RAPID COMMUNICATIONS

Selected for a Viewpoint in Physics

PHYSICAL REVIEW B 89, 241108(R) (2014)

Dynamical many-body corrections to the residual resistivity of metals

V. U. Nazarov,1 G. Vignale,2 and Y.-C. Chang1
1
Research Center for Applied Sciences, Academia Sinica, Taipei 11529, Taiwan
2
Department of Physics and Astronomy, University of Missouri, Columbia, Missouri 65211, USA
(Received 4 March 2014; published 30 June 2014)

The residual resistivity of metals at the absolute zero of temperature is usually understood in terms of electrons
scattering from random impurities. This mechanism, however, does not take into account dynamical many-body
effects, which cannot be described in terms of a static electron-impurity potential. Here we show that dynamical
corrections to the resistivity, already known to play a role in nanoscale conductors, are of quantitative importance
in the calculation of the residual resistivity of simple metals, and lead to a significantly improved agreement
between theory and experiment in the case of impurities embedded in an Al host. Our calculations are based
on a recently proposed form of the time-dependent many-body exchange-correlation potential, which is derived
from the time-dependent current density functional theory. Surprisingly, we find that the largest correction to the
residual resistivity arises from the real part of the exchange-correlation kernel of time-dependent current density
functional theory, rather than from its imaginary part. This unexpected result is shown to be consistent with recent
theories of the dynamical corrections to the resistivity of nanoscale conductors.

DOI: 10.1103/PhysRevB.89.241108 PACS number(s): 31.15.ee, 71.45.Gm, 72.15.−v

The calculation of the electrical resistance of conductors (TDDFT) [11,18]. Corrections to the resistance, which arise
is one of the basic problems in the theory of electronic from the dynamical character of the effective dressed potential,
transport [1]. At the absolute zero of temperature the crystal is what we mean when we talk of “dynamical many-body
lattice becomes irrelevant (no phonons) and the origin of the corrections.” The existence of dynamical corrections to the
resistance is traced to electron-impurity and electron-electron resistivity was first implicitly recognized in the nonequilibrium
scattering. Electron-electron scattering conserves momentum, Green’s function formalism of Meir and Wingreen [9], but the
so it is customary to assume that electron-impurity scattering correction was believed to vanish at zero temperature. It was
plays by far the dominant role. This is the basis for the standard only in later work, based on time-dependent current density
theory of residual resistivity in normal metals in the diffusive functional theory (TDCDFT) [19,20], that the dynamical
regime. Also in coherent nanoscale conductors, where the corrections were explicitly formulated in terms of a dynamical
diffusive model is no longer applicable, the resistance can xc potential and were shown to survive in the limit of zero
be expressed in terms of the transmission probability of temperature [21–26]. In these papers the additional dynamical
noninteracting electrons from one lead to another according resistance of a conductor was interpreted as a manifestation
to the Landauer-Büttiker formula [2–5]. Electron-electron of the many-body electronic viscosity [27,28] affecting the
interaction corrections to the resistivity are known to exist and inhomogeneous flow of electrons in a nanoconstriction or
can be included in principle by various methods, such as the around an impurity.
nonequilibrium Green’s function formalism [6–8], leading to Unfortunately, the complex behavior of the many-body
the Meir-Wingreen formula [9] for nanoscale conductors, and viscosity as a function of frequency and temperature in
more recently, the time-dependent current density functional an inhomogeneous electron gas, combined with ambiguities
theory [10,11]. Part of the correction is, in a sense, trivial—at in the characterization of experimental devices, has so far
least from the conceptual standpoint. This is the correction that frustrated attempts to reach definite conclusions concerning
arises from the many-body “dressing” of the static effective the importance of dynamical corrections in concrete situations
potential in which the electrons move. The dressing in question [22–24,29,30]. Very recently, however, an interesting paper
is essentially a generalization of the classical screening idea by Kurth and Stefanucci [31] has conclusively shown the
to include so-called exchange-correlation (xc) effects in the importance of dynamical corrections in an Anderson impurity
local environment of an electron [12,13]. If standard density model in the Coulomb blockade regime.
functional theory [14,15] is used to model these xc effects, In this Rapid Communication we consider the somewhat
great care must be exerted, in nanoscale systems, to ensure simpler problem of the residual resistivity of a normal metal
that the effective potential has the appropriate discontinuities (Al) with embedded impurities. The application of density
as a function of particle number [16,17]. This is because the functional theory to this system is not plagued by discontinu-
number of electrons in a nanoscale conductor typically fluc- ities in the xc potential and we can thus focus on dynamical
tuates between N and N + 1 during the conduction process. corrections in their pure form. We apply the method introduced
Besides the many-body dressing of the static effective potential in Refs. [32,33] to extract the dynamical effective potential of
(which exhibits the discontinuity as a function of N ), there is time-dependent DFT from the xc kernel of time-dependent
an additional effect arising from the fact that the conduction current DFT. In those papers, large many-body corrections to
process is a nonequilibrium one and therefore cannot be the stopping power of an electron gas for charged impurities
rigorously described in terms of a static effective potential: were found. The outcome of the present study is twofold: On
What is needed is a dynamical effective potential, such as the the one hand, we show that the dynamical corrections to the
one introduced in time-dependent density functional theory resistivity for this classic problem are indeed very significant

