Introduction to the “second quantization” formalism for non-

relativistic quantum mechanics

A possible substitution for Sections 6.7 and 6.8 of Feynman’s “Statistical
Mechanics”

Hal Tasaki∗

This is a self-contained and hopefully readable account of the method of creation and
annihilation operators (also known as the Fock space representation or the “second quan-
tization” formalism) for non-relativistic quantum mechanics of many particles.1 Assuming
arXiv:1812.10732v4 [cond-mat.stat-mech] 13 Sep 2022

knowledge only of conventional quantum mechanics in the wave function formalism, we

define the creation and annihilation operators, discuss their properties, and introduce cor-
responding representations of states and operators of many-particle systems.2 As the title
of the note suggests, we cover most topics treated in sections 6.7 and 6.8 of Feynman’s3
textbook4 [1].
We note that all the contents of the present note are standard, and the definitions and
derivations presented here have been known to many. Although the style of the present
note may be slightly more mathematical than standard physics literature, we do not try to
achieve full mathematical rigor.5

Contents
1 Wave functions of many particles 1

2 Creation and annihilation operators 5

3 The Fock space representation 10

4 Schrödinger equation and Hamiltonians 17

1 Wave functions of many particles

Single particle We start by recalling the standard quantum mechanical description of a
single particle, such as an electron or an atom. A state (at an instantaneous moment) of
a particle in the three-dimensional space is described by a wave function ϕ(r), which is a
∗
Department of Physics, Gakushuin University, Mejiro, Toshima-ku, Tokyo 171-8588, Japan.
hal.tasaki@gakushuin.ac.jp
1
Japanese translation is available: https://www.gakushuin.ac.jp/~ 881791/pdf/2ndQNoteJ.pdf
2
Note to experts: In particular, we here derive the (anti)commutation relations of the creation and an-
nihilation operators, rather than simply declaring them. In this sense, our approach is quite close to that of
Feynman. But we here focus on the action of creation/annihilation operators on general N body wave functions,
while Feynman heavily uses Slater-determinant-type states from the beginning. We hope that our presentation
provides a better perspective on this formalism.
3
A friend of mine pointed me out that this happens to be the centennial year of Richard Feynman’s birth.
Let me declare that this small article is to celebrate his 100th birthday!
4
I wrote this note for undergraduate students in our group studying Feynman’s textbook. The idea was that
they could skip these two somewhat complicated sections by studying this note (and they indeed did so).
5
The mathematically minded reader might be bothered by our heuristic treatment of the operators ψ̂(x) and
â(k). Our treatment can be made rigorous by using suitable advanced concepts.

1
complex-valued function of the position r = (x, y, z) ∈ R3 . The wave function ϕ(r) satisfies
the condition of square integrability:
Z
d3 r |ϕ(r)|2 < ∞ (1.1)

We denote by H1 the set of all wave functions satisfying (1.1), and call it the single-particle
Hilbert space. Recall that the function which is identically 0 is contained in H1 , but does not
describe any physical state.
Let us use “mathematicians’ notation” and denote the whole wave function ϕ(r) (where r
runs over the whole space R3 ) as ϕ. (We shall reserve the bra-ket notation for the Fock space
description.) For two wave functions ϕ, ψ ∈ H1 , we define their inner product by
Z
hϕ, ψi := d3 r {ϕ(r)}∗ ψ(r). (1.2)

p
The norm of a state ϕ ∈ H1 is defined as kϕk := hϕ, ϕi.

Distinguishable particles Before dealing with indistinguishable particles, we consider a

system of mutually distinguishable particles.
First consider a system of two particles, which we call particle 1 and particle 2. Suppose
that the particle 1 is in a single-particle state ϕ ∈ H1 and the particle 2 in ψ ∈ H1 . By
denoting the positions of the particles 1 and 2 as r 1 and r 2 , respectively, the whole state
may be described by the wave function ϕ(r 1 ) ψ(r 2 ), which is a complex-valued function of
(r 1 , r 2 ). Likewise one can take a state κ(r 1 ) η(r 2 ) of the same two-particle system. Then from
the principle of superposition, we see that states of the form α ϕ(r 1 ) ψ(r 2 ) + β κ(r 1 ) η(r 2 )
with α, β ∈ C are also allowed. Note that such a state is in general not written in the form
(function of r 1 ) × (function of r 2 ). Since any such superpositions are allowed, a general state
of the two-particle system is described by a wave function Φ(r 1 , r 2 ), which is an arbitrary
complex-valued function6 of (r 1 , r 2 ) ∈ R6 .
In exactly the same manner, we see that a quantum mechanical state of a system of N dis-
tinguishable particles is described by a wave function Φ(r 1 , . . . , r N ), a complex-valued function
of r 1 , . . . , r N , where r j is the position of the j-th particle. The wave function should satisfy
the condition of square integrability:
Z
d3 r 1 · · · d3 r N |Φ(r 1 , . . . , r N )|2 < ∞ (1.3)

We again denote the whole wave function Φ(r 1 , . . . , r N ) (where each of r 1 , . . . , r N runs
over the whole space) as Φ. The inner product of two wave functions Φ, Ψ is
Z
hΦ, Ψi := d3 r 1 · · · d3 r N {Φ(r 1 , . . . , r N )}∗ Ψ(r 1 , . . . , r N ), (1.4)

p
and the norm of a wave function Φ is kΦk := hΦ, Φi.
6
Recall that, with an arbitrary
P complete orthonormal system {ξ α }α=1,2,... of H1 , one can express any function
of (r1 , r2 ) as Φ(r 1 , r2 ) = ∞α,β=1 cα,β ξα (r 1 ) ξβ (r 2 ).

2
Two indistinguishable particles Let us first consider a system of two identical particles. A
state of the system is described by a wave function Φ(r 1 , r 2 ) satisfying the square integrability
condition (1.3). We shall see that the identity of particles imposes strict symmetry on the
wave function.
Note that when we write r 1 , r 2 , we are assigning labels (i.e., 1 and 2) to the particles.
It is known, however, that, in quantum physics, two identical particles are intrinsically in-
distinguishable. This, in particular, means that any physically meaningful observable must
be invariant when the labels of the two particles are exchanged.7 For example, the operator8
~2
− 2m △1 , which represents the kinetic energy of the first particle, is not allowed since it certainly
~2 ~2
distinguishes between the labels of the particles. The total kinetic energy − 2m △1 − 2m △2
is an example of a legitimate observable.9 In general, we require that any observable, when
expressed in the Schrödinger representation, satisfies
 d d   d d 
A r1 , ; r2 , = A r2, ; r1, . (1.5)
dr 1 dr 2 dr 2 dr 1

We abbreviate this identity as Â12 = Â21 .

