ARTICLE

On the Dynamics of Charging

in Nanoporous Carbon-Based
Supercapacitors
Clarisse Péan,†,‡,§ Céline Merlet,†,§ Benjamin Rotenberg,†,§ Paul Anthony Madden,^ Pierre-Louis Taberna,‡,§
Barbara Daffos,‡,§ Mathieu Salanne,†,§,* and Patrice Simon‡,§
†
Sorbonne Universités, UPMC Univ Paris 06, UMR 8234, PHENIX, F-75005 Paris, France, ‡CIRIMAT, UMR CNRS 5085, Université Paul Sabatier, Bat. 2R1, 118 route de
Narbonne, 31062, Toulouse Cedex 9, France, §Réseau sur le Stockage Electrochimique de l'Energie (RS2E), FR CNRS 3459, 80039 Amiens Cedex, France, and
^
Department of Materials, University of Oxford, Parks Road, Oxford OX1 3PH, United Kingdom

ABSTRACT Supercapacitors are electricity storage systems with high power performances. Their short
charge/discharge times are due to fast adsorption/desorption rates for the ions of the electrolyte on the
electrode surface. Nanoporous carbon electrodes, which give larger capacitances than simpler geometries,
might be expected to show poorer power performances because of the longer times taken by the ions to
access the electrode interior. Experiments do not show such trends, however, and this remains to be
explained at the molecular scale. Here we show that carbide-derived carbons exhibit heterogeneous and fast charging dynamics. We perform molecular
dynamics simulations, with realistically modeled nanoporous electrodes and an ionic liquid electrolyte, in which the system, originally at equilibrium in the
uncharged state, is suddenly perturbed by the application of an electric potential difference between the electrodes. The electrodes respond by charging
progressively from the interface to the bulk as ions are exchanged between the nanopores and the electrolyte region. The simulation results are then
injected into an equivalent circuit model, which allows us to calculate charging times for macroscopic-scale devices.

KEYWORDS: supercapacitor . room temperature ionic liquid . nanoporous carbon . molecular dynamics . power density .
transmission line model

S
upercapacitors are currently used in So far most of the simulation studies were
specific applications for which high dedicated to the understanding of the
power is required for a short time.1 origin of the “anomalous” increase of the
The electricity storage is achieved by (non- capacitance inside nanoporous carbon
faradaic) adsorption of ions from an electro- electrodes,14
21 without addressing the dy-
lyte onto electrodes with high-surface area; namic aspects. Nevertheless, it is the trans-
this surface-based mechanism, which in- port of the ions inside the pores which will
volves no redox reaction, is at the origin of control the power density of a device. It is
high power.2,3 As devices, supercapacitors thus of primary importance to well charac-
complement batteries, which have slower terize this transport on the molecular scale.
discharge time but which exhibit much great- A few studies have investigated the relaxa-
er stored energy densities.4 Many applications tion dynamics of ions close to electrified
therefore use this complementarity and com- planar surfaces.22
24 In a study directed
bine a supercapacitor with a Li-ion battery. at the nanoporous carbons, Kondrat and
Nevertheless, there is a tremendous need to Kornyshev have recently observed by using
increase the stored energy densities of super- a mean-field model that the charging of
capacitors, and various routes are currently initially empty pores proceeds in a front-like
explored.5
10 Recent advances were made way, while that of already filled pores is
when it was shown that using electrodes diffusive.25 * Address correspondence to
[email protected].
made of carbons with well-controlled nano- There are two main methods for model-
porosity could lead to large and unexpected ing the electrodes of a supercapacitors. The Received for review November 10, 2013
enhancements of the capacitance.11
13 How- first, simpler one consists in assigning con- and accepted January 13, 2014.
ever, these improvements must be made stant charges to every atom of an electrode,
Published online
while keeping good power performance. while the second one consists in applying a 10.1021/nn4058243
Energy density is proportional to the ca- constant potential and allowing the carbon
pacitance, which is an equilibrium property. charges to fluctuate in response to the local C XXXX American Chemical Society

PÉAN ET AL. VOL. XXX ’ NO. XX ’ 000–000 ’ XXXX A

www.acsnano.org
 ARTICLE
Figure 1. (Top) The simulation cell consists of a BMI-PF6 ionic liquid electrolyte surrounded by two porous electrodes
(CDC-1200) held at constant electrical potentials (blue, C atoms; red, the three sites of BMIþ; and green, PF6
ions). (Bottom)
Equivalent electric circuit used to model our simulation setup. Rbulk is the resistance of the electrolyte in the bulk region, Rl is
the resistance of the electrolyte adsorbed in an electrode slice, C1 and C2 are the capacitances of the two slices.

