You are on page 1of 37

# First to Second Quantization

1
I. QUANTUM MECHANICS

Quantum field theory is based on the same quantum mechanics that was invented by
Heisenberg, Born, Schrödinger, Pauli, and others in 1925-26. Therefore, for completeness
and continuity, we will review first quantization, that is, (ordinary) nonrelativistic quantum
mechanics, before going on to second quantization.

A. The postulates

In any quantum mechanical process, the object of interest to calculate is the probability
amplitude - a complex number. The square of the modulus of the amplitude is proportional
to the probability that the process will occur. The postulates of quantum mechanics are
rules on how to calculate the amplitude for any process.
For a given physical system, there is a complete set of states that describes all of the
possible configurations of the system. A state, denoted by |Ψi, is a vector in some Hilbert
space. c|Ψi, c a constant, represents the same physical state. The inner product of two
states, |Ψ1 i and |Ψ2 i, is written as hΨ2 |Ψ1 i. The amplitude for the process that takes the
system from the state |Ψ1 i to |Ψ2 i is just hΨ2 |Ψ1 i. The probability for this process to occur
is |hΨ2 |Ψ1 i|2 .
With every physical observable A, there is associated a Hermitian operator Â that acts
on the Hilbert space containing the states. If a measurement is made of the observable A,
the result of the measurement must be one of the real eigenvalues of Â - no other values
are possible. The expectation value of the physical quantity A in the state |Ψi is given by
hΨ|Â|Ψi, where the state vector is normalized to hΨ|Ψi = 1.
Last, quantum mechanics provides for time evolution by requiring the states of the system
to satisfy the Schrödinger equation,

d
ih̄ |Ψ(t)i = Ĥ|Ψ(t)i. (1)
dt

Ĥ is called the Hamiltonian operator and generates infinitesimal time translations. Suppose
it is time-independent, the formal solution to the Schrödinger equation is
µ ¶
i
|Ψ(t)i = exp − tĤ |Ψ(0)i ≡ Û (t)|Ψ(0)i. (2)

2
Û (t) is the time evolution operator. Substituting Eq.(2) into Eq.(1),
d d
ih̄ |Ψ(t)i = ih̄ Û (t)|Ψ(0)i
dt dt
= Ĥ Û (t)|Ψ(0)i
= Ĥ|Ψ(t)i (3)

we obtain " #
d d
Ĥ = ih̄ Û (t) Û † (t) = −ih̄Û (t) Û † (t). (4)
dt dt
If we have a classical Hamiltonian system (say, a particle) to quantize, we may obtain
the Hamiltonian operator Ĥ using its classical Hamiltonian H(x, p) as a guide. Since both
the position x and momentum p of a particle are observable quantities, upon quantization,
they become operators x̂ and p̂. We can turn H(x, p) into Ĥ by replacing x with x̂ and p
with p̂. To complete the quantization, we must specify quantum conditions for x̂ and p̂:

## [x̂, p̂] = ih̄. (5)

x̂ and p̂ are defined such that their commutators are equal to ih̄ times the corresponding
classical Poisson bracket. Since x̂ and p̂ do not commute, potential operator ordering prob-
lems arise in constructing Ĥ from H(x, p). For example, operator products x̂p̂ and p̂x̂ are
inequivalent while xp and px are the same in the classical system. In general, we choose a
symmetric ordering, xp → (x̂p̂ + p̂x̂)/2.
If the quantum mechanical system we are trying to describe has no classical analog, such
as a system with spin, then we must guess what the Hamiltonian operator is and what the
quantum conditions are.

B. Heisenberg picture

So far we have been describing the Schrödinger picture, where the states of the system
are time dependent while the operators are not. In the Heisenberg picture, the states are
time independent. The operators carry the time dependence. The transformation between
the two pictures is done with the time evolution operator Û (t). We want to obtain the same
results in both representations. In particular, the expectation of any operator Â should be
the same:
hΨ(t)|Â|Ψ(t)i = hΨ(0)|Û † (t)ÂÛ (t)|Ψ(0)i ≡ hΨ(0)|ÂH (t)|Ψ(0)i (6)

3
The time dependent operators in the Heisenberg picture are related to the Schrödinger
picture operators by
ÂH (t) = Û † (t)ÂÛ (t). (7)

Since the states are now time independent, they do not have to satisfy any Schrödinger
equation. The dynamics is locked into the time dependent operators. They must satisfy an
operator equation of motion,
Ã !
d d † d
ÂH = Û ÂÛ + Û † Â Û
dt dt dt
Ã !
d † d
= Û † Û Û ÂÛ + Û † Â Û Û † Û
dt dt
i † i †
= Û Ĥ ÂÛ − Û ÂĤ Û
h̄ h̄
i i
= ĤH ÂH − ÂH ĤH
h̄ h̄
i
= [ĤH , ÂH ] (8)

Note that if ÂH commutes with ĤH , it is a constant of the motion. If the operator ÂH has
an explicit time dependence, then we must add ∂ ÂH /∂t to the right-hand side of Eq.(8).
Equation (8) may also be obtained from the classical system by replacing ih̄ times the
Poisson bracket with a commutator, just as in the quantum conditions, Eq.(5).
In summary, to quantize a classical system in the Heisenberg picture, we elevate observ-
ables to operators, specify quantum conditions, choose an operator ordering, and specify
a Hilbert space containing the time independent state vectors. To compute the quantum
dynamics, we must solve the operator equation of motion. We note that since x̂H and p̂H
are time dependent, we specify the commutator at equal times,

## The classical Hamiltonian for the harmonic oscillator is

Ã r !Ã r !
1 1 1 i ω i ω
H = p2 + ω 2 x2 = h̄ω − √ p + x √ p+ x . (10)
2 2 2 h̄ω h̄ h̄ω h̄

4
To quantize this, we make x and p the operators x̂ and p̂, and demand them to satisfy the
commutation relation, [x̂, p̂] = ih̄. It follows that

1 1
Ĥ = p̂2 + ω 2 x̂2 , (11)
2 2
dx̂ i i 2
= [Ĥ, x̂] = [p̂ , x̂] = p̂, (12)
dt h̄ 2h̄
d2 x̂ dp̂ i iω 2 2
= = [Ĥ, p̂] = [x̂ , p̂] = −ω 2 x̂. (13)
dt2 dt h̄ 2h̄
We note that
x̂(t) = Â cos ωt + B̂ sin ωt (14)

## solves Eq.(13). And,

p̂(t) = −ω Â sin ωt + ω B̂ cos ωt. (15)

It follows that
√ √ √
ωx̂ = ω Â cos ωt + ω B̂ sin ωt,
1 √ √
√ p̂ = − ω Â sin ωt + ω B̂ cos ωt,
ω

and
√ i √
ωx̂ + √ p̂ = ω(Â + iB̂) exp(−iωt),
ω
√ i √
ωx̂ − √ p̂ = ω(Â − iB̂) exp(iωt).
ω

## Alternatively, we could express

µ ¶
† 1
Ĥ = h̄ω â â + , (16)
2

where
Ãr !
1 ω i
â = √ x̂ + √ p̂ ,
2 h̄ h̄ω
Ãr !
† 1 ω i
â = √ x̂ − √ p̂ , (17)
2 h̄ h̄ω
with r r
1 ω i ω i
[â, â† ] = [ x̂ + √ p̂, x̂ − √ p̂] = 1, (18)
2 h̄ h̄ω h̄ h̄ω

5
and

d i
â = [Ĥ, â] = iω[â† â, â] = −iωâ,
dt h̄
d † i
â = [Ĥ, â† ] = iω[â† â, â† ] = iωâ† . (19)
dt h̄
q q
Substituting x̂ = h̄/2ω(â + â† ) and p̂ = −i h̄ω/2(â − â† ) into the Hamiltonian operator,
we find

1 2 1 2 2
Ĥ = p̂ + ω x̂
2 2
1
= h̄ω[−(â − â† )2 + (â + â† )2 ]
4
1
= h̄ω(ââ† + â† â)
2 µ ¶
† 1
= h̄ω â â + (20)
2

Clearly if a state |ni is an eigenvector of Ĥ, then it must be an eigenvector of the number
operator N̂ = â† â and vice versa.

