arXiv:1212.3241v2 [physics.

chem-ph] 4 Apr 2014

Strong correlations in density-functional theory: A

model of spin-charge and spin-orbital separations
Daniel Vieira
Departamento de Fsica, Universidade do Estado de Santa Catarina, Joinville, 89219-710 SC,
Brazil
E-mail: [email protected]

Abstract
It is known that the separation of electrons into spinons and chargons, the spin-charge separation, plays a decisive role when describing strongly correlated density distributions in one
dimension [Phys. Rev. B 2012, 86, 075132]. In this manuscript, we extend the investigation by
considering a model for the third electron fractionalization: the separation into spinons, chargons and orbitons the last associated with the electronic orbital degree of freedom. Specifically, we deal with two exact constraints of exchange-correlation (XC) density-functionals:
(i) The constancy of the highest occupied (HO) Kohn-Sham (KS) eigenvalues upon fractional
electron numbers, and (ii) their discontinuities at integers. By means of one-dimensional (1D)
discrete Hubbard chains and 1D H2 molecules in the continuum, we find that spin-charge separation yields almost constant HO KS eigenvalues, whereas the spin-orbital counterpart can be
decisive when describing derivative discontinuities of XC potentials at strong correlations.
To

whom correspondence should be addressed

1 Introduction
Spin and charge use to be treated as fundamental properties of ordinary electrons. However, when
confined in one dimension, interacting electrons display the unusual property of separating their
spin and charge into two independent quasiparticles: spinons and chargons. 1 Both behave just
like ordinary electrons, but: spinons have spin-1/2 and no electrical charge, while chargons are
spinless charged electrons. Recently, 2,3 an additional fractionalization was shown to occur: The
spin-orbital separation, for which spin and orbital degrees of freedom are decoupled to form the
orbitons particles with no spin and charge, carrying solely the orbital information. Both, the spincharge and spin-orbital separations have recent evidences of experimental observation, 3,4 teaching
us that ordinary electrons can be considered bounded states of spinons, chargons and orbitons.
The Kohn-Sham (KS) formalism of density-functional theory (DFT), 5,6 by construction, retains the spin, charge and orbital degrees of freedom together, once it considers an auxiliary system of noninteracting particles. In contrast, it has been shown that the separation into spinons and
chargons are decisive when dealing with strongly correlated density distributions in one dimension. 7 W. Yang et. al. have assumed strongly correlated systems as one of the modern challenges
for DFT: The challenge of strongly correlated systems is a very important frontier for DFT. To
fully understand the importance of these systems, they must be looked at within the wider realm
of electronic structure methods. 8 Here, we intend to give a contribution into this challenge: By
means of one-dimensional (1D) discrete Hubbard chains and 1D H2 molecules in the continuum,
we extend the spin-charge investigation and propose a model for the spin-orbital separation in DFT.
Specifically, we consider two exact constraints of exchange-correlation (XC) density-functionals:
(i) the constancy of the highest occupied (HO) Kohn-Sham eigenvalues upon fractional electron
numbers, and (ii) their discontinuities at integers. These constraints are usually not satisfied even
by modern approaches, and are the cause of dramatic errors when describing any generic situation
involving transport of charges. 912
In detail, we shall compare the performance of local-density functionals and their spin-charge
separation corrections, including the spin-orbital fractionalization in both cases. We show that
2

spin-charge separation yields almost constant HO KS eigenvalues, whereas the separation into
orbitons can be decisive when dealing with derivative discontinuities of XC potentials at strong
correlations.

2 Theoretical background
2.1 Fractional electron numbers
In a system with M = N + w (0 w 1) electrons, the total ground-state energy is given by 13

E(N + w) = (1 w)E(N) + wE(N + 1).

(1)

Assuming that only the HO KS orbital can be fractionally occupied, Janak 14 has proved that

E
= HO = constant,
M

(2)

where HO is the HO KS eigenvalue. The fundamental energy gap at each integer N is given by the
difference between ionization potential (IP) and electronic affinity (EA): 15
Eg (N) = IP(N) EA(N) = N+1 (N + 1) N (N).

(3)

The Kohn-Sham gap is defined as

Eg,KS (N) = N+1 (N) N (N).

(4)

Therefore, by means of Eqs. (3) and (4), one can write

Eg (N) = Eg,KS (N) + xc ,

(5)

where xc = N+1 (N +1) N+1 (N) is defined as the derivative discontinuity of the XC potential. 15
For open shell systems, Eg (N) xc , since Eg,KS (N) = 0 (in a spin-restricted KS calculation).