structure of the electrolyte.26
28 It was shown that the 9.0, and 9.0 Å (with distributions ranging from 3 to
latter approach is much more realistic, and should 15
18 Å).30 In addition, the structure becomes more
always be used when studying the dynamics. Constant and more disordered when passing from CDC-1200 to
charge simulations should be kept for equilibration CDC-950 and then to CDC-800. Here after equilibration
purposes only, because imposing a sudden surface of the system at 0 V, we suddenly apply an electric
charge leads to an extremely large current at t = 0, and potential difference (a value of 1 V was used unless
consequently to unphysically fast dynamics and a large otherwise stated) and follow the charging dynamics of
increase of the temperature of the system.29 Here we the electrodes.
perform constant potential molecular dynamics simu- We have previously studied18,28 the equilibrium
lations involving realistic models of nanoporous car- aspects of the adsorption of ions in these systems over
bon electrodes with an ionic liquid electrolyte. We a range of applied potentials. Even at zero applied
study the response of the supercapacitors to a sud- potential, this electrolyte is found to wet the pores of
denly applied electrical potential difference. We show the carbon electrodes effectively such that equal num-
that the charge penetrates progressively from the bers of cations and anions are absorbed within the
carbon/electrolyte interface to the bulk of the electro- nanopores. When a potential is applied, there is an
des. The charging dynamics is, however, markedly exchange of ions between the bulk electrolyte and
heterogeneous at the nanoscale. Finally, by using an those absorbed within the electrode interiors such
equivalent circuit model, we show that a macroscopic that, after equilibration, there is a net excess of cations
electrode would charge on the time scale of 1 s, which within the electrode held at negative potential and the
means that the power density of such a device would electrode develops a net charge which balances the
be very good. excess charge of the absorbed ions.16,18,32 To a good
approximation, the total volume of electrolyte within
RESULTS AND DISCUSSION the electrode is not affected by the application of a
Charging Dynamics. The system we simulate is com- potential.
posed of an ionic liquid electrolyte (1-butyl-3-methyli- Figure 2 shows the time evolution of the average
midazolium hexafluorophosphate, BMI-PF6) confined charge per carbon inside the positive electrode for each
between a pair of nanoporous carbon electrodes, as of the three different porous materials studied after the
shown on Figure 1. For the latter, three structures potential difference is applied between the electrodes.
which were obtained by Palmer et al. by performing They all reach a plateau value of ≈0.007 e/carbon atom
quenched molecular dynamics at various thermal for CDC-1200 and CDC-950 and of ≈0.006 e/carbon
rates30 were used. They were shown to mimic very atom for CDC-800. It is probable that the fully charged
well carbide-derived carbons (CDCs)31 obtained under régime is not reached on the time scale of these
various synthesis temperatures, so that we will label simulations: for example, in CDC-1200, we expect to
them following the corresponding temperatures in C: reach a maximal charge of ≈0.011 e/carbon atom
CDC-800, CDC-950, and CDC-1200. They differ mostly which is obtained in calculations where we allow
by their local nanoporosity (pore sizes and shapes). sufficient time for full equilibration.18 This observation
In particular, their respective average pore sizes are 7.5, is consistent with an experimental observation that the

PÉAN ET AL. VOL. XXX ’ NO. XX ’ 000–000 ’ XXXX B

www.acsnano.org
 ARTICLE
Figure 2. Time evolution of the average charge per carbon
atom on the positive electrode for the three nanoporous
carbon materials. At t = 0, the potential difference between
the electrodes is suddenly switched from 0 to 1 V.