We next observe that any wave function Φ(r 1 , r 2 ) is decomposed into the symmetric and
the antisymmetric parts as

Φ(r 1 , r 2 ) = Φ+ (r 1 , r 2 ) + Φ− (r 1 , r 2 ), (1.6)

where
1
Φ± (r 1 , r 2 ) := Φ(r 1 , r 2 ) ± Φ(r 2 , r 1 ) . (1.7)
2
These wave functions, of course, satisfy

Φ± (r 1 , r 2 ) = ±Φ± (r 2 , r 1 ). (1.8)

Let us note that the two states Φ± are necessarily orthogonal (when nonzero). This is seen by
observing that
Z
hΦ+ , Φ− i = d3 r 1 d3 r 2 {Φ+ (r 1 , r 2 )}∗ Φ− (r 1 , r 2 )
Z
1
= d3 r 1 d3 r 2 |Φ(r 1 , r 2 )|2 − |Φ(r 2 , r 1 )|2
4


+ {Φ(r 2 , r 1 )}∗ Φ(r 1 , r 2 ) − {Φ(r 1 , r 2 )}∗ Φ(r 2 , r 1 ) = 0, (1.9)
R R
where we noted, e.g., that d3 r 1 d3 r 2 |Φ(r 1 , r 2 )|2 = d3 r 1 d3 r 2 |Φ(r 2 , r 1 )|2 from the change of
variables r 1 ↔ r 2 .
7
It should be stressed that we are not physically exchanging the two particles. We are not making any
changes to the physical state of the particles and merely changing the labels we assign (rather arbitrarily).
8 ∂2 ∂2 ∂2
△1 = ∂x 2 + ∂y 2 + ∂z 2 is the Laplacian corresponding to r 1 = (x1 , y1 , z1 ).
1 1 1
9
We can, of course, measure the kinetic energy of a single particle when one of the two identical particles is
in the measurement device and the other outside. Such a measurement may be modeled (somewhat artificially)
~2 ~2
by the observable − 2m △1 χ(r1 ) − 2m △2 χ(r2 ), where χ(r) is 1 inside the measurement device and 0 outside.
(The domain of the operator consists of wave functions that vanish at the boundary of the device.) Note that
we are distinguishing the two particles by their positions, not by their labels.

3
 Let Â12 = Â21 be any observable, and decompose its expectation value in the state Φ as

hΦ, Â12 Φi = hΦ+ , Â12 Φ+ i + hΦ− , Â12 Φ− i + hΦ+ , Â12 Φ− i + hΦ− , Â12 Φ+ i. (1.10)

By using the symmetry (1.5), we find that

Z
hΦ+ , Â12 Φ− i = d3 r 1 d3 r 2 {Φ+ (r 1 , r 2 )}∗ Â12 Φ− (r 1 , r 2 )
Z
1
= d3 r 1 d3 r 2 {Φ(r 1 , r 2 )}∗ Â12 Φ(r 1 , r 2 ) − {Φ(r 2 , r 1 )}∗ Â12 Φ(r 2 , r 1 )
4


+ {Φ(r 2 , r 1 )}∗ Â12 Φ(r 1 , r 2 ) − {Φ(r 1 , r 2 )}∗ Â12 Φ(r 2 , r 1 )
Z
1
= d3 r 1 d3 r 2 {Φ(r 1 , r 2 )}∗ Â12 Φ(r 1 , r 2 ) − {Φ(r 2 , r 1 )}∗ Â21 Φ(r 2 , r 1 )
4


+ {Φ(r 2 , r 1 )}∗ Â12 Φ(r 1 , r 2 ) − {Φ(r 1 , r 2 )}∗ Â21 Φ(r 2 , r 1 )
= 0, (1.11)

and similarly that hΦ− , Â12 Φ+ i = 0. We thus find for any physically meaningful observable
Â12 that
hΦ, Â12 Φi = hΦ+ , Â12 Φ+ i + hΦ− , Â12 Φ− i, (1.12)
which means that the state behaves as a statistical mixture rather than a pure state10 unless
Φ+ = 0 or Φ− = 0.
It may be reasonable to postulate that any physical system admits a description in terms
of pure states.11 This postulate implies that, for any two-particle wave function Φ(r 1 , r 2 ),
one must have either Φ+ (r 1 , r 2 ) = 0 or Φ− (r 1 , r 2 ) = 0 for any r 1 , r 2 . Equivalently, the wave
function must satisfy the symmetry

Φ(r 1 , r 2 ) = ζ Φ(r 2 , r 1 ), (1.13)

for any r 1 , r 2 ∈ R3 , where ζ is either 1 or −1.

This observation suggests that particles in our universe are classified into one of the two
classes with ζ = 1 and ζ = −1. This is indeed the case, and particles with ζ = 1 are known as
bosons, and with ζ = −1 as fermions. Electrons are fermions, and atoms may be fermions or
bosons. In the rest of the present note, we always understand that ζ is fixed to either 1 or −1
depending on the kind of particles that we are treating.

N indistinguishable particles Let us extend this consideration to a system of N identical

particles. We take a wave function Φ(r 1 , . . . , r N ) which satisfies the square integrability con-
dition (1.3). To take into account the symmetry with respect to the change of the labels of
the particles, we further assume that the wave function satisfies

Φ(r 1 , r 2 , . . . , r N ) = ζ P Φ(r P (1) , r P (2) , . . . , r P (N ) ), (1.14)

10
One can say that Φ+ and Φ− belong to distinct super-selection sectors.
11
This postulate is supported by the fact that, in the real world, any (elementary) particle is associated with
spin, whose spin quantum number is an integer for bosons and a half-odd-integer for fermions. Here we are
trying to justify (or, at least, to motivate) the symmetry requirement (1.13) without invoking spins.

4
for an arbitrary permutation P of {1, 2, . . . , N }. We here set
(
P 1 for bosons with ζ = 1,
ζ = P
(1.15)
(−1) for fermions with ζ = −1,

where (−1)P = ±1 denote the parity of the permutation P .

The inner product and the norm of these states are defined exactly as in the distinguishable
case. See (1.4). For any N = 1, 2, . . ., we denote by HN the space consisting of all such wave
functions. We call HN the N -particle Hilbert space. We also identify the 0-particle Hilbert
space H0 with C, the set of complex numbers.

Purpose of the present note What we have briefly described above provides a complete
quantum mechanical description of a system of multiple identical particles.12 One can for-
mulate various advanced problems in this language and perform calculations. It is however
awkward in some sense to first assing the labels 1, 2, . . . , N to particles (which cannot be
labeled indeed), write down the wave function, and finally impose additional symmetry con-
straint (1.14) to take into account the indistinguishability. This procedure sometimes makes
practical calculations complicated.
By using the creation and annihilation operators and the Fock representation, one can de-
scribe quantum states of indistinguishable particles and operators acting on them without ever
assigning labels to the particles. This formulation is not only theoretically beautiful but is also
useful in practice. It should be stressed, however, that this formulation is strictly equivalent
to the formulation in terms of wave functions described above; there is nothing new physically.
One might be tempted to imagine that the formulation in terms of the creation/annihilation
operators contains some novel physics compared with the wave function formulation, espe-
cially because the former is also known by the name “second quantization”. But this is a
misunderstanding.

2 Creation and annihilation operators

Let us discuss the creation and annihilation operators, which will play a central role in the
present note.

Creation operator For any ψ ∈ H1 and N = 1, 2, . . ., we wish to define the creation

operator ↠(ψ) : HN −1 → HN , which “adds” the state ψ to an arbitrary N − 1 particle state
Φ to generate a new N particle state ↠(ψ)Φ.
When N = 2 a natural definition of such an operator is
1 
(↠(ψ)ϕ)(r 1 , r 2 ) = √ ψ(r 1 )ϕ(r 2 ) + ζ ψ(r 2 )ϕ(r 1 ) , (2.1)
2

where ϕ ∈ H1 is an arbitrary single-particle state.13 Note that ψ(r 1 )ϕ(r 2 ) + ζ ψ(r 2 )ϕ(r 1 ) is
the only symmetric or antisymmetric two particle wave functions that one can generate from
12
The formalism can be automatically extended to include spin degrees of freedom.
13
We write the wave function representation Φ(r 1 , . . . , rN ) of a state Φ ∈ HN also as (Φ)(r1 , . . . , rN ).

