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

ARTICLE

pubs.acs.org/JPCC

Graphene-Based Supercapacitors: A Computer Simulation Study


Youngseon Shim,† YounJoon Jung,*,† and Hyung J. Kim*,‡,§,||

Department of Chemistry, Seoul National University, Seoul 151-747, Korea

Department of Chemistry, Carnegie Mellon University, Pittsburgh, Pennsylvania 15213, United States
§
School of Computational Sciences, Korea Institute for Advanced Study, Seoul 130-722, Korea

ABSTRACT: Energy density of supercapacitors based on a single-sheet graphene


electrode is studied via molecular dynamics (MD) computer simulations. Two electrolytes
of different types, pure 1-ethyl-3-methylimidazolium tetrafluoroborate (EMI+BF4) and
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

an 1.1 M solution of EMI+BF4 in acetonitrile, are considered as a prototypical room-


temperature ionic liquid (RTIL) and organic electrolyte, respectively. Structure of ions
near the electrode surface varies significantly with its charge density, especially in pure
RTIL. Specific capacitance normalized to the electrode surface area is found to be higher in
Downloaded via ZHEJIANG UNIV on October 9, 2023 at 12:27:02 (UTC).

EMI+BF4 than in acetonitrile solution by 5560%. This is due to strong screening of the
electrode charge by RTIL ions in the former. The RTIL screening behavior is found to be rather insensitive to temperature T. As a
result, the capacitance of supercapacitors based on pure EMI+BF4 decreases by less than 5% as T increases from 350 to 450 K. The
difference in size and shape between cations and anions and the resulting difference in their local charge distribution as counterions
near the electrified graphene surface yield cathodeanode asymmetry in the electrode potential in RTIL. As a consequence, specific
capacitance of the positively charged electrode is higher than that of the negatively charged electrode by more than 10%. A similar
degree of disparity in electrode capacitance is also found in acetonitrile solution because of its nonvanishing potential at zero charge.
Despite high viscosity and low ion diffusivity of EMI+BF4, its overall conductivity is comparable to that of the acetonitrile solution
thanks to its large number of charge carriers. The present study thus suggests that as a supercapacitor electrolyte, RTILs are
comparable in power density to organic electrolytes, while the former yield considerably better energy density than the latter at a
given cell voltage.

1. INTRODUCTION in activated carbon. Therefore, electrodes based on graphitic


Electric double layer capacitors (EDLCs), also referred to as materials hold immense potential to significantly improve both
supercapacitors, have emerged as an attractive energy storage power and energy densities of energy storage devices, such as
device to complement and even replace batteries in applications rechargeable lithium ion batteries12 and EDLCs,1318 and to
that require rapid energy draw, i.e., high power density.15 increase efficiency of energy conversion in, e.g., solar cells.19,20
Commonly used electrode materials for EDLCs are carbon- In this article, we investigate supercapacitors based on a single-
based, such as activated carbon, carbon nanofibers, and carbon sheet graphene electrode via MD simulations. While organic
nanotubes, because of their commercial availability at relatively electrolytes are commonly considered for supercapacitors, room-
low cost and good properties, such as large surface area, high temperature ionic liquids have also received significant attention,
capacitance, and long cycle life.4,6 However, the electrochemi- thanks to their attractive properties, in particular, wide electro-
cally available surface area of these materials during charging is chemical window, high ion density, nonvolatility, nonflammabil-
often limited because charge propagation via ion transport in and ity, and good thermal stability.6,21,22 For instance, RTILs’ large
out of small micropores is usually not efficient.4 Thus, in some electrochemical window allows high cell voltage that can lead to a
cases, small mesopores are incorporated in order to improve substantial increase in energy storage capacity of EDLCs, com-
charge transport in microporous materials.4 Low electrical con- pared to conventional electrolytes.13,23 Recently there have been
ductivity poses another difficulty in achieving high power density several computational efforts to obtain microscopic understand-
with activated carbon. ing of supercapacitors using RTILs as an electrolyte.2432 These
Two-dimensional graphene materials made of atomic carbon studies have shed light on, e.g., ion structure and distributions
sheets7 provide an exciting alternative to activated carbon for use near electrodes of various geometry and their influence on
in energy-related devices because of their excellent properties, capacitance. However, we are not aware of any prior analysis of
including large surface area, superior stiffness, high electrical RTILs vis-a-vis organic electrolytes in terms of their performance
conductivity, and chemical and thermal inertness.810 The large as supercapacitor electrolytes. In order to gain molecular-level
surface area of graphene, substantially higher than the BET insight into this important issue with attention focused on
surface area11 of activated carbon, can lead to a significant
increase in the energy storage capability of supercapacitors. In Received: April 13, 2011
addition, since its active surface forms effectively a large flat Revised: October 12, 2011
structure, ion transport in graphene is much more efficient than Published: October 12, 2011

r 2011 American Chemical Society 23574 dx.doi.org/10.1021/jp203458b | J. Phys. Chem. C 2011, 115, 23574–23583
The Journal of Physical Chemistry C ARTICLE

Figure 1. Model supercapacitor system employed in MD. A single-sheet graphene electrode is interfaced on both sides with an electrolyte (in the
present case, pure EMI+BF4). For convenience, the supercapacitor system is terminated on either side by introducing another graphene sheet as a
confining wall at a distance 6.43 nm from the central electrode. All graphene sheets, i.e., both the central electrode and confining walls, are rigid, flat, and
parallel to each other.

