Home Search Collections Journals About Contact us My IOPscience

Entanglement in N-harmonium: bosons and fermions

This content has been downloaded from IOPscience. Please scroll down to see the full text.

2014 J. Phys. B: At. Mol. Opt. Phys. 47 195503

(http://iopscience.iop.org/0953-4075/47/19/195503)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 130.108.121.217
This content was downloaded on 06/10/2014 at 15:11

Please note that terms and conditions apply.

 Journal of Physics B: Atomic, Molecular and Optical Physics

J. Phys. B: At. Mol. Opt. Phys. 47 (2014) 195503 (14pp) doi:10.1088/0953-4075/47/19/195503

Entanglement in N-harmonium: bosons and

fermions
C L Benavides-Riveros1,2, I V Toranzo3 and J S Dehesa3
1
Departamento de Física Teórica, Universidad de Zaragoza, 50009 Zaragoza, Spain
2
Instituto de Biocomputación y Física de Sistemas Complejos, Mariano Esquillor (Edificio I+D), 50018
Zaragoza, Spain
3
Instituto Carlos I de Física Teórica y Computacional and Departamento de Física Atómica Molecular y
Nuclear, Universidad de Granada, 18071 Granada, Spain

E-mail: carlosbe@unizar.es

Received 22 May 2014, revised 22 August 2014

Accepted for publication 27 August 2014
Published 24 September 2014

Abstract
The ground-state entanglement of a single particle of the N-harmonium system (i.e., a
completely integrable model of N particles where both the confinement and the two-particle
interaction are harmonic) is shown to be analytically determined in terms of N and the relative
interaction strength. For bosons, we compute the von Neumann entropy of the one-body reduced
density matrix by using the corresponding natural occupation numbers. A critical number, Nc, of
particles exists, and below it, for positive values of the coupling constant, the entanglement
grows when the number of particles increases; the opposite occurs for N > Nc . For fermions, we
compute the one-body reduced density matrix for the closed-shell spinned case. In the strong
coupling regime, the linear entropy of the system decreases when N grows. For fixed N, the
entanglement is found (a) to decrease (increase) for negatively (positively) increases values of
the coupling constant, and (b) to grow when the energy increases. Moreover, the spatial and spin
contributions to the total entanglement are found to be of comparable size.
Keywords: many-body systems, quantum information, entanglement, harmonium

(Some figures may appear in colour only in the online journal)

1. Introduction multiconfiguration, Hartree–Fock-type trial wave function,

which lacks physical reality.
The most precise determination of the properties of finite The application of quantum information ideas and tech-
many-electron systems is usually calculated using a full niques in electronic structure theory has recently led to the
configuration interaction (FCI) method, where the solution conclusion that the electron correlation is closely related to
of the corresponding Schrödinger equation in a given the entanglement of electrons. Indeed, it has been proved that
one-electron basis is expressed in terms of a linear while the single Slater determinant in the monoconfigura-
combination of all possible Slater determinants. Its tional Hartree–Fock approximation is a disentangled state, the
application is naturally reduced to a bunch of small wave function of the multiconfiguration Hartree–Fock
systems because of the enormous number of determinants approximations (such as FCI) accounts for the entanglement
involved [1, 2]. effects. Therefore, entanglement plays an essential role not
Let us highlight that for all FCI approaches, the corre- only in quantum communication between parties separated by
lation effects, which remain solely in the wave function, are macroscopic distances (see, e.g., [3, 4]), but also in char-
not described by any correlation operator. Moreover, these acterizing quantum correlations at short distances. The latter
effects are usually numerically computed as a difference of problem, where one should necessarily consider the indis-
two variational energy bounds. Thus, up to now the electron tinguishable character of the involved particles, has received
correlation has been widely and implicitly believed to be a relatively little attention until recently [5–10]. This has been a
purely methodical effect coming from the inadequate use of a serious omission because of its relevance in quantum

0953-4075/14/195503+14$33.00 1 © 2014 IOP Publishing Ltd Printed in the UK

J. Phys. B: At. Mol. Opt. Phys. 47 (2014) 195503 C L Benavides-Riveros et al

information processing in various physical systems (see, e.g., previously computed [26]. We obtain for the first time an
[5, 11]) it also allows us to gain deeper insight into non- explicit expression of the one-body reduced density matrix for
classical correlations of atomic and molecular systems, and to the closed-shell spinned fermionic case. This study will allow
fully understand the course of their dissociation processes and us to show that some entanglement features of finite many-
chemical reactions [10–12]. particle systems can be understood to a certain extent by
The main difficulty, however, stems from the fact that the purely kinematical considerations. The N-particle harmonium
Schrödinger equation of most quantum many-body systems is a completely integrable system with an arbitrary number of
cannot be solved analytically. Even numerically, the deter- particles, N, where both the confinement and the two-particle
mination of the wavefunction is, in general, a serious pro- interaction are harmonic. This model has been implicitly used
blem. In the last three years, an intense effort has been made to a certain extent to study cold atoms [27], to determine area
to determine the entanglement of some real atomic and laws of the ground-state entanglement of some one-dimen-
molecular species, such as helium-like atoms [13–17], a few sional chains [28], and to gain further insight into numerous
processes of diatomic molecules [18], and elementary che- phenomena of a variety of physical systems up to black holes
mical reactions [19]. This research has basically focused on (see, e.g., [21, 29–35]).
the entanglement of bipartite systems, mainly because the The paper is structured as follows. First, in section 2 we
characterization of this phenomenon for systems of many briefly formulate the quantum-mechanical problem of the N-
indistinguishable constituents is much less known, even at the particle harmonium, showing its separability by using the
notional level of entanglement measurement [20]. appropriate set of normal coordinates [36] and fixing our
Thus, the quantification of the entanglement of bound notational settings. In section 3 we discuss the general
states for model systems enabling analytic solutions of the mathematical structure of the one-body reduced density
associated Schrödinger equation is a promising way to matrix in the bosonic case, obtaining its explicit expression.
investigate correlation phenomena. Indeed, entanglement Moreover, we compute and discuss the analytical expression
between the constituents of any bound system is most con- of the von Neumann entropy of the one-body reduced density
veniently analyzed in such models, enabling us to relate it to matrix for the N-boson harmonium. Further, in sections 4 and
bosonic or fermionic characteristics. Up to now, however, the 5, we obtain explicit expressions for the one-body reduced
density matrix of the spinned N-fermion harmonium, we and
entanglement of certain models of two bound electrons is all
analyze the linear entropy of the one-body reduced density
that has been determined. We refer to the two-harmonium (or
matrix, both with and without the spin degree of freedom. The
Moshinsky) [21–25], Crandall, and Hooke [13] atoms. In all
paper ends with concluding remarks and two appendices.
these models, the electron confinement is harmonic, and the
electron-electron interaction is of a harmonic (two-harmo-
nium), r12−1 (Crandall), or Coulombic (Hooke) type. All of
2. The N-harmonium problem
them show that when the spin degree of freedom and the
indistinguishability of electrons are taken into account, new The N-harmonium model is a system of N interacting parti-
entanglement aspects [24] are encountered as compared to the cles (fermions or bosons) which interact harmonically in a
model of distinguishable particles, although some further three-dimensional harmonic well. It is characterized by the
clarification is needed. Hamiltonian
It is also worth noting that some entanglement features of
N N N
the two-harmonium atom are also qualitatively reproduced by 1 2 k 2 δ
the other two models, which give a good description for
H= ∑ pi + ∑ ri + ∑ rij2, (1)
2 i=1 2 i=1 2i<j
certain two-fermion systems (e.g., two electrons in a quantum
dot is approximately described by the Hooke model)— where rij := | ri − rj |, k is the coupling constant of the har-
namely the growth of the entanglement when either the monic well, and δ is the coupling of the harmonic interaction
relative strength or the excitation energy is increases. It is between the particles. The treatment of harmonically inter-
most interesting that recently, the entanglement of some real acting bosons by means of this Hamiltonian represents the
helium-like atoms has been numerically shown to have an first exact solution of an N-particle system using only con-
increases dependence on the energy. This has been done both ditions on the reduced space of two-particle density matrices
by using very accurate one-electron Hylleraas–Kinoshita type [37]. This system can be expressed in a separable form (i.e.,
basis functions [14, 17] and some Gaussian or Slater-type as a system of uncoupled oscillators) using the set of normal
orbital basis sets [6, 15]. It is also observed from the models coordinates {ξ1 ,..., ξ N }, given by
that the entanglement decrease of these systems, in terms of N
1
the nuclear charge Z, can also be understood as the result of ξ N := ∑ri and
the relative decrease of the electron-electron interaction. N i=1
In this paper, we will analytically study the entanglement m
1
of the one-body reduced density matrix of the N-harmonium ξ m := ∑ ( ri − rm +1), (2)
m (m + 1) i=1
model for bosons and fermions. Von Neumann (for bosons)
and linear (for fermions) entropies will be used. The one-body with m ∈ {1, … , N − 1}. This is an orthogonal transforma-
reduced matrix for the spinless fermionic case has been tion of the position variables. A similar change of coordinates

