8.

323 Problem Set 2 Solutions

February 21, 2023

Question 1: A Problem with Relativistic Quantum Mechanics (20 points)

The Schrödfinger equation for a free non-relativistic particle is:
1 2
i∂t ψ(~x, t) = − ∇ ψ(~x, t)
2m
The generalization of the above equation to a free relativistic particle is the so-called Klein-Gordon
equation

∂t2 ψ(~x, t) − ∇2 ψ(~x, t) + m2 ψ(~x, t) = 0

We emphasize that in both these equations, ψ(~x, t) is interpreted as a wave function for the dynamical
variable ~x(t), rather than a dynamical field.

(a) As a reminder, derive from the Schrödinger equation the continuity equation for the probability

∂t ρ + ∇ · J~ = 0

where
i
ρ = |ψ|2 , J~ = − (ψ ∗ ∇ψ − ψ∇ψ ∗ )
2m
We compute:
i  ~2 ∗ 
∂t ρ = ψ∂t ψ ∗ + ψ ∗ ∂t ψ = − ψ∇ ψ − ψ∗∇ ~ 2ψ
2m
i ~  ∗~ 
~ ∗ = −∇~ · J~
= ∇ · ψ ∇ψ − ψ ∇ψ
2m
where we use the Schrödinger equation in the second equality.

(b) Suppose ψ(~x, t) has the plane wave form, i.e.

~
ψ(~x, t) ∝ eik·~x

for some real vector ~k. Find the solutions to the Klein-Gordon equation above.
We substitute the ansatz ψ(x, t) = eik·x φ(t) into the Klein-Gordon equation to get an equation for φ(t):

∂t2 φ + (k2 + m2 )φ = 0

This has plane-wave solutions of positive and negative frequencies,

p
φ(t) = Ae−iωk t + Beiωk t , ωk = k2 + m2

1
Hence, the Klein-Gordon equation has solutions

ψ(~x, t) = Aei(−ωk t+k·x) + Bei(ωk t+k·x)

(c) Show that the Klein-Gordon equation also leads to a continuity equation, with ρ and J~ now given by

i i
ρ= (ψ ∗ ∂t ψ − ψ∂t ψ ∗ ). J~ = − (ψ ∗ ∇ψ − ψ∇ψ ∗ )
2m 2m
In the same way as in part (a), we compute:

i i
ψ ∗ ∂t2 ψ − ψ∂t2 ψ ∗ = (ψ ∗ ∇2 ψ − ψ∇2 ψ ∗ ) = −∇
~ · J~


∂t ρ =
2m 2m
where we use the Klein-Gordon equation in the second equality.

(d) Argue that this ρ cannot be interpreted as a probability density.

We write
i 1
ρ= (ψ ∗ ∂t ψ − ψ∂t ψ ∗ ) = Im(ψ∂t ψ ∗ )
2m m
Any proper probability density must be positive definite, i.e. ρ ≥ 0. This is not the case here. For
instance, for the plane wave solution A = 0, B from (b) we compute
 
ρ = Im Beiωk t+ik·x (−iωk )Be−iωk t−ik·x = −B 2 ωk < 0
√
Since ωk = k2 + m2 > 0, the existence of negative-frequency solutions means that ρ cannot be positive
definite, and cannot be interpreted as a probability density.

2
Question 2: Commutation relations of creation and annihilation operators (20 points)
For the real scalar field theory discussed in lecture,
1 1
L = − ∂µ φ∂ µ φ − m2 φ2
2 2
we showed that the time-evolution of the quantum operator φ(x, t) is given by
Z
1  
φ(x, t) = d¯2 k √ ak uk (x, t) + a†k u∗k (x, t)
2ωk

where
p
ωk = k2 + m2 , uk = e−iωk t+ik·x

We use π(x, t) to denote the momentum density conjugate to φ. The canonical commutation relations
among φ and π are

