Back and forth from Fock space to Hilbert space: a

guide for commuters

Andrea Beggi1 , Ilaria Siloi2 , Claudia Benedetti3 , Enrico

arXiv:1805.04552v2 [quant-ph] 31 Aug 2018

Piccinini4 , Luca Razzoli5 , Paolo Bordone5 , Matteo G. A.

Paris3,6
1 I.I.S.
Primo Levi, I-41058 Vignola, Italy.
2 Department of Physics, University of North Texas, 76201 Denton, Texas, USA.
3 Quantum Technology Lab, Dipartimento di Fisica Aldo Pontremoli, Università

degli Studi di Milano, I-20133 Milano, Italy.

4 Dipartimento di Ingegneria dell’Energia Elettrica e dell’Informazione Guglielmo

Marconi, Università di Bologna, I-40136 Bologna, Italy.

5 Dipartimento di Scienze Fisiche, Informatiche e Matematiche, Università di

Modena e Reggio Emilia, I-41125 Modena, Italy.

6 The Institute of Mathematical Sciences, CIT Campus, Taramani, Chennai,

600113, India.

Abstract. Quantum states of systems made of many identical particles, e.g.

those described by Fermi-Hubbard and Bose-Hubbard models, are conveniently
depicted in the Fock space. However, in order to evaluate some specific observables
or to study the system dynamics, it is often more effective to employ the Hilbert
space description. Moving effectively from one description to the other is thus a
desirable feature, especially when a numerical approach is needed. Here we recall
the construction of the Fock space for systems of indistinguishable particles, and
then present a set of recipes and advices for those students and researchers in
the need to commute back and forth from one description to the other. The
two-particle case is discussed in some details and few guidelines for numerical
implementations are given.
Back and forth from Fock space to Hilbert space: a guide for commuters 2

1. Introduction

The wave function describing the quantum state of a collection of identical bosons is
symmetric under the exchange of any two particles, and is thus naturally described in
the Hilbert space of symmetric functions, which is a subspace of the tensor product
of single-particle states. On the contrary, the wave function for a collection of
identical fermions is antisymmetric, which means that it must change sign when we
exchange any two particles. These wave functions are elements of a Hilbert space of
antisymmetric functions, which is another subspace of the tensor product of single-
particle states. Indistinguishability thus introduces correlations in the wave function,
and this is true even for non-interacting particles, a feature that prompted attempts
to do quantum information processing exploiting only the statistical properties of
quantum systems [1, 2].
The above facts are usually summarised by saying that for quantum particles with
a definite statistics not all available states are permitted, also in those situations where
one addresses free particles. In graduate physics courses, second quantization and the
Fock space [3, 4, 5, 6, 7], are presented as the natural framework where this constraint
may be naturally taken into account. Indeed, the Fock space is a crucial tool in the
description of systems made of a variable, or unknown, number of identical particles.
In particular, the Fock space allows one to build the space of states starting from
the single-particle Hilbert space. As a side effect, the usual introduction of the Fock
space may somehow give the impression that the Hilbert space description may be left
behind. On the other hand, the representation of operators in the Fock space is not
straightforward since the indexing of the basis set states, as well as the interpretation of
the number states in terms of particle states, are usually not trivial. A known example
is that of fermionic operators on a lattice system [8]: anti-commutation rules have to
be taken into account and additional phases appear in the components of hopping
operators from a site to another through periodic boundary conditions. Additionally,
one often encounters operators that are symmetrized or anti-symmetrized versions of
a distinguishable particle operator, e.g. the kinetic term in Hubbard models [8, 9, 10],
and in this case one may assume that those operators contain both bosonic and
fermionic features, which should be then discriminated (separated) using a suitable
transformation [11].
For all the above reasons, it is often more transparent to employ the Hilbert
space description and to study there the dynamics of a physical system [12], as done
in some recent works concerning the study of quantum walks of identical particles
[13, 14, 15, 16]. On the other hand, Fock number states appear quite naturally in
the description of systems of identical particles and thus a question arises on how and
whether we may go from Fock space to Hilbert space and vice versa with minimum
effort.
The main goal of this paper is to provide a gentle introduction to details of the
transformation rules between the different description of states and operators in the
two spaces. We start smoothly, by recalling the construction of the Fock space for
systems of indistinguishable particles, and then offer a set of recipes, guidelines, and
advices for those people interested in going back and forth from one description to
the other. We devote some attention to the two-particle case, which already contains
most of the interesting features related to indistinguishability, and briefly discuss how
to take care of the two different representations in numerical implementations. The
material presented in this paper is intended to be a concise reference about the different
Back and forth from Fock space to Hilbert space: a guide for commuters 3

representations employed in many-body physics, and it aims at being useful to students

and researchers working with systems of identical particles, ranging from photons in a
black-body cavity to interacting electrons in a lattice, and from neutrons in a neutron
star to helium atoms in a superfluid.
The paper is structured as follows. In Section 2.1 we recall few basic notions
about indistinguishable particles and the construction of the Fock space. In Section
3 we illustrate in details how to change description from Hilbert space to Fock space
and vice versa in the operator representation and the system evolution. Section 4
presents some specific applications, whereas Section 5 contains guidelines to numerical
implementations. Finally, Section 6 closes the paper with some concluding remarks.