2
J. Phys. B: At. Mol. Opt. Phys. 47 (2014) 195503 C L Benavides-Riveros et al

for momenta results in a canonical transformation, preserving Let us propose that a critical value exists for the number of
the symplectic form. Let us call the new set of momenta {Ξm }. particles. Below this critical value (around 3.5), we will show
A direct calculation shows that that for positive values of the coupling constant, δ, the first
N −1 N N occupation number decreases (and consequently the von
1
∑ ξm2 = ∑ rij2 = ∑ rm2 − ξ N2 , (3) Neumann entropy grows) as the number of particles increases.
m =1
Ni<j m =1 Above this value, the first occupation number increases as the
number of particles increases. Moreover, the first occupation
so that the Hamiltonian (1) can be expressed in the following number tends to 1 (and the entropy tends to 0) in the limit
separable form: when the number of particles tends to infinity.
N −1 For fermions, we use the one-body reduced density
1 2 1
H = N + ∑  m, where  N = ΞN + ω2ξ N2 and matrix to compute an explicit expression for both its purity
m =1
2 2
and its linear entropy. Let us propose that we will show that,
1 1 in the region of negative coupling δ, the entropy grows when
 m = Ξm2 + μ2 ξ m2, (4)
2 2 the number of particles increases. A similar situation is
where ω2 := k , and μ2 := k + Nδ , depending on the number observed in the attractive case for small values of the coupling
of particles. Then, the physical solutions of the associated constant. For large values of δ k , the situation is the opposite:
Schrödinger equation (i.e., the wave functions Ψ (x1 , ⋯ , xN )), When particles are added to the system, the entropy decrea-
can be readily obtained. Here, let us first note that (a) there is a ses. In both bosonic and fermionic cases, we will show that it
ground state whenever μ2 > 0 , and (b) the particles are no is possible to calculate these two measures of entanglement as
longer bound if the relative interaction strength a function of the coupling constant, δ k , and the number of
particles.
δ 1
⩽− . (5)
k N
On the other hand, let us comment here that the entan- 3. The N-boson harmonium: von Neumann entropy
glement problem is formulated in terms of reduced density
matrices. Either directly from the resulting one-particle In this section, we determine the von Neumann entropy of the
reduced density matrix or, since the particles are assumed to N-boson harmonium ground state in terms of N and the
interact pairwise, in terms of the two-particle density matrix, relative interaction strength, δ k . Taking into account the
Gaussian character of the ground-state oscillator wave func-
ρ ( x1, x2 ; y1 , y2 ) := ∫ dx3 ⋯dxN Ψ ( x1, x2 , x3 , ⋯ , xN ) tion, the ground state distribution on phase space is char-
acterized by the Wigner N-body density function,
× Ψ ( y1 , y2 ,, x3 , ⋯ , xN ) , (6)
1 −2 N ω −2 ∑N −1 m / μ
which carries all the necessary information required for cal- Wgsb ( r1, … , rN ; p1 , … , pN ) = e e m=1
π 3N
culating the quantum-mechanical properties of the whole
1 ⎡ N
system. The symbol x stands for spatial and spin coordinates, = exp ⎢ −(ω − μ) ξ N2 − μ ∑ri2
x := (r, ς ). In particular, the ground-state energy of the π 3N ⎢⎣ i=1
system can be computed by minimizing a simple linear N ⎤
functional of ρ. In passing, let us mention that the N-repre- ω−μ 2 1
+ ΞN − ∑pi2 ⎥ . (8)
sentability problem for this matrix has proved to be a major ωμ μ i = 1 ⎥⎦
challenge for quantum chemistry [38]. The one-body density
matrix, which is the basic variable in reduced density matrix The energy of the ground state is the total sum of the con-
functional theory [39], is then given as tributions of each individual oscillator (i.e.,
3
Egsb = 2 [ω + (N − 1) μ]). On the other hand, the one-body
ρ1 (x ; y):= ∫ ρ ( x, x2 ; y, x2 )dx2 . (7) density factorizes as a product of two separable quantities:
By means of the spectral theorem, ρ1 (x ; y) can be decom-
posed in terms of its natural spin orbitals, {ϕi (x)}, and its
d1b (r ; p) := ∫ Wgsb ( r, r2, …, rN ; p, p2 , …, pN )
eigenvalues, {ni }; the natural occupation numbers: N

ρ1 (x ; y) = ∑i ni ϕi (x) ϕi (y), where ∑i ni = 1, with (for fer- × ∏ drm dpm

mions) 0 ⩽ ni ⩽ 1 N . m =2
2 − 1 p2
For the case of an assembly of bosons, we will compute = e−μr μ Δr (r , μ , ω , N ) Δ p (p , μ , ω , N ). (9)
the von Neumann entropy of the one-body reduced density
matrix, explicitly finding the occupation numbers of the It can be shown that the Δ-function fulfills the following
system, as described in the next section. Recently, there has recursion relation:
been a renewed interest in formulating the reduced density ⎡
ω − μ⎛ ⎞2
N
exp ⎢ −
matrix theory using Wigner quasidistributions [25, 40, 41]. 1
r
Δ (r , μ , ω , N ):= ∫ ⎜r + ∑ rm ⎟⎟
Our treatment for bosons will use the Wigner function in π 3N 2 ⎢ N ⎜⎝ m =2 ⎠
phase space, which is a natural procedure for these systems. ⎣

3
J. Phys. B: At. Mol. Opt. Phys. 47 (2014) 195503 C L Benavides-Riveros et al

N ⎤ N
1 1 the Gaussian way as
− μ ∑ rm2 ⎥⎥ ∏ drm =
⎦ μ13 2 π 3(N − 1) 2 λ N3 2
m =2 m =2 d1b (U ) = e−λ N U , (16)
⎡ N −1 ⎤ π3
ω−μ μ⎛ ⎞2
N −1
× ∫ exp ⎢ − ⎜r + ∑ rm ⎟⎟ − μ ∑ rm2 ⎥⎥
N μ1 ⎜⎝
⎢ showing that it is actually a Gibbs state. Note that λ N ⩽ 1 for
⎣ m =2 ⎠ m =2 ⎦ all N, since ω + μ ⩾ 2 ωμ and  N ⩾ N 2ωμ; also note that
N −1
1 1 λ N → 1 when N → ∞.
× ∏ drm = It is known that there is a unitary operator, associated
m =2 π μ13 2 3(N − 1) 2
with any symplectic transformation (say, SN), which acts on
⎡ N −1 ⎤ the Hilbert space [42]. Let us use this transformation to find
⎢ ω − μ1 μ ⎛ N − 1 ⎞2
× ∫ exp − ⎜ r + ∑ rm ⎟ − μ ∑ rm2 ⎥
⎟ the set of occupation numbers on the basis of the Wigner
⎢ N − 1 μ ⎜⎝ ⎥
⎣ 1 m =2 ⎠ m =2 ⎦ eigenfunctions of the harmonic oscillator. Since the one-body
N −1 ⎛ μ ⎞ quasidensity function factorizes completely, from now on we
× ∏ drm = (const.) Δr ⎜⎜ r , μ1, ω , N − 1⎟⎟ , (10) will work in one dimension. From the series formula
m =2 ⎝ μ1 ⎠ ∞
1 +t x

