Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

E3438 Journal of The Electrochemical Society, 164 (11) E3438-E3447 (2017)

JES FOCUS ISSUE ON MATHEMATICAL MODELING OF ELECTROCHEMICAL SYSTEMS AT MULTIPLE SCALES IN HONOR OF JOHN NEWMAN
Improving Continuum Models to Define Practical Limits for
Molecular Models of Electrified Interfaces
Artem Baskinz and David Prendergast∗
The Joint Center for Energy Storage Research, The Molecular Foundry, Lawrence Berkeley National Laboratory,
Berkeley, California 94720, USA

We develop a continuum theory of electrolyte solutions in contact with a metal electrode, based on a generalized free energy
functional, and use it to explore the structure of the electric double layer at different electrochemical conditions. The model captures
the effects of specific adsorption of ions and solvent polarization, and can be applied on the same footing to cases with non-zero
faradaic current, beyond the classical double-layer regime. These advances permit the prediction of peculiar character in the ion
profiles at electrode potentials near the redox level, exploration of the electrochemical stability of the interface, and differentiation
between the mechanisms of electron and ion transport and associated time scales. The developed methodology enables us to self-
consistently determine the fundamental limits for a microscopic description of biased interfaces, in terms characteristic sizes and
time scales of relevant processes, within atomistic and ab initio molecular dynamics simulations.
© The Author(s) 2017. Published by ECS. This is an open access article distributed under the terms of the Creative Commons
Attribution 4.0 License (CC BY, http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse of the work in any
medium, provided the original work is properly cited. [DOI: 10.1149/2.0461711jes] All rights reserved.

Manuscript submitted April 4, 2017; revised manuscript received June 6, 2017. Published June 20, 2017. This paper is part of the
JES Focus Issue on Mathematical Modeling of Electrochemical Systems at Multiple Scales in Honor of John Newman.

As the efficiency of electrochemical conversion is dictated by the that, for an unknown system, the simulated electrochemical activity
energetics of elementary charge transfer reactions occurring at the may be an artifact of the methodology, which relies on an initial guess
electrode/electrolyte interface, the formulation of methods of control for the structure of the EDL that might prove inconsistent with the
over these reactions has become the pivotal goal of fundamental ma- aforementioned external conditions or some basic assumptions.
terial science for advanced energy storage systems. The ultimate goal The second problem is the experimental relevance of the compu-
of theoretical electrochemistry then is to explore the energetics of the tational electrode potential that is a long-standing generic issue for
relevant processes from an atomistic perspective by means of ab initio theoretical electrochemistry intimately related to the boundary con-
molecular dynamics (AIMD) where ionic and electronic components ditions of the half-cell setup. In conventional electrochemistry,5,6 the
of the system are described explicitly and are constrained by various absolute single electrode potential, E, is defined by two contributions
realistic conditions, such as the external bias, the temperature and the - the Galvani potential drop, ϕ, across the interface, which reflects
electrolyte concentration. However, despite more than a decade of the response of the electrolyte, and the chemical potential of elec-
efforts this goal is still far from being fulfilled. trons in the electrode μe . The relation between these two is not trivial
There are several interconnected fundamental problems to resolve and strongly depends on the regime at which the electrode operates.
before an atomistic approach could be routinely used to describe elec- Therefore, the computational electrode potential that is comparable to
trochemistry, if only at the level of a half-cell (i.e., a single electrode- experimental values is not known a priori, since at each potential the
electrolyte interface). First, there are at least three distinct regimes interface is characterized by a (unique) thermodynamic (steady) state
of performance of the electrochemical interface and each of these is which is unknown a priori and which must instead be determined
characterized by different sets of the time and space scales: a) the a posteriori. This, in turn, implies that in contrast to the currently
classical electric double layer (EDL) regime when no charge trans- adapted potentiostat schemes7–10 the electrode potential cannot be a
fer is assumed, b) the regime of an electrode at equilibrium with the free external parameter and should rather be the output of a self-
electrolyte (open circuit conditions, (OCP)) when only an exchange consistent procedure.
current in a form of mutually compensating fluxes of charged particles The necessity to account for the response of the electrolyte in
is possible, and c) the “true” electrochemical regime with a non-zero evaluating the electrode potential itself has a profound effect on the
net faradaic current. As long as the goal of atomistic simulations is atomistic modeling of biased interfaces and beyond. A distinct spatial
to reach a state when one would observe the actual electrochemical inhomogeneity of the EDL11 at different regimes challenges the very
reactions driven by the bias, the computational electrode potential idea of defining the redox potentials through overall estimates of the
scanning protocol should carry the system through a set of (quasi- free energy for bulk regions of the cell. Likewise, one needs extra mea-
)equilibrium points at which the atomistic structure is consistent with sures of caution when computationally assessing the electrochemical
the targeted external conditions. The key requirement for such simula- stability of an electrolyte. Currently, the AIMD approach suffers from
tions is thus thermodynamic stability of the interface that encompasses unrealistic configurations of the interfaces due to the limited number
all degrees of freedom. Currently, there is no such uniform atomistic of solvent molecules explicitly included in calculations which leads
methodology for the electrode potentials in a range of ±1 V around the to ambiguity in the definition of the electrode potential,12 insufficient
OCP since it would have to cover the processes whose characteristic accounting for the dielectric response of the liquid environment,13–16
times range from attoseconds (electron transfer (ET) time) to hun- erroneous assessments of the potential of zero charge (PZC),15,17,18–22
dreds of picoseconds (solvent dielectric relaxation) at length scales the spatial profile of the average electrostatic potential in the EDL
ranging from angstroms to tens of nanometers. Instead, we currently and its capacitance. Moreover, in practically realizable unit cells for
have methodologies that either elaborate the structure of the classical AIMD even an elementary fluctuation of the ion distribution or polar
EDL1–3 or study the charge transfer per se assuming consistency be- solvent molecules near the interface, caused by either thermal motion
tween the atomistic structure of the EDL and the external conditions.4 or by the loss of inventory accompanying reduction, causes a drastic
The crossover between the two approaches is not well defined, such change, on the order of ±1.0 V, in the calculated electrode potential.23
Such large fluctuations of the potential might then in turn trigger an
∗ Electrochemical Society Member. artificial spontaneous cascade of ion reductions that might lead to a
z
E-mail: abaskin@lbl.gov run-away regime. To circumvent this problem, one needs to limit stud-

Downloaded on 2017-11-09 to IP 190.111.30.254 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
Journal of The Electrochemical Society, 164 (11) E3438-E3447 (2017) E3439

