PHYSICAL REVIEW B 88, 155427 (2013)

Improved modeling of electrified interfaces using the effective screening medium method

Ikutaro Hamada,1,* Osamu Sugino,2 Nicéphore Bonnet,2,3 and Minoru Otani3

1
International Center for Materials Nanoarchitectonics (WPI-MANA), National Institute for Materials Science (NIMS),
Tsukuba 305-0044, Japan
2
Institute for Solid State Physics, University of Tokyo, Kashiwa 277-8581, Japan
3
Nanosystem Research Institute, National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba 305-8568, Japan
(Received 8 June 2013; published 21 October 2013)
An update of the effective screening medium (ESM) method [Otani and Sugino, Phys. Rev. B 73, 115407
(2006)] is presented, extending the ability to simulate electrified interfaces in an efficient and flexible way. The
need for an artificial vacuum buffer between the molecular system and the screening medium is removed by
defining a smooth transition of the ESM dielectric permittivity (smooth ESM), precluding numerical instabilities
when molecules come in contact with the ESM. Moreover, at short distances, the smooth ESM acts as a repulsive
wall, and thus the simulation cell can serve as a natural container for molecules in molecular-dynamics simulations.
Consequently, the smooth ESM method is a substantial advancement in modeling solid-liquid interfaces under
electric bias.

DOI: 10.1103/PhysRevB.88.155427 PACS number(s): 73.20.−r, 68.08.−p, 31.15.E−

I. INTRODUCTION located between the vacuum region and the ESM [original
ESM, Fig. 1(a)]. The discontinuous change causes numerical
There has been a growing interest in simulating electrified
instability when the electron density touches the boundary.
interfaces, e.g., the electrochemical double layer, because of
This is particularly problematic when simulating interfaces
their relevance to energy conversion and storage devices, such
between an electrode and a solution, because the molecules
as fuel cells, solar cells, and batteries. The thickness of the
of the solution would naturally come in close contact with
double layer of a water solution is simply estimated to be
the ESM. To avoid this problem, a vacuum region has been
1 nm (pH = 1) to 1 μm (pH = 7). This fact hinders first
inserted as a buffer between the molecules and the ESM, and an
principles treatment of the double layer, and prompts modeling
artificial potential wall has been used to prevent the molecules
of distant electrolyte ions. In earlier studies, the effect of the
and electrons from entering the buffer (see, e.g., Refs. 21 and
electrolyte ions has been incorporated as an external electric
25). It is important to remove this restriction to better model
field arising from a dipole layer1 or a charged layer,2–7 which
electrified interfaces.
is located in a vacuum region of electrode/solution/vacuum
Here, we improve upon the description of the screening
interface models. Alternatively, an electric field is applied by
medium by imposing a smooth transition of the dielectric
introducing compensating background charge.8–10 More recent
permittivity between the vacuum region and the ESM [smooth
studies11–16 have used electrode/solution/continuum interface
ESM, Fig. 1(b)], and thus lift the nonoverlap constraint
models, where the continuum is characterized by dielectric
on the electronic charge density. In line with the approach
constant and/or classical charges. Methods to calculate elec-
adopted in the polarizable continuum model of Fattebert and
trode potential, one of essential quantities in electrochemistry,
Gygi,30 the present method models the permittivity by a simple
have also been developed.10–12,15–19 Among other approaches,
analytical function, varying smoothly from  = 1 to  > 1 at
the effective screening medium (ESM) method11 has been
the boundary of the ESM region. The corresponding Green’s
developed as a way to simulate electrified interfaces within a
function, electrostatic potential, and energy are then derived.
first principles framework using periodic boundary conditions
After implementing this approach into our DFT code and
(PBC). Given a slab geometry standing for the interface,
conducting various tests, it turns out that, in case of an overlap
the ESM method applies a correction that lifts the periodic-
with the electronic charge of molecules, the smooth ESM
boundary condition in the surface normal direction, thereby
potential is numerically stable and acts naturally as a repulsive
allowing computation of the electrostatic potential of the
wall, thus removing the needs for an artificial vacuum buffer
isolated slab. Additionally, the method allows filling the region
and for an additional potential wall simultaneously.
away from the slab with a dielectric screening medium—
the ESM per se—as a simple way to include electrostatic
screening effects of the environment. Because the electrostatic II. METHOD
potential is obtained analytically using the Green’s function In the ESM method, the total energy functional has the
technique, the computational cost is comparable with that of following expression,
a conventional PBC calculation, opening the way to large- 
scale molecular-dynamics simulations of electrified interfaces Etot [ρe ] = Ts [ρe ] + Exc [ρe ] + d rd r  ρe (r)G(r,r  )ρI (r  )
within density functional theory (DFT).20–25 The method has 
also been applied successfully to surfaces26 and molecules27 1
+ d rd r  ρe (r)G(r,r  )ρe (r  )
isolated in vacuum, in place of the dipole correction.28,29 2
In the original version of the ESM method, the screening 
1
medium has a sharp boundary, that is, the relative permittivity + d rd r  ρI (r)G(r,r  )ρI (r  ), (1)
2
changes discontinuously from  = 1 to  > 1 at the boundary