2.2 The spin-charge separation correction

The total Hamiltonian of a 1D interacting system is known to separate into two independent terms,
of spin and charge, schematically represented by: 1,16,17

H = H 0 + H + H = H 0 + H I ,

(6)

where H 0, H and H stand for the kinetic, spin and charge terms, respectively. Here, the spinon
densities (r) are built from uncharged spin1/2 electrons, whereas the chargon densities (r)
are built from charged spinless electrons. Spinons and chargons are semions, 18,19 that is, particles
which follow a fractional occupation statistics. At temperature T = 0, a generalization of BoseEinstein and Fermi-Dirac statistics can be written as: 18

0 fi 1/g, for i HO

fi = 0, for i > HO ,

(7)

where fi indicates the average occupation of the i orbital. Fermions are characterized by g = 1,
while bosons by g 0. Semions are half way, with g = 1/2. This picture can be also associated
with a phase change (given by ei ) induced in the wave function upon exchange of two particles.
Fermionic wave functions have = , while bosonic = 0. Semions, half way, are described by

= /2. 18 Considering spinons and chargons as independent entities, their occupied states should
follow a semion distribution, where charge and spin are separated to form spin1/2 spinons and
spinless chargons. In Figure 1 (b) we display a strongly interacting semion distribution: spinons,
with no charge, doubly occupy each state, whereas chargons, with the entire charges, are characterized by single occupations. 20
As shown in Figure 1 (c), it has been proposed 7 that the occupied states of a noninteracting KS

system are built by retaining spin and charge together, at expense of the presence of holons (the
chargon antiparticles), whose densities are given by + (r). The KS potential can thus be written
as:
vKS [n](r) = vext (r) +

HI
HI

,
nKS (r) + (r)

(8)

with nKS (r) n(r) = (r) + (r).

In a non-magnetic LDA formulation, for Neven N (Neven + 2), we have that:
Neven /2

n(r) =

=, i=0

|i, (r)|2 +

=,

w |(Neven+2)/2, (r)|2 ,

Neven
1
|i, (r)|2
(r) =

2 =, i=0

(9)

1
+ w |Neven+1, (r)|2 + |Neven +2, (r)|2 ,
2 =,

(10)

with 0 w 1, allowing fractional occupation. Neven = 0, 2, 4, 6, ..., and i, (r) are the KS eigenvectors, with i=0, (r) = 0. Based on Eqs. (8) to (10), the spin-charge separation correction
(SCSC) is written as: 7
SCSC/approx

vKS

[n](r)
[n](r) = vext (r) + vH [n](r) + vapprox
xc
vH [ + ](r) vapprox
[ + ](r)
xc
vext (r) + vH [n](r) + vSCSC
[n](r),
xc

(11)

where vH and vxc label the Hartree and XC potentials, respectively. The XC SCSC potential of
Eq. (11) is not a functional derivative of a known XC energy functional, that is, it is a direct correction to the KS potential. Even though model potentials may suffer from conceptual drawbacks
when calculating the associated energy functionals, they are suggested to be a promising route
to new developments in DFT. 2123 For example, it is known that for KS potentials which are not
functional derivatives, different paths of assigning energy functionals give different results, 24,25

evidencing an impossibility of unambiguously assigning energy values. On the other hand, while
an ambiguous energy assignment represents a conceptual inconsistency, it is not necessarily meaningless, since the use of potentials as seeds can be also regarded as an interesting strategy for constructing new energy functionals. 24,25 In this sense, testing the SCSC model potential of Eq. (11)
under different paths of assigning energies is certainly a topic of investigation, which, however, we
judge to be out of the purposes of this manuscript.

(b)

Strongly
interacting system

(c) Kohn-Sham system

(a) Noninteracting
system

spins + charges
together

spinons + chargons
separated

spins + charges together

+ spinless holons

fractional charge occupation

Figure 1: (Color online) Schematic representation of occupied quantum states. (a) Nointeracting
system, with spins and charges together. (b) Strongly interacting system, with spins and charges
separated into spinons and chargons, respectively. (c) Noninteracting Kohn-Sham system: Spins
and charges together at expense of the presence of spinless holons (the chargons antiparticles).
Fractional charge occupations are indicated by the dashed squares.