2. Identical particles and the Fock space

2.1. From distinguishable to indistinguishable particles

Let us start by considering a collection of N identical but distinguishable particles,
each of which can be put in one of the K modes of a quantum system, e.g. the K
eigenstates of its Hamiltonian. The collective state describing the system is given by

|Ψi = |k1 i ⊗ |k2 i ⊗ ... ⊗ |kN i = |k1 , k2 , ..., kN i , (1)

or by any linear combination of states of this kind, which all belong to the K N -
dimensional N -particles Hilbert space HN = H1⊗N given by the tensor product of N
single-particle spaces H1 , each one with dimension K and basis {|ki i}i .
Let us now introduce the notion of indistinguishability [17]. We start by the
definition of the permutation operator P̂ij , whose effect is to exchange the states of
the particles i and j inside any state |{ki }i i:

P̂ij |k1 , k2 , ..., ki , ..., kj , ..., kN i = |k1 , k2 , ..., kj , ..., ki , ..., kN i . (2)

If the particles are indistinguishable, the overall state of the system |Ψiid will be given
by linear combination of states (of distinguishable particles) which is invariant under
action of P̂ij , e.g. states like |{ki }i i. It means that P̂ij |Ψiid is a state physically
indistinguishable from the previous one, i.e. they can differ only for a phase:

P̂ij |Ψiid = eiφ |Ψiid . (3)

Of course, two identical permutations must reproduce the initial state, i.e.:

P̂ij2 |Ψiid = e2iφ |Ψiid = |Ψiid , (4)

and thus the eigenvalues for P̂ij are given by eiφ = ±1.
According to the spin-statistics theorem we have two categories of identical
particles: fermions, which are characterized by half-integer spins and anti-symmetric
wavefunctions, and bosons, which are characterized by integer spins and symmetric
wavefunctions. High precision experiments have confirmed the spin-statistics and
established strict probability bounds for a violation to occur [18, 19, 20, 21].
Alternative para-statistics have been suggested earlier in the history of quantum
mechanics [22], however we are not discussing here the properties of those kind of
particles, e.g. anyons [23].
Back and forth from Fock space to Hilbert space: a guide for commuters 4

A state is symmetric or anti-symmetric under the action of P̂ij if, respectively, it

maintains or it changes its sign, i.e.:
P̂ij |ΨiF = − |ΨiF , (5)
P̂ij |ΨiB = + |ΨiB . (6)
A symmetric or anti-symmetric state can be built with the proper symmetrization
operator Ŝ or Â, acting on the distinguishable particle state |Ψi:
Ŝ |Ψi = |ΨiB , (7)
 |Ψi = |ΨiF . (8)
In general, the symmetrization operators can be built as
    r
Ŝ |ΨiB n1 !n2 !...nK ! X
|Ψi = = (±1)σ(P ) P̂ |Ψi , (9)
 |ΨiF N!
P̂

where we are applying on |Ψi all the possible distinct permutations P̂ of the N single-
particle states |ki i included in |Ψi, each one multiplied by the sign of the permutation
(−1)σ(P ) , where σ(P ) is the number of particle exchanges occurred in the permutation
P̂ , when we are dealing with fermions. Notice that any N -particle permutation P̂
can be built by composing a proper sequence of two-particle permutations P̂ij , and
the same is true for the operators  and Ŝ. The states should be then properly
normalized with the prefactor under square root, where nk indicates the number of
particles occupying the state k, i.e., the number of times that k occurs in |Ψi.
¿From the anti-symmetrization procedure for fermions, we derive the Pauli
exclusion principle, that forbids two fermions to occupy the same state: indeed, if
this was the case, e.g. for the particles i and j, we would have contemporarily
P̂ij |ΨiF = − |ΨiF , (10)
P̂ij |ΨiF = |ΨiF , (11)
where the first equality comes from Eq. (5), while the second one comes from the fact
that the two particles occupy the same state. The only possible conclusion is that our
state is the null vector.
In order to properly describe states and operators taking into account the
indistinguishability of particles we should move to the formalism of second
quantization [24], where states belong to the bosonic or fermionic Fock space F ,
that is a space containing states with a number of particles that in principle is not
fixed. The number states (basis states) of the Fock space can be represented as
|n1 , n2 , ..., nK iB(F ) , (12)

with the fundamental constraint ni ∈ {0, 1} holding only for fermions, because of
Pauli’s principle.
PK If we deal with a fixed number N of particles, there is the additional
constraint i=1 ni = N . In this last case, we are operating in the subspace of F
called FN . Indeed the Fock space is given by:
∞
M
F = FN . (13)
N =0
Back and forth from Fock space to Hilbert space: a guide for commuters 5

These number states coincide with the states of Eq. (1) except for the relabeling (and
the symmetrization). Indeed, while in the first quantization formalism we specify
for each particle i (i = 1, ..., N ) the state/mode ki that it occupies, in the second
quantization formalism we treat particles as excitation of the modes of a field, therefore
for each mode i (i = 1, ..., K) we specify how many excitation/particles ni it contains,
since we cannot distinguish among them (see Fig. 1).

Figure 1: First and Second Quantization compared for N = 4 particles and K = 6

modes available to each particle.

The total number of particles N can be changed through the application of

creation and annihilation (or destruction) operators, respectively denoted as â†i and
âi . These operators create or destroy a particle in the mode/state i, so they connect
the Fock subspaces with a different number of particles:
â†i : FN → FN +1 , (14)
âi : FN → FN −1 . (15)
In more detail, these operators are defined by the following commutation rules:
h i
[âi , âj ]∓ = â†i , â†j = 0, (16)
∓
h i
âi , â†j = δi,j , (17)
∓

where the upper sign, denoting the commutator, holds for bosons, while the lower
sign, denoting the anti-commutator, holds for fermions. It is quite important to stress,
therefore, that fermionic creation (annihilation) operators do anticommute: therefore,
when we exchange their order, we have to add a minus sign for each permutation we
perform.
Upon denoting by |0i the vacuum state, corresponding to a state with no particle,
i.e. |0, 0, ..., 0i (not to be confused with the null vector), we can build any number
state as
K
Y 1  † ni
|n1 , n2 , ..., nK iB(F ) = √ â |0i . (18)
i=1
ni ! i
Back and forth from Fock space to Hilbert space: a guide for commuters 6