1098-0121/2013/88(15)/155427(4) 155427-1 ©2013 American Physical Society

HAMADA, SUGINO, BONNET, AND OTANI PHYSICAL REVIEW B 88, 155427 (2013)

(a) Original ESM (b) Smooth ESM where g  , with norm g , is a generic vector of the reciprocal
lattice, parallel to the surface, and z is the real-space coordinate
in the direction perpendicular to the surface. Moreover, the
ε=1 ε=∞ ε=1 ε ∞ relative permittivity (z) is assumed to be a function of z
only.
z1 z1 Considering the case where the screening medium acts
as a metal, that is, a medium with an infinite permittivity,
FIG. 1. Schematic views of (a) the original ESM and (b) the we choose the following expression for the permittivity
smooth ESM. Here, the ESM acts as a metal, hence the infinite function,
permittivity. 
1 z < z1 ,
where ρe and ρI are the electronic and ionic charge densities, (z) = a(z−z1 )
(4)
e z > z1 ,
respectively; Ts is the kinetic energy; Exc is the exchange-
correlation energy. The third, fourth, and fifth terms are
the electron-ion interaction, Hartree, and ion-ion interaction where z = z1 is the boundary of the ESM region, and a
energies, respectively, and they are expressed in terms of the is a parameter controlling the smoothness of the transi-
Green’s function G(r,r  ) satisfying Poisson’s equation, tion. As required, the permittivity goes to infinity inside
the metallic screening region. Alternative functional forms
∇ · [(r)∇]G(r,r  ) = −4π δ(r − r  ). (2) could be introduced for cases where the ESM has a finite
Adopting Laue’s representation, the previous equation permittivity. Substituting the functional form given by Eq. (4)
becomes into Eq. (3), and solving for the Green’s function, we
obtain
{∂z [(z)∂z ](z) − g2 }G(g  ,z,z ) = −4π δ(z − z ), (3)

⎧ 2π −g |z−z | 2πκ g (z+z −2z1 )

⎪ e − ag e (z,z < z1 )
⎪
⎪ g
⎪
⎪
⎨ e  1 12 ξ (z1 −z )
8π g (z−z )+
(z < z1 < z )
 α
G(g  ,z,z ) =
⎪
⎪ 8π g (z −z1 )+ 12 ξ (z1 −z)
e (z < z1 < z)
⎪
⎪ α
⎪
⎩ 4πβ 1 ξ (2z1 −z−z ) 4π 12 a(2z1 −z−z )− 12 λ|z−z |
− αλ e 2 + λ
e (z,z > z1 )