N̂ |ni = n|ni,
µ ¶
1
Ĥ|ni = n+ h̄ω|ni. (21)
2

## hn|â† â|ni = nhn|ni ≥ 0 and hn|ni > 0 ⇒ n ≥ 0. (22)

We will normalize the states |ni such that hn|ni = 1. The ground state has n = 0. It is the
state of lowest energy h̄ω/2 and |0i satisfies

â|0i = 0. (23)

The first excited state has n = 1, the second n = 2, etc. The spectrum of Ĥ is En =
(n + 1/2)h̄ω. The energy levels are evenly spaced with separation h̄ω. To reach the excited
states from the ground state |0i, we observe that

[N̂ , â† ] = [â† â, â† ] = â† [â, â† ] = â† . (24)

Thus,
N̂ â† |ni = (â† N̂ + [N̂ , â† ])|ni = (n + 1)â† |ni = (n + 1)λ|n + 1i. (25)

6
Similarly, we have
[N̂ , â] = [â† â, â] = [â† , â]â = −â, (26)

## N̂ â|ni = (âN̂ + [N̂ , â])|ni = (n − 1)â|ni = (n − 1)µ|n − 1i. (27)

λ and µ are constants that are fixed by making sure the states generated are properly
normalized.

## ⇒ λ2 = hn|ââ† |ni = hn|(â† â + 1)|ni = n + 1

⇒ λ= n+1 (28)

µ|n − 1i = â|ni

⇒ µ2 = hn|â† â|ni = n

⇒ µ= n (29)

It follows that all the eigenstates of Ĥ can be generated by applying â† consecutively to |0i.
In general,
1
|ni = √ (â† )n |0i. (30)
n!
The operators â† and â are called the ladder operators, or raising and lowering operators,
because they make states that march up and down the ladder of excitations. As we shall soon
see, free quantum field theories reduce to a collection of independent harmonic oscillators,
one for each energy-momentum. The raising and lowering operators of the collection of
oscillators provide a particle interpretation. To find a particle interpretation of any quantum
field we look for operators similar to â† and â in the quantum field theory.

D. Coordinate representation

The position x of a particle is a physical observable. Let the associated operator be x̂.
The eigenvectors of x̂ satisfy
x̂|xi = x|xi. (31)

## hx0 |xi = δ(x0 − x), (32)

7
where δ is the Dirac distribution. They are assumed complete,
Z
dx|xihx| = 1, (33)

and form a basis for the Hilbert space of states. If we expand a state vector |Ψ(t)i in the
position basis |xi, the component of |Ψ(t)i in the |xi direction is hx|Ψ(t)i, which is just a
number Ψ(x, t) - the wave function representing the state |Ψ(t)i.
µZ ¶ Z Z
|Ψ(t)i = dx|xihx| |Ψ(t)i = dx|xihx|Ψ(t)i = dxΨ(x, t)|xi (34)

|xi is the state of the system where the particle is at position x. hx|Ψ(t)i = Ψ(x, t) is the
amplitude for the system in the state |Ψ(t)i to also be in the state |xi. Thus, |Ψ(x, t)|2 dx
is the probability that the particle will be found in (x − dx/2, x + dx/2) when the system is
in state |Ψ(t)i. Since the particle must be found somewhere,
Z Z
2
1= dx|Ψ(x, t)| = dxhΨ(t)|xihx|Ψ(t)i = hΨ(t)|Ψ(t)i. (35)

## In the coordinate representation, the Schödinger equation (1) becomes

ih̄ hx|Ψ(t)i = hx|Ĥ|Ψ(t)i,
∂t
Z

ih̄ Ψ(x, t) = dx0 hx|Ĥ|x0 ihx0 |Ψ(t)i
∂t Z
= dx0 hx|Ĥ|x0 iΨ(x0 , t). (36)

hx|Ĥ|x0 i is called the matrix element of the operator Ĥ in the position basis. To compute
hx|Ĥ|x0 i, we must determine the matrix elements hx|x̂|x0 i of x̂ and hx|p̂|x0 i of p̂. Since |xi
is an eigenvector of x̂,
hx|x̂|x0 i = xhx|x0 i = xδ(x − x0 ). (37)

## Suppose p̂ has eigenvectors |pi, then

Z
0
hx|p̂|x i = dphx|p̂|pihp|x0 i
Z
= dp phx|pihp|x0 i
Z Ã !

= dp −ih̄ hx|pihp|x0 i
∂x
∂ Z
= −ih̄ dphx|pihp|x0 i
∂x

= −ih̄ hx|x0 i
∂x

= −ih̄ δ(x − x0 ) (38)
∂x

8
If the Hamiltonian is
1
Ĥ = p̂2 + V (x̂), (39)
2
then " #
0 h̄2 ∂ 2
hx|Ĥ|x i = − + V (x) δ(x − x0 ). (40)
2 ∂x2
Substituting Eq.(40) into Eq.(36), we recover the familiar Schrödinger equation for the wave
function.
∂ h̄2 ∂ 2
ih̄ Ψ(x, t) = − Ψ(x, t) + V (x)Ψ(x, t). (41)
∂t 2 ∂x2
We note that

∂ h̄2 ∂2
ih̄Ψ∗ (x, t) Ψ(x, t) = − Ψ∗ (x, t) 2 Ψ(x, t) + V (x)Ψ∗ (x, t)Ψ(x, t),
∂t 2 ∂x
∂ ∗ h̄2 ∂2
−ih̄Ψ(x, t) Ψ (x, t) = − Ψ(x, t) 2 Ψ∗ (x, t) + V (x)Ψ∗ (x, t)Ψ(x, t),
∂t 2 ∂x
from which we derive
Ã !
∂ ∗ ih̄ ∂ ∂Ψ ∂Ψ∗
(Ψ Ψ) − Ψ∗ −Ψ = 0. (42)
∂t 2 ∂x ∂x ∂x

## To second quantize the Schrödinger equation,

∂ 1 ∂2
i ϕ(x, t) = − ϕ(x, t) + V (x)ϕ(x, t), (43)
∂t 2 ∂x2

we want to make the first quantized wave function ϕ(x, t) an operator ϕ̂(x, t), a time-
dependent field operator (so we will be working in the Heisenberg picture). Equation (43)
will then become an operator equation of motion,

∂ 1 ∂2
i ϕ̂(x, t) = − ϕ̂(x, t) + V (x)ϕ̂(x, t). (44)
∂t 2 ∂x2

The normalized eigenfunctions ϕn (x), with eigenvalue en , of the first quantized Hamilto-
nian,
1 ∂2
− + V (x), (45)
2 ∂x2

9
are assumed to form a complete set, thus any solution ϕ(x, t) to the Schrödinger equation
(43) can be expanded in terms of the ϕn ’s:
X
ϕ(x, t) = an (t)ϕn (x). (46)
n

In the first quantized system, ϕ(x, t) and ϕn (x) are wave functions and an (t) is just a number
times exp(−ien t). After second quantizing, ϕ(x, t) becomes an operator. Making an (t) an
operator and leaving ϕn (x) a function,
X
ϕ̂(x, t) = ân (t)ϕn (x) (47)
n

## clearly solves Eq.(44).

In quantum mechanics, operator equations of motion take the form of Eq. (8),

ϕ̂(x, t) = i[Ĥ, ϕ̂(x, t)]. (48)
∂t

Therefore, we must find a field Hamiltonian Ĥ and quantum conditions (equal-time com-
mutation relations) such that Eq.(48) reproduces Eq.(44). The commutator is the field
equivalent of [x̂(t), p̂(t)] = i for the first quantized system. In this field theory, ϕ̂(x, t) plays
the role of x̂(t). To write down the quantum commutator involving ϕ̂(x, t), we must find
what the momentum field π̂(x, t) conjugate to ϕ̂(x, t) is. Our strategy will be to treat the
first quantized system as a classical field theory, find an action principle that yields the
field equation (43), and use the action as a crutch to find the mometum field π(x, t) and
Hamiltonian H.