m N−m
(1 − t ) ∑L r (x ) e−x 2 t r = e− 1 −t 2 , (17)
where μm := ω + N
The constant (const.) will be
N
μ. r=0
determined by normalization so that it will depend on the
where Lr(x) is the Laguerre polynomial, it follows that
number of particles, as well as on the frequencies. Therefore,
∞
⎛ μ ⎞ (−1)r
Δr (r , μ , ω , N ) = (const.) Δr ⎜⎜ r , μ1, ω , N − 1⎟⎟
d1b (U ) = ∑ π
( )2
L r 2U 2 e−U n r , (18)
r=0
⎝ μ1 ⎠
where the occupation numbers are equal to
⎛ μ ⎞
= (const.) Δr ⎜⎜ r , μ2 , ω, N − 2⎟⎟ = … 2λ N ⎛ 1 − λ N ⎞r
⎝ μ2 ⎠ nr = ⎜ ⎟ =: ( 1 − t N ) t Nr , (19)
1 + λN ⎝ 1 + λN ⎠
⎛ ω−μ μ 2 ⎞
= (const.) exp ⎜⎜ − r ⎟⎟ , (11) fulfilling ∑r nr = 1, as one should expect.
⎝ N μ N −1 ⎠
For the sake of completeness in figure 1 we plot the
and in addition numerical behavior of the one-body quasidensity function of
the N-boson harmonium for different values of the number of
(
Δ p (p , μ , ω , N ) ≡ Δr p , μ−1, ω−1, N , ) (12) bosons. We observe that, as the number of bosons grows, the
profiles of the position and momentum densities become
so that the expression (9) can be transformed as narrower and wider, respectively. This clearly indicates that
the more precisely the particles are localized in position
N 3 ⎛ ωμ ⎞ ⎛
32
Nωμ
d1b (r ; p) = ⎜
3 ⎝ ⎠
⎟ exp ⎜ − r2 space, the larger the localization in momentum space, as one
π N ⎝ (N − 1) ω + μ should expect, according to the position-momentum uncer-
N ⎞ tainty principle. There is no relevant difference for negative
− p2 ⎟ , (13) values of the coupling constant, δ, except that in this case, it
ω + (N − 1) μ ⎠
can be plotted only when δ k ⩾ − 1 N .
where  N = [(N − 1) ω + μ][ω + (N − 1) μ]. Finally, we can compute the von Neumann entropy of the
Let us now diagonalize this expression by decomposing N-boson harmonium explicitly in terms of N and the relative
it into its set of natural orbitals, as well as into its occupation interaction strength δ k , obtaining the value
numbers. The task is made easier by noting that the one-body
t N log ( 1 − t N )
quasidensity is a linear combination of Laguerre polynomials, S (N ) := −∑n r log n r = −log ( 1 − t N ) −
which are the phase-space counterpart of the Hermite poly- r=0
1 − tN
nomials. To see that, let us define  N + N ωμ ⎡ 2N ωμ ⎤
= − log ⎢ ⎥, (20)
2N ωμ ⎢⎣  N + N ωμ ⎥⎦
⎡ ω + (N − 1) μ ⎤1 4 N ωμ
γN := (ωμ)1 4 ⎢ ⎥ , λ N := , (14)
⎣ (N − 1) ω + μ ⎦ N which complements and extends a similar formula encoun-
tered by other means in a black-hole context [31].
as well as the symplectic transformation This entanglement measure is numerically examined in
figure 2 in terms of the number of bosons, N, and the relative
⎛ γN 0 ⎞ interaction strength, δ k . As stated in (5), the minimum value
S N := ⎜⎜ ⎟ and U := S N u where uT = (r , p). (15)
−1 ⎟
of this relative strength is − 1 N . There is a critical point
⎝ 0 γN ⎠ around Nc ∼ 3.5 where the entanglement acquires its max-
imum value. This is more clearly seen in figure 3, where the
Then, the one-body quasidensity function can be rewritten in explicit dependence of the von Neuman entropy on N is

4
J. Phys. B: At. Mol. Opt. Phys. 47 (2014) 195503 C L Benavides-Riveros et al

1.0 1.0

4 4
0.5 0.5
2 2
0.0 0.0
-4 0
p -4 0 p
-2 -2
0 -2 0 -2
r 2 r 2
4 -4 4 -4
(a) (b)

1.0 1.0

4 4
0.5 0.5
2 2
0.0 0.0
-4 0
p -4 0
p
-2 -2
0 -2 -2
0
r 2 r 2
4 -4 4
-4

(c) (d)
Figure 1. The one-body quasidensity function (13) is plotted for different values of the number of bosons. Since the one-body quasidensity
function factorizes completely, we plot it in one spatial dimension (two dimensions in phase space). The strength, δ k , is taken to be equal to
1. Note that as the number of bosons grows, the profiles of the position and momentum densities become narrower and wider, respectively.

shown. To investigate it more closely, we study the depen-

dence of the occupation numbers on the number of bosons and
the relative interaction strength. In particular, in figure 4 we
plot the first occupation number separately in terms of N and
the relative strength. Therein, we note that below a critical 10
8
value of number of particles (around Nc ∼ 3.5), for positive N 6
values of the coupling constant δ k , the first occupation number 4
decreases as the number of particles increases. Beyond this 2
value, the situation is reversed: The value of the first occupation
0.4
number increases as the number of particles increases. This
implies that n 0 ∼ 1 in the limit when the number of particles S(N)
tends to infinite, which is a necessary condition for the exis- 0.2
tence of a Bose-Einstein condensation [43].
In summary, the critical point around N ∼ 3.5, where the 0.0
40 20 0
von Neumann entropy is maximum, is closely connected with
the minimum of the first occupation number occurring at this δ/k
position. We found that above this critical value, the spatial
Figure 2. Von Neumann entropy of the one-body reduced density
entanglement decreases when we add more and more particles matrix for the N-boson harmonium is plotted as a function of the
to the system, so that the degrowth rate increases when the number of bosons, N, and the relative interaction strength, δ k . It is
coupling constant increases. Moreover, it can be shown that apparent that there is a critical point around Nc ∼ 3.5, where the
the entropy vanishes when N goes to infinite. entanglement acquires its maximum value.

5
J. Phys. B: At. Mol. Opt. Phys. 47 (2014) 195503 C L Benavides-Riveros et al

Figure 3. Von Neumann entropy of the one-body reduced density matrix for the N-boson harmonium, whose explicit formula is given in
equation (20), is plotted as a function of the number of particles for eight different values of the relative interaction strength, δ k . Note that for
each value of the relative strength, there is a peak around the critical value, Nc.

The symbol J denotes an element of the permutation group,

SN of N elements (acting on the r-coordinates), and ϕnν is the
single-particle wave function given by a Hermite function of
degree n with frequency ν. The collective mode, ξN , is
symmetric under any exchange of the position coordinates.
Therefore, one has nN = 0; otherwise one would have nm = 0
for some m ≠ N , and the wave function would not be totally
antisymmetric. This eigenfunction can be rewritten as
Ψ gsf ( ξ1 ,..., ξN ) = (const.) ∑ ( − )J J
Figure 4. Plot of the first occupation number, n 0 = 1 − t N , as a J ∈SN
function of the number of particles for different positive values of
⎡ N −1 ⎤ ω 2 μ N −1 2
δ k . As it is displayed, n0 takes its minimum value around a critical
×⎢ ∏ Hn m ( μ ξm ) ⎥ e− 2 ξN − 2 ∑m=1 ξm , (23)
value in the number of particles (∼3.5). ⎢⎣ m =1
⎥⎦