√
where
√ α = a + 2g √ + a + 4g ;
2 2
β =√a + 2g − A. CO molecule
2
a √ + 4g ; ξ = a + a + 4g ; κ = −2g + a 2 + 4g2 ;
2 2 2 In the first set of calculations, a carbon monoxide (CO)
λ = a 2 + 4g2 . In the limit where a → ∞, it can be shown molecule is placed in a 0.5 × 0.5 × 2.4 nm3 unit cell, with its
that the Green’s function reduces to bond in the longitudinal direction (z). The ESM is placed at a
distance z1 = 1.0 nm from the center of the unit cell (z = 0).
2π −g |z−z | 2π g (z+z −2z1 )
G(z,z ) = e − e , (5) The total energy, computed at the point of the Brillouin
g g zone with the CO bond length held fixed, was calculated as a
which is consistent with the expression obtained for the function of the z coordinate of the O atom (zO ). It is plotted
original ESM scheme, given by Eq. 15 of Ref. 11 for boundary in Fig. 2, for different values of the smoothness parameter a,
condition (iii) [“vacuum/metal”]. It is moreover apparent, from going from 11.34 to 18.90 nm−1 . Initially, as the CO molecule
the obtained Green’s function, that the electron-ion and Hartree approaches the ESM, the total energy decreases owing to the
potentials will be smooth at z = z1 . attractive interaction with the image charge induced in the
ESM. However, at the point where the tail of the ionic potential
III. RESULTS AND DISCUSSIONS
starts overlapping with the ESM, the attractive ion-electron
interaction gets partially screened and the electronic energy
The smooth ESM method has been implemented into increases, such that the ESM acts as a repulsive wall for the
the STATE31 code (see Supplemental Material32 for details), molecule. Moreover, the strength of the repulsion is controlled
and applied to various test cases reported here. The calcula- by the smoothness parameter, that is, the larger the value of
tions have been performed with the Perdew-Burke-Ernzerhof a, the stronger the screening of the potential, and thus, the
functional33 (PBE) for the electronic exchange-correlation larger the repulsive force. When a is smaller than a certain
energy, ultrasoft pseudopotentials34 for the electron-ion inter- value, the attractive and repulsive forces balance, and give rise
action, and cutoff energies of 25 and 225 Ry for the plane-wave to an energy minimum. Thus, in general, the potential acting
basis sets of the wave functions and augmentation charge, on a molecule in the vicinity of the ESM can be tuned by the
respectively. smoothness parameter a.

155427-2
IMPROVED MODELING OF ELECTRIFIED INTERFACES . . . PHYSICAL REVIEW B 88, 155427 (2013)

z1 C. Metallic slab
8.0 Last, the smooth ESM method is used to study the
-1
7.0
a=11.34 nm charge and potential of an aluminum electrode under bias.
-1
a=13.23 nm
-1
The electrode is modeled by a six-layer Al slab in the
6.0 a=15.12 nm (111) direction, with 1 × 1 unit cell in the xy plane, and a
Relative energy (eV)

-1
5.0 a=17.01 nm PBE-optimized lattice constant of 0.41 nm. The thickness
-1
a=18.90 nm
4.0 of the vacuum region around the slab is equivalent to 13
ESM layers (3.27 nm), and the Brillouin zone is sampled using
3.0
a -centered 8 × 8 × 1 k-point mesh. The ESM is placed at
2.0 C O a distance of 0.82 nm from the surface of the slab, and the
1.0 value of the smoothness parameter is a = 9.45 nm−1 . The
system is put under bias by adding up to ±0.02 electrons to
0.0
the slab, and the excess/deficit charge density and potential
-1.0 are plotted (Fig. 4). Owing to the metallic screening, the
0.2 0.4 0.6 0.8 1.0 1.2 excess charge density accumulates at the surface of the slab,
zO (nm)
consistent with the behavior already observed in Ref. 11
using boundary condition (iii) [“vacuum/metal”]. However,
FIG. 2. (Color online) Total energy as a function of the longitu- a qualitatively different feature appears in the fact that inside
dinal position of the CO molecule zO and for different values of the
the ESM region, the electrostatic potential approaches zero
smoothness parameter a.
gradually as a result of the smooth permittivity transition, in
contrast to the original ESM version, in which the potential
is abruptly changed to 0 at z = z1 .11 Moreover, the charge
B. H2 O molecule induced in ESM can be calculated as

