Effects of pore surface oxidation on electrochemical and mass-transport properties of nanoporous carbon

Article abstract of DOI:10.1149/1.1843772

A new nanoporous (NP) carbon material with a high surface area and a narrow pore size distribution, around 8 A, has been used to investigate the effects that electrochemical oxidation at positive potentials exerts on the capacitance values and effective diffusion coefficients of ions inside the nanopores. An electroanalytical method, based on the single-particle microelectrode technique with micromanipulator, was applied to calculate the diffusion coefficients of 6 M KOH ions in NP carbon. The results were analyzed for short times using the Cottrell model and for long times using the spherical diffusion model. Using cyclic voltammetry, was found that different stages of oxidation took place between 0 and 0.5 V vs. Hg/HgO. After repeated cycling in the first region of oxidation (0-0.3 V), an activation leading to higher capacitance was observed, but the diffusion coefficients decreased from approximately 2 × 10^-9 to 0.5 × 10^-10 cm ² s^-1. In the second region of oxidation (0.3-0.5 V), where CO₂ and O₂ evolution can occur, both the capacitance and the diffusion coefficients decreased more dramatically. The effective diffusion coefficients of ions of an activated carbon particle were dependent on the operation potential, decreasing by an order of magnitude when going from -0.3 to +0.3 V. The results are discussed in terms of chemisorption of small oxygen functional groups (-OH or C=O) and ionic interaction with the pore wall.

Full text of DOI:10.1149/1.1843772

A270
Journal of The Electrochemical Society, 152 ͑2͒ A270-A276 ͑2005͒
0013-4651/2004/152͑2͒/A270/7/$7.00 © The Electrochemical Society, Inc.
Effects of Pore Surface Oxidation on Electrochemical
and Mass-Transport Properties of Nanoporous Carbon
z
¨
Marcelo Zuleta, Pehr Bjornbom, and Anders Lundblad
Department of Chemical Engineering and Technology, Royal Institute of Technology,
SE-100 44 Stockholm, Sweden
A new nanoporous ͑NP͒ carbon material with a high surface area and a narrow pore size distribution, around 8 Å, has been used
to investigate the effects that electrochemical oxidation at positive potentials exerts on the capacitance values and effective
diffusion coefficients of ions inside the nanopores. An electroanalytical method, based on the single-particle microelectrode
technique with micromanipulator, was applied to calculate the diffusion coefficients of 6 M KOH ions in NP carbon. The results
were analyzed for short times using the Cottrell model and for long times using the spherical diffusion model. Using cyclic
voltammetry, was found that different stages of oxidation took place between 0 and 0.5 V vs. Hg͉HgO. After repeated cycling in
the first region of oxidation ͑0-0.3 V͒, an activation leading to higher capacitance was observed, but the diffusion coefficients
decreased from approximately 2 ϫ 10Ϫ9 to 0.5 ϫ 10Ϫ10 cm2 sϪ1. In the second region of oxidation ͑0.3-0.5 V͒, where CO2 and
O2 evolution can occur, both the capacitance and the diffusion coefficients decreased more dramatically. The effective diffusion
coefficients of ions of an activated carbon particle were dependent on the operation potential, decreasing by an order of magnitude
when going from Ϫ0.3 to ϩ0.3 V. The results are discussed in terms of chemisorption of small oxygen functional groups ͑—OH
or CvO͒ and ionic interaction with the pore wall.
© 2004 The Electrochemical Society. ͓DOI: 10.1149/1.1843772͔ All rights reserved.
Manuscript submitted January 21, 2003; revised manuscript received July 2, 2004. Available electronically December 17, 2004.
Double layer capacitors ͑DLCs͒ are important electrical power
intercalation or insertion, may take place depending on, for instance,
the applied potential, kind of electrolyte, porosity, history of the
carbon material, cell design, and structure of the carbon material.
It is well known from the development of double-layer elec-
trodes and capacitors that, to achieve optimal performance ͑i.e., ca-
pacitance͒, an electrochemical pretreatment is required. Despite the
frequently used procedures for carbon electrodes, the knowledge
concerning how the positive polarization affects the mass transport
of ions in alkali solutions inside the pores is limited. Furthermore,
several of the existing studies, e.g., for activated carbons, have
mainly been performed using carbon materials with nonuniform
pore size distribution and complex carbon structures.6 However, it is
important to understand the chemistry behind electrochemical oxi-
dation of the carbon materials. Likewise, a comprehension of the
impact of the reactions taking place on the electrode is required for
adequately improving the performance of DLCs. These require-
ments are also valid for other energy power devices like fuel cells
and batteries, which contain carbon materials.
sources in systems such as electrically driven vehicles,1 computers,
and other electronic products.2 Theoretically, the driving force re-
sponsible for the charge and discharge processes in a DLC is the
difference in the electrochemical potential over the electrochemical
double layer. Therefore, the forces acting to reach a minimum in
Gibbs free energy are of pure electrostatic character, in contrast to
the situation for batteries and fuel cells, where energy conversion is
obtained by electrochemical redox reactions. As a result of the re-
versible processes occurring in capacitors, the DLC can be charged
and discharged more than 105 cycles with relatively short response
times.3 The specific capacitance, which for many power applications
must be as high as possible, is proportional to the electrode surface
area. For this reason, the pore size distribution of the carbon plays a
decisive role in achieving high energy storage. Carbon materials
have a relatively wide potential window in aqueous solutions and
are available with high specific surface area; however, the electro-
chemical properties of the carbon electrode are considerably af-
fected by the pore system. For instance, the mass-transport hin-
drance of ions, one of the most important power-limiting properties,
is caused by the morphology of the pore space. Nevertheless, the
potential-current response should, in galvanostatic mode, be sym-
metrical for both positive and negative polarization ͑see Fig. 1a͒.