From the commutation rules, we also deduce the action of the creation and annihilation
operators on the number states:
(√
† ni + 1 |n1 , n2 , ..., ni + 1, ..., nK iB
âi |n1 , n2 , ..., ni , ..., nK iB(F ) = (19)
(1 − ni )(−1)σi |n1 , n2 , ..., 1 − ni , ..., nK iF ,
(√
ni |n1 , n2 , ..., ni − 1, ..., nK iB
âi |n1 , n2 , ..., ni , ..., nK iB(F ) = (20)
ni (−1)σi |n1 , n2 , ..., 1 − ni , ..., nK iF ,

where the σi exponent is due to the anti-commuting rules and is given by

i−1
X
σi = nk . (21)
k=1

From Eqs. (19)-(20) we straightforwardly deduce that

âi |n1 , n2 , ..., ni , ..., nK i = 0 if ni = 0, (22)

â†i âi |n1 , n2 , ..., ni , ..., nK i = ni |n1 , n2 , ..., ni , ..., nK i . (23)

The last equation gives the definition of the number operator

n̂i = â†i âi , (24)

that counts the number of particles in the mode i.

Let us now consider a single-particle operator Ôj acting on the particle j. It is
apparent that we must have

P̂ij Ôj P̂ij = Ôi . (25)

Then, for a many-particle operator Ô, we will say that this operator is invariant under
particle exchange (permutation symmetry) if, for any couple of particles i and j, the
following relation holds:

P̂ij ÔP̂ij = Ô. (26)

However, the previous property does not mean that the operator Ô must be invariant
also under the action of  and Ŝ, but it implies that these symmetry operators must
commute with Ô. It means, therefore, that for Ô = Ĥ the evolution preserves the
symmetry of bosonic and fermionic states, i.e. their subspaces

FNB = ŜHN (27)

and
FNF = ÂHN (28)
are not mixed, since the Hamiltonian Ĥ and Ŝ (Â) have a common set of orthogonal
eigenstates. Therefore, this kind of operators can be used in the second quantization
formalism with identical particle states in the same way they were used in first
quantization, because they preserve the permutation symmetry. We remind here that
second quantization representation is strictly connected with the choice of the basis
{|ai i}i in which the operators â†i (âi ) are creating (destroying) particles: indeed, if
Back and forth from Fock space to Hilbert space: a guide for commuters 7

â†i (âi ) creates (destroys) a particle in the i -th eigenstate of the Hamiltonian, we can
define a change of basis and build a new set of operators b̂†i (b̂i ) that create (destroy)
particles in the i -th eigenstate of any other operator Ô (e.g. in the i -th site of a lattice
for the position operator):
 
h i
|bi i = b̂†i |0i = haj |bi i â†j  |0i .
X X
haj |bi i |aj i =  (29)
j j

2.2. Two-particle case

Here and in the rest of the article we apply the tools introduced in the previous
section to the specific case of N = 2 identical particles. Despite the small number of
particles, this example shows most of the peculiarities related to indistinguishability,
and illustrates how to deal with both Fock and Hilbert descriptions to conveniently
describe the physics of the system.
In the Hilbert space H2 (for distinguishable particles), the basis set is given by
{|i, ji}i,j . According to the symmetrization procedures described above for a 2-boson
state we have
(
√1 (|i, ji + |j, ii) for i < j
|i, jis = 2 (30)
|i, ii for i = j ,

while for a 2-fermion state the allowed basis set is given by:
1
|i, jia = √ (|i, ji − |j, ii) for i < j, (31)
2
where the constraint j > i avoids overcounting in the basis set, since

|j, iis = |i, jis , (32)

|j, iia = − |i, jia . (33)

It is easy to check that these two new basis sets are orthogonal and related to the
previous one by

{|i, ji}i,j = {|i, jis }i,j ∪ {|i, jia }i,j . (34)

Indeed, the dimension dB of the basis set for bosonic particles is given by

K(K − 1) K(K + 1)
dB = K + = , (35)
2 2
while the dimension dF of the fermionic basis set is given by

K(K − 1)
dF = , (36)
2
and

dB + dF = K 2 = dim(H2 ). (37)
Back and forth from Fock space to Hilbert space: a guide for commuters 8

Overall, the Hilbert space H2 is decomposed into two subspaces [25]: one contains
only symmetric states, while the other one only anti-symmetric states, see Fig. 2.
These subspaces are, in turn, the 2-particle restrictions of the Fock spaces for bosons
and fermions, namely F2B and F2F :

H2 = F2B ⊕ F2F . (38)

Figure 2: Hilbert space decomposition into symmetry-defined Fock subspaces for

N = 2 particles.

Each subspace is obtained, in general, by applying the proper symmetry operators

over H2 :

F2B = ŜH2 , (39)

F2F = ÂH2 , (40)

which are given by the following projectors:

X X 1
Ŝ = |i, ii hi, i| + (|i, ji + |j, ii) (hi, j| + hj, i|) , (41)
i i,j>i
2
X 1
 = (|i, ji − |j, ii) (hi, j| − hj, i|) . (42)
i,j>i
2

However, the action of Ŝ and  is not enough to write operators (or states) in the Fock
space. Indeed, if we consider the bosonic operator ÔB = Ŝ ÔŜ, in order to properly
represent it in the Fock space we still have to perform a change of basis. Therefore,
Back and forth from Fock space to Hilbert space: a guide for commuters 9

sometimes it could be useful to perform both operations in a single step, by writing

the symmetry operators in a mixed-basis representation (where the bras are states of
H2 , while kets are states of F2 ):
X X 1
Ŝ = |i, iis hi, i| + |i, jis √ (hi, j| + hj, i|) , (43)
i i,j>i
2
X 1
 = |i, jia √ (hi, j| − hj, i|) . (44)
i,j>i
2

so that we can write the operator ÔB in F2 directly as Ŝ ÔŜ † (a similar discussion
also holds for fermionic operators).
Even if in general this is not true, we observe that for the case N = 2 we have
ˆ
Ŝ + Â = I, (45)

in agreement with Eq. (34). In conclusion, a distinguishable-particle operator Ô

in H2 is invariant under particle-exchange symmetry if and only if the following
decomposition holds:

Ô = Ŝ ÔŜ + ÂÔÂ = Ôs + Ôa , (46)

i.e. the operator is the sum of its projections over the bosonic and fermionic subspaces
of H2 . This means that Ô does not mix states belonging to subspaces with different
symmetries (i.e., fermionic and bosonic), and it commutes with  and Ŝ. If this
property holds also for the Hamiltonian Ĥ, it is possible to perform a symmetrization
only over the state vectors without modifying the operators. Indeed, Ŝ and  are
projectors, Ŝ = Ŝ † and Ŝ n = Ŝ, and are orthogonal, i.e. Ŝ Â = ÂŜ = 0. The
dynamics thus conserves symmetries, and this is sufficient to get the right expectation
values:
Ĥs = Ŝ Ĥ Ŝ = Ĥ Ŝ 2 = Ĥ Ŝ, (47)
Ĥs |Ψis = Ĥ Ŝ 2 |Ψi = Ĥ |Ψis , (48)
Os = s hΨ|Ôs |Ψis = hΨ| Ŝ 2 ÔŜ 2 |Ψi = hΨ| Ŝ ÔŜ |Ψi = s hΨ|Ô|Ψis . (49)
Clearly for the 2-particle case, the evaluation of expectation values and dynamics
of the system can be conveniently obtained in the Hilbert space, starting with a
properly (anti-)symmetrized state[26], namely |Ψi(a)s . Then, with a proper reshaping
operation, observables may be recast in the Fock space, see Fig. 3. This procedure
presents advantages compared to a direct calculation in the Fock space where each
particle loses its identity. For instance, the operation of partial trace over the degrees
of freedom of one particle is straightforward in the Hilbert space, while it is more
delicate in the Fock space, where the natural basis set is the occupation number.
Moreover, the indexing of the basis in the Fock space is not trivial to handle [27],
and a convenient way to rank the basis vectors should be considered in the numerical
implementation (see Section 5). Therefore, it is often the case that the dynamics
and all the required observables are evaluated in the Hilbert space, after a proper
symmetrization of the initial state.
Back and forth from Fock space to Hilbert space: a guide for commuters 10

Figure 3: Pictorial representation of symmetrization and reshaping processes, which

transform states and operators, respectively, from Hilbert to Fock space and vice
versa. Notice that in the Hilbert space all particles have a clear identity (A or B in
the example) and may even possess well defined states (the colors in the example) if
they are distinguishable, whereas in the Fock space no particle has a clear identity
or a well defined state. Switching from Hilbert to Fock space requires to properly
(anti-)symmetrize states and to remove states with the wrong symmetry (reshaping),
while going back to Hilbert space requires to reshape both operators and states, but
the number of allowed states remains the same.

3. From Hilbert space to Fock space and vice versa

3.1. Operator representation: from Fock to Hilbert

Given the basis sets for the symmetric and antisymmetric subspaces, {|i, jis }i,j≥i
and {|i, jia }i,j>i , their union B ∗ = {|i, jis }i,j≥i ∪ {|i, jia }i,j>i is a basis set for the
whole Hilbert space H2 . As discussed in the previous section, we typically use the
distinguishable-particle basis set, hereafter labeled B = {|i, ji}i,j .
The bosonic operator ÔB may be represented with a dB × dB matrix defined on
F2B , but also with a K 2 × K 2 matrix acting on H2 . Adding dF rows and dF columns
F
full of zeros, we may extend the Fock matrix ÔB : the first dB rows and columns
involve only bosonic states, while the additional lines only the fermionic states (none,
Back and forth from Fock space to Hilbert space: a guide for commuters 11

since it is a bosonic operator):

 

ÔBF 0  .
 


H
ÔB |B ∗ = (50)


 
 
 
0 0
When the basis set is given by B ∗ , whose symmetric part coincides with the basis
H
of F2B , the representation ÔB |B∗ is valid in H2 . In order to represent operators in
the basis B of distinguishable particles, we need to reverse this transformation, that
is:
H H
ÔB |B = Ŝ † ÔB |B∗ Ŝ. (51)

The relation between the elements of a bosonic operator in the Fock (basis Bs∗ =
{|i, jis }i,j≥i ) and in the Hilbert space (basis B = {|i, ji}i,j ) is the following:

H

H 1 F


OB i,j;k,l
= hi, j| ÔB |B |k, li = OB i,j;k,l
(1 + δi,j ) (1 + δk,l ) , (52)
2
√
where  = 2 − 1.
Analogous arguments apply to the fermionic case, thus

ÔFH |B = † ÔFH |B∗ Â, (53)

where

 0 0
 

 
ÔFH |B∗
0 ÔFF
= . (54)
 
 
 

Again, the transformation of a fermionic operator in the Fock space (basis

Ba∗ = {|i, jia }i,j>i ) to the Hilbert space (basis B = {|i, ji}i,j ) is given by
1
OFH OFF i,j;k,l (1 − δi,j ) (1 − δk,l ) ςi,j ςk,l ,



i,j;k,l
= (55)
2
where ςi,j = sgn(j − i).
In the case of a 2-particle bosonic (fermionic) operator, the transformation laws
from Fock to Hilbert space of distinguishable particles are thus given by equations
(52) and (55), respectively. However, one should remind that
the Fock space elements
exist only for i ≤ j (i < j for fermions), so indices in OFF i,j;k,l must be exchanged if
i > j and/or k > l. Moreover, if i = j or k = l, the corresponding elements are zero.
Notice that if the dynamics is evaluated in the Hilbert space, the reshaping
operation needed to recast observables in the Fock space is given by the inverse
equations of (52) and (55).
Back and forth from Fock space to Hilbert space: a guide for commuters 12