3 Results
3.1 One-dimensional Hubbard chains
In one dimension, and in second-quantized notation, the Hubbard model 26 (1DHM) is defined as
L

j,

j,

H = t cj c j+1, + H.c. +U cj c j cj c j + V j cj c j, ,

(12)

where L is the number of sites, t is the amplitude for hopping between neighboring sites and V j
is an external potential acting on site j. Occupation of each site is limited to two particles, necessarily of opposite spin. 27 The Eq. (12) takes the role of a general Hamiltonian which describes
interacting electrons under two limits: The discrete space and the on-site electron-electron interaction U . The 1DHM is a very instructive many-body laboratory, which enables us to investigate the
effects of changing electronic correlation by just controlling the values of U (which can be varied
continuously). A different model, in the continuum and with long-ranged Coulomb interaction,
will be considered in Section 3.2.
With the density n(r) replaced by the on-site occupation n( j), the Hohenberg-Kohn and KS
theorems of DFT also hold for the 1DHM. 28 In terms of this variable, local-(spin)-density approximations for Hubbard chains and rings have been constructed, 2932 including a recent extension
to finite temperatures. 33 In this section, we chose the fully numerical Bethe-Ansatz local-density
approximation (BALDA-FN) 31 as the reference XC functional, considering only non-polarized
systems with N = N . There is no conceptual difference between the BALDA-FN functional and
other L(S)DA approximations applied to 3D systems. The only difference is that BALDA-FN is
based on exact solutions of 1D homogeneous Hubbard chains, and not on accurate solutions of 3D
homogeneous systems. Throughout this section, we shall denote the SCSC/BALDA-FN approach
simply as SCSC approximation.
Considering open Hubbard chains, in Figure 2 (a)-(b) we plot HO versus fractional charge
occupations for two values of U . The exact data, for all cases presented in this section, come
from a Lanczos diagonalization of the Hubbard Hamiltonian of Eq. (12). The BALDA yields a
quasi-linear dependence for HO . The slope of each straight line tend to increase as U is increased,
with Eg = 0 at odd integers (open shells). Since we are dealing with 1D systems, the BALDA
naturally yields Eg = 0 at even integers (closed shells), which, however, tend to be underestimated
as U is increased. The SCSC performs much better, yielding almost constant values for HO .
The associated energy gaps, on the other hand, are also (i) equal to zero (open shells) or (ii)
underestimated (closed shells), combined with severely incorrect ionization potentials (IP = -HO ).

-1.0
-1.5

(a)

L=15

(b)

U=1.0t

L=15

U=20.0t

-1.2
Exact

-1.6

BALDA

-1.7

-1.4

SCSC

-1.6

-1.8

-1.8

-1.9
-2.0

-2.0
0

Figure 2: (Color online) Open Hubbard chains of L = 15 sites: HO KS eigenvalues (HO ) for
0 < N 5.
3.1.1 Spin-orbital separation
The difficulty with energy gaps is intrinsic of most available density functionals. In this context,
beyond spinons and chargons, we propose here a model for the third electron fractionalization:
into orbitons. By definition, orbitons are excitations of the orbital degrees of freedom of electrons,
which behave like spinless and uncharged particles. As consequence, when dealing with the noninteracting KS electrons of Figure 1 (c), we propose the KS eigenvalues should be increased by a
constant , which is equivalent to include the presence of an anti-orbiton, the extension of the
SCSC idea of Eq. (11). Thus, we propose the XC potential to be given by
vapprox+
[n]( j) = vapprox
[n]( j) + ,
xc
xc

(13)

= (HO+ ) HO ; = 0, 1, 2, ...

(14)

with

As conceptually thought, the parameter is given by:

Neven
,
2

(15)

for
Neven 1 < N Neven + 1.

(16)

For example, in the strongly interacting limit, = 1 for 1 < N 3 and = 2 for 3 < N 5.
Noninteracting systems (or interacting systems with 0 < N 1) should have = 0, indicating the
absence of spin-orbital separation. A schematic representation of this orbital separation model is
shown in Figure 3.