1098-0121/2014/89(24)/241108(5) 241108-1 ©2014 American Physical Society

 RAPID COMMUNICATIONS

V. U. NAZAROV, G. VIGNALE, AND Y.-C. CHANG PHYSICAL REVIEW B 89, 241108(R) (2014)

and must be taken into account if a quantitative comparison we need to know the total electric field E, which is the sum of
with experiments is desired; on the other hand, we discover that the external electric field Eext and the field that is induced by
the dominant contribution to the dynamical correction is quite the current itself. From Maxwell’s equations, and since ∇ × B
different from the one that was proposed in earlier papers on averages out to zero, one has
the basis of a perturbative analysis of time-dependent CDFT:
It does not come from the imaginary part but from the real part 4π i
E(ω) = Eext (ω) + j(ω). (7)
of the xc kernel, and from this point of view, is more consistent ω
with the treatments of Refs. [30,31]. Equation (7) substituted into Eq. (5) implicitly determines the
We start by deriving an exact formula for the frequency- resistivity tensor ρij (ω) in the general case of finite frequency.
dependent resistivity in terms of quantities that can be As we show below, in the specific case of ω → 0, this relation
calculated entirely within density functional theory. Let us becomes explicit. Notice that both E and j vanish in the ω → 0
consider a monochromatic and uniform external electric field limit. The external electric field, on the other hand, is clearly
Eext (t) = Eext e−iωt applied to electron gas with impurities finite and given by Eext = − 4πi j in the same ω → 0 limit.
positioned at Rk , k = 1,2, . . . . We can write the (particle) ω
Substituting this expression for Eext in the second term on
current density averaged over the normalization volume V as the right-hand side of Eq. (5), putting j = 4πiω
(E − Eext ) on the
[34] (we use atomic units, i.e., the electron charge and mass left-hand side, and lastly, noting that
are set to unity unless otherwise indicated)

 ∂ Im χ (r,r ,ω)

c  
ji (ω) = dr dr χ̂ij (r,r ,ω)Eext,j , (1) χ (r,r ,ω) − χ (r,r ,0)  iω
, (8)
iωV V ∂ω ω=0

where χ̂ij (r,r ,ω) is the current-current response function of we arrive at our main formal result:
the inhomogeneous electron gas with impurities. A summation 
1
over the repeated Cartesian index j is implied. (Notice that the ρij = − 2 [∇i V0 (r)][∇j V0 (r )]
electric charge current is the negative of the particle current.) n̄0 V V

We transform Eq. (1) with the help of the sum rule [35] ∂ Im χ (r,r ,ω)


 ×
dr dr . (9)
 2  ∂ω
χ̂ij (r,r ,ω) dr
ω=0
ω − ωp2
Equation (9) is the formal solution to the problem of the

ω2 resistivity in terms of the density-density response function χ
= χ̂ik (r,r ,ω) ∇k ∇j V0 (r ) dr + n0 (r) δij , (2) of the interacting inhomogeneous electron gas with impurities.
c
We now substitute the operator equality
where


V0 (r) = v0 (r − Rk ), (3) ∂χ (ω)

∂χ −1 (ω)

