A Mathematical Model of An Electrochemical Capacitor With Double-Layer and Faradaic Processes

3168 Journal of The Electrochemical Society, 146 (9) 3168-3175 (1999)
S0013-4651(98)11-057-1 CCC: $7.00 The Electrochemical Society, Inc.

Electrochemical capacitors have been receiving increased atten-
tion during recent years
1-5
as high-power devices in energy-storage
systems. There are two mechanisms of energy storage in electro-
chemical capacitors: (i) a double-layer (DL) charging process due to
charge separation, and (ii) a faradaic process due to redox reactions.
Devices based on the DL phenomenon are referred to as electro-
chemical DL capacitors, and those based on faradaic reactions are
termed pseudocapacitors. Very high-surface-area carbon materials
6-9
are widely used for DL capacitors. On the other hand, relatively high-
surface-area transition metal oxides, such as amorphous
RuO
2
xH
2
O,
10
porous NiO
x
,
11,12
and CoO
x
,
12,13
have been identi-
fied as possible electrode materials for pseudocapacitors.
A number of models have been developed for analyzing the be-
havior of DL capacitors. Posey and Morozumi
14
developed a model
for potentiostatic and galvanostatic charging of the DL in porous
electrodes. Johnson and Newman
15
also developed a model for a
porous electrode to analyze desalting processes in terms of ionic
adsorption on porous carbon, and Tiedemann and Newman
16
used
the results from that model to evaluate the DL capacity of porous
electrodes. Recently, Srinivasan et al.
17
developed an analytic model
and used it to study constant-current discharging, cyclic voltage
sweeping, and the ac impedance of carbon xerogel DL capacitors.
Also, an analytic model was developed by Farahmandi
18
and used to
study the effects of both ionic and solid-phase conductivities on the
behavior of an electrochemical capacitor. More recently, Pillay and
Newman
19
modeled the influence of side reactions on the perfor-
mance of electrochemical capacitors. However, none of these capac-
itor models accounted for both pseudocapacitance and DL capaci-
tance. In fact, few capacitor models have considered faradaic reac-
tions, and for those that did,
20,21
only the approximation of linear
faradaic kinetics was considered.
Therefore, the objective of this work is to develop a mathemati-
cal model of an electrochemical capacitor with both DL and farada-
ic processes. These two capacitive processes can occur simultane-
ously with both contributing to the overall capacitance of the mater-
ial. This is especially true in the relatively high-surface area transi-
tion metal oxide pseudocapacitor materials.
5
The specific system
used to illustrate the complementary effects of DL and faradaic pro-
cesses is a symmetric capacitor comprised of uniformly sized spher-
ical hydrous ruthenium oxide (RuO
2
xH
2
O) particles with 30 wt %
sulfuric acid as the electrolyte. In the model, the diffusion of protons
into the solid RuO
2
xH
2
O particles is ignored for simplicity; conse-
quently, the DL and faradaic processes take place only on the exter-
nal surface of the particles. This simplification can be removed,
however, by following the procedure presented by Doyle et al.
22
for
ionic diffusion in the particles. The effects of particle size and cell
current density on the charge/discharge behavior are studied, and the
roles of the DL and Faradaic processes are delineated with specific
reference to Ragone plots.
Model Description
Figure 1 displays a schematic of a typical capacitor cell. Two
identical RuO
2
xH
2
O electrodes are separated by an ionically con-
ductive glass fiber and are contacted on one side, as shown in Fig. 1,
with tantalum current collectors. A solution of 30 wt % H
2
SO
4
is
used as the electrolyte, which completely fills the pores in the elec-
trodes and the separator. A one-dimensional model is developed
using the macrohomogeneous theory of porous electrodes, reviewed
by Newman and Tiedemann,
23
and De Vidts and White.
24
In the
model presented here, the electrolyte concentration is assumed to be
invariant and side reactions and thermal effects are ignored, along
with the variation of the DL capacitance with potential.
A Mathematical Model of an Electrochemical Capacitor with
Double-Layer and Faradaic Processes
Chuan Lin, James A. Ritter,* Branko N. Popov,* and Ralph E. White*
,z
Department of Chemical Engineering, Swearingen Engineering Center, University of South Carolina, Columbia,
South Carolina 29208, USA
A mathematical model of an electrochemical capacitor with hydrous ruthenium oxide (RuO
2
xH
2
O) electrodes including both dou-
ble-layer and surface faradaic processes is developed to predict the behavior of the capacitor under conditions of galvanostatic
charge and discharge. The effect of RuO
2
xH
2
O particle size is studied and shows that the smaller the particles the better the per-
formance because of the increased surface area per unit volume or mass. The model also predicts that the faradaic process increas-
es significantly the energy per unit volume of the capacitor for power densities of 100 kW/L or less.
1999 The Electrochemical Society. S0013-4651(98)11-057-1. All rights reserved.
Manuscript submitted November 16, 1998; revised manuscript received April 28, 1999.
* Electrochemical Society Active Member.
z
E-mail: [email protected]
Figure 1. Schematic of an electrochemical capacitor cell upon which the
model is based.
) unless CC License in place (see abstract). ecsdl.org/site/terms_use address. Redistribution subject to ECS terms of use (see 128.210.126.199 Downloaded on 2014-06-17 to IP
Journal of The Electrochemical Society, 146 (9) 3168-3175 (1999) 3169
The faradaic reactions occurring in the electrodes during
charge/discharge are expressed as
25
charge
H
0.8
RuO
2
xH
2
O oH
0.8
RuO
2
xH
2
O H
[1]
discharge
for the positive electrode, and as
charge
H
0.8
RuO
2
xH
2
O H
oH
0.8
RuO
2
xH
2
O [2]
discharge
for the negative electrode, where the state of charge, , is 0.5 for a
cell to be charged to 1 V. These reactions show that initially the
active materials in both electrodes are H
0.8
RuO
2
xH
2
O, which may
be obtained by discharging each RuO
2
xH
2
O electrode separately
using external electrodes. Then, during charge, the oxidation reac-
tion (Eq. 1) takes place in the positive electrode and the reduction
reaction (Eq. 2) occurs in the negative electrode. Finally, during dis-
charge, the opposite reactions take place in both electrodes, which
brings both electrodes to their original states after full discharge,
thereby completing a charge/discharge cycle. Thus, during charge,
the oxidation state of Ru changes from 3.2 to 3.7 in the positive elec-
trode and from 3.2 to 2.7 in the negative electrode, and during dis-
charge, it changes back from 3.7 to 3.2 in the positive electrode and
from 2.7 to 3.2 in the negative electrode.
In an electrode, the total current transferred from the matrix
phase to the electrolyte phase is expressed in terms of the DL current
and the faradaic current as
19,26
[3]
where i
2
is the superficial current density based on the projected
electrode area in the electrolyte phase, x is the position coordinate,
s
v
is the specific surface area per unit volume of the electrode, C
d
is
the DL capacitance,
1
and
2
are the potentials in the matrix and
electrolyte phases, respectively, and j
f
is the faradaic transfer current
density.
2
is defined as the potential
19,25
of a saturated calomel ref-
erence electrode (SCE) connected to the electrolyte phase at the
position of interest relative to an identical electrode connected to the
electrolyte phase just inside the left current collector (x 0). Thus,
2
is zero at x 0. The faradaic transfer current, j
f
, is taken in the
form of the Butler-Volmer equation
27
j
f
i
0
{exp[
a
f (
1