3.2. Symmetrized and antisymmetrized operators of distinguishable particles

Let us consider a native Hilbert operator T̂ H , i.e. an operator which arises naturally
in the distinguishable particles Hilbert space H2 where the basis set is given by B.
Such an operator conserves parity, being invariant under particle exchange, thus it
does not mix states with different symmetries; therefore it is equivalent to the sum
of its projections with defined symmetries: T̂ H = Ŝ T̂ H Ŝ † + ÂT̂ H † , see Fig. 4.
Overall, we have that H2s and H2a are invariant subspaces for T̂ H .
Since the operator acts on states of distinguishable particles, it contains
both bosonic and fermionic components, which can be isolated with a suitable
transformation. An interesting example is the kinetic term T̂ in the Hubbard model,
which can be derived from the discretization of the laplacian terms in the Schrödinger
equation, and describes the hopping of the two particles along a chain with K sites.
In the distinguishable particles framework we have
" K
# " K
#
H
X X
T̂ = −J (|ii hi + 1| + |i + 1i hi|) ⊗ I1 + I1 ⊗ −J (|ii hi + 1| + |i + 1i hi|) ,
i=1 i=1
(56)
PK
where I1 = i=1 |ii hi| is the single-particle identity operator, and J is a scale
factor that represents the tunneling amplitude between adjacent sites and depends
on physical parameters of the system, such as the particle mass and the distance
between the discrete sites. The form of T̂ for identical particles in the Fock space is
K
T̂ F = −J (ĉ†i ĉi+1 + ĉ†i+1 ĉi ),
X
(57)
i=1

where ĉi is an annihilation operator (for bosons or fermions) and ĉ†i the corresponding
creation operator for the mode i. Given the results of the previous section, we conclude
that the representation of the bosonic/fermionic operator in the Fock space can be
obtained from T̂ H by a simple change of basis, followed by a projection over the
subspace with the required symmetry. Considering T̂ a bosonic operator, we have

T̂BF = Ŝ T̂ H Ŝ † . (58)

Let us check the previous results in some significant cases:

F †
√
s hk, k + 1|T̂B |k, kis = −J s hk, k + 1|ĉk+1 ĉk |k, kis = − 2J, (59)
H † 1 √
s hk, k + 1|Ŝ T̂ Ŝ |k, kis = √ (hk, k + 1| + hk + 1, k|) T̂ H |k, ki = − 2J, (60)
2
F
s hk, l + 1|T̂B |k, lis = −J s hk, l + 1|ĉ†l+1 ĉl |k, lis = −J, (61)
l6=k,k−1

1 1
s hk, l + 1|Ŝ T̂ H Ŝ † |k, lis = √ (hk, l + 1| + hl + 1, k|) T̂ H √ (|k, li + |l, ki)
l6=k,k−1 2 2
= −J. (62)

Analogous results may be obtained for fermions, where additional attention to the
anticommutation relation is required to handle periodic boundary conditions (PBC).
Back and forth from Fock space to Hilbert space: a guide for commuters 13

Figure 4: Representation of a parity-conserving operator: symmetry-defined subspaces

are invariant.

If a state like |k, Kia is connected to a state like |1, kia , e.g. K jumps over the border,
we should account for an additional minus sign due to the reordering of the anti-
commuting fermionic operators. Indeed (we denote the state with no particle with
|0i):

|1, kia = ĉ†1 ĉ†k |0i , (63)

ĉ†K+1 ĉK |k, Kia = ĉ†1 ĉK |k, Kia = ĉ†1 ĉK ĉ†k ĉ†K |0i
P BC

= −ĉ†1 ĉ†k ĉK ĉ†K |0i = −ĉ†1 ĉ†k |0i = − |1, kia . (64)

The same results may be obtained without any change of basis, that is by simply
applying the operator T̂ H only over the proper (anti-)symmetrized states (this is
equivalent to applying the transformation described by Ŝ or Â). However, the
spectrum of the operator Ŝ T̂ H Ŝ † is substantially the spectrum of the bosonic operator,
while the spectrum of T̂ H contains also the fermionic eigenvalues of ÂT̂ H † .
Further, we observe that the representation of T̂BF in the basis B of H2 is quite
interesting. Indeed, it contains terms like −J|i + 1, jiss hi, j|, which can be rewritten
as:

−J|i + 1, jiss hi, j| = (65)

J
= − (|i + 1, ji hi, j| + |j, i + 1i hj, i| + |j, i + 1i hi, j| + |i + 1, ji hj, i|) .
2
This suggests that the Fock operator not only produces transitions where one particle
hops from a site to a nearest-neighbour one, but can also allows the two particles to
exchange their position: the third term in brackets in the RHS of the previous equation
sees the second particle in position i jumping on the site i + 1, and then exchanging
Back and forth from Fock space to Hilbert space: a guide for commuters 14

its position with the first particle, previously located on site j. These exchange terms
are a consequence of the fact that when we rewrite the operator in the Hilbert space
of distinguishable particles, it must bear signs of the exchange symmetry, due to
the fact that it is actually acting on identical particles (in this case, bosons). The
original operator T̂ H does not contain these terms, but they appear when we apply
the transformation Ŝ T̂ H Ŝ † .

3.3. Evolution of the system and symmetrization

Let us now consider our quantum system of N = 2 identical particles, whose evolution
is ruled by the Hamiltonian Ĥ F . Their dynamics can be directly calculated in the
Hilbert space of distinguishable particles H2 : we can properly (anti-)symmetrize the
initial state and get the final state of the evolution with the required symmetry, exactly
as we had carried over the evolution in the Fock space. This can be done either
using the Fock Hamiltonian Ĥ F rewritten in the Hilbert space (see Eqs. (52) and
(55)), or using directly the equivalent Hamiltonian Ĥ H for distinguishable particles,
provided that it conserves parity (i.e., it is invariant under particle exchange). Indeed,
projecting Ĥ H over the subspaces with the proper symmetry (i.e., using Ŝ or Â) -
and/or applying it only over properly symmetrized states - is equivalent to using Ĥ F ,
as we have just seen in the previous section.