similarities and differences in their roles in supercapacitors, we walls, were modeled as a rigid and flat layer of 448 sp2-hybridized
embark on a systematic analysis of pure EMI+BF4 and an carbon atoms with dimensions 3.432  3.398 nm2.
acetonitrile solution of EMI+BF4 as prototypes of these two To describe the positively and negatively charged electrodes as
liquid classes. In this initial study, we consider only half-cell well as the fully discharged case, three different uniform surface
properties, i.e., electric potential and capacitance of a single charge densities, σS = ( 0.86e/nm2 and 0, were considered for
graphene electrode in an electrolyte. Analysis of full super- the central graphene via partial charge assignments to its C
capacitor cells in parallel plate electrode geometry will be atoms. The corresponding total surface charges were (10e for
reported elsewhere.33 the charged electrode cases. The two confining graphene walls
The outline of this paper is as follows: In section 2, we give a were electrically neutral regardless of the charge state of the
brief account of the models and methods employed in the central electrode. The numbers of RTIL cations and anions were
present simulation study. In section 3, we examine the structure adjusted to meet the charge neutrality of the entire simulation
of, and charge distributions resulting from, EMI+ and BF4 ions system. For instance, in the case of the positive electrode with
(and also acetonitrile molecules in the case of the organic 10e, we employed 507 cations and 517 anions for pure
electrolyte) and their screening behaviors in the presence of a EMI+BF4, while the acetonitrile solution contained 95 cations
uniformly charged graphene electrode. A detailed comparison of and 105 anions. The Lennard-Jones parameters employed for C
the EMI+BF4 and acetonitrile solution cases is made for the atoms of the graphene are ε = 43.2 K and σ = 0.34 nm.34 During
electric potential and capacitance of supercapacitors. For insight the simulations, the graphenes were held rigid with carbon
into power density, the electrolyte conductivity is also ana- carbon bond length lCC = 0.1415 nm.35 For RTIL, the flexible
lyzed via linear response theory. The influence of temperature on all-atom description of refs 28 and 36, based on the EMI+ model
the energy and power densities of the RTIL-based supercapac- in refs 37 and 38 and BF4 in refs 39 and 40, was used. For
itor is also considered there. Concluding remarks are offered in CH3CN, we used the fully flexible six-site description of ref 41.
section 4. Electronic polarizability42 for both the graphene and electrolytes
was ignored in our present study.43,44
We simulated molecular dynamics of the system in the
2. SIMULATION METHODS canonical ensemble at 350 K using the DL_POLY program.45
The simulation cell was composed of the supercapacitor system
Our model supercapacitor system is composed of a flat, single-
described above, placed in an orthorhombic box of 3.432  3.398
sheet graphene electrode that interfaces with an electrolyte on
 30.0 nm3. The long-range electrostatic interactions were
either side (Figure 1). In the discharged configuration of the computed via the Ewald method, resulting in essentially no
supercapacitor, the electrode is immersed in either a pure RTIL truncation of these interactions. The trajectories were integrated
consisting of 512 pairs of EMI+ and BF4 or an organic via the Verlet leapfrog algorithm using a time step of 1 fs.
electrolyte composed of 100 pairs of EMI+ and BF4 and 1024 Simulations were carried out with 10 ns equilibration, followed
CH3CN molecules. Hereafter, the former and latter systems will by a 10 ns trajectory from which ensemble averages were
be simply referred to as the RTIL and organic electrolyte computed. To examine the temperature effect on energy storage,
supercapacitors, respectively. We note that the organic electro- we also studied the RTIL supercapacitor at 450 K. Compared to
lyte considered here models a 1.1 M solution of EMI+BF4 with the 350 K case, the density of EMI+BF4 was reduced by 5% to
mole fraction 0.089 in acetonitrile. The combined electrode account for thermal expansion.46
electrolyte system was placed between two confining graphene
walls in the xy plane situated at z = (6.43 nm, such that the 3. RESULTS AND DISCUSSION
electrode surface positioned at z = 0 was parallel to the walls
(Figure 1). Outside of the confining walls was a vacuum. All Structure. We begin with electrolyte structures and their
graphene sheets, viz., the central electrode and two confining variations with the electrode surface charge density σS. We
23575 dx.doi.org/10.1021/jp203458b |J. Phys. Chem. C 2011, 115, 23574–23583
The Journal of Physical Chemistry C ARTICLE

Figure 3. Average number density nα(z) of (a) EMI+, (b) BF4 and (c)
CH3CN in the organic electrolyte supercapacitor using a 1.1 M
acetonitrile solution of EMI+BF4 at 350 K. (d) P(θ) of CH3CN in
the first peak ofR the first solvation layer, i.e., z < 0.4 nm, in (c) with
normalization dθP(θ) = 1. The notation is the same as in Figure 2. In
(d), θ is the angle between the normal to the electrode surface and the
molecular orientation of acetonitrile defined as the C-to-C direction
from its nitrile to methyl groups.

are not shown because it is invariant under z f z due to


symmetry of our supercapacitor system.
Figure 2. Average number density of (a) EMI+ and (b) BF4 ions in the
We first consider the results for the RTIL supercapacitor in
RTIL supercapacitor at 350 K. The corresponding results at 450 K are Figure 2. Regardless of the electrode charge density σS, electro-
shown in (c) and (d). z is the distance from the graphene electrode lyte structures show significant fluctuations along z. These
measured in nanometers. N represents the fully discharged case with an fluctuations, particularly strong in the vicinity of the electrode
electrically neutral electrode, while (+) and () denote, respectively, the surface, extend up to z ≈ 3 nm. Similar spatial fluctuations are
positively and negatively charged electrode configurations with the also present near the confining walls at z = (6.43 nm. This
surface charge density σS = (0.86e/nm2, where e is the elementary indicates that on average ions form layered structures near flat
charge. In the inset, the same results for z e 1.5 nm are shown for clear surfaces24,27,32,47 and that this ordering persists over a consider-
exposition of nα(z) near the electrode. In (e) and (f), nα(z) around the able distance from the surfaces. Nevertheless, the ion densities
electrically neutral and negative electrodes at 350 K is presented, become essentially constant in the region 3.5 nm j z j 4.5 nm,
respectively.
irrespective of σS. This suggests that RTIL there is nearly bulklike
and thus the presence of confining walls at z = (6.43 nm would
mention at the outset that the electrode surface is fully solvated
not affect our analysis based on the results obtained for |z| j
(viz., wetted) in the present study, irrespective of T, σS, or elec- 4 nm. It is also worthwhile to mention that RTIL distributions
trolytes we employed. For convenience, we employ a Cartesian
near the central electrode and near the confining walls differ even
coordinate system, where the graphene electrode surface spans
when the former is electrically neutral with σS = 0. This is due to
the xy plane (x0 < x < x0 and y0 < y < y0) with the origin at its the fact that both sides of the central electrode are exposed to
center and its normal defines the z direction (cf. Figure 1). We
RTIL, while only one side of the confining walls interfaces with
introduce the ion densities averaged over x and y
the electrolyte. Comparison of panels a and b and panels c and d
Z x Z y
n̅ α ðzÞ ¼ A0 1
0 0
dx0 dy0 nα ðx0 , y0 , zÞ of Figure 2 reveals that temperature has little influence on RTIL
ð1Þ distributions. Except for a very minor reduction in their structure,
x0 y0
nα(z) of both cations and anions remain essentially unchanged as
A0 ¼ 4x0 y0 T increases from 350 to 450 K.
With this in mind, we consider ion structures in the first
where A0 is the surface area of the graphene electrode with x0 = solvation shell (more accurately, solvation layer), which is
1.716 nm and y0 = 1.699 nm, nα(x,y,z) is the local number density defined as the region associated with the local density maximum
of solvent species α (α = EMI+, BF4, or CH3CN) at (x,y,z), and located closest to the electrode surface. In the case σS = 0, the first
the center-of-mass of ions and molecules is used to represent the peak in local densities of the cations and anions result at z = 0.37
position of α. The results for the average ion and molecule and 0.38 nm, respectively (Figure 2ad). The average ring
densities nα(z) are displayed in Figures 2 and 3. nα(z) for z < 0 orientation of EMI+ in the first solvation shell, i.e., z j 0.45 nm,
23576 dx.doi.org/10.1021/jp203458b |J. Phys. Chem. C 2011, 115, 23574–23583
The Journal of Physical Chemistry C ARTICLE

