QFT PS4 Solutions: Free Quantum Field Theory (22/10/18) 1

Solutions 4: Free Quantum Field Theory

1. Heisenberg picture free real scalar field

We have
d3 p
Z
1 −iωp t+ip·x † iωp t−ip·x

φ(t, x) = a p e + ap e (1)
(2π)3 2ωp
p

(i) By taking an explicit hermitian conjugation, we find our result that φ† = φ. You need to note that
all the parameters are real: t, x, p are obviously real by definition and if m2 is real and positive
semi-definite then ωp is real for all values of p. Also (↠)† = â is needed.

(ii) Since π = φ̇ = ∂t φ classically, we try this on the operator (1) to find

d3 p
Z r
ωp −iωp t+ip·x † iωp t−ip·x

π(t, x) = −i ap e − a p e (2)
(2π)3 2
(iii) In this question as in many others it is best to leave the expression in terms of commutators. That
is exploit
[A + B, C + D] = [A, C] + [A, D] + [B, C] + [B, D] (3)
as much as possible. Note that here we do not have a sum over two terms A and B but a sum over
an infinite number, an integral, but the principle is the same.
The second way to simplify notation is to write e−iωp t+ip·x = e−ipx so that we choose p0 = +ωp .
bp = âp e−iωp t+ip·x and †p = â†p e−iωp t+ip·x . You can quickly
The simplest way however is to write A
check these satisfy the same commutation relations as âp and â†p . In fact Abp = eiHt âp e−iHt = âp (t)
for a free field, i.e. we have just applied time evolution to the free field Heisenberg picture operators.
We have the usual commutation relations for the annihilation and creation operators with contin-
uous labels (here the three-momenta) 1
h i h i
ap , a†q = (2π)3 δ 3 (p − q) , [ap , aq ] = a†p , a†q = 0 . (4)
The first field commutation relation is then
d3 p d3 q
Z Z r
1 ωq
[φ(t, x), π(t, y)] = 3
p 3
(−i)
(2π) 2ωp (2π) 2
h i
(ap e−iωp t+ip·x + a†p eiωp t−ip·x ), (aq e−iωq t+iq·y − a†q eiωq t−iq·y ) (5)
d3 p d3 q
Z
ωp
r
= (−i) 3 3
(2π) (2π) 4ωq
h i h i
−e−i(ωp −ωq )t+ip·x−iq·y ap , a†q + ei(ωp −ωq )t−ip·x+iq·y a†p , aq (6)
d3 p d3 q
Z
ωp −i(ωp −ωq )t+ip·x−iq·y
r
= (−i) −e (2π)3 δ 3 (p − q)
(2π)3 (2π)3 4ωq

+ei(ωp −ωq )t−ip·x+iq·y (2π)3 (−δ 3 (p − q)) (7)
d3 p 1 +ip·(x−y)
Z
−ip·(x−y)
= i e + e (8)
(2π)3 2
i 3
= 2δ (x − y) (9)
2
h i
1
Note that this gives the annihilation and creation operators units of energy−3 (in natural units). The usual ai , a†j = iδij
is for the case of discrete labels.
QFT PS4 Solutions: Free Quantum Field Theory (22/10/18) 2
Thus we find the canonical equal-time commutation relation

[φ(t, x), π(t, y)] = iδ 3 (x − y) . (10)

For the commutators h i

φ̂(t, x), φ̂(t, y) = [π̂(t, x), π̂(t, y)] = 0 . (11)
we see that they are of similar form
h i hZ d3 p 1
φ̂(t = 0, x), φ̂(t = 0, y) = (âp e+ip·x + â†p e−ip·x ),
(2π)3 2ωp
p

d3 q
Z
1 +iq·y † −iq·y
i
(â q e + âq e ) (12)
(2π)3 2ωq
p

d3 p
Z r
h ωp
[π̂(t = 0, x), π̂(t = 0, y)] = 3
(−i) (âp e+ip·x − â†p e−ip·x ),
(2π) 2
d3 q
Z r
ωq +iq·y † −iq·y
i
(−i) (âq e − âq e )
(2π)3 2
(13)

Thus all we need to show that

d3 pd3 q h
Z i
+ip·x † −ip·x −iq·y † +iq·y
Cs = (âp e + sâp e ), (âq e + sâ q e ) , (14)
(ωp ωq )s/2
is zero where s = ±1 as the field (momenta) commutator is proportional to C+ (C− ). Using (3)
and (4) we find exactly as before that

d3 pd3 q
Z h i
+ip·x+iq·y †
Cs = s/2
[âp , âq ] e + s âp , âq e+ip·x−iq·y
(ωp ωq )
h i h i
† −ip·x+iq·y † † −ip·x−iq·y
+s âp , âq e + âp , âq e (15)
Z 3 3
d pd q
= sδ(p − q)e+ip·(x−y) − sδ(p − q)e−ip·(x−y) (16)
(ωp )s
d3 p +ip·(x−y)
Z
−ip·(x−y)
= s e − e (17)
(ωp )s

If you now change integration variable in the second term to p0 = −p you will find it is of exactly
the same form as the first term and these cancel giving zero as required.
Aside - Trying to be Careful
This is really rubbish isn’t it! The integrals here look divergent as they scale as energy (=momen-
tum=mass) to the power (3 − s). The oscillatory factor just means we are adding +∞ and −∞ as
we integrate out at high energy scales, at large |p|. The whole thing looks (and is) badly defined.
QFT is not very reasonable about divergences.
Riemann-Lebesgue lemma2 is the real answer here but unlikely to be helpful to Imperial Physics
students who aren’t given the relevant mathematical tools. If we set up the problem more carefully
we could apply this lemma and you’d be happy. In practice I haven’t set up the maths carefully. I
think you will find most QFT text books are the same. We really should define everything carefully
to make sure that such integrals are always defined properly.
2
See https://bit.ly/2KOsAkA.
QFT PS4 Solutions: Free Quantum Field Theory (22/10/18) 3
The quick and dirty way here is to remember that these integrals here are part of an operator
expression coming from commutators even if it is a unit operator. To evaluate these expressions the
operator must act on something. So you should let your expression act on a dummy function f (p).
Now this dummy function has to be of the right type, some well behaved function. Basically that
has to be something that falls off for large |p| “nicely”. The functions needed will always respect
our space-time symmetries, so here will always be functions of |p|. A suitable example might be
something that falls off as exp(−p.p/(2σ 2 )) where you can choose sigma to be as big as you like
(take it to infinity only after everything else is done). Now your integrals are of the right form for
Riemann-Lebesgue lemma to apply. Alternatively, as each integral is now well behaved and finite,
you can start to manipulate them. In our case we would need f (p) = f (−p) for our odd/even
arguments to work but the space-time symmetries guarantee that any function we have in practice
will have that symmetry.
To be more precise we should set everything up carefully. This is what the Axiomatic QFT3 is very
careful about.
p
(iv) Given ωp = p2 + m2 then ωp = ω−p and hence