2
U
1
)]
exp[
c
f (
1

2
U
1
)]} [4]
where i
0
is the exchange current density of the faradaic reaction, f is
equal to F/RT, F is Faradays constant, R is the universal gas con-
stant, T is the absolute temperature,
a
and
c
are the anodic and
cathodic transfer coefficients for the reactions, respectively, and U
1
is the equilibrium potential of the faradaic reaction. U
1
is a function
of the state of charge,
25
which can be expressed by , the fraction of
oxidized species in the electrode. By introducing another variable,
Q
f
, the faradaic charge of the electrode, , , and j
f
are related by the
following equations
[5]
0.5 [6]
[7]
where Q
f,red
is the faradaic charge of a fully reduced electrode, and
Q
f,oxd
is the faradaic charge of a fully oxidized electrode. Equation 7
relates the rate of accumulation of the faradaic charge within the
electrode to the faradaic transfer current. For simplicity, Eq. 4, 5, and
7 are combined into a single equation
Q
t
s j
f
v f
Q Q
Q Q
f f, red
f,oxd f, red
i
x
s C
t
s j
2 1 2
v d v f
( )
exp[
c
f (
1

2
U
1
)]} [8]
Conservation of charge leads to
[9]
where i
1
is the superficial current density based on the projected
electrode area in the matrix phase of the porous electrode. With the
assumption of no concentration gradients in the electrodes, Ohms
law applies in both the matrix and electrolyte phases as
[10]
[11]
where is the electronic conductivity of the matrix, and
p
is the
ionic conductivity of the electrolyte inside the pores of the elec-
trodes. In order to solve the system of equations more efficiently,
Eq. 3, 4, 9, 10, and 11 are combined into a single equation
exp[
c
f (E U
1
)]} [12]
with the local electrode potential, E, defined as
E
1

2
[13]
and the DL time constant as
[14]
where L is the thickness of one electrode (L L
), and the
dimensionless distance is defined as x/L. In the separator, because
the electronic conductivity in the glass fiber is zero
i
1
0 [15]
and thus the total transfer current is zero
[16]
However, Ohms law still applies to the current in the electrolyte
within the pores of the separator, as
[17]
where
s
is the ionic conductivity in the pores of the separator.
To solve this system of partial differential and algebraic equa-
tions, the following boundary and initial conditions are used. For the
case of a constant-current charge/discharge between 0 and 1 V, the
boundary conditions for the positive electrode are
at 0, t > 0 [18]
exp[
c
f (E U
1
)]} [19]
at 1, t > 0 [20]
E i L