4. Expectation values and projections

4.1. Density operator

In order to simplify the notation, let us define a factor g to distinguish between bosons
(g = +1) and fermions (g = −1). If needed, we will use the subscript |...ig to denote
symmetrized or anti-symmetrized states.
The density operator is the fundamental quantity for evaluating the expectation
values of all the observables characterizing the system. The density operator can be
calculated in the Hilbert space as usual:

ρH (t) = |Ψ(t)i hΨ(t)| , (66)

and then recast in the Fock space using Eq. (52) for bosons (j ≥ i, l ≥ k):
2
ρF
i,j;k,l = ρH , (67)
(1 + δi,j ) (1 + δk,l ) i,j;k,l

and Eq. (55) for fermions (j > i, l > k):

ρF H
i,j;k,l = 2ρi,j;k,l , (68)

where, within
P the index constraint, we have ςi,j = ςk,l = 1. Notice that if
|Ψ(t)i = i,j βi,j (t) |i, ji, we should remember that for exchange symmetry

βi,j (t) = g · βj,i (t), (69)

and then we have

ρH (t) =
XX
∗
βi,j (t)βk,l (t) |i, ji hk, l| . (70)
i,j k,l
Back and forth from Fock space to Hilbert space: a guide for commuters 15

If in the Fock space we have

ρF (t) =
X X
∗
αi,j (t)αk,l (t)|i, jig g hk, l|, (71)
i,j≥i k,l≥k

it is easy to show that the proper (anti-)symmetrization of the Hilbert matrix elements

βi,j (t) + g · βj,i (t) √



 √ = 2βi,j (t) ∀ i < j
2

αi,j (t) = (72)
βi,i (t) (1 + g)

∀i = j

2
gives exactly the expected results in the Fock space (see Eqs. (67) and (68)).

4.2. Occupation numbers

The expectation value of the number operator n̂k , corresponding to the average
number of particles in the mode k, can be calculated as follows:

hnk i = Tr[ρF (t)n̂k ] = Tr[ρF (t)ĉ†k ĉk ]

F † F †
X X
= g hi, i|ρ (t)ĉk ĉk |i, iig + g hi, j|ρ (t)ĉk ĉk |i, jig
i i,j>i
√ √
= 2 2ρF F † F †
X X
k,k;k,k (t) + g hi, k|ρ (t)ĉk ĉk |i, kig + g hk, j|ρ (t)ĉk ĉk |k, jig
i<k j>k

2ρF ρF ρF
X X
= k,k;k,k (t) + i,k;i,k (t) + k,j;k,j (t). (73)
i<k j>k

Upon recalling that in the Hilbert space the symmetry exchange requires

ρH H H
i,j;k,l (t) = g · ρj,i;k,l (t) = g · ρi,j;l,k (t), (74)

we may rewrite hnk i in the Hilbert space, also using Eq. (68), as follows:

hnk i = 2ρH 2ρH 2ρH

X X
k,k;k,k (t) + i,k;i,k (t) + k,j;k,j (t)
i<k j>k
 

= 2 ρH ρH ρH
X X
k,k;k,k (t) + i,k;i,k (t)
=2 i,k;i,k (t), (75)
i6=k i

since ρH 2 H H
k,j;k,j = g ρj,k;j,k = ρj,k;j,k .

4.3. Entropies
Given a quantum system, it is natural to ask how to measure the amount of
quantum correlations in it. Besides representing an intriguing trait of quantum
mechanics, quantum entanglement has turned into a fundamental resource for
quantum information theory and quantum computing, since it can be used to
implement protocols and tasks that could not be accomplished within the classical
framework [28]. The term entanglement refers to an intrinsic relation between
subsystems of a composite quantum system: in an entangled state, each subsystem
Back and forth from Fock space to Hilbert space: a guide for commuters 16

cannot be described independently of the state of the other one, or, in other words,
what we know (ignore) about A, is what we know (ignore) about B, and vice-versa.
For a system composed of two subsystems A and B (bipartite) described by the density
matrix ρAB , the entanglement among A and B can be quantified in different ways
[29], depending on the reduced state of the subsystem ρA(B) and the size dA(B) of the
subsystem A(B). In the case of pure states, the entanglement can always be measured
with the von Neumann entropy S(ρA ) = −ρA log2 ρA , with S(ρA ) = S(ρB ), [30].
Here we consider a compound system described the total density matrix ρSE , where
a quantum system S is coupled to an external bath E, acting as a noise source. In
this case the entanglement between system and environment gives a measure of the
decoherence, which quantifies the loss of coherence in the quantum correlations of
the system [31]. In this picture, decoherence can be evaluated via the von Neumann
entropy of the quantum system S(ρS ), with ρS = TrE ρ. Whenever the quantum
system S contains indistinguishable particles, this quantity should be evaluated in the
Fock space, which is the natural space for the system since it accounts for the exchange
symmetry. Indeed, we have
1 1
S(ρF
S )=− Tr[ρF F
S ln ρS ] > − Tr[ρH H H
S ln ρS ] = S(ρS ), (76)
ln(dg ) ln(K 2 )
since ρHS and ρF S have the same eigenvalues: they only differ for a unitary
transformation, and the additional eigenvalues of ρH S are zeros that do not contribute
to the entropy. We therefore conclude that Tr[ρH H F F
S ln ρS ] = Tr[ρS ln ρS ], and
F H
the only difference between S(ρS ) and S(ρS ) is given by different normalization
(dg = K(K+g)
2 < K 2 ): so we conclude that S(ρHS ) underestimates the loss of quantum
correlations with respect to S(ρF
S ). The reason is intuitively obvious: since the system
always possesses a residual amount of correlations due to exchange symmetry, these
correlations are seen as quantum correlations by the entropy of the Hilbert space,
which is devised for distinguishable particles. On the other hand, they are correctly
not counted by the von Neumann entropy evaluated in the Fock space. Indeed, they
are not genuine quantum correlations - like entanglement or quantum discord [32]-
which may be exploited to perform quantum information tasks.

