Homogeneous Electron Liquid in Arbitrary Dimensions Exchange and Correlation Using the Singwi-Tosi-Land-Sj olander Approach L. V. Duc Pham1Pascal Sattler1Miguel A. L. Marques1and Carlos L. Benavides-Riveros2 3

2025-04-29 0 0 4.32MB 13 页 10玖币
侵权投诉
Homogeneous Electron Liquid in Arbitrary Dimensions:
Exchange and Correlation Using the Singwi-Tosi-Land-Sj¨olander Approach
L. V. Duc Pham,1Pascal Sattler,1Miguel A. L. Marques,1and Carlos L. Benavides-Riveros2, 3,
1Institut f¨ur Physik, Martin-Luther-Universit¨at Halle-Wittenberg, 06120 Halle (Saale), Germany
2Pitaevskii BEC Center, CNR-INO and Dipartimento di Fisica, Universit`a di Trento, I-38123 Trento, Italy
3Max Planck Institute for the Physics of Complex Systems, N¨othnitzer Str. 38, 01187, Dresden, Germany
(Dated: March 3, 2023)
The ground states of the homogeneous electron gas and the homogeneous electron liquid are
cornerstones in quantum physics and chemistry. They are archetypal systems in the regime of
slowly varying densities in which the exchange-correlation energy can be estimated with a myriad of
methods. For high densities, the behaviour the energy is well-known for 1, 2, and 3 dimensions. Here,
we extend this model to arbitrary integer dimensions, and compute its correlation energy beyond
the random phase approximation (RPA), using the celebrated approach developed by Singwi, Tosi,
Land, and Sj¨olander (STLS), which is known to be remarkably accurate in the description of the
full electronic density response for 2Dand 3D, both in the paramagnetic and ferromagnetic ground
states. For higher dimensions, we compare the results obtained for the correlation energy using the
STLS method with the values previously obtained using RPA. We found that at high dimensions
STLS tends to be more physical in the sense that the infamous sum rules are better satisfied by the
theory. We furthermore illustrate the importance of the plasmon contribution to STLS theory.
I. INTRODUCTION
The ground states of the homogeneous electron gas
(HEG) and the homogeneous electron liquid (HEL) have
played a prominent role in the modelling and under-
standing of a wide range of interacting electronic sys-
tems [1–12]. These systems are some of the most im-
portant models of choice to develop, improve and bench-
mark many approximate approaches to the full many-
electron problem, including some of the most popu-
lar exchange-correlation functionals of density functional
theory (DFT) [13–16].
An important question about these models is the de-
pendency of the correlation energy, and other physical
quantities, on D, the dimension of the physical space
in which the gas/liquid is embedded. Indeed, there is
a wealth of experimental electronic setups in which one
or two of the physical dimensions are much smaller than
the remaining ones. They can thus be modeled as one-
or two-dimensional quantum systems [17–21]. Further-
more, reduced dimensional systems often exhibit notable
physical properties, ranging from Luttinger physics [22]
to Moir´e superlattices [23]. More recently, progress in
the fabrication of artificial materials is paving the way
for the realization of non-integer dimensions, as fractal
substrates (e.g., Sierpi´nski carpets of bulk Cu) confining
electron gases [24–26]. The possibility to circumvent the
von Neumann-Wigner theorem or to produce unconven-
tional topological phases by engineering (or mimicking)
additional synthetic dimensions [27–30] also highlights
the importance of realizing, studying and understanding
interacting fermionic and electronic systems embedded in
non-conventional dimensions.
cl.benavidesriveros@unitn.it
For the high-density spin-unpolarized HEG in D=
1,2, and 3, the energy per electron can be expanded in
terms of rs, the Wigner-Seitz radius, as follows [31]:
εD(rs0) = aD
r2
sbD
rs
+cDln rs+O(r0
s).(1)
The constants aD,bDand cDare independent of rs, and
their functional form in terms of the dimension Dis well
known [8]. Quite remarkably, the logarithm that appears
in Eq. (1) for the 3D case is due to the long range of the
Coulomb repulsion, and cannot be obtained from stan-
dard second-order perturbation theory [32]. In a recent
work [33], the HEG was extended to arbitrary integer
dimensions. It was found a very different behavior for
D > 3: the leading term of the correlation energy does
not depend on the logarithm of rs[as in Eq. (1)], but