= −χ (0) χ (0)
k ∂ω
ω=0 ∂ω
ω=0
v0 (r) is the bare potential of one impurity centered at√origin, into Eq. (9), and apply again the static sum rule of Eq. (4).
n0 (r) is the ground-state electron density, and ωp = 4π n̄0 Recalling that [18]
is plasma frequency of the homogeneous electron gas without
impurities. Applying Eq. (2) twice with respect to the integra- 1
χ −1 (r,r ,ω) = χs−1 (r,r ,ω)−fxc (r,r ,ω) − , (10)
tion over r and r in Eq. (1), and using the expression for the |r − r |
density-response function
where χs is the “Kohn-Sham” density-density response func-
c
χ (r,r ,ω) = − 2 ∇i · χ̂ij (r,r ,ω) · ∇j , tion of the noninteracting system with ground-state density
ω n0 (r) and fxc is the dynamical exchange and correlation kernel,
together with the static sum rule [36] we conveniently rewrite Eq. (9) in terms of χs and the fxc as
 follows:
χ (r,r ,0)∇i V0 (r ) dr = ∇i n0 (r), (4)
ρij = ρ1,ij + ρ2,ij , (11)
we eventually write the current density as 
 1
ω 1 ρ1,ij =− 2 [∇i Vs (r)][∇j Vs (r )]
ji (ω) =  2  n̄0 Eext,i +   n̄0 V
i ω − ωp 2 ω − ωp2 V
2 V

 ∂ Im χs (r,r ,ω)


×
dr dr , (12)
× dr dr [∇i V0 (r)][χ (r,r ,ω) ∂ω ω=0
V
 where Vs (r) is the static Kohn-Sham (KS) potential,

− χ (r,r ,0)][∇j V0 (r )] Eext,j . (5) 
1
ρ2,ij = − 2 [∇i n0 (r)][∇j n0 (r )]
In order to calculate the frequency-dependent complex n̄0 V V
resistivity tensor ρij (ω) defined as

∂ Im fxc (r,r ,ω)

×
dr dr . (13)
Ei (ω) = −ρij (ω)jj (ω), (6) ∂ω ω=0

241108-2
 RAPID COMMUNICATIONS

DYNAMICAL MANY-BODY CORRECTIONS TO THE . . . PHYSICAL REVIEW B 89, 241108(R) (2014)

Notice that in Eq. (12) we

 have used the “Kohn-Sham version” the hydrodynamics of inhomogeneous viscous electron liquid.
of the static sum rule χs (r,r ,0)∇i Vs (r ) dr = ∇i n0 (r) to A simplified form for this kernel (neglecting the bulk viscosity
introduce the Kohn-Sham potential Vs . and keeping only the shear viscosity) is [40]
Equation (12) is the single-particle (KS) contribution to  1 1 1 1
the resistivity. Under the assumption that the electrons feel iωfˆxc,ij = ∇j η̃xc (ω)∇j + ∇j η̃xc (ω)∇i
only one impurity at a time, (i.e., the coherent scattering j
n0 n0 n0 n0
of an electron from more than one impurity is neglected),
2 1 1
we replace the full KS potential Vs (r) by the KS potential − ∇i η̃xc (ω)∇j , (16)
associated with a single impurity in the electron gas, which we d n0 n0
further assume to be spherically symmetric, i.e., Vs (r) = Vs (r).
where η̃xc (ω) = ηxc (ω) − μxciω(ω) is the complex viscoelastic
The KS response function χs (r,r ,ω) is computed accordingly.
coefficient of the homogeneous electron liquid, with ηxc being
The normalization volume is taken to be equal to the volume
the xc shear viscosity and μxc the xc shear modulus (ηxc and
per impurity, i.e., V = 1/ni . It can be proved without further
μxc are both real; d is the number of space dimensions).
assumptions [33,36] that Eq. (12), thus modified, yields exactly
We therefore use Eq. (13) with fxc given by Eq. (15)
kF ni and fˆxc given by Eq. (16) with the viscoelastic constants
ρ1 = σtr (kF ), (14)
n̄0 of the homogeneous electron liquid evaluated at the local
density n0 (r). The values of these constants are taken from
where σtr is the transport cross section, which can be calculated
Ref. [28]. Notice that fˆxc has a real part controlled by the
by means of the standard T -matrix (phase-shift) technique in
shear modulus μxc and an imaginary part controlled by the
the potential Vs (r) [12,13].
shear viscosity ηxc . The perturbative treatment suggested in
Equation (13) is the dynamical exchange-correlation con-
Ref. [29] led to the conclusion that the dynamical correction
tribution. If the frequency dependence of fxc is neglected—
to the resistivity would be controlled exclusively by the shear
as one does, for example, in the adiabatic approximation
viscosity. However, we see that this is not the case in the
to TDDFT—then fxc is purely real and Eq. (13) yields
present treatment. Because the Kohn-Sham response function
ρ2 = 0. This result, combined with the discussion of the
χs has both a real and an imaginary part, the imaginary part
previous paragraph, leads to the important conclusion that
of fxc , as given by Eq. (15), is now controlled by both μxc
the adiabatic approximation to TDDFT is equivalent to the
and ηxc .
classical potential-scattering (T -matrix) approach as far as the
In Fig. 1, we present results for resistivity for substitutional
calculation of the resistivity is concerned.
impurities of atomic number Z from 11 through 32 in an
To find the many-body contribution to the resistivity from
aluminum host. The latter is modeled as a jellium with Wigner-
Eq. (13), we need a good approximation to the dynamical
Seitz radius rs = 2.07. The single-particle contribution ρ1
exchange and correlation kernel fxc . It is known [38] that
calculated from Eq. (14) and represented by the dashed
fxc (r,r ) is strongly nonlocal (i.e., a long-ranged function of
chained curve (black) is found to be in agreement with earlier
|r − r |) and this nonlocality is crucial to a proper description
calculations [12,13]. The total resistivity, including dynamical
of many-body effects in transport phenomena, even on a
exchange and correlation contributions, is represented by the
qualitative level [32]. This immediately poses the problem of
construction of a reasonably accurate nonlocal approximation
for fxc . In a recent paper [32] we have shown how this can Na Mg Al Si P S Cl Ar K Ca Sc Ti V Cr Mn Fe Co Ni Cu Zn Ga Ge
be done starting from an exact representation of the scalar 9
ρ = ρ1 + ρ2
fxc kernel in terms of the tensor exchange-correlation kernel 8 ρ1
ρ/ni (μΩ cm/at. %)

