Final exam solutions, Phys 221B, W15

1. In HW5, #4 (Srednicki 48.5), you had the interaction

L1 = 2gN γ µ PL M∂µ φ + h.c. ,

with mφ > mM and mN = 0. Now let mM > mφ and mN = 0.

a) Calculate the decay rate for M → φN when none of the spins is measured.
b) Calculate the decay rate with N having measured helicity, positive or negative.
c) Calculate the differential decay rate 11.48 when the initial M is known to have
sz = + 12 and the spin of the N is not measured.
d) Which of P , T, and C is violated by this interaction? For each violation, can it be
seen in the calculation you’ve done in (b)? in (c)? (Explain precisely).

a) Labeling the momenta M(p) → φ(k)N (p0 ), and m ≡ mM > M ≡ mφ and mN = 0.

Using the Dirac equation, we have

T = igu(p0 )k/ (1 − γ5 )u(p)

= igu(p0 )(p/ − p/ 0 )(1 − γ5 )u(p)
= imgu(p0 )(1 + γ5 )u(p)
. (1)

Using
(ū(p0 )(1 + γ5 )u(p))∗ = ū(p)(1 + γ5 )u(p0 ) = ū(p)(1 − γ5 )u(p0 ) .
this becomes
X
|T |2 = −m2 g 2 Tr[(m − p/ )(1 − γ5 )p/ 0 (1 + γ5 )]
spins

= −2m2 g 2 Tr[(m − p/ )p/ 0 (1 + γ5 )]

= −2g 2 m2 Tr[(m − p/ )p/ 0 ]
= −8g 2 m2 p · p0 = 4g 2 m2 (m2 − M 2 ) .

Note the properties (1 ± γ5 )a/ = a/ (1 ∓ γ5 ), (1 + γ5 )(1 − γ5 ) = 0, (1 ± γ5 )γ5 = (1 ± γ5 )(±1).

These help to simplify things. Then
1 0 1
X
Γ= |~
p | |T |2
8πm2 2 spins
g 2 (m2 − M 2 )2
=
8πm
1
 1
Note the factor 2
to average over the initial spin.
b) Now
X g 2 m2 
|T |2 = − Tr (m − p/ )(1 − γ5 )(1 + s0 γ5 )p/ 0 (1 + γ5 ) .

initial
2
spin

Noting that (1 − γ5 )(1 + s0 γ5 ) = (1 − γ5 )(1 − s0 ), we get immediately that

0 0 g 2 (m2 − M 2 )2
Γ(s = +1) = 0 , Γ(s = −1) = .
8πm
This is the property of the weak interaction that only left-handed fermions couple to it.
Many people missed the s-dependence due to a subtlety. According to 38.28,
1
us ūs = (1 − sγ5 z/ )(m − p/ ) ,
2
and then
1 1
(1 − sγ5 z/ )(m − p/ ) = (1 − sγ5 )(−p/ ) .
lim
m→0 2 2
However, some people replaced this with
1
us ūs = (1 − sγ5 z/ )(−p/ ) ,
2
which is not the same. The point is that as m → 0, z/ is order 1/m (38.30), so there is
another term.
c) Now
X g 2 m2 
|T |2 = − Tr (1 − sγ5 z/ )(m − p/ )(1 − γ5 )p/ 0 (1 + γ5 )

final
2
spin

= −g 2 m2 Tr (1 − sz/ )(m − p/ )p/ 0 (1 + γ5 )

 

= −g 2 m2 Tr[(1 − sz/ )(m − p/ )p/ 0 ]

= −4g 2 m2 (msz · p0 + p · p0 )
= 2g 2 m2 (m2 − M 2 )(1 − s cos θ) ,

where cos θ is the angle between the z-axis and the N momentum, and the problem asks
for s = +1. The differential decay rate is then

dΓ g 2 (m2 − M 2 )2
= (1 − cos θ)
dΩ 32π 2 m
2
 We can understand the angle-dependence as follows. When θ = 0 the momentum
is in the z-direction, and the spin must also be in the z-direction by angular momentum
conservation. But the neutrino has negative helicity (part b) so the amplitude must vanish.
d) From Srednicki 40.37, 40.41, and 40.47, we see that the γ µ and γ µ γ5 interactions
have opposite P and C transformations, so these symmetries must be violated no matter
how φ transforms. They have the same T transformation; to cancel the minus signs in
40.41 we need T −1 φ(x)T = −φ(T x).
For P , momentum is a vector and spin is a pseudovector, so the helicity changes sign.
The helicity-dependent amplitude in (b) therefore implies violation of P . Similarly in (c)
the correlation between initial spin and final direction implies violation of P . C.-S. Wu
discovered parity violation by a correlation between spin and momentum in nuclear β
decay.
We can’t probe C in this one calculation. We would have to look also at M → φN ,
and in fact we would find that antiparticles have the opposite correlation between spin and
momentum, violating C but preserving CP . To probe T directly we would need to study
the inverse process φN → M.
2. Consider a theory with two triplets of real scalars, φi and χi , so i runs from 1 to 3.
Let
1
L = − (∂µ φ·∂ µ φ + ∂µ χ·∂ µ χ + m2 φ·φ + m2 χ·χ)
2
λ λ0
− (φ·φ φ·φ + χ·χ χ·χ) − φ·χ φ·χ ,
8 2
where the dot products are with respect to i.
a) Find all continuous and discrete internal symmetries. Internal means acting trivially
on spacetime, so excludes Lorentz, P , and T , which are obvious.
b) Is this the most general renormalizable Lagrangian in d = 4 with these symmetries?
If not, what is missing?
c) For m2 < 0 and λ > 0, find all classical minima of the potential. The coupling λ0 is
allowed to have either sign: consider all values of λ0 .
d) Determine the unbroken continuous symmetry and the number of Goldstone bosons,
as a function of λ0 .

