Quantum Electrodynamics

Feynman rules for QED

Electrodynamics essentially describes interaction between the electrically charged particles in terms
of electromagnetic field. Classical electrodynamics is given by Maxwell’s equations and the corre-
sponding Lagrangian density is,
1
L = − F µν Fµν − j µ Aµ , where, Fµν = ∂µ Aν − ∂ν Aµ and ∂µ F µν = j ν . (1)
4
In quantum theory of electrodynamics, the interaction of gauge field with charged matter fields
(fermions or complex scalar) is introduced by demanding gauge invariance of the theory. The
starting equation for quantum theory of spin-1/2 fermions is the Dirac equation,

L = ψ (i∂/ − m) ψ ⇒ (i∂/ − m)ψ = 0, where ∂/ ≡ γ µ ∂µ . (2)

The un-normalized plane wave solutions of Dirac equation are,

us (p) e−ip·x

positive energy
ψ(x) = s = 1, 2 (3)
vs (p) eip·x negative energy
0
   
1
 0   1 
where, u1 = 
 E+m pz  and u2 =  px −ipy 
  
E+m
px +ipy −pz
E+m E+m
 pz   px −ipy 
E+m E+m
 px +ipy   −pz 
v1 =  E+m  and v2 =  E+m  (4)
 1   0 
0 1

Here, us and vs are 4-component Dirac spinors satisfying,

/ − m)us (p) = 0
(p and (p/ + m)vs (p) = 0 (5)

The adjoint spinors are defined as,


us (p) (p/ − m) = 0
where ψ ≡ ψ † γ 0 (6)
v s (p) (p
/ + m) = 0

The normalization of the spinors is achieved with the desired 2E particle per unit volume,

u†r (p) us (p) = vr† (p) vs (p) = 2E δrs or ur (p) us (p) = v r (p) vs (p) = 2m δrs (7)
X X
us (p) us (p) = p / + m and vs (p) v s (p) = p/ − m (8)
s s

Fourier decomposition of the Dirac fields in terms of creation and annihilation operators are,
X 1 Xh i
ψ(x) = p fs (p) us (p) e−ipx + fˆs† (p) vs (p) eipx ≡ ψ+ + ψ− (9)
p
2Ep V s
X 1 Xh i
ψ(x) = p fs† (p) us (p) eipx + fˆs (p) v s (p) e−ipx ≡ ψ − + ψ + (10)
p
2Ep V s
eipx eipx
ψ − |0i ≡ p us (p)|0i = |ls (p)i, ψ− |0i ≡ p vs (p)|0i = |¯ls (p)i
2Ep V 2Ep V
e−ipx e−ipx
ψ+ |ls (p)i ≡ p us (p)|ls (p)i = |0i, ψ + |¯ls (p)i ≡ p v s (p)|¯ls (p)i = |0i,(11)
@Ep V 2Ep V

1
where l, ¯l are the representative of fermionic particle and anti-particle. The gauge field i.e. electro-
magnetic field is included in the theory of fermion by demanding U(1) local gauge invariance of the
Dirac Lagrangian (2). The Dirac Lagrangian is expected to be invariant under the transformation,
ψ(x) → ψ 0 (x) = eie α(x) ψ(x) (12)
which is achieved by minimal substitution of ∂µ in (2) by Dµ ,
L = ψ (iD/ − m) ψ where Dµ = ∂µ + ie Aµ (x) (13)
and Aµ (x) is the gauge field transforming as,
Aµ (x) → A0µ (x) = Aµ (x) + ∂µ α(x). (14)
Therefore, the full Dirac Lagrangian density describing fermion-gauge field interaction is given by,
1
L = ψ (iγ µ ∂µ − m) ψ − eψγ µ ψ Aµ − F µν Fµν . (15)
4
The fermion-gauge field interaction term is thus ψψAµ , which is also the basic QED interaction
vertex. The gauge field Aµ (x) is a 4-vector but not all the components are independent, only two
are. The two Maxwell’s equations ∇ · B = 0 and ∇ × E = −∂B/∂t when expressed in terms of vector
potential Aµ will constrain two of the four components. It is customary to use the two transverse
polarization states as the relevant degrees of freedom. Maxwell equation (1) when written in terms
of Aµ yields,
∂µ (∂ µ Aν − ∂ ν Aµ ) = j ν ⇒ ∂µ ∂ µ Aν − ∂ ν ∂µ Aµ = j ν . (16)
Since gauge transformation keeps physics invariant, we are at liberty to impose certain condition on
Aµ to simplify (16), say,
∂µ Aµ = 0 ⇒ ∂µ ∂ µ Aν = j ν (17)
by choosing α(x) i.e. ∂µ ∂ µ α(x) in (14) appropriately. The condition ∂µ Aµ = 0 is called Lorentz
gauge, which in momentum space reads kµ Aµ (k) = 0 and constrained one of the components of
Aµ (k). Additionally, we can further choose any α(x) that satisfies ∂µ ∂ µ α(x) = 0 without violating
Lorentz gauge ∂µ Aµ = 0: choose (i) Temporal gauge A0 = 0, (ii) axial gauge A3 = 0 etc. The
temporal gauge in the momentum space leads to constraining another component of Aµ , i.e. A0 (k) =
0. Temporal gauge implies A · k = 0 and assuming the direction of momentum of EM field is along
z-axis, the longitudinal component gets fixed as A3 = 0, leaving only the transverse components
A1,2 to be independent. The Fourier decomposition of gauge field Aµ (x), therefore, is
1 Xh i
aλ (k) λµ (k) e−ikx + a†λ λ?
X
ikx
Aµ (x) = √ µ (k) e (18)
k
2ωk V λ

where λµ are the polarization vectors. Since only the transverse polarization states matter, λ = 1, 2,
   
0 0
1
 1  2
 0 
µ = 
 0  and µ =  1 
   : plane polarization (19)
0 0
The polarization state 0µ is known as scalar polarization and is zero because of temporal gauge.
Similarly, the 3µ is known as longitudinal polarization. It is possible to have other choices for
transverse polarization vectors,
   
0 0
1   1  1   1 
L
 
circular polarization : µ = √  i  and √  −i  (20)
2 2
0 0
   
0 0
 cos θ   sin θ 
elliptical polarization : L
µ =  i sin θ  and  −iθ 
    (21)
0 0