where Hn is the Hermite polynomial of degree n, and now

4. The spinless N-fermion harmonium: linear entropy n m ∈ {1, ⋯ , N − 1}. The exponential power of (23) is the
bosonic wave function, which is the counterpart of the
In this section, we study the spatial entanglement of the N- Wigner function in (8).
fermion harmonium (i.e., entanglement without taking into As an illustrative case, let us consider the three-harmo-
account the spin degree of freedom of the constituents) using nium system. To produce an antisymmetric wave function,
the linear entropy of the one-particle density matrix as a there is only one possibility for the degree of the polynomials,
measure of the entanglement of one particle. This measure is a namely n1 = 1 and n 2 = 2. Then, the ground-state eigen-
first-order approximation of the von Neumann entropy of the function is
reduced one-body density of the system. In fact, it is a lower
bound of this entropy. We will consider the case in which the Ψ gsf ( ξ1, ξ2, ξ3 ) = (const.) ∑ ( − )J J
J ∈ S3
spinless fermions are confined in a one-dimensional well,
× [ψ (ξ ) ] e− 2 ξ3 − 2 ( ξ1 + ξ2 ),
ω 2 μ 2 2
mainly because the antisymmetric fermion case requires more (24)
elaborate computations. Here, we must consider the anti-
symmetry under the exchange of position coordinates where ψ (ξ ) := H1 ( μ ξ1 ) H2 ( μ ξ2 ) and ξ := (ξ1, ξ2 ). Note
that the six elements of the permutation group S3 are
{r1 ,..., rN }.
Taking into account the Hamiltonian (1) for the N-fer- p1 = (1 2 3), p2 = (2 1 3), p3 = (2 3 1),
mion harmonium, one sees that the ground-state energy of the p4 = (1 3 2), p5 = (3 1 2), p6 = (3 2 1).
system is
N −1
The coordinate, ξ, transforms under the representation of the
1 ⎛ 1⎞ 1 1 permutation group in the following form:
Egsf = ω+
2
∑ μ ⎜⎝ m + ⎟⎠ = ω + μ N 2 − 1
2 2 2
( ) (21)
j=1 R ( p1 ) ξ = ξ , R ( p2 ) ξ = M1ξ , R ( p3 ) ξ = M1M2 ξ ,
and the corresponding eigenfunction can be expressed as R ( p4 ) ξ = M2 ξ , R ( p5 ) ξ = M1M3 ξ , R ( p6 ) ξ = M3 ξ ,
1 where
Ψ f
( ξ1 ,..., ξN ) = ∑ ( − )J J
N!
⎡
J ∈SN
N −1 ⎤ 0 1( )
M1 = − 1 0 , M2 = ( cos π 3 sin π 3
sin π 3 − cos π 3 )
× ⎢ ϕnωN ( ξN ) ∏ ϕnμm ( ξm ) ⎥ , (22)
⎢⎣ m =1
⎥⎦ (
M3 = cos π 3 −sin π
− sin π 3 − cos π
3 .
3 )
where n m ∈ {0, ⋯ , N − 1}, so that ni ≠ n j whenever i ≠ j .

6
J. Phys. B: At. Mol. Opt. Phys. 47 (2014) 195503 C L Benavides-Riveros et al

Note that det Mi = −1 and M tM = 1. Any matrix function corresponding to the Gaussian one on the right-hand
⎛ cos ϕ sin ϕ ⎞ side of (27) is nothing but the Wigner N-body quasidensity
Mϕ = ⎜ ⎟ given by (8).
⎝ sin ϕ −cos ϕ ⎠
Let us now calculate the one-body density of the one-
is a reflection in the axis θ = ϕ 2. In our case, this means dimensional spinless N-fermion harmonium, which is defined
θ = π 2, π 6, and − π 6 (i.e., three reflection axes at 60- by
degree angles to each other). The rotations R (p3 ) and R (p5 )
are products of two reflections each. Thus, the eigenfunction ρ1 ( r ; r ′) = ∫ dr2 ⋯drN Ψ f ( r, r2, ⋯, rN ) Ψ f
transforms as follows:
× ( r ′, r2, ⋯ , rN ). (29)
Ψ gsf ( ξ1, ξ2, ξ3 )
6 For this purpose, it is convenient to rewrite the eigen-
= (const.) e − 2 ξ3 − 2 ( ξ1 + ξ2 )
ω 2 μ 2 2
∑( − )i+1ψ ( R ( pi ) ξ) function (27) as
i=1
6
ω −μ 2 μ
( r +r +r ) ∑( − )i+1ψ ( R ( pi ) ξ)
2 2 2 Ψ f ( r1, ⋯ , rN ) = (const.)
= (const.) e − 2 ξ3 − 2 1 2 3

× ( r1, ⋯ , rN ) e−a( r1 +⋯+ rN ) + b N ( r1+⋯+ rN ) ,

2 2 2
i=1 (30)
= (const.) e − 2 ξ3 − 2 ( r1 + r2 + r3 ) ( r1 − r2 )
ω −μ 2 μ 2 2 2

μ μ−ω
× ( r1 − r3 )( r2 − r3 ). (25) with a = 2 and b N = 2N . Moreover, we will use the Hub-
bard-Stratonovich transformation
Note that the last equality was possible because
α ∞

1 1
6
2
eαζ =
π −∞
∫ 2
dz e−αz + 2αzζ , with
∑( − )i+1ψ μ R ( pi ) ξ
12 2 μ3 2 i = 1
( ) α ∈  and Re (α ) > 0, (31)
1
= −
2
( )
ξ13 − 3ξ1ξ22 = ( r1 − r2 )( r1 − r3 )( r2 − r3 ) (26) where ζ = r2 + ⋯ + rN and α = 2b N . Then, the one-body
density reads
is a Vandermonde determinant. This three-fermion result can 2
bN
ρ1 ( r ; r ′) = k N e−( a − b N )( r )∫ 2
dz e−2b N z e(N − 1) 2a ( r + r ′+ 2z )
2 + r ′2 2
be extended to the N-fermion system. In fact, the general
results of the theory of antisymmetric functions ensure that
the wave function of a system of spinless fermions is equal to × ∫ dz2 ⋯dz N ( z ,z ,⋯,z r 2 N)  ( z r′, z 2, ⋯ , z N )
the wave function of the corresponding bosonic system, N
2
multiplied by the N-variable Vandermonde determinant (see × ∏ e−z j , (32)
section 3.1 of [44]). So, the ground-state wave function of the j=2
spinless N-fermion harmonium described by the Hamiltonian
(1) has the form where kN denotes the normalization constant (to be calculated
later), and z j := 2a ⎡⎣ rj − 2aN (r + r ′ + 2z ) ⎤⎦ is a convenient
b

Ψ f ( r1, ⋯ , rN ) = (const.) ∏ ( ri − r j )
change of coordinates. A key observation [26] is that the last
i<j
expression on the right-hand side is the multiplication of two
⎡ ω−μ μ
N ⎤ Slater determinants, and therefore the computation of the
exp ⎢ − ξN2 − ∑ri2 ⎥⎥, (27)
⎢⎣ 2 2 i=1 ⎦ reduced density matrix can be completed using well-known
methods in quantum chemistry. Indeed, with the notation
where the product-like symbol on the right-hand side of this
μ−ω
expression denotes the Vandermonde determinant, β N := ,
(N − 1) ω + μ
1 ⋯ 1
q( r , r ′) := μ ⎡⎣ r − 2 βN2 ( r + r ′) ⎤⎦
1
r1 ⋯ rN
( r1, ⋯ , rN ) := ⋮ ⋱ ⋮
(μ − ω)2 N−1
r1N − 1 ⋯ rNN − 1 and cN := , (33)
(N − 1) ω + μ 2N
= ∏ ( ri − r j ). (28)
one can express the one-body density as
1 ⩽i < j⩽N
N −1
Concrete calculations [30, 34] have borne out this fact several
times. On the other hand, it has been pointed out by several
ρ1 ( r ; r ′) = k N e−a N ( r
2 + r ′2
) + 2c N rr ′
∫ du e−u ∑ 21jj! H j
2

j=0
authors that this function is actually a generalization of
⎡ ⎤ ⎡ ⎤
Laughlinʼs wave function for the fractional quantum Hall × ⎣ q( r , r ′) − βN u⎦ H j ⎣ q( r ′, r ) − βN u⎦ , (34)
effect [45]. This is not surprising since, in the Hall effect, the
magnetic field can be understood as a harmonic potential with a N := a − b N − cN . This expression can be calculated
acting on the electrons. Let us also point out that the Wigner analytically with the help of the following general formula,

7
J. Phys. B: At. Mol. Opt. Phys. 47 (2014) 195503 C L Benavides-Riveros et al

derived in appendix A: Keeping all this in mind and determining the normal-
∞ ization constant, kN, by imposing that ∫ ρ1 (r; r ) dr = 1, one
∫−∞ e−u Hk ( q( r,r′) − βN u) Hk ( q( r′,r ) − βN u) du
2
finally has the expression
k 1 1 ωμ
e−a N ( r + r ′ ) + 2cN rr ′
2 2

= ∑ ( 2β N )
n1, n 2 = 0
n1+ n 2 even
n1+ n 2
( )( )
k
n1
k
n 2 H k − n1
ρ1 ( r ; r ′) =
π N
N − 1 2r
(N − 1) ω + μ