## A classical mechanical system is described by a set of generalized coordinates q1 , · · · , qN ,

the associated velocities, q̇1 , · · · , q̇N , and a Lagrangian L[qi (t), q̇i (t), t]. The dot denotes the
time derivative d/dt. The Lagrangian governs the dynamics and is, at most, a quadratic
function of q̇i . The time integral
Z tb
A[qi (t)] = dt L[qi (t), q̇i (t), t] (49)
ta

10
over the Lagrangian along an arbitrary path qi (t) is called the action of this path. The path
qi (t) that is actually chosen by the system as a function of time is called the classical orbit,
qicl (t). It has the property of extremizing the action in comparison with all neighboring paths

## qi (t) = qicl (t) + δqi (t) (50)

with fixed end points: δqi (ta ) = δqi (tb ) = 0. To express this property formally, we consider
the variation of A[qi (t)],

## δA[qi (t)] ≡ {A[qi (t) + δqi (t)] − A[qi (t)]}lin

Z tb
= dt {L[qi (t) + δqi (t), q̇i (t) + δ q̇i (t), t] − L[qi (t), q̇i (t), t]}lin
ta
Z tb " #
∂L ∂L
= dt δqi (t) + δ q̇i (t)
ta ∂qi ∂ q̇i
Z tb " Ã !# ¯t
∂L d ∂L ∂L ¯b
¯
= dt − δqi (t) + δqi (t)¯
ta ∂qi dt ∂ q̇i ∂ q̇i ¯
ta
Z tb " Ã !#
∂L d ∂L
= dt − δqi (t) (51)
ta ∂qi dt ∂ q̇i

Here, repeated indices are understood to be summed - Einstein’s summation convention. For
the classical orbit qicl (t), δA[qi (t)] = 0 (Hamilton’s principle of least action) and we obtain
the Euler-Lagrange equations: Ã !
∂L d ∂L
− = 0. (52)
∂qi dt ∂ q̇i
There is an alternative formulation of classical dynamics which is based on a Legendre
transformed function of the Lagrangian called the Hamiltonian

H[qi (t), pi (t), t] ≡ q˙i (t) L[qi (t), q̇i (t), t] − L[qi (t), q̇i (t), t]. (53)
∂ q̇i

Its value at any time is identified with the energy of the system. The natural variables in H
are no longer qi (t) and q̇i (t), but qi (t) and the canonically conjugate momenta pi (t) defined
by the equations

pi (t) ≡ L[qi (t), q̇i (t), t]. (54)
∂ q̇i
In order to specify the Hamiltonian H[qi (t), pi (t), t] in terms of its proper variables qi (t) and
pi (t), the equations (54) for pi (t) have to be solved for q̇i (t),

## q̇i (t) = vi [qi (t), pi (t), t], (55)

11
and inserted into Eq.(53), giving

H[qi (t), pi (t), t] = vi [qi (t), pi (t), t]pi (t) − L[qi (t), vi [qi (t), pi (t), t], t]. (56)

## In terms of H, we have the canonical form of the action, Eq.(49),

Z tb
A[qi (t), pi (t)] = dt {q̇i (t)pi (t) − H[qi (t), pi (t), t]}. (57)
ta

## The variation of A[qi (t), pi (t)],

Z tb " #
∂H ∂H
δA[qi (t), pi (t)] = dt δ q̇i (t)pi (t) + q̇i (t)δpi (t) − δqi (t) − δpi (t)
ta ∂qi ∂pi
Z tb (" # " # )
∂H ∂H
= dt q̇i (t) − δpi (t) − ṗi (t) + δqi (t) (58)
ta ∂pi ∂qi
since δqi (ta ) = δqi (tb ) = 0. For the classical orbit, qicl (t) and pcl
i (t), δA[qi (t), pi (t)] = 0 and

## we obtain the Hamilton equations:

∂H
q̇i (t) = ,
∂pi
∂H
ṗi (t) = − . (59)
∂qi
These agree with the Euler-Lagrange equations (52) via Eq.(54), as can easily be verified.
Ã !
∂L ∂L d ∂L
pi = = ⇒ ṗi = ,
∂ q̇i ∂vi dt ∂ q̇i

∂H ∂vj ∂L ∂L ∂vj
= pj − −
∂qi ∂qi ∂qi ∂vj ∂qi
∂vj ∂L ∂vj
= pj − − pj
∂qi ∂qi ∂qi
∂L
= − .
∂qi
The 2N -dimensional space of all qi and pi is called a phase space. An arbitrary function
F [qi (t), pi (t), t] changes along an arbitrary path as follows
d ∂F ∂F ∂F
F [qi (t), pi (t), t] = q̇i + ṗi + . (60)
dt ∂qi ∂pi ∂t
If the path is a classical orbit, we may insert Eq.(59) and find
d ∂F ∂H ∂F ∂H ∂F
F [qi (t), pi (t), t] = − +
dt ∂qi ∂pi ∂pi ∂qi ∂t
∂F ∂H ∂H ∂F ∂F
= − +
∂qi ∂pi ∂qi ∂pi ∂t
∂F
≡ {F, H} + , (61)
∂t

12
where we have introduced the Poisson bracket,
∂A ∂B ∂B ∂A
{A, B} ≡ − , (62)
∂qi ∂pi ∂qi ∂pi
again with the Einstein summation convention for the repeated index i. From Eq.(61), a
function F [qi (t), pi (t)] which has no explicit dependence on time and which commutes with
H, i.e., {F, H} = 0, is a constant of motion along the classical orbit. In particular, H itself
is often of this type, i.e., H = H[qi (t), pi (t)]. In this case, the energy is a constant of motion.
Obviously, the original Hamilton equations themselves are a special case of Eq.(61).
d ∂qi ∂H ∂H ∂qi ∂H
qi = {qi , H} = − = , (63)
dt ∂qj ∂pj ∂qj ∂pj ∂pi
d ∂pi ∂H ∂H ∂pi ∂H
pi = {pi , H} = − =− . (64)
dt ∂qj ∂pj ∂qj ∂pj ∂qi
We also note that at each time t,
∂qi ∂pj ∂pj ∂qi
{qi , pj } = − = δij ,
∂qk ∂pk ∂qk ∂pk
∂qi ∂qj ∂qj ∂qi
{qi , qj } = − = 0,
∂qk ∂pk ∂qk ∂pk
∂pi ∂pj ∂pj ∂pi
{pi , pj } = − = 0. (65)
∂qk ∂pk ∂qk ∂pk
The methods of Lagrange and Hamilton provide for an elegant and flexible description
of dynamical systems. They can be applied to any chosen set of generalized coordinates qi ,
the only prerequisite being the knowledge of the Lagrangian or Hamiltonian.

B. Canonical quantization

In a quantum mechanical system we may identify a set of coordinate operators {q̂i }, acting
on a Hilbert space whose vectors are identified with states of the system. The eigenvectors
of the coordinate operators satisfy

## q̂i |q1 , q2 , · · · , qi , · · ·i = qi |q1 , q2 , · · · , qi , · · ·i, (66)

and are states in which cooridnate i has value qi . Operators p̂i corresponding to classical
conjugate momenta pi = ∂L/∂ q̇i obey canonical commutation relations with the q̂i ,

## [q̂i , p̂j ] = ih̄δij ,

[q̂i , q̂j ] = 0,
[p̂i , p̂j ] = 0. (67)

13
The commutation relations are closely related to the classical Poission bracket.

1
{A, B} → [Â, B̂]. (68)
ih̄

In particular, we have

dA ∂A dÂ i ∂ Â
= {A, H} + → = [Ĥ, Â] + . (69)
dt ∂t dt h̄ ∂t

## IV. CLASSICAL FIELD THEORY

In field theory, the analog of the generalized coordinates qi , is a field ϕ(~x, t), in which
the discrete index i has been replaced by the continuous position vector ~x. The position ~x
is not a coordinate, but rather a parameter that labels the field coordinate ϕ at point ~x at
a particular time t. There may be more than one field at each point in space, in which case
the fields may carry a distinguishing subscript, as in ϕa (~x, t). To qualify as a mechanical
system, the fields must be associated with a Lagrangian
Z
L(t) = d3~xL[ϕa (~x, t), ∂µ ϕa (~x, t)], (70)

## which determines their time development.

In Eq.(70), we employ the conventions

~ = g µν Aν ,
Aµ = (A0 , A)
~ = gµν Aν ,
Aµ = (A0 , −A) (71)

## for any vector Aµ , where the Minkowski metric

 
 +1 0 0 0 
 

 0 −1 0 0 

gµν = 

.
 (72)
 0 0 −1 0 
 
 
0 0 0 −1

For derivatives with respect to the coordinate vector xµ = (x0 , ~x), we use the notation