Figure 4. Snapshots of EMI+ and BF4 ions close to (a) electrically


neutral (σS = 0), (b) positively charged (σS = σ(+)), and (c) negatively Figure 5. Snapshots of ions and solvent molecules close to the electrode
charged (σS = σ()) graphene electrodes in the RTIL supercapacitor at with (a) σS = 0, (b) σS = σ(+), and (c) σS = σ() in the organic electrolyte
350 K. Only the ions with their center-of-mass located within 0.45 nm supercapacitor at 350 K. As in Figure 4, only the ions and molecules with
from the electrode surface are shown. their center-of-mass located within 0.45 nm of the electrode surface are
shown here.
is nearly parallel to the graphene surface, forming a π-stacking
structure (cf. Figure 4a below); the average angle between the We turn to the organic electrolyte supercapacitor case in
normal vector of the electrode surface, i.e., z axis, and that of EMI+ Figure 3. One of the most salient features is that for the neutral
rings there is found to be about 10. electrode the first solvation shells of ions are formed consider-
Also notable is that RTIL structures vary markedly with σS, ably further from the electrode than that of acetonitrile. As a
especially near the electrode surface. Electrostatic interactions consequence, there are virtually no ions present with their center-
with the surface charge induce reorganization of ions, which leads of-mass located within 0.5 nm of the neutral graphene surface,
to major structural enhancement and reduction for counterions whereas acetonitrile distribution exhibits a main peak at about z =
and co-ions, respectively, near the charged electrode, compared 0.36 nm. This is well illustrated by the snapshot in Figure 5a. Even
to the σS = 0 case. Therefore, around the positive electrode with when the graphene is charged with a significant surface charge
σS = 0.86e/nm2(σ(+)) that yields a total surface charge 10e, the density σ((), only a very small number of counterions are present
distribution of counterions BF4 exhibits a strong first solvation in the close neighborhood of the electrode, compared to CH3CN
shell peak at z = 0.37 nm. By contrast, near the negatively charged molecules there (Figure 5b,c). Thus the “covering” (i.e., wetting)
electrode with σS = 0.86e/nm2(σ()) and a total surface of flat graphene electrode surface is effected primarily via acetoni-
charge 10e, BF4 ions, which are now co-ions, are characterized trile molecules irrespective of σS, at least in the charge density
by the first solvation shell peak that is much lower in height and range we studied. This is attributed to the low ionic concentration
located further from the electrode (z = 0.54 nm) than the σS = of the electrolyte solution, which is 1.1 M in the present case, as
σ(+) case. EMI+ ions show essentially the same trend even though well as to the highly dipolar character of acetonitrile. We note that
the extent of their peak location variations is considerably less the ionic concentration in this range is typically used in real
than that of smaller anions. The first solvation shell peak supercapacitors.
positions of the former are z = 0.37 and 0.35 nm for the σ(+) In Figure 3d, we have analyzed the orientation of acetonitrile,
and σ() cases, respectively. Another noticeable feature is that, on in particular, its probability distribution P(θ), where θ is the
average, counterions and co-ions form alternating layers around angle between the z axis and the orientation of the CC bond of
the electrified graphene surfaces (Figure 2f).24,27,32,47 This is CH3CN in the direction from the nitrile to methyl groups. We
reminiscent of alternating concentric shell structures of counter- considered only the acetonitrile molecules with their center-of-
ions and co-ions around a central charge found in bulk RTILs.48,49 mass located in the region z < 0.4 nm, corresponding to the first
For further insight, snapshots of RTIL ions close to the peak of the first solvation shell in Figure 3c. The average θ value is
graphene electrode surface are shown in Figure 4. It clearly θ ≈ 70 in the presence of the neutral electrode, indicating that
illustrates the population changes of the cations and anions the nitrile C of acetonitrile is closer to the graphene surface than
with σS. It is noteworthy that there is a non-negligible number of its methyl C. Thus on average, the CC axis of CH3CN in the
co-ions present close to the electrifed graphene surfaces. We also z < 0.4 nm region is slightly tilted from the orientation parallel to
mention in passing that in the presence of a neutral electrode, the electrode surface, probably due to steric hindrance arising
cations are mainly surrounded by anions and vice versa in their from its methyl group. As expected, the tilted orientation is
immediate vicinity. favored even more near the positively charged electrode because
23577 dx.doi.org/10.1021/jp203458b |J. Phys. Chem. C 2011, 115, 23574–23583
The Journal of Physical Chemistry C ARTICLE

Figure 6. Average local charge density Fα(z) (units: e/nm3) of (a)


EMI+ and (b) BF4 in the RTIL supercapacitor at 350 K. In (c), the total Figure 7. Average local charge density Fα(z) (units: e/nm3) of (a)
charge density, F+(z) + F(z), is shown. EMI+, (b) BF4, and (c) CH3CN in the organic electrolyte super-
capacitor at 350 K. The total charge density, F+(z) + F(z) + FCH3CN(z),
is displayed in (d).
of Coulomb attraction between the electrode and the nitrile
group of CH3CN; the resulting average value for θ is about 60.
In the case of the negatively charged electrode with σ(),
however, Coulomb repulsion between the two pushes out the
nitrile group, resulting in the average orientation of acetonitrile
parallel to the electrode surface in the first peak of the first
solvation layer. This orientation change is accompanied by a
significant reduction in the first peak height.
Electric Potential. We proceed to the electric potential of the
supercapacitors. For convenience, we decompose the total
electric potential Φ into two components, Φσ and ΦIL, arising
from the graphene surface charge and RTIL ions, respectively. In
the case of the organic electrolyte supercapacitor, there is an
additional contribution ΦCH3CN from acetonitrile. We calculated Figure 8. ΦIL(z) (solid line) and Φσ(z) (dotted line) in (a) pure
ΦIL(z) by integrating the Poisson equation24,26 EMI+BF4 and (b) the acetonitrile solution at 350 K. In (b), the
Z z contribution to Φ from acetonitrile, ΦCH3CN(z), is shown in a dashed-
ΦIL ðzÞ ¼  4π ∑
α¼( 0
ðz  z0 ÞF α ðz0 Þ dz0 ð2Þ dotted line. For easy comparison of different components, the results for
Φσ(z) rather than Φσ(z) are exhibited, so that it has the same sign as
ΦIL(z). Units for electric potential: V.
Z x Z y
Fα ðzÞ ¼ A0 1 dx0 dy0 Fα ðx0 , y0 , zÞ
0 0