However, carbon materials, such as the nanoporous carbon ͑NC͒
material used here, do not follow ideal capacitor behavior when
Oxidation of Carbon
Ross and Sattler found that a high degree of ordered graphite-like
structure leads to less oxidation of the carbon material ͑i.e., the
oxidation occurs at the edge planes rather than on the basal planes
of the graphite structure͒.7 They also found that the formation of
carbon corrosion products ͑CO and CO2) together with oxygen
evolution on acetylene black started at positive potentials larger than
0.4 V against Hg͉HgO in 30% KOH solution.8 Several investiga-
tions, made in alkaline solutions and using different carbon
materials,9-14 have found that a controlled polarization leads princi-
pally to moderate oxidative processes that do not change the surface
morphology of the carbon material. Instead, the moderate oxidation
leaves different oxygen functional groups on the carbon surface,
predominantly hydroxyl groups ͑i.e., phenolic͒ and lower quantities
Ͼ
used as a positive electrode ͑defined by E
Ϫ0.3 V vs. Hg͉HgO
in Fig. 1b͒. However, the NC acting as a negative electrode follows
an idealized behavior. Comparing the slopes, dE/dt, from which
the specific capacitance, Cs values can be calculated
C
s
͓
ϭ i/(⌬E/⌬t) , it may be seen that these values are not the same
͔
for this carbon material for all potentials. Thus, this is an indication
that other processes besides simple electrostatic forces are at hand.
The deviations from the ideal behavior may be explained as the
result of an extra storage of charges due to faradaic reactions and/or
specific adsorption processes taking place on the surface of the car-
bon material. Thus, the total capacitance is made up of both a con-
tribution from the double layer and another, as pseudocapacitance,
from the functional groups added to the carbon surface. The electro-
chemical events occurring at positive polarization in carbon materi-
als are very complex.4,5 Different types of adsorption ͑physisorption,
chemisorption͒, oxidation reactions, and other processes, such as
of carbonyl groups ͑e.g., of quinone type R͑CvO͒—R ͑CvO͒͒,
Ј
but even some minor amounts of carboxyl groups ͑R—͑OH͒CvO͒
and lactones ͑R—O—͑R ͒CvO͒. The mild oxidation, produced
Ј
from electrochemical or even thermal treatments, gives, apart from
improved wettability,15 a positive effect on the electrochemical
properties of the carbon material, particularly enhanced specific
capacitance.16 The analyses of carbon oxides by thermal desorption
of CO2 and CO from activated carbons17-19 have shown that the
groups responsible for the higher specific capacitance come mainly
from oxidized sites which contain only one oxygen, i.e., carbonyl,
hydroxyl, and quinone groups, rather than from the functional
z E-mail: anders.lundblad@ket.kth.se
Downloaded on 2015-03-08 to IP 132.174.255.116 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 152 ͑2͒ A270-A276 ͑2005͒
A271
Figure 1. The potential-time response of a DLC during galvanostatic cy-
cling: ͑a͒ ideal behavior of porous carbon materials acting as positive and as
negative electrode, respectively, and ͑b͒ charging and discharging an NC
particle polarized at positive and negative potentials. Current: Ϯ1 ␮A, par-
ticle size: 150 ␮m.
Figure 2. Pore size distribution measurements of carbon powder from SiC,
by gas porosimetry, according to Zheng et al.37
groups containing two oxygen atoms, like carboxyl, lactones, and
anhydrides. A stronger oxidation causes degradation processes of the
carbon surface, involving evolution of CO and CO2 . Apart from the
fact that corrosion reactions result in a change of the morphology of
the carbon surface, they negatively affect the charge and discharge
properties because carbon dioxide can easily be absorbed in the
KOH electrolyte and form carbonate ions, CO23Ϫ . These ions might
cover the microporosity of the carbon material and obstruct or retard
the access of ions in and out of the pores, causing a loss of energy
storage capacity. In comparison with alkaline electrolytes, the oxi-
dation of carbon in acid solutions has shown greater complexity of
the processes occurring at positive polarization, for instance, a
higher tendency to further oxidation, leaving carbonyl and carboxyl
groups and CO2 ,12,20 but also intercalation of ions inside the carbon
structure,21 a process which might in turn cause carbon corrosion.9
This fact reflects the advantage of working with alkaline solutions
when studying the effects that positive polarization exerts on mass-
transport properties of ions in aqueous solutions inside carbon nan-
opores, principally when oxygen functional groups responsible for
higher capacitance values are present on the carbon surface.
size. The total porosity can reach up to 70% of the total solid vol-
ume and the specific surface area is larger than 1200 m2/g. More
information regarding the synthesis pathway of skeleton carbon ma-
terials obtained from carbides ͓also called carbide-derived carbons
͑CDCs͔͒ can be found in Ref. 24 and, particularly for the nanopo-
rous carbon material obtained from silicon carbide and the electrode
manufacturing, in Ref. 25 and 26. The structure analysis and appli-
cation of this material as a DLC can be found in the article by Zuleta
et al.22 and the references therein.
