∂q 2
∂ Q α (2 − D)(2A0 − q 2 B0 ) + 4m2 B0 
2
=− 2 q 2 =0
∂q π 2(D − 1)
h   i  
2 ∂ 2 2 ∂ 2
Q2 α (2 − D) B0 (0; m, m) + q ∂q2 B0 (q ; m, m) q2 =0 − 4m ∂q2 B0 (q ; m, m) q2 =0
 
=
π 2(D − 1)
2 0
α 4 2 (1 − ε)B0 (0; m, m) + 2m B0 (0; m, m)
=− Q
4π 3 1 − 23 ε
 
α 4 2 1
  2 
µ 2 1
=− Q (1 − 3 ε) ∆ + log m2 + 2m + O(ε)
4π 3 6m2
α 4 2h  2 i
µ
=− Q ∆ + log m 2 . (133)
4π 3
In this computation we used that
 
2 ∂ 2

q B0 (q ; m, m) 2 =0
q =0
(134)
∂q 2
and
Z 1 
∂ Γ(ε)
B00 (0; m, m) 2 ε

= 2 (4πµ ) dx 2 2 ε

∂q 0 (x(x − 1)q + m ) q2 =0
ε Z 1
4πµ2

Γ(ε)
= −ε 2
dx x(x − 1) 2
m m
 0
2 ε
1 4πµ 1
= Γ(ε) ε = + O(ε). (135)
6m2 m2 6m2
| {z }
1+O(ε)

f ) (iii)
By plugging in Eq. (75) into the relation for the unrenolmalized mass we obtain
 α 2  MS  α 2   2 
m(0) = 1 − Q 3∆ m (µ) = 1 − Q 3 ∆ + 34 + log (mµOS )2 mOS (136)
4π 4π
Therefore, at O(α) we have
 α 2  α 2   2 
mM S (µ) = 1 + Q 3∆ 1 − Q 3 ∆ + 34 + log (mµOS )2 mOS
4π 4π
 α 2 4  2 
= 1− Q 3 3 + log (mµOS )2 mOS
4π 
 α 2  2 
= 1− Q 4 + 3 log (mµOS )2 mOS . (137)
4π
For the electric charge we have
    2 
α 2 2 α 2 2
(0)
e = 1+ MS
Q ∆ e (µ) = 1 + Q ∆ + log (mµOS )2 eOS , (138)
4π 3 4π 3

19
which leads to
  2 
α 2 2
eMS OS
=e 1+ Q log (mµOS )2 . (139)
4π 3

Exercise groups: time and locations

Group Tutor Time Room

1 Jorge Segovia Tuesday, January 12th and 19th, 10-12 PH 3344
2 Javad Komijani Wednesday, January 13th and 20th, 12-14 PH 3343

20