5
ψ(r) and ϕ(r). The norm of the new state is readily found to be
Z
2 1  ∗ 
†
kâ (ψ)ϕk = d3 r 1 d3 r 2 ψ(r 1 )ϕ(r 2 ) + ζ ψ(r 2 )ϕ(r 1 ) ψ(r 1 )ϕ(r 2 ) + ζ ψ(r 2 )ϕ(r 1 )
2
= kψk2 kϕk2 + ζ |hψ, ϕi|2 . (2.2)

Suppose that ψ and ϕ are normalized, i.e., kψk = kϕk = 1. We see from (2.2) that the new
state ↠(ψ)ϕ is normalized only when hψ, ϕi = 0. Note also that, for fermions with ζ = −1,
we have ↠(ψ)ϕ = 0 when ψ = eiθ ϕ. This is a mathematical expression of the Pauli exclusion
principle, which inhibits two fermions to occupy the same single-particle state.
Let us consider the most straightforward extension of (2.1) to general N . For an arbitrary
Φ ∈ HN −1 , we define
N
† 1 X j−1
(â (ψ)Φ)(r 1 , . . . , r N ) := √ ζ ψ(r j ) Φ(r 1 , . . . , r̆ j , . . . , r N ), (2.3)
N j=1

for any r 1 , . . . , r N ∈ R3 . Here and in what follows we denote by r 1 , . . . , r̆ j , . . . , r N the sequence

without r j , i.e., r 1 , . . . , r j−1 , r j+1 , . . . , r N . Thus (2.3) with N = 3 reads

1 
(↠(ψ)Φ)(r 1 , r 2 , r 3 ) = √ ψ(r 1 ) Φ(r 2 , r 3 ) + ζ ψ(r 2 ) Φ(r 1 , r 3 ) + ψ(r 3 ) Φ(r 1 , r 2 ) . (2.4)
3
Note that the right-hand side of (2.3) or (2.4) satisfy the symmetry (1.14). For N = 1 and
Φ = 1 ∈ H0 , the definition (2.3) gives

(↠(ψ)1)(r) = ψ(r). (2.5)

It is clear from the definition (2.3) that the creation operator ↠(ψ) is linear in ψ, i.e.,
n
X  Xn
↠cℓ ψ ℓ = cℓ ↠(ψ ℓ ), (2.6)
ℓ=1 ℓ=1

for any states ψ 1 , . . . , ψ n ∈ H1 and coefficients c1 , . . . , cn ∈ C.

Annihilation operator We define the annihilation operator â(ψ) : HN → HN −1 for any

N = 1, 2, . . . by â(ψ) := {↠(ψ)}† . It should be the operator that “removes” the state ψ from
an arbitrary N particle state Φ to generate a new N −1 particle state â(ψ)Φ. Let us determine
the action of â(ψ) from the identity

h↠(ψ)Ξ, Φi = hΞ, â(ψ)Φi, (2.7)

for arbitrary Ξ ∈ HN −1 and Φ ∈ HN with N = 1, 2, . . .. From (2.3), we find

h↠(ψ)Ξ, Φi
Z N
1 X j−1  ∗
= d3 r 1 · · · d3 r N √ ζ ψ(r j ) Ξ(r 1 , . . . , r̆ j , . . . , r N ) Φ(r 1 , . . . , r N )
N j=1

6
Let us fix j, and write the sequence (r 1 , . . . , r̆ j , . . . , r N ) as (s1 , . . . , sN −1 ) (or, in other words,
set si = r i for i < j and si−1 = r i for i > j). Then we have Φ(r 1 , . . . , r N ) =
= Φ(s1 , . . . , sj−1 , r j , sj , . . . , sN −1 ) = ζ j−1 Φ(r j , s1 , . . . , sN −1 ), where we used the symmetry
(1.14). Since (ζ j−1 )2 = 1, we now have
Z N Z
3 3 1 X  ∗
= d s1 . . . d sN −1 √ d3 r j ψ(r j ) Ξ(s1 , . . . , sN −1 ) Φ(r j , s1 , . . . , sN −1 )
N j=1

By rewriting r j as q, we see that the summands with different j produce exactly the same
expression, and hence
Z √ Z
= d3 s1 . . . d3 sN −1 {Ξ(s1 , . . . , sN −1 )}∗ N d3 q {ψ(q)}∗ Φ(q, s1 , . . . , sN −1 ). (2.8)

By comparing this expression with the right-hand side of (2.7), we find for any Φ ∈ HN with
N = 1, 2, . . . that
√ Z
(â(ψ)Φ)(r 1 , . . . , r N −1 ) = N d3 q {ψ(q)}∗ Φ(q, r 1 , . . . , r N −1 ), (2.9)

which is the desired characterization of the annihilation operator. The wave function
(â(ψ)Φ)(r 1 , . . . , r N −1 ) clearly satisfies the desired symmetry (1.14). We also define

â(ψ)c0 = 0, (2.10)

for any c0 ∈ H0 = C since ↠(ψ)Φ 6∈ H0 for all possible Φ. Note that the annihilation operator
â(ψ) is antilinear in ψ, i.e.,
n
X  Xn
â cℓ ψ ℓ = (cℓ )∗ â(ψ ℓ ), (2.11)
ℓ=1 ℓ=1

for any states ψ 1 , . . . , ψ n ∈ H1 and coefficients c1 , . . . , cn ∈ C. This is clear from the expression
(2.9), but follows readily from (2.6).

Anticommutation relations We shall derive the (anti)commutation relations (2.15), (2.16),

and (2.23), which provide the essential characterization of the creation and annihilation op-
erators. Let ϕ, ψ ∈ H1 . For any Ξ ∈ HN wit N = 2, 3, . . ., we find by using (2.9) twice
that
Z
√
(â(ϕ)â(ψ)Ξ)(r 1 , . . . , r N −2 ) = N − 1 d3 q{ϕ(q)}∗ (â(ψ)Ξ)(q, r 1 , . . . , r N −2 )
p Z
= N (N − 1) d3 q d3 q ′ {ϕ(q) ψ(q ′ )}∗ Ξ(q ′ , q, r 1 , . . . , r N −2 ),
(2.12)

and
p Z
(â(ψ)â(ϕ)Ξ)(r 1 , . . . , r N −2 ) = N (N − 1) d3 q d3 q ′ {ψ(q ′ ) ϕ(q)}∗ Ξ(q, q ′ , r 1 , . . . , r N −2 ).
(2.13)

7
Since Ξ(q, q ′ , r 1 , . . . , r N −2 ) = ζ Ξ(q ′ , q, r 1 , . . . , r N −2 ), we find

â(ϕ)â(ψ)Ξ = ζ â(ψ)â(ϕ)Ξ, (2.14)

for any Ξ ∈ HN wit N = 2, 3, . . .. Let us define [Â, B̂]−ζ := ÂB̂ − ζ B̂ Â for any operators
 and B̂. For bosons, where ζ = 1, this defines the standard commutator, and for fermions,
where ζ = −1, this defines the anticommutator. Recalling that Ξ is arbitrary in (2.14), we
find the following (anti)commutation relations for the annihilation operators

[â(ϕ), â(ψ)]−ζ = 0, (2.15)

for any ϕ, ψ ∈ H1 . By taking the adjoint, we find for the creation operators that

[↠(ϕ), ↠(ψ)]−ζ = 0. (2.16)

For fermions, where we set ζ = −1, one finds by setting ϕ = ψ in (2.15) and (2.16) that

{â(ϕ)}2 = {↠(ϕ)}2 = 0 (only for fermions!) (2.17)

This is another mathematical expression of the Pauli exclusion principle.