The micromanipulated single-particle microelectrode tech-
nique.—The single-particle microelectrode technique, developed by
23
¨
Bursell and Bjornbom, is a relatively new and advantageous tech-
nique for the study of electrochemical phenomena. The implemen-
tation of a micromanipulator provides good control of a carbon fiber
probe by which a small particle ͑20-200 ␮m͒ is contacted, thus
enabling the formation of a single-particle microelectrode. One of
the most important advantages using this technique is the elimina-
tion of resistances, i.e., reduced macrodiffusion and resistance po-
larization, but also the avoidance of supplementary electrode mate-
rials ͑e.g., binder or current collector͒. Ura et al. have demonstrated
the utility of this technique for determining diffusion coefficients of
hydrogen in metal hydrides.27
Objectives
In an earlier work, we modified a single-particle microelectrode
method using carbon particles in the micrometer scale in aqueous 6
M KOH.22 This technique was originally developed for characteriz-
ing agglomerates of fuel cell electrodes.23 We applied it together
with mathematical models as an electroanalytical method for calcu-
lation of effective diffusion coefficients. Here we investigate the
effects that the positive polarization has on the specific capacitance
values and on the effective diffusion coefficients for ions inside the
nanopores of a relatively new carbon material obtained from silicon
carbide. The carbon particles consist entirely of nanopores with a
narrow pore size distribution, around 8 Å in pore size. This unique
feature makes this material very useful as a model system for inves-
tigations of phenomena related to the microporosity of carbon.
The cell, electrolyte, reference electrode, and measurement
system.—The cell used in the microelectrode technique consists of a
12 cm tray made of plexiglass, with a 6 mm diam hole cover for
nitrogen outflow and a supplementary smaller hole for nitrogen in-
flow. This construction permits an inert atmosphere inside the cell as
well as a border for external contact with the carbon fiber and the
probe. A nickel wire was used as counter electrode and the regis-
tered potentials referred to the Hg͉HgO reference electrode im-
mersed in 6 M KOH; all the other potential values in this work
are referred to this electrode. Figure 3a describes the principle of
the experimental setup. The electrolyte used in the cell was 6 M
KOH bubbled previously in nitrogen. The electrochemical mea-
surement system used in the experiments was Autolab II supplied
by Eco Chemie BV. All the experiments were carried out at room
temperature ͑22°C͒.
Experimental
The NC particles.—The NC material used was obtained from
silicon carbide powder, which was provided in different particle
sizes by Norton AS, and exposed to thermal treatment, in which it
undergoes a heterogeneous gas reaction with chlorine at 700-
1000°C. The result is the reaction of the silicon atoms with chlorine
according to the reaction
The thermal pretreatment and capacitance values.—To wet the
internal surface of the material and to remove any gases present in
its interior, the carbon particles were boiled four times in 6 M KOH
for 15 min and exposed to vacuum conditions for 15 min. Deminer-
alized water was added to compensate for the water lost during
boiling. We assume for our model that the pretreatment leads to
complete wetting of the particles. The capacitance values are ob-
tained by integration of the I-E response from the cyclic voltammo-
gram ͑CV͒, thus giving the charge in coulombs divided by the po-
tential interval for integration. Finally, the capacitance value divided
by the particle volume gives the volumetric capacitance.
SiC ϩ 2Cl2 g͒ ϭ SiCl g͒ ϩ C ͑s)
͑
͑
4
Argon gas is subsequently used to remove silicon chloride from the
remaining solid product: a rigid skeleton material with carbon atoms
forming
a disordered amorphous nanoporous structure. Gas-
porosimetry measurements, illustrated in Fig. 2, have shown that the
pores in the carbon are 8.2 Ϯ 0.2 Å wide and extremely uniform in
Downloaded on 2015-03-08 to IP 132.174.255.116 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A272
Journal of The Electrochemical Society, 152 ͑2͒ A270-A276 ͑2005͒
toward the center of the particle. According to Fick’s second law, the
diffusion equation in the case of linear diffusion can be written as
ץCKϩ
ץt
ץ2CKϩ
ץz2
Ϫ Deff
ϭ 0
͓1͔
Regarding a potential step at negative potentials, for instance, E
ϭ 0.0 to Ϫ0.1 V vs. Hg͉HgO, it is possible to express the change in
concentration of potassium ions, ⌬CϩK ϭ C1ϩ,K Ϫ C0ϩ,K at equilib-
rium, in terms of the specific capacitance per volume of particle, Cs ,
and the change in potential (⌬V) divided by the Faraday constant
͑F͒
Cs⌬V
⌬CKϩ
ϭ
͓2͔
F
The solution of Eq. 1 deduced with the appropriate initial and
boundary conditions defined along the pore length, coordinate z,
from the pore mouth (z ϭ 0), can be expressed as follows
Figure 3. Schematic drawing of the equipment used in the experimental
part. ͑a͒ Single-particle experimental setup: ͑1͒ microscope, ͑2͒ Hg͉HgO ref-
erence electrode, ͑3͒ gas inflow, ͑4͒ micromanipulator, ͑5͒ gold probe, ͑6͒
carbon fiber connected to the microparticle, ͑7͒ the cell, ͑8͒ Ni wire, ͑9͒
nitrogen gas outflow, ͑10͒ separator. ͑b͒ Sketch of the cell used for the
Raman measurements: ͑1͒ laser source, ͑2͒ plexiglass plate with perforation,
͑3͒ current collectors, ͑4͒ plexiglass plate with channel for reference elec-
trode, ͑5͒ sealing, ͑6͒ Ni wire, ͑7͒ separator, ͑8͒ NC electrode, ͑9͒ glass.