[φ(x, t), φ(x0 , t)] = [π(x, t), π(x0 , t)] = 0, [φ(x, t), π(x0 , t)] = iδ (3) (x − x0 )

(a) Show that it is enough to impose the canonical commutation relations at t = 0. That is, once we
impose them at t = 0, then the relations at general t are automatically satisfied.
Note: this statement in fact applies not only to V (φ) = 21 m2 φ2 , but any potential V (φ).
In the Heisenberg picture we have:

[A(x, t), B(x0 , t)] = [eiHt A(x, 0)e−iHt , eiHt B(x0 , 0)e−iHt ] = eiHt [A(x, 0), B(x0 , 0)]e−iHt

Now let us impose the canonical commutation relations at t = 0. Then, it follows that

[φ(x, t), φ(x0 , t)] = [π(x, t), π(x0 , t)] = eiHt 0e−iHt = 0
[φ(x, t), π(x0 , t)] = eiHt iδ (3) (x − x0 )e−iHt = iδ (3) (x − x0 )

These are again precisely the canonical commutation relations, now at generic t.

(b) Express ak and a†k in terms of φ(k) and π(k), where φ(k) and π(k) are Fourier transforms of
φ(x, t = 0) and π(x, t = 0), e.g.
Z
φ(k) = d3 xe−ik·x φ(x, t = 0)

We start with the mode expansions for φ(x, t) and π(x, t)

Z
1  
φ(x, t) = d¯3 k √ ak e−iωk t+ik·x + a†k eiωk t−ik·x
2ωk
Z r
ωk  
3
π(x, t) = −i d¯ k ak e−iωk t+ik·x − a†k eiωk t−ik·x
2

This is almost of the form of a Fourier transform, and by changing variables of one of the terms from
k → −k we have
1  
φ(k, t) = √ ak e−iωk t + a†−k eiωk t
2ωk
r
ωk  
π(k, t) = −i ak e−iωk t − a†−k eiωk t
2

3
Now, observe that the equations are decoupled in k. We can take t = 0 and solve this as a regular system
of equations for ak and a†k .
r r
ωk 1
ak = φ(k) + i π(k)
2 2ωk
r r
† ωk 1
ak = φ(−k) − i π(−k) (1)
2 2ωk

(c) Using the expressions derived in part (b), deduce the commutation relations

[ak , ak0 ], [a†k , a†k0 ], [ak , a†k0 ]

from the commutation relations above at t = 0.

It is useful to take the Fourier transform F (from position to momentum space) of the t = 0 canonical
commutation relations:

[φ(k), φ(k0 )] = [π(k), π(k0 )] = Fx→k ◦ Fx0 →k0 (0) = 0

Z
0 0
[φ(k), π(k )] = Fx→k ◦ Fx0 →k0 (iδ (x − x )) = i d3 xd3 x0 e−ik·x e−ik ·x δ (3) (x − x0 )
0 (3) 0

Z
0
= i d3 xe−i(k+k )·x = i(2π)3 δ (3) (k + k0 )

Now we compute commutators of creation and annihilation operators using the results in (b)

i 1
[φ(k), π(k0 )] + [π(k), φ(k0 )] = − (2π)3 (δ (3) (k + k0 ) − δ (3) (k0 + k))


[ak , ak0 ] = + =0
2 2
i 1
[a†k , a†k0 ] = − [φ(−k), π(−k0 )] + [π(−k), φ(−k0 )] = − (2π)3 (δ (3) (−k − k0 ) − δ (3) (−k0 − k)) = 0


2 2
† i 1
[ak , ak0 ] = + −[φ(k), π(−k0 )] + [π(k), φ(−k0 )] = − (2π)3 (−δ (3) (k − k0 ) − δ (3) (k0 − k))


2 2
= (2π)3 δ (3) (k − k0 ) (2)

