Quantum Field Theory 1 Homework 1

Arjun Dey (arjun.dey@weizmann.ac.il)

May 4, 2020

For Question 2 and Question 3, we don’t consider the normalization constants. In Question 2 part
c, the solutions to the independent modes of SHO’s are found by analogy to part b, not by exactly
solving. In reality, the opeators in this case are non-Hermitian, so we should define new operators
which are Hermitian and then find out the wavefunction using those. But the question ask us to
solve it by analogy.

1. Gaussian Integrals and Wick’s Theorem

Z
1
Z= dn x e− 2 xi Aij xj (1)

where A is symmetric and strictly positive matrix.

(a)
∞
Z r
− 21 ax2 2π
dxe = (2)
−∞ a
[For doing this integration, substitute 12 ax2 = y to convert the gaussian integral into a Gamma
function.]
In a similar fashion, for integration over many variables x1 , ...., xn , we get

∞ n
(2π)n/2
Z Z
1X
ai x2i = √


··· dx1 . . . dxn exp − (3)
−∞ 2 a1 . . . an
i=1

In our question, the matrix A is a symmetric matrix so we can always find a orthogonal matrix
U that diagonalizes the matrix A s.t.

A = U T DU (4)

Since U −1 = U T . D is a diagonal matrix.

Using 4 in 1, we get:

1
Quantum Field Theory 1 Arjun Dey (arjun.dey@weizmann.ac.il) Homework 1

Z
1
Z= dn x e− 2 xi Aij xj
Z
1 T Ax
= dn x e− 2 x (5)
Z
1 T
= dn x e− 2 (Ux) D(Ux)

We can substitute Ux = y. We also need to change the volume element accordingly. Since
our transformation is linear and the linear transformation matrix is non-singular, our Jacobian
will simply be the determinant of the transformation matrix.
Therefore,

dy1 . . . dyn = det(U)dx1 . . . dxn (6)

Z
1 T
Z= dn x e− 2 (Ux) D(Ux)

1 T
e− 2 (y) D(y)
Z
n
= d y
∴ det(U) (7)
Since D is diagonal we can simply use the result of 3
(2π)n/2
= p
det(U) det(D)

det(A) = det(UDUT )
(8)
= det(D) det(U)2

Substituting 8 into 7, we get

(2π)n/2  
Z=p P ROV ED (9)
det(A)
− 12 ax2 +bx
R∞
(b) We start by considering the integral −∞ e
To write this integral in the form of 2, we need to complete the square.

Z ∞ Z ∞ b 2 b2
− 12 ax2 +bx 1
e = e− 2 a(x− a ) + 2a
−∞ −∞
r (10)
2π b2
= e 2a
a

2
Quantum Field Theory 1 Arjun Dey (arjun.dey@weizmann.ac.il) Homework 1

Now we generalize this to a symmetric positive matrix A which is not necessarily diagonal.

A = UT DU
=⇒ A−1 = U−1 D−1 (UT )−1 (11)
 
=⇒ A−1 = UT D−1 U Since U is orthogonal i.e. U−1 = UT

Let us consider the exponent of our integrand:

1
θ = − xT Ax + bT x
2
1
= − xT UT DUx + bT x
2
1 T
= − y Dy + bT UT y (Substituting y = Ux)
2
(x = UT y since U is orthogonal)
1
= − Dii yi2 + (U b)i yi
2 (12)
(Since D is diagonal)
1 1 1
= − Dii yi2 + (U b)i yi − (D−1 )ii (U b)2i + (D−1 )ii (U b)2i
2 2 2
1 1
= − Dii (yi − (U b)i )2 + (D−1 )ii (U b)2i
2 2
1 T 1
= − z Dz + (Ub) (D)−1 (Ub)
T
2 2
1 T 1
= − z Dz + bT ∆b
2 2

Where ∆ = A−1 = UT D−1 U (Using 11).

Notice that we have substituted zi = yi − (U b)i . Also observe that the Jacobian for this
substitution is 1 i.e.

dz1 . . . dzn = dy1 . . . dyn (13)

Z
1 T 1 T
Z(b) = e 2
b ∆b
dn z e− 2 z Dz

1 T ∆b (2π)n/2 (14)
= e2b p
det(A)
(Using the result from the last part.)

3
Quantum Field Theory 1 Arjun Dey (arjun.dey@weizmann.ac.il) Homework 1

Hence Proved.

(c) The expectation value of a function F (x) is defined as

Z
1
−1
hF (x)i = Z dn xF (x)e− 2 xi Aij xj (15)

Z
1
−1
∴ he b i xi
i=Z dn xe− 2 xi Aij xj +bi xi
Z(b) (16)
=
Z
1
= e 2 bi ∆ij bj

So our m-point correlator becomes:

!
∂ ∂ 1
e 2 bi ∆ij bj


hxk1 . . . xkm i = ... (17)
∂bk1 ∂bkm
b=0

• Since A is a symmetric non-singular positive matrix, its inverse is also a symmetric matric. So
∆ is a symmetric matrix.
• For any odd m, the correlation function vanishes.