d3 p
Z Z
3 iq·x 1 ip·x † −ip·x

d x φ(t = 0, x)e = d3 x ap e + a p e eiq·x (18)
(2π)3 2ωp
p

d3 p
Z
1 i(p+q)·x † i(q−p)·x

= d3 x ap e + a p e (19)
(2π)3 2ωp
p

d3 p
Z
1 3 (3) † 3 (3)

= a p (2π) δ (q + p) + a p (2π) δ (q − p) (20)
(2π)3 2ωp
p

1
= p a−q + a†q (21)
2ωq

d3 p
Z Z r
3 iq·x 3 ωp ip·x † −ip·x

d x π(t = 0, x)e = −i d x a p e − a p e eiq·x (22)
(2π)3 2
d3 p
Z r
3 ωp i(p+q)·x † i(q−p)·x

= −i d x a p e − a p e (23)
(2π)3 2
d3 p
Z r
ωp
3 (3) † 3 (3)

= −i a p (2π) δ (q + p) − a p (2π) δ (q − p) (24)
(2π)3 2
r
ωq
= −i a−q − a†q (25)
2

Thus solving for a−q and a†q gives

Z r !
ωq 1
aq = d3 x e−iq·x φ(0, x) + i p π(0, x) (26)
2 2ωq
Z r !
ωq 1
a†q = d3 x eiq·x φ(0, x) − i p π(0, x) (27)
2 2ωq
(28)
3
See https://bit.ly/2wrg1ID.
QFT PS4 Solutions: Free Quantum Field Theory (22/10/18) 4
Hence since [φ(x), φ(y)] = [π(x), π(y)] = 0
h i Z
†
ap , aq = d3 x e−ip·x d3 y eiq·y (29)
"r r #
ωp 1 ωq 1
× φ(0, x) + i p π(0, x), φ(0, y) − i p π(0, y) (30)
2 2ωp 2 2ωq
Z r
ωp ωq
r
3 3 −ip·x+iq·y
= d xd ye −i [φ(0, x), π(0, y)] + i [π(0, x), φ(0, y)] (31)
4ωq 4ωp
Z r
ωp (3) ωq (3)
r
3 3 −ip·x+iq·y
= d xd ye δ (x − y) + δ (y − x) (32)
4ωq 4ωp
Z r
ωp ωq
r
3 i(q−p)·x
= d xe + (33)
4ωq 4ωp
= (2π)3 δ (3) (p − q) (34)

(v) Taking the derivative we have

d3 p ip ip·x
Z
† −ip·x
∇φ(x) = ap e − a p e (35)
(2π)3 2ωp
p

so4
Z
P = − d3 x π(x)∇φ(x) (36)

d3 p d3 q
Z r
3 ωp ip·x † −ip·x
q
iq·x † −iq·x

= − d x a p e − a p e aq e − aq e (37)
(2π)3 (2π)3
p
2 2ωq
√
d3 p d3 q ωp q
Z
= − d3 x 3 3 √ (38)
(2π) (2π) 2 ωq
n o
× ei(p+q)·x ap aq − ei(p−q)·x ap a†q − e−i(p−q)·x a†p aq + e−i(p+q)·x a†p a†q (39)
√
d3 p d3 q ωp q
Z
= − 3 3 √ (40)
(2π) (2π) 2 ωq
n o
× (2π)3 δ (3) (p + q) ap aq + a†p a†q − (2π)3 δ (3) (p − q) ap a†q + a†p aq (41)
d3 p p
Z
† † † †
= − −a a
p −p − a a
p −p − a a
p p − a a
p p (42)
(2π)3 2

Now p ap a−p + a†p a†−p is an odd function under p → −p and so integrates to zero. Thus

d3 p p
Z
† †
p = a p ap + a p ap (43)
(2π)3 2
d3 p p †
Z
3 (3)
= 2a a
p p + (2π) δ (0) (44)
(2π)3 2
d3 p
Z
= p a†p ap (45)
(2π)3

since again pδ (3) (0) is an odd function (albeit it poorly defined!).

4
NOTE THIS IS FOR t = 0 should add in more exponentials for non-tirvial times!
QFT PS4 Solutions: Free Quantum Field Theory (22/10/18) 5

The interpretation is that a†p ap d3 p is the operator giving the number of quanta in a small volume
d3 p centred at momentum p. This will indeed contribute p to the total momenta. We see the same
type of term for the energy, the Hamiltonian operator H, but we get a factor of ωp not p in that
case.

(vi) From (2) we have

d3 p
Z Z Z r
3 2 3 ωp −iωp t+ip·x † iωp t−ip·x

d xΠ = d x −i a p e − ap e
(2π)3 2
d3 q
Z r
ωq −iωq t+iq·x † iωq t−iq·x

.−i a q e − aq e (46)
(2π)3 2
d3 p d3 q
Z Z Z r
ωp ωq −iωp t+ip·x † iωp t−ip·x

= − d3 x a p e − ap e
(2π)3 (2π)3 4

−iωq t+iq·x † iωq t−iq·x
aq e − aq e (47)

Using that
Z
d3 xeip·x = (2π)3 δ(p) (48)

d3 x and use the resulting delta function of δ 3 (p ± q) to eliminate the q integral.