4
Question 3: Noether charges in terms of creation and annihilation operators (20 points)
In problem set 1, we obtained the conserved charges associated with spacetime translational symmetries
for a complex scalar field theory. The results there can be easily converted to the corresponding expressions
for a real scalar field theory.
1 1
L = − ∂µ φ∂ µ φ − m2 φ2
2 2
(a) Express the Hamiltonian H of this theory in terms of ak and a†k .
From problem set 1, we quote
Z
1
H= d3 x(π 2 + (∇φ)2 + m2 φ2 )
2
It is convenient to first convert this expression into momentum space, before using the decomposition into
creation and annihilation operators. We use the identity:
Z Z
0
d xf (x)g(x) = d3 xd3 kd3 k0 ei(k+k )·x f (k)g(k0 )
3

Z Z
1 3 3 0 (3) 0 0
= d kd k δ (k + k )f (k)g(k ) = d¯3 kf (k)g(−k) (3)
(2π)3
More generally if there are derivatives acting on f or g, each derivative acting on f drags down a factor
of ik, while each derivative actin on g drags down a factor of ik0 , which becomes −ik after performing
the d3 x integral. We further use the shorthand d¯x = dx/2π (d¯ is to d as ~ is to h).

Hence, we can now write

Z
1
d¯3 k π(k, t)π(−k, t) + (k2 + m2 )φ(k, t)φ(−k, t)


H=
2
Z
1
d¯3 k π(k, t)π(−k, t) + ωk2 φ(k, t)φ(−k, t)


=
2
Z  Z Z
1 3 ωk

† † † 1
= d¯ k ak (t)ak (t) + ak (t) ak (t) = d¯ kωk ak ak + (2π) δ(0) d¯3 kωk
3 3
2 2 2
In the third equality we use the relations (1) from problem 2(b). In the last equality we use the commutator
(2) from problem 2(c), as well as the expressions ak (t) = e−iωk t ak and a†k (t) = eiωk t a†k . Note that in the
above calculation, we showed that the time-dependence cancels explicitly. We could have also used that
H is conserved to remove the time-dependence immediately by evaluating all fields at t = 0.

This can be written as

Z
H= d¯3 kωk Nk + E0
Z
1
for the number operator Nk = a†k ak , and zero-point energy E0 = (2π)3 δ(0) d¯3 kωk .
2

(b) Express the conserved charges P i for spatial translations, in terms of ak and a†k .
Again we quote the charges from problem set 1, and use (3) to write it in momentum space.
Z Z
P = d x π∂ φ = −i d¯3 kπ(k, t)φ(−k, t)k i
i 3 i

Z
1   
= d¯3 kk i ak (t) − a−k (t)† a−k (t) + ak (t)†
2
Z
1  
= d¯3 kk i ak a−k e−2iωk t + ak a†k − a†−k a−k − a†−k a†k e2iωk t
2

5
In the third equality we use the relations (1) from problem 2(b). Observe that due to the k i factor and
the commutation relations (2), the first and fourth terms in our final expression are odd under the change
of variables k → −k, so they must vanish. Therefore,
Z  Z Z Z
1  1
Pi = d¯3 kk i ak a†k + a†k ak = d¯3 kk i a†k ak + d¯3 k(2π)3 δ (3) (0)k i = d¯3 kk i Nk
2 2

In the second equality we use the commutator (2), and in the third equality we note that the last term
vanishes because the integrand is k-odd.

We may combine this with the expression for H in part (a) to write
Z
P = d¯3 k k µ Nk + δ µ0 E0 ,
µ
k 0 = ωk (4)

(c) Starting with

Z
1  
φ(0, 0) = d¯3 k √ ak + a†k
2ωk

show that under the action of translation operators,

i xi i xi
φ(x, t) = eiHt−iP φ(0, 0)e−iHt+iP

Hint: this problem becomes trivial using the following formula for a harmonic oscillator,

eiαN ae−iαN = e−iα a, N = a† a

The formula in the hint follows from the Baker-Campbell-Hausdorff (BCH) formula,
1
eX Y e−X = Y + [X, Y ] + [X, [X, Y ]] + · · ·
2!
We check:

[iαN, a] = iα[a† a, a] = −iαa

1
eiαN ae−iαN = a + (−iα)a + (−iα)2 a + · · · = e−iα a
2!
and similarly, e−iαN a† e−iαN = eiα a† .