cell
p
t
s i
Q Q
f E U
v
f,oxd f, red
a
0
1
{exp[ ( )]
E i L

cell
i
x
2
2

s
i
x
2
0

s C L
v d
p
2
1 1
2
2
0
1
E E
t
i
C
f E U

d
a
{exp[ ( )]
i
x
2

p
i
x
1

i
x
i
x
1 2
0
t
s i
Q Q
f U

v
f,oxd f, red
a
0
1 2 1
{exp[ ( )]
[21]
The boundary conditions are set such that at the interface between
the current collector and the electrode ( 0), the current arises
from the matrix phase, and at the interface between the electrode and
separator ( 1), the current arises from the solution phase. The ini-
tial conditions are
25
at t 0 for charge E V
o
and 0 [22]
at t 0 for discharge E 2V
o
and 1 [23]
where V
o
is the initial potential before charge and is taken as 0.5 V
(SCE) for a cell charge/discharge between 0 and 1 V. Similar bound-
ary and initial conditions can be set for the negative electrode. How-
ever, in this model only the positive electrode is simulated; because
of the symmetrical feature of the two electrodes, results for the neg-
ative electrode are calculated based on the simulation results for the
positive electrode.
The entire system of equations to be solved is comprised of two
equations and two unknowns, which are Eq. 8 and 12, and E and ,
respectively. The system of equations for the positive electrode is
solved using a backward finite difference method
28
with Newmans
BAND(J) algorithm.
29
It is noteworthy that Eq. 17 is used to evalu-
ate the potential in the solution phase of the separator, which is also
solved simultaneously with the system of equations for the positive
electrode. Once E and are obtained, the other unknowns are calcu-
lated using these above equations; for example, Eq. 11, 13, and 4 are
used to calculate
1
,
2
, and j
f
, respectively. The DL and faradaic
currents in each electrode are defined by and obtained from the fol-
lowing two equations, respectively
[24]
i
f
s
v
j
f
[25]
Model Parameters
In this study, the RuO
2
xH
2
O particles are assumed to be non-
porous spheres with a uniform size; thus, the pores in the electrodes
are composed of the spaces between the particles, and the surface
area per unit volume, s
v
, of the electrode is expressed as
[26]
where D is the diameter of the particles and is the electrode poros-
ity (taken as 0.25). Many of the other parameters associated with the
RuO
2
xH
2
O are unavailable in the literature; therefore, the following
reasonable assumptions are made. The DL capacitance on the RuO
2
surface is assumed to be 2.0 10
5
F/cm
2
, which is the same for
carbon materials.
30
Both of the anodic and cathodic transfer coeffi-
cients,
a
and
c
, are taken as 0.5; and the exchange current density
of the faradaic reactions, i
0
, is taken as 10
5
A/cm
2
. The equilibrium
potential, U
1
, for the faradaic reactions is estimated from a linear
relationship between the potential and the state of charge (), as
reported by Jow and Zhang.
25
At 0, U
1
is equal to 0.5 V for both
the electrodes, and at 0.5, U
1
is equal to 1 and 0 V for the posi-
tive and negative electrodes, respectively. Also, according to the
relationship between and (Eq. 6), U
1
is equal to 0.5(1 ) and
0.5 for the reactions in the positive (Eq. 1) and negative electrodes
(Eq. 2), respectively. Here is the fraction of oxidized species in the
faradaic reaction.
In addition, it is assumed that the faradaic reactions take place
only on the surface of the RuO
2
xH
2
O particles inside the electrodes.
As a result, the faradaic charge of a fully oxidized electrode, Q
f,oxd
,
is estimated from
[27] Q
s F
h L
f,oxd
v
A

2
s
D
v

6 1 ( )
i s C
E
t
c v d

0
where s
v
is obtained from Eq. 26, is equal to 0.5 for the fully
charged cell, h is the length of the crystal lattice on the surface and
taken approximately as 4 ,
31
and L
A
is Avogadros number. Equa-
tion 27 implies that there is only one Ru atom in the area of h
2
on the
surface of the RuO
2
xH
2
O particles and that Q
f,oxd
is a function of
the particle size. The faradaic charge of a fully reduced electrode,
Q
f,red
, is taken as zero.
The electronic conductivity of the matrix, , is taken as
10
5
S/cm,
32
and the ionic conductivity of bulk 30 wt % H
2
SO
4
,
0
,
is 0.8 S/cm.
33
The ionic conductivity of the electrolyte inside the
pores of the electrode is frequently expressed as
29
p

0
1.5
[28]
Similarly, the ionic conductivity of the electrolyte inside the pores of
the separator is given by
s