R
we apply the
This gives us

d3 p ωp
Z Z
3 2 −2iωp t † † † † +2iωp t
d xΠ = − a p a −p e − ap a p − ap a p + a a
p −p e (49)
(2π)3 2

From (2) we have

d3 p
Z Z Z
3 2 3 1 −iωp t+ip·x † iωp t−ip·x

d x (∇φ) = d x ipa p e − ipa p e
(2π)3 2ωp
p

d3 q
Z
1 −iωq t+iq·x † iωq t−iq·x

. iqa q e − iqa q e (50)
(2π)3 2ωq
p

d3 p d3 q
Z Z Z
3 p.q −iωp t+ip·x † iωp t−ip·x

= d x ia p e − iap e
(2π)3 (2π)3 4ωp ωq
p

. iaq e−iωq t+iq·x − ia†q eiωq t−iq·x (51)

Again we apply the d3 x and use the resulting delta function of δ 3 (p±q) to eliminate the q integral
R

to find
d3 p p2
Z Z
3 2 −2iωp t † † † † 2iωp t
d x (∇φ) = ap a −p e + ap a p + ap a p + a a
p −p e (52)
(2π)3 2ωp

Finally, in the same manner we find that

d3 p 1
Z Z
3 2 −2iωp t † † † † 2iωp t
d x (φ) = a a
p −p e + a a
p p + a a
p p + a a
p −p e (53)
(2π)3 2ωp
QFT PS4 Solutions: Free Quantum Field Theory (22/10/18) 6
Putting this together we find
Z
1
H = d3 x Π2 + (∇φ)2 + m2 φ2 (54)
2
d3 p ωp2 + p2 + m2
Z
1 † †

= a a
p p + a a
p p
2 (2π)3 2ωp
d3 p −ωp2 + p2 + m2
Z
1 −2iωp t † † 2iωp t

a a
p −p e + a a
p −p e (55)
2 (2π)3 2ωp
d3 p d3 p
Z Z
1
† †

† 1 3 3
= ωp ap ap + ap ap = ωp ap ap + (2π) δ (0) (56)
2 (2π)3 (2π)3 2
p
with ωp = + p2 + m2

(vii)

d3 p
Z
1
[H, a†k ak ] = †
ωp ap ap + †
, ak ak (57)
(2π)3 2
d3 p
Z
† †
= ωp ap ap , ak ak (58)
(2π)3
d3 p
Z h i
† †
= ωp ap p k k =0
a , a a (59)
(2π)3

Note that this result is not true for any interesting i.e. interacting theory. The interactions mix the
modes of different momenta leading to a lack of conservation of particle number. Only a continuous
symmetry can guarantee conserved numbers and those are usually linked to total numbers of various
particles, not the individual quanta of one particle at one momentum.

2. Time evolution of annihilation operator

(i) From (56) and using the commutation relation

h i h i
ap , a†q = (2π)3 δ(p − q) , [ap , aq ] = a†p , a†q = 0 , (60)

we have that
d3 q
Z
† 1
[H, ap ] = ωq aq aq + , ap (61)
(2π)3 2
Z 3
d q
† †
= ωq a q a q ap − a p aq aq (62)
(2π)3
d3 q
Z
† † 3
= ωq a a
q q p a − (a a
q q pa + (2π) δ(p − q)a−p ) (63)
(2π)3
Z
= − d3 qωq δ(p − q))ap = −ωp ap (64)

as required.

(ii) We want to prove that

(itH)n ap = ap [it(H − ωp )]n (65)
QFT PS4 Solutions: Free Quantum Field Theory (22/10/18) 7
using [H, ap ] = −ωp ap . First note that trivially the relation holds for n = 1. Now if we assume the
relation holds for n = m then

(itH)m+1 ap = (itH)(itH)m ap (66)

= itHap [it(H − ωp )]m (67)
m
= it(ap H − ωp ap ) [it(H − ωp )] (68)
= ap [it(H − ωp )]m+1 . (69)

So the relation holds for n = m + 1 if it holds for n = m. Thus, by induction, it holds for all n ≥ 0.

(iii) Start by expanding the first exponential

∞
!
−iHt
X 1
e iHt
ap e = (iHt)n ap e−iHt (70)
n!
n=0
!
X 1
= (iHt)n ap e−iHt (71)
n
n!
!
X 1
= ap [it(H − ωp )]n e−iHt (72)
n
n!
!
X 1
= ap [it(H − ωp )]n e−iHt (73)
n
n!
= ap eit(H−ωp ) e−iHt (74)
it(H−ωp −H)
= ap e (75)
(76)

Thus we see that5

eiHt ap e−iHt = ap e−itωp (77)
where we have used the obvious facts that [H, H] = 0 and [ωp , H] = 0.
The corresponding equation for a†p follows from hermitian conjugation
† †
eiHt ap e−iHt = ap e−itωp (78)
−iHt † iHt †
†
(ap )† e e−itωp (ap )†

⇒ e = (79)
⇒ e+iHt a†p e−iHt = e+itωp a†p (80)
⇒ eiHt a†p e−iHt = a†p e+iωp t . (81)

3. Delta Functions

The Dirac delta function is defined through

Z +∞
dx δ(x − x0 )f (x) = f (x0 ) . (82)
−∞

You should always start from this equation when using a delta function.
5
Another neat way to prove this by taking the derivative of the equation with respect to t and solving the resulting
operator valued differential equation.
QFT PS4 Solutions: Free Quantum Field Theory (22/10/18) 8
(i) Consider Z
J= dy δ(g(y))f (y) . (83)

Assume that the zero’s of g are at Z = {y0 |g(y0 ) = 0} and are widely spaced. Then the only places
where the integral (83) has a non-zero contribution is in the region of one of these zeros as we need
the argument of the delta function to be zero from (82). So we can write J as
X Z y0 +
J= dy δ(g(y))f (y) . (84)
y0 ∈Z y0 −

So consider one of these zeros, say y0 , and expand the function g around this zero to find that
g(y) = g(y0 ) + (y − y0 )g 0 (y0 ) + O((y − y0 )2 ). By definition G(y0 ) = 0 so we have for small that
X Z y0 +
J= dy δ((y − y0 )g 0 (y0 ))f (y) . (85)
y0 ∈Z y0 −