Now we generalize. For α(k0 ) a real-valued function,

d̄3 k0 α(k0 )Nk0 d̄3 k0 α(k0 )Nk0 d̄3 k0 δ(k−k0 )α(k0 )Nk0 d̄3 k0 δ(k−k0 )α(k0 )Nk0
R R R R
ei ak e−i = ei ak e−i
= eiα(k)Nk ak e−iα(k)Nk = e−iα(k) ak (5)

In the first equality, we use that the 2 sets of operators {αk , αk† , Nk } and {αk0 , αk† 0 , Nk0 } commute with
each other for k 6= k0 . This allows us to move all but the k0 = k exponentials on the left-hand side past the
ak factor, where it cancels out with the exponentials on the right. Note that for this, it is essential that
α(k0 ) is a real-valued function. The last equality follows from the instance of the BCH formula derived
above. In the same way, we have that
d̄3 k0 α(k0 )Nk0 † −i d̄3 k0 α(k0 )Nk0
R R
ei ak e = eiα(k) a†k (6)

6
Using our expressions for H and P i in part (a) and (b), identities (5)-(6) allow us to compute
i xi ) i xi ) d̄3 k0 (ωk0 t−k0 ·x)Nk0 +iE0 t d̄3 k0 (ωk0 t−k0 ·x)Nk0 −iE0 t
R R
ei(Ht−P ak e−i(Ht−P = ei ak e−i
d̄3 k0 (ωk0 t−k0 ·x)Nk0 d̄3 k0 (ωk0 t−k0 ·x)Nk0
R R
= ei ak e−i = e−i(ωk t−k·x) ak
i xi ) i xi ) d̄3 k0 (ωk0 t−k0 ·x)Nk0 † −i d̄3 k0 (ωk0 t−k0 ·x)Nk0
a†k e−i(Ht−P
R R
ei(Ht−P = ei ak e = ei(ωk t−k·x) a†k

Finally, we get

d¯3 k i(Ht−P i xi )
Z
i xi ) i xi ) i i
ei(Ht−P φ(0, 0)e−i(Ht−P = √ e (ak + a†k )e−i(Ht−P x )
2ωk
d¯3 k
Z
= √ (ak e−i(ωk t−k·x) + a†k ei(ωk t−k·x) ) = φ(x, t)
2ωk

7
Question 4: Noether charges for Lorentz symmetries of a real scalar (20 points + 10 bonus)
In this problem we work out the conserved currents corresponding to the Lorentz symmetries of a real
scalar theory,
1 1
L = − ∂µ φ∂ µ φ − m2 φ2
2 2
(a) Consider an infinitesimal Lorentz transformation

Λµ ν = δµ ν + ωµ ν

where ωµν = −ωνµ are infinitesimal numbers. Show that this satisfies

Λµ ρ ηρλ Λν λ = ηµν

to first order in ωµν , so this does give a Lorentz transformation.

We compute:

Λµ ρ ηρλ Λν λ = (δµ ρ + ωµ ρ )ηρλ (δν λ + ων λ ) = (ηµλ + ωµλ )(δ λ ν − ω λ ν )

= ηµν + ωµν − ωµν + ωµλ ω λ µ = ηµν + O(ω 2 )

(b) Write down how φ transforms under an infinitesimal Lorentz transformation, and show that the
conserved Noether current for this transformation can be written as

J µλν = xλ T µν − xν T µλ

where T µν is the conserved energy-momentum tensor derived in problem set 1.