At t ϭ 0 for all z: C ϭ Cϩ0,K ϭ 0
ϩ
Ͼ
At t
0 z ϭ 0: C ϭ ⌬CK
,
and
at z ϭ ϱ: C ϭ 0
Expressing concentration in specific capacitance according to Eq. 2
leads, for linear diffusion, to the Cottrell relation, Eq. 3. More de-
tails regarding the mathematics are given in the paper by Zuleta
et al.22
Raman spectrophotometry measurements.—A Raman spectro-
photometer, Renishaw 2000, with a laser wavelength of 514 nm was
used with a microscope, Leica DMLM; the penetration depth of the
laser is approximately 100 nm. The software WIRE was used to
register the spectra. The samples were electrodes, consisting of na-
noporous carbon with pyrolitic carbon as binder; after polishing, the
electrodes were pretreated in potassium hydroxide as described pre-
viously for the carbon particles. The electrodes were later activated
from 0.0 to 0.3 V against Hg͉HgO in a separate cell. The cell con-
sisted of two plexiglass plates ͑see Fig. 3b͒; one of the plates had a
perforation of 15 mm in the middle to make focusing the laser at the
NC electrode easier. The working and counter electrodes were
pressed together with the polypropylene separator, the electrical
conductors, which consisted of sheets of stainless steel, the objective
glass, and finally a sealing.
1/2
Deff
I ϭ AC ⌬V
͓3͔
ͩ ͪ
s
␲t
Deff in this equation represents the effective diffusion coefficient in
cm2 sϪ1, t the time elapsed, and A the outer surface area. Thus, by
plotting the current I vs. tϪ1/2, the effective diffusion coefficient may
be determined from the slope of the linear relation.
The Cottrell equation is applicable for the diffusion layer gener-
ated at or close to the interface between the carbon particle and the
electrolyte solution. This asymptotic solution of Eq. 1 is valid only
when the diffusing ions still have penetrated only a small fraction of
the radius of the particle. In our potential step experiments, this
typically occurs in the time interval 10-90 s ͑before 10 s initial
background processes invalidate measured data͒. However, in poten-
tial regions where the electrode material does not behave totally
capacitively, the charging or discharging processes may easily be
affected by other events, e.g., faradaic reactions occurring simulta-
neously during the charge of the double layer. Under these circum-
stances, the diffusion layer may be perturbed by the interionic at-
traction or repulsion between the ion species in the solution and the
added surface groups on the carbon surface. Hence, determining the
diffusion coefficients in the long-time region, using the diffusion
theory of a single sphere with radius a, is considered
The mathematical model used for determination of diffusion
coefficients.—Perturbations of the potential of a carbon particle
cause an electrical current to flow followed by a movement of ions
toward the particle surface. Both diffusion and migration are estab-
lished mechanisms for the ions to approach the external particle
surface. According to Koresh and Soffer,16 the ions must deplete,
partially or completely, the hydration shell, to penetrate the nanop-
ores. A stronger adsorption of the ions is expected in the nanopores,
because the adsorption energy in carbon micropores is higher com-
pared with that on a flat surface. This effect is particularly appre-
ciable in pores smaller than 10 Å.28 Inside the pores, two cases of
transport can be conceived, one in which adsorbed ions, associated
to a mirror charge on the material side, move along the pores as
neutral species against a concentration gradient. In a second case,
ions penetrate the pores driven by a potential gradient. This is the
topic of a separate paper. In this work, we assume a transport of ions
corresponding to the first case, i.e., the ions become adsorbed on the
pore wall and move inside the pores by diffusion.
2ADeff Cs⌬V
Ϫ ␲2Deff
t
I ϭ
exp
͓4͔
ͩ
ͪ
a2
a
Equation 4 is an asymptotic solution of Eq. 5 and is valid from the
time when the diffusing ions have reached the center of the spherical
particle (r ϭ 0)
ץCKϩ
ץt
ץ2CKϩ
ץr2
2 ץCKϩ
The electroanalytical method used for calculation of diffusion
coefficients, when using the potential step method, starts from the
assumption of semi-infinite mass transport from and to a thin layer
inside the nanopores located adjacent to the outer surface. As men-
tioned, we regard the adsorbed ions as neutralized by charges at and
near the nanopore walls. Thus, the principal mechanism for charging
of the DLC is diffusion of adsorbed species along the pores and
Ϫ Deff
ϩ
ϭ 0
ͩ
ͪ
r
ץr
At t ϭ 0 for all r: C ϭ Cϩ0,K ϭ 0
ץCKϩ
ץr
·
ϩ
Ͼ
At t
0 r ϭ a: C ϭ ⌬CK
,
and
at r ϭ 0:
ϭ 0
͓5͔
Downloaded on 2015-03-08 to IP 132.174.255.116 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 152 ͑2͒ A270-A276 ͑2005͒
A273
Figure 5. Voltammetric curves recorded for particle A, activated from 0 to
0.3 V, and particle B, activated from 0 to 0.5 V. Scan rate 0.2 mV sϪ1
Figure 4. CV for an NC particle approximately of size 70 ␮m. Scan rate
0.1 mV sϪ1
.
.