0
s
1.5
[29]
where the separator porosity,
s
, is taken as 0.7. The electrode thick-
ness for both the positive and negative electrodes is 50 m, the sep-
arator thickness is 25 m, and the temperature is 25C. All these val-
ues are summarized in Tables I and II, along with the values of other
parameters used in this model.
Results
At the outset, it is useful to verify the accuracy of the numerical
method, which is done by comparing the numerical results for the
Table I. Model parameters that describe the Faradaic kinetics
and open-circuit potentials of the electrodes.
Parameter Value Reference
i
0
10
5
A/cm
2
Assumed
a
0.5 Assumed
c
0.5 Assumed
U
1
(positive electrode) 0.5(1 ) V (vs. SCE) Jow and Zhang
25
U
1
(negative electrode) 0.5 V (vs. SCE) Jow and Zhang
25
Table II. Model parameters that describe the electrodes
and separator.
Parameter Value Reference
D 5, 10, 30 nm Assumed
0.25 Assumed
s
0.7 Pillay and Newman
19
L L
5 10
3
cm Assumed
L
L
s
2.5 10
3
cm Assumed
C
d
2 10
5
F/cm
2
Assumed
s
v
1.5 10
6
, 4.5 10
6
, and From Eq. 26
9.0 10
6
cm
2
/cm
3
for the
particle sizes of D 30, 10,
and 5 nm, respectively
h 4.0 10
8
cm Pollak and OGrady
31
Q
f,oxd
150.2, 450.6, and 901.2 C/cm
3
for From Eq. 27
the particle sizes of D 30, 10,
Q
f,red
0 Assumed
10
5
S/cm Trasatti and Lodi
32
0
0.8 S/cm Darling
33
p
0.1 S/cm From Eq. 28
s
0.469 S/cm From Eq. 29
T 298.15 K Assumed
7.5 10
3
, 2.25 10
2
, 4.5 10
2
s From Eq. 14
for the particle sizes of D 30, 10,
V
o
0.5 V (vs. SCE) Assumed
DL process alone (i.e., faradaic process is shut off) with those gen-
erated from a completely analytical model for DL capacitors devel-
oped by Srinivasan et al.
17
Some dimensionless variables are also
introduced, such as dimensionless cell potential, *
cell
, dimension-
less local potential, E*, dimensionless time, *, dimensionless cell
current, i*
cell
, etc. The definitions of these dimensionless variables are
given in the List of Symbols. Finally, the results presented here are
for discharge only, because the results for charge are simply the mir-
ror image of those for discharge.
Figure 2A, B, and C compares the results for the change in the
dimensionless cell potential with time at two different cell currents,
the dimensionless potential distribution at different discharge times,
and the dimensionless DL current distribution at different discharge
times, respectively. In Fig. 2A, the dimensionless cell potential,
*
cell
, is defined as
cell
/2V
o
, where
cell
represents the matrix po-
tential difference between the two current collectors. Also, the rela-
tive utilization,
17
*i*
cell
, represents the fraction of the available
capacity extracted from the cell. The RuO
2
xH
2
O particles are 5 nm
in diameter. The results show that the numerical solutions agree very
well with the analytical solutions up to the fourth digit, which is suf-
ficient for numerical simulations. It is noteworthy that this numeri-
cal accuracy is achieved by taking 101 space node points and a time
step of 10
6
s up to 2 s and 10
4
s thereafter. All of the results pre-
sented were obtained from the numerical model using these dis-
cretization grid sizes.
Figure 3A shows discharge curves for the combined DL and
faradaic processes at a constant cell current of i*
cell
0.001 for three
different RuO
2
xH
2
O particle sizes, and Fig. 3B compares the dis-
charge behavior of each of their contributions for a capacitor with
5 nm diam RuO
2
xH
2
O particles at a constant cell current of i*
cell

0.001 to examine how the DL and faradaic processes separately
affect the cell potential. Linear relationships of the cell potential
with discharge time are observed in Fig. 3A for all three particle
Figure 2. Comparisons of the numerical solutions with the analytic ones for
the DL process alone and 5 nm diam RuO
2
xH
2
O particles: (A) constant-cur-
rent discharge curves at two dimensionless currents; (B) potential distribution
through the positive electrode at different times during discharge at a constant
current of i*
cell
0.5; and (C) DL current distribution through the positive elec-
trode at different times during discharge at a constant current of i*
cell
0.5.
Figure 3. (A) Effect of RuO
2
xH
2
O particle size on constant-current dis-
charge behavior for the combined DL and faradaic processes at i*
cell
0.001
and (B) comparison of constant-current discharge behavior for the DL
process alone, faradaic process alone, and combined DL and faradaic pro-
cesses at i*
cell
0.001 and with 5 nm diam RuO
2
xH
2
O particles.
sizes, and the discharge time increases with a decrease in the parti-
cle size. The actual discharge time is used here instead of relative uti-
lization (*i*
cell
), because the DL time constant () depends on the
particle size, which makes the discharge curves incomparable in
terms of relative utilization. Linear relationships are again observed
in Fig. 3B for the DL process alone, faradaic process alone, and the
combined DL and faradaic processes; and the DL process alone has
the shortest discharge time, the faradaic process alone has a dis-
charge time five times longer than the DL process alone, and the
combined DL and faradaic processes has the longest discharge time,
appearing as a linear combination of the two individual processes.
The linear behavior of the change in cell potential with time is char-
acteristic of a DL capacitor when the DL capacitance is independent
of the potential. As for the faradaic process simulated here, the lin-
ear behavior is most likely due to the linear relationship between the
equilibrium potential and the state of charge, which makes this
faradaic material of RuO
2
xH
2
O behave like a simple DL capacitor.
Figure 4A and B illustrates the effect of cell current on the cell
potential for the combined DL and faradaic processes and for the
faradaic process alone, respectively. The RuO
2
xH
2
O particles are
5 nm in diam. For the combined DL and faradaic processes
(Fig. 4A), at the smaller current (i.e., i*
cell
0.001), the cell poten-
tial decreases linearly with time. However, at the larger current (i.e.,
i*
cell
0.5), a sharp drop in potential at the start of discharge is ob-
served, which is followed by a linear decrease in potential for the
remainder of the discharge. These discharge curves are similar in
shape to those for the DL process alone, as shown in Fig. 2A, but the
potential drop is larger for the combined DL and faradaic processes.
For the faradaic process alone (Fig. 4B), at the smaller current, the
cell potential also decreases linearly as those observed for the com-
bined DL and faradaic processes (Fig. 4A) and for the DL process
alone (Fig. 2A). At the larger current, however, the cell potential
behaves differently from that for the DL process alone. The potential
decreases almost linearly with time but with a marked drop at the
start of discharge, unlike the initial potential drops for the DL
process alone that exhibit curvature. This marked potential drop is
believed to be associated with the faradaic kinetic resistance; the
result is consistent with a larger drop in potential for the combined
DL and faradaic processes than that for the DL process alone.
To examine how potentials and currents inside the electrode
change during discharge, the distributions of the potential through-
out the positive electrode at different discharge times at a cell current
of i*
cell
0.5 are shown in Fig. 5A and B for the combined DL and
faradaic processes and for the faradaic process alone, respectively.
The corresponding distributions of the currents are illustrated in
Fig. 6A and B for the combined DL and faradaic processes and in
Fig. 6C for the faradaic process alone. For the combined DL and
faradaic processes (Fig. 5A), the potential distribution changes grad-
ually from the initial uniform distribution at the start of discharge to
a steady-state nonuniform profile at about *i*
cell
0.222. Then the
cell potential decreases with this steady-state profile until the cell is
totally discharged. The nonuniform potential distribution is due
mainly to the ohmic potential drop in the solution phase (the con-
ductivity in the matrix phase is much greater than that in the solution
phase in this modeling simulation), and the transition time to reach
the steady-state profile is determined by the DL time constant.
Figure 4. Constant-current discharge curves at two dimensionless currents:
(A) for the combined DL and faradaic processes and (B) for the faradaic
process alone.
Figure 5. Potential distributions throughout the positive electrode at different
times during discharge at a constant current of i*
cell
0.5: (A) for the com-
bined DL and faradaic processes and (B) for the faradaic process alone.
Again, these distributions are similar in shape to those at the same
discharge times for the DL process alone, as shown in Fig. 2B, but
the decrease in potential is slower for the combined DL and farada-
ic processes because it holds more energy compared with the DL
process alone. For the faradaic process alone (Fig. 5B), the potential
distribution changes immediately to a steady-state nonuniform pro-
file after the start of the discharge, then the potential decreases with
this profile until the cell is fully discharged. The reason for this
immediate change in potential distribution is that in this case there is
no DL time constant that controls the transition of the potential dis-
tribution to gradually reach a steady-state profile. Again, the nonuni-
form potential distribution is largely due to the ohmic resistance in
the solution phase.
For the combined DL and faradaic processes, the DL current dis-
tribution throughout the electrode (Fig. 6A) is very nonuniform at
short times, and the degree of this nonuniformity decreases with
time. The DL current attains a uniform profile at about *i*
cell