To evaluate the (anti)commutator [â(ϕ), ↠(ψ)]−ζ is interesting but a little complicated.
We encourage the reader to explicitly write down the following derivation in the case with
N = 2.14 Take an arbitrary Ξ ∈ HN with N = 1, 2, . . .. First it is easy to see from (2.3) and
(2.9) that
N
1 X j−1
(↠(ψ)â(ϕ)Ξ)(r 1 , . . . , r N ) = √ ζ ψ(r j )(â(ϕ)Ξ)(r 1 , . . . , r̆ j , . . . , r N )
N j=1
N
X Z
= ζ ψ(r j ) d3 q {ϕ(q)}∗ Ξ(q, r 1 , . . . , r̆ j , . . . , r N ).
j−1
(2.18)
j=1

Next we simply use (2.9) to observe that

Z
√
†
(â(ϕ)â (ψ)Ξ)(r 1 , . . . , r N ) = N +1 d3 q {ϕ(q)}∗ (↠(ψ)Ξ)(q, r 1 , . . . , r N ). (2.19)

To rewrite the right-hand side, we note that (2.3) implies

(↠(ψ)Ξ)(q, r 1 , . . . , r N )
n N o
1 X
=√ ψ(q) Ξ(r 1 , . . . , r N ) + ζ j ψ(r j ) Ξ(q, r 1 , . . . , r̆ j , . . . , r N ) . (2.20)
N +1 j=1

Note that j in this expression corresponds to j − 1 in (2.3). Substituting this back to (2.19),
we get

(â(ϕ)↠(ψ)Ξ)(r 1 , . . . , r N )
Z n N
X o
3 ∗
= d q {ϕ(q)} ψ(q) Ξ(r 1 , . . . , r N ) + ζ j ψ(r j ) Ξ(q, r 1 , . . . , r̆ j , . . . , r N ) . (2.21)
j=1
14
One of the undergraduate students in our group indeed worked this out on the blackboard. The demon-
stration turned out to be quite useful.

8
From (2.18) and (2.21), we finally obtain
 
â(ϕ)↠(ψ)Ξ − ζ ↠(ψ)â(ϕ)Ξ (r 1 , . . . , r N ) = hϕ, ψi Ξ(r 1 , . . . , r N ), (2.22)

for any Ξ. This leads to the (anti)commutation relation

[â(ϕ), ↠(ψ)]−ζ = hϕ, ψi, (2.23)

for any ϕ, ψ ∈ H1 .

Examples of annihilation and creation operators Take an arbitrary complete orthonor-

mal system {ξ α }α=1,2,... of the single particle Hilbert space H1 . We then define

âα := â(ξ α ), â†α := ↠(ξ α ), (2.24)

which are annihilation and creation operators of the state ξ α . We see immediately from (2.15),
(2.16), and (2.23) that

[âα , âβ ]−ζ = [â†α , â†β ]−ζ = 0, [âα , â†β ]−ζ = δα,β , (2.25)

for any α, β = 1, 2, . . .. It is remarkable that, for bosons, these are nothing but the commutation
relations for the raising and lowering operators in a system of harmonic oscillators.
For an arbitrary x ∈ R3 , consider a state η x (r) := δ(r − x), in which the particle is
completely localized at x.15 Recall that the inner product of two such states is
Z
hη , η i = d3 r δ(r − x) δ(r − y) = δ(x − y).
x y
(2.26)

Consider the annihilation and creation operators corresponding the state η x , i.e.,

ψ̂(x) := â(η x ), ψ̂ † (x) := ↠(η x ), (2.27)

which annihilates or creates a particle at x.16 From (2.15), (2.16), (2.23), and (2.26), we see
that they satisfy the (anti)commutation relations

[ψ̂(x), ψ̂(y)]−ζ = [ψ̂ † (x), ψ̂ † (y)]−ζ = 0, [ψ̂(x), ψ̂ † (y)]−ζ = δ(x − y), (2.28)

for any x, y ∈ R3 .
Let us see how the operator ψ̂ † (x) is related to the standard operators
R 3 ↠(ϕ) with ϕ ∈ H1 .
First, takeR an arbitrary single-particle state ϕ ∈ H1 . Since ϕ(r) = d x ϕ(x) δ(r − x), we see
that ϕ = d3 x ϕ(x) η x . Then from the linearity (2.6) of the creation operator, we have
Z
â (ϕ) = d3 x ϕ(x) ψ̂ † (x).
†
(2.29)

15
In physics literature, including Feynman’s “Statistical Mechanics” [1], the state η x is denoted as |xi.
16
In Feynman’s notation [1], ψ̂(x) and ψ̂ † (x) become â(x) and ↠(x). Our notation is also standard.
The careful reader might have noticed that η x does not satisfy the square integrability condition (1.1), and is
not a proper single-particle state. We nevertheless define the operators ψ̂(x) and ψ̂ † (x) (rather formally), since
they turn out to be useful.

9
Next P
take an arbitrary complete orthonormal Psystem {ξ α }α=1,2,... of H1 , and note that17 δ(r −
∞ ∞
x) = α=1 {ξα (x)} ξα (r), and hence η = α=1 {ξα (x)}∗ ξα . Again by using the linearity, we
∗ x

find
∞
X
ψ̂ † (x) = {ξα (x)}∗ ↠(ξ α ), (2.30)
α=1
which will turn out to be useful later.
For an arbitrary wave number vector k ∈ R3 , define the standard plane-wave state18 by
uk (r) = (2π)−3/2 eik·r . (2.31)
From the standard expression of the delta function
Z 3
d w iw·z
δ(z) = e for z ∈ R3 , (2.32)
(2π)3
one finds that the inner product of two plane wave states is given by
′
huk , uk i = δ(k − k′ ), (2.33)
for any k, k′ ∈ R3 . We then define the corresponding annihilation and creation operators as19
â(k) := â(uk ), ↠(k) := ↠(uk ), (2.34)
which satisfies the (anti)commutation relations
[â(k), â(k′ )]−ζ = [↠(k), ↠(k′ )]−ζ = 0, [â(k), ↠(k′ )]−ζ = δ(k − k′ ), (2.35)
for any k, k′ ∈ R3 . From (2.31) and (2.32), we see that
Z
d3 k
δ(r − x) = 3/2
e−ik·x uk (r). (2.36)
(2π)
This equality, with the linearity (2.6), implies
Z
† d3 k
ψ̂ (x) = e−ik·x ↠(k). (2.37)
(2π)3/2

3 The Fock space representation

We now discuss the description, which is often called the Fock space representation, of many-
particle quantum mechanics in terms of the creation and annihilation operators. This formalism
is also known by the name “second quantization” formalism. But one should note that this
formalism is nothing more than a clever way of rewriting the standard quantum mechanics
based on wave functions. The unfortunate and misleading name “second quantization” simply
reflects the (constructive) confusion in early history.
For convenience let us summarize the (anti)commutation relations (2.15), (2.16), and (2.23)
obeyed by the creation and annihilation operators:
[â(ϕ), â(ψ)]−ζ = [↠(ϕ), ↠(ψ)]−ζ = 0, [â(ϕ), ↠(ψ)]−ζ = hϕ, ψi for any ϕ, ψ ∈ H1 (3.1)
17
This is the condition of completeness of {ξ α }α=1,2,...
P . To see this relation, take the inner product of the
state ξα (r) and the assumed expansion δ(r − x) = ∞ ∗
β=1 cβ ξβ (r) to find that cα = {ξα (x)} .
18
This state does not satisfy the square integrability condition (1.1).
19
In Feynman’s notation [1], â(k) and ↠(k) become â(k) and ↠(k). In this case, his notation seems to be
standard.