instead scales polynomially as cD/rγD
s, with the expo-
nent γD= (D3)/(D1). In the large-Dlimit, the
value of cDwas found to depend linearly with the di-
mension. This result was originally obtained within the
random-phase approximation (RPA) by summing all the
ring diagrams to infinite order.
While RPA is known to be exact in the limit of the
dense gas, includes long-range interactions automati-
cally, and is applicable to systems where finite-order
many-body perturbation theories break down [34], it
has well-known deficiencies at the metallic (intermedi-
ate) and low densities of the typical HEL (1 rs6).
Quantum Monte-Carlo (QMC) is an option for those
regimes [12, 35], but there are other high-quality ap-
proaches such as the celebrated Singwi, Tosi, Land, and
Sj¨olander (STLS) method that provides results compa-
rable to QMC [36]. This method attempts to tackle in
an approximate manner both the exchange and electronic
correlations through a local-field correction. As such, this
scheme is often surprisingly accurate in the description
of the full electronic density response and is commonly
arXiv:2210.03024v3 [cond-mat.str-el] 1 Mar 2023
2
used to investigate quasi-one-dimensional [37], inhomo-
geneneuos [38–40] and warm dense electron liquids [41–
43]. It has also inspired the development of new func-
tional methods that explicitly retain the dynamical and
non-local nature of electronic correlations, while properly
accounting for the exchange contribution [44–48].
The purpose of this paper is to calculate the correla-
tion energy of the HEL for arbitrary integer dimensions
(in particular for D > 3) with the STLS method. As ex-
pected, the value of this portion of the energy improves
significantly with respect to the RPA result (that tends
to over-correlate the liquid). The remainder of this paper
is organized as follows. In Section II we review the main
STLS equations and rewrite them explicitly in arbitrary
Ddimensions. As a result, we can compute the Lindhard
polarizability, the structure factor, and the so-called local
field correction in the Hartree-Fock approximation, pro-
viding explicit formulae for some representative systems.
In Section III we explain how the correlation energy is
computed in this scheme, and discuss the fully polarized
case. In Sections V and VI we present and discuss our
numerical results for the exchange and correlation ener-
gies, the pair density, and its Fourier transform. We also
discuss the fulfillment of the compressibility sum rule.
Finally, in Section VII we present our main conclusions.
Two Appendices that give further technical details on
our calculations are presented at the end of the text.
II. STLS THEORY IN DDIMENSIONS
In this section we review the main STLS equations
and write them explicitly in arbitrary Ddimensions. We
follow the standard notation, namely, nis the D-dimen-
sional particle density, Ω is the volume occupied by the
electronic liquid, and qFis the usual Fermi wavelength.
The STLS theory departs from RPA (to include short-
range correlation between electrons) by writing the two-
particle density distribution f2(r,p,r0,p0, t) as follows:
f2(r,p,r0,p0, t) = f(r,p, t)f(r0,p0, t)g(rr0),(2)
where f(r,p, t) is the one-particle phase-space density
and g(r) is the equilibrium, static pair distribution func-
tion. Eq. (2) can be seen as an ansatz that terminates
the hierarchy that otherwise would write the two-particle
distribution function in terms of the three-particle distri-
bution function, and so on. This leads to the following
density-density response function [36]:
χD(q, ω) = χ0
D(q, ω)
1Φ(q)[1 GD(q)]χ0
D(q, ω).(3)
Here χ0
D(q, ω) is the Lindhard polarizability, i.e., the in-
homogeneous non-interacting density response function
of an ideal Fermi gas in Ddimensions, GD(q) is the local
field correction, and Φ(q) is the D-dimensional Fourier
transform of the Coulomb potential:
Φ(q) = (4π)D1
2ΓD1
2
qD1,(4)
where Γ denotes the gamma function. The presence of
GD(q) in Eq. (3) is the key feature of the STLS equations
that gives the “beyond RPA” flavor to the theory. This
local field correction is a direct result of the short-range
correlation between the electrons. In arbitrary dimen-
sions, it is given by:
GD(q) = 1
nZq0·qqD3
q0D1[SD(qq0)1] dDq0
(2π)D,
(5)
with SD(q) being the structure factor. We can simplify
this integral to a two-dimensional one by substituting