Next, we consider the case of a water molecule in a 1.0 × (r) − 1
1.0 × 2.8 nm3 unit cell. The position of the ESM is unchanged ρind (r) = −∇ · P(r) = ∇ · ∇V (r) , (6)
4π
(z1 = 1.0 nm), and the value of the smoothness parameter
is fixed at a = 5.67 nm−1 . Three configurations of the water where P(r) and V (r) are polarization and electrostatic poten-
molecule are computed: (i) “H-down”, in which the H atoms tial, respectively. In Fig. 4(d), it can be seen that the image
point away from the ESM; (ii) “H-para”, in which the molecule charge is induced in ESM, which screens the electrostatic
lies parallel to the surface; (iii) “H-up”, in which the H atoms potential. This demonstrates the role of the smooth ESM.
point toward the ESM. The total energy computed as a function
of the longitudinal position of the water molecule (Fig. 3)
turns out to be only slightly affected by the specific molecular
(a) ESM
orientation, and exhibits a minimum around 0.2–0.3 nm, before
the onset of the repulsive force. 2.0
z1
+0.02e
+0.01e
(b)
Δρe (e nm )

1.0
-1

-0.01e
-0.02e
0.0

-1.0
z1
-2.0
4.0
(c)
8.0 0.4
2.0
ΔV (eV)
Relative energy (eV)

H-down 0.2 0.0

H-para 0.0 -2.0
6.0
Relative energy (eV)

-0.2 -4.0
H-up
-0.4 2.0
Δρind (e nm )

(d)
-1

0.0 0.1 0.2 0.3 0.4 0.5

4.0 zO (nm) 0.0

-2.0
ESM
2.0 -2.0 -1.0 0.0 1.0 2.0
z (nm)

0.0 FIG. 4. (Color online) Planar average of electronic charge density

difference ρe , electrostatic potential difference V , and induced
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 charge in ESM ρind , upon charging the slab. (a) Longitudinal view
zO (nm) of the setup with the Al(111) slab and the ESM. (b) Electronic
charge density difference compared to that of the neutral slab. A
FIG. 3. (Color online) Total energy as a function of the longi- positive (negative) sign indicates an excess (deficit) of electrons.
tudinal position of the water molecule and for different molecular (c) Electrostatic potential difference compared to that of the neutral
orientations. The inset is an enlargement around the energy minimum. slab. (d) Image charge induced in ESM. The vertical dotted line stands
a = 5.67 nm−1 is used. for the surface of the ESM.

155427-3
HAMADA, SUGINO, BONNET, AND OTANI PHYSICAL REVIEW B 88, 155427 (2013)

IV. SUMMARY electrode potential,35 we believe that the simulation of solid-

The ESM method has been improved by letting the liquid interfaces under electric bias can be greatly advanced.
dielectric permittivity vary smoothly between the vacuum
region and the ESM, thereby lifting the need for an artificial
ACKNOWLEDGMENTS
vacuum buffer between the molecular system and the ESM.
The Green’s function has been derived for the new functional This work was partly supported by a Grant-in-Aid for Scien-
form of the permittivity, and implemented into our plane-wave tific Research on Innovative Area “Materials Design through
DFT code. Various tests conducted on small molecules have Computics: Complex Correlation and Non-equilibrium Dy-
shown that the smooth ESM acts as a repulsive wall at short namics” from the Ministry of Education, Culture, Sports,
distances, and that the strength of the repulsion can be tuned by Science and Technology (MEXT), Japan (No. 23104501), the
the smoothness parameter of the permittivity. Thus, the smooth Strategic Program for Inovative Research (SPIRE), MEXT,
ESM is not only a numerically robust screening medium, but it Japan, the Computational Materials Science Initiative (CMSI),
also acts as a natural container wall for molecules in molecular- Japan, and the World Premier International Research Center
dynamics simulations, thereby extending the ability to model Initiative (WPI), MEXT, Japan. Numerical calculations were
electrified interfaces in an efficient and flexible manner. By performed at the Supercomputer Center, Institute for Solid
combining the presently upgraded ESM method with a scheme State Physics, University of Tokyo, and at the Information
recently developed to perform calculations at a constant Technology Center, University of Tokyo.