∂ϕa
∂µ ϕa ≡ ,
∂xµ
∂ϕa
∂ µ ϕa ≡ . (73)
∂xµ

14
Here x0 ≡ ct, where c is the speed of light, so that all the xµ have dimensions of length. From
hereon, we set c = 1. Generally, we shall use lower case Greek letters (α, β, · · · , µ, ν, · · ·) for
space-time vector indices (0, 1, 2, 3), and lower case italic letters (i, j, · · ·) for purely spatial
vector indices (1, 2, 3).
Now, let us demand that the action
Z t2 Z
A[ϕa (~x, t)] = dt d3~xL[ϕa (~x, t), ∂µ ϕa (~x, t)], (74)
t1 R

## within some region R of space to be extremal:

Z t2 Z
δA[ϕa (~x, t)] = δ dt d3~xL[ϕa (~x, t), ∂µ ϕa (~x, t)]
t1 R
Z t2 Z
= dt d3~xδL[ϕa (~x, t), ∂µ ϕa (~x, t)]
t1 R
Z t2 Z " #
∂L ∂L 3
= dt d ~x δϕa + δ∂µ ϕa
t1 R ∂ϕa ∂(∂µ ϕa )
Z t2 Z ( " #)
3 ∂L ∂L
= dt d ~x − ∂µ δϕa
t1 R ∂ϕa ∂(∂µ ϕa )
Z Z t2 Z ¯t
∂L ¯2 ∂L
+ d ~x 3
δϕa ¯¯ + dt dσi δϕa
R ∂(∂0 ϕa ) ¯ t1 ∂R ∂(∂i ϕa )
t1
= 0, (75)

and obtain the equations of motion at every point inside R for every time between t1 and
t2 , " #
∂L ∂L
− ∂µ = 0. (76)
∂ϕa ∂(∂µ ϕa )
Here, the variation is over all possible fields ϕa (~x, t) inside R,

## for all ~x in R, and

δϕa (~y , t) = 0 (79)

for all ~y on the surface ∂R of R, but are otherwise arbitrary. But, Eq.(76) must hold at
every point in space-time since t1 , t2 and R were chosen arbitrarily.

15
The field momentum conjugate to ϕa (~x, t) is
∂L
πa (~x, t) = . (80)
∂(∂0 ϕa (~x, t))
The field Hamiltonian is given by the Legendre transform of the Lagrangian,
Z
H= d3~x{πa (~x, t) · ∂0 ϕa (~x, t) − L[ϕa (~x, t), ∂µ ϕa (~x, t)]}. (81)

If we are given a Lagrangian, we may quickly find the field equations from Eq.(76), the
Hamiltonian from Eq.(81), and quantize by making ϕa (~x, t) and its conjugate momentum
πa (~x, t) operators satisfying equal-time commutation relations,

[ϕ̂a (~x, t), π̂b (~x0 , t)] = iδab δ 3 (~x − ~x0 ). (82)

In the absence of difficult operator ordering ambiguities, the field Hamiltonian operator and
quantum conditions, will reproduce the field equations as operator equations of motion.

A. Noether’s theorem

If the action A[ϕa (~x, t)] is unchanged by a re-parametrization of xµ and ϕa , i.e., is in-
variant under some group of transformations on xµ and ϕa , then there exist one or more
conserved quantities, i.e., combinations of fields and their derivatives which are invariant un-
der the transformations. This crucial result is known as Noether’s theorem. It accounts for
conservation of energy, momentum, angular momentum, and any other “quantum” number
which particles happen to carry, like charge, color, isospin, etc.
To prove Noether’s theorem, consider
Z
δA[ϕa (~x, t)] = δ d4 xL[ϕa (~x, t), ∂µ ϕa (~x, t)]
Z R
= d4 xδL[ϕa (~x, t), ∂µ ϕa (~x, t)]
R
Z " #
4∂L ∂L
= dx δϕa + δ∂µ ϕa + ∂µ Lδxµ
R ∂ϕa ∂(∂µ ϕa )
Z ( " #) Z " #
4 ∂L ∂L ∂L µ ν
= dx − ∂µ δϕa + dσµ δϕa + δ ν Lδx
R ∂ϕa ∂(∂µ ϕa ) ∂R ∂(∂µ ϕa )
Z " # Z " #
∂L ∂L ∂L
= dσµ δϕa + ∂ν ϕa δxν − dσµ ∂ν ϕa − δ µν L δxν
∂R ∂(∂µ ϕa ) ∂(∂µ ϕa ) ∂R ∂(∂µ ϕa )
Z " #
∂L µ ν
= dσµ ∆ϕa − θ ν δx ,
∂R ∂(∂µ ϕa )

16
where the total variation
∆ϕa ≡ δϕa + ∂ν ϕa δxν , (83)

## and the enery-momentum tensor

∂L
θµν ≡ ∂ν ϕa − δ µν L. (84)
∂(∂µ ϕa )
We will first discuss the isospin symmetry, where the fields ϕa vary according to some
small parameter δ²a .
∂ϕa
ϕa → ϕa + δϕa = ϕa + δ²b (85)
∂²b
That is, ∆ϕa = δϕa = (∂ϕa /∂²b )δ²b . If the action A[ϕa (~x, t)] is invariant under this trans-
formation, then
Z Z
∂L
δA[ϕa ] = dσµ ∆ϕa = dσµ Jaµ δ²a = 0. (86)
∂R ∂(∂µ ϕa ) ∂R

## Here, the current

∂L ∂ϕb
Jaµ ≡ . (87)
∂(∂µ ϕb ) ∂²a
It follows from
Z Z
dσµ Jaµ δ²a = d4 xδ²a ∂µ Jaµ = 0 (88)
∂R R

that
∂µ Jaµ = 0. (89)

From this conserved current, we can also establish a conserved charge, given by the integral
over the time-component of the current:
Z
Qa ≡ d3~xJa0 . (90)

## Now let us integrate the conservation equation:

Z
0 = d3~x∂µ Jaµ
Z Z
= d3~x∂0 Ja0 + d3~x∂i Jai
d Z 3 0 Z
= d ~xJa + dSi Jai
dt
d
= Qa + surface term (91)
dt
Let us assume that the fields appearing in the surface term vanish sufficiently rapidly at
infinity so that the last term can neglected. Then,
d
∂µ Jaµ = 0 ⇒ Qa = 0 (92)
dt

17
In summary, the symmetry of the action implies the conservation of a current Jaµ , which in
turn implies a conservation principle:

## symmetry → current conservation → conservation principle (93)

The second case, when the action is invariant under the space-time symmetry of the
Lorentz and Poincaré groups, will be discussed later.

B. Schrödinger field

## The Lagrangian density of the Schrödinger field

1
L = iϕ∗ ∂t ϕ − ∂x ϕ∗ ∂x ϕ − V (x)ϕ∗ ϕ. (94)
2

Here, ∂t ≡ ∂/∂t and ∂x ≡ ∂/∂x. We can determine the equations of motion by treating ϕ
and ϕ∗ as independent objects, so that for instance

∂L
= 0,
∂(∂t ϕ∗ )

∂L 1

= − ∂x ϕ,
∂(∂x ϕ ) 2
∂L
= i∂t ϕ − V (x)ϕ,
∂ϕ∗
give us the Schrödinger equation (43). On the other hand,

∂L
= iϕ∗ ,
∂(∂t ϕ)

∂L 1
= − ∂x ϕ ∗ ,
∂(∂x ϕ) 2
∂L
= −V (x)ϕ∗ ,
∂ϕ
give us the complex-conjugate of Eq.(43).
From Eq.(94), we find that the conjugate momentum field is

∂L
π(x, t) = = iϕ∗ (x, t). (95)
∂(∂t ϕ)

Since the field conjugate to ϕ∗ is found to vanish, there are only two independent fields,
ϕ(x, t) and π(x, t).