0.222, and then decreases continuously with this uniform profile
until the cell is totally discharged at *i*
cell
3.7 and it reaches 0.2.
These distributions are also similarly shaped to those for the DL
process alone, as shown in Fig. 2C. Meanwhile, the faradaic current
(Fig. 6B) increases with time and its distribution reaches a maximum
of nonuniformity at about *i*
cell
0.222. Finally, a uniform distri-
bution is attained and the faradaic current reaches 0.8 at *i*
cell

3.7, which is four times greater than the final current for the DL
process alone, as shown in Fig. 6A. For the faradaic process alone
(Fig. 6C), the faradaic current distribution is at a maximum of non-
uniformity at the start of discharge, then the degree of this nonuni-
formity decreases with time, and finally it reaches a uniform profile
at *i*
cell
2.8.
Because local potentials are related to local currents, a change in
local potential must result in a change in local current. However, for
the DL and faradaic processes alone, they are related in different
ways. The DL current is proportional to the derivative of the local
potential with respect to time (Eq. 24), but the faradaic current is
nonlinearly related to the local potential with an exponential depen-
dence (Eq. 25 and 4). These different features for the two processes
determine their different current distributions in response to a
change in local potential. For the DL process alone (Fig. 2B and C),
the change in local potential is the greatest at the start of discharge;
as a result, the DL current distribution reaches a maximum of
nonuniformity. As the change in local potential becomes smaller
with time, the DL current distributions become more uniform. Final-
ly, as the potential distribution attains a steady-state profile, the
change in local potential with time is the same throughout the elec-
trode, which leads to a uniform DL current distribution. For the
faradaic process alone (Fig. 5B and 6C), the change in local poten-
tial is also the greatest at the start of discharge, which leads to a very
nonuniform, faradaic current distribution; meanwhile, the potential
distribution reaches a steady-state profile instantly. This steady-state
potential profile does not translate into a uniform faradaic current
distribution because of the nonlinear relationship between the local
potential and the local current as shown in Eq. 4. The final uniform
faradaic current distribution is likely reached because the difference
between the two exponential terms approaches a constant. For the
combined DL and faradaic processes, the potential and the DL cur-
rent distributions are similar to those for the DL process alone; its
faradaic current distributions are similar to those for the faradaic
process alone. However, the relative magnitudes of the DL and fara-
daic currents are determined by how they are related in the govern-
ing equations (Eq. 4, 24, and 25).
The results in Fig. 3A show that the particle size is extremely im-
portant to the discharge behavior of the capacitor. It is therefore of in-
terest to investigate the performance of this model capacitor in terms
of Ragone plots, i.e., in terms of the energy and power densities. The
energy and power densities are calculated as follows
[30]
[31]
where I
cell
is the total cell current, which is obtained by multiplying
the cell current density (i
cell
) by the cross-sectional area of the elec-
trode, A.
ave
is the average cell potential, t
d
is the discharge time,
and V
cell
is the volume of the cell including the electrodes, elec-
trolyte, and separator. The average cell potential is given by
Power density
cell ave
cell
I
V
Energy density
cell ave d
cell
I t
V
Figure 6. Current distributions throughout the positive electrode at different