10
The representation of states We start by introducing a new description of the states in
HN with N = 1, 2, . . .. First, let us take 1 ∈ H0 = C, which is a “state” without any particles,
and write it as |Φvac i. Here “vac” stands for the “vacuum”. From (2.10), the vacuum state
satisfies
â(ϕ)|Φvac i = 0, (3.2)
for any ϕ ∈ H1 . This relation is repeatedly used in most applications of the Fock space
representation.
For an arbitrary ϕ ∈ H1 , we see from (2.5) that ↠(ϕ)|Φvac i is the state ϕ itself. Then
for an arbitrary ψ ∈ H1 one sees that ↠(ψ) ↠(ϕ)|Φvac i represents the state (2.1) with two
particles. Although not all states in H2 is written in the form (2.1), one can recover any state
in H2 by considering arbitrary superpositions of two-particle states of the form (2.1).
This consideration can be readily generalized to an arbitrary N = 1, 2, . . .. Let ϕ1 , . . . , ϕN ∈
H1 be arbitrary states. Then ↠(ϕ1 ) · · · ↠(ϕN )|Φvac i (if non-vanishing) is a state in HN . The
N particle Hilbert space HN is recovered by considering arbitrary superpositions of states of
the form ↠(ϕ1 ) · · · ↠(ϕN )|Φvac i.
Take arbitrary ϕ1 , . . . , ϕN , ψ 1 , . . . , ψ N ∈ H1 , and consider two states

|Φi = ↠(ϕ1 ) · · · ↠(ϕN )|Φvac i, |Ψi = ↠(ψ 1 ) · · · ↠(ψ N )|Φvac i. (3.3)

To state the following fundamental and useful result, we define, for any N × N matrix A =
(ai,j )i,j=1,...,N , X
|A|ζ := ζ P a1,P (1) a2,P (2) · · · aN,P (N ) , (3.4)
P

where the sum runs over all N ! permutations of {1, 2, . . . , N }. Thus |A|− is the standard
determinant, and |A|+ is the quantity known as permanent.

Theorem 3.1 The inner product of the two states defined in (3.3) is given by
 
 hϕ1 , ψ 1 i hϕ1 , ψ 2 i · · · hϕ1 , ψ N i 
 
 hϕ , ψ i hϕ , ψ i · · · hϕ , ψ i 
 2 1 2 2 2 N 
hΦ|Ψi =  .. .. .. ..  . (3.5)

 . . . . 

hϕ , ψ i hϕ , ψ i · · · hϕ , ψ i
N 1 N 2 N N ζ

Proof : The relation can be shown by noting that

hΦ|Ψi = hΦvac |â(ϕN ) · · · â(ϕ1 )↠(ψ 1 ) · · · ↠(ψ N )|Φvac i, (3.6)

and repeatedly using â(ϕ)↠(ψ) = ζ ↠(ψ)â(ϕ)+hϕ, ψi, which is the first of the (anti)commutation
relations (3.1), and â(ϕ)|Φvac i = 0. First, observe that
N
X
† †
â(ϕ1 )â (ψ 1 ) · · · â (ψ N )|Φvac i = ζ j−1 hϕ1 , ψ j i ↠(ψ 1 ) · · · ↠(ψ j−1 )↠(ψ j+1 ), · · · ↠(ψ N )|Φvac i
j=1
(3.7)
where ↠(ψ j ) is missing in the summand. We can repeat this process in
â(ϕN ) · · · â(ϕ1 )↠(ψ 1 ) · · · ↠(ψ N ) until all ↠and â are gone. The reader should work out the

11
case with N = 3 explicitly to check that the desired (3.5) follows. For general N , we clearly
see
X N
Y
â(ϕN ) · · · â(ϕ1 )↠(ψ 1 ) · · · ↠(ψ N )|Φvac i = η(P ) hϕj , ψ P (j) i|Φvac i, (3.8)
P j=1

and hence
X N
Y
hΦ|Ψi = η(P ) hϕj , ψ P (j) i, (3.9)
P j=1

where P is summed over all permutations of {1, . . . , N }, and η(P ) = ±1 is a certain (still
undetermined) sign factor which depends only on P . For bosons with ζ = 1, it is clear that
η(P ) = 1, and we are done. We still need to show η(P ) = (−1)P for fermions with η = −1.
This can be done by examining the sign appearing in relations like (3.7), but there is an easier
trick. Fix a permutation P0 , and note that the anticommutation relation (3.1) implies

hΦ|Ψi = (−1)P0 hΦvac |â(ϕN ) · · · â(ϕ1 )↠(ψ P0 (1) ) · · · ↠(ψ P0 (N ) )|Φvac i. (3.10)

Exactly as before the right-hand side can be rewritten as in (3.9), Q but we immediately see
(without any calculations) that the result contains the term (−1)P0 N j=1 hϕj , ψ P0 (j) i, which
has the desired sign factor. Since all the states ϕ1 , . . . , ϕN , ψ 1 , . . . , ψ N are arbitrary, and P0
is also arbitrary, we have shown that η(P ) = (−1)P .

Let us make a short remark on the wave function representation of the above basic states.
We have already seen that the two-particle √ state ↠(ψ)↠(ϕ)|Φvac i corresponds to the wave
function {ψ(r 1 )ϕ(r 2 ) + ζ ψ(r 2 )ϕ(r 1 )}/ 2. Similarly one can show that the wave function
representation of the state ↠(ϕ1 ) · · · ↠(ϕN )|Φvac i is
1 X P
Φ(r1 , . . . , r N ) = √ ζ ϕ1 (r P (1) ) ϕ2 (r P (2) ) · · · ϕN (r P (N ) ), (3.11)
N! P

where P is summed over all permutations of {1, . . . , N }.

Proof : We give inductive proof.20 The statement is already shown for N = 1 and 2. We
assume that it is valid for N − 1, i.e., the wave function representation of the (N − 1) particle
state ↠(ϕ2 ) · · · ↠(ϕN )|Φvac i is
1 X ′
Φ′ (r 2 , . . . , r N ) = p ζ P ϕ2 (r P ′ (2) ) ϕ3 (r P ′ (3) ) · · · ϕN (r P ′ (N ) ), (3.12)
(N − 1)! P′

where P ′ is summed over permutations of {2, 3, . . . , N }. By using the basic definition (2.3),
the desired wave function representation of ↠(ϕ1 )↠(ϕ2 ) · · · ↠(ϕN )|Φvac i is written as
N
1 X j−1
Φ(r 1 , . . . , r N ) = √ ζ ϕ1 (r j ) Φ′ (r 1 , . . . , r̆ j , . . . , r N )
N j=1
N
1 X X j−1 P ′
=√ ζ ζ ϕ1 (r j ) ϕ2 (r Pj′ (2) ) ϕ3 (r Pj′ (3) ) · · · ϕN (r Pj′ (N ) ). (3.13)
N ! j=1 P ′
20
The proof is not interesting. It may be skipped.

12
Here, in order to incorporate the relabelling (r 2 , . . . , r N ) → (r 1 , . . . , r̆ j , . . . , r N ) of the vari-
ables, we defined (
P ′ (k) − 1 if P ′ (k) ≤ j,
Pj′ (k) = (3.14)
P ′ (k) if P ′ (k) > j.
We can now define permutation P of {1, . . . , N } by P (1) = j and P (k) = Pj′ (k) for k =
′
2, . . . , N . Since one finds ζ j−1 ζ P = ζ P by inspection, we get the desired expression (3.11).

Slater determinant states for fermions For fermions, where we set ζ = −1, the state
↠(ϕ1 ) · · · ↠(ϕN )|Φvac i is called the Slater determinant state. This is because its wave function
representation (3.11) has a precise form of determinant. Slater determinant states, although
being very special N particle states, play important roles in the theory of many fermions. Let
us state two important properties of Slater determinant states.
The first property is about the essential role of linear independence of the single-particle
states ϕ1 , . . . , ϕN .

Theorem 3.2 A Slater determinant state |Φi = ↠(ϕ1 ) · · · ↠(ϕN )|Φvac i is nonzero if and only
if the states ϕ1 , . . . , ϕN are linearly independent.