ies to those with a large portion of concentrated electrolyte16 where the effects of ET on the structure of the EDL. The same holds true
both fluctuations in the charge profile of solvent and ions are negligible for any aspect of dynamics at the interface. This limits the scope of
with respect to the thermodynamic average. However, calculations at continuum models to the classical EDL regime. However, the theory
this scale may not be currently achievable due to computational over- does not provide estimates of its limits in terms of the applied external
heads. bias, which may lead to an erroneous picture if the applied potential
Lastly, the lack of a computationally tractable unified description for unknown system extends beyond the safe potential (or surface
of the continuous electronic and discrete solute/solvent degrees of charge) range of the EDL, when only reorganization of the electrolyte
freedom constitutes another fundamental problem for the atomistic is assumed. Another problem of continuum models is the incomplete
description of biased interfaces. Customarily, the typical condensed- connection to the equilibrium thermodynamics of the metal-liquid in-
phase AIMD setup, which ideally should be open at either end away terface, whose description is usually done in terms of the unique (for
from the interface, for electrons at the electrode and for dissolved given conditions) redox potential as a starting point. In contrast, for
ions in the electrolyte, is constrained by periodic boundary condi- continuum models, the starting point is the state of zero charge of the
tions (PBC) in conjunction with the requirement of global charge surface (zero potential with respect to its value in the bulk) character-
neutrality needed to define the potential drop across the interface. A ized by flat ion density profiles which, from a thermodynamic stand
specified concentration is realizable within the atomistic picture only point, is a non-equilibrium state defined by the so-called potential of
as integer ratios of solute species to solvent molecules. However, in zero charge (PZC). The proper evaluation of the offset between these
the conventional picture of continuous densities of electronic states starting points for the electrode potential requires the introduction of
in metals, the variation of chemical potential of electrons is realiz- the equilibration mechanism, whether it is due to dipolar screening that
able via essentially non-integer electron population. Obviously, such induces the accumulation of extra surface charge in the pure solvent
a treatment is not possible for discrete ions within a finite simula- or the actual equilibration of the metal electrode with the electrolyte
tion. In order to represent realistic situations, the boundary conditions containing its own ions. In the framework of the hybrid continuum
(applied bias) and the molecular content of the AIMD setup must be models that treat metal electrodes exposed to dielectric media quan-
mutually compatible in terms of system size as well as sampling time. tum mechanically47 this problem is solved only heuristically and the
For a given upper bound on system size, only a discrete and limited ambiguity of mapping the electrode potentials of continuum models
set of boundary conditions is then possible and vice versa. Therefore, onto the thermodynamic scale remains.
in practice, the AIMD setup for a biased interface must be subjected To go beyond the classical EDL regime, a combination of ET
to additional serious limitations suiting the goals of simulations. For Tafel-like kinetic equations9,53 with ion transport kinetics of contin-
instance, the steady state regime at (or beyond) the redox potential uous media54 is used. Despite some progress in thermodynamically
may become inaccessible to AIMD as it would require simulations at consistent continuum transport theory,55,56 this approach lacks self-
constant chemical potential of ions24 to replenish the lost ones, while consistency and does not resolve the effect of ET on the structure of
the actual goal usually is to study the nascent “just forming” states the double layer due to the application of a linearized version of Ohm’s
resulting from irreversible ET. law to the steady-state regime for the homogeneous system. At this
As a conceptual alternative to the AIMD schemes, continuum ap- point, the input from AIMD that provides immediate access to con-
proaches represent the electrolyte as a structureless medium and its figurational and even chemical/electronic degrees of freedom would
response to the external polarization reflects a thermodynamically av- be very valuable, but within current limits on length and time scales,
eraged estimate of the ion distribution with respect to distance from such atomistic scale ET events may cause changes which would push
the interface that is consistent with externally imposed boundary con- the system far from thermodynamic equilibrium (e.g., significantly
ditions (e.g. the electrolyte concentration and the potential difference) altering concentration) thus devaluating further estimates of its char-
at the expense of the loss of information about local molecular-scale acteristics from AIMD.
spatial and dynamic variations. A class of such hybrid continuum Due to the obvious complementary character of the AIMD ap-
models that combine the continuum description of an electrolyte and proach and continuum models, a self-consistent combination of these
the atomistic DFT-based treatment of the electrode is based on the two would permit a more realistic description of the electrode-
generalized Poisson equation (GPE)25 electrolyte interface. In particular, instead of relying on the spon-
  taneous equilibration of the slow molecular subsystem, from an atom-
∇ · ε (r ) ∇ −  ϕ (r ) = 0 [1] istic computational stand point it would be much more efficient to
have it pre-equilibrated by means of a continuum model for each
with a smooth spatially dependent dielectric function ε(r ) 23,26–28 and value of the external electrode potential. Likewise, by extending the
various polarizable continuum models.29–31 These mean-field models applicability of a continuum model to the regime when the electrode
are generalized in the framework of classical DFT32–35 in terms of potential is close to (or beyond) the redox potential would allow the
free energy functionals [ϕ; ρ+ , ρ− ] whose variation with different formulation of a robust computational protocol for AIMD simulations
boundary conditions for the electrostatic potential provides the ther- that would avoid the pitfalls of having the molecular system far from
modynamically optimal spatial characteristics of the interface. The (quasi-)equilibrium. At the same time, using continuum models with
accounting for steric effects,36–41 short-range coulomb correlations, correct boundary conditions for ϕ(r ), one can evaluate useful criteria
and many-body interactions introduced by the self-energy corrections relevant to AIMD, such as the minimum required size of the system
to the mean-field approximation42–45 enables going beyond the stan- and the realistic range of the electrode potential for a given electrolyte
dard Poisson-Boltzmann approximation (PBA)46 and explains over- concentration.
screening and crowding phenomena as well as the “camel-bell” tran- In pursuit of developing such a self-consistent methodology for
sition in differential capacitance for ionic liquids.47,48 Depending on modeling biased interfaces, in this work we propose a continuum
a particular implementation, the corresponding ad hoc modifications phenomenological model to describe the effects of ET on the structure
are introduced either directly into an action-like free energy functional of the EDL as well as the effects of specific adsorption of ions. With
resulting in the Helmholtz-like form of GPE49,50 or lead to specific properly tuned input parameters, this advanced model should permit
changes in the boundary conditions for ϕ(r ).8,51,52 both design and validation of thermodynamically consistent AIMD
Despite these advances, these continuum models do not include the simulations and enable exploration of the limits of a microscopic
effects of specific adsorption, thus predicting unrealistic electrostatic description of biased interfaces.
potential profiles known to be dominated by specifically adsorbed
ions. More importantly, by virtue of the general theorems of elec- Theory
trostatics, the boundary conditions of the GPE for the electrostatic
We start with the PBA of the free energy functional
potential at the electrode, ϕ0 , and the electrode surface charge density, 2
σ, are uniquely connected, which makes it impossible to account for [ρ+ (r ), ρ− (r )], where only the coulomb potential φ = 4πε0 eε|r −r  |

Downloaded on 2017-11-09 to IP 190.111.30.254 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
E3440 Journal of The Electrochemical Society, 164 (11) E3438-E3447 (2017)

between monovalent ions is incorporated in the energy term and Here, in contrast to other studies where the dielectric function is
core-core repulsions are neglected. With a self-consistent mean-field explicitly dependent on either the electric field ε(E)61 or the elec-
electrostatic potential ϕ(r ), the entropy term associated with solvent tronic density of solutes26 or metal electrodes ε(ρ)49,50 causing di-
molecules,39,57 and a mass conservation term defined by the chem- electric saturation in regions of high electric field or electronic den-
ical potentials of positive and negative ions μ± ,36,37 the functional sity, we use a spatially dependent dielectric function ε(r). As has been
[ϕ(r ); ρ+ , ρ− ], that from now on we will call the Gouy-Chapman demonstrated,62 a smooth “step-like” function ε(z) which decays from
with the exclusion volume (GCEV) model, gives rise to a modi- its bulk value to the high frequency limit in the vicinity of the elec-
fied Poisson-Boltzmann equation (MPBE) with Neumann or Dirichlet trode (at sufficiently high surface electron density σ) can accurately
boundary conditions: ∂ϕ |
∂z z→∞
= 0, ∂ϕ |
∂z z→0
= − εσ0 ε or ϕ(0) =ϕ0 . reproduce a response consistent with Booth’s ansatz for ε(E). In this
Here we propose to generalize this GCEV model by adding terms way, we ignore the dependence of ε(z) on the surface charge density,
describing non-electrostatic specific interactions 1 (r ) and 2 (r ) which will, however, be reflected in an underestimation of the differ-
of monovalent (z = 1) ions of effective size a with the electrode ential capacitance in the small regime of electrode potentials around
surface: PZC.
The variation δ = 0 and δρδ± = 0 leads to the modified Poisson-
  δϕ
  ε0 ε (r ) Boltzmann equation (MPBE) that reads as follows:
 ϕ (r ) ; ρ± (r ) = dr − |∇ϕ|2 + eϕ (ρ+ − ρ− )
2 eϕ−1 −eϕ−2
 en b e kB T
− e kB T
∇ · ε (r ) ∇ϕ (r ) =  eϕ− 
+ 1 (r ) ρ− + 2 (r ) ρ+ − μ− ρ− − μ+ ρ+ ε0 1 −eϕ−2
1 − 2a 3 n b + a 3 n b e k B T + e k B T