A Lorentz scalar field transforms in a way obeying φ0 (x0 ) = φ(x). Therefore, under an infinitesimal
Lorentz transformation, the scalar field φ transforms as

δφ = φ0 (x) − φ(x) = φ((Λ−1 )µ ν xν ) − φ(xν ) = φ((δ µ ν − ω µ ν )xν ) − φ(xν ) = −ω µ ν xν ∂µ φ

where in the last equality we Taylor expand φ(xµ − ω µ ν xν ) = −ω µ ν xν ∂µ φ.

Using this, the Lagrangian density transforms as

∂L
δL = L[φ0 ] − L[φ] = L[φ − ω λ ν xν ∂λ φ] − L[φ] = −ω λ ν xν ∂λ φ = −ω λ ν xν ∂λ L = −∂λ (ω λ ν xν L)
∂φ
We expand only to first order in ω. In the final equality, we use that ωµν is antisymmetric, i.e.

∂λ (ω λ ν xν f (x)) = ω λ ν δλν f (x) + ω λ ν xν ∂λ f (x) = ω λ λ f (x) + ω λ ν xν ∂λ f (x) = ω λ ν xν ∂λ f (x)

Hence, δL = ∂µ F µ , for F µ = −ω µ ν xν L
The Noether current for the transformation parameterized by ωλν is given by
∂L
jµ = − δφ − F µ = ω λ ν xν ∂ µ φ∂λ φ + ω µ ν xν L
∂(∂µ φ)
= ωλν xν (∂ µ φ∂ λ φ + η µλ L) = ωλν xν T µλ

for the energy-momentum tensor from problem set 1 (now with a real scalar):
1
T µν = ∂ µ φ∂ ν φ + η µν L = ∂ µ φ∂ ν φ − η µν (∂ρ φ∂ ρ φ + m2 φ2 )
2

8
Note that ωλν is an arbitrary antisymmetric tensor which parameterizes our infinitesimal transformation.
In total we have an antisymmetric tensor worth of conserved currents, which we can package in J µλν :
1
J µλν = xλ T µν − xν T µλ , j µ = − ωλν J µλν
2
In writing J µλν we have made the antisymmetry in λ and ν manifest by explicitly antisymmetrizing over
these indices.

(c) Using conservation of the energy-momentum tensor, verify that the current in (b) is conserved, i.e.

∂µ J µλν = 0

We compute:

∂µ J µλν = ∂µ (xλ T µν − xν T µλ ) = δµ λ T µν + xλ ∂µ T µν − δµ ν T µλ − xν ∂µ T µλ
= T λν − T νλ = 0

In the 3rd equality we used the conservation law ∂µ T µν = 0, and in the 4th equality we used from problem
set 1 that T µν = T νµ is symmetric.

(d) (Bonus problem) Consider the conserved charges associated with J µλν ,
Z
M = d3 xJ 0λν
λν

Express the conserved charges M µν for the Lorentz symmetries of this theory in terms of ak and a†k .
From part (b), we have
Z Z
M µν = d3 xJ 0µν = d3 x(xµ T 0ν − xν T 0µ )

This is antisymmetric in µ and ν, so we need to compute M 0i and M ij . To do this, we need to expand

T 0µ in terms of creation and annihilation operators:
1
T 0µ = −π∂ µ φ − η 0µ −π 2 + (∇φ)2 + m2 φ2


 2 
1 2
= (π + (∇φ)2 + m2 φ2 ), −π∂ i φ = (H(x), P i (x))
2

Since M µν are conserved currents, we can compute them at t = 0.

First, we need the identity
Z Z
0
d3 xxi f (x)g(x) = d3 xd3 kd3 k0 ei(k+k )·x xi f (k)g(k0 )
Z Z
1 3 0 0 0
= d kd k (−i∂ki δ (k + k ))f (k)g(k ) = i d¯3 k∂ki f (k)g(−k)
3 (3)
(7)
(2π)3

where we have used integration by parts in the last equality. More generally if there are derivatives acting
on f or g, each derivative acting on f drags down a factor of ik, while each derivative actin on g drags
down a factor of ik0 , which becomes −ik after performing the d3 x integral.