times during discharge at a constant current of i*
cell
0.5: (A) and (B) for the
combined DL and faradaic processes, and (C) for the faradaic process alone.
[32]
Figure 7A shows Ragone plots for the combined DL and farada-
ic processes and for three different particle sizes. The curves are all
similarly shaped, but the energy density decreases with an increase
in particle size over the entire range of power densities, and it begins
to drop off sooner but more slowly as the power density increases.
To examine how the DL and faradaic processes independently affect
the energy and power densities in a capacitor, Fig. 7B shows the con-
tributions from each process for a capacitor with 5 nm diam
RuO
2
xH
2
O particles. The Ragone plots exhibit similar trends for
the combined DL and faradaic processes and the DL and faradaic
processes alone, except at very high powers. For the DL process
alone, the energy density is essentially independent of the power
density up to about 20 kW/L, and then it drops sharply at higher
power densities. In contrast, the energy density decreases a little
faster with an increase in power density for the faradaic process
alone. At very high power densities (above 100 kW/L), however, the
faradaic process fails because the potential in the positive electrode
drops instantly below 0.5 V at the start of discharge, and no energy
can be drawn from the cell at such high power densities. For the
combined DL and faradaic processes, a higher energy density over a
broader range of power densities is due largely to the faradaic pro-
cess; however, it approaches the energy density obtained for the DL
process alone at very high power densities, again indicating that the
contribution from the faradaic process becomes negligibly small at