5. Guidelines for Numerical implementation

5.1. Base ordering and indexing

One of the main problems in implementing numerically the calculations of operators
is the different indexing in Hilbert and Fock spaces. This situation is made more
involved by the differences between allowed states for fermions and bosons. Let us see
this with an example. Let us consider a system with N = 2 identical particles, which
can occupy K = 4 sites, or modes. The allowed states in the Hilbert and Fock spaces
are given in Table 1.
Since any vector or matrix must be indexed with a progressive index m, we have
to define a global index m that depends on the single-particle states i and j and follows
the correct ordering when basis set states |mi are |i, ji, |i, jis or |i, jia . It turns out
that we have
H2 : m = K(i − 1) + j, (77)
B/F
F2 : m = K(i − 1) + j − s(g, i) , (78)
Back and forth from Fock space to Hilbert space: a guide for commuters 17

Space Basis set Dimension

|1, 1i |1, 2i |1, 3i |1, 4i
H2 |2, 1i |2, 2i |2, 3i |2, 4i
16
(distinguishable) |3, 1i |3, 2i |3, 3i |3, 4i
|4, 1i |4, 2i |4, 3i |4, 4i
|1, 1is |1, 2is |1, 3is |1, 4is
F2B |2, 2is |2, 3is |2, 4is
10
(bosons) |3, 3is |3, 4is
|4, 4is
|1, 2ia |1, 3ia |1, 4ia
F2F
|2, 3ia |2, 4ia 6
(fermions)
|3, 4ia

Table 1: Basis sets for Hilbert and Fock spaces of N = 2 identical particles, which can
occupy K = 4 sites.

where s(g, i) is a correction term that takes into account the fact that states with
indices exchanged must not be counted again in Fock space, and also that states with
identical indices are forbidden for fermions (g = −1). From an intuitive point of view,
we can think that i and j in |i, ji are two numbers living on a ring ZK = {1, 2, ..., K}:
i plays the role of the tens, while j plays the role of units and, overall, we have
m = K(i − 1) + j. By a simple combinatorial reasoning we find:

i(i − g)
s(g, i) = . (79)
2
Indeed, for a fixed value of i, denoted as i∗ , the number of forbidden states which
must be subtracted from m is
∗ ∗
i
X −1
iX
∗
B: card{|i, ji | i ≤ i ∧ j < i} = (i − 1) = i, (80)
i=1 i=0
∗ ∗
i
X i
X
∗
F : card{|i, ji | i ≤ i ∧ j ≤ i} = i= i. (81)
i=1 i=0
Pn
In both cases, we calculate the result with the Gauss formula i=0 i = 21 n(n+1). One
can easily verify that {|i, jis(a) | i ≤ i∗ ∧ j (≤) i} are exactly the states not appearing
in Table 1 since forbidden.
So, according to Eqs. (78) and (79), the state |3, 4i, e.g., is the basis state
|mi = |12i in H2 , the basis state |mi = |9i in F2B , and the basis state |mi = |6i in
F2F . This allows us to scan all the elements of vector states and operators in terms of
the single-particle states i and j, and it also lets us to switch easily from their Hilbert
representation to the Fock one and vice versa.
Back and forth from Fock space to Hilbert space: a guide for commuters 18

5.2. Reshaping cycle

Now, it is worth noting that, in order to fill-in properly the elements of an operator
B/F
O in the space F2 , starting from the corresponding operator in the Hilbert space
(the so-called reshaping operation), we must use cycles like
for i=1,N
for j=i+∆,N
for k=1,N
for l=k+∆,N
OF (i, j; k, l) = OH (i, j; k, l) · ...
end
end
end
end
where the correction
1−g
∆= (82)
2
is 0 for bosons (i.e., states with i = j are allowed) and 1 for fermions (i.e., states with
i = j are forbidden).

5.3. Computational and storage considerations

Working in the Hilbert space offers an obvious advantage from the physical point of
view, since one has a clear identification of the degrees of freedom associated to each
particle, and a better indexing of states. On the other hand, a couple of issues arises
from the point of view of numerical implementation. The first is linked to the larger
dimension of the space and may be properly addressed by opportunely inverting Eqs.
(78) and using reshaping cycles as those presented in Section 5.2, so that the number
of operations is not significantly larger in the Hilbert space [33]. Let us define the
function
K−∆

X
m − 1 − n for r ∈ N+

fKg (r) := n=K−∆−r+1

m−1

for r = 0
r
= m − 1 − (2K + g − r) for r ∈ N , (83)
2
where ∆ is defined in Eq. (82), g = ± 1 for bosons (fermions), and K is the number
( )

of modes of the quantum system. The expression in Eq. (83), which is in principle
the result for r ∈ N+ , returns m − 1 for r = 0, thus it already summarizes the two
distinct cases. Let r̄Kg be the greater value of r ∈ {0, 1, ..., K − 1} such that fKg (r) ≥ 0,
i.e.
r̄Kg = max{r | r ∈ {0, 1, ..., K − 1} ∧ fKg (r) ≥ 0}. (84)
Hence, the inverse formulae of Eqs. (78) are given by

 ig (m) = 1 + r̄g
K K
(85)
 j g (m) = ∆ + ig (m) + f g (r̄g ) .
K K K K
Back and forth from Fock space to Hilbert space: a guide for commuters 19

The other and major issue is instead related to the storage of the matrix elements of
states and operators, since in both cases using the Hilbert space description amounts
to storing several empty cells, i.e. those corresponding to states with the wrong
symmetry. This problem may be addressed by exploiting the above mapping, and
also noticing that the involved matrices are often sparse, e.g. when systems with only
nearest-neighbour interactions are considered [34], so that sparse matrix declarations
and algorithms may be exploited to reduce the storage space.