Equation 4 is valid after times of typically 100 s in our potential step
experiments for particles of 50-100 ␮m. Some authors have sug-
gested that this relation is most usable for the calculation of the
diffusion coefficients. This is because the slope from the linear re-
lation, when plotting ln I vs. t, is less sensitive to background cur-
rents compared with the total current at the short-time region.29
When the potential steps are performed at negative potentials, where
the capacitor follows an ideal behavior, the diffusion coefficients
obtained from the experimental data using the Cottrell and the radial
diffusion equations have shown good agreement.22
evolution, as well as other corrosion processes of the carbon surface,
start for many carbon materials in this domain. The occurrence of
these processes can explain the increase of the current peak; how-
ever, no perceivable evidence of gas evolution could be detected in
the microscope at 0.5 V at room temperature.
Activation of the NC.—The CV for the particle described above
clearly shows that different processes are occurring in the three do-
mains. Consequently, it is expected that, after repeated polarization,
the electrochemical properties of the NC have to be affected in a
certain way, depending on in which of the domains, shown in Fig. 4,
the capacitor electrode is working. In all three cases, the perfor-
mance and charge-discharge responses of the DLC must be optimal
and stable for the given potential window. Figure 5 shows some of
the CVs for the activation of two selected carbon particles, particle
A, with an average diam of 82 ␮m, and particle B, of average diam
62 ␮m. The first-mentioned particle was activated during 23 cycles
to 0.3 V, that is, a potential below the limits of domain 2, while
particle B was activated, 13 cycles, to 0.5 V, the end potential of
domain 3. The scan rate, 0.2 mV/s for both particles A and B, is
twice that used in the previous experiment, illustrated in the CV in
Fig. 4, as it allows the particles to be activated many cycles in a
reasonable experimental time. However, as seen in the first cycle,
the plateau that was previously present in domain 2 is less pro-
nounced at this sweep rate.
After repeated cycling, the current peaks decrease for both acti-
vation procedures. This current reduction indicates a consumption
of, for instance, active sites on the carbon surface similar to a chemi-
sorption process, rather than only a reversible faradaic reaction. This
consumption seems to be more significant at 0.5 than at 0.3 V. The
current decreases by 66% from the first to the 23rd cycle for particle
A, whereas it decreases by more than 80% after only the 13th cycle
for particle B.
Results
Current-potential response.—Cyclic voltammetry, combined
with microelectrode technique, provides an exceptional way to
monitor different physicochemical processes and faradaic reactions
generated at specific potentials. The use of microelectrodes involves
two key aspects for cyclic voltammetry. The first is the small mea-
sured currents (10Ϫ6-10Ϫ12 A͒, which lead to a reduction of the
effects of IR drop. The second aspect is the charging current; it
decreases exponentially, while the time-dependent faradaic current
decreases linearly with electrode size. Consequently, an improve-
ment of the faradaic/nonfaradaic current ratio can be reached.30,31
Furthermore, a very slow sweep rate was applied in the experiments,
which permits a more pronounced reduction of the capacitive cur-
rent (Ic ϰ v) in comparison with the reduction of the possible fara-
daic currents (If ϰ v1/2).31 In addition, the low sweep rates allow for
the ion species to penetrate the pores and interact with the inner
surface of the pores. The advantage of using a single particle instead
of a conventional electrode for voltammetric studies has been illus-
trated by Kim et al.,32 who obtained clearly separated peaks for the
single particle as opposed to the signals of a film electrode.
Figure 4 shows the first CV taken after the pretreatment of the
NC particles ͑i.e., after the particles are boiled in 6 M KOH and
vacuum treated͒. The selected particle had a size of approximately
70 ␮m, and it was exposed to positive polarization between 0 and
0.5 V at a scan rate of 0.1 mV/s. As seen in the figure, three different
I-E domains can be distinguished from the CV. In domain 1, for the
polarization between 0 to 0.1 V, the current is more or less constant
and stems principally from the double-layer capacitance. In domain
2, the current increases dramatically up to a potential of 0.36 V vs.
Hg͉HgO. It remains constant thereafter in a potential interval be-
tween 0.36-0.41 V. This plateau in domain 2 is not observable at
higher sweep rates than 1 mV/s. In domain 3, a new increase of the
current takes place. It can be seen that the forward sweep yields a
lower current compared with that of the backward sweep. This sug-
gests that the equilibrium potential has not been completely reached
inside the pores, and that the process continues occurring during the
reverse sweep. It is well known from the literature that oxygen
Effects of positive polarization on specific capacitance.—The ca-
pacitance values were determined between 0 and Ϫ0.2 V, a potential
window where CVs have shown the typical electrochemical behav-
ior of a DLC. The specific capacitance was obtained from the mea-
sured value divided by the particle volume. Note from Fig. 6 that the
activation processes for particles A and B had different impacts on
the capacitive behavior of the particles. Comparing the correspond-
ing I-E area of the first and last measured cycle for each particle,
shown respectively in Fig. 6a and b, one sees that, qualitatively, the
activation to 0.3 V increased the capacitance value. However, the
CVs for the activation performed to 0.5 V showed a dramatic reduc-
tion of the I-E area. The specific capacitance values of the particles
during cycling are shown quantitatively in Fig. 6c. For instance, the
increase of capacitance due to the activation of particle A reaches a
maximum value after the fourth activation cycle and remains con-
Downloaded on 2015-03-08 to IP 132.174.255.116 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A274
Journal of The Electrochemical Society, 152 ͑2͒ A270-A276 ͑2005͒
Figure 8. Dependence of the diffusion coefficients on the potential of an NC
particle of size 130 ␮m. The change in potential was 0.1 V for all the steps.
The arrow shows the decreasing trend of Deff along the positive potentials.