Kohn-Sham system (h=1)

Kohn-Sham system (h=2)

e(HO+h)
e(HO+h)
eHO

eHO

anti-orbiton

Figure 3: (Color online) Noninteracting Kohn-Sham systems: Spins and charges together, at expense of the presence of spinless holons and the conjectured anti-orbitons. The anti-orbitons are
represented by excitations between orbital levels.
as the HO KS eigenvalues yielded by the v
Let us define HO
xc

approx+

potential. By means of

can be determined by means of the following KS equation:

Eqs. (13) and (14), HO

{ts +vext ( j) + vH [n]( j) +vapprox

[n]( j)} i ( j) = i i ( j),
xc

(17)

= (HO+ ) .
HO

(18)

with

The orbital degrees of freedom we mention here come from the solution of the Hubbard-like KS
equations, under the Hubbard Hamiltonian of Eq. (12), and does not come from a multi-band
Hubbard model.
In Figure 4 (a)-(d) we plot the results for the BALDA+ and SCSC+. The inclusion of

yields Eg = 0 for open and closed shells. At strong correlations (when U is increased), the curves
of SCSC+ fit in very good agreement with the exact data, including the constancy of HO , the
derivative discontinuities of the XC potentials and the ionization potentials. The BALDA+ also
yields correct energy gaps, but combined with incorrect IPs and linear dependence of HO between
integers.

-1.0
-1.0 (a)

L=15

(b)

U=1.0t
-1.2

-1.2

L=15

Exact
BALDA +

-1.4

U=5.0t

-1.4

SCSC +

-1.6

-1.6

-1.8

-1.8
-2.0

-2.0
0

-1.0

-1.0

(c)
-1.2

(d)

L=15

-1.2

U=10.0t

-1.4

-1.4

-1.6

-1.6

-1.8

-1.8

-2.0

L=15

U=20.0t

-2.0
0

Figure 4: (Color online) Open Hubbard chains of L = 15 sites: HO KS eigenvalues (HO ) obtained
by including the factor of Eqs. (13) and (14), with given by Eqs. (15) and (16).

3.2 One-dimensional H2 molecule in the continuum

In dissociation processes, to preserve neutrality of isolated atoms, it is known that the increment
of nuclear separation uses to be followed by an increment of correlation energy density. 8 In order to test the spinon-chargon-orbiton approach under this type of delimitation weakly-strongly
10

correlated systems, in this section we consider the dissociation of a 1D H2 molecule.

In contrast with the discrete Hubbard chains, a different way of describing interacting electrons in one-dimension may take into account (a) continuum space and (b) long-ranged Coulomb
interaction. This is a special case of interest, since chemistry in general is described by using both
ingredients. The electronic Hamiltonian of a 1D H2 molecule can be written as:
2

H =

j=1

1
1 d2
+ vext (x j ) +
2
2 dx j
2

vint (x j , xk ).

(19)

j,k=1
( j=k)

A common choice, which avoids singularities of the Coulomb interaction, is the soft-Coulomb
potential: 3437
vint (x j , xk ) =

q j qk
b2 + (x j xk )2

(20)

where q j and qk are the electron charges placed at positions x j and xk , respectively, and b is a
softening parameter. The same idea holds for the electron-nucleus interactions:
2

Zi q j

i=1

b2 + x j Xi

vext (x j ) =

(21)

labelling the nuclear charge Zi placed at position Xi . The repulsive nucleus-nucleus interaction
(not showed) is also described by means of a soft-Coulomb potential (under the same parameter
b). Specifically for b = 1, a local-density approximation (1DLDA) has been used to describe
one-dimensional atoms and molecules. 3537 Here we intend to use the 1DLDA and implement
the corresponding SCSC+ correction to it. In this section, we shall denote the SCSC/1DLDA
approach simply as SCSC approximation.
In Figure 5 (a), considering 0 < N 2, we plot HO for a 1D H2 molecule with nuclear separation d = X2 X1 = 1.6 a.u., which is the exact equilibrium distance. 36 The values of HO follow the
same trend as observed in the previous figures, displaying typical errors of weakly correlated situations. As the separation d is increased, on the other hand, the SCSC+ yields very accurate data
for HO , as seen from Figure 5 (b), valid for d = 4.0 a.u.: It is a clear delimitation between weakly
11

and strongly correlated systems, under a fixed value of b. These results are in accordance with the
observation that correlation energy density is zero in isolated H atoms but substantial around each
H atom in H2 at long distances. 38,39

0.2

(a)

d = 1.6 a.u.

-0.2

0.0

(b)

d = 4.0 a.u.