kB T [3]
+ dr {a 3 ρ− lna 3 ρ− + a 3 ρ+ lna 3 ρ+ where n b is the bulk concentration of the ions and a is the characteristic
a3
size of the solvated ion. Its solution defines the extreme saddle point
+ (1 − a 3 ρ− − a 3 ρ+ )ln(1 − a 3 ρ− − a 3 ρ+ )}. of the functional  and spatial profiles ρ± for counterions that become
dependent on each other. As both ρ+ and ρ− are usually complex func-
[2]
tions of coordinates parametrized by the electrode potential, it is more
convenient to explore the properties of the model by analyzing the
By definition, 1 (r ) and 2 (r ) adsorption potentials, as a part of “phase relationship” f (ρ+ , ρ− ) = 0. For the Gouy-Chapman model
the free energy functional, correspond to the interactions of ions with (as a → 0), the phase relation is well known ρ+ ρ− = n 2b .46,47 As we
the neutral electrode in the infinitely diluted limit. Therefore, these found, in the GCEV model36 ρ+ and ρ− are dependent through the
potentials do not include any dependencies on surface coverage or
relation ρ+nρ2 − = ( 1−a1−2a
3 ρ −a 3 ρ
+ − 2
bias in contrast to the adsorption energies used in the phenomenolog- 3n ) . Due to the fundamental symmetry of
b b
ical electrochemistry (e.g. Frumkin isotherm58,59 ). Indirectly, 1 (r ) the coulomb interactions, the common feature of this class of models
and 2 (r ) should depend on the bias as it tunes the overlap of the is that f (ρ+ , ρ− ) = 0 is independent on the boundary conditions, as it
ion frontier orbitals with the electrode surface states. However, since is exemplified in Fig. 1a. The inversion symmetry ρ+ ↔ ρ− is guaran-
we disregard the electronic structure of the electrode, these quantum teed by the equal size of the ions. The introduction of non-electrostatic
effects will be ignored here. However, the proposed model keeps the interactions 1 (r ) and 2 (r ) results in the concomitant modifications
dependencies of the free energy of adsorption of ions on the bias and of f (ρ+ , ρ− ) = 0 which now reflects the interplay between adsorption
surface coverage since it operates with the self-consistent electrostatic potentials, electrostatic forces and the effects of volume exclusion. As
potential that takes into account finite size effects and all interactions a consequence, the phase relation becomes dependent on the electrode
between ions except non-Coulombic ones. We also ignore repulsion potential ϕ0 (see Fig. 1b).
between solvent molecules and thus ignore the layering in the double Eq. 2 describes the double layer operating as an ideal non-
layer.60 linear capacitor. As the electrochemical potential of electrons

Figure 1. “Phase relationships” f (ρ+ , ρ− ) = 0 for ion concentrations. a) The original GCEV model. Intervals of potentials are shown by dashed arrows. Maximum
concentration ρ+ = ρ− = 13.28 M. b) Continuum model developed in this paper (see Eqs. 7–8). A symmetrical case when 1 = 2
= 0 is shown, the adsorption
energy E ads = 0.1 eV (for details of potentials see text). As the electrode potential |ϕ| increases, the phase relation resembles more and more the relation
for the GCEV model. Orange, red and pink colors are for positive electrode potentials; light blue, blue and violet are for negative potentials. For both models,
a = 5Å, n b = 2M, ε = 80. Green and red shaded areas mark regions that are electrostatically allowed and forbidden, respectively.

Downloaded on 2017-11-09 to IP 190.111.30.254 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
Journal of The Electrochemical Society, 164 (11) E3438-E3447 (2017) E3441

approaches the reduction level defined for the reaction O + e− → R whose variation gives a new MPBE:
as −eφr ed +  O (φr ed ) =  R (φr ed ), at a certain overpotential one
would expect non-zero faradaic current and significant changes in the ∇ · ε (r ) ∇ϕ (r )
⎛ ⎞
structure of the double layer associated with the “leakage” of the ca- eϕ−1
en b ⎝ e kB T ρ ∗
(z)
pacitor. However, with a single self-consistent MPBE it is impossible = 1 − a 3 ρ∗+ (z) − + ⎠
to capture these effects as explained above as a general property of ε0 eϕ−1
1 − 2a 3 n b + a 3 n b e k B T nb
equations of electrostatics.
To resolve this issue, we introduce a simplified kinetic equation [8]
which describes the time evolution of the concentration of the re-
ducing species, ρ+ , that is characterized by two different time scales. with the same boundary conditions as for Eq. 3. Eqs. 3, 5, 6 and
The kinetics of the forward reduction process ρ+ → ρ0 is character- 8 constitute the formalism which describes the effects of ET on the
ized by τ→ whereas the kinetics of the backward oxidation processes structure of the double layer.
ρ0 → ρ+ is associated with τ← . Here, ρ0 is the time and spatially de-
pendent concentration of the reduced species. Finally, TD introduces Results and Discussion
the kinetics related to the migration/diffusion of ions in the double
layer. For simplicity’s sake, here we consider only the effect of cation Effects of specific adsorption and electron transfer.—Eq. 3 cap-
reduction. The equation reads as follows: tures several important features. The double-layer capacitance is dom-
inated by the effects of adsorption of ions.63 In Fig. 2 we show
∂ρ+ (z, t) c̃ (z) ρ+ (z, 0) − ρ+ (z, t) the progression of differential capacitance, Cdi f = dϕ dσ
, for two
=− ρ+ (z, t) +
∂t τ→ TD different adsorption energies of anions with E ads = 3k B T (blue
solid line) and E ads = 6k B T (red solid curves) for concentrations
c̃ (z)
+ (ρ+ (z, 0) − ρ+ (z, t)) , [4] n b = 0.1, 0.2, 0.4 and 0.8 M. Here we assume that the electrode
τ← is “the sticky surface” for anions with the adsorption energy profile
where c̃(z) describes the portion of ρ+ (z, t) or ρ0 (z, t) that gets re- defined by 1 = E ads (1 − e−k(z−z0 ) )2 − E ads where k = 2Å−1 and
duced/oxidized at coordinate z. Here we assume a tunneling regime z 0 = 0, and no specific adsorption for cations 2 = 0. The effec-
− z
when c̃(z) = e a0 , and a0 specifies the “depth” of the layer where tive size of ions is set as a = 3Å far enough from the saturation
ρ+ (z) effectively reduces/oxidizes. The function c̃(z) reflects the limit ∼10Å at n b = 0.8 M which should guarantee the minimum
mechanism of the electrochemical process. ρ+ (z, 0) is the initial cation of Cdif , at the PZC in the absence of ion adsorption, dielectric con-
concentration profile corresponding to the ideal capacitor regime (as stant ε = 80 and temperature T = 300K . For sufficiently negative
defined by the solution of the MPBE above). Thus, the first two terms electrode potentials ϕ < −0.15 eV, the effects of specific adsorption
in Eq. 4 describe the competition between the spatially dependent ET vanish, and both pictures give the same results. However, at higher
and the thermodynamic force bringing new cations from the bulk elec- adsorption energy of anions, one sees not only the shift of the min-
trolyte to restore the initial profile that corresponds to the extremum ima of Cdif , but also a peculiar “camel-bell” transition.37,38,40 Similar
for the free energy functional of Eq. 2. The last term in Eq. 4 describes trends were experimentally observed for aqueous solutions of simple
the oxidation rate of newly reduced ρ0 (z, t). Here we also neglect the anions for polycrystalline Ag-electrodes.63,64 The values of Cdif are
effect of the accumulated ρ+ (z, t) and do not consider any mechanism overestimated. However, Eq. 3 with advanced adsorption and steric
of diffusion from the surface. Note, that the PBA does not take into repulsion potentials65,66 may easily produce, without introducing the
account any aspects of dynamics. The combination of Eqs. 3 and 4 ad hoc term rε0 , the effect of the compact Helmholtz layer with a linear
is not self-consistent, yet provides an approximate way to include the profile of the electrostatic potential resulting in more realistic values
effects of ET on the structure of the EDL. of Cdif .
At steady state, Another useful feature of Eq. 3 is that it allows us, in princi-
ple, to estimate the Esin-Markov coefficient and the electrosorption
c̃(z)
τ→
+ 1 valency l B = −z( ∂σ M
) , where σ M is the charge density on the metal
∂σi E
ρ+ (z, ∞) ≡ ρ∗+ (z) = ρ+ (z, 0)
TD