amount of charge stored depends on the rate of

charging: For example, in nanopowders of CDCs,
the specific capacitance was shown to decrease by
15
40% when the current density is switched from
0.1 to 5 A g
1 during galvanostatic cycling.33 Our
simulation setup corresponds to an instantaneous
change of potential, so that we expect our results to
match experimental ones in the limit of high current
densities. Note that even if smaller than the equilibri-
um one, the final charge of the CDCs in the present
simulations remains much larger than in the case of
planar graphite electrodes, for which 0.004 e are stored
per carbon atom for the same applied electric poten-
tial of 1 V.34 The comparison of the different porous
materials shows that CDC-1200 and CDC-950 seem to
have similar charging profiles on Figure 2 while that of
CDC-800 is markedly different: it takes a much longer
Figure 3. Evolution of the local charge with time as a
time to reach the plateau régime for the latter. This function of depth (z, see Figure 1) inside the positive
behavior can be linked to the average pore sizes, which electrodes for the three CDCs. At each time step, the charge
is calculated for the whole slab of carbon atoms with
are identical for CDC-1200 and CDC-950, with those of
positions between z
δ and z þ δ (δ = 0.5 Å here). Dark
CDC-800 being smaller.30 We will now further analyze regions have a null charge, while yellow-colored ones have
the dynamics on the local scale. a high positive charge. The dashed lines on the top panel
delimit the slice which is further analyzed on Figure 4.
Charging Heterogeneity. In constant potential simula-
tions, the charge carried by each carbon atom is
calculated every time step. It is, thus, possible to electrode as the diffusive ion exchange process
examine the evolution of the local charge for any reaches greater depths. This is consistent with the
well-defined region of the electrode. To illustrate the transmission line models which are used for represent-
charge penetration inside the electrode, Figure 3 dis- ing supercapacitors at the macroscopic scale as equiva-
plays the charging profiles as a function of depth (z, as lent electric circuits,35 and is in qualitative agreement
defined on Figure 1) inside the positive electrodes for with the results of Kondrat and Kornyshev, who
the three CDCs. The charge is calculated for the whole showed by using a mean-field model that the charging
slab of carbon atoms with positions between z
δ and of initially wetted single pores occurs by a diffusive ion
z þ δ (δ = 0.5 Å here); it reflects the local imbalance in exchange.25 In the present case, the complex network
the number of cations and anions induced by the of pores is initially wetted by the ionic liquid. The
applied potential, which is strongly affected by the diffusive charging of the nanoporous carbon is, how-
local topology and connectivity of the nanoporous ever, very heterogeneous at the nanometric scale,
material. The origin (z = 0) corresponds to the frontier since some regions very deep inside the electrode start
between the electrolyte and the electrode. There is a to charge relatively quickly after the electric potential is
general tendency for the charge to penetrate progres- applied. For example, for CDC-950, there is a large
sively from the boundary toward the interior of the increase of the charge in the middle of the electrode

PÉAN ET AL. VOL. XXX ’ NO. XX ’ 000–000 ’ XXXX C

www.acsnano.org
 ARTICLE
Figure 4. (Top) Repartition of the local charge inside a 11 Å-thick slice of the CDC-1200 positive electrode (corresponding to
the zone with 19.5 Å < z < 30.5 Å which is delimited on Figure 3), for different times of simulation. (Bottom) Snapshots showing
the electrode structure and the instantaneous ionic positions for this slice (red, cations centers of mass; green, anions) at the
corresponding times. The white circle and square on the bottom left panel indicate regions which are discussed in the text.

for 1 ns < t < 2 ns. This is due to the presence of a large various pores: their local charge increases on the
pore ending close to this region (see the snapshot nanosecond time scale but it is subject to fluctuations
provided in the Supporting Information), which acts as due to the diffusion of the ions inside the electrodes. In
an electrolyte reservoir. Another indicator of this het- addition to the pore sizes, the connectivity between
erogeneity in charging can be seen in CDC-800, for the pores should play an important role in the charging
which the region around z = 40 Å starts to charge long process and influence its heterogeneous character.
before the one at 35 Å. Equivalent Circuit. The interpretation of experiments
To further illustrate the importance of the local on supercapacitors usually relies on the use of equiva-
structure on the heterogeneous charging, the rearran- lent electric circuits. Here we examine the relevance of
gement of the local charge inside a 11 Å-thick slice such macroscopic models in the light of our micro-
centered 25 Å away from the interface (i.e., correspond- scopic simulations and the implications of the latter for
ing to 19.5 Å < z < 30.5 Å on Figure 3) is shown at transport on larger scales. The most commonly used
various simulation times for the positive CDC-1200 model in the case of nanoporous electrodes is that
electrode on Figure 4. The structure differs substan- of a transmission line, in which the charge penetrates
tially from pictures based on regular, long pores such progressively into the electrode.35,39 This model is
as carbon nanotubes. It is better described as a com- based on an infinite succession of slices connected in
plex network of interconnected edge, planar, hollow or parallel (via the constant potential solid surface), each
pocket sites of size barely larger than a few ionic radii, of them consisting of a capacitor, accounting for the
and the ions sample various adsorption modes.36,37 possibility to store charge locally via the polarization of
The amount of charge which is stored locally increases the electrolyte at the surface of the pores, and of a
with the degree of confinement of these sites.38 A resistance describing the dissipative electrolyte trans-
highly confined site close to the center of the slice is port through the pores.
indicated with a white circle on Figure 4. At t = 0.2 ns, The link between such effective macroscopic mod-
this site is occupied by a cation and the local charge on els and the underlying microscopic mechanisms re-
the electrode is negative. At t = 1.0 ns, there is an ion mains to be established. A key question, discussed
pair, and the cation is then expelled, resulting in the below, is the definition of the equivalent circuit slices
formation of a high positive charge on the carbon from the microscopic structure of the electrode, which
electrode. At the bottom right corner of the slice, a should be as large as possible to ensure a good
wider pore is present (it is indicated with a white square representation of the chosen regions. It should be
on Figure 4). It is already well charged at t = 0.2 ns, large enough to ensure that the sum of all the hetero-
with 3 anions lying close to the surface. Subsequently, geneous charging events results in an homogeneous
though, one of the anions leaves the pore for another charging at the scale of the slice. Given the small size of
region. It is replaced by a cation, decreasing the local the simulated system, we have thus divided each
charge. There is, therefore, an interplay between the electrode into two slices of equal thickness l, resulting