Figure 6. CVs and calculated capacitances. ͑a͒ Particle A, cyclically acti-
vated from 0 to 0.3 V, scan rate 10 mV/s, ͑b͒ particle B, activated from 0 to
0.5 V, scan rate 10 mV/s, and ͑c͒ calculated specific capacitances.
applied potential for an NC particle of diam 45 ␮m. The particle was
activated from 0 to 0.3 V in similarly as for particle A. After the 14th
CV, potential steps of 0.1 V, between Ϫ0.3 to 0.3 V, were applied.
For this plot, the diffusion coefficients were calculated using the
equation for radial diffusion ͑Eq. 4͒ instead of Eq. 3, the Cottrell
relation. This was done because both faradaic and nonfaradaic pro-
cesses occurred on the carbon surface between 0 and 0.3 V. The
trend observed in Fig. 8 is that the diffusion coefficients become
lower as the applied potential becomes more positive. A similar
tendency of Deff reduction with increasing positive potentials has
been reported for lithium insertion into disordered carbons, using a
single-particle electrode technique33 and conventional electrodes.34
The presence of oxygen functional groups was confirmed in the
carbons used in the latter reference.
stant with further cycling. Activation of particle B caused an in-
crease of the capacitance after the first cycle, but it dropped consid-
erably after the second cycle.
The distortion of the square-shaped CV observed at the reversal
of the potential in Fig. 6b revealed that the transport of ions inside
the pores became strongly hindered inside the pores of particle B.
However, there is no clear indication from the CV in Fig. 6a that the
mass-transport limitations increased after the activation process for
particle A.
Effects of positive polarization on effective diffusion coeffi-
cients.—For comparison, the effective diffusion coefficients against
the number of polarization cycles are plotted in Fig. 7 for particles A
and B. The values shown in this figure were calculated from a po-
tential step between 0 and Ϫ0.1 V ͑where the NC material shows an
ideal double layer behavior͒, after each complete activation cycle,
using the Cottrell equation ͑Eq. 3͒ for semi-infinite planar diffusion.
By plotting It1/2 against t1/2 from the experimental data obtained
from the potential step and checking where the product It1/2 is con-
stant, it is possible to estimate the Cottrellian domain.22 The values
of the diffusion coefficients, during activation of particle A, decrease
and reach a level around 0.5 ϫ 10Ϫ9 cm2 sϪ1 after being activated
20 times, whereas the diffusion coefficients for particle B, activated
to 0.5 V, decrease in comparison, by almost two orders of magnitude
after the 10th cycle.
Raman measurements.—According to Kinoshita,5 the corrosion
of carbon materials may arise as a principal anodic process at tem-
peratures below 50°C and at potentials below 0.5 V vs. Hg͉HgO.
The evolution of CO or CO2 implies that the carbon surface is
modified when the carbon atoms close to the surface are consumed
in the corrosion reaction. However, the stable values of capacitance
obtained after electro-oxidation of the NC in domain 2 showed, no
evidence of carbon consumption when compared with the results
obtained from oxidation in domain 3, where the loss of both capac-
ity and ion transportation is more evident. To verify that the electro-
chemical oxidation taking place in domain 2 is moderate ͑i.e., no
production of CO or CO2) and that no modifications of the carbon
structure occurred after the cycling, an in situ spectrophotometric
method was used. Raman spectroscopy has been widely used for
characterizing carbon materials, because Raman spectra can provide
structural information on different forms of carbon. The literature
concerning this matter is extensive. The NC material used here
shows in principle Raman response signals similar to those of poly-
crystalline graphites, namely, the peak typical for graphite approxi-
mately 1580 cmϪ1, known as the G ͑graphite͒ peak, and an addi-
Effects of potential on effective diffusion coefficients.—Figure 8
shows the dependence of the effective diffusion coefficients on the
tional signal known as the D ͑disorder͒ band, at about 1360 cmϪ1
.
The D band ͑a forbidden transition͒ results from a relaxation of the
symmetry rule for Raman scattering due to the presence of disorder
parts like grain edges. Thereby, the occurrence of the peak at 1360
cmϪ1 has been explained as a crystallite size ͑La͒ effect. Using the
peak intensity, i, the relative intensity ratio, R, of the D to the G
peak, is correlated to the reciprocal of the crystalline size along the
basal plane in the surface layer (R ϭ iD /iG ϰ 4.4 nm/La͒.35 Thus,
changes in the R value may be interpreted as changes of nanocrys-
talline graphite or related to the number of defects such as edges and
lattice vacancies of the graphite basal plane. Nakahara et al.10 stud-
ied the oxidation of pyrolytic graphite, comparing the R values be-
tween the D and G peaks in both alkaline and acid media. They
found that the oxidation to hydroxyl groups did not damage the
carbon structure and, therefore, did not change the R value ͑before
Figure 7. Dependence of the effective diffusion coefficients on the degree of
activation for particles A and B, activated as shown in Fig. 4.
Downloaded on 2015-03-08 to IP 132.174.255.116 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 152 ͑2͒ A270-A276 ͑2005͒
A275
Figure 9. Raman spectra of NC electrode ͑a͒ pristine, ͑b͒ boiled in 6 M
KOH, and ͑c͒ electrochemically activated to 0.3 V vs. Hg͉HgO.
Figure 10. Possible oxygen groups on the carbon surface after mild and
strong activations and ͑see asterisks͒ a suggested mechanism of enhancement
of the total capacitance by a reversible redox system.5,9,17
and after the oxidation treatment͒ of the carbon spectra, whereas a
stronger oxidation, producing principally carboxyl groups, caused
an increase of the R values.