Exact

-0.2

1DLDA
SCSC+

-0.4

1D H

-0.4

molecule
-0.6

-0.6
-0.8

-0.8

-1.0
0.0

0.5

1.0

1.5

0.0

2.0

0.5

1.0

1.5

2.0

Figure 5: (Color online) One-dimensional H2 molecule in the continuum. (a) HO KS eigenvalues

(HO ) for 0 < N 2 and nuclear separation d = 1.6 a.u. (b) The same as before, but for d = 4.0 a.u.
Remember: for 1DLDA, = 0 for 0 < N 2; in the case of SCSC+, = 0 for 0 < N 1 and
= 1 for 1 < N 2. The exact data come from total energies, directly extracted from the
literature. 36
In summary, we can conclude that: (l) spin-charge separation, when included by means of the
SCSC XC potential, yields almost constant highest occupied KS eigenvalues; (ll) in the limits of
strong correlations, the model we proposed here for the spin-orbital separation yields accurate energy gaps for both, open and closed shells (in association with correct derivative discontinuities
of the XC potentials). Considering (l) and (ll), we can argue that when dealing with strong correlations, electrons should not be treated as unique particles. Instead, the separation into spinons,
chargons and orbitons can be crucial. The way to include it in a noninteracting KS calculation,
which by construction retains spin, charge and orbital degrees of freedom together, is a combination of (l) and (ll).

4 Conclusions
Constancy of HO : The BALDA-FN and 1DLDA yielded typical results attributed to delocalization
12

errors of density functionals: Incorrect linear behavior of HO upon fractional charge occupation.
On the other hand, the SCSC/BALDA-FN and SCSC/1DLDA, which have been especially conceived to deal with strong correlations, yielded almost constant values for HO .
Derivative discontinuity: By means of Eq. (13), we proposed the inclusion of the extra electron fractionalization into orbitons which yielded accurate energy gaps at strong correlations.
We observed a delimitation between weak and strong correlations by simply changing the on-site
interaction U , in open Hubbard chains, or changing the nuclear separation (at fixed interaction
parameter b) in the dissociation of a 1D H2 molecule. It has been recently pointed out that, beyond
strong correlations, it is a basic challenge to understand whether the KS orbitals and eigenvalues
have any further significance. 8 The spinon-chargon-orbiton separation can therefore be a step into
this direction.
Energy functionals: Testing the SCSC and SCSC+ potentials under different paths of assigning energy functionals 24,25 is a topic of future investigation. The resulting energy functionals may
then be used in the derivation of even more accurate XC potentials, which, for example, could
suitably link the weakly and strongly correlated regimes.
Extensions: Possible direct generalizations to higher dimensions, especially to three-dimensional
(3D) systems, depend on a particular question: Is electron fractionalization also possible in 3D? At
our knowledge, this is still a topic of debate, and therefore deserves further investigation. Nevertheless, indirect generalizations are certainly possible, as the case of quasi one-dimensional systems.
It should be noted that a successful alternative route to obtain accurate derivative discontinuities
and constancy of HO the SCE approach has been described in recent letters. 40,41 In this sense,
understanding possible connections between using SCSC+ and other accurate approaches is also
a topic of future investigation.
Acknowledgments: We thank Vivaldo L. Campo Jr. for his code of Lanczos exact diagonalization and for the original version of the BALDA-FN code.

13

References
(1) T. Giamarchi, Quantum Physics in One Dimension (Oxford University Press, Oxford, 2003).
(2) Wohlfeld, K.; Daghofer, M.; Nishimoto, S.; Khaliullin, G.; van den Brink, J. Phys. Rev. Lett.
2011, 107, 147201.
(3) Schlappa, J.; Wohlfeld, K.; Zhou, K. J.; Mourigal, M.; Haverkort, M. W.; Strocov, V. N.;
Hozoi, L.; Monney, C.; Nishimoto, S.; Singh, S.; Revcolevschi, A.; Caux, J.-S.; Patthey, L.;
Rnnow, H. M.; van den Brink, J.; Schmitt, T. Nature 2012, 485, 82.
(4) Jompol, Y.; Ford, C. J. B.; Griffiths, J. P.; Farrer, I.; Jones, G. A. C.; Anderson, D.; Ritchie,
D. A.; Silk, T. W.; Schofield, A. J. Science 2009, 325, 597.
(5) Kohn, W. Rev. Mod. Phys. 1999, 71, 1253.
(6) Capelle, K. Braz. J. Phys. 2006, 36, 1318.
(7) Vieira, D. Phys. Rev. B 2012, 86, 075132.
(8) Cohen, A. J.; Mori-Snchez, P.; Yang, W. Chem. Rev. 2012, 112, 289.
(9) Ruzsinszky, A.; Perdew, J. P.; Csonka, G. I.; Vydrov, O. A.; Scuseria, G. E. J. Chem. Phys.
2006, 125, 194112.
(10) Kmmel, S.; Kronik, L. Rev. Mod. Phys. 2008, 80, 3.
(11) Cohen, A. J.; Mori-Snchez, P.; Yang, W. Science 2008, 321, 792.
(12) Mori-Snchez, P.; Cohen, A. J.; Yang, W. Phys. Rev. Lett. 2008, 100, 146401.
(13) Perdew, J. P.; Parr, R. G.; Levy, M.; Balduz Jr., J. L. Phys. Rev. Lett. 1982, 49, 1691.
(14) Janak, J. Phys. Rev. B 1978, 18, 7165.
(15) Capelle, K.; Vignale, G.; Ullrich, C. A. Phys. Rev. B 2010, 81, 125114.
14