PÉAN ET AL. VOL. XXX ’ NO. XX ’ 000–000 ’ XXXX D

www.acsnano.org
 ARTICLE
TABLE 1. Parameters Employed in the Equivalent Circuit
Modelsa

CDC-1200 CDC-950 CDC-800

Rbulk(10 Ω)
8
1.0 1.1 1.2
Rl(108Ω) 1.2 0.7 4.9
C1(10
18F) 4.3 3.6 3.9
C2(10
18F) 2.7 3.2 2.6
a
In all cases, the bulk electrolyte region thickness is 50 Å for each half-cell, while
the considered electrode slices have a l = 21 Å thickness.

in the equivalent electric circuit illustrated on Figure 1.

Rbulk is the resistance of the electrolyte in the bulk
region, Rl is the resistance of each electrode slice, C1
and C2 are the capacitances of the two slices. Among
these quantities, the only unknown is Rl, which can be
determined by fitting the total electrode charge evolu-
tion with time. It is given by
"

#
t t
Q(t) ¼ Qmax 1
A1 exp

A2 exp

τ1 τ2
(1)
where the expressions for Qmax, A1, A2, τ1 and τ2 are
given as a function of the electric circuit components in
the Supporting Information. The fitting was made on
the bulk part of the electrode, i.e., by excluding the first
5 Å of the electrode in order to avoid interfacial effects.
Note that including this region, although decreasing
the quality of the fit, does not change the results
qualitatively.
The values of all the calculated and fitted param-
eters are given in Table 1 (see the Supporting Informa-
tion for additional details), and results of the fits are
shown on Figure 5 for the three CDC electrodes. In all
cases eq 1 underestimates the charge at short times
but the matching is very good for intermediate and
long times. To check the transferability of this ap- Figure 5. Total charge of the electrodes as a function of
time for the three systems. Molecular dynamics simulation
proach, we have made two different tests. First results are well described by eq 1, which corresponds to the
we have calculated the charges in the two slices equivalent circuit displayed on Figure 1, with the param-
(expressions provided in the Supporting Information); eters of Table 1. The only parameter adjusted to reproduce
these results is the resistance of the electrode slices, Rl. The
they correspond very well with the data extracted from charge of each slice predicted with the equivalent circuit is
the simulations, without further parameter fitting. also in good agreement with the simulation data, without
Second, we have compared the charging dynamics further parameter fitting.
for an applied potential of 2 V with the prediction from
eq 1 using the Rl values fitted at 1 V. Again, a good electrodes, which contain interconnected nanopores
agreement was obtained (the data is shown on Sup- with an average pore size of 9.0 Å. For CDC-800, the
plementary Figure 4). Therefore the proposed equiva- transport inside the smaller pores is slower, by a factor
lent circuit is clearly able to capture the charging of approximately 4
8. The low resistivity of the elec-
behavior of our modeled supercapacitor. Given the trolyte inside the electrodes may arise from the ex-
relative widths of the electrolyte (50 Å for each half- istence of collective transport effects, as suggested by
cell) and electrode regions (2l = 42 Å for each a recent study from Kondrat et al. in which the diffusion
electrode) in our simulation cells, we can conclude coefficients of the ions from an ionic liquid adsorbed
from the fact that Rbulk is of the same order of magni- inside electrified slit pores were shown to be strongly
tude as Rl for both CDC-1200 and CDC-950 that the dependent on the charged state of the electrode.40
transport of the ions is only slightly affected inside the Charging on Larger Scales. The good agreement be-
pores. This is consistent with the structure of these tween the simple two-slice model and the molecular

PÉAN ET AL. VOL. XXX ’ NO. XX ’ 000–000 ’ XXXX E

www.acsnano.org
 ARTICLE
simulation results further indicates that the size of the charge starts from the electrode/electrolyte interface and
regions defining the slices is adequate. More precisely, penetrates progressively into the bulk of the carbon
this length scale is sufficiently large to average out material. At the nanometer scale, though, strong hetero-
some of the local heterogeneity of the charging pro- geneities are observed due to the particular structure of
cess, while allowing the diffusive behavior to be cap- CDCs. These materials consist in a complex network of
tured on a coarser scale. Identifying the relevant length pores of various sizes, and differ substantially from the
scales for the definition of coarse-grained ionic and carbon nanotubes or the slit pores which are often used
charge densities might also be useful in the frame- to model them. Depending on the local organization of
work of mesoscopic models such as time-dependent the porous networks, some regions may start to charge
Density Functional Theory.25 In addition, we can also before others despite being located more deeply inside
extract from this effective macroscopic model a crude the electrode. For example, in CDC-950, a region in the
estimate of the characteristic charging time constant, middle of the electrode charges very quickly because it is
which would probably need to be improved by using located near to a large pore of the carbon.
larger simulation cells. In the full transmission line By comparing three CDCs with different structures,
model, the diffusive charging occurs over the thickness we have shown that the charging time depends on the
lmacro of the electrodes with a characteristic time scale average pore size. CDC-1200 and CDC-950, which have
RCl2macro, with R and C the resistance and capacitance similar average pore sizes, are charged on similar time
per unit length, respectively. This time can be deter- scales, even though the details of the pore connectivity
mined from our simulation results as and the distribution of pore throats may also play a