2
The corresponding orthonormal and completeness conditions are,

orthonomral : rµ (k) sµ? (k) = −δrs r, s = 1, 2 (22)

X
completeness : rµ (k) rν (k) = −gµν . (23)
r=1,2

With is brief introductory remarks on the QED Lagrangian, which is the gauge invariant Dirac
Lagrangian (15), the Feynman rules for QED are listed below,

The QED processes are all in 2nd order and above of the S-matrix since it is impossible to have a
lone ψψAµ vertex with Aµ being massless k 2 = 0. The essential QED processes are,

1. Møller scattering: e− e− → e− e−
2. Bhaba scattering: e+ e− → e+ e−
3. Compton scattering; γ e ± → γ + e±
4. Electron-muon process: e+ e− → µ+ µ−

5. Pair production / annihilation: e+ e− ↔ γ γ

The first three are elastic scattering processes while the last two are inelastic. There is another
scattering process e X → e X, which is called Mott scattering when mX  me which reduces to
Rutherford scattering in the non-relativistic limit v  c. Calculating QED scattering cross-sections
we need to calculate |Mf i |2 as usual and for un-polarized cross-sections, i.e. to say not caring about
the initial and final spin states, we must
• average over initial spins: ⇒ Πi 2si1+1 |Mf i |2

• sum over all spins: ⇒ Πi 2si1+1 si ,sf |Mf i |2 .

P

The above QED processes, apart from Compton and Mott / Rutherford scattering, are based only
on a couple of different vertices corresponding to s and t channels. Since the Feynman rules for them

3
are standard, we list them below and will be used without worrying about creation / annihilation
operation any further.

The invariant amplitude of the following QED processes are worked out first before a detailed
calculation of differential scattering cross-section dσ/dΩ for them. An alternate notation in terms
of vector currents ψγ µ ψ is also used to express the invariant amplitudes. For instance in the first
case, we can write
jeµ ≡ [u(p3 ) γ µ u(p1 )]e and jlν ≡ [u(p4 ) γ ν u(p3 )]l
j µ gµν j ν je · jl
iMf i = ie2 e 2 l = ie2 (24)
k k2
Møller scattering:

(t)  −igµν
us01 (p01 ) (−ieγ µ ) us1 (p1 ) us02 (p02 ) (−ieγ ν ) us2 (p2 )
  
iMf i = (25)
k 2 + i
(u) −igµν
us02 (p02 ) (−ieγ µ ) us1 (p1 ) us01 (p01 ) (−ieγ ν ) us2 (p2 )
   
iMf i = (26)
k 02 + i
(t) (u)
iMf i = iMf i − iMf i (27)
Since the two diagrams corresponding to t and u channels are related by an exchange of final particles,
the corresponding Mf i are added with a relative minus sign between the amplitudes as this involves
exchange of fermions.
Bhaba scattering:

4
 (s) −igµν 
[v s2 (p2 ) (−ieγ µ ) us1 (p1 )] us01 (p01 ) (−ieγ ν ) vs02 (p02 )

iMf i = (28)
k 2 + i
(t)  −igµν 
= us01 (p01 ) (−ieγ µ ) us1 (p1 ) v s2 (p2 ) (−ieγ ν ) vs02 (p02 )
 
iMf i 02
(29)
k + i
(s) (t)
iMf i = iMf i − iMf i (30)

Compton scattering:

(s)
h 0
ii(q/ + m)  r
us0 (p0 ) (−ieγ ν ) νr ? (k 0 ) µ (k) (−ieγ µ ) us (p)

iMf i = (31)
q2 2
− m + i
0
(u) i(q
/ + m) h r0 ? 0 i
us0 (p0 ) (−ieγ µ ) rµ (k) 02 ν
 
iMf i =  µ (k ) (−ieγ ) us (p) (32)
q − m2 + i
(s) (u)
iMf i = iMf i + iMf i (33)

e+ e− → µ+ µ− :

−igµν
iMf i = [v s2 (p2 ) (−ieγ µ ) us1 (p1 )] [ur1 (q1 ) (−ieγ ν ) vr2 (q2 )] (34)
k 2 + i

QED scattering cross-section: Møller scattering

The invariant amplitude for Møller scattering is (27)
(t) (u)
iMf i = iMf i − iMf i
[us01 (p01 )γ µ us1 (p1 )] [us02 (p02 )γµ us2 (p2 )] [us02 (p02 )γ µ us1 (p1 )] [us01 (p01 )γµ us2 (p2 )]
 
2
= ie −
(p1 − p01 )2 (p1 − p02 )2
1 X
⇒ |M |2 = |Mf i |2
4 0 0
s1 ,s2
s1 ,s2
†
e2 X (u01 γ µ u1 ) (u02 γµ u2 ) (u02 γ µ u1 ) (u01 γµ u2 )

= − ×
4 0 0 t u
s1 ,s2
s1 ,s2

(u01 γ ν u1 ) (u02 γν u2 ) (u02 γ ν u1 ) (u01 γν u2 )

 
−
t u
2
 
e T11 T22 T12 + T21
≡ + 2 − . (35)
4 t2 u tu

5
In the above expression, the Mandelstam variable t and u are used to replace (p1 −p01 )2 and (p1 −p02 )2
respectively. We will calculate each term separately, explicit calculation of T11 is shown and the
remaining terms follow similarly,
†
X
T11 = [(u01 γ µ u1 ) (u02 γµ u2 )] [(u01 γ ν u1 ) (u02 γν u2 )]
s01 ,s02
s1 ,s2
X
= [(u2 γµ u02 ) (u1 γ µ u01 )] [(u01 γ ν u1 ) (u02 γν u2 )]
s01 ,s02
s1 ,s2
Xh i
αβ 0β a µab 0b
(uα u1 ) (u0λ u1 ) (u0m
νλσ σ mn n