18
The Hamiltonian is easily computed to be
Z
H = dx[π(x, t)∂t ϕ(x, t) − L]
Z
1
= dx ∂x ϕ∗ ∂x ϕ + V (x)ϕ∗ ϕ
2 " #
Z 2
1 ∂
= dxϕ∗ − + V (x) ϕ (96)
2 ∂x2

## From Eqs.(42) and (87), we have

∂L ∂ϕ ∂L ∂ϕ∗
J0 = +
∂(∂0 ϕ) ∂² ∂(∂0 ϕ∗ ) ∂²
∂ϕ
= iϕ∗
∂²

= ϕ ϕ. (97)

It follows that
ϕ(²) = ϕ(0) exp(−i²). (98)

And,

∂L ∂ϕ ∂L ∂ϕ∗
J1 = +
∂(∂1 ϕ) ∂² ∂(∂1 ϕ∗ ) ∂²
1 1
= ∂x ϕ∗ (iϕ) − ∂x ϕ(iϕ∗ )
2 2
i ∗
= − (ϕ ∂x ϕ − ϕ∂x ϕ∗ ). (99)
2

## V. SECOND QUANTIZATION (CONT’)

In canonical field quantization, the classical fields ϕ(x, t) and π(x, t) are replaced by
operators ϕ̂(x, t) and
π̂(x, t) = iϕ̂† (x, t), (100)

where ϕ∗ (x, t) is replaced by the Hermitian conjugate field operator ϕ̂† (x, t). We impose on
these operators the equal-time commutation relations:

## [ϕ̂† (x, t), ϕ̂† (x0 , t)] = 0. (103)

19
Using the Hamiltonian
Z
1
Ĥ(t) = dx[ ∂x ϕ̂† ∂x ϕ̂ + V (x)ϕ̂† ϕ̂], (104)
2

## and the above equal-time commutation relations, we find

Z
1
[Ĥ(t), ϕ̂(x, t)] = dy[ ∂y ϕ̂† (y, t)∂y ϕ̂(y, t) + V (y)ϕ̂† (y, t)ϕ̂(y, t), ϕ̂(x, t)]
Z
2
½ ¾
1 † †
= dy ∂y [ϕ̂ (y, t), ϕ̂(x, t)]∂y ϕ̂(y, t) + V (y)[ϕ̂ (y, t), ϕ̂(x, t)]ϕ̂(y, t)
2
Z ½ ¾
1
= − dy [∂y δ(y − x)]∂y ϕ̂(y, t) + δ(y − x)V (y)ϕ̂(y, t)
2
1
= ∂x2 ϕ̂(x, t) − V (x)ϕ̂(x, t) (105)
2

## Alternatively, we could use

Z µ ¶
1
[Ĥ(t), ϕ̂(x, t)] = dy[ϕ̂† (y, t) − ∂y2 − V (y) ϕ̂(y, t), ϕ̂(x, t)]
2
Z µ ¶
† 1 2
= dy[ϕ̂ (y, t), ϕ̂(x, t)] − ∂y − V (y) ϕ̂(y, t)
2
Z µ ¶
1 2
= − dyδ(y − x) − ∂y − V (y) ϕ̂(y, t)
2
1 2
= ∂x ϕ̂(x, t) − V (x)ϕ̂(x, t)
2

Therefore,

ϕ̂(x, t) = i[Ĥ, ϕ̂(x, t)]
∂t
implies
∂ 1
i ϕ̂(x, t) = − ∂x2 ϕ̂(x, t) + V (x)ϕ̂(x, t).
∂t 2
We note that we have picked a particular ordering for the operators in Eq.(104), called
normal ordering, to avoid the vacuum energy divergence.

A. Particle interpretation

## Recall the expansion (47),

X
ϕ̂(x, t) = âi (t)ϕi (x).
i

It follows that
X †
ϕ̂† (x, t) = âi (t)ϕ∗i (x). (106)
i

20
Substitute Eq.(47) into the above equal-time commutation relations (101),

## δ(x − x0 ) = [ϕ̂(x, t), ϕ̂† (x0 , t)]

X X
= [ ϕi (x)âi (t), ϕ∗j (x0 )â†j (t)]
i j
X
= ϕi (x)ϕ∗j (x0 )[âi (t), â†j (t)] (107)
i,j

only if
[âi (t), â†j (t)] = δij . (108)

This is because the first quantized energy eigenfunctions are assumed to be complete,
X X
ϕi (x)ϕ∗i (x0 ) = hx|ϕi ihϕi |x0 i
i i
Ã !
X
= hx| |ϕi ihϕi | |x0 i
i
0
= hx|x i
= δ(x − x0 ) (109)

Alternatively, we could express âi (t) and â†i (t) in terms of ϕ̂(x, t) and ϕ̂† (x, t) respectively.
Z
âi (t) = dxϕ∗i (x)ϕ̂(x, t),
Z
â†i (t) = dxϕi (x)ϕ̂† (x, t). (110)

It follows that
Z Z
[âi (t), â†j (t)] = dx dyϕ∗i (x)ϕj (y)[ϕ̂(x, t), ϕ̂† (y, t)]
Z Z
= dx dyϕ∗i (x)ϕj (y)δ(x − y)
Z
= dxϕ∗i (x)ϕj (x)
Z
= dxhϕi |xihx|ϕj i
= hϕi |ϕj i

= δij (111)

Similarly,
Z Z
[âi (t), âj (t)] = dx dyϕ∗i (x)ϕ∗j (y)[ϕ̂(x, t), ϕ̂(y, t)] = 0,
Z Z
[â†i (t), â†j (t)] = dx dyϕi (x)ϕj (y)[ϕ̂† (x, t), ϕ̂† (y, t)] = 0. (112)

21
Next, substitute the expansion (47) into the Hamiltonian operator.
Z " #
1 ∂2 †
Ĥ(t) = dxϕ̂ − + V (x) ϕ̂
2 ∂x2
Z " #
X
∗ † 1 ∂2 X
= dx ϕi (x)âi (t) − + V (x) ϕj (x)âj (t)
i 2 ∂x2 j
X Z
= ej â†i (t)âj (t) dxϕ∗i (x)ϕj (x)
i,j
X
= ei â†i (t)âi (t) (113)
i

## The time dependence of the operators âi (t) is determined by

d X
âi (t) = i[Ĥ(t), âi (t)] = iej [â†j (t)âj (t), âi (t)] = −iei âi (t), (114)
dt j

which is solved by
âi (t) = exp(−iei t)âi (0) ≡ exp(−iei t)âi . (115)

It follows that
â†i (t) = exp(iei t)â†i , (116)
P †
and Ĥ = i ei âi âi . We note that the use of the eigenfunction basis {ϕi (x)} has the effect
that the time dependence of the operators âi (t) becomes trivial, being characterized by a
simple phase factor.
For fixed i, we note that â†i and âi look identical to the raising and lowering operators of
the harmonic oscillator. The field Hamiltonian is just an infinite sum of harmonic oscillator
Hamiltonians. The expansion (47) has reduced the quantum field theory to an infinite set
of harmonic oscillators. Following the discussion given on the harmonic oscillator, we can
now develop a particle interpretation. In the following we will construct the state vectors,
making use of the time-independent operators âi . This is possible within the Heisenberg
picture where the operators are time dependent while the state vectors are constant.
The lowest energy state of Ĥ, the ground state or bare vacuum, is the one that is empty,

## âi |0i = 0. (117)

The destruction operator for any i, âi , finds no particle (excitation) to annihilate in the
empty vacuum |0i, so the result is the null vector.
â†i |0i is a state of energy ei . It describes one particle of energy ei , created by â†i , the
creation operator for mode i. â†j |0i is also a one-particle state except that the energy of the

22
particle is ej . We have only one field in this theory, so there is only one type of particle.
The difference in energy between the two one-particle states must be due to a difference in
momentum between the two.
â†i â†j |0i is a two-particle state with energy ei + ej . The collection of all the states spanned
by the states formed by operating on |0i with any number of creation operators for any
mode i is called a Fock space. In general,

1
√ (a†1 )n1 (a†2 )n2 · · · |0i ≡ |n1 , n2 , · · ·i. (118)
n1 !n2 ! · · ·

## and the particle-number operator

X X †
N̂ ≡ n̂i = âi âi , (119)
i i
Ã !
X
N̂ |n1 , n2 , · · · , ni , · · ·i = ni |n1 , n2 , · · · , ni , · · ·i = N |n1 , n2 , · · · , ni , · · ·i,
i