Now change variable to x = (y − y0 )g 0 (y0 ) to match the form given in the definition of the delta
function (82). The change of variables is gives us
X Z η dx
J= 0 (y )
δ(x)f (y) , η = g 0 (y0 ) . (86)
−η g 0
y0 ∈Z

Now in order to apply the formula (82) we note that we must be running from below the zero of
the argument of the delta function, from below x0 in (82), to above it. For the case of g 0 (y0 ) > 0
there is no problem as η is positive and we have the same form as (82). The range of integration
can be extended to infinity without any problem. When g 0 (y0 ) < 0 however, η is negative and we
are running past the zero in the wrong direction. However easy to switch direction but we get an
overall minus sign in this case. This then cancels the negative sign of g 0 (y0 ) in the denominator.
Thus we have that
X Z η0 dx
J= 0 (y |)
δ(x)f (y) , η 0 = |g 0 (y0 )| . (87)
−η 0 |g 0
y0 ∈Z

Now we apply (82) to find that

Z X f (y0 )
J= dy δ(g(y))f (y) = , (88)
|g 0 (y0 )|
y0 ∈Z

(ii) By inspection
d4 k
Z
I= δ(k 2 − m2 )f (k 2 , p2 , (k − p)2 ) (89)
(2π)4
must be Lorentz invariant if k and p are four vectors as the expression is made up of Lorentz scalars.
The arguments k 2 , p2 , (k −p)2 and m2 are all Lorentz scalars. The measure, d4 k is Lorentz invariant
because any boost to new variables k 0µ = Λµν k ν produces a Jacobian factor in the transformation
|Λ| but this is 1 by definition of the Lorentz transformations. Thus all elements of this integral are
invariant.
There is only one variable p in this problem so we can only be a function of the only remaining
variable p (and of course it can depend on m or other constant parameters). The only invariant we
can build out of this is p2 so the result must be a function of p2 .
QFT PS4 Solutions: Free Quantum Field Theory (22/10/18) 9
The result
X Z d3 k p
I= f (m2 , p2 , (k − p)2 ) , ω = | k2 − m2 | . (90)
2ω
k0 =±ω

follows from (88). Consider the k0 integration where we have that g(k0 ) = (k0 )2 − ω 2 . This has two
zeros at k0 = ±ω. We find that g 0 (k0 ) = 2k0 so at the zeros we have g 0 (k0 ) = 2|k0 | = 2ω.
Since I is Lorentz invariant as a whole, and all the other terms in the form (90) are Lorentz invariant,
we deduce that d3 k/(2ω) is also Lorentz invariant. You could also prove this directly by changing
variables to a boosted frame k 0µ = Λµν k ν .

(iii) Consider the integral

Z +∞
i i
K= dp0 f (p0 ) − (91)
−∞ p0 − ω + i p0 − ω − i
The poles and contours used in the two terms are shown in figure 1.
For the first term we can distort the integration path so that near the pole at p0 − ω + i we run
along a semi circle centred on ω of radius η > 0 running above the axis (positive imaginary part),
see figure 2. We will assume η is infinitesimal so that we do not encounter any poles in the function
f . We run along the real axis for the rest of the path. We can now take the limit → 0 to place
the pole on the real axis as for η > 0 the path does not run through the pole.
Z
i
K+ = dp0 f (p0 ) (92)
C+ p0 − ω

For the second term we move the contour of integration so near its pole at p0 − ω − i so that now
this path runs along the real axis except for a semi-circle centred on ω of radius η > 0 but this time
the semi-circle is below the axis (negative imaginary part), see figure 2.
Z
i
K− = − dp0 f (p0 ) (93)
C− p0 − ω

If we reverse the direction of integration in this second case we absorb the overall minus sign. Since
the whole result is K = K+ + K− when we put the two together, the contributions coming from
two integrations along the real axes now cancel. The only part remaining is an integration around
a small circle of radius η centred on the pole at k0 = ω, see figure 3. Note we are running round
this pole in a negative sense so that the residue there tells us this integral is equal to −2πi times
the residue. The residue is simply if (p0 = ω) with the the factor of i coming from the numerator.
Thus we find that we have
K = 2πf (p0 = ω) (94)
However using the definition of the Dirac delta function (82) we also have that this may be written
as Z +∞
K = 2π dp0 f (p0 )δ(p0 − ω) (95)
−∞
Now comparing (91) and (95) we can identify that
i i
2πδ(p0 − ω) = − (96)
p0 − ω + i p0 − ω − i
QFT PS4 Solutions: Free Quantum Field Theory (22/10/18) 10

w-ie w-ie

Figure 1: Figure showing two contours used for two terms in the delta function representation K. Here
the poles are placed off the real axis and contours run along the real line.

C+ C+
w w

C- -C-
Figure 2: Figure showing two contours used for two terms in the delta function representation, K+ of
(92) uses C+ shown on the left while on the right is C− used by K− of (93). The poles of the integrand
are now on the real axis but the contours follow semicircles above or below the real line to avoid the the
poles.

Figure 3: Absorbing the minus sign of the second term of (91) by reversing the direction of the countour
C− , the only non-zero contribution now comes from a small circle around the pole running in the negative
direction..
QFT PS4 Solutions: Free Quantum Field Theory (22/10/18) 11
4. The Advanced Propagator

We have
DA (x) = −θ(−x0 ) h0|[φ(x), φ(0)]|0i (97)

(i) We start by evaluating DA . We have

d3 p d3 q
Z
1 h
†
i
−ip·x
h
†
i
ip·x
[φ(x), φ(0)] = a ,
p q a e + a ,
p q a e (98)
(2π)3 (2π)3 4ωp ωq
p 0
p =ωp
Z 3
d p d q3 1
(2π)3 δ (3) (p − q) e−ip·x − eip·x p0 =ωp

= 3 3
p (99)
(2π) (2π) 4ωp ωq
d3 p 1
Z
−ip·x ip·x

= e − e p0 =ωp
(100)
(2π)3 2ωp

Hence since h0|0i = 1 we have6

DA (x) = −θ(−x0 ) h0|[φ(x), φ(0)]|0i (101)