[5] and σi is the charge density connected with the specifically adsorbed
c̃ (z) 1
τ→
+ 1
τ←
+ 1
TD species,64,67 and to compare that with experimental values of l B . That
opens the way to theoretically evaluate the charge transfer coeffi-
Finally, the ET times, τ→ and τ← depend on the difference between the cient, λ, and refine the model based on experimental results.68,69 With
reduction level eφr ed and the electrochemical potentials of electron. both the effects of lattice saturation and specific adsorption accounted
We assume that for, the concentration profiles ρ− (z) and ρ+ (z) can be viewed as a
generalization of a Frumkin isotherm. The model presented above
eφr ed −μe (ϕ) τ20
τ→ (ϕ) = τ0 e kB T
and τ← (ϕ) = [6] also naturally includes the interactions with the image charge in-
τ→ (ϕ) duced on the electrode surface but does not account for the contri-
bution from the “quantum capacitance”. That affects the estimates of
In order to find the self-consistent solution for the electrostatic
the PZC accordingly. Lastly, a smoothed step-like dielectric function
potential as well as for the corresponding density profiles for anions, ε−1
ε(z) = 1 + 1+ex p(−α(z−z 26,49
similar to that found in approximate so-
we rewrite the free energy functional ∗ [ϕ(r ); ρ− (r )] with the fixed 0
 ))

cation spatial profile ρ∗+ (z) lutions of the Poisson equation with the Booth’s-like ε(E),61,62 where
  α = 5 and z 0 = 0.5Å in Eq. 3 renormalizes the overall capacitance
  ε0 ε (r ) making it more realistic as compared to that from the GCEV model,
 ϕ (r ) ; ρ± (r ) = dr − |∇ϕ|2 + eϕ ρ∗+ − ρ−
2 as shown in Fig. 2 by green and orange curves corresponding to the
 anion adsorption energies E ads = 3k B T and 6k B T , respectively.
+ 1 (r ) ρ− + 2 (r ) ρ∗+ − μ− ρ− − ρ∗+ The analysis of the model that combines the simple ET kinetic
Eq. 4 and the self-consistent Eq. 3 allows us to draw the following
 observations. At μe  φr ed (ideal capacitor regime), τ→  τ←
kB T
+ dr {a 3 ρ− lna 3 ρ− + a 3 ρ∗+ lna 3 ρ∗+ but at this electrode potential not enough ρ0 is created to drive the
a3 current in the anodic direction. The second term in Eq. 5 induces
+ (1 − a 3 ρ− − a 3 ρ∗+ )ln(1 − a 3 ρ− − a 3 ρ∗+ )}, asymmetry between ρ+ (z) and ρ0 (z) that leads to non-zero current at
μe (ϕ) = φr ed , where τ→ = τ← = τ0 . Further decrease of the electrode
[7] potential promotes a cathodic current by decreasing τ→ but at the same

Downloaded on 2017-11-09 to IP 190.111.30.254 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
E3442 Journal of The Electrochemical Society, 164 (11) E3438-E3447 (2017)

Figure 2. Progression of differential capacitances with specific adsorption of anions. a) Adsorption energies Eads = 3k B T (blue solid line), dielectric constant
ε = 80, b) E ads = 6k B T (red solid curves) dielectric constant ε = 80, c) Eads = 3k B T (solid green line), ε = ε(z), d) E ads = 6k B T , ε = ε(z) (orange solid line)
for bulk concentrations n = 0.1, 0.2, 0.4 and 0.8 M. Black dashed curves - differential capacitance of the GCEV model.36

time it depletes the cation concentration next to the electrode, slowing the dashed curves). Next, as the electrochemical activity, 0 , of the
down reduction. On the other hand, the reduced ρ0 will be less and less electrode increases progressively (red curves), we see changes in Cdif
likely oxidized due to the increase of τ← . This competition of rival in the same voltage range. The inclusion of the effects of specific ad-
factors results in a steady state current which depends on the initial sorption and solvent polarization mixed with ET will have a profound
mutual alignment of the reduction level φr ed and initial E f . However, effect on the EDL structure affecting the PZC. In principle, the param-
at the finite scanning rate of voltammetry, the current may appreciably eters for the presented model can be extracted from EIS experiments
deviate from the steady magnitude and should exhibit hysteresis. mapping the characteristics of ET resistance and ion diffusion with
In Fig. 3 we show the effects of ET on the ion concentration pro- corresponding characteristic time scales TD , τ→ and τ← .
files and the electrostatic potential as compared to the regular EDL
predicted by the GCEV model. No effects of specific adsorption are AIMD simulations of the biased interface: fundamental limits
included. Different electrochemical conditions are defined by parame- and implications from continuum models.—Thermodynamic consis-
ter 0 (see the inset of Fig. 3) which specifies the offset between Ef of tency with interfacial electrochemical boundary conditions requires
the neutral electrode and the reduction potential φr ed and thus its elec- that AIMD simulations meet several interconnected requirements. Let
trochemical activity. As τ0 and T D are set as 1 and 100, respectively, us make some common assumptions on typical AIMD calculations
τ→ = τ← = TD at the potential corresponding to 0 = 0.115eV. relevant to condensed phase interfacial studies of half-cells: (1) We
At the potentials φred − μe ≤ 0.115 eV, τ→ ≤ TD and the EDL employ periodic boundary conditions (PBCs), where the semi-infinite
behavior starts deviating from that of the ideal capacitor. A peculiar solid electrode surface is modeled as a sufficiently thick slab of atoms
non-monotonic concentration profile of anions (Fig. 2c) is a result extended periodically in the plane of the surface – typically less than
of the competition between attraction to cations, repulsion from the 10 atomic layers may suffice for most simple metals. These PBCs
negatively charged electrode and steric factors which preclude high define two relevant characteristics of the associated orthorhombic su-
local concentration of both types of ions at the same distance. If the percell - its overall length in the direction of the surface normal, l, and
backward oxidation of ρ0 is suppressed (τ←  τ→ at μe = φred ), its area in the plane of the surface, S; (2) The liquid/electrolyte region
the model predicts the nearly full depletion of ρ+ at z = 0 which is defined by another “slab” sufficiently thick to preserve a particular
at small 0 may lead to the depolarization of the electrode surface volume with bulk behavior (material density, molecular diffusivity, av-
(σ > 0 at ϕ < 0). The model also predicts the regimes at which the erage concentration and spatial correlation of dissolved species, etc.)
electrode surface is covered by anions even at ϕ < 0. at some distance from the electrode surface; (3) overall the length
The effects of ET impact the differential capacitance of the dou- of the AIMD simulation cell is sufficient to also include interfacial
ble layer. Though the differential capacitance is strictly speaking well phenomena different from the bulk behavior in both the electrode and
defined only in the range of potentials within the classical double electrolyte, for example, to encompass the electric double layer. If we
layer, here we show the “renormalization” of surface charge density assume the electrode to be metallic, then these interfacial differences
in the presence of non-zero current. In Fig. 4 we show the correspond- will be limited (excluding the outermost layer of metal atoms) to the
ing modifications of Cdif as the electrode potential approaches φr ed . electrolyte – this would not be so for any insulating or semiconducting
Focusing first on the effect of varying solution concentration (blue layers formed on the electrode surface, perhaps as models of a solid
curves) at constant offset 0 = 0.22 and at φr ed − μe ≈ k B T (below electrolyte interphase. So, the volume of the AIMD supercell can be
−0.15 V ), we notice Cdif drops dramatically (compare with Fig. 1 or obtained as the product S (lsolid +linterface +lbulk ). Within this volume we

Downloaded on 2017-11-09 to IP 190.111.30.254 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
Journal of The Electrochemical Society, 164 (11) E3438-E3447 (2017) E3443

Figure 3. Structure of the double layer. Concentration profiles of cations (red curves), anions (blue curves) and electrostatic potential (black curves) at different
electrode potentials and electrochemical conditions defined by 0 (see the inset) with (solid curves) and without (dashed curves) effects of ET. The electrolyte
concentration n b = 6 M, ion size a = 5Å, The “depth” length of the ET function c̃(z) is a0 = 0.2Å, the ET time τ0 = 1, migration time TD = 100. The
ration of τ+ /TD and the surface charge density σ are shown for each electrode potential w/o the effects of ET. a) and b) ϕ = −0.1, −0.185, −0.3 V . c) and d)
ϕ = −0.01, −0.025, and −0.05 V .