x0 y0 contributions to local charge density. While the degree of charge
separation is not as pronounced as BF4, the charge distribution
where Fα(x,y,z) is the local charge density arising from the of EMI+ has a similar extended character. Comparison with
atomic charge distribution of ionic species α, F α(z) is its average Figure 2 discloses that for z j 1 nm, charge densities of
at z, and α = ( denotes sum over ionic species. ΦCH3CN was electrolytes show much more rapid oscillations with z than their
obtained in a similar way. number densities. By contrast, the oscillatory behavior of the
The results for Fα(z) of the RTIL supercapacitor at 350 K are former nearly disappears for z J 1 nm, while structural order
exhibited in Figure 6. The corresponding results for 450 K are persists well beyond z = 1 nm. The finite character of atomic
nearly the same as 350 K (cf. Figure 2ad) and thus are not charge distributions of ions is also responsible for this interesting
shown there. Both cation and anion charge densities are char- difference between the electrolyte charge and number densities.
acterized by strong oscillations in z near the electrode surface. We note that the relaxation of charge oscillations over ∼1 nm
Surprisingly, the oscillations of the anion charge density are observed here is in reasonable accord with the long-range
accompanied by sign changes, especially in the presence of a charge screening length, 0.51 nm, obtained for dipolar solva-
positively charged or neutral electrode. This is attributed to the tion in other EMI+-based RTIL systems in different model
extended nature of ion charge distributions.28 Specifically, de- descriptions.48,49
spite its overall negative charge, the B and F sites of BF4 ions Several interesting aspects of Figure 6, in particular, rapid
have, respectively, partial positive (+1.1504e) and negative oscillations for z j 1 nm and flattening for z J 1 nm, are shared
(0.5376e) charges and thus play antagonistic roles in their by the corresponding cation and anion charge densities in the
23578 dx.doi.org/10.1021/jp203458b |J. Phys. Chem. C 2011, 115, 23574–23583
The Journal of Physical Chemistry C ARTICLE

Figure 10. Number of (a) cations and (b) anions in EMI+BF4,


Figure 9. Profile of total electric potential Φ(z) (in units of V) in the contained in the volume spanning from the graphene electrode surface
supercapacitor. Electrolytes employed are (a) pure EMI+BF4 at 350 K, to z in the RTIL supercapacitor at 350 K. Their difference N+(z) 
(b) pure EMI+BF4 at 450 K, and (c) 1.1 M acetonitrile solution of N(z) is presented in (c).
EMI+BF4 at 350 K. For clarity, Φ(z) at the electrode is set at 0 V for
all cases. For comparison, MD results for Φ(z) for a conventional
capacitor with pure acetonitrile used as a dielectric material at 350 K nα(z) and Nα(z), whereas their extended atomic charge distribu-
are displayed in (d). tion is employed for Fα(z) and QIL(z). Panels a and b of Figure 10
show that populations of counterions increase in a stepwise
acetonitrile solution in Figure 7. The charge density arising from manner with z, viz., a very abrupt and steep rise near z = 0.35 nm,
CH3CN in Figure 7c shows a similar oscillatory behavior. One followed by a relatively plateau behavior in the 0.4 nm j z j
important difference is that while F+(z) and F(z) approach a 0.7 nm region and a rapid increase for z J 0.75 nm. This is more
positive and a negative value as z increases, the acetonitrile charge evidence that counterions form a well-defined layered structure
density converges to 0. This is fully expected since acetonitrile close to the charged electrodes in the RTIL supercapacitor
molecules are dipolar, i.e., electrically neutral. Because of the low (cf. Figure 2).
ionic concentration, spatial fluctuations of the electrolyte charge The results for the difference between cation and anion
density are governed primarily by acetonitrile for z j 0.7 nm. populations, ΔN(z) = N+(z)  N(z), in the RTIL supercapacitor
The results for Φσ(z) and ΦIL(z) are presented in Figure 8, at 350 K are shown in Figure 10c. Pronounced oscillations of
while those for Φ(z) are in Figure 9. In the RTIL supercapacitor ΔN(z) up to z ≈ 3 nm in the presence of charged electrodes are the
case, ΦIL(z) tracks Φσ(z) very closely except for z j 0.3 nm. direct consequence of alternating layered structures of counterions
Also their magnitudes are vey similar in nearly the entire and co-ions mentioned above (cf. Figure 2f). Since the total surface
electrolyte region. Thus screening of the electrode surface charge charge of the graphene is (10e, the full screening occurs when
is effected essentially completely via RTIL ions located mainly in ΔN(z) = -5, viz., the total charge in the volume spanning from
the first solvation shell. The resulting total electric potential Φ(z) z to z, including the electrode charge, is 0. The results in
is very small in magnitude and varies little with z outside of the Figure 10c confirm that the complete screening obtains at z ≈
first solvation shell region (Figure 9a). In other words, the RTIL 0.35 nm for both the positive and negative electrodes. Interestingly,
supercapacitor behaves like an ideal electric double layer capa- counterions continue to accumulate more than co-ions in the RTIL
citor for z J 0.5 nm. As expected from nα(z) in Figure 2, a rise in as z increases beyond 0.35 nm until |ΔN(z)| reaches a maximum
T from 350 to 450 K has nearly a negligible effect on Φ(z) in of ∼15 around z = 0.4 nm. Therefore, the total charge in the
RTIL (Figure 9b). volume 0.4 nm < z < 0.4 nm, including the electrode charge,
For additional insight, we have analyzed the ion numbers and becomes positive in the case of the negatively charged electrode and
charges in the electrolyte volume extending from the electrode negative in the positively charged electrode case! This is responsible
surface to a surface at z (cf. eqs 1 and 2) for |QIL(z)| > 5e for 0.4 j z j0.5 nm in Figure 11a.
Z z
Returning to Figure 9a, we notice a significant gap in the
Nα ðzÞ ¼ A0 dz0 n̅ α ðz0 Þ ð3Þ
0 magnitude of Φ(z) between the positive and negative electrodes
for z J 1 nm. The magnitude of the potential drop ΔΦS(= Φ(z = 0)
Z z
 Φ(z ≈ 4 nm)) at the electrode with respect to the bulk
QIL ðzÞ ¼ A0 ∑
α¼( 0
dz0 Fα ðz0 Þ
electrolyte shows a sizable difference between the two. MD yields
ΔΦ(+) = 1.12 V with σS = σ(+) and ΔΦ() = 1.42 V with σ() at
NCH3CN(z) and QCH3CN(z) are defined similarly. The reader is 350 K, while PZC (potential of zero charge), i.e., the potential drop
reminded that the position of the center-of-mass of α is used for for the neutral electrode, is 0.07 V (Table 1). The corresponding
23579 dx.doi.org/10.1021/jp203458b |J. Phys. Chem. C 2011, 115, 23574–23583
The Journal of Physical Chemistry C ARTICLE