6. Concluding remarks

In graduate physics courses, second quantization and the Fock space are presented as
the natural framework to deal with quantum systems made of many indistinguishable
particles, leaving the impression that the Hilbert space description may be left behind.
While this is certainly true for the description of quantum states of those systems, the
evaluation of some specific observable or the study of the system dynamics may be
often more conveniently pursued using the Hilbert space description.
A research-oriented teaching of these topics should therefore reflect the
importance of both descriptions, and provide tools to connect them in the most
straightforward way. To this aim, we have provided here a gentle and self-contained
introduction to details of the transformation rules between the different description of
states and operators in the two spaces. In particular, we have devoted some attention
to the two-particle case, since this already contains most of the interesting features
related to indistinguishability. The paper aims at being a concise reference about the
different representations for students and researchers working with systems made of
many identical particles, especially those interested in numerical approaches to the
system dynamics.

Acknowledegements

This work has been supported by JSPS through FY2017 program (grant S17118) and
by SERB through the VAJRA award (grant VJR/2017/000011). PB and MGAP are
members of GNFM-INdAM.

References
[1] Y. Omar, N. Paunkovic, S. Bose, and V. Vedral Phys. Rev. A 65, 062305 (2002).
[2] S. Bose, A. Ekert, Y. Omar, N. Paunkovic, and V. Vedral Phys. Rev. A 68, 052309 (2003).
[3] V. Fock, Z. Phys. 75, 622 (1932).
[4] A. L. Fetter, J. D. Walecka, Quantum Theory of Many Particle System (McGraw-Hill, New
York, 1971).
[5] M. C. Reed, B. Simon, Methods of Modern Mathematical Physics, Vol. II (Academic, New York,
1975).
[6] G. D. Mahan, Many Particle Physics (Plenum, New York, 1981).
[7] J. W. Negele, H. Orland, Quantum many-particle systems (Addison-Wesley, Redwood, 1988).
[8] J. Hubbard, Proc. R. Soc. A 276, 238 (1963).
[9] H. A. Gersch, G. C. Knollman, Phys. Rev. 129, 959 (1963).
[10] M. P. A. Fisher, P. B. Weichman, G. Grinstein, D. S. Fisher, Phys. Rev. B 40, 546 (1989).
[11] F. Töppel, A. Aiello, Phys. Rev. A 88, 012130 (2013).
[12] It may be useful to remind here an aphorism by Asher Peres:“Quantum phenomena do not occur
in a Hilbert space. They occur in a laboratory”.
[13] C. Benedetti, F. Buscemi, P. Bordone, Phys. Rev. A 85, 042314 (2012).
[14] A. Beggi, F. Buscemi, P. Bordone, Quantum Inf. Process. 15, 3711 (2016).
Back and forth from Fock space to Hilbert space: a guide for commuters 20

[15] I. Siloi, et al., Phys. Rev. A 95, 022106 (2017).

[16] A. Beggi, L. Razzoli, P. Bordone, M. G. A. Paris, Phys. Rev. A 97, 013610 (2018).
[17] S. Chaturvedi, S. Biswas, Resonance 19, 45 (2014).
[18] K. Deilamian et al., Phys. Rev. Lett. 74, 4787 (1995).
[19] D. DeMille et al., Phys. Rev. Lett. 83, 3978 (1999).
[20] G. Modugno et al., Phys. Rev. Lett. 81, 4790 (1998).
[21] D. Javorsek II et al., Phys. Rev. Lett. 85, 2701 (2000).
[22] H. S. Green, Phys. Rev. 90, 270 (1953).
[23] F. Wilczek, Phys. Rev. Lett. 49, 957 (1982).
[24] V. S. Shchesnovich, arXiv:1308.3275 [cond-mat.quant-gas] (2013).
[25] This decomposition holds only for N = 2, since for higher dimensions there is an additional
subspace N containing states that do not possess any symmetry, but are required to produce
a complete basis set for HN .
[26] Considering that [Ĥ, Ŝ] = 0, we can obtain the same results with an alternative approach:
we begin with a non-symmetrized state |Ψi, evolve it with the distinguishable-particle
Hamiltonian Ĥ, and then perform the symmetrization at the end of the evolution, since
Ĥs |Ψis = Ĥ |Ψis = Ĥ Ŝ |Ψi = Ŝ Ĥ |Ψi.
[27] J. M. Zhang, R. X. Dong, Eur. J. Phys. 31, 591 (2010).
[28] R. Horodecki, P. Horodecki, M. Horodecki, K. Horodecki, Rev. Mod. Phys. 81, 865 (2009).
[29] G. Otfried, G. Tóth, Phys. Rep. 474, 1 (2009).
[30] M. A. Nielsen, I. Chuang, Quantum computation and quantum information (Cambridge Series
on Information and the Natural Sciences, 2004).
[31] W. H. Zurek, Rev. Mod. Phys. 75, 715 (2003).
[32] H. Ollivier, W. H. Zurek, Phys. Rev. Lett. 88, 017901 (2001).
[33] The issue concerning the larger number of operations cannot be avoided when using built-in
subroutines, such as matrix multiplication and diagonalization.
[34] E. Piccinini, C. Benedetti, I. Siloi, M. G. A. Paris, P. Bordone, Comp. Phys. Comm. 215, 235
(2017).