require global charge neutrality, based on the three kinds of charge to the interfacial region, Slinterface . We assume that thermal fluctuations
carriers which meet at the interface, such that n e = N+ − N− , where are insufficient to drive the system out of this metastable state toward
N+ and N− are the number of cations and anions in the liquid, respec- the lower energy state with the excess charge split evenly on either
tively, and n e is the amount of excess electronic charge (electrons if side of the slab.
positive, holes if negative). By virtue of the necessarily discrete nature of the atomistic descrip-
In all that follows, we will refer to one specific interface as the tion of the ions, the charge neutrality requirement limits n e to integer
active electrochemical interface. Clearly, a slab of finite thickness values and sets some minimum limits on the surface change density at
has two sides that meet the electrolyte under PBCs, but we assume the electrode (described below). In the absence of an explicit reference
that at finite bias, inversion symmetry is broken and only one surface electrode, an internal reference for the half-cell AIMD setup is pro-
becomes electrified. This can be achieved by the initial arrangement vided by the calculated, self-consistent, thermodynamically-averaged,
of the ion distribution, which should mimic the profile determined electrostatic potential in the bulk region of the explicitly included elec-
from the continuum model. Specifically, within the “bulk” electrolyte trolyte. Vital characteristics of the supercell are: its ability to screen
volume, S lbulk , there are an equal number of cations and anion, while electric fields, such as that generated by the electronic surface charge
the mismatch defining the number of surface electrons, ne , is limited density, within a certain length lscr and a particular time scale. At ther-
mal equilibrium, material specifics define these length and time scales
as the Debye length, δ D , and the slowest relaxation time τx 70 (e.g. the
inverse rotational diffusion coefficient), respectively. Under an exter-
nal bias, the Gouy-Chapman length l GC 71 dictates the variation of the
distributions of ions and defines the number of ionic species within a
layer next to the electrode as compared to its value in the bulk. Thus
l GC via the link with the surface charge density limits the minimum
electrolyte concentration and the electrode surface area, S, connecting
the efficiency of the time averaging over the simulation length and the
“static” (or relaxed) picture given by continuum models. Though the
applicability of these lengths and time scales is, strictly speaking, lim-
ited by the idealized diluted electrolytes (ionic strength below 0.2 M,
electrode potentials below 80 mV72 ), they can still be used to estimate
the critical parameters of AIMD setups.
The combination of the aforementioned conditions leads to the
following limitations for the AIMD simulations of biased interfaces
(here we consider the charging of the electrode with electrons).

1. Due to the discrete nature of the ionic charge imbalance, both


Figure 4. Effects of ET on the differential capacitance. Dashed curves – no
the surface charge density, σ, and the electrode potential, ϕ, are
ET, solid lines – with ET. Blue curves – progression of Cdif for ion con-
centrations n b = 0.1, 0.2, 0.4 and 0.8 M. φr ed − E f = 0.22 eV . Ion quantized values with the minimum “quanta” σ = ±1/S and
size a = 3Å. Red curves – the effect of the electrochemical activity of the ϕ = L(σ) defined by the capacitance of the EDL. For the
surface. Ion size a = 5Å. The concentration n b = 6M. The offsets are AIMD setups with a fixed number of ions, the maximum im-
0 = 0.25, 0.24, 0.23, 0.22 eV . balance n e ≤ N−T − 1, where N−T is the total number of anions

Downloaded on 2017-11-09 to IP 190.111.30.254 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
E3444 Journal of The Electrochemical Society, 164 (11) E3438-E3447 (2017)

defined by the bulk electrolyte concentration, n b , and the size of supercell will generate unphysical long-ranged electrostatic er-
the interfacial region. rors that may also prevent fair comparisons with measurements.
2. The surface area has to be within Smin ≤ S ≤ Smax range. This example also shows that for a certain concentration the sur-
Smax is limited by the computational feasibility of the simulation face area S may not be too small as it would require an excessive
(maximum number of solvent molecules). For aqueous solution, length of the simulation cell L to accommodate the appropriate
Smax [Å2 ] ≤ Nwater 29.89L[Å]
, where L is the length of the solution part number of ions to support the interfacial ionic imbalance which
of the cell. The lowest limit for Smin is dictated by the minimum is again tantamount to a very long simulation time.
electrode potential ϕmin (in electrochemical scale) that has to be
in the safe EDL region at least several k B T away from reduc- Here, we propose to use our continuum model to assess the critical
tion potential φr ed for the minimum ionic imbalance (n e = 1). limits for the AIMD setup to meet the requirements of thermody-
However, for an AIMD practitioner Smin may rather be limited namic stability from the standpoint of the bias. Specifically, to de-
by other factors such as incompatibility of the bulk electrolyte termine practicable AIMD simulations, we focus on the size of the
concentration with the target bias (see below). AIMD setup in terms of number of atoms, which, for condensed phase
3. In either conditions (constant surface charge density73 or constant systems, is directly linked to spatial dimensions via material density.
chemical potential of electrons74 ), the size of the cell must be such For the moment, we omit discussion of the necessary time scales for
that thermal fluctuations of polar solvent molecules and ionic converged thermodynamic averaging, with the expectation that this is
species do not exceed a certain limit. As has been shown,73 for merely an overall multiplicative factor in determining computational
a pure polar solvent, the probability distribution of the electrode cost and tractability.
potential derived from Marcus theory for linear solvent polariza- For comparison, we use both the GCEV model and our model that
tion has a Gaussian shape with a standard deviation independent includes the effects of the specific adsorption of anions and spatial
of the actual electrode potential. In the realistic AIMD setup with dependence of the dielectric function. Other effects, such as ET can be
water (L = 10Å and S = 100Å2 , 32 water  molecules), the included in a similar way. The input parameters (bulk concentration,
standard deviation of the electrode potential ϕ2  = 0.51 V , dielectric function and surface charge density) are used to obtain
which is much larger than the thermal energy. Clearly, for such a the information about the self-consistent electrode potential and the
small AIMD setup, correspondingly large fluctuations might eas- ion density distributions that are further used as an input for the
ily induce electron transfer events across the solid-liquid interface AIMD simulations. In this way, we can try to reproduce the pre-
with little correspondence to reality. Explicit inclusion of ionic equilibrated state compatible with the boundary conditions and skip
species8,9 may introduce even more serious problems, since the the slow equilibration of the molecular subsystem. The AIMD is the
interfacial fluctuations of discrete ion distributions may induce then used to obtain anew the ion profiles as an updated input for
even larger fluctuations of the electrode potential or may shift the the continuum model. In Fig. 5 we show the conceptual scheme of
energy of a LUMO orbital for a molecular species in the elec- input/output parameters.
trolyte into resonance with the electrode Fermi level. The rigorous To assess the screening length, lscr , we require that at each induced
estimations of the effects of fluctuations is a non-trivial prob- surface charge density, σ, the deviation of the average electrostatic
lem that requires one to use not only the fluctuation-dissipation potential at that distance from the overall potential drop meets the
theorem70 which relates the differential capacitance and the vari- condition | ϕ(lϕ
scr )
| ≤ 0.02 which guarantees the approximate neutrality
ance of the total charge q 2  = Cdi f k B T or the electrode po- of the interfacial region. To define the length of the bulk region, lb ,
tential ϕ2  = Ck BdiTf but some other kinetic characteristics of the we require it to be long enough and contain enough ions to support
solution, e.g. diffusion constants, viscosity, etc. To the best of the ionic imbalance and preserve approximately the bulk solution
our knowledge the general expression for the spatially dependent concentration in the overall cell: N+T = 2N+c , N−T = N+T − n e , where
standard deviation of ion densities ρ2± (z; σ) is unknown for the N±T – total number of ions of each sign in the cell L = lscr + lb ,
advanced continuum models and will be a subject of future stud- N±c = S ∫l0scr ρc± (z)dz – is the number of ions in the interfacial region
ies. Note that the necessary constants in the ET model described predicted by the continuum model. For a given surface area S we
above may be determined from consistent AIMD simulations, es- evaluate the length of the cell L as a function of the concentration for
pecially the position in energy of available electronic levels in the each value of the charge imbalance n e and thus evaluate the minimum
electrolyte and estimates of the time constants for charge transfer size of the cell for each value of the electrode potential. For the model
based on constrained electronic structure calculations.75 that includes the effect of specific adsorption of anions and spatial
4. To properly describe the interface, the cell has to comprise the variance of dielectric function we use the parameters described below.
interfacial region characterized by lscr (screening length) and the 1 = E ads (1 − e−k(z−z0 ) )2 − E ads (k = 2Å−1 , E ads = 0.5 eV and
bulk region whose dimensions allow for the√proper description of ε−1 
z 0 = 0). ε(z) = 1 + 1+ex p(−α(z−z  )) , where α = 5, and z 0 = 0.5Å. The
ion-ion correlations at finite concentration ( S ≥ 2lc , lbulk ≥ 2lc , 0