ave
d
0
cell
d
1
t
dt
t
very high power densities. Overall, these Ragone plots show very
clearly that for a capacitor utilizing both DL and faradaic processes,
a higher energy density results from the faradaic process and a high-
er power density results from the DL process, as compared to either
of the individual processes alone.
Discussion
Both of the DL and faradaic processes studied here are surface
phenomena which take place at the interface between the electrode
and electrolyte. For the DL process, more surface area and a higher
specific capacitance both give rise to more energy storage at the inter-
face. Also, for the surface faradaic process, more surface area gives
rise to more energy storage at the interface. Moreover, the more active
the surface material is toward reversible redox reactions (determined
by the value of the exchange current density, i
0
), the higher the power
density. In this study, however, the specific capacitance of the DL
process and the exchange current density of the faradaic process are
all fixed; so the only variable is the surface area, which is changed by
varying the size of the RuO
2
xH
2
O particles in the electrode. In fact,
some researchers
34
reported much higher DL capacitance of 60 F/
cm
2
for oxide electrode systems based on a proton-exchange mecha-
nism. In this case, the calculated DL current would be expected to be
tripled, which in turn would result in a threefold increase in the ener-
gy density contributed from the DL process.
According to Eq. 26, the surface area is inversely proportional to
the particle size. As the particle size decreases from 30 to 10 to 5 nm,
the surface area increases three times and six times, respectively,
which is proportionally the same as the increase in the discharge
times shown in Fig. 3A. In addition, the results in Fig. 3B show clear-
ly that the discharge time for the faradaic process alone is five times
higher than that for the DL process alone, indicating the importance
of the faradaic process for increasing the energy density of a capaci-
tor. Furthermore, Eq. 30 and 31 state that the energy density is pro-
portional to the total cell current, I
cell
, average cell potential,
ave
,
and discharge time, t
d
, but it is inversely proportional to the cell vol-
ume, V
cell
. The power density is the same as the energy density except
that it is independent of t
d
(which may not be true because
ave
is a
function of t
d
). Although a first glance at these equations suggests that
an increase in cell current results in an increase in energy density,
actually this is not true because the discharge time decreases with an
increase in cell current. The overall effect is that the energy density
decreases with an increase in power density. The results in Fig. 7A
show that the ratio of the increase in energy density at the same power
density is roughly the same as the ratio of the increase in surface area
for the three different sizes of RuO
2
xH
2
O particles.
In addition, in order to increase the power density, the cell cur-
rent density must be increased, as shown in Fig. 7A and B. Howev-
er, the resulting increase in power density is obtained at the expense
of decreasing energy density, mainly because of the ohmic resistance
in the solution phase for the DL process alone. For the faradaic
process alone, the energy loss at higher power densities is due to the
ohmic resistance in the solution phase as well as the faradaic kinet-
ic resistance. The faradaic kinetic resistance even makes the farada-
ic process alone fail, because the cell potential cannot hold at the
working potential range. What is even more interesting, however, is
that for the combined DL and faradaic processes, the energy loss is
less compared with that for the faradaic process alone at the same
power density. The reason is that the DL process has the ability to
hold the cell potential, which allows the faradaic process to deliver
some energy even beyond the power at which the faradaic process
alone would fail. These results are consistent with those shown in
Fig. 2A, 4A, and 4B, which indicate that the energy loss at the larg-
er current (based on relative utilization) is 20, 44, and 37% for the
DL process alone, faradaic process alone, and combined DL and
faradaic processes, respectively. Although there is always a trade-off
between the energy and power densities, a capacitor with both DL
and faradaic processes can store more energy and deliver more
power compared with that obtained from either the DL or faradaic
processes alone.
Figure 7. Regone plots: (A) for the combined DL and faradaic processes for
three RuO
2
xH
2
O particle sizes and (B) for the DL process alone, faradaic
process alone, and combined DL and faradaic processes with 5 nm diam
RuO
2
xH
2
O particles.
In this study, it is also assumed that the faradaic reaction takes
place only at the surface of the particle with one-electron exchange
between 0 and 1 V (SCE). In practice, however, the inner layers
within a particle may also participate in the redox reaction through a
proton diffusion mechanism;
3,10
other authors
35
have suggested a
faradaic reaction of RuO
2
with a two-electron exchange. As a result,
the faradaic charge of a fully oxidized electrode can double accord-
ing to Eq. 27, which means that the energy density from the farada-
ic process can also double. Nevertheless, the power density may de-
crease because of slow proton diffusion in the innerlayers of the
solid phase. In addition, in this study the electronic conductivity of
the matrix is taken as 105 S/cm, which is 3-4 orders of magnitude
higher than the value suggested by some researchers
36
based on
boundary effects associated with powder materials
37
and less crys-
tallinity
38
in hydrous RuO
2
. However, if the conductivity is taken as
this lower value, the simulated results will not be affected much
because the limiting conductivity is in the solution.
Conclusions
A mathematical model of an electrochemical capacitor comprised
of hydrous ruthenium oxide (RuO
2
xH
2
O) electrodes, including both
DL and surface faradaic processes, was developed to predict the
behavior of the capacitor under the conditions of galvanostatic charge
and discharge. A comparison of various numerical solutions from the
model with analytical ones for the DL process alone revealed that the
numerical simulations were accurate up to the fourth digit. The
results show that the energy density can be increased by decreasing
the particle size, whereas the power density can only be increased by
increasing the cell current density. However, there is always a trade-
off between the energy and power densities. The results also indicate
that a capacitor with both DL and faradaic processes has higher ener-
gy densities compared with the DL process alone and higher power
densities compared with the faradaic process alone. The model pre-
dicts that the faradaic process increases significantly the energy per