 role. The process is slower by a factor of 4
8 in the case
lmacro 2
τ(lmacro ) ¼ Rl  Ctot  (2) of CDC-800, which has a smaller average pore size.
lsim Experimental supercapacitors involving nanoporous
where Ctot is given by the total capacitance of the carbon show good power performances in addition
electrodes and lsim their thickness in our simulations. to their excellent energy densities,11 a result which
For a typical commercial supercapacitor, the electrode seemed rather counterintuitive and had not been
size is lmacro ∼ 100 μm, which gives us characteristic explained at the microscopic scale previously. From
charging times ranging between 0.5 and 4 s for our our molecular dynamics simulations data, we have
systems, in good agreement with the experimental fitted a macroscopic model based on an equivalent
data: for example, in reference 11, relaxation time electric circuit. The transmission line model which is
constants, defined as the minimum time needed to used in experiments is clearly able to capture the
discharge all the energy from the supercapacitor cell charging behavior. This fit provides us with a value
with an efficiency greater than 50%, of 5
20 s were for the resistance of the electrolyte inside the pores.
measured with similar CDCs to the ones we simulate This quantity is of the same order of magnitude as that
(although the electrolyte is different, its electrical con- of the bulk electrolyte (for similar dimensions), which
ductivity is of the same order of magnitude as the leads to charging times of one to ten seconds time
ionic liquid in the present study, and so would be the scale for a 100 μm thick electrode. This confirms that
corresponding Rbulk). Our estimate does not take into the transport of the ions is not much affected in the
account the thickness of the electrolyte region and the porous materials. In future work we will investigate
resistance of other components of the supercapacitor whether the fast charging that we observe can be
such as the separator, but these should not impact the associated with the existence of collective transport
charging time significantly. effects, as suggested by the recent work from Kondrat
et al.40 If the existence of such a mechanism leading to
CONCLUSION fast transport of the charge is shown to be universal for
In conclusion, we have used molecular dynamics a series of electrolytes and electrode structures, it
simulations in order to study the charging of electrified means that nanoporous-carbon based supercapacitors
nanoporous CDC electrodes impregnated with a BMI-PF6 have inherently high power densities, so that the quest
electrolyte. We have shown that it occurs in a diffusive for new carbon materials should mainly focus on
way, in agreement with the recent literature.25 The enhancing their energy density.

METHODS yields the correct structural, thermodynamic and dynamic

properties of BMI-PF6.41 Three and one sites describe the cation
The simulation cells consists of [BMI][PF6] ionic liquid (600 and the anion, respectively.34,42 The sizes of the simulation cells
ion pairs) surrounded by two identical and symmetric nanopor- are chosen in order to reproduce the experimental densities of
ous carbon electrodes, obtained by quenched molecular the bulk electrolyte. Supporting Information Table 1 gathers the
dynamics.30 Following our previous works, we use for the ionic number of carbon atoms and the lengths for all the simulation
liquid the coarse-grained model of Roy and Maroncelli, which cells for each type of studied carbon.