∑∑ctN, s r 2t−sr′s
( )
× − q( r , r ′) Hk− n 2 − q( r ′, r ) Γ ( ) ( n1 + n 2 + 1
2 ). (35) ×
t=0 s=0
(39)

Then, the integral on the right-hand side of (34) can be for the one-body density of the one-dimensional spinless N-
expressed as the following power sum: fermion harmonium.
N −1
Let us now quantify the entanglement of the system by
∫ du e−u ∑ 21jj! H j ⎡⎣ q( r,r′) − βN u⎤⎦ H j ⎡⎣ q( r′,r ) − βN u⎤⎦
2 means of the linear entropy associated with the one-body
density [46]:
j=0

SL = 1 − N Tr ⎡⎣ ρ12 ⎤⎦ ,
N − 1 2t
(40)
= π ∑∑ctN, s r 2t−sr′s , (36)
t=0 s=0
where ΠN = N Tr [ρ12 ] is the purity of the system. This
where the expansion coefficients, depend on both the crN, s , entanglement measure, which is a non-negative quantity that
frequencies ω, μ and the number of particles. When r + s is vanishes if and only if the state has Slater rank 1 and is
an odd number or is lager than 2(N − 1), crN, s = 0. By sym- therefore separable, has been recently used in various two-
metry it is clear that ctN, s = csN, t . In the one-fermion case, there fermion systems [13, 24] as well as for various helium-like
is only one coefficient c0,1 0 = 1. For N fermions, the rank of systems [14, 16, 17] in both ground and excited states.
the quadratic form of the cr, s is 2N − 1. In particular, when Moreover, in our systems it turns out that
N = 2, we have the expansion coefficients:
Tr ⎡⎣ ρ12 ⎤⎦ = ∫ ρ1 ( r; r′) ρ1 ( r′; r ) dr′ dr
c0,2 0 =
2μ
, c1,2 1 =
μ μ + 2μω + 5ω ( 2 2
) and
N − 1 2t 2e

ω+μ (ω + μ)2 = k N2 ∑ ∑∑ ctN, s ceN, f (2t − s + f ,2e− f + s) , (41)

t,e=0 s=0 f =0
μ ( ω − μ) ( 3 ω + μ)
c2,2 0 = c0,2 2 = . (37)
2(ω + μ)2 where, as shown in appendix B,
(n, m) = (m, n)
For completeness, let us mention that the reader will have no
∞
difficulty in verifying that ρ1 (r , r )—the diagonal of (29)— = ∫−∞ r nr′m e−2a ( r +r′ )+4C rr′ dx dy
N
2 2
N

coincides (except for the difference in the normalization

n m
conventions) with the one obtained in [25 Section V] by the
Wigner function method. Moreover, for N = 3 the non van- = 2 ∑∑
i=0 j=0
( ni)( mj)(−1) λ j −(n + m − i − j + 1) 2
1

ishing expansion coefficients are

× λ 2−(i + j + 1) 2
3⎡ (ω − μ)2 ⎤
c0,3 0 = ⎢1 +
2⎣
⎥,
(2ω + μ)2 ⎦
×Γ ( n+m−i−j+1
2 )Γ ( n+m+1
2 )
× 4 ⎡⎣ 1 + (−1)n + m − i − j ⎤⎦ ⎡⎣ 1 + (−1)i + j ⎤⎦ ,
1
(42)
c1,3 1 =
(
μ −15ω3 + 51μω2 + 15μ2 ω + 3μ3 ),
( 2 ω + μ) 3 with λ1 = 4(a N − cN ) and λ 2 = 4(a N + cN ).
Using expressions (40), (41), and (42), we have investi-
c2,3 2 =
(
μ2 363ω4 + 168μω3 + 90μ2 ω2 + 24μ3ω + 3μ4 ), gated the dependence of the linear entropy of the system on
4 the relative interaction strength δ k , as well as on the energy.
4 ( 2 ω + μ)
The results are given in figure 5. A first basic observation is
c2,3 0 = c0,3 2 = −
(
μ 39ω3 − 3μω2 + 15μ2 ω + 3μ3 ), that for fixed N, the entanglement decreases (increases) when
3
2 ( 2 ω + μ) the relative interaction strength increases in the negative
μ (5ω + μ) (ω − μ)2
2 2 (positive) region. Moreover, for negative values of the cou-
c4,3 0 = c0,3 4 = and c3,3 1 = c1,3 3 pling constant, the spatial entanglement of the N-fermion
8 ( 2 ω + μ) 4
harmonium grows when N increases. For small positive

=
(
μ2 (ω − μ) 65ω3 + 33μω2 + 9μ2 ω + μ3 ). (38)
values of the coupling constant (i.e., when the fermions attract
2 ( 2 ω + μ) 4 each other), we find that for very small values of δ k , the
entanglement grows again with increases N. In general, we
Except for the difference in the normalization and notation observe that for positive values of the strength, we have
conventions, these coefficients coincide with the ones found various regimes of dependence on N. In the strong-coupling
in [26]. regime, the linear entropy grows as the number of particles

8
J. Phys. B: At. Mol. Opt. Phys. 47 (2014) 195503 C L Benavides-Riveros et al

Figure 5. Linear entropy of the one-particle density matrix of the N-fermion harmonium as a function of the coupling constant for five
different values of N. Note the change in behavior when the strength grows.

increases since the purity decreases with an increases number

of particles. For instance, for δ k = 10 , the purity is 0.92 for
N = 2 and 0.94 for N = 10, whereas for δ k = 1, the purity is
0.998 and 0.992, respectively.
Figure 6 displays the dependence of the linear entropy of
the system on the ground state-energy, Egsf ω , for various N 2
values of the number of fermions, N. We should keep in mind
N 3
that the relation between the relative interaction strength and
N 4
the energy is given by
N 5
⎡ 2E f ω − 1⎤
2

δ /k = ⎢ ⎥ − 1.
gs
(43)
⎢⎣ N 2 − 1 ⎥⎦

Figure 6. Linear entropy, SL, of the N-fermion harmonium versus the

We find that the entanglement grows when the energy
ground-state energy, Egsf ω , for different values of N. The relation
increases. For N = 2, this behavior has been recently observed
between the relative strength, δ k , and the energy is stated in (43).
not only in the Moshinsky or two-fermion harmonium atom,
The entanglement grows when the energy increases.
but also in other two-fermion systems [13–17]. The novelty is
that this behavior appears to also be true for heavier N-fer-
antisymmetric spin state. For a system of fermions with spin
mion harmonium atoms.
It is worth noting that our results can be considered as a in the presence of a strong magnetic field, the spin part of the
generalization of previous results for the entanglement eigenfunction factorizes, so the spatial part must be anti-
entropy of the fractional quantum Hall effect, based on an symmetric [26]. In general, however, for the isolated fer-
exact matrix-product representation of the Laughlin wave mionic system, the spatial and spin parts of the ground-state
function [47–49]. This well-known wave function, formulated wave function are not separable, and antisymmetry must take
by R Laughlin to explain the fractional quantum Hall effect, into consideration both spin and position coordinates.
consists of a lone Slater determinant and reads as in (27), Let us now calculate the total entanglement of the N-
except that there is no coupling between the electrons; that is, fermion harmonium (i.e., the entanglement of both spatial and
ω = μ or δ = 0 , and the domain is the complex space. spin degrees of freedom) using the linear entropy as a measure
of entanglement. For simplicity, we consider one-dimensional
models with an even number (say, 2N ) of spinned electrons
5. Total entanglement of the spinned N-fermion from now on. The total spin of the system is zero, and all the
harmonium spatial orbitals are doubly occupied (restricted configuration).
This will eventually allow for an easy generalization to closed-
So far, we have considered only spinless particles. This means shell three-dimensional systems. We define the set of spin
that in the two previous sections, we have determined the natural orbitals in the following way:
spatial entanglement of the N-boson and N-fermion harmonia.
Nevertheless, our conclusions for the bosonic system also ⎧
⎪ ϕ 2 (r ) ↑ , if j is odd,
j −1

apply to the total entanglement from a qualitative point of φ j (x ) = ⎨

⎪ ϕ j (r ) ↓ , if j is even.
view, because the spin part of the ground-state eigenfunction ⎩ 2
fully factorizes. For fermionic systems this is no longer true,
at least when they are isolated. In fact, the ground state of two The symbols ↑ and ↓ mean spin up and spin down, respectively,
fermions is the product of a two-boson state and an and we denote x := (r , ς ), ς as the spin coordinate.