Proof : We note that (3.5) implies hΦ|Φi = det[G], where G is the Gramm matrix associated
with ϕ1 , . . . , ϕN , i.e., an N ×N matrix defined by (G)j,k = hϕj , ϕk i. It is a well-known theorem
in linear algebra that det[G] is nonzero if and only if ϕ1 , . . . , ϕN are linearly independent.

The second property shows that a Slater determinant state ↠(ϕ1 ) · · · ↠(ϕN )|Φvac i is fully
determined by the subspace spanned by the states ϕ1 , . . . , ϕN .

Theorem 3.3 Suppose that the two sets {ϕ1 , . . . , ϕN } and {ψ 1 , . . . , ψ N } of states in H1 span
the same N -dimensional subspace of H1 . Then there is a nonzero constant c ∈ C, and we have

↠(ϕ1 ) · · · ↠(ϕN )|Φvac i = c ↠(ψ 1 ) · · · ↠(ψ N )|Φvac i, (3.15)

i.e., the two Slater determinant states are the same.

PN
Proof : From the assumption we have ϕj = k=1 βj,k ψ k with some βj,k ∈ C for any j =
1, . . . , N . Then we have
N
X
† †
â (ϕ1 ) · · · â (ϕN ) = β1,k1 · · · βN,kN ↠(ψ k1 ) · · · ↠(ψ kN ). (3.16)
k1 ,...,kN =1

Since {↠(ψ k )}2 = 0, the product ↠(ψ k1 ) · · · ↠(ψ kN ) is either zero or ±â† (ψ 1 ) · · · ↠(ψ N ).
We have thus proved that ↠(ϕ1 ) · · · ↠(ϕN ) = c ↠(ψ 1 ) · · · ↠(ψ N ) with some c ∈ C. But
Theorem 3.2 guarantees that c 6= 0.

Complete orthonormal systems We fix an arbitrary complete orthonormal system {ξ α }α=1,2,...

of H1 , and again set â†α := ↠(ξ α ). Then any state in HN is written as a superposition of the
states21
|Ξα1 ,α2 ,...,αN i := â†α1 â†α2 · · · â†αN |Φvac i, (3.17)
21
In Feynman’s notation [1] this is |α1 , α2 , . . . , αN i.

13
with α1 , α2 , . . . , αN = 1, 2, . . .. To avoid overcounting, we only consider α1 , α2 , . . . , αN such
that α1 ≤ α2 ≤ · · · ≤ αN for bosons, and α1 < α2 < · · · < αN for fermions. We find from (3.5)
that hΞα1 ,...,αN |Ξα′1 ,...,α′N i = 0 unless (α1 , . . . , αN ) = (α′1 , . . . , α′N ), and also that
Y
 nα ! for bosons,
hΞα1 ,...,αN |Ξα1 ,...,αN i = α (3.18)

1 for fermions.
P
Here nα is the number of j such that αj = α, which clearly satisfies ∞ α=1 nα = N . We here
use the convention 0!=1. Q
Thus a complete orthonormal system of HN is formed by the states ( α nα !)−1/2 |Ξα1 ,...,αN i
for any α1 , . . . , αN = 1, 2, . . . such that α1 ≤ α2 ≤ · · · ≤ αN for bosons, and by the states
|Ξα1 ,...,αN i for any α1 , . . . , αN = 1, 2, . . . such that α1 < α2 < · · · < αN for fermions. Interest-
ingly the completeness condition of these systems can be compactly written as
∞
1 X
|Ξα1 ,...,αN ihΞα1 ,...,αN | = 1̂, (3.19)
N!
α1 ,...,αN =1

where the right-hand side denotes the identity operator on HN . Here we are intentionally
overcounting the states by summing over all α1 , . . . , αN = 1, 2, . . ., without restricting their
ordering.
Both for fermions and bosons, the state (3.17) can be rewritten as
∞
Y 
|Ξα1 ,α2 ,...,αN i = (â†1 )n1 (â†2 )n2 (â†3 )n3 · · · |Φvac i = (â†α )nα |Φvac i, (3.20)
α=1

where nα is defined as above. We have nα = 0, 1, 2 . . . for bosons and nα = 0, P

1 for fermions.
Note that only a finite number of nα can be nonzero because of the constraint ∞ α=1 nα = N .
This motivates us to define another representation of the basis state 22

∞
Y 
(â†α )nα
|Γn1 ,n2 ,n3 ,... i := √ |Φvac i, (3.21)
α=1
nα !

where
P∞ the sequence n1 , n2 , n3 , . . . takes any combinations of allowed values with the constraint
α=1 nα = N . These states clearly form a complete orthonormal system of HN . Note that
nα can be interpreted as the number of particles occupying the single-particle state ξ α . The
description of many-particle states in terms of the basis (3.21) is known as the occupation
number representation.

The “second quantization” of operators We still fix an arbitrary complete orthonormal

system {ξ α }α=1,2,... of H1 , and write â†α := ↠(ξ α ). Let ô be an arbitrary operator on H1 . We
then define an operator B̂(ô) on HN with any N by
∞
X
B̂(ô) = â†α hξ α , ô ξ β i âβ . (3.22)
α,β=1

22
In Feynman’s notation [1] this is |n1 , n2 , n3 , . . .i.

14
The operator B̂(ô) is often called (again misleadingly) the “second quantization” of ô, although
there is nothing “quantized”. It is crucial to note that the operator B̂(ô) does not depend
on the choice of a complete orthonormal system. To Psee this take another arbitrary complete
orthonormal system {κµ }µ=1,2,... of H1 . Since ξ α = ∞ κ hκ , ξ i, we see from the linearity
†
P∞ † µ=1 µ µ α
(2.6) of the creation operator that â (ξ α ) = µ=1 â (κµ ) hκµ , ξ α i. Thus we have
∞
X
B̂(ô) = ↠(ξ α ) hξ α , ô ξ β i â(ξ β )
α,β=1
∞
X
= ↠(κµ ) hκµ , ξ α i hξ α , ô ξ β i hξ β , κν i â(κν )
α,β,µ,ν=1
X∞
†
= â (κµ ) hκµ , ô κν i â(κν ), (3.23)
µ,ν=1

which shows the desired independence.

For an arbitrary ϕ ∈ H1 , we find from the (anti)commutation relations (3.1) that
∞
X
B̂(ô) ↠(ϕ) = â†α hξ α , ô ξ β i âβ ↠(ϕ)
α,β=1
∞
X ∞
X
=ζ â†α ↠(ϕ) hξ α , ô ξ β i âβ + â†α hξ α , ô ξ β i hξ β , ϕi
α,β=1 α,β=1
∞
X X∞
= ↠(ϕ) â†α hξ α , ô ξ β i âβ + â†α hξ α , ô ϕi. (3.24)
α,β=1 α=1

From the linearity (2.6), we see that the second term in the right-hand side is
∞
X ∞
X 
↠(ξ α ) hξ α , ô ϕi = ↠ξ α hξ α , ô ϕi = ↠(ôϕ). (3.25)
α=1 α=1

We have thus found an interesting and useful commutation relation

[B̂(ô), ↠(ϕ)] = ↠(ôϕ), (3.26)

both for bosons and fermions. It is also a good exercise to show the relation

[B̂(ô), B̂(ô′ )] = B̂([ô, ô′ ]), (3.27)

for any operators ô and ô′ on H1 .