Z 3
d p 1
= −θ(−x0 ) e−ip·x − eip·x p0 =ωp

(102)
(2π)3 2ωp
d3 p 1 −iωp x0 ip·x
Z
iωp x0 −ip·x
= −θ(−x0 ) e e − e e (103)
(2π)3 2ωp
d3 p 1 −iωp x0 ip·x
Z
iωp x0 ip·x
= −θ(−x0 ) e e − e e p ↔ −p in second term(104)
(2π)3 2ωp
!
3p
Z
d 1 1
= −θ(−x0 ) e−ip·x e−ip·x

+ (105)
(2π)3 2ωp p0 =ωp (−2ωp ) p0 =−ωp

We now need to introduce a p0 integration and rewrite the expression in terms of a contour inte-
gration. There are two standard ways to do this. In the first approach we shift the poles of the
integrand, introducing a small positive infinitesimal into the integrand which is taken to zero
(from the positive side) at the end of the calculation. This is the approach used in the lectures and
it is common practice to use this notation, especially in the case of the time-ordered (Feynman)
propagator. The second approach is to make small distortions in the contour away from the real
p0 axis near the poles. This is used by Tong in his derivation of the Feynman propagator (sec.2.7.1
page 38) though Tong reverts to the first and standard notation later on (see Tong equation (3.37)).
Both methods are equivalent in the → 0+ limit.
First approach
Consider
d4 p
Z
i
I1 (x) = e−ip·x (106)
(2π) (p0 − i) − p2 − m2
4 2
Z ∞
d3 p dp0
Z
1
= − e−ip·x (107)
(2π) −∞ 2πi (p0 − ωp − i)(p0 + ωp − i)
3

6
R +∞
For the p ↔ −p in second term don’t forget that in changing variable in each momentum component −∞
dpi means
the range of integration changes from −∞ to +∞ to the other way round giving another minus sign.
QFT PS4 Solutions: Free Quantum Field Theory (22/10/18) 12
p
where, as usual, ωp = p2 + m2 . The p0 integration is along the real axis with poles in the
integrand as shown here

-wp+ie +wp+ie

(108)

The dp0 integrand

1
f (p0 , p) = e−ip·x (109)
(p0 − ωp − i)(p0 + ωp − i)
has simple poles at
p0 = ±ωp + i. (110)
Near these poles the integrand looks like f ≈ R± /(p0 ∓ ωp + i) with residues R± given by

1 −ip·x
R± = ± e p0 =±ωp +i
(111)
2ωp

The idea is that we think of our expression for the advanced propagator in (105) as being of the
form
d3 p
Z
0
DA (x) = −θ(−x ) (R+ + R− ) . (112)
(2π)3
In order for this to match I1 (x) of (107) we need to find a closed contour C such that by using the
residue theorem we can deduce that
dp0
Z
f (p0 , p) = θ(−x0 ) R+ + R−

(113)
C 2πi
0 0
If x0 > 0 then e−ip x → 0 as =(p0 ) = −i∞. This means that an integration of this integrand f
round a large semi-circle running around the lower half plane is equal to zero

dp0
Z
f (p0 , p) = 0 ifx0 > 0 as =(p0 ) → −i∞ . (114)
C− 2πi

We can therefore add this integration of f around the C− semi-circle to our p0 integration along
the real axis in I1 without changing the result for I1 . So we produce an expression for I1 which uses
a closed contour for the p0 integration by adding this lower semi-circle. Now no poles are enclosed
QFT PS4 Solutions: Free Quantum Field Theory (22/10/18) 13
within this closed contour so the residue theorem tells us the result is zero

C+
-wp+ie +wp+ie
dp0
Z
f (p0 , p) = 0 if x0 > 0 . (115)
C 2πi

C-

0 0
If x0 < 0 then e−ip x → 0 as =(p0 ) = +i∞. This means that an integration of this integrand
f round a large semi-circle running around the upper half plane will give zero. We can therefore
add this to our existing p0 integration along the real axis in I1 without changing the result. So we
produce a closed contour by adding the semi-circle above and now the residue theorem tells us that
we pick up contributions from both poles. This gives us

C+
-wp+ie +wp+ie
dp0
Z
f (p0 , p) = R+ + R− if x0 < 0 . (116)
C 2πi
C-

Putting the two cases together gives us the desired result

dp0
Z
f (p0 , p) = θ(−x0 ) R+ + R− .

(117)
C 2πi

Alternative approach
The second approach to these types of problem is to distort the contour away from the real p0 axis
near the poles by a small amount. So now consider
d3 p dp0
Z Z
i
I(x) = e−ip·x (118)
(2π) C 2π p − m2
3 2

d3 p dp0
Z Z
1
= − e−ip·x (119)
(2π) C 2πi (p − ωp )(p0 + ωp )
3 0

p
where, as usual, ωp = p2 + m2 . The dp0 integrand
1
f (p0 , p) = e−ip·x (120)
(p0 − ωp )(p0 + ωp )
QFT PS4 Solutions: Free Quantum Field Theory (22/10/18) 14
has simple poles at
p0 = ±ωp (121)
with residues f ≈ R± /(p0 ∓ ωp ) near p0 = ±ωp given by

1 −ip·x
R± = ± e p0 =±ωp
(122)
2ωp

Observing that we can rewrite

d3 p
Z
0
DA (x) = −θ(−x ) (R+ + R− ) (123)
(2π)3

in order for this to match I(x) we need to find a contour C such that

dp0
Z
f (p0 , p) = θ(−x0 ) R+ + R−

(124)
C 2πi

Consider the following contour

−Ep Ep

C
(125)
0 x0
If x0 > 0 then e−ip → 0 as =(p0 ) = −i∞ and we close the contour below and pick up no poles

−Ep Ep

C
dp0
Z
f (p0 , p) = 0 (126)
C 2πi

0 x0
If x0 < 0 then e−ip → 0 as =(p0 ) = +i∞ and we close the contour below and picking up both
poles

dp0
Z
f (p0 , p) = R+ + R− (127)
C 2πi
−Ep Ep

so that indeed
dp0
Z
f (p0 , p) = θ(−x0 ) R+ + R−

(128)
C 2πi
QFT PS4 Solutions: Free Quantum Field Theory (22/10/18) 15
To give a bit more detail on how we close the contour, consider x0 > 0. We have (remember p0 is
complex)
0 0 0 0
eIm p x eIm p x