## where N is the total number of particles.

Last, we note that ϕ̂(x, t) is expanded in terms of âi (t) only, while ϕ̂† (x, t) is expanded in
terms of â†i (t) only. Thus, ϕ̂(x, t) is a destruction operator and ϕ̂† (x, t) is a creation operator.
ϕ̂† (x, t)|0i is a one-particle state where the particle is located at position x at time t.
X
h0|ϕ̂(x, t)ϕ̂† (x0 , t)|0i = ϕi (x)ϕ∗j (x0 )h0|âi (t)â†j (t)|0i
i,j
X
= ϕi (x)ϕ∗j (x0 )h0|{[âi (t), â†j (t)] + â†j (t)âi (t)}|0i
i,j
X
= ϕi (x)ϕ∗j (x0 )δij
i,j
X
= ϕi (x)ϕ∗i (x0 )
i
= δ(x − x0 ) (120)

So,
|x; ti ≡ ϕ̂† (x, t)|0i. (121)

23
B. Particles in a box

## Consider the case where V (x) = 0 with 0 ≤ x ≤ L. The normalized eigenfunctions,

µ ¶
1 2nπx
ϕn (x) = √ exp i , (122)
L L

are plane waves or momentum eigenfunctions. The expansions (47) and (106) turn into
Fourier series,
µ ¶
1 X 2nπx
ϕ̂(x, t) = √ exp i ân (t),
L n L
µ ¶
† 1 X 2nπx †
ϕ̂ (x, t) = √ exp −i ân (t), (123)
L n L

## and Eq.(110) becomes

µ ¶
1 Z 2nπx
ân (t) = √ dx exp −i ϕ̂(x, t),
L L
µ ¶
† 1 Z 2nπx
ân (t) = √ dx exp i ϕ̂† (x, t). (124)
L L

## It follows that the state

µ ¶
1 Z 2nπx
â†n (t)|0i = √ dx exp i ϕ̂† (x, t)|0i. (125)
L L

Since we are integrating over x, â†n (t) creates a particle via ϕ̂† (x, t) at every point x with

amplitude exp(i2nπx/L)/ L. In other words, the state â†n (t)|0i is a one-particle state with

position probability amplitude at each point given by exp(i2nπx/L)/ L. That is, â†n (t)

creates a particle with wave function exp(i2nπx/L)/ L. The square of this wave function is
independent of x, so â†n (t) creates a one-particle state and we do not know where the particle
is. It has equal probability of being anywhere. The particle has a definite momentum and
we have no idea what its position is, in accordance with the uncertainty principle.
Likewise, ϕ̂† (x, t) creates a particle at x, a definite position. Since we know its position,
we have no idea what its momentum and energy are. To create a particle definitely located
at x, we need to use â†n (t) for every n. The particle has equal probability of having any
momentum and hence n.

24
C. First quantized one-particle system

Now consider the general case where V (x) 6= 0. From above, a state containing one
particle described by wave function f1 (x1 , t) is given by
Z
dx1 f1 (x1 , t)ϕ̂† (x1 , t)|0i. (126)

Suppose this state has a definite energy, then it must be an eigenstate of Ĥ.
Z
Ĥ dx1 f1 (x1 , t)ϕ̂† (x1 , t)|0i
Z Z " #
1 ∂2

= dx dx1 f1 (x1 , t)ϕ̂ (x, t) − + V (x) ϕ̂(x, t)ϕ̂† (x1 , t)|0i
2 ∂x2
Z Z " #
† 1 ∂2
= dx dx1 f (x1 , t)ϕ̂ (x, t) − + V (x) δ(x − x1 )|0i
2 ∂x2
Z Z " #
† 1 ∂2
= dxϕ̂ (x, t) dx1 f (x1 , t) − + V (x) δ(x − x1 )|0i
2 ∂x2
Z " #
1 ∂2
= dx1 − 2
+ V (x1 ) f (x1 , t)ϕ̂† (x1 , t)|0i
2 ∂x1
Z
= dx1 E1 f (x1 , t)ϕ̂† (x1 , t)|0i
Z
= E1 dx1 f (x1 , t)ϕ̂† (x1 , t)|0i

## which is true provided

" #
1 ∂2
− + V (x1 ) f1 (x1 , t) = E1 f1 (x1 , t). (127)
2 ∂x21

So, we see how the first quantized one-particle system emerges from the quantum field theory.

## D. First quantized two-particle system

A two-particle state with one particle at x1 and another at x2 is ϕ̂† (x1 , t)ϕ̂† (x2 , t)|0i. If
the two particles have wave function f2 (x1 , x2 , t), then the state is
Z
dx1 dx2 f2 (x1 , x2 , t)ϕ̂† (x1 , t)ϕ̂† (x2 , t)|0i. (128)

## Applying Ĥ(t) to this state, we have

Z Z Z " #
† 1 ∂2
dx dx1 dx2 f2 (x1 , x2 , t)ϕ̂ (x, t) − + V (x) ϕ̂(x, t)ϕ̂† (x1 , t)ϕ̂† (x2 , t)|0i
2 ∂x2

25
Z Z Z " #
† 1 ∂2
= dx dx1 dx2 f2 (x1 , x2 , t)ϕ̂ (x, t) − + V (x) δ(x − x1 )ϕ̂† (x2 , t)|0i
2 ∂x2
Z Z Z " #
† 1 ∂2
+ dx dx1 dx2 f2 (x1 , x2 , t)ϕ̂ (x, t) − 2
+ V (x) ϕ̂† (x1 , t)ϕ̂(x, t)ϕ̂† (x2 , t)|0i
2 ∂x
Z Z Z " #
† 1 ∂2
= dx ϕ̂ (x, t) dx2 dx1 f2 (x1 , x2 , t) − + V (x) δ(x − x1 )ϕ̂† (x2 , t)|0i
2 ∂x2
Z Z Z " #
† 1 ∂2
+ dx ϕ̂ (x, t) dx1 dx2 f2 (x1 , x2 , t) − + V (x) δ(x − x2 )ϕ̂† (x1 , t)|0i
2 ∂x2
Z " #
1 ∂2
= dx1 dx2 − + V (x1 ) f2 (x1 , x2 , t)ϕ̂† (x1 , t)ϕ̂† (x2 , t)|0i
2 ∂x21
Z " #
1 ∂2
+ dx1 dx2 − + V (x2 ) f2 (x1 , x2 , t)ϕ̂† (x1 , t)ϕ̂† (x2 , t)|0i
2 ∂x22

## This state will be an eigenstate of Ĥ if f2 satisfies

" #
1 ∂2 1 ∂2
− − + V (x1 ) + V (x2 ) f2 (x1 , x2 , t) = E2 f2 (x1 , x2 , t). (129)
2 ∂x21 2 ∂x22

## If we apply the field Hamiltonian operator to its N -particle eigenstate,

Z
dx1 · · · dxN fN (x1 , · · · , xN , t)ϕ̂† (x1 , t) · · · ϕ̂† (xN , t)|0i, (130)

## we find that the N -body wave function must satisfy

" N N
#
1X ∂2 X
− + V (xi ) fN (x1 , · · · , xN , t) = EN fN (x1 , · · · , xN , t). (131)
2 i=1 ∂x2i i=1

In this way, all of the N -body first quantized systems are contained in the corresponding
quantum field theory. The operators ϕ̂(x, t) and ϕ̂† (x, t) change the number of particles
present so we can treat physical processes where the particle number is changing in a unified
manner. One way to exactly solve a quantum field theory is to solve the N -body Schrödinger
equation (131) for general N . Then any amplitude for any process will amount to a multi-
dimensional integral over the initial and final state wave functions.

26
F. Bosons and fermions

The quantum conditions we have chosen specify that ϕ̂† (x, t) commutes with itself. It
follows that
Z
dx1 dx2 f2 (x1 , x2 , t)ϕ̂† (x1 , t)ϕ̂† (x2 , t)|0i
Z
= dx2 dx1 f2 (x2 , x1 , t)ϕ̂† (x2 , t)ϕ̂† (x1 , t)|0i
Z
= dx1 dx2 f2 (x2 , x1 , t)ϕ̂† (x1 , t)ϕ̂† (x2 , t)|0i

Therefore,
f2 (x1 , x2 , t) = f2 (x2 , x1 , t). (132)

f2 (x2 , x1 , t) is symmetric under the exchange of x1 and x2 . The symmetric exchange is due
to the use of commutators in the quantum conditions. By the Pauli principle, this exchange
symmetry implies that we are dealing with bosons.
The wave function for two identical fermions must be antisymmetric under the exchange
of coordinates:
ψ2 (x1 , x2 , t) = −ψ2 (x2 , x1 , t). (133)

It follows that
Z
dx1 dx2 ψ2 (x1 , x2 , t)ϕ̂† (x1 , t)ϕ̂† (x2 , t)|0i
Z
= dx2 dx1 ψ2 (x2 , x1 , t)ϕ̂† (x2 , t)ϕ̂† (x1 , t)|0i
Z
= − dx1 dx2 ψ2 (x1 , x2 , t)ϕ̂† (x2 , t)ϕ̂† (x1 , t)|0i

## If ϕ̂† (x1 , t) and ϕ̂† (x2 , t) commute, then

Z
dx1 dx2 ψ2 (x1 , x2 , t)ϕ̂† (x1 , t)ϕ̂† (x2 , t)|0i = 0.