9
J. Phys. B: At. Mol. Opt. Phys. 47 (2014) 195503 C L Benavides-Riveros et al

Let us begin with the consideration of the simpler, non- the expression
interacting case; that is, when δ = 0 , and hence μ = ω. The
⎛ ρ ↑↑ r , r ′ ⎞
1 ( )
corresponding ground-state eigenfunction is then given by the 0
ρ1 (x , x′) = ⎜⎜ ⎟. (46)
) ⎟⎠
expression ↓↓
1 (
⎝ 0 ρ r , r ′
Ψ f ( x1, … , x N )
1 ⎡ 2N ⎤ Moreover, it is clear that
= ∑ ( − )J ⎢⎢ ∏ φi − 1 ( x J (i ) ) ⎥
⎥⎦
2N ! J ∈ S2N ⎣ i=1 ρ1↑↑ ( r , r ′) = ρ1↓↓ ( r , r ′)
1
=
2N !
∑ ( − )J J = ∫ dx2 ⋯dx2N Ψ f ( r, ς1, x2, …, x N )
J ∈ S2N
⎡ N ⎤ Ψ f
( r ′, ς1, x2 , …, x N ) ς =↑
×⎢ ∏ ↑2i − 1 ↓2i ϕiω− 1 ( r2i − 1) ϕiω− 1 ( r2i ) ⎥ . (44) 1

⎢⎣ ⎥⎦ 1 1
i=1 =
2N (N − 1)!
dr2 ⋯drN ∫
In the noninteracting case, the basis of the one-particle Hilbert
space is the set of Hermite functions whose degree is less than φ0 (r ) ⋯ φ0 ( rN )
N. Moreover, let us define × ⋮ ⋱ ⋮
φ N (r ) ⋯ φ N ( rN )
H = { j ∈ S2N : j fixes 1, … , N} ≃ S N
φ0 ( r ′) ⋯ φ0 ( rN )
K = { j ∈ S2N : j fixes N + 1 ,..., 2N} ≃ S N ,
× ⋮ ⋱ ⋮
where S˜N = H × K is the direct product of these two sub- φ N ( r ) ⋯ φ N ( rN )
′
groups. The set S2′N = S2N S̃N is the set of right cosets in S2N, N −1
giving the following equivalence relation: J ∼ J ′ if and only 1
= ∑ φi (r ) φi ( r ′). (47)
if there exists (j, j′) ∈ S˜N such that J ′ = (j, j′) J . 2N i=0
Therefore, we can reorganize the expression (44) in the
following form: The occupation numbers appear twice, which is a well-known
result for atomic and molecular scientists.
N 2N
1 Let us now consider the general interacting case, for
Ψ f ( x1, … , x N ) = ∑( − ) J ∏ ↑J (i ) ∏ ↓ J (i )
2N ! which δ ≠ 0 , and hence μ ≠ ω . The ground-state eigen-
i=1 i=N +1
function for the Hamiltonian (1) is similar to (22), except that
ϕ0ω ( rJ (1) ) ⋯ ϕ0ω ( rJ (N ) )
each spatial orbital is doubly occupied. For the same reasons
× ⋮ ⋱ ⋮ as in the spinless case, the collective mode, ξ2N , occupies the
ϕNω ( rJ (1) ) ⋯ ϕNω ( rJ (N ) ) Hermite function of degree zero. The other coordinates,
ϕ0ω ( rJ (N + 1) ) ⋯ ϕ0ω ( rJ (2N ) ) {ξm } 2mN=−11, occupy the other spatial orbitals in such a way that
the total wave function is totally antisymmetric under inter-
× ⋮ ⋱ ⋮ , (45)
change with coordinates {x m } 2mN= 1. The ground-state eigen-
ϕNω ( rJ (N + 1) ) ⋯ ϕNω ( rJ (2N ) ),
function schematically reads

where the sum runs over a representative of each coset. The 1

Ψ f ( x1 ,..., x2N ) = ∑ ( − )J
choice of the representative J ∈ [J ] ∈ S2′N is immaterial. The 2N ! J ∈ S2N
number of summands is
⎡ N 2N ⎤
×⎢ ∏ ∏ ↓ J (i ) ⎥
S2N S˜N =
S2N
˜
SN
=
(2N )!
(N !)2
= 2N .
N ( ) ⎢⎣ i=1
↑J (i )
i=N +1
⎥⎦
× J ⎡⎣ ϕ0ω ( ξ2N ) ϕ0μ ( · )
For instance, for the four-harmonium (N = 4), the eigen-
function can be written in the alternative form N ⎤
× ∏ ϕmμ ( · ) ϕmμ ( · ) ⎥ . (48)
ω2 1 − ω ( r12+ r22+ r32+ r42 ) m =1
⎥⎦
f
Ψ ( x1, x2 , x 3, x 4 ) = e 2
π 3!
For instance, in the particular case, N = 4, the eigen-
× ⎡⎣ ( ↑1 ↑2 ↓3 ↓4 + ↓1 ↓2 ↑3 ↑4 ) (1,2) (3,4)
function has the form
− ( ↑1 ↑3 ↓2 ↓4 + ↓1 ↓3 ↑2 ↑4 ) (1,3) (2,4)
Ψ f ( x1, x2 , x 3, x 4 )
+ ( ↑1 ↑4 ↓2 ↓3 + ↓1 ↓4 ↑2 ↑3 ) (1,4) (2,3) ⎤⎦.
1
The one-body reduced density matrix corresponding to the
= ∑ ( − )J ↑J (1) ↑J (2) ↓J (3) ↓J (4)
4! J ∈ S4
wave function (45) is diagonal in spin space and is given by

10
J. Phys. B: At. Mol. Opt. Phys. 47 (2014) 195503 C L Benavides-Riveros et al

× J ⎡⎣ ϕ0ω ( ξ4 ) ϕ1μ ( ξ1) ϕ0μ ( ξ3 ) ϕ1μ ( ξ2 ) ⎤⎦

= (const.) ∑ ( − )J ↑J (1) ↑J (2) ↓ J (3) ↓ J (4)
J ∈S4

× J [ ξ1ξ2 ] e− 2 ξ4− 2 ( ξ1 + ξ2 + ξ3 )
ω μ 2 2 2

= (const.) ∑( − )J ↑J (1) ↑J (2) ↓ J (3) ↓ J (4)

S 4′

× (J (1), J (2)) (J (3), J (4)) e− 2 ξ4− 2 ( ξ1 + ξ2 + ξ3 ).

ω μ 2 2 2

Assuming again that an antisymmetric polynomial is equal to

a symmetric polynomial multiplied by a Vandermonde
determinant [44], we can write the ground-state eigenfunction Figure 7. Linear entropy of the spinned N-fermion harmonium as a
of the general interacting spinned N-fermion system as function of the relative coupling constant for various values of N.
Qualitatively, it is similar to the spinless case in figure 5.
ω −μ 2 μ 2N
Ψ f ( x1, … , x2N ) = (const.)e − 2 ξ2 N − 2 ∑i = 1 ri2
⎡ N 2N ⎤
∫ dr2 ⋯drN e−2a( r +⋯+r )+b ( η +η′ )+d ( η+η′)
2 2 2 2 2

× ∑( − )J ⎢ ∏ ↑J (i ) ∏ ↓ J (i ) ⎥ × 2 N 2N N

⎢⎣ i = 1 ⎥⎦
S2′N i=N +1 × r ,, 2 ,..., N r ′,2, … , N
× (J (1), … , J (N )) (J (N + 1), … , J (2N )) . (49)
= (const.) e −a( r )+b ( r +r′ )+d ( r + r ′)2
2 + r ′2 2 2
2N N