Figure 11. (a) QIL(z) in the RTIL supercapacitor at 350 K. (b) QIL(z) Figure 12. Nα(z) of (a) EMI+, (b) BF4 and (c) acetonitrile in the
and QCH3CN(z) and (c) their sum in the organic electrolyte super- organic electrolyte supercapacitor. In (d), the corresponding difference
capacitor. In (b), QIL(z) and QCH3CN(z) are plotted in the solid and in Nα(z) between the cations and anions is shown.
dotted lines, respectively.
electrode charges by ions in the organic electrolyte, compared to
the RTIL. The low ionic concentration of the former is directly
Table 1. Electrode Potential ΔΦS and Specific Capacitance cS responsible for the weak screening. It is also noteworthy that
acetonitrile makes an important contribution to Φ for small z.
solvent σS (e/nm2) ΔΦS (V)a cS (μF/cm2)
Through alignment of its dipole moment (cf. Figures 7c,d and
EMI+BF4 at 350 K 0.86 1.12 5.78 11b,c), acetonitrile mainly governs the electrolyte charge density
0.86 1.42 5.09 and thus the shielding in the region z j 0.7 nm as noted above
EMI+BF4 at 450 K 0.86 1.14 5.54 and reduces |ΔΦS| by 24 V. It should nonetheless be stressed
0.86 1.48 4.98
that its cumulative charge QCH3CN(z) vanishes as z increases
beyond ∼1 nm because acetonitrile molecules are electrically
CH3CN/EMI+BF4 at 350 K 0.86 2.00 3.62
neutral. This means that the screening of the electrode charge at
0.86 2.00 3.28
large distances is governed by ions as expected from Figure 8b.
a
ΔΦS at PZC is 0.07 and 0.10 V for the RTIL and organic electrolyte According to Figure 12d, full screening occurs close to z = 1 nm
supercapacitors, respectively.
for the organic electrolyte supercapacitor.
Another difference from the RTIL case is that the positive and
values at 450 K are nearly the same, viz., 1.14, 1.48, and 0.10 V, negative electrodes of the organic electrolyte supercapacitor are
respectively. The difference in size and molecular structure characterized by the same |ΔΦS| (=2 V) (Figure 9c) despite their
between the cations and anions is believed to be mainly respon- substantial difference in ion charge density (Figure 7a,b) and
sible for the |ΔΦ(+)|  |ΔΦ()| disparity, which is about 0.3 V populations (Figure 12a,b) at small z. Analogous to the RTIL
irrespective of T. Smaller BF4 anions shield the positively charged capacitor, however, ΔΦS of the organic electrolyte supercapa-
electrode more efficiently via their F sites with a partial negative citor does not vanish at PZC. Specifically, ΔΦS = 0.1 V at PZC
charge (0.5376e) than bulkier EMI+ cations screen the nega- with respective contributions of 0.3 and 0.2 V from ions and
tively charged electrode. This is manifested as a large peak of acetonitrile. By contrast, PZC is 0 in neat acetonitrile, indicating
height 13e in QIL(z) at z ≈ 0.2 nm around the positively charged that the nonvanishing PZC in the organic electrolyte super-
electrode in Figure 11a. By contrast, the corresponding value capacitor is induced by the ions. We ascribe this to a significant
around the negatively charged electrode is QIL(z) = 3.5e. This difference between EMI+ and BF4 in nα(z) (Figure 3a,b) and
enables anions to reduce the electric field at short distances arising Fα(z) (Figure 7a,b) in the acetonitrile solution, arising from their
from the electrode surface charge better than cations. Since ΔΦS is differing size and shape.
given by the integration of the electric field, better reduction of the For completeness, we have also analyzed a conventional
electric field at short distances yields smaller magnitude for ΔΦS. capacitor, consisting of a single-sheet graphene electrode and a
The results for the organic electrolyte supercapacitor in dielectric material modeled as acetonitrile. Its MD result for
Figures 8b and 9c exhibit interesting differences from those for Φ(z) in the presence of a positive electrode charged with σ(+) is
the RTIL supercapacitor. To be specific, while ΦIL(z) tracks shown in Figure 9d. One prominent feature is that Φ(z) shows a
Φσ(z) analogous to the RTIL supercapacitor case, the differ- bimodal character, i.e., a rapid drop near the electrode and a
ence in their magnitude in the organic electrolyte supercapacitor, linear decrease with z in the region z J 0.5 nm. We notice that
which is about 46 V for z J 0.5 nm, is much bigger than the the Φ(z) behavior for z j 0.5 nm is very similar, though lesser
RTIL supercapacitor. This exposes the weak screening of the in extent, to that of the organic electrolyte supercapacitor in
23580 dx.doi.org/10.1021/jp203458b |J. Phys. Chem. C 2011, 115, 23574–23583
The Journal of Physical Chemistry C ARTICLE

Figure 9b. This provides additional evidence that acetonitrile


plays a crucial role near the electrode in the organic supercapacitor.
Linear decay in Φ(z) for z J 0.5 nm in the conventional capacitor,
on the other hand, is a manifestation of incomplete screening of
the electrode charge due to the dipolar character of the solvent,
revealing its clear distinction from EDLCs.
Specific Capacitance. We turn to specific capacitance of the
RTIL and organic electrolyte supercapacitors, normalized to the
total electrode surface area exposed to the electrolyte
 
1 ∂qS  ∂σS  σS
c ¼
S
S  ¼ 1=2 S  ≈ ð4Þ Figure 13. Time correlation function of the ionic current normal to the
2A0 ∂ΔΦ ∂ΔΦ 2ΔΔΦS electrode surface in (a) RTIL and (b) organic electrolyte supercapaci-
PZC PZC
tors at 350 K.