where lc is the ion-ion correlation length and lbulk is the length effective size of ions is set as a = 4Å, dielectric constant ε = 80 and
of the bulk electrolyte region, defined already). With a fixed temperature T = 300K . Empirically we found that lscr ∼ √1nb and
number of ions in the simulation cell, the bulk region will ef- has a very weak dependence on the boundary conditions similarly to
fectively work as a buffer to damp oscillation in the ion pro- what one would expect from the Debye screening.
file at the interface which is in fact what may define its dimen- In Fig. 6a we show the results obtained with the GCEV model
μC μC
sion. To withstand a certain potential drop across the interface, for the surface charge density 8 cm 2 ≤ σ ≤ 40 cm 2 keeping the over-

there will be corresponding ion density profiles with a spatial all neutrality of the cell. The estimates show that for concentrations
ionic imbalance. For a monovalent binary aqueous electrolyte n b ≤ 0.5 M AIMD simulations would require excessively long cells
(ε = 80, T = 300 K , n b = 1M, z = 1, δ D ≈ 3Å) con- that for the higher charge imbalance (e.g. n e = 5) can easily ex-
tained in the simulation cell with a size L = 10Å and S = 150Å2 , ceed computational feasibility (150Å → 1000 water molecules, black
the minimum surface charge density σmin = 10.7 μC/ cm 2 . How- dashed line). Alternatively, these limitations can be seen in terms of
ever, if the concentration is n b = 1M, this unit cell contains only maximum applied potential (see Fig. 5b) that cell can support. For con-
one ion pair which along with the estimates of l GC = 3.4Å at centration 0.5 M, 1 M and 2 M, ϕmax is 0.12 V, 0.2 V and 0.38 V ,
σmin = 10.7 μC/ cm 2 suggests that this AIMD setup is a very respectively. As long as the GCEV continuum model captures well
rough representation of the physical system where the screening the electrolyte response, these estimates seriously challenge the at-
of the surface charge may be obtained only at the expense of very tempts to simulate atomistically thermodynamically stable interfaces.
long simulation time. Furthermore, the strong implicit correlation To meet the stability requirements, the short AIMD cells (≤ 20Å)
between charged species in neighboring periodic replicas of the can be used to simulate interfaces only at the potentials smaller or

Downloaded on 2017-11-09 to IP 190.111.30.254 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
Journal of The Electrochemical Society, 164 (11) E3438-E3447 (2017) E3445

Figure 5. The scheme of the hybrid methodology. Input parameters and the flow of input/output for both the continuum model and the AIMD simulations are
shown. Integer numbers of ions N±c (z i ) in the interfacial layers of width a are used to restore the spatial profile of ions in the AIMD setup. The schematics of
the ion-ion pair correlation function with the characteristic correlation length lc is shown to emphasize the critical length scale of the bulk region of the AIMD
setup.

Figure 6. Estimates of critical parameters of AIMD setups based on the continuum models. (a,b) The GCEV model. (c,d) The continuum model developed in
this paper (for parameters see the text). (a,c) The minimum length of the liquid part of the cell (surface area S = 200Å2 ) as a function of bulk concentration of
monovalent electrolyte for different charge imbalance n e = 1, 2, 3, 4, 5 ē. (b,d) The minimum length of the liquid part of the cell as a function of maximum applied
potential for different bulk concentrations n b = 0.5, 1, 2, 3, 4, 5 M. Lengths are evaluated at discrete values of potentials corresponding to the charge imbalance
n e . Solid lines are drawn to guide the eye. Black dashed line shows the computational “feasibility” limit of 1000 water molecule per cell (∼ =150Å).

Downloaded on 2017-11-09 to IP 190.111.30.254 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
E3446 Journal of The Electrochemical Society, 164 (11) E3438-E3447 (2017)

comparable to kT = 0.025 V which may not be relevant for any 5) These estimates indicate that for realistic electrolyte concentra-
practical applications. tions and currently used AIMD cell sizes, the validity of ab
The consideration of more realistic continuum models helps to initio approaches is limited to very small electrode potentials
alleviate these severe size limitations. In Figs. 6c, 6d we show the (ϕmax < 0.05 V). Simulations at the higher biases would im-
estimates of the cell sizes when specific adsorption of ions and spatial ply strongly non-equilibrium configurations of the solution and
variance of dielectric function are accounted for. The strong specific incompatibility with the targeted external conditions.
adsorption (E ads = 0.5 eV ) of anions reduces the variation of cell 6) Estimates of the critical parameters for AIMD simulations based
lengths within the same range of the surface charge density (see Fig. on continuum models that include the effects of specific adsorp-
5c) since at these conditions the capacity is dominated by the ad- tion and solvent polarization show that simulations of practical
sorption. Therefore, the biggest effect of it is expected to be well voltages (ϕ ∼ 0.1 V) are within reach of atomistic simulations but
seen at lower potentials as it is shown in Fig. 5d. Indeed, for con- require special computational protocols to avoid time-consuming
centration 0.5 M, 1 M and 2 M, ϕmax is 0.21 V, 0.42 V and 0.65 V , pre-equilibration stages.
respectively. This result indicates that AIMD simulations of stable
interfaces at ϕ ∼ 0.5 V can be practicable at realistic concentrations We believe that our findings pave the way toward a self-consistent
as they naturally include such effects. methodology for modelling biased interfaces with the continuum level
The attempts to evaluate the size limits for interfaces at higher description as an indispensable part.
biases encounter serious problems inherent to continuum models per
se. In the absence of ion-ion correlations, this simplified descrip-
tion predicts the over-packing of cations and the full depletion of
anions in the interfacial region. The problems become especially ev- Acknowledgments
ident at lower concentrations at which N−c < 1 in the large region This work was supported by the Joint Center for Energy Storage
(lscr ∼ 15 − 20Å). At this point, AIMD outputs should serve to re- Research, an Energy Innovation Hub funded by the U.S. Department
fine the continuum model in a self-consistent fashion to construct of Energy, Office of Science, Basic Energy Sciences. Portions of this
eventually more realistic interfaces. work were supported by a User Project at The Molecular Foundry
and its compute cluster (vulcan), managed by the High-Performance
Computing Services Group, at Lawrence Berkeley National Labora-
Conclusions tory (LBNL), and portions of this work used the computing resources
In the presented work we develop the model that combines the of the National Energy Research Scientific Computing Center, LBNL,
classical continuum theory of the electric double layer with kinetic both of which are supported by the Office of Science of the U.S. De-
equations to incorporate the effects of specific adsorption and electron partment of Energy under contract no. DE-AC02-05CH11231. Au-
transfer within the formalism based on the free energy functional. thors thank Prof. J. Newman and Dr. Ph. Ross for useful discussions.
Specifically, we achieved the following results. The authors declare no competing financial interest.