fˆxc of time-dependent current density functional theory. This 7

representation reads
6
ω2      −1 Aluminum host
fxc = − ∇ −2 ∇ · fˆxc + χ̂s−1 − fˆxc T̂ χ̂s−1 − fˆxc T̂ 5 rs = 2.07
c
 −1   −1  4
× χ̂s − fˆxc − χ̂s−1 T̂ χ̂s−1 T̂ χ̂s−1 · ∇∇ −2 , (15)
3
where χ̂s is the KS current-current response function, T̂ is 2
the projector operator onto the subspace of transverse vector
fields (i.e., divergence-free fields), and ∇ −2 is the inverse 1
f (r )
of the Laplace operator: ∇ −2 f (r) = − dr 4π|r−r  | in three
0
10 12 14 16 18 20 22 24 26 28 30 32
dimensions [39].
Z
By making use of the local density approximation (LDA)
for the tensor fˆxc [20] in the right-hand side of Eq. (15), FIG. 1. (Color online) Residual resistivity of aluminum due to
we obtain a nonlocal approximation for the scalar fxc , which substitutional impurities of atomic number Z, as a function of Z. The
satisfies the static zero-force sum-rule requirements [32,38], solid chained curve (red online) is our result with inclusion of the
and can, therefore, be considered a promisingly accurate dynamical exchange and correlations [the sum of ρ1 and ρ2 obtained
approximation for transport problems. At the same time, in with use of Eqs. (14) and (13), respectively]. The dashed chained
Ref. [40], the LDA to the exchange and correlation kernel fˆxc curve (black) is the result of the single-particle theory (ρ1 only).
of the TDCDFT has been worked out within the framework of Solid symbols are experimental data compiled from Ref. [37].