To prove Wick’s theorem, we will use induction. First we will prove it for m = 2.

!
∂ ∂ 1
e 2 bi ∆ij bj


hxk1 xk2 i =
∂bk1 ∂bk2
b=0
!
1
b∆ b

1 (18)
= e 2 i ij j (δik1 ∆ij δjk2 + δjk1 ∆ij δik2 )
2
b=0
= ∆k1 k2 (∵ ∆k1 k2 = ∆k2 k1 )

In the last line we have put b = 0 to make the exponential 1.

By the induction hypothesis, we assume that the Wick’s theorem holds for m = n − 1. We
have to prove that it holds for n also.
Notations for the following calculations:
• I l (ka , kb ) denotes the correlation function hxka . . . xkl−1 xkl+1 . . . xb i.
• In is the n-point correlation function.
1
• A is e 2 bi ∆ij bj

4
Quantum Field Theory 1 Arjun Dey (arjun.dey@weizmann.ac.il) Homework 1

!
∂ ∂ ∂ ∂ 1
e 2 bi ∆ij bj


hxk1 . . . xkn−2 xkn−1 xkn i = ... (19)
∂bk1 ∂bkn−2 ∂bkn−1 ∂bkn
b=0

Among the n partial derivatives, we can choose any one of the derivatives to act first and that
will give us:
!
∂ ∂ ∂ ∂ 1 1
In = e 2 bi ∆ij bj (δka i ∆ij bj + bi ∆ij δka j )


... ... (20)
∂bk1 ∂bka−1 ∂bka+1 ∂bkn 2
b=0

Observe that
Xk−1
∂n−k [ ∆kn−i ,kn I i (n − k + 1, n) + ∆kn ,α bα (∂n−k+1 . . . ∂n−1 A)]
i=1
k−1
X
= ∆kn−i ,kn ∂n−k I i (n − k + 1, n) + ∆kn−k ,kn I k (n − k, n) + ∆kn ,α bα (∂n−k . . . ∂n−1 A) (21)
i=1
Xk
= ∆kn−i ,kn I i (n − k, n) + ∆kn ,α bα (∂n−k . . . ∂n−1 A)
i=1

So if we use 21 in 20, we will get

n
" #
X
In = ∆kn−i ,kn I i (1, n) + ∆kn ,α bα (In−1 )
i=1 b=0
" n
# (22)
X
i
= ∆kn−i ,kn I (1, n)
i=1 b=0

Since by our Induction hypothesis, Wick’s theorem holds for any m = n − 1 variables, I i (1, n)
gives the right-hand-side of the Wick’s theorem statement for n − 1 variables. Applying this
fact in 22, we get that:
X
In = ∆kP1 ,kP2 . . . ∆kPn−1 ,kPn (23)
all pairings P of k1 , . . . kn

Hence Proved.

2. The wave functional of the vacuum 1 - canonical quantization

5
Quantum Field Theory 1 Arjun Dey (arjun.dey@weizmann.ac.il) Homework 1

(a) Upon acting on the ground state wavefunction of the harmonic oscillator, the annihilation
operator gives 0 i.e. â|0i = 0 where
1
â = √ (mωx̂ + ip̂) (24)
2~mω
∂
Since p̂ = −i~ ∂x , the equation for the ground state wavefunction is a first order differential
equation. If |0i = ψ(x), then:
∂ψ(x)
~ = −mωxψ(x)
∂x
2
Let’s take an ansatz ψ(x) = e−αx
2 2
(25)
=⇒ ~(−2αx)e−αx = −mωxe−αx
mω
=⇒ α =
2~
So, the unnormalized ground state wavefunction is given by:
mω 2
ψ(x) = e− 2~ x (26)

(b) φ(t, ~x) is a real scalar field with mass m and action:
Z  
1 1
S = dtd3 x ∂µ φ∂ µ φ − m2 φ2
2 2
Z  
3 1 2 1 ~ 2 1 2 2
= dtd x π − (∇φ) − m φ (27)
2 2 2
∂L
where π̂ =
∂ φ̇
The field operator φ̂ is analogous to x̂ and π̂ is analogous to p̂.
The field operators follow the following commutation relations:
[φ(x), π(y)]Equal T ime = iδ (3) (x − y)
(28)
[φ(x), φ(y)]Equal T ime = [π(x), π(y)]Equal T ime = 0