The Raman spectra obtained here for a circular 20 mm diam, 3
mm thick electrode are shown in Fig. 9. All the spectra have the
same peaks. The spectrum of the pristine electrode is shown in Fig.
9a and corresponds to a relatively disordered carbon material.36 The
slight increase of R (R ϭ 0.95) for the spectrum of the boiled elec-
trode may be interpreted as an increase in the disorder of the carbon
structure due to a mild etching of the carbon surface. However, the
difference is small, and no further experimental data was collected
to estimate its significance. Nevertheless, as shown in Fig. 9c, the
value of R remains constant, R ϭ 0.95, for the carbon electrode
after the electrochemical oxidation. The interpretation of the con-
stant R value is that the structure continues unmodified after the
electrochemical oxidation, thus suggesting that the faradaic reac-
tions involved in domain 2 are relatively moderate and do not affect
the carbon structure.
mechanism of pseudocapacitance in alkaline solutions is elucidated
in Fig. 10. The picture shows oxidation occurring to different de-
grees, depending on the extent of the positive polarization. The ac-
tivation to 0.3 V for particle A may be the result of a chemisorption
of hydroxyl ions on the wall of the nanopores, which leads to the
formation of proposed functional groups like phenol or quinone.3,5
These groups may act as reversible redox systems bound to the
surface and, therefore, carry a charge. At positive polarization, this
charge may be incorporated chemically on the surface but released
and associated with a potassium ion when the system is maintained
at negative potentials.
The current transients at the reversal of the potential from the
CVs and the higher capacitance values obtained for particle A do not
reflect, in a straightforward way, the effects of the electro-oxidation
on the transport properties of ions inside the nanopores. But the
results from the potential steps in Fig. 7 show that the diffusion
coefficients for particle A decreased to a value of around 0.5
ϫ 10Ϫ9 cm2 sϪ1 after 23 activation cycles. Hence, this is an indi-
cation that the nanopores become partially blocked and that the ad-
dition of hydroxyl/carbonyl groups on the walls of the pores distorts
the path of the ions and causes slower mass transport. The diffusion
coefficients calculated from the potential steps performed between
Ϫ0.3 to 0.3 V for the activated particle shown in Fig. 8 indicate that
a more pronounced lowering of ion transport rate inside the nanop-
ores takes place at positive polarization. A possible explanation for
the reduction of Deff with increasing potential may be an increase of
tortuosity in the pore system, blocking of pores and changed path-
ways of ions, caused by pore constriction. This may be due to the
formation of surface groups, for instance, OHϪ, which occupy a
certain fraction of the cross-sectional area of the pore, or a possible
increase of the diffusion resistance due to the interactions between
surface groups and the ions.
At positive potentials up to 0.5 V, the unfavorable effects on both
the mass transport and capacitance are more easily seen, as was
shown for particle B in Fig. 6b and 7. Apart from the corrosion
products, surface groups with two oxygen atoms are more dominant
at these potentials. These groups do not give a high contribution to
the total capacitance at a difference from, for instance, quinones.17,18
The severe decrease of Deff values may be caused by a substantial
obstruction of the nanopores, i.e., the pores become more blocked
by bigger functional groups such as carboxyl and carbonate ions.
For instance, a carboxyl surface group can occupy a surface area of
approximately 17 Å2. The formation of carbon dioxide inside the
pores may also explain the dramatic decrease of the diffusion coef-
ficients through the formation of carbonate ions according to the
reaction
Discussion
The I-E response of the NC when subjected to an increasing
positive polarization has shown that the faradaic processes taking
place around 0.3 V are of a different character compared to the
processes that take place at 0.5 V. No gas evolution was observed
during the polarization to 0.3 V, and according to the Raman spectra
for particle A, no changes in the morphology of the carbon surface
were produced by the electrochemical treatment. Hence, the results
suggest that the faradaic processes occurring at 0.3 V relate to an
oxidation below the potentials for corrosion processes and oxygen
evolution. Electrochemical oxidation of carbon materials in alkaline
solutions, particularly using carbons with a disordered microtexture,
can easily bind oxygen to the edges of the graphene planes and form
carbon-oxygen complexes at positive potentials. The principal func-
tional groups, which are present after the oxidation at relatively low
positive potentials, correspond to the oxidized species that during
thermal desorption preferentially evolve as CO, such as surface hy-
droxyl ͑e.g., phenolic͒ and carbonyl ͑e.g., quinone-type͒ groups.17,18
These functional groups involving one oxygen atom bound to the
carbon surface enhance the capacitance of the electrode materials,
whereas the functional groups involving two oxygen atoms show an
opposite effect. The results in Fig. 5 showing increased capacitance
when the carbon particle was oxidized to 0.3 V support this discus-
sion. However, when the polarization reaches higher positive poten-
tials than 0.3 V, a decrease of capacitance is observed. It is believed
that a stronger oxidation gives rise to carboxyl and other highly
oxygenated functional groups. Also, gaseous O2 and CO2 may be
formed, the latter forming carbonate.
The mechanism by which the formation of chemisorbed func-
tional groups contributes to the total capacitance of the carbon elec-
trode is known as pseudocapacitance. Our hypothesis regarding the
Downloaded on 2015-03-08 to IP 132.174.255.116 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A276
Journal of The Electrochemical Society, 152 ͑2͒ A270-A276 ͑2005͒
2 OHϪ ϩ CO2 → CO32Ϫ ϩ H2O
Raman measurements. The Swedish Energy Agency and the Swed-
ish Research Council are gratefully acknowledged for financial sup-
port.