a) Since all the i indices are dotted, there is an obvious SO(3),

φi → Oij φj , χi → Oij χj .

3
The orthogonal matrix O must be the same for both, as there is a term where φi is dotted
into χi . Since every term has an even number of φ’s and and even number of χ’s, there are
both of
Z : (φi , χi ) → (−φi , χi ) , Z 0 : (φi , χi ) → (φi , −χi ) .
By this notation I mean that these act on all i at once. i-dependent operations would be
part of SO(3). Finally, since the two m’s are equal and the two λ’s are equal, there is the
swap
Z 00 : (φi , χi ) → (χi , φi ) .
Of course there are also products of these, for example (φi , χi ) → (−φi , −χi ) is ZZ 0 .
b) In the potential, the SO(3) requires that the quadratic terms be of one of the forms

φ·φ , χ·χ , φ·χ .

The symmetry Z or Z 0 forbids φ · χ, and Z 00 requires that the remaining two have equal
coefficients m2 . In the quartic term the SO(3) invariants are

(φ·φ)2 , (χ·χ)2 , (φ·χ)2 , φ·φ χ·χ , φ·χ φ·φ , φ·χ χ·χ .

The last two are forbidden by Z and Z 0 . The first four are allowed, with the first two
having equal coefficients due to Z 00 . So we see that a term −λ00 φ·φ χ·χ/4 should have been
included. In fact, there is a one-loop graph, with one λ vertex, and one λ0 vertex, whose
divergence requires a λ00 counterterm.
c) The only term that depends on the relative orientation φi and χi is the λ0 term,

φ·χ φ·χ = φ·φ χ·χ cos2 θ

where θ is the angle between φi and χi .

For λ0 > 0 it is favorable for them to be perpendicular,

φi = vni , χi = v 0 n0i , n·n = n0 ·n0 = 1 , n·n0 = 0 . (2)

The potential becomes

1 1
V = m2 (v 2 + v 02 ) + λ(v 4 + v 04 ) ,
2 8
which is minimized at
v 2 = v 02 = −2m2 /λ .
For λ0 < 0 it is favorable for them to be parallel (or antiparallel),

φi = vni , χi = v 0 ni , n·n = 0 . (3)

4
The potential becomes
1 1 1
V = m2 (v 2 + v 02 ) + λ(v 4 + v 04 ) + λ0 v 2 v 02 ,
2 8 2
which is minimized at
2m2
v 2 = v 02 = − .
λ + 2λ0
However, for λ + 2λ0 < 0 there is no minimum, we can decrease the energy without bound
by increasing v and v 0 .
When λ + 2λ0 = 0 the potential determines only v 2 − v 02 . This case thus has degenerate
vacua not related by any symmetries. Such a situation is unnatural (fine-tuned) here, but
can arise in supersymmetric theories.
d) For λ0 < − 12 λ there is no vacuum at all so the question is moot. For − 21 λ < λ0 < 0,
the vacuum (3) is invariant under the SO(2) or U (1) rotation around ni , so SO(3) is broken
to SO(2). There are 3 − 1 = 2 Goldstone bosons, from the possible choices of n on the
two-sphere. For λ0 > 0, the symmetry is fully broken: the SO(2) that leaves ni invariant
rotates n0i around ni . The breaking is SO(3) → I and there are three Goldstone bosons,
two from the direction of ni and one from the angle of n0i around ni . For completeness,
when λ0 is exactly zero there is an O(3) × O(3) symmetry and the relative orientation of
φi and χi is undetermined. The breaking O(3) × O(3) → O(2) × O(2) leaves 4 Goldstone
bosons.
3. Consider the renormalizable d = 2 theory
∞
1 µ 1 2 2 X λn 2n
L = − ∂µ φ∂ φ − m φ − φ .
2 2 n=2
(2n)!

a) Considering V2k , show that there is a divergent one-loop graph that is first-order in
couplings (just one vertex). Calculate Zλk and Zm to this order and then calculate βλk and
γm in MS.
b) Defining
∞
1 2 2 X λn 2n
V (φ) = m φ + φ ,
2 n=2
(2n)!
write the equation for the running m(µ) and λn (µ) in terms of V (φ, µ).
c) If V (φ, µ1 ) = g cos βφ for constants g and β, what is V (φ, µ2 )? If V (φ, µ1 ) = g cosh βφ
for constants g and β, what is V (φ, µ2 )?
d) This also provides a nice example of Wilsonian renormalization. Consider the theory
with bare potential V1 (φ) = g1 cos β1 φ, and a UV cutoff on the propagator momenta,