9
Now we are ready to compute
Z Z
0i 3 i i 1
M = d x(tP (x) − x H(x))|t=0 = − d3 xxi (π 2 + (∇φ)2 + m2 φ2 )
2
Z
i
d¯3 k ∂ki π(k)π(−k) + ωk2 ∂ki φ(k)φ(−k)


=−
2
√ √
Z
i 
=− d¯3 k −∂ki ( ωk (ak − a†−k )) · ( ωk (a−k − a†k ))
4

2 1 † 1 †
+ ωk ∂ki ( √ (ak + a−k )) · ( √ (a−k + ak ))
ωk ωk
Z
i 3

i † † i † †

=+ d¯ k (k + ωk ∂ki )(ak − a−k ) · (a−k − ak ) + (k − ωk ∂ki )(ak + a−k ) · (a−k + ak )
4
Z Z
i 
† †
 i  
=− 3
d¯ kωk (∂ki ak )ak + (∂ki a−k )a−k = − d¯3 kωk (∂ki ak )a†k − (∂ki a†k )ak
2 2
Z
i  
=− d¯3 kωk a†k (∂ki ak ) + (2π)3 ∂ki δ(k − k0 )|k0 =k − (∂ki a†k )ak
2
Z
i  
=− d¯3 kωk a†k (∂ki ak ) − (∂ki a†k )ak
2

Line 2 follows from identity (7). Line 5 is obtained by noting that all but 2 terms in the integrand of
line 4 either cancel out or are odd. In line 6 we use the identity obtained by taking the k-derivative of
[ak , a†k0 ] = (2π)3 δ (3) (k − k0 ), and evaluating at k0 = k. Finally, to reach line 7 we use that the δ (3) -term
in line 6 is odd.

In a similar way, we can also compute for i 6= j (since M ii = 0 by asymmetry)

Z Z
M = d x(x P (x) − x P (x)) = − d3 x(xi π∂ j φ − xj π∂ i φ)
ij 3 i j j i

Z
1
d¯3 k k j π(k)∂ki φ(−k) − (i ↔ j)


=
2
 
j√
Z
i 3 † 1 †
=− d¯ k k ωk (ak − a−k ) · ∂ki ( √ (a−k + ak )) − (i ↔ j)
4 ωk
i
Z    
i 3 j † k †
=− d¯ k k (ak − a−k ) · − + ∂ki (a−k + ak ) − (i ↔ j)
4 ωk
Z
i  
=− d¯3 kk j ak ∂ki a−k + ak ∂ki a†k − a†−k ∂ki a−k − a†−k ∂ki a†k − (i ↔ j)
4
Z
i  
=− d¯3 kk j ak ∂ki a†k − a†−k ∂ki a−k − (i ↔ j)
4
Z
i  
=− d¯3 kk j −(∂ki ak )a†k − a†k (∂ki ak ) − (i ↔ j)
4
Z
i  
=− d¯3 kk j (2π)3 ∂ki δ(k − k0 )|k0 =k − 2a†k (∂ki ak ) − (i ↔ j)
4
Z
i 3

i † j †

=− d¯ k k ak (∂kj ak ) − k (∂ki ak )ak
2
where we make ample use of integration by parts, drop total-derivative terms, and use i 6= j to completely
ignore having to deal with ∝ ∂ki k j terms.

Physically, the conserved quantities are the center of mass velocities M 0i and angular momenta M ij .

10
Altogether, we may write
Z
i
M µν
=− d¯3 kk µ (a†k ∂kν ak − ∂kν a†k ak ) − (µ → ν), k 0 = ωk
2

11
MIT OpenCourseWare
https://ocw.mit.edu

8.323 Relativistic Quantum Field Theory I

Spring 2023

For information about citing these materials or our Terms of Use, visit: https://ocw.mit.edu/terms.