= 2 γµ u2 ) (u1 γ 1 γ 2 γν u2 )
s01 ,s02
s1 ,s2

u0β
X X X X
= uσ1 ua1 γ µab u0b 0λ νλσ
1 u1 γ un2 uα αβ
2 γµ
0m mn
2 u2 γν
s1 s01 s2 s02
h ih i
σa bλ nα βm
= (p/1+ m) γ µab (p/01+ m) γ νλσ (p/2+ m) γµαβ (p/02+ m) γνmn
or, T11 = Tr [(p/1+ m) γ µ (p/01+ m) γ ν ] Tr [(p/2+ m) γµ (p/02+ m) γν ] (36)
T22 = Tr [(p/1+ m) γ µ (p/02+ m) γ ν ] Tr [(p/2+ m) γµ (p/01+ m) γν ] (37)
T12 = Tr [(p/1+ m) γ (p/01+ m) γν
µ
(p/2+ m) γµ (p/02+ m) γ ν ] (38)
T21 = Tr [(p/1+ m) γµ (p/01+ m) γ ν (p/2+ m) γ µ (p/02+ m) γν ] (39)

To evaluate these traces we recall some basic γ-matrix algebra,

{γµ , γν } = 2 gµν and{γm u, γ5 } = 0

µ
γµ γ = 4
γν γµ γ ν = −2γµ
γλ γµ γν γ λ = 4 gµν
γρ γµ γν γλ γ ρ = −2 γλ γν γµ (40)

The traces of series of γ-matrices can be determined from the following sets of identities together
with the above algebra,

Tr (odd number of γ) = 0
Tr (γµ γν ) = 4 gµν
Tr (γµ γν γρ γσ ) = 4 (gµν gρσ − gµρ gνσ + gµσ gνρ ) (41)

Using the above γ-matrix algebra and trace formulas, we evaluate the following,

Tr [(p/1+ m) γ µ (p/01+ m) γ ν ]
= Tr (γ λ p1λ + m) γ µ (γ σ p01σ + m) γ ν
 

= Tr p1λ p01σ γ λ γ µ γ σ γ ν + m p1λ γ λ γ µ γ ν + m p01σ γ µ γ σ γ ν + m2 γ µ γ ν

 

= 4 p1λ p01σ g λµ g σν − g λσ g µν + g λν g µσ + 4m2 g µν

 

= 4 pµ1 p0ν 0 µν
+ pν1 p0µ 2 µν
 
1 − p1 · p1 g 1 + 4m g
= 4 p1 p1 + pν1 p0µ
 µ 0ν µν 2 0

1 + g (m − p1 · p1 ) . (42)

Using this expression and a similar such for the second term of T11 we obtain,
ν 0µ
T11 = 4 pµ1 p0ν µν 2 0 0 0 2 0
   
1 + p1 p1 + g (m − p1 · p1 ) × 4 p2µ p2ν + p2ν p2µ + gµν (m − p2 · p2 )

= 16 (p1 · p2 )(p01 · p02 ) + (p1 · p02 )(p01 · p2 ) + p1 · p01 (m2 − p2 · p02 )+



(p1 · p02 )(p01 · p2 ) + (p1 · p2 )(p01 · p02 ) + p1 · p01 (m2 − p2 · p02 ) +

2 p2 · p02 (m2 − p1 · p01 ) + 4 (m2 − p1 · p01 )(m2 − p2 · p02 ) .

(43)

6
The 4-momentum conservation implies,
p1 + p2 = p01 + p02 → p1 · p2 = p01 · p02
⇒ p1 − p01 = p02 − p2 → p1 · p01 = p2 · p02
⇒ p1 − p02 = p01 − p2 → p1 · p02 = p01 · p2 (44)
When these results are substituted in (43), we get
T11 = 16 2(p1 · p2 )2 + 2(p1 · p02 )2 + 4p1 · p01 (m2 − p1 · p01 ) + 4(m2 − p1 · p01 )2
 

= 32 (p1 · p2 )2 + (p1 · p02 )2 + 2m2 (m2 − p1 · p01 )

 
(45)
T22 = 32 (p1 · p2 )2 + (p1 · p01 )2 + 2m2 (m2 − p1 · p02 )
 
(46)
T12 + T21 = 64 −(p1 · p2 )2 + 2m2 (p1 · p2 )
 
(47)
where T22 is obtained by the substitution p01
↔ p02
in (45). Derivation of the sum T12 + T21 is a
lengthy but straight forward calculation involving only γ-matrix algebra. Substituting every thing
back in (35), the invariant amplitude for Møller scattering is,
e4 32 

|M |2 = (p1 · p2 )2 + (p1 · p02 )2 + 2m2 (m2 − p1 · p01 ) +
4 t2
32 
(p1 · p2 )2 + (p1 · p01 )2 + 2m2 (m2 − p1 · p02 ) +
u2 
64 
(p1 · p2 )2 − 2m2 (p1 · p2 ) . (48)
tu
We will calculate the differential scattering cross-section in CM frame, p~1 + p~2 = 0 = p~01 + p~02
where each particle has the same mass m and consequently same energy E. This also implies
|~
p1 | = |~ p01 | = |~
p2 | = |~ p02 | = p. Therefore, we can write,
p1 ), p01 = (E, p~01 ), p02 = (E, −~
p1 = (E, p~1 ), p2 = (E, −~ p01 )
p1 · p2 = E 2 + p2 , p1 · p01 = E 2 − p2 cos θ, p1 · p02 = E 2 + p2 cos θ
s = (p1 + p2 )2 = 4E 2
t = (p1 − p01 )2 = 2m2 − 2(E 2 − p2 cos θ) = −2p2 (1 − cos θ)
u = (p1 − p02 )2 = 2m2 − 2(E 2 + p2 cos θ) = −2p2 (1 + cos θ)
tu = 4p4 sin2 θ (49)
Substituting the above results in |M |2 yields,
 2
(E + p2 )2 + (E 2 + p2 cos θ)2 + 2m2 (m2 − E 2 + p2 cos θ)
|M |2 = 8e4 +
4p4 (1 − cos θ)2
(E 2 + p2 )2 + (E 2 − p2 cos θ)2 + 2m2 (m2 − E 2 − p2 cos θ)
+
4p4 (1 + cos θ)2
(E 2 + p2 )2 − 2m2 (E 2 + p2 )

2
4p4 sin2 θ
2e4 (E 2 + p2 )2 + (E 2 + p2 cos θ)2 − 2m2 p2 (1 − cos θ)

= +
p4 (1 − cos θ)2
(E 2 + p2 )2 + (E 2 − p2 cos θ)2 − 2m2 p2 (1 + cos θ)
+
(1 + cos θ)2
(E 2 + p2 )2 − 2m2 (E 2 + p2 )