PÉAN ET AL. VOL. XXX ’ NO. XX ’ 000–000 ’ XXXX F

www.acsnano.org
 ARTICLE
Molecular dynamics simulations are conducted on the three Porous Carbon Nanofibers as an Efficient Electrode
different systems, with a time step of 2 fs which is consistent Material for Supercapacitors. ACS Nano 2012, 6, 7092–7102.
with the coarse-grained model used. The forces are calculated 6. Wu, Z.-S.; Sun, Y.; Tan, Y.-Z.; Yang, S.; Feng, X.; Mullen, K.
as the sum of site
site Lennard-Jones potential and Coulombic Three-Dimensional Graphene-Based Macro- and Meso-
interactions. Parameters for the ions and carbon atoms are the Porous Frameworks for High Performance Electrochemi-
same as in our previous work.18,34,42 In all the simulations, the cal Capacitive Energy Storage. J. Am. Chem. Soc. 2012, 134,
Coulomb interactions are calculated through a two-dimensional 19532–19535.
Ewald summation.43,44 7. Wu, Z.-S.; Parvez, K.; Feng, X.; Müllen, K. Graphene-Based
The procedure to simulate the charging process follows two In-Plane Micro-Supercapacitors with High Power and En-
steps: the system is first equilibrated for a few nanoseconds, first ergy Densities. Nat. Commun. 2013, 4, 2487.
with a constant charge of 0 e on all carbon atoms and then with 8. Hao, G.-P.; Lu, A.-H.; Dong, W.; Jin, Z.-Y.; Zhang, X.-Q.;
a 0 V potential difference between the two electrodes. At t = 0, Zhang, J.-T.; Li, W.-C. Sandwich-Type Microporous Carbon
this potential difference Δψ0 is suddenly set to 1 V (ψþ = 0.5 V Nanosheets for Enhanced Supercapacitor Performance.
and ψ
=
0.5 V or 2 V. The electrodes are held at constant Adv. Energy Mater. 2013, 3, 1421–1427.
potential using a method developed by Reed et al.43 from an 9. Yang, X.; Cheng, C.; Wang, Y.; Qiu, L.; Li, D. Liquid-Mediated
original proposal by Siepmann.45 Simulations are continued for Dense Integration of Graphene Materials for Compact
sufficiently long to reach equilibrium, i.e., to exhibit a long Capacitive Energy Storage. Science 2013, 341, 534–537.
plateau for the value of the charge on carbon atoms. The con- 10. DeBlase, C. R.; Silberstein, K. E.; Truong, T.-T.; Abruna, H. D.;
stant potential approach is computationally expensive com- Dichtel, W. R. β-Ketoenamine-Linked Covalent Organic
pared to the constant charge approach but is essential to study Frameworks Capable of Pseudocapacitive Energy Storage.
realistically dynamic processes such as the charging process.38 J. Am. Chem. Soc. 2013, 135, 16821–16824.
Typically, 50 000 CPU hours were needed to gather a trajectory 11. Chmiola, J.; Yushin, G.; Gogotsi, Y.; Portet, C.; Simon, P.;
of 1 ns for a given system. Taberna, P. L. Anomalous Increase in Carbon Capacitance
Controlling the temperature is an important concern for such at Pore Sizes Less Than 1 Nanometer. Science 2006, 313,
simulations. In a macroscopic experiment, the supercapacitor is 1760–1763.
thermostated in order to control its temperature. It is therefore 12. Raymundo-Pi~ nero, E.; Kierzek, K.; Machnikowski, J.; Béguin,
preferable to simulate the systems in the NVT ensemble, keep- F. Relationship between the Nanoporous Texture of Acti-
ing a weak coupling to the heat-bath in order to avoid the vated Carbons and their Capacitance Properties in Differ-
dynamics being dictated by the thermostat relaxation time. ent Electrolytes. Carbon 2006, 44, 2498–2507.
Here we have used a Nosé-Hoover thermostat46 with a relaxa- 13. Largeot, C.; Portet, C.; Chmiola, J.; Taberna, P. L.; Gogotsi, Y.;
tion time of 10 ps and a target temperature of 400 K. The Simon, P. Relation Between the Ion Size and Pore Size for
variations of temperature with time are shown in the Support- an Electric Double-Layer Capacitor. J. Am. Chem. Soc.
ing Information for Δψ0 = 1 V: the largest increase is observed in 2008, 130, 2730–2731.
the case of CDC-1200, for which the temperature switches from 14. Wu, P.; Huang, J. S.; Meunier, V.; Sumpter, B. G.; Qiao, R.
400 to 470 K during the first nanosecond due to the Joule Complex Capacitance Scaling in Ionic Liquids-Filled Nano-