1 −ip0 x0

(p0 )2 − ω 2 e = (p0 )2 − ω 2 ≤ |p0 |2 − ω 2 (129)

p p p

where the last step comes from writing |p0 |2 = |(p0 )2 − ωp2 + ωp2 | ≤ |(p0 )2 − ωp2 | + |ωp2 | using the
0 0
triangle inequality. Closing the integral below we have Im p0 ≤ 0 so that eIm p x ≤ 1. Thus,
evaluating first at finite |p0 | we have, for the infinite semi-circular path C− on the lower half plane

C-

(130)

we have
Z Z
0 1 −ip0 x0
0
1 −ip0 x0

dp 0 2 2
e ≤ dp 0 2 2
e (131)

C− (p ) − ωp
C − (p ) − ω p

0 0
eIm p x
Z
0
≤ dp 0 2 (132)
C− |p | − ωp2
Z
1
≤ dp0 0 2 (133)
C − |p | − ωp2
π|p0 |
= (integrate on C − with dp0 = |p0 |dθ) (134)
|p0 |2 − ωp2
→ 0 as p0 → ∞ (135)

Hence we see that the contribution from the integration along C − is zero and we can close the
contour C along C − in order to evaluate I(x) when x0 > 0. A similar argument holds for x0 < 0.

(ii) We have, taking the derivative of e−ip·x

d4 p
Z
i
2 2
−p2 + m2 e−ip·x

∂ +m DA (x) = 4 2 2
(136)
C (2π) p − m
d4 p −ip·x
Z
= − 4
ie (137)
C (2π)
= −iδ (4) (x) (138)

(Note the integrand has no poles on the real axis, so none of the subtleties in the path C relevant
when defining DA (x) appear.)
QFT PS4 Solutions: Free Quantum Field Theory (22/10/18) 16
∗5. Time evolution and propagators of a complex scalar field

(i) Consider two sets of annihilation and creation operators â†1 , â†2 , a1 and a2 obeying
h i h i
âip , â†jq = (2π)3 δ 3 ((p − q))δij , [âip , âjq ] = â†ip , â†jq = 0 , i, j = 1, 2 . (139)

We define
1 1
b̂p = √ (â1p + iâ2p ) , ĉp = √ (â1p − iâ2p ) . (140)
2 2
so that
1 1
b̂†p = √ â†1p − iâ†2p , ĉ†p = √ â†1p + iâ†2p . (141)
2 2
We require several commutators to be proved. The key identity here is that for any operators
A, B, C, D we have
[A + B, C + D] = [A, C] + [A, D] + [B, C] + [B, D] (142)
which you should prove if this is not familiar.
The commutators between a b or c annihilation and a b̂ or ĉ creation operator are all of the form
1h i
(â1p + isp â2p ) , â†1q + isq â†2q (143)
2
where sp , sq = ±1. Expanding we have that

1h i
(â1p + isp â2p ) , â†1q + isq â†2q (144)
2
1h i 1h i 1h i 1h i
= â1p , â†1q + isq â1p , â†2q + isp â2p , â†1q − sp sq â2p , â†2q (145)
2 2 2 2
1 3 1 3
= δ (p − q) + 0 + 0 − sp sq δ (p − q) (146)
2 2
1 3
= (1 − sp sq )δ (p − q) . (147)
2
So we find

[b̂p , ĉ†q ] = 0 sp = +1, sq = +1 (148)

[b̂p , b̂†q ] 3
= δ (p − q) sp = +1, sq = −1 (149)
[ĉp , ĉ†q ] = δ 3 (p − q) sp = −1, sq = +1 (150)
[ĉp , b̂†q ] = 0 sp = −1, sq = −1 (151)

Note the last is the hermitian conjugate of the first so you could avoid calculating it for that reason.
The commutators between a pair of b or c annihilation operators are all of the form
1
[(â1p + isp â2p ) , (â1q + isq â2q )] (152)
2
where sp , sq = ±1. Since annihilation operators commute with each other even if they are of the
same type and at the same momentum, all these commutators are clearly zero. Taking the hermitian
conjugate, (or using a similar argument for the creation operators), we see that all the commutators
between a pair of b̂† or ĉ† creation operators are also zero.
QFT PS4 Solutions: Free Quantum Field Theory (22/10/18) 17

(ii) Given the question below for Q we can save time by considering b̂†p b̂p + sĉ†p ĉp where s = ±1. We
then have that
1 †
b̂†p b̂p + sĉ†p ĉp = â1p − iâ†2p (â1p + iâ2p )
2
s
+ â†1p + iâ†2p (â1p − iâ2p ) (153)
2
1 †

= â â1p + iâ†1p â2p − iâ†2p â1p + â†2p â2p
2 1p
s
+ â†1p â1p − iâ†1p â2p + iâ†2p â1p + â†2p â2p (154)
2
1
= (1 + s) â†1p â1p + â†2p â2p + i(1 − s) â†1p â2p − â†2p â1p (155)
2
When s = +1 we have the desired result that

b̂†p b̂p + ĉ†p ĉp = â†1p â1p + â†2p â2p . (156)

From this we see that (here Zp is the zero point energy for mode p)
XZ
H =
b d3 p ωp â†ip âip + Zp (157)
i=1,2
Z
= d3 p ωp b̂†p b̂p + ĉ†p ĉp + 2Zp (158)

(iii) With the Hamiltonian given by (158), to show that

d3 p
Z
1
Φ̂H (t, x) = 3
p (b̂p e−ipx + ĉ†p e+ipx ) (159)
(2π) 2ωp
given the form at t − 0 (where px ≡ pµ xµ and p0 = +ωp ) we can use the same method as used for
the real field. That is since like all operators7

Φ̂H (t, x) = exp{iHt}

b Φ̂H (t = 0, x) exp{−iHt}
b (160)

all we need to do is look at the behaviour of

exp{iHt}
b b̂p exp{−iHt}
b and b †p exp{−iHt}
exp{iHt}ĉ b . (161)