## ϕ̂† (x1 , t)ϕ̂† (x2 , t) = −ϕ̂† (x2 , t)ϕ̂† (x1 , t),

or
ϕ̂† (x1 , t)ϕ̂† (x2 , t) + ϕ̂† (x2 , t)ϕ̂† (x1 , t) ≡ {ϕ̂† (x1 , t), ϕ̂† (x2 , t)} = 0. (134)

27
From hereon, we will use braces, {}, to denote an anticommutator. Therefore, to quantize
the field ϕ̂(x, t) satisfying Eq.(44) and obtain fermions, we must use anticommutators for
the quantum conditions,

## VI. LORENTZ INVARIANCE

A. Lorentz transformations

It follows that

## = gµν Λµρ xρ Λνσ xσ

= gµν Λµρ Λνσ xρ xσ

= gρσ xρ xσ

= x2

## gµν Λµρ Λνσ = gρσ . (138)

If we take the determinant of Eq.(138), we get (det Λ)2 = 1, which implies that det Λ = ±1.
Transformations with det Λ = +1 are proper, and transformations with det Λ = −1 are
improper. Equation (138) implies

## gµν Λµ0 Λν0

28
= Λ00 Λ00 − Λi 0 Λi 0

= g00

= 1

This means that (Λ00 )2 − Λi 0 Λi 0 = 1; thus, either Λ00 ≥ +1 or Λ00 ≤ −1. Transformations
with Λ00 ≥ +1 are orthochronous, while those with Λ00 ≤ −1 are non-orthochronous.

B. Lorentz group

The set of all Lorentz transformations forms a group. The product of any two Lorentz
transformations is another Lorentz transformation. The product is associative. There is an
identity transformation,
Λµν = δ µν .

## ⇒ g ρα Λνα Λνσ = g ρα gασ

⇒ Λνρ Λνσ = δ ρσ

It follows that
(Λ−1 )ρν = Λνρ = gνµ Λµα g αρ , (139)

## For an infinitesimal Lorentz transformation, we can write

Λµν = δ µν + δω µν . (140)

## gµν Λµρ Λνσ = gµν (δ µρ + δω µρ )(δ νσ + δω νσ )

≈ gµν (δ µρ δ νσ + δ µρ δω νσ + δ νσ δω µρ )

29
= gρσ + δωρσ + δωσρ

= gρσ

It follows that
δωρσ = −δωσρ . (141)

Thus there are six independent infinitesimal Lorentz transformations (in four spacetime
dimensions). These can be divided into three rotations and three boosts:

## for a boost in the direction n̂ by rapidity δη. For example,

 
 cos θ − sin θ 0 
 
Rz (θ) = 
 sin θ cos θ 0 

 
0 0 1

and  
 1 −δθ 0 
 
Rz (δθ) = 
 δθ 1 0 

 
0 0 1
And,  
 cosh η 0 0 − sinh η 
 
 
 0 1 0 0 
Bz (η) = 

,

 0 0 1 0 
 
 
− sinh η 0 0 cosh η
q
where cosh η = γ ≡ 1/ 1 − v 2 /c2 , sinh η = γv/c, and rapidity η = tanh−1 (v/c). So,
 
 1 0 0 −δη 
 

 0 1 0 0  
Bz (δη) = 

.

 0 0 1 0 
 
 
−δη 0 0 1

30
We observe that
 
 1 −δθ 0 
 
Rz (δθ) = 
 δθ 1 0 

 
0 0 1
   
1 0 0  0 −i 0 
   
=   
 0 1 0  − iδθ  i 0 0 

   
0 0 1 0 0 0
= 1 − iδθ(ζ1 |ζ1 ihζ1 | + ζ0 |ζ0 ihζ0 | + ζ−1 |ζ−1 ihζ−1 |)

## = (1 − iζ1 δθ)|ζ1 ihζ1 | + (1 − iζ0 δθ)|ζ0 ihζ0 | + (1 − iζ−1 δθ)|ζ−1 ihζ−1 |,

where |ζ1 i, |ζ0 i, and |ζ−1 i are eigenvectors of the Hermitian matrix with eigenvalues ζ1 = 1,
ζ0 = 0, and ζ−1 = −1 respectively. It follows that

lim [Rz (δθ)]n = lim (1 − iζ1 δθ)n |ζ1 ihζ1 | + |ζ0 ihζ0 | + lim (1 − iζ−1 δθ)n |ζ−1 ihζ−1 |
n→∞ n→∞ n→∞

= Rz (θ)

## Here, δθ = limn→∞ θ/n.

Not all Lorentz transformations can be reached by compounding infinitesimal ones. In-
finitesimal transformations of the form Λ = 1 + δω are clearly proper and orthochronous:

## det Λ = ²κλµν Λκ0 Λλ1 Λµ2 Λν3

= ²κλµν (δ κ0 + δω κ0 )(δ λ1 + δω λ1 )(δ µ2 + δω µ2 )(δ ν3 + δω ν3 )
≈ ²κλµν (δ κ0 δ λ1 δ µ2 δ ν3 + δω κ0 δ λ1 δ µ2 δ ν3 + δ κ0 δω λ1 δ µ2 δ ν3 + δ κ0 δ λ1 δω µ2 δ ν3 + δ κ0 δ λ1 δ µ2 δω ν3

= 1,

Λ00 = δ 00 + δω 00 = 1.

Since the product of any two proper orthochronous Lorentz transformations is proper or-
thochronous:
det Λ = 1, det Λ0 = 1 ⇒ det(ΛΛ0 ) = det Λ det Λ0 = 1,

Λ00 ≥ 1, Λ000 ≥ 1 ⇒ (ΛΛ0 )00 = Λ0µ Λ0µ0 = Λ00 Λ000 + Λ0i Λ0i0 ≥ 1,

any transformation that can be reached by compounding infinitesimal ones is proper or-
thochronous. Thus, the Lorentz transformations that can be reached by compounding in-
finitesimal ones are both proper and orthochronous, and they form a subgroup. Generally,

31
when a theory is said to be Lorentz invariant, this means under the proper orthochronous
subgroup only.
Two discrete transformations that take one out of the proper orthochronous subgroup
are parity and time reversal. The parity transformation is
 
 +1 0 0 0 
 

 0 −1 0 0 

P µν = 

 = (P −1 )µ .
 ν (144)
 0 0 −1 0 
 
 
0 0 0 −1

## It is orthochronous, but improper. The time-reversal transformation is

 
 −1 0 0 0 
 

 0 +1 0 0 

T µν = 

 = (T −1 )µ .
 ν (145)
 0 0 +1 0 
 
 
0 0 0 +1

It is nonorthochronous and improper. From hereon, in this chapter, we will treat the proper
orthochronous subgroup only. Parity and time reversal will be treated separately in later
chapters.