241108-3
 RAPID COMMUNICATIONS

V. U. NAZAROV, G. VIGNALE, AND Y.-C. CHANG PHYSICAL REVIEW B 89, 241108(R) (2014)

solid chained curve (red). An improved agreement between collisionless shear modulus multiplied by the collision time. It
theory and experiment can be clearly seen from the figure. The now appears that the xc kernel constructed via Eq. (15) does
dynamical corrections are sizable and indeed of the right order take into account the effect of electron-impurity collisions. It
of magnitude to account for the observed discrepancy between does so via the imaginary part of the noninteracting response
available theoretical calculations and experimental data. function χs which, indeed, is calculated in the presence of the
Further analysis shows that the dominant contribution to impurity. It is the imaginary part of χs that, combined with the
the dynamical corrections does not come from the viscosity, real part of fˆxc (i.e., μxc ), generates the main contribution to
ηxc , but from the shear modulus, μxc , contribution to the xc the dynamical resistivity. We interpret this fact as an indication
kernel of Eq. (16). In fact, the contribution from ηxc is one that Eq. (15) is dressing the homogeneous kernel (16) by
to two orders of magnitude smaller than that from μxc . This including the effect of electron-impurity collisions.
is apparently in contrast with the discussion of Ref. [29], in In conclusion, we have developed the nonadiabatic time-
which the dynamical correction was expressed solely in terms dependent density functional formalism for a systematic
of ηxc . calculation of the dc residual resistivity of metals with
Physically, however, the emergence of μxc as the main impurities. The contribution to the resistivity arising from
contributor to the resistivity makes physical sense. The the many-body interactions has been expressed through the
homogeneous kernel of Eq. (16) is “collisionless,” i.e., does dynamical exchange and correlation kernel fxc . We have
not include the effect of electron-impurity collisions on shown that all the dynamical effects of the electron-electron
the viscosity and the shear modulus. Mathematically, these interaction are contained in the frequency dependence of fxc .
quantities were obtained from a “high-frequency” calculation The adiabatic approximation, which neglects this frequency
[28] in which the frequency remains higher than the electron- dependence, is exactly equivalent to the conventional single-
impurity collision rate. Therefore, we expect that the true particle potential-scattering theory of the resistivity, provided
zero-frequency limit of the viscosity, which should ultimately the coherent scattering from multiple impurities is neglected
control the excess dissipation and hence the dynamical as well. Our representation of the single-particle and the many-
correction to the resistivity, should be quite different from body contributions to the resistivity through the Kohn-Sham
the collisionless ηxc used in Eq. (16). In fact, according response function and the dynamical xc kernel, respectively, is
to a dispersion relation argument put forth in Ref. [30], a fit to use in band-structure calculations [41], and can provide
reasonable expression for the zero-frequency limit of the guidance in calculating the resistance of nanoscale conductors.
viscosity (i.e., the true collisional viscosity) is ηxc (0) = μxc τ ,
where μxc is the high-frequency shear modulus, and 1/τ is the V.U.N. acknowledges support from National Science Coun-
electron-impurity scattering rate. The same conclusion was cil, Taiwan, Grant No. 100-2112-M-001-025-MY3. G.V.
reached in Ref. [28] [see, in particular, Eq. (5.28)] where the acknowledges support from DOE Grant No. DEFG02-
collisional (i.e., ω = 0) viscosity was found to be equal to the 05ER46203.

[1] N. F. Mott and H. Jones, The Theory of the Properties of Metals [13] M. J. Puska and R. M. Nieminen, Phys. Rev. B 27, 6121 (1983).
and Alloys (Dover, New-York, 1958). [14] P. Hohenberg and W. Kohn, Phys. Rev. 136, B864 (1964).
[2] R. Landauer, IBM J. Res. Dev. 1, 223 (1957). [15] W. Kohn and L. J. Sham, Phys. Rev. 140, A1133 (1965).
[3] R. Landauer, J. Phys.: Condens. Matter 1, 8099 (1989). [16] J. P. Perdew, R. G. Parr, M. Levy, and J. L. Balduz, Phys. Rev.
[4] M. Büttiker, Y. Imry, R. Landauer, and S. Pinhas, Phys. Rev. B Lett. 49, 1691 (1982).
31, 6207 (1985). [17] C. Toher, A. Filippetti, S. Sanvito, and K. Burke, Phys. Rev.
[5] M. Büttiker, Phys. Rev. Lett. 57, 1761 (1986). Lett. 95, 146402 (2005).
[6] S. Datta, Electronic Transport in Mesoscopic Systems [18] E. K. U. Gross and W. Kohn, Phys. Rev. Lett. 55, 2850 (1985).
(Cambridge University Press, New York, 1997). [19] S. K. Ghosh and A. K. Dhara, Phys. Rev. A 38, 1149 (1988).
[7] M. D. Ventra, Electrical Transport in Nanoscale Systems [20] G. Vignale and W. Kohn, Phys. Rev. Lett. 77, 2037 (1996).
(Cambridge University Press, Cambridge, 2008). [21] S. Stefanucci and C.-O. Almbladh, Europhys. Lett. 67, 14
[8] G. Stefanucci and R. van Leeuwen, Nonequilibrium Many- (2004).
Body Theory of Quantum Systems: A Modern Introduction [22] N. Sai, M. Zwolak, G. Vignale, and M. Di Ventra, Phys. Rev.
(Cambridge University Press, Cambridge, 2013). Lett. 94, 186810 (2005).
[9] Y. Meir and N. S. Wingreen, Phys. Rev. Lett. 68, 2512 (1992). [23] J. Jung, P. Bokes, and R. W. Godby, Phys. Rev. Lett. 98, 259701
[10] G. Vignale, in Fundamentals of Time-Dependent Density Func- (2007).
tional Theory, edited by M. A. L. Marques, N. T. Maitra, F. M. [24] N. Sai, M. Zwolak, G. Vignale, and M. Di Ventra, Phys. Rev.
S. Nogueira, E. K. U. Gross, and A. Rubio (Springer, Berlin, Lett. 98, 259702 (2007).
2012), p. 457. [25] M. Koentopp, K. Burke, and F. Evers, Phys. Rev. B 73,
[11] C. A. Ullrich, Time-Dependent Density-Functional Theory: 121403(R) (2006).
Concepts and Applications (Oxford University Press, Oxford, [26] P. Bokes, J. Jung, and R. W. Godby, Phys. Rev. B 76, 125433
2012). (2007).
[12] R. M. Nieminen and M. Puska, J. Phys. F: Met. Phys. 10, L123 [27] S. Conti, R. Nifosı̀, and M. Tosi, J. Phys.: Condens. Matter 9,
(1980). L475 (1997).