As the c and b operators commute, the only part that matter that matters is the b̂† b̂ term in the
first case and the ĉ† c term in the second case. More formally we can commute one part of the
Hamitonian through the b̂p or the ĉ†p operator to be cancelled. The problem reduces to that of a
single type of operator, i.e. we are left with

exp{iH
b b t}b̂p exp{−iH
b b t} and b c t}ĉ† exp{−iH
exp{iH b c t} , (162)
p

where Z Z
H
bb = 3
d p ωp b̂†p b̂p , H
bc = d3 p ωp ĉ†p ĉp . (163)

We have already derived these so from (77) we know that

exp{iH
b b t}b̂p exp{−iH
b b t} = exp{−iωp t} exp{iH b c t}ĉ† exp{−iH
and b c t} = exp{+iωp t} .
p
(164)
and we find the desired time evolution for the free complex scalar field in the Heisenberg picture.
7
As this is a free Hamiltonian it is the same in Heisenberg and Scroödinger pictures and so it needs no subscript.
QFT PS4 Solutions: Free Quantum Field Theory (22/10/18) 18
(iv) (No answer for this part so far).
Consider two real8 scalar fields
d3 p
Z
1 −iωp t+ip·x † iωp t−ip·x

φ̂Hi (t, x) = âip e + â ip e (165)
(2π)3 2ωp
p

where âip and â†ip are the operators given in (139). Show that if (??) is true then the field operators
obey
1 1
Φ̂H (t, x) = √ φ̂H1 (t, x) + iφ̂H2 (t, x) , Φ̂†H (t, x) = √ φ̂H1 (t, x) − iφ̂H2 (t, x) . (166)
2 2

(v) The Wightman function for a complex scalar field is defined as

D(x − y) := h0|Φ̂(x)Φ̂† (y)| 0i . (167)

Substituting (159) into (167) give us

d3 p d3 q
Z
1 1
D(x − y) = h0|(b̂p e−ipx + ĉ†p e+ipx )(b̂†q e+iqy + ĉq e−iqy )| 0i (168)
(2π)3 2ωp (2π)3 2ωq
p p

d3 p d3 q
Z
1 1
= h0|b̂p b̂†q | 0ie−ipx+iqy (169)
(2π)3 2ωp (2π)3 2ωq
p p

as the c (ĉ† ) operator annihilates the ket (bra) vacuum. Commuting the two b operators gives a
non-zero term containing a delta function in momentum (which will kill one integral) plus a second
term b̂†q b̂p which again annihilates the vacuum states so does not contribute. This leaves us with

d3 p 1 −ip(x−y)
Z
D(x − y) = e (170)
(2π)3 2ωp

exactly we found before for real scalar fields.

Note that from the definition we have that for any type of field

(D(x − y))∗ := h0|Φ̂(y)Φ̂† (x)| 0i = D(y − x) . (171)

For the free field case in (170) we can see this explicitly.

(vi) For free complex scalar fields by substituting (159) into h0|Φ̂(x)Φ̂(y)| 0i gives us

d3 p d3 q
Z
1 1
h0|Φ̂(x)Φ̂(y)| 0i = h0|(b̂p e−ipx + ĉ†p e+ipx )(b̂q e−iqy + ĉ†q e+iqy )| 0i
(172)
(2π)3 2ωp (2π)3 2ωq
p p

d3 p d3 q
Z
1 1
= 3
p 3
p h0|b̂p ĉ†q | 0ie−ipx+iqy (173)
(2π) 2ωp (2π) 2ωq

as the b (ĉ† ) operator annihilates the ket (bra) vacuum. However the two remaining operators
commute and again they then annihilate the vacuum so this case is zero for any time.
8
These represent classical real scalar fields but their quantised versions are hermitian not real.
QFT PS4 Solutions: Free Quantum Field Theory (22/10/18) 19
(vii) The Feynman propagator for complex scalar fields is defined as

∆F (x − y) := h0|TΦ̂(x)Φ̂† (y)| 0i (174)

where T is the time ordering operator. We can express ∆F in terms of the Wightman functions for
the complex field (167)

∆F (x − y) = θ(x0 − y0 )h0|Φ̂(x)Φ̂† (y)| 0i + θ(y0 − x0 )h0|Φ̂† (y)Φ̂(x)| 0i (175)

Following the same steps as above we can show that

d3 p d3 q
Z
† 1 1
h0|Φ̂ (y)Φ̂(x)| 0i = 3
p 3
p h0|(b̂†q e+iqy + ĉq e−iqy )(b̂p e−ipx + ĉ†p e+ipx )| (176)
0i
(2π) 2ωp (2π) 2ωq
d3 p d3 q
Z
1 1
= 3
p 3
p h0|ĉq ĉ†p | 0ie−iqy+ipx (177)
(2π) 2ωp (2π) 2ωq
d3 p d3 q
Z
1 1
= 3
p 3
p (2π)3 δ 3 (q − p)e−iqy+ipx (178)
(2π) 2ωp (2π) 2ωq
d3 p 1 −ip(y−x)
Z
= e (179)
(2π)3 2ωp

By comparing with (170) we see that

h0|Φ̂† (y)Φ̂(x)| 0i = D(y − x) . (180)

This means that the Feynman propagator for complex scalar fields is

∆F (x − y) = h0|TΦ̂(x)Φ̂† (y)| 0i = θ(x0 − y0 )D(x − y) + θ(y0 − x0 )D(y − x) . (181)

This is exactly the same form as we had for the real free field scalar propagator so the Feynman
propagator is of exactly the same form i.e. in momentum space it is equal to
i
∆F (p) = . (182)
p2 − m2 + i
Clearly these are also solutions of the Klein-Gordon equation as they were in the real free scalar
field case. Again the poles at p2 = m2 indicate that we have particle-like solutions of mass m
dominating the behaviour. The presences of two distinct types of annihliation/creation operator in
this complex field indicates there are two independent degrees of freedom with the same mass, here
distinct particle and anti-particles.
Note that h0|TΦ̂† (y)Φ̂(x)| 0i = h0|TΦ̂(x)Φ̂† (y)| 0i if the times are unequal because the T operator
defines the order so how we write the fields under the T operator has no effect.
QFT PS4 Solutions: Free Quantum Field Theory (22/10/18) 20
6. Charge of a complex scalar field