We know that H = x̄ p(x̄)q̇(x̄) − L. Since we know our Lagrangian density L from the
P
action and also we know the equivalent of p and q, we can find the Hamiltonian density.
H = π φ̇ − L
 
2
=π −L ∵ For the given Lagrangian φ̇ = π (29)
1 1 ~ 2 1 2 2
= π 2 + (∇φ) + m φ
2 2 2
This is the Hamiltonian density of a free scalar field with mass m. So the Hamiltonian for the
free scalar field is: Z Z
1 1 ~ 2 1 2 2
H = d xH = d3 x π 2 + (∇φ)
3
+ m φ (30)
2 2 2

6
Quantum Field Theory 1 Arjun Dey (arjun.dey@weizmann.ac.il) Homework 1

(c) Let us write the field φ(x, t) in momentum space:

d3 p ip·x e
Z
φ(x, t) = e φ(p, t)
(2π)3
(31)
d3 p ip·x ∂ φ(p,
e t) Z d3 p
Z
π(x, t) = φ̇(x, t) = e = eip·x π
e(p, t)
(2π)3 ∂t (2π)3

If we substitute this in our Hamiltonian, we get

d3 p 1
Z
2 2 2 e 2



H= |e
π (p, t)| + |p| + m| |φ(p, t)|
(2π 3 ) 2
(32)
d3 p 1
Z
π (p, t)|2 + ωp2 |φ(p,
e t)|2 ∵ ωp2 = |p|2 + m2



= 3
|e
(2π ) 2

Note that the fields φe and πe are appearing as modulo-squares in the Hamiltonian. This is
because of the delta function from the x-integration. That leaves only counter-propagating
modes. And since φe∗ (p, t) = φ(−p,
e t), we get the modulo. This is a very standard calculation
so we are not doing it explicitly here.
Now we see that each individual momentum mode is an independent harmonic oscillator. So
the ground state wave functional will be the product of all the ground state wave functions
of the independent harmonic oscillators. For each mode the ground state wavefunction is
given by 26. So the overall ground state wavefunctional is (Note that we are not deriving
the wavefunction or the wavefunctional explicitly, we are using analogy with the harmonic
oscillator to write down the wavefunctional):
d3 p ωp e
|φ(p,t)|2
R
ωp e −
e t)] = Πp e− |φ(p,t)|2 (2π)3 2
Ψ[φ(p, 2 =e (33)

Where ωp2 = |p|2 + m2 .

Now if we want the ground state wave functional in the position space we need to do the
inverse Fourier transform on the fields.

d 3 p ωp
d3 x1 d3 x2 e−ip·(x1 −x2 ) φ(x1 ,t)φ(x2 ,t)
R R R
− (34)
Ψ[φ(x, t)] = e (2π)3 2

(d) Unlike the quantum mechanical harmonic oscillator where we have M and ω as parameters, in
quantum field theory (for the given case in the question) the action only have one parameter
~
m. So our length scale is given by λ = mc where m is the relativistic mass. At this length scale,
single particle quantum mechanics breaks down because of pair production (From energy-time
uncertainty relation, you can see pairs of particles created and annihilated at this length scales).
This is the Compton wavelength.

7
Quantum Field Theory 1 Arjun Dey (arjun.dey@weizmann.ac.il) Homework 1

3. The wave functional of the vacuum 2 - path integral derivation

Z
φ(t )=φ
Ψφ1 [φ2 ] = Dφφ(t21 )=φ21 eiS[φ] (35)

This is the wavefunctional for a given action if we take t1 −→ ∞. First we simplify the action by
expanding it around the classical solution φ. e So we use φ = φe + δφ.
Z
1 1
S[φ] = d4 x( ∂µ φ∂ µ φ − m2 (1 − i)φ2 )
2 2
Z
1 1
= d4 x( ∂µ (φe + δφ)∂ µ (φe + δφ) − m2 (1 − i)(φe + δφ)2 )
2 2
Z 
1 e µe 1 1
= d4 x ∂µ φ∂ φ + ∂µ δφ∂ µ δφ + ∂µ δφ∂ µ φe − m2 (1 − i)δφ2 −
2 2 2

1 2
m (1 − i)φe2 − m2 (1 − i)φδφe
2
Z 
1 4 e µ φ)
e − φ∂ e 2 φe − m2 (1 − i)φe2 + 2∂µ (δφ∂ µ φ)
e − 2δφ∂ 2 φe + ∂µ (δφ∂µ δφ) − δφ∂ 2 δφ−
= d x ∂µ (φ∂
2

2 2 1 2 2 2
m (1 − i)δφ − m (1 − i)φ − 2m (1 − i)φδφ
e e
2
Z
1
= d4 x − φ(∂
e 2 φe + m2 (1 − i)φ)e − 2δφ(∂ 2 φe + m2 (1 − i)φ)e + S[δφ]+
2
!
µ µ µ
e φe + δφ∂ φe + δφ∂ ∂φ)
∂µ (φ∂