ΔΔΦS ¼ ΔΦS  ΔΦS ðat PZCÞ advantage over organic electrolytes in energy density; viz., use of
RTILs would improve the capacitance of the supercapacitors
where qS (=σSA0) is the total electrode charge, A0 is the graphene significantly, compared to organic electrolytes, even at the same
surface area (eq 1), and cS is evaluated at PZC. The factor 2 in the cell potential. Another important finding is that specific capaci-
denominator of eq 4 arises because both sides of the graphene tance of RTIL supercapacitors is rather insensitive to tempera-
electrode interface with the electrolyte (cf. Figure 1). ture, provided that the electrode surface is fully wetted by ions. As
The results for cS are compiled in Table 1. For the RTIL such, RTILs would provide a promising class of electrolytes for
supercapacitor, c(+) and c() are, respectively, 5.78 and 5.09 μF/ efficient energy storage in a broad temperature range.
cm2 at 350 K and 5.54 and 4.98 μF/cm2 at 450 K. A few Ion Conductivity. To obtain insight into power density of the
comments are in order here: First, since the RTIL distribution RTIL and organic electrolyte supercapacitors, we briefly consider
(Figure 2ad) and thus potential drop hardly change with T as electrolyte conductivity in the direction perpendicular to the
noted several times above, so does the electrode capacitance. For electrode surface. In the GreenKubo formulation based on
the present supercapacitor system based on EMI+BF4, an linear response theory, ion conductivity along the z direction, i.e.,
increase in temperature by 100 K leads to a reduction of specific normal to the electrode surface, is related to the time correlation
capacitance by only j4%. Second, the aforementioned difference function of collective ionic current Jz(t) as
in solvation of the positively and negatively charged electrodes,
1 Z ∞
induced by the difference in size and shape of cations and anions, σ GK ¼ dt CJJ ðtÞ ð5Þ
and resulting discrepancy in the magnitude of their ΔΦS yield VkB T 0
non-negligible cathode-anode asymmetry50,51 in cS. Our finding
that c(+) > c() is in good agreement with prior theory50,51 and CJJ ðtÞ ¼ ÆJz ð0ÞJz ðtÞæ
MD study,28,27,32 but is at variance with ref 26, where the
opposite result (c(+) < c()) was obtained for a similar system N
with a lower electrode surface charge density, 0.5125e/nm2. Jz ðtÞ ¼ ∑ qi vz, i ðtÞ
i¼1
Third, despite the complete neglect of electronic polarizability
in our model description for both the electrode and electrolytes, where V is the volume of the system, kB is Boltzmann’s constant,
the MD results we obtained for our graphene-based supercapacitors qi and vz,i are the charge and z-component of center-of-mass
are comparable to experimental results for closely related systems. velocity of ith ion and Æ...æ represents an equilibrium ensemble
For example, the specific capacitance of a highly oriented pyrolytic average. For comparison, we also consider the conductivity
graphite (HOPG) electrode in N,N-diethyl-N-methyl-N-(2-meth- estimated via the NernstEinstein equation
oxyethyl)ammonium bis(trifluoromethanesulfonyl)imide was found
to be 2.25 μF/cm2.15

1
σ NE ¼ nα q2α DGK, α ð6Þ
As in the RTIL supercapacitor case, the organic electrolyte kB T α ¼ (
supercapacitor also exhibits cathodeanode asymmetry in capa-
citance. The simulation results for T = 350 K are c(+) = 3.62 μF/ Z ∞
cm2 and c() = 3.28 μF/cm2. The nonvanishing PZC discussed DGK, α ¼ Nα1 ∑
i∈α 0
dt Ævz, i ð0Þvz, i ðtÞæ
above is responsible for this asymmetry in the organic electrolyte
supercapacitor. Interestingly, the relative difference in electrode where qα, nα, and Nα are, respectively, the charge, number
capacitance is ∼10% for both capacitors. Nevertheless, it should be density, and total number of ions of species α, i ∈ α means
noticed that specific capacitance of the organic electrolyte super- sum over all ions of species α, and DGK,α is their translational
capacitors is smaller than that of the RTIL supercapacitors by diffusion coefficient along z. We note that σGK reduces to σNE if
1.82.2 μF/cm2. This corresponds to a ∼35% diminution in the the contribution of cross correlation in CJJ(t) is ignored com-
acetonitrile solution, compared to the pure EMI+BF4 case. As pletely in eq 5.
analyzed above, this decrease is attributed primarily to the weak The results for CJJ(t) are displayed in Figure 13 and those for
ionic screening of the electrode charges in the acetonitrile solution. the conductivities and diffusion coefficients are compiled in
The result that RTIL supercapacitors exceed organic electro- Table 2. We first consider the RTIL supercapacitor case.
lyte supercapacitors in electrode capacitance is one of the major Figure 13a shows that its CJJ(t) varies little with the electrode
findings of the present work. This means that in addition to their charge density σS. Though not presented there, we mention that
wide electrochemical window, RTILs offer another important the temporal behaviors of CJJ(t) at 450 K, including the
23581 dx.doi.org/10.1021/jp203458b |J. Phys. Chem. C 2011, 115, 23574–23583
The Journal of Physical Chemistry C ARTICLE

Table 2. Translational Diffusion Coefficient and Ion Conductivity Normal to the Electrode Surfacea
solvent σS (e/nm2) DGK,+ DGK, σNE σGK

EMI+BF4 0.86 0.20 (0.52) 0.21 (0.44) 0.59 (1.0) 0.27 (0.29)
0.0 0.20 (0.47) 0.21 (0.42) 0.59 (0.96) 0.27 (0.28)
 0.86 0.20 (0.47) 0.21 (0.41) 0.59 (0.95) 0.26 (0.27)
CH3CN/EMI+BF4 0.86 1.3 1.4 0.73 0.32
0.0 1.3 1.3 0.72 0.33
 0.86 1.3 1.3 0.72 0.32
a
The system temperature is T = 350 K. Results for the RTIL supercapacitor at 450 K are given in parentheses. Diffusion coefficients and conductivities
are measured in units of 109 m2 s1 and S/m, respectively.