unit volume of the capacitor for power densities of 100 kW/L or less.
Acknowledgments
The authors are grateful for the financial support provided in part
by the U.S. Army Research Office under grant no. DAAH04-96-1-
0421 and in part by the U.S. Department of Energy under Coopera-
tive Agreement no. DE-FC02-91ER75666.
The University of South Carolina assisted in meeting the publication
costs of this article.
List of Symbols
A cross-sectional area of the electrode, cm
2
C
d
DL capacitance per unit surface area, F/cm
2
D diameter of the RuO
2
xH
2
O particles, cm
E local electrode potential,
1
-
2
, V
E* dimensionless local potential, E/2V
o
f F/RT, V
1
F Faradays constant, 96487 C/equiv
h length of the crystal lattice on the RuO
2
xH
2
O surface, cm
i
0
exchange current density for the faradaic reaction, A/cm
2
i
1
superficial current density in the matrix phase, A/cm
2
i
2
superficial current density in the electrolyte phase, A/cm
2
i
c
DL current per unit volume of the electrode, A/cm
3
i
cell
applied cell current density, A/cm
2
I
cell
total cell current, A
i*
c
dimensionless DL current, i
c
L/i
cell
i*
cell
dimensionless cell current, i
cell
L(
p
)/(
p
V
o
)
i
f
faradaic current per unit volume of the electrode, A/cm
3
j
f
faradaic transfer current density, A/cm
2
i*
f
dimensionless faradaic current, i
f
L/i
cell
L thickness of the electrode, L L
, cm
L
A
Avogadros number, 6.0226 10
23
mol
1
L
s
thickness of the separator, cm
Q
f
faradaic charge per unit volume of the electrode, C/cm
3
Q
f,oxd
faradaic charge per unit volume of the fully oxidized electrode, C/cm
3
Q
f,red
faradaic charge per unit volume of the fully reduced electrode, C/cm
3
R universal gas constant, 8.3143 J/mol K
s
v
specific surface area per unit volume of the electrode, cm
2
/cm
3
t time, s
t
d
discharge time, s
T temperature, K
U
1
equilibrium potential for the faradaic reaction, V (vs. SCE)
V
cell
volume of the cell, cm
3
V
o
initial potential before charge, 0.5 V (vs. SCE)
x position coordinate, cm
Greek
a
anodic transfer coefficient of the faradaic reaction
c
cathodic transfer coefficient of the faradaic reaction
state of charge in the faradaic reaction
electrode porosity
s
separator porosity
1
potential in the matrix phase, V
2
potential in the electrolyte phase, V
ave
average cell potential, V
cell
cell potential or the matrix potential difference between the two cur-
rent collectors, V
*
cell
dimensionless cell potential,
cell
/2V
o
0
ionic conductivity of the bulk electrolyte, S/cm
p
ionic conductivity of the electrolyte inside the pores of the electrode,
S/cm
s
ionic conductivity of the electrolyte inside the pores of the separator,
S/cm
fraction of oxidized species in the faradaic reaction
electronic conductivity of the matrix phase, S/cm
DL time constant, s
* dimensionless time, t/
dimensionless position coordinate, x/L
References
1. B. E. Conway, J. Electrochem. Soc., 138, 1539 (1991).
2. B. E. Conway, in Third International Seminar on Double Layer Capacitors and
Similar Energy Storage Devices, Vol. 3, Florida Educational Seminar, Inc., Boca
Raton, FL (1993).
3. S. Trasatti and P. Kurzweil, Platinum Metals Rev., 38, 46 (1994).
4. S. Sarangapani, B. V. Tilak, and C. P. Chen, J. Electrochem. Soc., 143, 3791 (1994).
5. I. D. Raistrick, in Electrochemistry of Semiconductors and Electrodes, J. McHardy
and F. Ludwig, Editors, pp. 297-355, Noyes Publications, Park Ridge, NJ (1992).
6. H. Shi, Electrochim. Acta, 41, 1633 (1996).
7. M. Aoki, K. Sato, and Y. Kobayashi, IEICE Trans. Fundam., E77-A, 208 (1994).
8. S. T. Mayer, R. W. Pekala, and J. L. Kaschmitter, J. Electrochem. Soc., 140, 446
(1993).
9. I. Tanahashi, A. Yoshida, and A. Nishino, Carbon, 29, 1033 (1991).
10. J. P. Zhang, P. J. Cygan, and T. R. Jow, J. Electrochem. Soc., 142, 2699 (1995).
11. K. C. Liu and M. A. Anderson, J. Electrochem. Soc., 143, 124 (1996).
12. V. Srinivasan and J. W. Weidner, J. Electrochem. Soc., 144, L210 (1997).
13. C. Lin, J. A. Ritter, and B. N. Popov, J. Electrochem. Soc., 145, 4091 (1998).
14. F. A. Posey and T. Morozumi, J. Electrochem. Soc., 113, 176 (1966).
15. A. M. Johnson and J. Newman, J. Electrochem. Soc., 118, 510 (1971).
16. W. Tiedemann and J. Newman, J. Electrochem. Soc., 122, 70 (1975).
17. V. Srinivasan, C. Lin, J. A. Ritter, and J. W. Weidner, in Electrochemical Capaci-
tors II, F. M. Delnick, D. Ingersoll, X. Andrieu, and K. Naoi, Editors, PV 96-25,
p. 153, The Electrochemical Society Proceedings Series, Pennington, NJ (1996).
18. C. J. Farahmandi, in Electrochemical Capacitors II, F. M. Delnick, D. Ingersoll, X.
Andrieu, and K. Naoi, Editors, PV. 96-25, p. 167, The Electrochemical Society Pro-
ceedings Series, Pennington, NJ (1996).
19. B. Pillay and J. Newman, J. Electrochem. Soc., 143, 1806 (1996).
20. O. S. Ksenzhek, Russ. J. Phys. Chem., 37, 1089 (1963).
21. S. Sarangapani, B. V. Tilak, and C.-P. Chen, J. Electrochem. Soc., 143, 3791 (1996).
22. M. Doyle, T. F. Fuller, and J. Newman, J. Electrochem. Soc., 140, 1526 (1993).
23. J. Newman and W. Tiedemann, AIChE J., 21, 25 (1975).
24. P. De Vidts and R. E. White, J. Electrochem. Soc., 144, 1343 (1997).
25. T. R. Jow and J. P. Zhang, J. Electrochem. Soc., 145, 49 (1998).
26. T. Yeu and R. E. White, J. Electrochem. Soc., 137, 1327 (1990).
27. R. E. White, S. E. Lorimer, and R. Dardy, J. Electrochem. Soc., 130, 1123 (1983).
28. M. E. Davis, Numerical Methods and Modeling for Chemical Engineers, John
Wiley & Sons, New York (1984).
29. J. S. Newman, Electrochemical Systems, 2nd ed., Prentice-Hall, Inc., Englewood
Cliffs, NJ (1991).
30. K. Kimoshita, Carbon: Electrochemical and Physicochemical Properties, John
Wiley & Sons, New York (1988).
31. F. H. Pollak and W. E. OGrady, J. Electrochem. Soc., 132, 2385 (1985).
32. S. Trasatti and G. Lodi, in Electrodes of Conductive Metallic Oxides-Part A, p. 301,
S. Trasatti, Editor, Elsevier, New York (1980).
33. H. E. Darling, J. Chem. Eng. Data, 9, 421 (1964).
34. S. Levine and A. L. Smith, Disc. Faraday Soc., 52, 290 (1971).
35. D. Michell, D. J. A. Rand, and W. Woods, J. Electrochem. Soc., 89, 11 (1978).
36. S. Pizzini, G. Buzzanca, C. Mari, L. Rossi, and S. Torchio, MRS Bull., 7, 449
(1972).
37. S. Trasatti, Electrochim. Acta, 36, 225 (1991).
38. W.-C. Shin and S.-G. Yoon, J. Electrochem. Soc., 144, 1055 (1997).

A Mathematical Model of An Electrochemical Capacitor With Double-Layer and Faradaic Processes

Uploaded by

Copyright:

Available Formats

A Mathematical Model of An Electrochemical Capacitor With Double-Layer and Faradaic Processes

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Mathematical Model of An Electrochemical Capacitor With Double-Layer and Faradaic Processes

Uploaded by

Copyright:

Available Formats

3168 Journal of The Electrochemical Society, 146 (9) 3168-3175 (1999)

S0013-4651(98)11-057-1 CCC: $7.00 The Electrochemical Society, Inc.

Figure 6. Current distributions throughout the positive electrode at different

You might also like