1) Introduction of non-electrostatic (e.g., specific adsorption) inter-


actions 1 and 2 of ions with the electrode permits treating the References
system at the level of the differential modified Poisson-Boltzmann
1. W. Schmickler, Electrical Double Layers: Theory and Simulations. Encyclopedia of
equation and leads to significant modifications of the “phase re- Electrochemistry, Wiley & Sons (2007).
lationship” f(ρ+ , ρ− ) = 0 and the differential capacitance of the 2. E. Spohr, Computer Simulations of Electrochemical Interfaces in Adv. Electrochem.
double layer as compared to the GCEV model. Sci. Eng, Vol. 6, R. C. Alkire and D. M. Kolb, Editors, Wiley-VCH Verlag GmbH,
2) Accounting for the non-linear solvent polarization effect results Weinheim (1999).
3. E. Spohr, Molecular Dynamics Simulations of water and Ion Dynamics in the Elec-
in the appreciable renormalization of differential capacitance. trochemical Double Layer. Solid State Ionics, 150, 1 (2002).
4. C. Hartnig and M. T. M. Koper, Solvent Reorganization in Electron and Ion Transfer
These advances permit direct comparison of our model to the ex- Reactions near a Smooth Electrified Surface: a Molecular Dynamics Study, J. Am.
perimental measurements of the electrode surface coverage by specifi- Chem. Soc., 125, 9840 (2003).
cally adsorbed ions (e.g. Hurwitz-Parson method). This in turn enables 5. S. Trasatti, The Concept of Absolute Electrode Potential an Attempt at a Calculation.
J. Electroanal. Chem., 52, 313 (1974).
an access to the details of ion adsorption potential. 6. S. Trasatti, The Absolute Electrode Potential: an Explanatory Note, J. Electroanal.
Chem., 209, 417 (1986).
3) We show that the description of the electric double layer at elec- 7. I. Tavernelli, R. Vuilleumier, and M. Sprik, Ab Initio Molecular Dynamics for
trode potentials close to the redox potentials can be efficiently Molecules with Variable Numbers of Electrons. Phys. Rev. Lett., 88, 231002 (2002).
8. N. Bonnet, T. Morishita, O. Sugino, and M. Otani, First-Principles Molecular Dy-
performed with a formalism based on a pair of coupled modified namics at a Constant Electrode Potential. Phys. Rev. Lett., 109, 266101 (2012).
Poisson-Boltzmann equations. This scheme predicts the peculiar 9. O. Sugino, I. Hamada, M. Otani, Y. Morikawa, T. Ikeshoji, and Y. Okamoto, First-
structure of the double layer in the presence of non-zero faradaic principles molecular dynamics simulation of biased electrode/solution interface. Surf.
current and significant changes in the differential capacitance at Sci., 601, 5237 (2007).
10. T. Ikeshoj and M. Otani, Toward full simulation of the electrochemical oxygen
potentials near the redox levels. Our model is equally suitable for reduction reaction on Pt using first principles and kinetic calculations. Phys. Chem.
either const − ϕ or const − σ boundary conditions. Chem. Phys., 19, 4447 (2017).
11. M. Favaro, B. Jeong, Ph. N. Ross, J. Yano, Z. Hussain, Z. Liu, and E. J. Crumlin,
When the formalism is used with the parameters obtained from ab Unravelling the electrochemical double layer by direct probing of the solid/liquid
initio calculations and/or relevant experiments, e.g. EIS, it allows us interface. Nat. Commun, 7, 12695 (2016).
12. V. Tripkovic, M. E. Björketun, E. Skúlason, and J. Rossmeisl, Standard Hydrogen
to merge the continuum models self-consistently with AIMD. Using Electrode and Potential of Zero Charge in Density Functional Calculations. Phys.
continuum models, we perform an analysis of the thermodynamic Rev. B, 84, 115452 (2011).
consistency and the limiting characteristics of an AIMD setup in terms 13. J. Rossmeisl, E. Skúlason, E. Björketun, Marten, V. Tripkovic, and J. K. Nørskov,
of the cell sizes and the applied surface electron density/electrode Modeling the Electrified Solid - Liquid Interface. Chem. Phys. Lett., 466, 68 (2008).
14. M. E. Björketun, Z. Zeng, R. Ahmed, V. Tripkovic, K. S. Thygesen, and J. Rossmeisl,
potential that leads us to the following results. Avoiding Pitfalls in the Modeling of Electrochemical Interfaces. Chem. Phys. Lett.,
555, 145 (2013).
4) We provide estimates of AIMD cell sizes as functions of applied 15. M. Nielsen, M. E. Björketun, M. H. Hansen, and J. Rossmeisl, Towards First Princi-
potentials that meet the requirements of thermodynamic stability ples Modeling of Electrochemical Electrode-Electrolyte Interfaces. Surf. Sci., 631, 2
of biased interfaces and compatibility of the content of the elec- (2015).
16. F. Ambrosio, G. Miceli, and A. Pasquarello, Redox levels in aqueous solution: Effect
trolyte (ion concentration, number of solvent molecules) with the of van der Waals interactions and hybrid functionals. J. Chem. Phys., 143, 244508
target conditions. (2015).

Downloaded on 2017-11-09 to IP 190.111.30.254 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
Journal of The Electrochemical Society, 164 (11) E3438-E3447 (2017) E3447