frequency and relative amplitude of oscillations, are very close to electrode was found to be considerably larger than that of the
those at 350 K. Thus regardless of σS and T we considered, the negatively charged electrode in both EMI+BF4 and acetonitrile
ion conductivity along z determined via eq 5 remains largely solution. This cathodeanode asymmetry50,51 is ascribed to
unchanged (σGK = 0.260.29 S/m). DGK,α, on the other hand, differing screening effciency arising from the difference in size
shows a marked increase with T. This difference in the T and molecular structure between the cations and anions.
dependence arises primarily from the presence of the T1 factor To gain insight into power density, we analyzed ion conduc-
in σGK in eq 5, which is absent in DGK,α. Another noteworthy tivity. It was found that ion conductivity in the direction normal
aspect is that σNE obtained with the neglect of the cross to the electrode surface is larger in the acetonitrile solution than
correlation in CJJ(t) overestimates the actual conductivity, in pure EMI+BF4 but only by ∼20%. This result suggests that all
viz., σGK, by more than a factor of 2. This reveals an important other things being equal, the two electrolytes would be largely
role played by the cross correlation in the determination of comparable in power density.
conductivity; it cancels a significant part of the contribution The effect of temperature on the supercapacitor performance
from the self-correlation component of CJJ(t) and thus re- was also examined. Interestingly and importantly, specific capa-
duces ion conductivity substantially. It is worthwhile to note citance and ion conductivity of the RTIL supercapacitors were
that the relaxation behavior of CJJ(t) in Figure 13a is very found to vary little with T. This finding, together with the results
similar to that in pure EMI+PF6 studied in ref 52 even though summarized above in this section, indicates that RTILs are a
anionic species are different. viable candidate to replace conventional electrolytes in energy
CJJ(t) of the acetonitrile solution in Figure 13b shows an storage devices, which has promising potential for a good and
interesting departure from that of RTIL. Specifically, relaxation reliable performance in energy storage and delivery over a
dynamics of the former become decelerated and its librational significant temperature range. It would thus be very worthwhile
character attenuated with increasing t, compared to the latter. in the future to extend the present study to other supercapacitor
While this kind of relaxation behavior sometimes leads to systems composed of differing RTILs and/or electrodes to find
“superdiffusion”,53,54 we found that when integrated over t, CJJ(t) optimal conditions, configurations, and combinations of electro-
for the present organic electrolyte supercapacitor system yields a lytes and electrode materials for efficient energy storage.
proper plateau behavior and thus well-defined conductivity. As in
the RTIL supercapacitor case, σGK is considerably smaller than ’ AUTHOR INFORMATION
σNE, confirming the importance of the cross-correlation effect.
Comparison of the RTIL and organic electrolyte supercapa- Corresponding Author
citor results at 350 K shows a couple of noteworthy features. *E-mail: yjjung@snu.ac.kr; hjkim@cmu.edu.
First, the ion translational diffusion coefficients along z in Present Addresses
)

acetonitrile solution are more than 6-fold greater than those in Carnegie Mellon University.
EMI+BF4 because the latter is considerably more viscous than
the former. Second, the conductivities of the RTIL and organic
electrolyte are comparable. This is due to the high ion density, ’ ACKNOWLEDGMENT
i.e., large number of charge carriers, in the former, which com-
This work was supported by the National Research Founda-
pensates for low mobility of its individual charge carriers.
tion of Korea (NRF) grants funded by the Korean Government
(MEST) (Nos. 2011-0001212 and 2011-0018038). Y.S. ac-
knowledges the financial support from the BK 21 Program
4. CONCLUDING REMARKS of Korea.
In this article, we have studied supercapacitors based on a
single-sheet graphene electrode via MD. Two different electro- ’ REFERENCES
lytes, i.e., neat EMI+BF4 and a 1.1 M acetonitrile solution of
EMI+BF4, were considered. One of our key findings is that (1) Conway, B. E. Electrochemical Supercapacitors: Scientific Funda-
mentals and Technological Applications; Plenum: New York, 1999.
specific capacitance of the electrode normalized to its surface area
(2) K€otz, R.; Carlen, M. Electrochim. Acta 2000, 45, 2483–2498.
is 5560% higher when pure EMI+BF4 is employed as an (3) Pandolfo, A. G.; Hollenkamp, A. F. J. Power Sources 2006,
electrolyte than the acetonitrile solution. Strong and effective 157, 11–27.
screening of electrode charges by RTIL ions close to the (4) Frackowiak, E. Phys. Chem. Chem. Phys. 2007, 9, 1774–1785.
electrode is mainly responsible for high capacitance of the RTIL (5) Abru~ na, H. D.; Kiya, Y.; Henderson, J. C. Phys. Today 2008,
supercapacitor. Specific capacitance of the positively charged 61, 43–47.