We have by definition Z
Q=i d3 x Φ† Π† − ΠΦ (183)

(i) We have

d3 p d3 q
Z Z
ωp ip·x
r
† −ip·x † −iq·x
i d3 x Π(x)Φ(x) = d3 x cp e − bp e bq eiq·x
+ cq e (184)
(2π)3 (2π)3 4ωq
√
d3 p d3 q ωp
Z
3
= d x
(2π)3 (2π)3 4ωq
p
n o
× ei(p+q)·x cp bq + ei(p−q)·x cp c†q − e−i(p−q)·x b†p bq − e−i(p+q)·x b†p c†q (185)
√
d3 p d3 q ωp
Z
= 3 3
p
(2π) (2π) 4ωq
n o
× (2π)3 δ (3) (p + q) cp bq − c†p b†q + (2π)3 δ (3) (p − q) cp c†q − b†p bq (186)
d3 p
Z
1 † † † †

= cp b−p − c p b−p + cp cp − bp bp (187)
2 (2π)3
Now
Z Z †
3 † † 3
i d x Φ (x)Π (x) = − i d x Π(x)Φ(x) (188)

d3 p
Z
1 † † † †

= c−p bp − c b
−p p − cp cp + bp bp (189)
2 (2π)3
and hence
d3 p
Z
1 † † † † † †

Q = c −p bp − cp b−p + c b
p −p − c b
−p p + 2bp bp − 2cp cp (190)
2 (2π)3
d3 p †
Z
†
= bp bp − cp cp (191)
(2π)3
d3 p †
Z
† 3 (3)
= b bp − c c
p p − (2π) δ (0) (192)
(2π)3 p
d3 p †
Z
†
= b bp − c cp + infinite constant (193)
(2π)3 p p

where we have used the fact that odd functions integrate to zero.

(ii) Substitute in the form of the fields

d3 p
Z
1
Φ(x) = (bp eip·x + c†p e−ip·x ) (194)
(2Π)3 2ωp
p

d3 p
Z r
ωp
Π(x) = 3
(−i) (cp eip·x − b†p e−ip·x ) (195)
(2Π) 2

and hence confirm that the ETCR (equal time commutation relations) for complex scalar fields are
[Tong (2.71), p.34]
h i
[Φ(x), Π(y)]x0 =y0 = (2π)3 δ 3 (x − y) , Φ† (x), Π† (y) = (2π)3 δ 3 (x − y) , (196)
x0 =y0
QFT PS4 Solutions: Free Quantum Field Theory (22/10/18) 21
along with the other zero commutators of the ETCR:
h i
0 = [Φ(x), Φ(y)]x0 =y0 = Φ(x), Φ† (y) (197)
x =y
h i0 0
†
= [Π(x), Π(y)]x0 =y0 = Π(x), Π (y) (198)
x0 =y0
h i
†
= Φ(x), Π (y) = 0, (199)
x0 =y0

and so forth.
Note that the ETCR for the complex fields should match what we found for the real field case
exactly. These ETCR are true for ANY field. Since we are meant to treat Φ and (Φ)† as separate
fields, they should each obey exactly the same equal time commutation relations. It is trivial to
check the as we can take the hermitian conjugate of [Φ, Π] and should get the [Φ† , Π† ] for free and it
should look the same. The factor of i is critical for that. This is really i~δ 3 (x − y) but with ~ = 1,
the same factor as seen in the QM [q, p] = i~ commutator.
Then we can prove the charge-field commutator by calculating (not all operators are at equal times)
Z h i
[Q, Φ(x)] = i d3 y Φ† (y)Π† (y) − Φ(y)Π(y), Φ(x) (200)
Z
= −i d3 y [Φ(y)Π(y), Φ(x)] (201)
Z
= −i d3 y −iΦ(y)δ 3 (x − y)

(202)
= −Φ(x) (203)

Alternatively, we can prove this as follows:

d3 p d3 q
Z
1 h † †
i
iq·x
h
† † †
i
−iq·x

[Q, Φ(x)] = c b
p p − b c ,
p p q b e + c b
p p − b c ,
p p q c e (204)
(2π)3 (2π)3 2ωq
p

d3 p d3 q
Z
1 3 (3)

iq·x † −iq·x

= − (2π) δ (p − q) bp e + b p e (205)
(2π)3 (2π)3 2ωq
p

d3 p
Z
1 ip·x † −ip·x

= − bp e + b p e (206)
(2π)3 2ωp
p

= −Φ(x) (207)

where we have used [ab, c] = a [b, c] + [a, c] b and hence

h i h i
b†p bp , bq = c†p , bq bp = −(2π)3 δ (3) (p − q)bp (208)
h i
b†p bp , b†q = 0 (209)
h i h i
c†p cp , b†q = b†p cp , b†q = (2π)3 δ (3) (p − q)b†p (210)
h i
c†p cp , bq = 0 (211)

(iii) We have that [Q, η(x)] = qη(x) for some operator η with c-number charge q. Consider

η → η 0 = eiθQ ηe−iθQ (212)

QFT PS4 Solutions: Free Quantum Field Theory (22/10/18) 22
We proceed exactly as we did for this type of transformation for the Hamiltonian and annihilation
operator. In fact as all we used was the commutation relation, a simple substitution into the answer
given above is sufficient to convince that indeed η 0 = eiθq η.
However let us illustrate this with a different proof.

dη 0
= iQeiθQ .η.e−iθQ + QeiθQ .η.(−iQe−iθQ ) (213)
dθ
= ieiθQ .Qη.e−iθQ − ieiθQ .ηQ.(e−iθQ ) (214)
= ieiθQ .[Q, η].e−iθQ = ieiθQ .qQ.e−iθQ = iqeiθQ Q.e−iθQ (215)
dη 0
⇒ = iqQ (216)
dθ
Integrating this and using the fact that if θ = 0 we have η = η 0 as a boundary condition, we see
that we get our answer η 0 = eiθq η. If integration of an operator equation worries you, note that
if you substitute the solution into the differential equation, the operators are not involved in the
differentiations, just the c-number parts with which you are familiar.