241108-4
 RAPID COMMUNICATIONS

DYNAMICAL MANY-BODY CORRECTIONS TO THE . . . PHYSICAL REVIEW B 89, 241108(R) (2014)

[28] S. Conti and G. Vignale, Phys. Rev. B 60, 7966 (1999). [36] V. U. Nazarov, J. M. Pitarke, C. S. Kim, and Y. Takada, Phys.
[29] G. Vignale and M. Di Ventra, Phys. Rev. B 79, 014201 (2009). Rev. B 71, 121106(R) (2005).
[30] D. Roy, G. Vignale, and M. Di Ventra, Phys. Rev. B 83, 075428 [37] E. Babic, Z. Vucic, I. Zoric, M. Ocko, B. Leontic, E. Girt, and
(2011). R. Krsnik, Solid State Commun. 10, 691 (1972); G. Boato, M.
[31] S. Kurth and G. Stefanucci, Phys. Rev. Lett. 111, 030601 (2013). Bugo, and C. Rizzuto, Nuovo Cimento 45, 226 (1966); Y. Fukai,
[32] V. U. Nazarov, J. M. Pitarke, Y. Takada, G. Vignale, and Y.-C. Phys. Rev. 186, 697 (1969).
Chang, Phys. Rev. B 76, 205103 (2007). [38] G. Vignale, Phys. Lett. A 209, 206 (1995).
[33] V. U. Nazarov, J. M. Pitarke, Y. Takada, G. Vignale, and Y.-C. [39] In the case of the one-dimensional inhomogeneity, Eq. (15)
Chang, Int. J. Mod. Phys. B 22, 3813 (2008). simplifies to an expression which does not involve χ̂s [42,43].
[34] In Eq. (1), the integration over r gives the microscopic current This, however, is not true in the general case.
density by virtue of the Kubo formula, while the integration over [40] G. Vignale, C. A. Ullrich, and S. Conti, Phys. Rev. Lett. 79,
r with the division by V represents the averaging in the system 4878 (1997).
uniform on the macroscopic scale. All many-body and possible [41] N. Papanikolaou, R. Zeller, P. H. Dederichs, and N. Stefanou,
quantum interference effects are contained in the current-density Comput. Mater. Sci. 8, 131 (1997).
response function χ̂ . [42] G. Vignale and W. Kohn, in Electronic Density Functional
[35] Equation (2) is a generalization for infinite systems of the Theory: Recent Progress and New Directions, edited by J.
corresponding sum rule of Ref. [42] accurately taking into Dobson, M. P. Das, and G. Vignale (Plenum, New York,
account the potential of the uniform positive background, which 1998).
as well as the bare potential of the impurities V0 (r), is external [43] M. Dion and K. Burke, Phys. Rev. A 72, 020502(R)
to the electron subsystem. (2005).

241108-5