By using (3.26) repeatedly and noting B̂(ô)|Φvac i = 0, we see for any ϕ1 , . . . , ϕN ∈ H1 that
N
X
B̂(ô) ↠(ϕ1 ) · · · ↠(ϕN ) |Φvac i = ↠(ϕ1 ) · · · ↠(ϕj−1 ) ↠(ôϕj ) ↠(ϕj+1 ) · · · ↠(ϕN ) |Φvac i.
j=1
(3.28)
Note that the operator ô is applied to each of the N particles on the right-hand side. Thus,
when acting on HN , the “second quantization” B̂(ô) is interpreted as

B̂(ô) = ô1 + ô2 + · · · + ôN , (3.29)

15
where ôj is the copy of ô acting on the j-th particle.
A simple but important example is obtained by setting ô = 1̂, the identity operator on H1 .
Then, (3.28) reads B̂(1̂) ↠(ϕ1 ) · · · ↠(ϕN ) |Φvac i = N ↠(ϕ1 ) · · · ↠(ϕN ) |Φvac i, which means
B̂(1̂) |Φi = N |Φi for any |Φi ∈ HN . We therefore write N̂ := B̂(1̂), and call it the number
operator. From the definition (3.22), we have
∞
X ∞
X
†
N̂ = â (ξ α ) â(ξ α ) = n̂(ξ α ), (3.30)
α=1 α=1

where we defined
n̂(ϕ) = ↠(ϕ)â(ϕ), (3.31)
for any ϕ ∈ H1 with kϕk = 1. It is easily found that the eigenvalues of n̂(ϕ) are 0, 1, 2, . . .
for bosons, and 0, 1 for fermions.23 Thus n̂(ϕ) is interpreted as the operator that counts the
number of particles in the single-particle state ϕ. By using (2.30), one can also check that
Z Z
N̂ = d x ψ̂ (x) ψ̂(x) = d3 x ρ̂(x),
3 †
(3.32)

where ρ̂(x) = ψ̂ † (x) ψ̂(x) is called the density operator at x ∈ R3 .

Fock space The creation and annihilation operators ↠(ϕ), â(ϕ) and the “second quantized”
operator B̂(ô) act on the N particle Hilbert space HN with any N = 0, 1, . . .. Then one may
also regard them as operators acting on the Hilbert space

F = H0 ⊕ H1 ⊕ H2 ⊕ H3 ⊕ · · · , (3.33)

which contains states with various particle numbers. The space F is known as the Fock space.
The Fock space F consists of states of the form c0 + |Φ(1) i + |Φ(2) i + |Φ(3) i + · · · , where
c0 ∈ C = H0 is an arbitrary complex number, and |Φ(N ) i is an arbitrary state (which may be
vanishing) in HN .24 For any N 6= M , we define hΦ(N ) |Φ(M ) i = 0 for any |Φ(N ) i ∈ HN and
|Φ(M ) i ∈ HM . Equivalently, we can also say that the Fock space F consists of all possible
linear combinations of states of the form ↠(ϕ1 ) · · · ↠(ϕN )|Φvac i with any N = 0, 1, 2, . . . and
any ϕ1 , . . . , ϕN ∈ H1 .
Note that, in a system of electrons or atoms, it is unphysical to consider superpositions of
states with different particle numbers. Such states can never be realized or, more precisely, can
never be observed. The Fock space therefore should be regarded as a purely theoretical object.
Nevertheless, states in the Fock space can be useful in some problems, notably mean-field
theories for superconductivity and Bose-Einstein condensation.
23
For bosons, we only need to note that (3.1) implies [â(ϕ), ↠(ϕ)] = 1, and repeat the standard argument
for the quantum harmonic oscillator. For fermions, we first notes that {n̂(ϕ)}2 = ↠(ϕ)â(ϕ)↠(ϕ)â(ϕ) =
↠(ϕ){1 − ↠(ϕ)â(ϕ)}â(ϕ) = ↠(ϕ)â(ϕ) = n̂(ϕ). Since any eigenvalue n should satisfy the same relation
n2 = n, we see that n must be 0 or 1.
P
24
To be rigorous the space consists of all possible infinite sums such that |c0 |2 + ∞ n=1 kΦ
(n) 2
k < ∞. The
(N)
reader familiar with the standard formulation of direct sum may better interpret |Φ i in the above sum as
(0, . . . , 0, |Φ(N) i, 0, . . .).

16
4 Schrödinger equation and Hamiltonians
The Schrödinger equation (in the wave function representation) of interacting (non-relativistic)
particles is
(Ĥ0 + Ĥint ) Φ(r 1 , . . . , r N ) = E Φ(r 1 , . . . , r N ). (4.1)
The Hamiltonian of a non-interacting system is
N n
X (p̂j )2 o
Ĥ0 = + V (r̂ j ) , (4.2)
2m
j=1

where V (r) is the single-particle potential. The interaction Hamiltonian is

N
X
Ĥint = Vint (r̂ j − r̂ j ′ ), (4.3)
j,j ′=1
(j<j ′ )

where Vint (r−r′ ) denotes the two-body interaction potential. Let us rewrite these Hamiltonians
using the creation and annihilation operators.

Non-interacting Hamiltonian Comparing (4.2) with (3.29), one immediately finds that
Ĥ0 = B̂(ĥ), where
p̂2
ĥ = + V (r̂) (4.4)
2m
is the single-particle Hamiltonian. We shall derive the standard expression of Ĥ0 in terms of
ψ̂(x). Note first that
Z n ~2 o
hξ α , ĥ ξ β i = d3 x {ξα (x)}∗ − △ +V (x) ξβ (x). (4.5)
2m

Then, from (3.22), we find

∞
X
Ĥ0 = ↠(ξ α ) hξ α , ĥ ξ β i â(ξ β )
α,β=1
X∞ Z n ~2 o
= d3 x ↠(ξ α ) {ξα (x)}∗ − △ +V (x) ξβ (x) â(ξ β )
2m
α,β=1
Z n ~2 o
= d3 x ψ̂ † (x) − △ +V (x) ψ̂(x)
2m
Z Z
~2
=− d x ψ̂ (x) △ ψ̂(x) + d3 x V (x) ρ̂(x),
3 †
(4.6)
2m

where we used (2.30) to get the third line.

17
 Let us also derive the expression in terms of â(k). Substituting (2.37) into the third
expression of (4.6), we get
Z Z 3 3 ′ n ~2 o ′
3 d kd k †
Ĥ0 = d x â (k) e−ik·x
− △ +V (x) eik ·x â(k′ )
(2π)3 2m
Z Z 3 3 ′
d k d k −i(k−k′ )·x ~2 |k′ |2 †
= d3 x e â (k)â(k′ )
(2π)3 2m
Z Z 3 3 ′
3 d k d k −i(k−k′ )·x
+ d x e V (x) ↠(k)â(k′ )
(2π)3
Z 2 2 Z 3 3 ′
3 ~ |k| † d kd k
= d k â (k)â(k) + Ṽ (k − k′ ) ↠(k)â(k′ ), (4.7)
2m (2π)3
where we used the expression (2.32) of the delta function, and defined the Fourier transforma-
tion of the potential by Z
Ṽ (k) := d3 x e−ik·x V (x). (4.8)

It is illuminating to rewrite the second term as

Z 2 2 Z 3 3
3 ~ |k| † d kd q
Ĥ0 = d k â (k)â(k) + Ṽ (q) ↠(k + q)â(k). (4.9)
2m (2π)3
This expression shows that a particle with wave number k is annihilated and that with k + q
is created, with the amplitude Ṽ (q).
Let us assume that the complete orthonormal system {ξ α }α=1,2,... consists of the eigenstates
of the single-particle Hamiltonian ĥ, i.e.,

ĥ ξ α = ǫα ξ α (4.10)

for any α = 1, 2, . . ., where ǫα is the single-particle energy eigenvalue. Then it follows from
(3.26) that
[Ĥ0 , â†α ] = ǫα â†α (4.11)
and hence
N
X 
[Ĥ0 , â†α1 · · · â†αN ] = ǫαj â†α1 · · · â†αN . (4.12)
j=1

This relation, with Ĥ0 |Φvac i = 0, implies

N
X 
Ĥ0 |Ξα1 ,...,αN i = ǫαj |Ξα1 ,...,αN i, (4.13)
j=1

where the state |Ξα1 ,...,αN i is defined in (3.17). We have thus found that the state |Ξα1 ,...,αN i
is the energy eigenstate when there is no interaction between particles. The same conclusion
follows from the expression
X∞ ∞
X
Ĥ0 = ǫα â†α âα = ǫα n̂α , (4.14)
α=1 α=1

which follows from (4.10) and the definition (3.22). Here n̂α = n̂(ξ α ) is the number operator
for the state ξα . The expression (4.14) is the well-known diagonalized form of H0 .