1 2
The corresponding energy is Egsf = 2 (ω + μ) + μ (N 2 − 1). gN 2
× ∫ dz e−2g z N
2
e(N − 1) 2a ( r + r ′+ 2z )
Let us define η = r + r2 + ⋯ + rN and η′ =
N
r ′ + r2 + ⋯ + rN . To compute the one-density, we use twice 2

the Hubbard-Stratonovich identity, both to compute N inte-

× ∫ dz2 ⋯dz N z ,, z ,..., z r 2 N  z r′, z 2, … , z N ∏ e−z ′ j , (51)
j=2
grals with ζ′ = rN + 1 + ⋯ + r2N , and also to compute N − 1
2a ⎡⎣ rj −
(r + r ′ + 2z ) ⎤⎦, d N = 81N (μμ−+ωω) ,
integrals with ζ = r2 + ⋯ + rN . Each diagonal element of the gN 2
where z j′ := 2a
one-body density matrix reads μ−ω μ
and gN = 2N μ + ω . As in the spinless case, let us define
ρ1↑↑ ( r , r ′) = ρ1↓↓ ( r , r ′) = (const.) e −a( r ) q˜( r , r ′) = μ ⎡⎣ r − 2 β22N ( r + r ′) ⎤⎦ and
2 + r ′2
1

∫ dr2 ⋯drN e−2a( r +⋯+r )+b ( η +η′ )

2 2 2 2
× 2 N 2N
(μ − ω)2 2N − 1
c˜2N = (52)
b 22N (2N − 1) ω + μ 8N
× ∫ dz e −2b 2N z 2 N
e 2a ( η + η′+ 2z )2
with βN as defined in (33). Then,
× (r ,, 2 ,..., N )  ( r ′,2, … , N )
1 1 2ωμ
× ∫ dz N +1⋯dz2N  2( z N + 1, … , z 2N ) ρ1↑↑ ( r , r ′) = ρ1↓↓ ( r , r ′) =
2π N (2N − 1) ω + μ
2N
× e−a N′ ( r )+2c˜ ∫ du e−u
2 + r ′2 2
2 2N rr ′
× ∏ e−z j , (50)
j=N +1 N −1
1 ⎡ ⎤
× ∑ H j ⎣ q˜( r , r ′) − β2N u⎦ H j
where z j := 2a ⎡⎣ rj − 22aN (η + η′ + 2z ) ⎤⎦. Once again, the
b j
j=0 2 j!
expression on the right is the total integral of the product of ⎡ ⎤
× ⎣q˜( r ′, r ) − β2N u⎦ , (53)
two Slater determinants; that is,
with a N′ := a − b 2N − c˜2N .
∫ dz N +1⋯dz2N  2( z N + 1, … , z 2N ) The expression (53) is similar to (34), except that we
2N have an even number of fermions: N → 2N and cN → c̃2N . It
2
× ∏ e−z j is now clear that the conclusions for the spinless fermionic
j=N +1 case also hold for the full spin case from a qualitative point of
view. Figure 7 plots the linear entropy for the spinless and the
is a real constant. Therefore, spinned cases as functions of the relative strength.
For completeness, we have numerically investigated the
ρ1↑↑ ( r , r ′) = ρ1↓↓ ( r , r ′) dependence of the linear entropy on the number of particles
for some specific values of the relative coupling in both
= (const.) e −a( r )
2 + r ′2

spinless and spinned cases (i.e., for spatial and total

11
J. Phys. B: At. Mol. Opt. Phys. 47 (2014) 195503 C L Benavides-Riveros et al

Figure 8. Linear entropy of the one-body reduced density matrix for the spinless and spinned N-fermion harmonium as a function of the
number of particles for different values of δ k .

entanglements). Figure 8 exhibits the results for

δ k ∈ {−1 15, 2, 4, 22}. In the spinned case, the behavior is
similar to the spinless system, so that the total entanglement
behaves similar to the spatial entanglement. In fact, for small
values of the coupling, the entropy grows when the number of
particles increases, while in the strong-coupling regime the
opposite is true. The inclusion of the spin tends to enhance the
entropy of the system; this phenomenon is most important for
small values of the coupling constant. Note that in turn, the
contributions of the spin and the spatial entropy are of com-
parable size.
Finally, let us investigate the relative behavior of entan-
glement and energy. This is done in figure 9, where the linear
entropy as a function of the ground-state energy is shown for
Figure 9. Linear entropy, SL, of the N-fermion spinned harmonium
some values of the number of particles. Let us recall here that versus the ground-state energy, Egsf ω , for different values of N. The
the relation between the relative coupling and the energy is
relation between the relative strength, δ k , and the energy is stated
given by in (54).
⎡ 2E f ω − 1 ⎤2
δ k=⎢ ⎥ − 1.
gs
(54)
⎢⎣ 2N 2 − 1 ⎥⎦ 6. Concluding remarks

As in the spinless case, the entropy grows when the ground- In this work, we have shown that some entanglement features
state energy increases. For very large values of the dimen- of finite many-particle systems can be understood, to a certain
sionless energy, Egsf ω (i.e., very large values of the relative extent, by purely kinematical considerations. This is done by
strength δ k ), the gap between the spinless and spinned cases explicitly analyzing the entanglement of the N-boson and N-
decreases. fermion harmonium systems. This has been possible because

12
J. Phys. B: At. Mol. Opt. Phys. 47 (2014) 195503 C L Benavides-Riveros et al

for these harmonic systems, we have been able to analytically Appendix A. Calculation of the
∞
integral ∫ −∞ e−u H k (a − cu)H k (b − cu)du
2
calculate not only the one-body reduced matrix for bosons
and fermions, but also the von Neumann entropy in the
bosonic case and the linear entropy in the fermionic case. In In this section, we calculate the integral
doing so, we complement and extend to harmonic systems ∞
with an arbitrary number of particles the study of entangle- ∫−∞ e−u Hk (a − cu) Hk (b − cu)du,
2
(A.1)
ment recently done for various two-electron models
[13, 21–25], as well as some helium-like systems
where a, b, and c are real numbers and Hk(x) is the Hermite
[6, 14, 15, 17] and certain quantum complex networks [50].
polynomial of degree k. We use the following property of the
We have analytically determined the entanglement of
Hermite polynomials:
these harmonic systems for both spatial and spin degrees of
freedom in terms of the number of particles and the relative Hn (−z) = (−1)n Hn (z),
interaction strength (or coupling constant). We have used the
von Neumann entropy and the linear entropy as entangle- as well as the power series of these polynomials around
ment quantifiers in the bosonic and fermionic systems, z = z0,
respectively. We have found that for positive couplings, the
n
entanglement of the three-boson harmonium atom is bigger ⎛ n⎞
Hn (z) = ∑ 2k ⎜⎝ ⎟H k
n − k ( z 0 )( z − z 0) .
than the one in the two-boson system, but in general the
k=0
k⎠
entanglement of the N-boson harmonium decreases when N
increases. Moreover, the entanglement of a given N-boson This expression allows us to rewrite each Hermite polynomial
system grows when the positive coupling constant increases in the following form:
that is, when the positive value of the strength of the
interparticle interaction relative to the confinement well Hk (a − cu) = ( −1)k
z 0 =−a
increases. On the other hand, globally speaking, the spatial k
entanglement of the N-fermion harmonium grows when N
increases for both negative and sufficiently small positive
× ∑ (2c)n
n1= 0
1
( )
k n1
n1 Hk− n1 ( −a) u . (A.2)

values of the coupling constant. Moreover, the entanglement

behavior is opposite in the positive strong-coupling regime. Inserting (A.2) into (A.1) we obtain
On the other hand, entanglement of a given N-fermion ∞
system grows when the coupling constant increases. The ∫−∞ e−u Hk (a − cu) Hk (b − cu)du
2

contribution of the spin degree of freedom to the entangle-

∞ k
ment of the N-fermion system is shown to be of positive
comparable size to the contribution of the spatial degrees of
= ∫−∞ e−u ( −1)k ∑ (2c)n ( nk1 )Hk−n ( −a) un ( −1)k
2
1
1
1

n1= 0
freedom. k
In summary, we have shown that in the repulsive and
attractive cases for relatively small values of the coupling
× ∑ (2c)n
n 2 =0
2
( )
k
n 2 H k − n 2 ( − b) u
n2

constant, the entanglement of both bosonic and fermionic N- k

harmonium atoms grows when the number of particles
increases, basically because purity decreases. However, in the
= ∑
n1, n 2 = 0
n1+ n 2 even
(2c)n1+ n 2 nk
1 ( )( ) k
n 2 H k − n1
regime of strong coupling, the situation is exactly the oppo- ⎛ n1 + n2 + 1 ⎞
site: purity increases (and hence entanglement decreases) by × ( − a ) H k − n 2 ( − b) Γ ⎜ ⎟. (A.3)
⎝ 2 ⎠
adding particles to the system.