5
k 2 < Λ21 . Consider the contribution to V2k from all graphs that have just one vertex but
any number of loops. Show that you get the same result with a lower cutoff Λ2 and a new
potential g2 cos β2 φ. Give g2 , β2 in terms of g1 , β1 . Assume that m  Λ1 , Λ2 .

a) There is a divergent contribution to V2k in which two of the lines from a λk+1 vertex
join to form a loop. Then

d2− q̄
Z
1 1
V2k = −λk Zk − λk+1
2 (2π) q̄ + m2
2− 2

1
= −λk Zk − λk+1 Γ(/2)(4π/D)/2
8π
1
= −λk Zk − λk+1 + finite .
4π
So in MS,
λk+1
Zk = 1 −
4πλk 
to this order. A similar calculation yields
λ2
Zm = 1 −
4πm2 
from the usual graph.
Now,

0 = µ∂µ bare
ln λk0
= µ∂µ bare
ln(λk Zk µ̃(k−1) ) (4)

In the second line, we have used that the dimension of φ is 21 (d − 2) = −/2, so the
dimension of λk is 2 while that of λk0 is 2 + (k − 1). Also, I’ve set Zφ = 1 to this order.
This becomes
∞
X
0 = (k − 1) + β̂λj ∂λj ln(λk Zk )
j=2

β̂k β̂k+1 β̂k λk+1

= (k − 1) + − + . (5)
λk 4πλk  4πλ2k 

Expanding in , we have from order 1

β̂k = −(k − 1)λk + βk .

6
Order 0 then gives

βk kλk+1  (k − 1)λk λk+1

0= + − . (6)
λk 4πλk  4πλ2k 

and so
λk+1
βk = − .
4π
Similarly one finds
λ2
γm = − .
8πm2
b) We have
µ∂µ m2 = 2γm m2 , µ∂µ λk = βk .
Then
∞
X βn 2n
µ∂µ V (φ, µ) = γm m2 φ2 + φ
n=2
(2n)!
∞
1 X λn+1 2n
= − φ
4π n=1 (2n)!
∞
m=n+1 1 ∂ 2 X λm 2m
= − φ + constant
4π ∂φ2 m=1 (2m)!
1 ∂2
= − V (φ, µ) + constant . (7)
4π ∂φ2

We have defined λ1 = m2 for uniformity. As usual, the constant is not important. If you
did not find this exact form, that’s fine, but this makes the next part easier.
c) Using part (b), we get for g cos bφ (I’re renamed β to b to avoid confusion)

b2
µ∂µ g = g, µ∂µ b = 0 .
4π
So
2 /4π 2 /4π
g(µ2 ) = (µ2 /µ1 )b g(µ1 ) , V (φ, µ2 ) = (µ2 /µ1 )b V (φ, µ1 ) .
for the cosine potential. Similarly
2 /4π 2 /4π
g(µ2 ) = (µ2 /µ1 )−b g(µ1 ) , V (φ, µ2 ) = (µ2 /µ1 )−b V (φ, µ1 ) .

for the cosh potential.

7
 Note, by the way, that the dimensionless coupling ĝ(µ) = g(µ)/µ2 satisfies
2 /4π−2
ĝ(µ2 ) = (µ2 /µ1 )b ĝ(µ1 )

in the cosine case. This coupling is relevant, meaning that the dimensionless size grows at
low energy, only for b2 < 8π, and it is irrelevant for larger values. The point b2 = 8π is
known as the Kosterlitz-Thouless phase transition.
d) A vertex λ2k+2l gives a divergent contribution to V2k in which 2l of the external
lines pair up into l loops, each a copy of the one-loop divergence above:
∞ l
d2 q̄
Z
X 1 1
V2k = − λk+l l .
l=0
2 l! (2π) q̄ + m2
2 2

The symmetry factor 2l comes from interchanging the two ends of each loop. The symmetry
factor l! comes from permuting the loops.
Now, the Wilsonian idea is that we focus on the effect of modes between Λ1 and Λ2 . If
we only include those in the integral, we get
Z Λ1 2
d q̄ 1 1
2 2 2
= ln(Λ1 /Λ2 )
Λ2 (2π) q̄ + m 2π

for negligible m2 . We can then write

V2k = −λk,2

where ∞  l
X 1 1
λk,2 = λk+l,1 ln(Λ1 /Λ2 ) .
l=0
l! 4π
That is, we absorb the virtual effects in this range into the effective coupling at lower scale
Λ2 . Now, for V1 (φ) = g1 cos b1 φ, λk,1 = (−b1 )2k . Then we can do the sum,
2 2
λk,2 = (−b1 )2k (Λ1 /Λ2 )−b1 /4π

as found above.
It’s a bit different because here we’re talking about cutoffs Λ, and before about the
reference scale µ. But we can think about setting the cutoff just slightly above the scale of
interest, so as to integrate out all higher energy effects, so it comes to the same thing.