18
 Assume, for simplicity, that the energy eigenvalues of ĥ are non-degenerate and ordered
as ǫα < ǫα+1 for α = 1, 2, . . .. Then the ground state PNof the non-interacting Hamiltonian
Ĥ0 is obtained by minimizing the energy eigenvalue j=1 ǫαj . For bosons, this is realized
when αj = 1 for j = 1, . . . , N , and hence the ground state and the ground state energy are
|Ξ1,1,...,1 i and N ǫ1 , respectively. For fermions, the minimum is realized if we take αj = j for
P
j = 1, . . . , N , and the ground state and the ground state energy are |Ξ1,2,...,N i and N α=1 ǫα ,
respectively.

Time evolution of the annihilation operator in non-interacting systems Let t be the

time variable, and consider the annihilation operator in the Heisenberg representation given
by
ψ̂(x, t) := eiĤ0 t ψ̂(x) e−iĤ0 t . (4.15)
d iĤ0 t d −iĤ0 t
Recalling that dt e = i eiĤ0 t Ĥ0 and dt e = −i Ĥ0 eiĤ0 t , we find that the time derivative
of (4.15) can be written as

∂
i ψ̂(x, t) = eiĤ0 t [ψ̂(x), Ĥ0 ] e−iĤ0 t . (4.16)
∂t
By using the (anti)commutation relations (2.28) and the expression in the third line of (4.6),
one finds that
~2
[ψ̂(x), Ĥ0 ] = − △ ψ̂(x) + V (x) ψ̂(x). (4.17)
2m
Note that this commutation relation is valid for both bosonic and fermionic systems. Substi-
tuting this into (4.16), and using (4.15), we observe that

∂ ~2
i ψ̂(x, t) = − △ ψ̂(x, t) + V (x) ψ̂(x, t). (4.18)
∂t 2m
Needless to say, the equation has exactly the same form as the Schrödinger equation for a single
particle. This similarity, which holds only in non-interacting systems, may be a reason which
caused (and still causes) the misunderstanding that ψ̂(x, t) is obtained by “second quantizing”
the wave function ψ(x, t).

Interaction Hamiltonian The interaction Hamiltonian (4.3) can be written in terms of the
creation and annihilation operators as
Z
1
Ĥint = d3 x d3 y ψ̂ † (x) ψ̂ † (y) Vint (x − y) ψ̂(y) ψ̂(x). (4.19)
2

To see that this expression correctly recovers the desired (4.3), we shall examine its action
on the state ψ̂ † (x1 ) · · · ψ̂ † (xN )|Φvac i, in which N particles have definite positions x1 , . . . , xN .
Note first that
N
X
† †
ψ̂(x) ψ̂ (x1 ) · · · ψ̂ (xN )|Φvac i = ζ j−1 δ(x − xj ) ψ̂ † (x1 ) · · · ψ̂ † (xN ) |Φvac i (4.20)
| {z }
j=1
no ψ̂† (xj )

19
where we used the (anti)commutation relations (2.28). Similarly we find
N
X
† †
ψ̂(y) ψ̂(x) ψ̂ (x1 ) · · · ψ̂ (xN )|Φvac i = ηj,j ′ δ(x − xj ) δ(y − xj ′ ) ψ̂ † (x1 ) · · · ψ̂ † (xN ) |Φvac i,
| {z }
j,j ′ =1
no ψ̂† (xj ), ψ̂† (xj ′ )
(j6=j ′ )
(4.21)
′ ′
where ηj,j ′ = ζ j−1 ζ j −1 if j > j ′ and ηj,j ′ = ζ j−1 ζ j if j < j ′ . Noting that ψ̂ † (x) δ(x − xj ) =
ψ̂ † (xj ) δ(x − xj ), we have

ψ̂ † (x) ψ̂ † (y) ψ̂(y) ψ̂(x) ψ̂ † (x1 ) · · · ψ̂ † (xN )|Φvac i

N
X
= ηj,j ′ δ(x − xj ) δ(y − xj ′ ) ψ̂ † (xj ) ψ̂ † (xj ′ ) ψ̂ † (x1 ) · · · ψ̂ † (xN ) |Φvac i
| {z }
j,j ′ =1
no ψ̂† (xj ), ψ̂† (xj ′ )
(j6=j ′ )
N
X
= δ(x − xj ) δ(y − xj ′ ) ψ̂ † (x1 ) · · · ψ̂ † (xN )|Φvac i. (4.22)
j,j ′ =1
(j6=j ′ )

Thus, by applying Ĥint in the form of (4.19) onto the state ψ̂ † (x1 ) · · · ψ̂ † (xN )|Φvac i, and per-
forming the integration, we get
N
1 X
Ĥint ψ̂ † (x1 ) · · · ψ̂ † (xN )|Φvac i = Vint (xj − xj ′ ) ψ̂ † (x1 ) · · · ψ̂ † (xN )|Φvac i, (4.23)
2 ′
j,j =1
(j6=j ′ )

which clearly coincides with the action of (4.3).

Let us finally rewrite (4.19) by using ↠(k) and â(k). Substituting (2.37) to (4.19), we get
Z 4
Y 
1
Ĥint = d3 x d3 y d3 kν e−ik1 ·x−ik2 ·y+ik3 ·y+ik4 ·x Vint (x − y) ↠(k1 ) ↠(k2 ) â(k3 ) â(k4 )
2(2π)6 ν=1

By changing the variable of integration from x to w = x − y, this becomes

Z 4
Y 
1 3 3
= d wd y d3 kν e−i(k1 +k2 −k3 −k4 )·y−i(k1 −k4 )·w Vint (w) ↠(k1 ) ↠(k2 ) â(k3 ) â(k4 )
2(2π)6 ν=1
4
Z Y 
1
= d3 kν δ(k 1 + k2 − k3 − k4 ) Ṽint (k1 − k4 ) ↠(k1 ) ↠(k2 ) â(k3 ) â(k4 ), (4.24)
2(2π)3
ν=1

where we defined the Fourier transformation of the interaction potential as

Z
Ṽint (k) = d3 x e−ik·x Vint (x). (4.25)

Note that the final expression in (4.24) shows that the total momenta before and after the
action of Vint are conserved, i.e., p1 + p2 = p3 + p4 , where pν = ~kν . By performing the

20
integration over k1 , writing k = k4 , k′ = k3 , q = k3 − k2 , and changing the variable of
integration from k2 to q, the expression (4.24) is rewritten as
Z
1
Ĥint = d3 k d3 k′ d3 q Ṽint (q) ↠(k + q) ↠(k′ − q) â(k′ ) â(k), (4.26)
2(2π)3

which shows that momentum ~q is transformed from one particle to the other with amplitude
Ṽint (q).

I wish to thank Daisuke Ida for his valuable comments and suggestions and the undergraduate stu-
dents in our theoretical physics group for their careful readings of the manuscript and useful comments.
I also thank Mizuki Yamaguchi for a useful comment on the argument in Section 1 about the symmetry
of two-particle wave functions.

References
[1] R.P. Feynman, “Statistical Mechanics: A Set of Lectures” (Westview Press, 1988).

21