## In quantum theory, symmetries are represented by unitary (or antiunitary) operators.

This means that we associate a unitary operator U (Λ) to each proper, orthochronous Lorentz
transformation Λ. These operators must obey the composition rule

## For an infinitesimal transformation, we can write

i
Û (1 + δω) = Iˆ + δωµν M̂ µν , (147)
2
where M̂ µν = −M̂ νµ is a set of Hermitian operators called the generators of the Lorentz
group.
Let Λ0 = 1 + δω 0 , then to linear order in δω 0 ,

## Û (Λ)−1 Û (Λ0 )Û (Λ) = Û (Λ−1 Λ0 Λ)

32
µ ¶
i 0
⇒ Û (Λ) −1
Iˆ + δωµν M̂ µν Û (Λ) = Û ((Λ−1 )µρ (δ ρσ + δω 0ρσ )Λσν )
2
i 0
⇒ Iˆ + δωµν Û (Λ)−1 M̂ µν Û (Λ) = Û (δ µν + (Λ−1 )µρ δω 0ρσ Λσν )
2
= Û (1 + Λ−1 δω 0 Λ)

Therefore,
i 0 i
Iˆ + δωµν Û (Λ)−1 M̂ µν Û (Λ) = Iˆ + (Λ−1 δω 0 Λ)ρσ M̂ ρσ
2 2
where
0
δωµν Û (Λ)−1 M̂ µν Û (Λ) = (Λ−1 )ρµ δω 0µν Λνσ M̂ ρσ
= gµα Λαρ δω 0µν Λαρ M̂ ρσ
0
= δωµν Λµρ Λνσ M̂ ρσ
0
Since δωµν is arbitrary, we have

## Û (Λ)−1 M̂ µν Û (Λ) = Λµρ Λνσ M̂ ρσ (148)

We observe that each vector index on M̂ µν undergoes its own Lorentz transformation. This
is a general result: any operator carrying one or more vector indices should behave similarly.
For example, for the energy-momentum four-vector P̂ µ , where P̂ 0 is the Hamiltonian Ĥ and
P̂ i are the components of the total three-momentum operator,

## Now, let Λ = 1 + δω in Eq.(148), then to linear order in δω,

µ ¶ µ ¶
i i
Iˆ − δωρσ M̂ ρσ M̂ µν Iˆ + δωρσ M̂ ρσ = (δ µρ + δω µρ )(δ νσ + δω νσ )M̂ ρσ
2 2
i i
M̂ µν − δωρσ M̂ ρσ M̂ µν + δωρσ M̂ µν M̂ ρσ = M̂ µν + δω µρ δ νσ M̂ ρσ + δ µρ δω νσ M̂ ρσ
2 2
i
− δωρσ [M̂ ρσ , M̂ µν ] = g µα δωαρ M̂ ρν + g να δωασ M̂ µσ
2
i
δωρσ [M̂ µν , M̂ ρσ ] = g µρ δωρσ M̂ σν + g νσ δωσρ M̂ µρ
2
1
= (g µρ δωρσ M̂ σν + g µσ δωσρ M̂ ρν + g νσ δωσρ M̂ µρ + g νρ δωρσ M̂ µσ )
2
1
= δωρσ (g µρ M̂ σν − g µσ M̂ ρν − g νσ M̂ µρ + g νρ M̂ µσ )
2
1
= δωρσ (−g µρ M̂ νσ + g µσ M̂ νρ − g νσ M̂ µρ + g νρ M̂ µσ )
2
1
= − δωρσ (g µρ M̂ νσ − g νρ M̂ µσ − g µσ M̂ νρ + g νσ M̂ µρ )
2

33
Therefore,
[M̂ µν , M̂ ρσ ] = i(g µρ M̂ νσ − g νρ M̂ µσ − g µσ M̂ νρ + g νσ M̂ µρ ). (150)

These commutation relations specify the Lie algebra of the Lorentz group. We can identify
the components of the angular momentum operator as Jˆi ≡ −1/2²ijk M̂ jk , and the compo-
nents of the boost operator K̂i ≡ M̂ i0 .

1
[Jˆi , Jˆj ] = ²imn ²jpq [M̂ mn , M̂ pq ]
4
i
= ²imn ²jpq (g mp M̂ nq − g np M̂ mq − g mq M̂ np + g nq M̂ mp )
4
i
= − (²imn ²jmq M̂ nq − ²imn ²jnq M̂ mq − ²imn ²jpm M̂ np + ²imn ²jpn M̂ mp )
4
i
= − (²min ²mjq M̂ nq + ²imn ²jqn M̂ mq + ²inm ²jpm M̂ np + ²imn ²jpn M̂ mp )
4
= −iM̂ ij
= i²ijk Jˆk (151)

since
1 1
²ijk Jˆk = − ²kij ²klm M̂ lm = − (δil δjm − δim δjl )M̂ lm = −M̂ ij .
2 2

1
[Jˆi , K̂j ] = − ²imn [M̂ mn , M̂ j0 ]
2
i
= − ²imn (g mj M̂ n0 − g nj M̂ m0 − g m0 M̂ nj + g n0 M̂ mj )
2
i
= (²ijn M̂ n0 − ²imj M̂ m0 )
2
= i²ijk M̂ k0

## [K̂i , K̂j ] = [M̂ i0 , M̂ j0 ]

= i(g ij M̂ 00 − g 0j M̂ i0 − g i0 M̂ 0j + g 00 M̂ ij )

= iM̂ ij

## Similarly, we can let Λ = 1 + δω in Eq.(149), then to linear order in δω,

µ ¶ µ ¶
i i
Iˆ − δωρσ M̂ ρσ P̂ µ Iˆ + δωρσ M̂ ρσ = (δ µν + δω µν )P̂ ν
2 2

34
i
P̂ µ − δωρσ [M̂ ρσ , P̂ µ ] = P̂ µ + g µα δωαν P̂ ν
2

i
δωρσ [P̂ µ , M̂ ρσ ] = g µρ δωρσ P̂ σ
2
1
= (g µρ δωρσ P̂ σ + g µσ δωσρ P̂ ρ )
2
1
= δωρσ (g µρ P̂ σ − g µσ P̂ ρ )
2

Therefore,
[P̂ µ , M̂ ρσ ] = i(g µσ P̂ ρ − g µρ P̂ σ ). (154)

It follows that

1
[Ĥ, Jˆi ] = − ²ijk [P̂ 0 , M̂ jk ]
2
i
= − ²ijk (g 0k P̂ j − g 0j P̂ k )
2
= 0, (155)

1
[P̂i , Jˆj ] = ²jmn [P̂ i , M̂ mn ]
2
i
= ²jmn (g in P̂ m − g im P̂ n )
2
i
= (−²jmi P̂ m + ²jin P̂ n )
2
= i²ijk P̂k , (156)

## [Ĥ, K̂i ] = [P̂ 0 , M̂ i0 ]

= i(g 00 P̂ i − g 0i P̂ 0 )

= −iP̂i (157)

## [P̂i , K̂j ] = −[P̂ i , M̂ j0 ]

= −i(g i0 P̂ j − g ij P̂ 0 )
= −iδij Ĥ (158)

## Also, the components of P̂ µ should commute with each other.

[P̂ i , P̂ j ] = 0,

35
[P̂ i , Ĥ] = 0. (159)

In summary,
[Ĥ, P̂ i ] = 0,

[P̂ i , P̂ j ] = 0,

[Ĥ, Jˆi ] = 0,

## E. Noether’s theorem (cont’)

Now let us investigate the case when the action is invariant under the space-time sym-
metry of the Lorentz and Poincaré groups. First, consider the translation

xµ → x0µ = xµ + aµ , (160)

## where aµ is a constant. a0 represents time displacements, and ai represents space displace-

ments. From Eq.(84),
∂L
θµν ≡ ∂ν ϕa − δ µν L
∂(∂µ ϕa )
is the energy-momentum tensor. By integrating the energy-momentum tensor, we can gen-
erate conserved quantities. Define
Z
Pν ≡ d3~xθ0ν . (161)

Then

∂µ θµν = 0
⇒ ∂0 θ0ν + ∂i θi ν = 0
d
⇒ Pν = 0. (162)
dt

36
We observe that
Z Z
∂L
P0 = d3~xθ00 = d3~x ∂0 ϕa − L = H. (163)
∂(∂0 ϕa )
Therefore, displacement in time (space) leads to the conservation of energy (momentum).
Here,
Z Z
3 ∂L
Pi = d ~xθ0i = d3~x ∂i ϕa . (164)
∂(∂0 ϕa )
Next, consider the Lorentz transformation

xµ → x0µ = xµ + δω µν xν . (165)

It follows that

∂µ (θµν δω να xα ) = 0

⇒ ∂µ (θµν δωνα xα ) = 0

## ⇒ ∂µ (θµν δωνα xα + θµα δωαν xν ) = 0

⇒ δωνα ∂µ (θµν xα − θµα xν ) = 0
⇒ ∂ρ (θρµ xν − θρν xµ ) = 0 (166)

Define
Z
M µν ≡ d3~x(θ0µ xν − θ0ν xµ ). (167)

Then

d µν
⇒ M =0 (168)
dt

37