qq0=tand using the fact that SD(q) = SD(q) in
homogeneous systems. Afterwards, we rewrite the inte-
gral using D-dimensional spherical coordinates, where q
is parallel to the Dth-axis, and integrate over all angles
except the angle θbetween qand tto obtain:
GD(q) = qD3
(2π)Dn
2πD1
2
ΓD1
2Z
0Zπ
0
[SD(t)1]×
[q2tD1qtDcos θ](sin θ)D2
(q2+t22qt cos θ)D1
2
dt . (6)
Note in passing that in the 2Dcase we must use po-
lar coordinates instead of the spherical coordinates and
obtain:
G2(q) = 1
(2π)2nZ
0Z2π
0
[S2(t)1]×
qt t2cos θ
(q2+t22qt cos θ)1/2dt . (7)
These equations, together with the equation for the
dielectric function, 1/D(q, ω) = 1 + Φ(q)χD(q, ω), lead
to an equation for the dielectric function within the STLS
theory, namely:
D(q, ω)=1Φ(q)χ0
D(q, ω)
1 + GD(q)Φ(q)χ0
D(q, ω).(8)
Finally, the relation between the structure factor and
the dielectric function D(q, ω) (see, for instance, [49])
can be easily generalized for arbitrary dimensions:
SD(q) = 1
πnΦ(q)Z
0
Im 1
D(q, ω)dω . (9)
We can improve the readability of this equation by sep-
arating the contributions of the single-particle and plas-
mon excitations [50]:
SD(q) = 1
πnΦ(q)ZqvF+q2
2
0
Im 1
D(q, ω)
+1
Φ(q)nReD(q, ω)
ω 1
δ(ωωp(q)) ,(10)
3
where vFis the Fermi velocity and ωp(q) is the plasmon
dispersion [51]. Equation 10, Eq. 6 and Eq. 8 form the
core of the STLS set of equations. In this framework,
they can be evaluated self-consistently. Notice, indeed,
that GD(q) and SD(q) can be written symbolically as
GD=F1[SD] and SD=F2[GD], indicating the existence
of the mutual functional relations introduced above [47].
We should reiterate here that the crucial aspect of the
STLS method is the appearance of the density-density
pair distribution g(rr0) in the decoupling of the equa-
tion of motion (2). Despite the crudeness of such a fac-
torization scheme, the formalism gives excellent correla-
tion energies for both 3D and 2D homogeneous electron
gases. Unfortunately, the factorization also results in a
number of well-known shortcomings, including negative
values of g(rr0) for sufficiently large values of rs, and
the failure to satisfy the compressibility sum rule (i.e., at
long wavelengths the exact screened density response is
determined by the isothermal compressibility). The first
problem, the un-physical behavior of the pair distribu-
tion, is counter-balanced by the fact that the exchange-
correlation energy is in reality an integral over g(r): it
turns out that its value is still quite reasonable in the
metallic range as it benefits from an error cancellation
[43]. As we will see below, this result, known for the
case of 2D and 3D, does also hold for larger dimen-
sions. To tackle the second problem, Vashista and Singwi
[52] provided a correction to the STLS method that
established the correct compressibility sum rule in the
metallic-density regime. This is however a partial solu-
tion as both the original STLS method and the Vashista-
Singwi extension under-estimate the exchange energy, as
pointed out by Sham [53].
A. Real part of the Lindhard function for D= 5,7
The expressions for χ0
2(q, ω) and χ0
3(q, ω) are widely
known since long ago (see for instance [31, 54]), but
higher dimensional expressions are missing in the litera-
ture. In Appendix A we detail their calculation for higher
dimensional settings by performing a linear perturbation
from equilibrium. We sketch here the main results.
The real part of χ0
D(q, ω) can be written explicitly as:
Reχ0
D(q, ω) = 2
(2π)DPZΘ(qF− |p1
2q|)Θ(|p+1
2q| − qF)
ωp·qdDp
2
(2π)DPZΘ(|p1
2q| − qF)Θ(qF− |p+1
2q|)
ωp·qdDp,(11)
where Pdenotes the principal value. By evaluating these integrals, it is possible to obtain analytical expressions for
specific cases. For instance, for D= 5 and D= 7 one gets:
Reχ0
5(q, ω) = q3
F
8π3(1
96˜q5"3
2˜q22˜ω4+ 24˜q412˜q2(˜q22˜ω)2ln
2˜q˜q2+ 2˜ω
2˜q+ ˜q22˜ω
+3
2˜q2+ 2˜ω4+ 24˜q412˜q2(˜q2+ 2˜ω)2ln
2˜q˜q22˜ω
2˜q+ ˜q2+ 2˜ω
+ 12˜q716˜q5+ 144˜q3˜ω2#2
3)
(12)
and
Reχ0
7(q, ω) = q5
F
368640π4˜q5(60(16˜q4+ 3(˜q2+ 2˜ω)412(˜q3+ 2˜q˜ω)2)15(˜q2+ 2˜ω)6
˜q2ln
2˜q˜q22˜ω
2˜q+ ˜q2+ 2˜ω
+60(16˜q4+ 3(˜q22˜ω)412(˜q32˜q˜ω)2)15(˜q22˜ω)6
˜q2ln
2˜q˜q2+ 2˜ω
2˜q+ ˜q22˜ω
4224˜q5+ 1280˜q3(˜q4+ 12˜ω2)120˜q(˜q8+ 40˜q4˜ω2+ 80˜ω4)).(13)
A similar calculation gives a closed expression for the imaginary part:
Imχ0
D(q, ω) =
h(D)1
˜q1ν2
D1
21ν2
+D1
2˜ω < ˜q˜q2
2and ˜q < 2
0 ˜ω < ˜q˜q2
2and ˜q > 2
h(D)1
˜q1ν2
D1
2˜q˜q2
2˜ω˜q+˜q2
2
0 ˜ω > ˜q+˜q2
2
(14)
摘要:

HomogeneousElectronLiquidinArbitraryDimensions:ExchangeandCorrelationUsingtheSingwi-Tosi-Land-SjolanderApproachL.V.DucPham,1PascalSattler,1MiguelA.L.Marques,1andCarlosL.Benavides-Riveros2,3,1InstitutfurPhysik,Martin-Luther-UniversitatHalle-Wittenberg,06120Halle(Saale),Germany2PitaevskiiBECCenter...

展开>> 收起<<
Homogeneous Electron Liquid in Arbitrary Dimensions Exchange and Correlation Using the Singwi-Tosi-Land-Sj olander Approach L. V. Duc Pham1Pascal Sattler1Miguel A. L. Marques1and Carlos L. Benavides-Riveros2 3.pdf

共13页,预览3页

还剩页未读, 继续阅读

声明:本站为文档C2C交易模式,即用户上传的文档直接被用户下载,本站只是中间服务平台,本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私,请立即通知玖贝云文库,我们立即给予删除!
分类:图书资源 价格:10玖币 属性:13 页 大小:4.32MB 格式:PDF 时间:2025-04-29

开通VIP享超值会员特权

  • 多端同步记录
  • 高速下载文档
  • 免费文档工具
  • 分享文档赚钱
  • 每日登录抽奖
  • 优质衍生服务
/ 13
客服
关注