2 (50)
sin2 θ
This can be be inserted in the general formula for differential scattering cross-section in CM frame
that we have derived earlier,
dσ 1 p?f 1 2e4 α2
?
= 2 ? |M |2 = 2 2 4
[· · · ] = [· · · ] (51)
dΩ 64π s pi 256π E p 8E 2 p2

7
In the ultra-relativistic limit E  m, the differential scattering cross-section is,
" #
dσ α2 1 1
= + +1 . (52)
dΩ? 4E 2 sin4 θ2 cos4 θ2
"  2  2  2 #
α2 2 1 1 t t
= s + + + . (53)
2s t u u u

QED scattering cross-section: Bhaba scattering

The invariant amplitude for Bhaba scattering is,

e2 32 

2
|M | = (p1 · p2 )2 + (p1 · p02 )2 + 2m2 (m2 − p1 · p01 ) +
4 t2
32 
(p1 · p01 )2 + (p1 · p02 )2 + 2m2 (m2 − p1 · p2 ) +
u2 
64  0 2 2 0 0
(p1 · p2 ) + m [(p1 · p2 ) + (p1 · p2 ) − (p1 · p1 )] . (54)
tu

The derivation of the above and the subsequent results for Bhaba scattering are left as an exercise. In
the CM frame, various dot products and Mandelstam variables in terms of energy and 3-momentum
of the e± are the same as (49), using which yields
 2 2 2 2 2 2 2 2
2 4 (E + p ) + (E + p cos θ) − 2m p (1 − cos θ)
|M | = 2e +
p4 (1 − cos θ)2
E 4 + p4 cos2 θ + 2m2 E 2 (E 2 + p2 cos θ)2 + 2m2 (E 2 + p2 cos θ)

− (55)
2E 4 E 2 p2 (1 − cos θ)

In the ultra-relativistic limit E  m, differential cross-section for Bhaba scattering is,

" #
dσ α2 1 + cos4 θ2 1 + cos2 θ 2 cos4 θ2
= + − (56)
dΩ? 8E 2 sin4 θ2 2 sin2 θ2
"  2  2   #
α2 2 1 1 t s 2
= u + + + . (57)
2s s t s t

QED scattering cross-section: Compton scattering

The invariant amplitude for Compton scattering is (33),
(s) (u)
iMf i = iMf i + iMf i
γ µ (p/ − /k 0+ m)γ ν r
 ν
/ + k/ + m)γ µ

γ (p 0
= −ie2 us0 (p0 ) 2 2
+ 0 2 2
µ (k) rν ? (k 0 )us (p) (58)
(p + k) − m (p − k ) − m
(s) (u) (s)† (u) (s) (u)†
|Mf i |2 = |Mf i |2 + |Mf i |2 + Mf i Mf i + Mf i Mf i (59)

(s)
We begin by calculating |Mf i |2 , employing usual averaging over initial spins and polarization, sum

8
over all spins and polarization (using q = p + k),

(s) e4 X X [u0 γ ν (q/ + m)γ µ u]† [u0 γ α (q/ + m)γ β u] r? 0 0

|Mf i |2 = 2 − m 2 ]2
µ (k) rν (k 0 ) rα (k)rβ ? (k 0 )
4 0 0
[(p + k)
s,s r,r

e4 X
i 0 0 j
X
r r?
X 0 0
= [uΓ u ] [u Γ u]  α  µ rν rβ ?
4(s − m2 )2 0 r 0
s,s r

e4 X
gαµ gνβ ua Γiab u0b u0m Γjmn un

= 2 2
4(s − m ) 0 s,s
4
e
/ 0+ m)Γj (p/ + m)
Tr gαµ gνβ Γi (p
 
= 2
16(p · k)
e4 h i
= 2
/ 0+ m)Γ†i (p/ + m)
Tr Γi (p (60)
16(p · k)
where, Γi ≡ γ ν (q/ + m)γ µ , Γj ≡ γ α (q/ + m)γ β and gαµ gνβ Γj = Γ†i
(s − m2 )2 = (p2 + k 2 − 2p · k − m2 )2 = (m2 + 0 − 2p · k − m2 )2 = 4(p · k)2
The calculation of trace follows expanding the terms inside the square bracket in (60) and writing
only the terms having even number of γ-matrices,
h i
/ 0+ m)Γ†i (p
Tr Γi (p / + m) = Tr [{γ ν (q/ + m)γ µ (p / 0+ m)} {γµ (q/ + m)γν (p/ + m)}]
= Tr γ ν /q γ µ/p0+ mγ ν γ µ/p0+ mγ ν /q γ µ + m2 γ ν γ µ ×


γµ /q γν p/ + mγµ γν p/ + mγµ /q γν + m2 γµ γν
 

= Tr [γ ν /q γ µ/p0γµ /q γν p/] + m2 Tr [γ ν /q γ µ/p0γµ γν ] + m2 Tr [γ ν γ µ/p0γµ γν p/]

| {z } | {z } | {z }
I II III
ν µ 0
+ m Tr [γ γ /p γµ /q γν ] + m Tr [γ /q γ γµ γν p/] + m Tr [γ /q γ µ γµ /q γν ]
2 2 ν µ 2 ν
| {z } | {z } | {z }
IV V VI
+ m2 Tr [γ ν γ µ γµ /q γν p/] + m4 Tr [γ ν γ µ γµ γν ] (61)
| {z } | {z }
VII VIII

Calculation of individual traces are straightforward using the γ matrix algebra in (41),
I = 16[2(q · p)(q · p0 ) − q 2 (p · p0 )]
= 32[(p + k) · p][(p + k) · p0 ] − 16(p + k)2 (p · p0 )
= 32(p2 + k · p)(p · p0 + k · p0 ) − 16(p2 + k 2 + 2p · k)(p · p0 )
= 32(m2 + k · p)(p · p0 + k · p0 ) − 16(m2 + 2p · k)(p · p0 )
II = −32m2 q · p0 = −32m2 (p + k) · p0
III = 16m2 p · p0
IV = −32m2 q · p0 = −32m2 (p + k) · p0
V = −32m2 p · q = −32m2 p · (p + k) = −32m2 (p2 + p · k) − 32m2 (m2 + p · k)
VI = 64m2 q 2 = 64m2 (p + k)2 = 64m2 (p2 + k 2 + 2p · k) = 64m2 (m2 + 2p · k)
VII = −32m2 q · p = −32m2 p · (p + k) = −32m2 (p2 + p · k) − 32m2 (m2 + p · k)
VIII = 64m4 (62)
Here we have used p2 = m2 and k 2 = 0, the k being photon 4-momentum. Putting everything back,