* 20
[email protected] O. Sugino, I. Hamada, M. Otani, Y. Okamoto, and T. Ikeshoji, Surf.
1
J. Rossmeisl, J. K. Nørskov, C. D. Taylor, M. J. Janik, and Sci. 601, 5237 (2007).
21
M. Neurock, J. Phys. Chem. B 110, 21833 (2006). M. Otani, H. Hamada, O. Sugino, Y. Morikawa, Y. Morikawa,
2
C. L. Fu and K. M. Ho, Phys. Rev. Lett. 63, 1617 (1989). Y. Okamoto, and T. Ikeshoji, J. Phys. Soc. Jpn 77, 024802
3
K. P. Bohnen and D. M. Kolb, Surf. Sci. 407, L629 (1998). (2008).
4 22
A. Y. Lozovoi and A. Alavi, Phys. Rev. B 68, 245416 (2003). M. Otani, H. Hamada, O. Sugino, Y. Morikawa, Y. Morikawa,
5
J. G. Che and C. T. Chan, Phys. Rev. B 67, 125411 (2003). Y. Okamoto, and T. Ikeshoji, Phys. Chem. Chem. Phys. 10, 3609
6
S. Kajita, T. Nakayama, and M. Kawai, J. Phys. Soc. Jpn. 76, 044701 (2008).
23
(2007). I. Hamada and Y. Morikawa, J. Phys. Chem. C 112, 10889 (2008).
7 24
S. Schnur and A. Groß, Catal. Today 165, 129 (2011). T. Ikeshoji, M. Otani, I. Hamada, and Y. Okamoto, Phys. Chem.
8
A. Y. Lozovoi, A. Alavi, J. Kohanoff, and R. M. Lynden-Bell, Chem. Phys. 13, 20223 (2011).
25
J. Chem. Phys. 115, 1661 (2001). T. Ohwaki, M. Otani, T. Ikeshoji, and T. Ozaki, J. Chem. Phys. 136,
9
J.-S. Filhol and M. Neurock, Angew. Chem. Int. Ed. 45, 402 (2006). 134101 (2012).
10 26
C. D. Taylor, S. A. Wasileski, J.-S. Filhol, and M. Neurock, Phys. I. Hamada, M. Otani, O. Sugino, and Y. Morikawa, Phys. Rev. B
Rev. B 73, 165402 (2006). 80, 165411 (2009).
11 27
M. Otani and O. Sugino, Phys. Rev. B 73, 115407 (2006). Y. Takimoto, M. Otani, and O. Sugino, Phys. Rev B 81, 153405
12
R. Jinnouchi and A. B. Anderson, Phys. Rev. B 77, 245417 (2008). (2010).
13 28
I. Dabo, B. Kozinsky, N. E. Singh-Miller, and N. Marzari, Phys. J. Neugebauer and M. Scheffler, Phys. Rev. B 46, 16067 (1992).
29
Rev. B 77, 115139 (2008). L. Bengtsson, Phys. Rev. B 59, 12301 (1999).
14 30
I. Dabo, N. Bonnet, Y. L. Li, and N. Marzari, Fuel Cell Science: J.-L. Fattebert and F. Gygi, J. Comput. Chem. 23, 662 (2002).
31
Theory, Fundamentals, and Biocatalysis, edited by A. Wieckowski, Y. Morikawa, H. Ishii, and K. Seki, Phys. Rev. B 69, 041403(R)
and J. Nørskov (Wiley, New York, 2010), Chap. 13. (2004).
15 32
K. Letchworth-Weaver and T. A. Arias, Phys. Rev. B 86, 075140 See Supplemental Material at http://link.aps.org/supplemental/
(2012). 10.1103/PhysRevB.88.155427 for the implementation of the
16
Y. H. Fang, G. F. Wei, and Z. P. Liu, Catal. Today 202, 98 (2013). smooth ESM.
17 33
J. Rossmeisl, E. Skúlason, M. E. Björketun, V. Tripkovic, and J. K. J. P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett. 77, 3865
Nørskov, Chem. Phys. Lett. 466, 68 (2008). (1996).
18 34
V. Tripkovic, M. E. Björketun, E. Skúlason, and J. Rossmeisl, Phys. D. Vanderbilt, Phys. Rev. B 41, 7892 (1990).
35
Rev. B 84, 115452 (2011). N. Bonnet, T. Morishita, O. Sugino, and M. Otani, Phys. Rev. Lett.
19
J. Cheng and M. Sprik, Phys. Chem. Chem. Phys. 14, 11245 (2012). 109, 266101 (2012).

155427-4