Acknowledgments
Appendix B. The Gaussian integral, (n;m)
CLBR was supported by a Francisco José de Caldas scho-
larship, funded by Colciencias. He very much appreciates the In the following, we calculate the integral
warm atmosphere of the Departamento de Física Atómica
Molecular y Nuclear at the Universidad de Granada. IVT ∞ ∞

and JSD gratefully acknowledge the MINECO grant

(n, m) = ∫−∞ ∫−∞ x ny me−a( x +y )+2cxy dx dy.
2 2
(B.1)
FIS2011–24540 and the excellence grant FQM-7276 of the
Junta de Andalucía. In addition, the authors are most grateful Here, we perform the change of coordinates
to A R Plastino, J L Alonso, A Botero, J M Gracia-Bondía,
and J C Várilly for helpful and illuminating discussions. x = u + v and y = u − v , (B.2)

13
J. Phys. B: At. Mol. Opt. Phys. 47 (2014) 195503 C L Benavides-Riveros et al

and the integral now reads [13] Manzano D, Plastino A R, Dehesa J S and Koga T 2010
n m J. Phys. B: At. Mol. Opt. Phys. 43 275301
(n, m) = 2 ∑∑
i=0 j=0
( ni)( mj)( −1) j [14] Dehesa J S, Koga T, Yáñez R J, Plastino A R and Esquivel R O
2012 J. Phys. B: At. Mol. Opt. Phys. 45 015504
Dehesa J S, Koga T, Yáñez R J, Plastino A R and Esquivel R O
∞ ∞
2012 J. Phys. B: At. Mol. Opt. Phys. 45 239501
× ∫−∞ ∫−∞ un+m−i−jvi+j e−λ u −λ v 1
2
2
2
du dv , (B.3)
(corrigendum)
[15] Benenti G, Siccardi S and Strini G 2013 Eur. J. Phys. D 67 83
with λ1 = 2(a − c ) and λ 2 = 2(a + c ). From the form of the [16] Hofer T S 2013 Front. Chem. 1 24
(double) integral, on using polar coordinates in the (u,v)- [17] Lin C H, Lin Y C and Ho Y K 2013 Few-Body Syst. 54 2147
plane, u = R cos ϕ and v = R sin ϕ, the double integral [18] Esquivel R O, Flores-Gallegos N, Molina-Espìritu M,
Plastino A R, Angulo J C, Antolin J and Dehesa J S 2011
vanishes if i + j is odd, since the exponential is an even J. Phys. B: At. Mol. Opt. Phys. 44 175101
function of ϕ. Switching u and v in the polar coordinates, we [19] Esquivel R O, Molina-Espiritu M, Plastino A R and Dehesa J S
likewise see that it vanishes if (m + n) − (i + j ) is odd. 2014 Quantum entanglement features of elementary
The integral thus becomes chemical reactions (in preparation)
[20] Lévay P and Vrana P 2008 Phys. Rev. A 78 022329
(n, m) [21] Heisenberg W 1926 Z. Physik 38 411
n m [22] Moshinsky M 1968 Am. J. Phys. 36 52
= 2 ∑∑
i=0 j=0
( ni)( mj)( −1) λ j −(n + m − i − j + 1) 2 −(i + j + 1) 2
1 λ2 Moshinsky M 1968 Am. J. Phys. 36 763 (erratum)
[23] Pipek J and Nagy I 2009 Phys. Rev. A 79 052501
∞ ∞ [24] Yáñez R J, Plastino A R and Dehesa Eur J S 2010 Phys. J. D
× ∫−∞ ∫−∞ un+m−i−jvi+j e−u −v 2 2
du dv 56 141
[25] Benavides-Riveros C L, Gracia-Bondía J M and Várilly J C
n m
2012 Phys. Rev. A 86 022525
= 2 ∑∑
i=0 j=0
( ni)( mj)( − 1) λ j −(n + m − i − j + 1) 2 −(i + j + 1) 2
1 λ2 [26] Schilling C 2013 Phys. Rev. A 88 042105
[27] Armstrong J R, Zinner N T, Fedorov D V and Jensen A S 2011
⎛ n + m − i − j + 1⎞ ⎛ n + m + 1⎞ J. Phys. B: At. Mol. Opt. Phys. 44 055303
× Γ⎜ ⎟Γ ⎜ ⎟ [28] Eisert J, Cramer M and Plenio M B 2010 Rev. Mod. Phys. 82
⎝ 2 ⎠ ⎝ 2 ⎠ 277 see specially Section III
1 [29] Bombelli L, Koul R K, Lee J and Sorkin R D 1986 Phys. Rev.
× ⎡⎣ 1 + ( − 1)n + m − i − j ⎤⎦ ⎡⎣ 1 + ( − 1)i + j ⎤⎦ . (B.4) D 34 373
4 [30] Johnson N F and Payne M C 1991 Phys. Rev. Lett. 67 1157
[31] Srednicki M 1993 Phys. Rev. Lett. 71 666
[32] Armstrong J R, Zinner N T, Fedorov D V and Jensen A S 2012
Phys. Rev. E 85 021117
[33] Kościk P and Okopińska A 2013 Few-Body Syst. 54 1637
References [34] Wang Z, Wang A, Yang Y and Li X 2012 Commum. Theor.
Phys. 58 639
[1] Szabo A and Ostlund N 1996 Modern Quantum Chemistry [35] Schilling C, Gross D and Christandl M 2013 Phys. Rev. Lett.
(New York: Dover) 110 040404
[2] Helgaker T, Jorgensen P and Olsen J 2000 Molecular [36] Calles A and Moshinsky M 1970 Am. J. Phys. 38 456
Electronic Structure Theory (Chichester: Wiley) [37] Gidofalvi G and Mazziotti D A 2004 Phys. Rev. A 69 042511
[3] Nielsen M A and Chuang I L 2010 Quantum Computation and [38] Mazziotti D A 2012 Phys. Rev. Lett. 108 263002
Quantum Information (Cambridge: Cambridge University [39] Gritsenko O, Pernal K and Baerends E J 2005 J. Chem. Phys.
Press) 122 204102
[4] Horodecki R, Horodecki P, Horodecki M and Horodecki K [40] Dahl J P 2009 Can. J. Chem. 87 784
2009 Rev. Mod. Phys. 81 865 [41] Blanchard P, Gracia-Bondía J M and Várilly J C 2012 Int. J.
[5] Schliemann J, Cirac J I, Kuś M, Lewenstein M and Loss D Quant. Chem. 112 1134
2001 Phys. Rev. A 64 022303 [42] Littlejohn R 1986 Phys. Reports 138 193
[6] Huang Z and Kais S 2005 Chem. Phys. Lett. 413 1 [43] Pethick C J and Pitaevskii L P 2000 Phys. Rev. A 62 033609
[7] Wang H and Kais S 2007 Isr. J. Chem. 47 59 [44] Procesi C 2007 Lie Groups. An Approach Through Invariants
[8] Benatti F, Floreanini R and Marzolino U 2010 J. Phys. B: At. and Representations (New York: Springer)
Mol. Opt. Phys. 44 091001 [45] Laughlin R B 1983 Phys. Rev. Lett. 50 1395
[9] Benatti F, Floreanini R and Marzolino U 2014 Phys. Rev. A 89 [46] Plastino A R, Manzano D and Dehesa J S 2009 Europhys. Lett.
032326 86 20005
[10] Tichy M, Mintert F and Buchleitner A 2011 J. Phys. B: At. [47] Haque M, Zozulya O and Schoutens K 2007 Phys. Rev. Lett.
Mol. Opt. Phys. 44 192001 98 060401
[11] Amico L, Fazio R, Osterloh A and Vedral V 2008 Rev. Mod. [48] Iblisdir S, Latorre J I and Orús R 2007 Phys. Rev. Lett. 98
Phys. 80 517 060402
[12] Esquivel R O, Molina-Espíritu M, Salas F, Soriano C, [49] Latorre J I and Riera A 2009 J. Phys. A: Math. Theor. 42
Barrientos C, Dehesa J S and Dobado J A 2013 Decoding 504002
the building blocks of life from the perspective of quantum [50] Cardillo A, Galve F, Zueco D and Gómez-Gardeñes J 2013
information Advances in Quantum Mechanics ed P Bracken Phys. Rev. A 87 052312
doi:10.5772/55160

14