(s) e4
|Mf i |2 32 2m4 + m2 (−p · p0 − k · p0 + 2k · p) + (k · p)(k · p0 )
 
= 2
16(p · k)
2e4  4
2m + m2 (−p · p0 − k · p0 + 2k · p) + (k · p)(k · p0 )

= 2
(63)
(p · k)

9
Similarly, we can write down the invariant amplitude for the u-channel Compton scattering by
replacing k with −k 0 ,
(u) 2e4  4
|Mf i |2 = 2m + m2 (−p · p0 + k 0 · p0 − 2k 0 · p) + (k 0 · p)(k 0 · p0 )

0 2
(64)
(p · k )
The cross terms in the calculation of |Mf i |2 is,

(s)† (u) (s) (u)† 2e4

2(p · p0 ) {(k · p) − (k · k 0 ) − (p · k 0 )} − m4

Mf i Mf i + Mf i Mf i =
(k · p)(k 0 · p)
+ m2 {−2(k · p) − (k · p0 ) + (k · k 0 ) − (p · p0 ) + 2(p · k 0 ) + (p0 · k 0 )} (65)

.
In order to simplify the above expression first we consider the 4-momentum conservation which gives,
 p · k = p0 · k 0

0 0
p+k =p +k = p · k 0 = p0 · k (66)
p · p0 = p · (p + k − k 0 )


Secondly, Compton scattering is studied in the rest frame of electron for obvious reason, which gives
p = (m, 0, 0, 0), k = (k ≡ |~k|, ~k), k 0 = (k0 ≡ |~k 0 |, ~k 0 )
p · k = mk, p · k 0 = mk0 ,
p · p~0 = p · (p + k − k 0 ) = m2 + mk − mk0 ,
k
k · k 0 = kk0 (1 − cos θ) = m(k − k0 ) ⇒ k0 = (67)
1 + k/m(1 − cos θ)
The last expression follows from squaring p0 = p + k − k 0 ,
p0 = p + k − k 0 ⇒ p02 = p2 + 2p · (k − k 0 ) − 2k · k 0 ⇒ k · k 0 = p · (k − k 0 ) = m(k − k0 ) (68)
Using these results for dot products in the above expressions for various |Mf i |2 yield,

(s) 2e4  2
|Mf i |2 m + mk + kk0

= 2
(69)
k
(u) 2e4  2
|Mf i |2 m − mk0 + kk0

= 02
(70)
k
(s) (u) 2e4  2
⇒ |Mf i |2 + |Mf i |2 (k + k02 )(m2 + kk0 ) − mkk0 (k − k0 )

= 2 02
k k
2e4  2
(k + k02 )(m2 + kk0 ) − k2 k02 (1 − cos θ)

= 2 02
(71)
k k
(s)† (u) (s) (u)† 2e4 
−2m2 kk0 − k2 k02 (1 − cos θ)

Mf i Mf i + Mf i Mf i = (72)
k2 k02
(s) (u) (s)† (u) (s) (u)†
⇒ |Mf i |2 = |Mf i |2 + |Mf i |2 + Mf i Mf i + Mf i Mf i
2e4  2
(k + k02 )(m2 + kk0 ) − 2k2 k02 (1 − cos θ) − 2m2 kk0

=
k2 k02
k0

k
= 2e4 0 + − sin2 θ . (73)
k k
We recall that the differential scattering cross-section in Lab frame is,
dσ 1 1
= |Mf i |2 for m1 = 0
dΩ 64π (m + E1 − E1 cos θ)2
2

dσ 1 1
⇒ = |Mf i |2
dΩ 64π 2 (m + k − k cos θ)2
e4 k0
 
1 k 2
= + − sin θ
32π 2 (m + m[k/k0 − 1)]2 k0 k
 0 2 
α2 k0

k k 2
= + − sin θ (74)
2m2 k k0 k

10
This is the Klein-Nishina formula for the differential cross-section of the Compton scattering. If we
substitute back the expression for k0 in the above formula, it gives,
2 
α2
 
dσ 1 1 2
= 1 + k/m(1 − cosθ) + − sin θ
dΩ 2m2 1 + k/m(1 − cosθ) 1 + k/m(1 − cosθ)
2
 
α 2k
(1 − cos θ) 1 + k/m(1 − cosθ) + 1 − k/m(1 − cosθ) − sin2 θ
 
≈ 2
1−
2m m
2
 
α 2 2k 2 2
= 1 + cos θ − (1 − cos θ)(1 + cos θ) + O(1/m )
2m2 m
2
α
= (1 + cos2 θ) as m → ∞. (75)
2m2
This is the Thompson scattering cross-section for classical electromagnetic radiation by free electron.

QED scattering cross-section: e+ e− → µ+ µ−

In Bhaba scattering the process is elastic in which the energy of the electron positron are not
too high. But if the energy of the pair is high enough E  m then the scattering need not be
elastic, in the final state particles of higher masses can appear. Here we study the high energy
e− (s1 , p1 ) e+ (s2 , p2 ) → µ− (r1 , q1 ) µ+ (r2 , q2 ), which serves as a prototype of e± → x± , which is at
the basic of e+ e− -collider. In this case, there is only one s-channel diagram and the corresponding
Feynman amplitude is given in (34),
−igµν
iMf i = [v s2 (p2 ) (−ieγ µ ) us1 (p1 )]e 2 [ur1 (q1 ) (−ieγ ν ) vr2 (q2 )]µ
k + i
[v 2 γ µ u1 ]e [u1 γµ v2 ]µ ie2
 