(36)

The first two terms becomes zero because of equation of motion/Euler Lagrange equations of the
classical solution. S[δφ] doesn’t depend of φe so it gives just a scalar and only contributes as a constant
term outside the path integral. The last term is a total derivative and using Stokes’ theorem we
can take it to the surface. If we assume that at the spatial boundary of our integration the fields
vanish, we can easily see that the second and third terms inside the total derivative vanish. So the
only remaining term that we need to consider is

Z Z
4 e µ φ)
d x∂µ (φ∂ e = d3 xdt∂µ (φ∂
e µ φ)
e
ZV (37)
3 e 0 φ|
e t=0
= d xφ∂ t=−T

So our wavefunctional is given by

8
Quantum Field Theory 1 Arjun Dey (arjun.dey@weizmann.ac.il) Homework 1

Z
1
d3 xφ∂
e 0 φ|
e t=0
R
φ(t )=φ
Ψφ1 [φ2 ] = constant ∗ Dφφ(t21 )=φ21 eiS[φ] = constant ∗ e 2 i t=−T (38)

Notice that our Jacobian for the path integral was 1.

The only thing left to do is to find the classical solution given the boundary conditions.
Our classical equation of motion is:

(∂ 2 + m2 (1 − i))φ(x)
e =0 (39)

with boundary conditions:

φ(−T
e ) = φ1
(40)
φ(0)
e = φ2

This is a very common differential equation in QFT and we will take the solution of this differential
equation from Peskin (without the boundary conditions). To solve this differential equation, the
standard practice is to perform a Fourier transform on the position variables and the differential
equation becomes just a quadratic equation. We solve the quadratic equation and transform it back
to the position space. The solution that we get is:

d3 p
Z
φ(x,
e t) = (C1 (p)eiωp t + C2 (p)e−iωp t ) (41)
(2π)3
p
Where ωp = p2 + m2 (1 − i).
Putting the boundary conditions, we get

d3 p φ1 (p)e−iωp T − φ2 (p) iωp t φ2 (p)e−2iωp T − φ1 (p)e−iωp T −iωp t

Z
φ(x,
e t) = ( e + e ) (42)
(2π)3 e−2iωp T − 1 e−2iωp T − 1
Putting T −→ ∞, we will get:

d3 p
Z
φ(x,
e t) = ((φ2 (p))eiωp t + (0)e−iωp t ) (43)
(2π)3
So now our wavefunctional is:

1
d3 xφ∂
e 0 φ|
e t=0
R
Ψ[φ2 (p)] = limT −→∞ lim−→0 ; constant ∗ e 2 i t=−T

d3 p d3 p0 iω t
1
d3 x φ2 (p)eiωp t d3 x φ2 (p0 )(iωp0 )e p0
R R R R
i
= lim−→0 ; constant ∗ e 2 (2π)3 (2π)3
(44)
d3 p ωp
|φ2 (p,t)|2
R
−
= constant ∗ e (2π)3 2

Since φ(−p) = φ∗ (p).

9
Quantum Field Theory 1 Arjun Dey (arjun.dey@weizmann.ac.il) Homework 1

This is the wavefunctional in the momentum space that we got in canonical quantization.
Now if we do the inverse Fourier transform we get back the wavefunctional that we got in the position
space in the canonical quantization.

4. Gaussian Integration of Grassmann Numbers

(a)
Z
∗
dθ∗ dθ θθ∗ e−θ aθ
Z
1
= dθ∗ dθ θθ∗ (1 − θ∗ aθ + θ∗ aθθ∗ aθ + . . . )
2
But all the terms of the Taylor expansion containing θ∗ in the left gives 0 because of the
θ∗ outside bracket since (θ∗ )2 = 0
Z Z
= dθ∗ dθ θθ∗

=1
(45)

(b) N complex Grassmann variables θi and a Hermitian matrix A.

N
Z !  
X
N ∗ ∗ ∗
Πi=1 dθi dθi θj θk exp − θl Alm θm
l,m=1
N
Z !  !
∂ X
= ΠN
i=1 dθi∗ dθi θj θk∗ exp − θl∗ Alm θm
∂Akj
l,m=1
∂

= det(A) ref: tutorial notes
∂Akj
(46)
∂
= i ...i A1i1 . . . AN iN
∂Akj 1 N
= i1 ...iN A1i1 . . . δik ,j . . . . . . AN iN
 
= (adjugate(A))jk By definition of adjugate. ref:adjugate
 
−1 −1 adjugate(A)
= det(A)(A )jk ∵A =
det(A)

Hence Proved.

10