heating effect,29 then up to 490 K during the next 5 ns before pores. ACS Nano 2011, 5, 9044–9051.
starting to decrease slightly. The diffusion coefficients increase
15. Shim, Y.; Kim, H. J. Nanoporous Carbon Supercapacitors in
by a factor of 2
3 in this temperature range in the bulk ionic
an Ionic Liquid: A Computer Simulation Study. ACS Nano
liquid,41 which has to be taken into account when analyzing the
2010, 4, 2345–2355.
results. The increase of temperature is larger for the two other
16. Kondrat, S.; Kornyshev, A. A. Superionic State in Double-
applied potentials; all the results presented here, therefore,
Layer Capacitors with Nanoporous Electrodes. J. Phys.:
correspond to the 1 V case, although the charging dynamics
Condens. Matter 2011, 23, 022201.
is qualitatively the same for Δψ0 = 2 V.
17. Kondrat, S.; Perez, C. R.; Presser, V.; Gogotsi, Y.; Kornyshev,
Conflict of Interest: The authors declare no competing
A. A. Effect of Pore Size and its Dispersity on the Energy
financial interest.
Storage in Nanoporous Supercapacitors. Energy Environ.
Sci. 2012, 5, 6474–6479.
Acknowledgment. We acknowledge the support of the
18. Merlet, C.; Rotenberg, B.; Madden, P. A.; Taberna, P.-L.;
French Agence Nationale de la Recherche (ANR) under grant
Simon, P.; Gogotsi, Y.; Salanne, M. On the Molecular Origin
ANR-2010-BLAN-0933-02 ('Modeling the Ion Adsorption in
of Supercapacitance in Nanoporous Carbon Electrodes.
Carbon Micropores'). The research leading to these results
Nat. Mater. 2012, 11, 306–310.
has received funding from the European Research Council
19. Wu, P.; Huang, J.; Meunier, V.; Sumpter, B. G.; Qiao, R.
under the European Union's Seventh Framework Programme
Voltage Dependent Charge Storage Modes and Capacity
(FP/2007-2013)/ERC Grant Agreement n. 102539. We are grate-
in Sub-Nanometer Pores. J. Phys. Chem. Lett. 2012, 3,
ful for the computing resources on JADE (CINES, French
1732–1737.
National HPC) obtained through the project c2013096728. We
20. Xing, L.; Vatamanu, J.; Borodin, O.; Bedrov, D. On the
acknowledge PRACE for awarding us access to resource CURIE
Atomistic Nature of Capacitance Enhancement Generated
based in France at TGCC.
by Ionic Liquid Electrolyte Confined in Subnanometer
Supporting Information Available: Additional computational Pores. J. Phys. Chem. Lett. 2013, 4, 132–140.
details, variation of temperature due to Joule effect, details of 21. Feng, G.; Li, S.; Presser, V.; Cummings, P. T. Molecular
the equivalent circuit modeling and charging dynamics for an Insights into Carbon Supercapacitors Based on Room
applied potential of 2 V. This material is available free of charge Temperature Ionic Liquids. J. Phys. Chem. Lett. 2013, 4,
via the Internet at http://pubs.acs.org. 3367–3376.
22. Lanning, O.; Madden, P. A. Screening at a Charged Surface
by a Molten Salt. J. Phys. Chem. B 2004, 108, 11069–11072.
REFERENCES AND NOTES 23. Pinilla, C.; Del Pópolo, M. G.; Kohanoff, J.; Lynden-Bell, R. M.
1. Miller, J. R.; Simon, P. Electrochemical Capacitors for Polarization Relaxation in an Ionic Liquid Confined between
Energy Management. Science 2008, 321, 651–652. Electrified Walls. J. Phys. Chem. B 2007, 111, 4877–4884.
2. Simon, P.; Gogotsi, Y. Materials for Electrochemical Capac- 24. Vatamanu, J.; Borodin, O.; Smith, G. D. Molecular Simula-
itors. Nat. Mater. 2008, 7, 845–854. tions of the Electric Double Layer Structure, Differential
3. Frackowiak, E.; Abbas, Q.; Béguin, F. Carbon/Carbon Super- Capacitance, and Charging Kinetics for N-Methyl-N-pro-
capacitors. J. Energy Chem. 2013, 22, 226–240. pylpyrrolidinium Bis(fluorosulfonyl)imide at Graphite Elec-
4. Armand, M.; Tarascon, J.-M. Building Better Batteries. trodes. J. Phys. Chem. B 2011, 115, 3073–3084.
Nature 2008, 451, 652–657. 25. Kondrat, S.; Kornyshev, A. A. Charging Dynamics and
5. Chen, L.-F.; Zhang, X.-D.; Liang, H.-W.; Kong, M.; Guan, Q.-F.; Optimization of Nano-Porous Supercapacitors. J. Phys.
Chen, P.; Wu, Z.-Y.; Yu, S.-H. Synthesis of Nitrogen-Doped Chem. C 2013, 117, 12399–12406.