= ie2 2
≡ je · jµ (76)
(p1 + p2 ) s
† 
e4 X [v 2 γ µ u1 ]e [u1 γµ v2 ]µ [v 2 γ ν u1 ]e [u1 γν v2 ]µ
 
|M |2 =
4 s1 ,s2 s s
r1 ,r2

e4 X
= [{v 2 γµ u1 }µ {u1 γ µ v2 }e ] [{v 2 γ ν u1 }e {[u1 γν v2 }µ ]
4s2 s1 ,s2
r1 ,r2
4
e X h α αβ β i
= 2
{v 2 γµ u1 }µ {ua1 γ µab v2b }e {v m
2 γ u1 }e {uρ1 γνρσ v2σ }µ
νmn n
4s s1 ,s2
r1 ,r2
"( ) ( )#
e4 X X X X
= γ µab
v2b v m
2 γ νmn
un1 ua1 γµαβ uβ1 uρ1 γνρσ v2σ v α
2
4s2 s2 s1 r1 r2
e
4
e h µab i
γ (p/2− me )bm γ νmn (p/1+ me )na e γµαβ (q/1+ mµ )βρ γνρσ (q/2− mµ )σα

= 2 µ
4s
e4
= Tr [γ µ (p/2− me )γ ν (p/1+ me )] Tr [γµ (q/1+ mµ )γν (q/2− mµ )] (77)
4s2
These traces of γ-matrices are almost similar to (42), which gives us,
Tr [γ µ (p/2− me )γ ν (p/1+ me )] = 4 pµ1 pν2 + pν1 pµ2 − g µν (m2e + p1 · p2 )
 
(78)
Tr [γµ (q/1+ mµ )γν (q/2− mµ )] = 4 q1µ q2ν + q1ν q2µ − gµν (m2µ + q1 · q2 )
 
(79)
Since the energy involves here is much higher E  me , the electron mass is neglected subsequently
keeping, however, only the muon mass mµ . Putting back the traces in (77),
e4
|M |2 4 [pµ1 pν2 + pν1 pµ2 − g µν p1 · p2 )] · 4 q1µ q2ν + q1ν q2µ − gµν (m2µ + q1 · q2 )
 
= 2
4s
8e4 
(p1 · q1 ) (p2 · q2 ) + (p1 · q2 ) (p2 · q1 ) + m2µ (p1 · p2 )

= 2
(80)
s

11
In CM frame various kinematical identities can be used to express this in terms of measurable
quantities. In CM frame, if θ is the angle between incoming electron and outgoing muon i.e.
between p~1 and ~q1 , then we have

1. p21 = p22 = 0 ⇒ p1 = (E, p~), p2 = (E, −~

p) where E = |~
p| = p
2. q12 = q22 = m2µ ⇒ q1 = (E, ~q), q2 = (E, −~q) where E 2 − |~q|2 = E 2 − q 2 = m2µ
3. p1 + p2 = q1 + q2 ⇒ p1 · q1 = p2 · q2 and p1 · q2 = p2 · q1
2 2 2
4. s = (p1 + p2 ) = (q1 + q2 ) = 4E p1 · p2 = 2E 2 , s = 4E 2
⇒
1
5. p1 · q1 = E(E − q cos θ), p1 · q2 = E(E + q cos θ) and q 2 = (s − 4m2µ ) = E 2 − m2µ .(81)
4
The above set of identities can be used to simplify invariant amplitude in (80),

8e4 
|M |2 (p1 · q1 )2 + (p1 · q2 )2 + m2µ p1 · p2

=
s2
8e4  2
E (E − q cos θ)2 + E 2 (E + q cos θ)2 + 2m2µ E 2

= 4
16E
e4  4
2E + 2E 2 q 2 cos2 θ + 2m2µ E 2

=
2E4   
4 1 2 2 2 4 1 2 2 2 2
= e 1 + 2 (q cos θ + mµ ) = e 1 + 2 {(E − mµ ) cos θ + mµ }
E E
" #
2
mµ
= e4 1 + cos2 θ + 2 (1 − cos2 θ) (82)
E

Therefore, the differential and total scattering cross-section in CM frame is,

" #
dσ 1 p?f 2 e4 q 2
4m2µ 2
= |M | = 1 + cos θ + (1 − cos θ)
dΩ 64π 2 s p?i 64π 2 s p s
q
s − 4m2µ
" #
2 2
α 4m µ
= 1 + cos2 θ + (1 − cos2 θ)
4s 2E s
r " #
α2 4m2µ 2
4m2µ 2
= 1− 1 + cos θ + (1 − cos θ) (83)
4s s s
r " #
4πα2 4m2µ 2m2µ
σ = 1− 1+ . (84)
3s s s
α2 α2 +1 4πα2
Z
dσ 2
for E  mµ , = (1 + cos θ) ⇒ σ = 2π × (1 + cos2 θ) d(cos θ) = (85)
dΩ 4s 4s −1 3s

If the initial energy is high s > 4m2µ then the cross section of inelastic scattering is increased and
makes production of heavier particles possible through QED processes. An important issue is the
total cross-section for production of quark-antiquark pairs that finally ends in hadrons,

e+ e− → q q̄ (86)

Two modifications are necessary, (i) replace muon charge e with Qe, where Q = 2/3 for u, c, t quarks
and Q = −1/3 for d, s, b quarks, (ii) multiply the answer by 3 to account for 3 colors. As long
as the energy is high enough compared to the quark masses, the result of (85) will hold subject to
the
√ above two modifications, but will depart significantly if any of the quark mass is approximately
s/2. Defining a quantity,

σ(e+ e+ → q q̄) X
R= + − + −
=3 (Qi e)2 (87)
σ(e e → µ µ ) i

12
As σ(e+ e− → µ+ µ− ) is known, the ratio R gives the direct count of number of colors,
"   2  2 #
2
2 1 1
R = 3 + + = 2 for u, d, s
3 3 3
 2
2 10
= 2+3 = for u, d, s, c
3 3
 2
10 1 11
= +3 = for u, d, s, c, b (88)
3 3 3
Whatever departures observed are all due to strong interaction becoming significant at the quark
mass threshold.