17. N. Luque, S. Woelki, D. Henderson, and W. Schmickler, A model for the electrical 47. K. B. Oldham, A Gouy-Chapman-Stern model of the double layer at a (metal)/(ionic
double layer combining integral equation techniques with quantum density functional liquid) interface. J. Electroanal. Chem., 613, 131 (2008).
theory. Electrochimica Acta, 56, 7298 (2011). 48. D. -E. Jiang, D. Meng, and J. Wu, Density functional theory for differential capac-
18. J. Cheng and M. Sprik, Alignment of Electronic Energy Levels at Electrochemical itance of planar electric double layers in ionic liquids. Chem. Phys. Lett., 504, 153
Interfaces. Phys. Chem. Chem. Phys., 14, 11245 (2012). (2011).
19. I. Rungger, X. Chen, U. Schwingenschlögl, and S. Sanvito, Finite-bias Electronic 49. K. Letchworth-Weaver and T. A. Arias, Joint Density Functional Theory of the
Transport of Molecules in a Water Solution. Phys. Rev. B, 81, 235407 (2010). Electrode-Electrolyte Interface: Application to Fixed Electrode Potentials, Inter-
20. N. Kharche, J. T. Muckerman, and M. S. Hybertsen, First-Principles Approach to facial Capacitances, and Potentials of Zero Charge. Phys. Rev. B, 86, 075140
Calculating Energy Level Alignment at Aqueous Semiconductor Interfaces. Phys. (2012).
Rev. Lett., 113, 176802 (2014). 50. D. Gunceler, K. Letchworth-Weaver, R. Sundararaman, K. A. Schwarz, and
21. T. A. Pham, D. Lee, E. Schwegler, and G. Galli, Interfacial Effects on the Band Edges T. A. Arias, The importance of nonlinear fluid response in joint density-functional
of Functionalized Si Surfaces in Liquid Water. J. Am. Chem. Soc., 136, 17071 (2014). theory studies of battery systems. Modell. Simul. Mater. Sci. Eng., 21, 074005 (2013).
22. K. Leung, Predicting the Voltage Dependence of Interfacial Electrochemical Pro- 51. M. Otani and O. Sugino, First-principles Calculations of Charged Surfaces and
cesses at Lithium-Intercalated Graphite Edge Planes. Phys. Chem. Chem. Phys., 17, Interfaces: A Plane-Wave Nonrepeated Slab Approach. Phys. Rev. B, 73, 115407
1637 (2015). (2006).
23. R. Jinnouchi and A. B. Anderson, Electronic Structure Calculations of Liquid- 52. I. Hamada, O. Sugino, N. Bonnet, and M. Otani, Improved Modeling of Electrified
Solid Interfaces: Combination of Density Functional Theory and Modified Poisson- Interfaces Using the Effective Screening Medium Method. Phys. Rev. B, 88, 155427
Boltzmann Theory. Phys. Rev. B, 77, 245417 (2008). (2013).
24. C. Perego, M. Salvalaglio, and M. Parrinello, Molecular dynamics simulations of 53. V. Viswanathan, J. K. Norskov, A. Speidel, R. Scheffler, S. Gowda, and A. C. Luntz,
solutions at constant chemical potential. J. Chem. Phys., 142, 144113 (2015). LiO2 Kinetic Overpotentials: Tafel Plots from Experiment and First-Principles The-
25. G. Fisicaro, L. Genovese, O. Andreussi, N. Marzari, and S. Goedecker, A generalized ory. J. Phys. Chem. Lett., 4, 556 (2013).
Poisson and Poisson-Boltzmann solver for electrostatic environments. J. Chem. Phys., 54. J. Newman and K. E. Thomas-Alyea, Electrochemical Systems, 3rd Ed. Wiley &
144, 014103 (2016). Sons (2004).
26. J. -L. Fattebert and F. Gygi, First-principles molecular dynamics simulations in a 55. A. Latz and J. Zausch, Thermodynamic consistent transport theory of Li-ion batteries.
continuum solvent. Int. J. Quantum Chem., 93, 139 (2003). J. Power Sources, 196, 3296 (2011).
27. Y. -H. Fang, G. -F. Wei, and Z. -P. Liu, Theoretical modeling of electrode/electrolyte 56. S. Braun, C. Yada, and A. Latz, Thermodynamically Consistent Model for Space-
interface from first-principles periodic continuum solvation method. Catalysis Today, Charge-Layer Formation in a Solid Electrolyte. J. Phys. Chem. C, 119, 22281 (2015).
202, 98 (2013). 57. J. J. Bikerman, Structure and capacity of the Electrical Double Layer. Philos. Mag.,
28. A. A. Kornyshev and M. A. Vorotyntsev, Nonlocal electrostatic approach to the Dou- 33, 384 (1942).
ble Layer and adsorption at the Electrode-electrolyte Interface. Surf. Sci., 101, 23 58. A. N. Frumkin, Trans. Faraday Soc., 55, 156 (1955)
(1980). 59. A. N. Frumkin et al. Kinetics of Electrode Processes, Moscow State University Press,
29. J. Tomasi and M. Persico, Molecular Interactions in Solution: An Overview of Meth- Moscow (1952) (in Russian).
ods Based on Continuous Distributions of the Solvent. Chem. Rev., 94, 2027 (1994). 60. T. Biben, J. P. Hansen, and Y. Rosenfeld, Generic density functional for electric
30. A. Klamt and G. Schuurmann, COSMO: a new approach to dielectric screening in double layer in molecular solvent. Phys. Rev. E, 57, R3727 (1997).
solvents with explicit expressions for the screening energy and its gradient. J. Chem. 61. F. Booth, The Dielectric Constant of Water and the Saturation Effect. J. Chem. Phys.,
Soc., Perkin Trans., 2, 799 (1993). 19, 391 (1951).
31. C. J. Cramer and D. G. Truhlar, General parameterized SCF model for free energies 62. J. E. Curry and D. A. McQuarrie, On the Effect of Dielectric Saturation on the
of solvation in aqueous solution. J. Am. Chem. Soc., 113, 8305 (1991). Swelling of Clays. Langmuir, 8, 1026 (1992).
32. N. D. Mermin, Thermal Properties of the Inhomogeneous Electron Gas. Phys. Rev., 63. D. Larkin, K. L. Guyer, J. T. Hupp, and M. J. Weaver, Determination of specific
137, A1441 (1965). adsorption of some simple anions at a polycrystalline silver aqueous interface using
33. Y. Rosenfeld and N. W. Ashcroft, Theory of simple classical fluids: Universality in differential capacitance and kinetic probe techniques. J. Electroanal. Chem. Interfa-
the short-range structure. Phys. Rev. A, 20, 1208 (1979). cial Electrochem., 138, 401 (1982).
34. J. P. Hansen and I. R. McDonald, Theory of simple Liquids, 2nd ed. Academic Press, 64. L. Falciola, P. R. Mussini, S. Trasatti, and L. M. Doubova, Specific adsorption of
London (1986). bromide and iodide anions from nonaqueous solutions on controlled-surface poly-
35. S. K. Reed, O. J. Lanning, and P. A. Madden, Electrochemical Interface Between an crystalline silver electrodes. J. Electroanal. Chem., 593, 185 (2006).
Ionic Liquid and a Model Metallic Electrode. J. Chem. Phys., 126, 084704 (2007). 65. L. M. C. Pinto, E. Spohr, P. Quaino, E. Santos, and W. Schmickler, Why Silver
36. I. Borukhov, D. Andelman, and H. Orland, Steric Effects in Electrolytes: A Modified Deposition is so Fast: Solving the Enigma of Metal Deposition. Angew. Chem. Int.
Poisson-Boltzmann Equation. Phys. Rev. Lett., 79, 435 (1997). Ed., 52, 7883 (2013).
37. A. A. Kornyshev, Double-Layer in Ionic Liquids: Paradigm Change? J. Phys. Chem. 66. L. M. C. Pinto, P. Quaino, E. Santos, and W. Schmickler, On the Electrochemical
B, 111, 5545 (2007). Deposition and Dissolution of Divalent Metal Ions. Chem. Phys. Chem., 15, 132
38. M. V. Fedorov and A. A. Kornyshev, Towards understanding the structure and ca- (2014).
pacitance of electrical double layer in ionic liquids. Electrochimica Acta, 53, 6835 67. M. L. Foresti, M. Innocenti, F. Forni, and R. Guidelli, Electrosorption Valency and
(2008). Partial Charge Transfer in Halide and Sulfide Adsorption on Ag(111). Langmuir, 14,
39. M. Z. Bazant, M. S. Kilic, B. D. Storey, and A. Ajdari, Towards an understanding 7008 (1998).
of induced-charge electrokinetics at large applied voltages in concentrated solutions. 68. V. Marichev, Experimental Approach to the Anion Problem in DFT Calculation of
Adv. Colloid Interface Sci., 152, 48 (2009). the Partial Charge Transfer During Adsorption at Electrochemical Interfaces. Chem.
40. N. Georgi, A. Kornyshev, and M. Fedorov, The anatomy of the double layer and ca- Phys. Lett., 411, 434 (2005).
pacitance in ionic liquids with anisotropic ions: Electrostriction vs. lattice saturation. 69. V. Marichev, Partial Charge Transfer During Anion Adsorption: Methodological
J. Electroanal. Chem., 649, 261 (2010). Aspects. Surf. Sci. Rep., 56, 277 (2005).
41. A. Abrashkin, D. Andelman, and H. Orland, Dipolar Poisson-Boltzmann Equation: 70. R. Kubo, Statistical Mechanics, North-Holland Publishing Company, Amsterdam
Ions and Dipoles Close to Charge Interfaces. Phys. Rev. Lett., 99, 077801 (2007). (1965).
42. D. F. Calef and P. G. Wolynes, Classical Solvent Dynamics and Electron Transfer. 1. 71. D. Andelman, Handbook of Physics of Biological Systems, Vol. I., Chap. 12, p. 603,
Continuum Theory. J. Phys. Chem., 87, 3387 (1983). Amsterdam, Elsevier Science, (1994).
43. J. Forsman, A Simple Correlation-Corrected Poisson-Boltzmann Theory. J. Phys. 72. H. -J. Butt, K. Graf, and M. Kappl, Physics and Chemistry of Interfaces, John Wiley
Chem. B, 108, 9236 (2004). & Sons, Weinheim (2003).
44. M. Z. Bazant, B. D. Storey, and A. A. Kornyshev, Double Layer in Ionic Liquids: 73. N. Bonnet, O. Sugino, and M. Otani, Effect of thermal motion on catalytic activity
Overscreening versus Crowding. Phys. Rev. Lett., 106, 046102 (2011). of nanoparticles in polar solvent. J. Chem. Phys., 140, 044703 (2014).
45. M. Ma, S. Zhao, and Z. Xu, Investigation of Dielectric Decrement and Correlation Ef- 74. D. T. Limmer, C. Merlet, M. Salanne, D. Chandler, P. A. Madden, R. van Roij, and
fects on Electric Double-Layer Capacitance by Self-Consistent Field Model. Comput. B. Rotenberg, Charge Fluctuations in Nanoscale Capacitors. Phys. Rev. Lett., 111,
Phys. Commun., 20, 441 (2016). 106102 (2013).
46. G. Bolt, Analysis of the validity of the Gouy-Chapman theory of the electric double 75. Q. Wu and T. van Voorhis, Direct Calculations of electron Transfer Parameters
layer. J. Colloid Sci., 10, 206 (1955). through Constrained Density Functional Theory. J. Phys. Chem. A, 110, 9212 (2006).

Downloaded on 2017-11-09 to IP 190.111.30.254 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

You might also like