23582 dx.doi.org/10.1021/jp203458b |J. Phys. Chem. C 2011, 115, 23574–23583


The Journal of Physical Chemistry C ARTICLE

(6) Simon, P.; Gogotsi, Y. Nat. Mater. 2008, 7, 845–854. (42) Schmickler, W. Chem. Rev. 1996, 96, 3177–3200.
(7) Novoselov, K. S.; Geim, A. K.; Morozov, S. V.; Jiang, D.; Zhang, Y.; (43) Yan, T.; Burnham, C. J.; Del Popolo, M. G.; Voth, G. A. J. Phys.
Dubonos, S. V.; Grigorieva, I. V.; Firsov, A. A. Science 2004, 306, 666–669. Chem. B 2004, 108, 11877–11881.
(8) Meyer, J. C.; Geim, A. K.; Katsnelson, M. I.; Novoselov, K. S.; (44) For the effect of solute polarizability in RTIL, see, e.g.:Jeong,
Booth, T. J.; Roth, S. Nature 2007, 446, 60–63. D.; Shim, Y.; Choi, M. Y.; Kim, H. J. J. Phys. Chem. B 2007, 111,
(9) Gomez-Navarro, C.; Weitz, R. T.; Bittner, A. M.; Scolari, M.; 4920–4925.
Mews, A.; Burghard, M.; Kern, K. Nano Lett. 2007, 7, 3499–3503. (45) Forester, T. R.; Smith, W. DL_POLY user manual; CCLRC,
(10) Allen, M. J.; Tung, V. C.; Kaner, R. B. Chem. Rev. 2010, Daresbury Laboratory: Daresbury, Warrington, U.K., 2001.
110, 132–145. (46) Taguchi, R.; Machida, H.; Sato, Y.; Smith, R. L. J. Chem. Eng.
(11) Brunauer, S.; Emmett, P. H.; Teller, E. J. Am. Chem. Soc. 1938, Data 2009, 54, 22–27.
60, 309–319. (47) Pinilla, C.; Del Popolo, M. G.; Lynden-Bell, R. M.; Kohanoff, J.
(12) Ishikawa, M.; Sugimoto, T.; Kikuta, M.; Ishiko, E.; Kono, M. J. Phys. Chem. B 2005, 109, 17922–17927.
J. Power Sources 2006, 162, 658–662. (48) Shim, Y.; Choi, M. Y.; Kim, H. J. J. Chem. Phys. 2005,
(13) Vivekchand, S. R. C.; Rout, C. S.; Subrahmanyam, K. S.; 122, 044510.
Govindaraj, A.; Rao, C. N. R. J. Chem. Sci. 2008, 120, 9–13. (49) Shim, Y.; Jeong, D.; Manjari, S. R.; Choi, M. Y.; Kim, H. J. Acc.
(14) Stroller, M. D.; Park, S.; Zhu, Y.; An, J.; Ruoff, R. S. Nano Lett. Chem. Res. 2007, 40, 1130–1137.
2008, 8, 3498–3502. (50) Fedorov, M. V.; Kornyshev, A. A. J. Phys. Chem. B 2008,
(15) Islam, M. M.; Alam, M. T.; Okajima, T.; Ohsaka, T. J. Phys. 112, 11868–11872.
Chem. C 2009, 113, 3386–3389. (51) Oldham, K. B. J. Electroanal. Chem. 2008, 613, 131–138.
(16) Zhu, Y.; Stoller, M. D.; Cai, W.; Velamakanni, V.; Piner, R. D.; (52) Shim, Y.; Kim, H. J. J. Phys. Chem. B 2008, 112, 11028–11038.
Chen, D.; Ruoff, R. S. ACS Nano 2010, 4, 1227–1233. (53) Liu, B.; Goree, J. Phys. Rev. E 2007, 75, 016405.
(17) Liu, C.; Yu, Z.; Neff, D.; Zhamu, A.; Jang, B. Z. Nano Lett. 2010, (54) Ott, T.; Bonitz, M.; Donk o, Z.; Hartmann, P. Phys. Rev. E 2008,
10, 4863–4868. 78, 026409.
(18) Kim, T. Y.; Lee, H. W.; Stoller, M.; Dreyer, D. R.; Bielawski,
C. W.; Ruoff, R. S.; Suh, K. S. ACS Nano 2011, 5, 436–442.
(19) Wang, X.; Zhi, L. J.; M€ullen, K. Nano Lett. 2008, 8, 323–327.
(20) Wang, X.; Zhi, L. J.; Tsao, N.; Tomovic, Z.; Li, J. L.; M€ullen, K.
Angew. Chem., Int. Ed. 2008, 47, 2990–2992.
(21) Ue, M. In Electrochemical Aspects of Ionic Liquids; Ohno, H., Ed.;
John Wiley & Sons: Hoboken, NJ, 2005; Chapter 17.
(22) Armand, M.; Endres, F.; MacFarlane, D. R.; Ohno, H.; Scrosati,
B. Nat. Mater. 2009, 8, 621–629.
(23) Balducci, A.; Dugas, R.; Taberna, P. L.; Simon, P.; Plee, P. D.;
Mastragonstino, M.; Passerini, S. J. Power Sources 2007, 165, 922–927.
(24) Pinilla, C.; Del Popolo, M. G.; Kohanoff, J.; Lynden-Bell, R. M.
J. Phys. Chem. B 2007, 111, 4877–4884.
(25) Yang, L.; Fishbine, B. H.; Migliori, A.; Pratt, L. R. J. Am. Chem.
Soc. 2009, 131, 12373–12376.
(26) Kislenko, S. A.; Samoylov, I. S.; Amirov, R. H. Phys. Chem.
Chem. Phys. 2009, 11, 5584–5590.
(27) Feng, G.; Zhang, J. S.; Qiao, R. J. Phys. Chem. C 2009, 113,
4549–4559.
(28) Shim, Y.; Kim, H. J. ACS Nano 2010, 4, 2345–2355.
(29) Lauw, Y.; Horne, M. D.; Rodopoulos, T.; Nelson, A.; Leermakers,
F. A. M. Phys. Rev. Lett. 2009, 103, 117801.
(30) Lauw, Y.; Horne, M. D.; Rodopoulos, T.; Nelson, A.; Leermakers,
F. A. M. J. Phys. Chem. B 2010, 114, 11149–11154.
(31) Trulsson, M.; Algotsson, J.; Forsman, J.; Woodward, C. E.
J. Phys. Chem. Lett. 2010, 1, 1191–1195.
(32) Vatamanu, J.; Borodin, O.; Smith, G. D. J. Am. Chem. Soc. 2010,
132, 14825–14833.
(33) Shim, Y.; Kim, H. J.; Jung, Y. Faraday Discuss. DOI: 10.1039/
C1FD00086A.
(34) Hummer, G.; Rasaiah, J. C.; Noworyta, J. P. Nature 2001,
414, 188–190.
(35) Odom, T. W.; Huang, J.-L.; Kim, P.; Lieber, C. M. Nature 1998,
391, 62–64.
(36) Shim, Y.; Kim, H. J. ACS Nano 2009, 3, 1693–1702.
(37) Canongia Lopes, J. N.; Deschamps, J.; Padua, A. A. H. J. Phys.
Chem. B 2004, 108, 2038–2047.
(38) Canongia Lopes, J. N.; Deschamps, J.; Padua, A. A. H. J. Phys.
Chem. B 2004, 108, 11250.
(39) de Andrade, J.; B€oes, E. S.; Stassen, H. J. Phys. Chem. B 2002,
106, 13344–13351.
(40) Wu, X.; Huang, S.; Wang., W. Phys. Chem. Chem. Phys. 2005,
7, 2771–2779.
(41) Nikitin, A. M.; Lyubartsev, A. P. J. Comput. Chem. 2007, 28,
2020–2026.

23583 dx.doi.org/10.1021/jp203458b |J. Phys. Chem. C 2011, 115, 23574–23583

You might also like