Helicity structure: So far we have calculated the unpolarized cross-sections for various QED
processes, which involves averaging and summing over electron spins and photon polarizations. But
what if we want to calculate cross-sections for each set of spin orientations? We will use the QED
process e+ e− → µ+ µ− to compute this simply because we have only one Feynman diagram to worry
about. However, the Dirac spinors u1 , u2 , v1 , v2 given in (4) are the energy eigenstates and not the
eigenstates of Ŝz , unless the (anti-) particles are traveling in the z-direction, px = py = 0, pz 6= 0,
~ 6= 0 but, H, Σ ~ · p~ = 0 and, γ µ pµ − m u = 0, γ µ pµ + m v = 0
   


H, S (89)
     ∓pz   
1 0 E+m
0
 0   1   0   ±pz 
u1 =   E+m 
 ±pz  , u2 =  0  , v1 =  1  , v2 =  0  (90)
    
E+m
∓pz
0 E+m 0 1
 
  1 0 0 0
1 1 σz 0 1 0 −1 0 0 
Ŝz = Σz = =   (91)
2 2 0 σz 2 0 0 1 0 
0 0 0 −1
1 1 1 1
Ŝz u1 = u1 , Ŝz u2 = − u2 , (−Ŝz ) v1 = v1 , (−Ŝz ) v2 = − v2 (92)
2 2 2 2
The reason why we cannot have simultaneous energy and spin eigenstates of Dirac equation is fairly
~ 6= 0. However, the helicity operator Σ
obvious, [H, S] ~ · p̂ does commute with Dirac hamiltonian H.
~ ~
From the definition, the operator Σ · p̂ = 2S · p̂ is the component of spin in the direction of the
momentum p̂. The eigenvalues of the helicity operator for spin-1/2 particles are, obviously, ±1. The
+1 helicity states are called right-handed (R) and the −1 helicity states are left-handed (L). What
would then be the solutions of Dirac equation that are also eigenstates of helicity. Defining helicity
projection operator
1 ~ · p~


ΠR,L ≡ 1 ± Σ → ΠR,L u = uR,L , ΠR,L v = vR,L (93)
2
where uR,L and vR,L are obviously 4-component spinors. Then it follows that
~ · p̂) uR.L = ± uR,L
(Σ and ~ · p̂) vR.L = ∓ vR,L .
(Σ (94)
The helicity states
 uR,L etc. are further splitted for convenience into 2-component objects ua,b such

ua
that uR,L = . The action of helicity operator thus is,
ub
       
~ ua ~σ · p̂ 0 ua ua (~σ · p̂)ua = ±ua
(Σ · p̂) = =± ⇒ (95)
ub 0 ~σ · p̂ ub ub (~σ · p̂)ub = ±ub
Consider a spin-1/2 particle moving in (θ, φ) direction, p̂ = (sin θ cos φ, sin θ sin φ, cos θ).
sin θ e−iφ
   
pz px − ipy cos θ
~σ · p̂ = ⇒ ~σ · p̂ = . (96)
px + ipy −pz sin θ eiφ − cos θ

13
  
a
Writing ua,b = , the helicity eigenvalue equation (95) give,
b

sin θ e−iφ
    
cos θ a a b ±1 − cos θ iφ
=± ⇒ = e (97)
sin θ eiφ − cos θ b b a sin θ

For the R-handed helicity states (helicity = +1), we have,

sin θ2 iφ cos θ2
 
b 1 − cos θ iφ
= e = e ⇒ uR
a,b ∝ (98)
a sin θ cos θ2 e sin θ2
iφ

Using proportionality constants λ1,2 , the R-handed helicity state is written as,

λ1 cos θ2
 
 λ1 eiφ sin θ 
uR =  2  (99)
 λ2 cos θ 
2
λ2 eiφ sin θ2

The undetermined constants can be fixed by noting that Dirac equation gives us,
  
E−m −~σ · p~ ua
γ µ pµ − m u =


=0
~σ · p~ −E − m ub
~σ · p~ |~
p| |~
p|
⇒ ub = ua = (~σ · p̂) ua = + ua (100)
E+m E+m E+m
where +-sign follows from relative
√ normalization of ua and ub determined by (95). The overall
normalization is as usual N = E + m giving 2E particles per unit volume. In exactly the same
way the uL and the antiparticle spinors can be obtained. For the anti-particle spinors of course, the
equation (94) will be,
~ · p̂)vR,L = ∓vR,L and va = ~σ · p~ vb
(Σ (101)
E+m
Then the particle and anti-particle helicity eigenstates are,
   
cos θ2 − sin θ2
 eiφ sin θ2   eiφ cos θ2 
uR = N  , u = N ,
   
|~p| θ L |~
p | θ
 E+m cos 2  E+m sin 2
 
 
|~
p| iφ θ |~
p| iφ θ
E+m e sin 2 − E+m e cos 2
   
|~
p| θ |~
p| θ
E+m sin 2 E+m cos 2
|~
p| iφ θ   |~ p| iφ θ 
 − E+m

vR = N  e cos 2

, v = N
 E+m e sin 2 .

(102)
θ
 L  θ
 sin 2   cos 2 
eiφ cos θ2 eiφ sin θ2

For the purpose of discussion in the context of high energy e+ e− → µ+ µ− , we consider the limit
E  m in CM frame,

p1 = (E, 0, 0, E), p2 = (E, 0, 0, −E), q2 = (E, −E sin θ, 0, −E cos θ)

q1 = (E, E sin θ, 0, E cos θ),
(103)
In such case, the R and L-handed particle / antiparticle spinors in (102) become,
       
c −s s c
√  seiφ  √  iφ  √  iφ  √  iφ 
uR = E   , uL = E  ce  vR = E  −ce  , vL = E  se 
 c   s   −s   c 
seiφ −ceiφ ceiφ seiφ
(104)

14
where s = sin θ/2 and c = cos θ/2. The initial state electron can either be in a R or L-handed
helicity state with θ = 0 and choosing φ = 0,
   
1 0
√  0  √  1 
e− : uR (p1 ) = E   1  or uL (p1 ) = E  0 
   (105)
0 −1

Similarly, the initial state positron helicity spinors can be obtained using θ = π and choosing φ = 0,
   
1 0
√ 0 √
e+ : vR (p2 ) = E   1 
   
 −1  or vL (p2 ) = E  0  (106)


0 1

The corresponding helicity states for µ− are uR,L (q1 ) with some θ but φ = 0 and for µ+ are vR,L
with θ = π − θ but φ = π,
   
c −s
√  s  √  c 
µ− : uR (q1 ) = E  c  or uL (q1 ) = E  s 
   (107)
s −c
   
c s
√  s  √  −c 
µ+ : vR (q2 ) = E  −c  or vL (q2 ) = E  s 
   (108)
−s −c

So in the initial e+ e− state there are 4 possible combinations of helicity and likewise 4 possible
combinations in the final µ+ µ− helicity states:
− − − −
|ii → e+
R eR , e+
R eL , e+
L eR , e+
L eL (109)
|f i → µ+
R µ−
R, µ+
R µ−
L, µ+
L µ−
R, µ+
L µ−
L (110)
1
|M |2unpolarized |MRR→RR |2 + |MRR→RL |2 + · · · |MLL→LL |2