PÉAN ET AL. VOL. XXX ’ NO. XX ’ 000–000 ’ XXXX G

www.acsnano.org
 ARTICLE
26. Reed, S. K.; Madden, P. A.; Papadopoulos, A. Electrochemi-
cal Charge Transfer at a Metallic Electrode: A Simulation
Study. J. Chem. Phys. 2008, 128, 124701.
27. Vatamanu, J.; Borodin, O.; Smith, G. D. Molecular Insights
into the Potential and Temperature Dependences of the
Differential Capacitance of a Room-Temperature Ionic
Liquid at Graphite Electrodes. J. Am. Chem. Soc. 2010,
132, 14825–14833.
28. Merlet, C.; Rotenberg, B.; Madden, P. A.; Salanne, M.
Computer Simulations of Ionic Liquids at Electrochemical
Interfaces. Phys. Chem. Chem. Phys. 2013, 15, 15781–15792.
29. Merlet, C.; Péan, C.; Rotenberg, B.; Madden, P. A.; Simon, P.;
Salanne, M. Simulating Supercapacitors: Can We Model
Electrodes as Constant Charge Surfaces? J. Phys. Chem.
Lett. 2013, 4, 264–268.
30. Palmer, J. C.; Llobet, A.; Yeon, S.-H.; Fisher, J. E.; Shi, Y.;
Gogotsi, Y.; Gubbins, K. E. Modeling the Structural Evolu-
tion of Carbide-Derived Carbons Using Quenched Molec-
ular Dynamics. Carbon 2010, 48, 1116–1123.
31. Gogotsi, Y.; Nikitin, A.; Ye, H.; Zhou, W.; Fischer, J. E.; Yi, B.;
Foley, H. C.; Barsoum, M. W. Nanoporous Carbide-Derived
Carbon with Tunable Pore Size. Nat. Mater. 2003, 2, 591–594.
32. Kondrat, S.; Georgi, N.; Fedorov, M. V.; Kornyshev, A. A. A
Superionic State in Nano-Porous Double-Layer Capacitors:
Insights from Monte Carlo Simulations. Phys. Chem. Chem.
Phys. 2011, 13, 11359–11366.
33. Perez, C. R.; Yeon, S.-H.; Segalini, J.; Presser, V.; Taberna,
P.-L.; Simon, P.; Gogotsi, Y. Structure and Electrochemical
Performance of Carbide-Derived Carbon Nanopowders.
Adv. Funct. Mater. 2013, 23, 1081–1089.
34. Merlet, C.; Salanne, M.; Rotenberg, B.; Madden, P. A. Imi-
dazolium Ionic Liquid Interfaces with Vapor and Graphite:
Interfacial Tension and Capacitance from Coarse-Grained
Molecular Simulations. J. Phys. Chem. C 2011, 115, 16613–
16618.
35. Conway, B. E. Electrochemical Supercapacitors: Scientific
Fundamentals and Technological Applications; Kluwer:
New York, 1999.
36. Wang, H.; Koster, T. K. J.; Trease, N. M.; Segalini, J.; Taberna,
P. L.; Simon, P.; Gogotsi, Y.; Grey, C. P.; Real-Time, N. M. R.
Studies of Electrochemical Double-Layer Capacitors. J. Am.
Chem. Soc. 2011, 133, 19270–19273.
37. Deschamps, M.; Gilbert, E.; Azais, P.; Raymundo-Pi~ nero, E.;
Ammar, M. R.; Simon, P.; Massiot, D.; Béguin, F. Exploring
Electrolyte Organization in Supercapacitor Electrodes with
Solid-State NMR. Nat. Mater. 2013, 12, 351–358.
38. Merlet, C.; Péan, C.; Rotenberg, B.; Madden, P. A.; Daffos, B.;
Taberna, P.-L.; Simon, P.; Salanne, M. Highly Confined Ions
Store Charge More Efficiently in Supercapacitors. Nat.
Commun. 2013, 4, 2701.
39. Taberna, P.-L.; Simon, P.; Fauvarque, J.-F. Electrochemical
Characteristics and Impedance Spectroscopy Studies of
Carbon-Carbon Supercapacitors. J. Electrochem. Soc.
2003, 150, A292–A300.
40. Kondrat, S.; Wu, P.; Qiao, R.; Kornyshev, A. Accelerating
Charging Dynamics in Sub-Nanometer Pores. arXiv
2013, No. 1311.7529.
41. Roy, D.; Maroncelli, M. An Improved Four-Site Ionic Liquid
Model. J. Phys. Chem. B 2010, 114, 12629–12631.
42. Merlet, C.; Salanne, M.; Rotenberg, B. New Coarse-Grained
Models of Imidazolium Ionic Liquids for Bulk and Inter-
facial Molecular Simulations. J. Phys. Chem. C 2012, 116,
7687–7693.
43. Reed, S. K.; Lanning, O. J.; Madden, P. A. Electrochemical
Interface Between an Ionic Liquid and a Model Metallic
Electrode. J. Chem. Phys. 2007, 126, 084704.
44. Gingrich, T. R.; Wilson, M. On the Ewald Summation of
Gaussian Charges for the Simulation of Metallic Surfaces.
Chem. Phys. Lett. 2010, 500, 178–183.
45. Siepmann, J. I.; Sprik, M. Influence of Surface-Topology and
Electrostatic Potential on Water Electrode Systems. J. Chem.
Phys. 1995, 102, 511–524.
46. Nosé, S. A Unified Formulation of the Constant Tempera-
ture Molecular Dynamics Methods. J. Chem. Phys. 1984,
81, 511–519.

PÉAN ET AL. VOL. XXX ’ NO. XX ’ 000–000 ’ XXXX H

www.acsnano.org