(16) Voit, J. Rep. Prog. Phys. 1995, 58, 977.

(17) Miranda, E. Braz. J. Phys. 2003, 33, 3.
(18) A. Khare, Fractional Statistics and Quantum Theory (World Scientific Publishing, 2005).
(19) Bernevig, B.A.; Giuliano, D.; Laughlin, R. B. Phys. Rev. Lett. 2001, 87, 177206.
(20) For example, consider a two-electron system of opposite spins. In the limit of strong electronelectron interaction, the situation is equivalent to a system of noninteracting spinless electrons, which singly occupy the two lowest quantum states (in the ground-state).
(21) Karolewski, A.; Armiento, R.; Kmmel, S. J. Chem. Theory Comput. 2009, 5, 712.
(22) Oliveira, M. J. T.; Rsnen, E.; Pittalis, S.; Marques, M. A. L. J. Chem. Theory Comput.
2010, 6, 3664.
(23) Boguslawski, K.; Jacob, C. R.; Reiher, M. J. Chem. Phys. 2013, 138, 044111.
(24) Gaiduk, A. P.; Staroverov, V. N. J. Chem. Phys. 2012, 136, 064116.
(25) Elkind, P. D.; Staroverov, V. N. J. Chem. Phys. 2012, 136, 124115.
(26) Hubbard, J. Proc. Roy. Soc. (London) A 1963, 276, 238.
(27) For this reason, in spite of using the XC nomenclature, we shall deal only with correlation.
(28) Schnhammer, K.; Gunnarsson, O.; Noack, R. M. Phys. Rev. B 1995, 52, 2504.
(29) Lima, N. A.; Silva, M. F.; Oliveira, L. N.; Capelle, K. Phys. Rev. Lett. 2003, 90, 146402.
(30) Frana, V. V.; Vieira, D.; Capelle, K. New J. Phys. 2012, 14, 073021.
(31) Xianlong, G.; Polini, M.; Tosi, M. P.; Campo, V. L.; Capelle, K.; Rigol, M. Phys. Rev. B
2006, 73, 165120.
(32) Capelle, K.; Campo, V. L. Phys. Rep. 2013, 528, 91.
15

(33) Xianlong, G.; Chen, A-H.; Tokatly, I. V.; Kurth, S. Phys. Rev. B 2012, 86, 235139.
(34) Tempel, D. G.; Martnez, T. J.; Maitra, N. T. J. Chem. Theory Comput. 2009, 5, 770.
(35) Helbig, N.; Fuks, J. I.; Casula, M.; Verstraete, M. J.; Marques, M. A. L.; Tokatly, I. V.; Rubio,
A. Phys. Rev. A 2011, 83, 032503.
(36) Wagner, L. O.; Stoudenmire, E. M.; Burke, K.; White, S. R. Phys. Chem. Chem. Phys. 2012,
14, 8581.
(37) Stoudenmire, E. M.; Wagner, L. O.; White, S. R.; Burke, K. Phys. Rev. Lett. 2012, 109,
056402.
(38) Sle, P.; Gritsenko, O. V.; Nagy, A.; Baerends, E. J. J. Chem. Phys. 1995, 103, 10085.
(39) Baerends, E. J. Phys. Rev. Lett. 2001, 87, 133004.
(40) Malet, F.; Gori-Giorgi, P. Phys. Rev. Lett. 2012, 109, 246402.
(41) Mirtschink, A.; Seidl, M.; Gori-Giorgi, P. Phys. Rev. Lett. 2013, 111, 126402.

16