= (111)
4
that is the unpolarized invariant amplitude is spin average over all 16 possible helicity (or spin)
combinations. To explicitly calculate the individual Feynman amplitudes, we need to evaluate the
electron and muon currents (76). Explicitly calculation of the electron current is worked out here.

je = v s2 (p2 ) γ µ us1 (p1 ) = vs†2 (p2 )γ 0 γ µ us1 (p1 ) ⇒ (112)

†
v R (p2 )γ 0 uR (p1 ) = vR (p2 )(γ 0 )2 uR (p1 ) = 0
1 †
v R (p2 )γ uR (p1 ) = vR (p2 )(γ 0 γ 1 )uR (p1 ) = 0
2 †
v R (p2 )γ uR (p1 ) = vR (p2 )(γ 0 γ 2 )uR (p1 ) = 0
3 †
v R (p2 )γ uR (p1 ) = vR (p2 )(γ 0 γ 3 )uR (p1 ) = 0
µ
⇒ v R (p2 )γ uR (p1 ) = (0, 0, 0, 0) (113)
Similarly, v L (p2 )γ µ uL (p1 ) = (0, 0, 0, 0) (114)
0 †
v R (p2 )γ uL (p1 ) = vR (p2 )(γ 0 )2 uL (p1 ) = 0
†
v R (p2 )γ 1 uL (p1 ) = vR (p2 )(γ 0 γ 1 )uL (p1 ) = −2E
2 †
v R (p2 )γ uL (p1 ) = vR (p2 )(γ 0 γ 2 )uL (p1 ) = −2iE
3 †
v R (p2 )γ uL (p1 ) = vR (p2 )(γ 0 γ 3 )uL (p1 ) = 0
µ
⇒ v R (p2 )γ uL (p1 ) = 2E (0, −1, −i, 0) (115)
Similarly, v L (p2 )γ µ uR (p1 ) = 2E (0, −1, i, 0) (116)

15
This is a very interesting result – only 2 out of possible 4 initial helicity states are non-zero corre-
− + − + − + −
sponding to e+R eL and eL eR meaning we cannot have eR eR or eL eL initial helicity states.

Similarly for muons,

jµ = ur1 (q1 ) γ µ vr2 (q2 ) = u†r1 (q1 )γ 0 γ µ vr2 (q2 ) ⇒ (117)

µ µ
uR (q1 )γ vR (q2 ) = uL (q1 )γ vL (q2 ) = (0, 0, 0, 0) (118)
µ
uR (q1 )γ vL (q2 ) = 2E (0, − cos θ, i, sin θ), (119)
µ
uL (q1 )γ vR (q2 ) = 2E (0, − cos θ, −i, sin θ) (120)

Thus in the limit E  m only 2 helicity combinations are nonzero for final muon states as well.
This is an important feature of QED and it applies equally well to QCD. Consequently, out of 16
possible helicity combinations only four RL → RL, RL → LR, LR → LR, LR → RL transitions
contribute. The calculation of Feynman amplitude follows from the expression in eqn. (76), using
je and jµ obtained above from eqn. (112) through eq. (120),

− + − e2
M (e+
R eL → µR µL ) [2E (0, −1, −i, 0)] [2E (0, − cos θ, i, sin θ)]
=
s
e2
= 4E 2 (1 + cos θ) = e2 (1 + cos θ) (121)
s
− + − e2
M (e+ e
R L → µ µ
L R ) = [2E (0, −1, −i, 0)] [2E (0, − cos θ, −i, sin θ)]
s
e2
= − 4E 2 (1 − cos θ) = −e2 (1 − cos θ) (122)
s
2
− + − e
M (e+
L eR → µR µL ) = [2E (0, −1, i, 0)] [2E (0, − cos θ, i, sin θ)]
s
e2
= − 4E 2 (1 − cos θ) = −e2 (1 − cos θ) (123)
s
2
− + − e
M (e+
L eR → µL µR ) = [2E (0, −1, i, 0)] [2E (0, − cos θ, −i, sin θ)]
s
e2
= 4E 2 (1 + cos θ) = e2 (1 + cos θ) (124)
s
The situation is depicted in the figure below.

Hence the invariant amplitude in CM frame in the limit E  m is the average of the sum of modulus

16
square of the individual amplitudes,
1
|M |2 |M (RL → RL)|2 + M (RL → LR)|2 + M (LR → RL)|2 + M (LR → RL)|2

=
4
e4 
2(1 + cos θ)2 + 2(1 − cos θ)2 = e4 (1 + cos2 θ)

= (125)
4
This expression is exactly what we would get from (82) in the limit E  mµ . In the ultra-relativistic
limit that we are working, the helicity eigenstates are also eigenstates of γ 5 ,

γ 5 uR = + uR , γ 5 uL = − uL , γ 5 vR = − vR , γ 5 vL = + vL (126)
 
0 I
where, γ 5 ≡ iγ 0 γ 1 γ 2 γ 4 = , γ 5 γ µ = −γ µ γ 5 , (γ 5 )2 = 1 and (γ 5 )† = γ 5 (127)
I 0

Defining the projection operators PR,L as,

1 + γ5 1 − γ5
PR = and PL = (128)
2 2
2 2
with PR + PL = 1, PR = PR , PL = PL , PR PL = 0. (129)

These PR,L project out the chiral eigenstates when acting on general eigenstate (similar to helicity
projection operators, which we did not use here),

PR,L ψ = ψR,L and ψPR,L = ψ L,R , ψ = ψR + ψL (130)

In light of this, vector current ψγ µ ψ can be decomposed as follows,

ψγ µ ψ = ψ(PL + PR )γ µ (PL + PR )ψ
= (ψ R + ψ L )γ µ (ψL + ψR )
= ψ R γ µ ψR + ψ L γ µ ψL (131)

are the only non-zero terms. But we know from (126) that right-handed spinor actually corresponds
−
to left-handed anti-particle, therefore the non-zero terms of vector current corresponds to e+
R eL and
+ −
eL eR as is obtained before with helicity states.

17