The CO23Ϫ ions are bigger than the CO2 molecules and may strongly
block the nanopores. It could be interesting to eliminate the carbon-
ate ions from the porous structure and investigate the influence of
solely the more oxidized forms of carbon on the capacitance and the
diffusion coefficients, and likewise, to investigate the effects of the
presence of carbonate ions for the same purpose. This would require
another experimental system and may be handled in a future work.
The results have practical implications for the performance of
carbon electrodes for DLCs. The carbon double-layer particles in
this study have been exposed a to potential cycling, which mimics
that of a positive electrode during electrochemical activation of a
DLC. It is clear from this study that activation by potential cycling
yields a higher capacitance, but also a decreased diffusivity. Thus,
the high power capacitance of the electrode becomes reduced by
activation. This effect is especially important for electrodes made of
nanoporous carbon with its very small and uniform pore size distri-
bution, but it is also seen for other activated carbon electrodes.6
The diffusivity of an activated electrode also shows a strong
potential dependence. A consequence of this is that the negatively
The Royal Institute of Technology assisted in meeting the publication
costs of this article.
References
1. E. Faggioli, P. Rena, V. Danel, X. Andrieu, R. Mallant, and H. Kahlen, J. Power
Sources, 84, 261 ͑1999͒.
¨
2. R. Kotz and M. Carlen, Electrochim. Acta, 45, 2483 ͑2000͒.
3. B. E. Conway, Electrochemical Supercapacitors: Scientific Fundamentals and
Technological Application, Kluwer Academic, New York ͑1999͒.
´
4. A. Rolla, A. Sadakowski, J. Wild, and P. Zoltowski, J. Power Sources, 5, 189
͑1980͒.
5. K. Kinoshita, Carbon: Electrochemical and Physicochemical Properties, John
Wiley & Sons, New York ͑1988͒.
6. C.-H. Kim, S.-II. Pyun, and H.-C. Shin, J. Electrochem. Soc., 149, A93 ͑2002͒.
7. P. N. Ross and M. Sattler, J. Electrochem. Soc., 135, 1464 ͑1988͒.
8. P. N. Ross and H. Sokol, J. Electrochem. Soc., 131, 1742 ͑1984͒.
9. T. Arikado, C. Iwakura, H. Yoneyama, and H. Tamura, Electrochim. Acta, 21, 551
͑1976͒.
10. M. Nakahara and K. Shimizu, J. Mater. Sci., 27, 1207 ͑1992͒.
11. J.-B. Donet, T. K. Wang, J. M. C. Peng, and S. Rebouillat, Carbon Fibres, 3rd ed.,
Marcel Dekker, New York ͑1998͒.
polarized electrode should have
a faster response ͑smaller
resistance-capacitance time constant͒ than the positive electrode.
The results also point to the importance of cell balancing when
using a multicell supercapacitor pack. The high power performance
of a cell drifted to higher potentials significantly decreases when the
positive electrode is exposed to potentials above 0.3 V vs. Hg͉HgO.
12. W. W. Wright, Comp. Polym., 3, 231 ͑1990͒.
13. S.-J. Park and K.-D. Kim, J. Colloid Interface Sci., 218, 331 ͑1999͒.
14. S. Ahn and B. J. Tatarchuk, J. Electrochem. Soc., 142, 782 ͑1995͒.
15. K. Horita, Y. Nishibori, and T. Ohshima, Carbon, 34, 217 ͑1996͒.
16. J. Koresh and A. Soffer, J. Electrochem. Soc., 124, 1379 ͑1977͒.
17. Y.-R. Niar and H. Teng, J. Electroanal. Chem., 540, 119 ͑2003͒.
18. Y.-R. Nian and H. Teng, J. Electrochem. Soc., 149, A1008 ͑2002͒.
19. P.-Z. Cheng and H. Teng, Carbon, 41, 2057 ͑2003͒.
20. A. Dekanski, J. Stevanovic, R. Stevanovic, B. Z. Nikolic, and V. M. Jovanovic,
Carbon, 39, 1195 ͑2001͒.
Conclusions
1. An electroanalytical method, based on the single-particle mi-
croelectrode technique, was used successfully to study the effects of
positive polarization on mass-transport properties of ions inside the
pores of an NC material.
´
21. A. Harrach and A. Metrot, Electrochim. Acta, 34, 1877 ͑1989͒.
¨
22. M. Zuleta, M. Bursell, A. Lundblad, and P. Bjornbom, J. Electroanal. Chem., 549,
2. The deviations from the ideal behavior, observed in the stud-
ied nanoporous carbon when acting as a positive electrode, are
caused by several processes, particularly oxidation of the carbon
surface. The oxidation processes that take place up to 0.3 V vs.
Hg͉HgO enhance the electrochemical properties of the NC as elec-
trode material for DLCs. Oxidation at higher potentials has an un-
favorable impact on the electrode material performance.
3. The results indicate that the transport properties of the nan-
oporous system are very sensitive to the incorporation of species on
the pore surfaces. The oxygen surface groups may cause pore con-
striction, i.e., reduction of the pore size, but may also increase the
tortuosity of the pore system for the ions moving in and out during
the charge and discharge processes. These groups can block the
pores, forcing the ions to move longer pathways to other openings in
the pore structure.
101 ͑2003͒.
¨
23. M. Bursell and P. Bjornbom, J. Electrochem. Soc., 137, 363 ͑1990͒.
24. Y. Gogotsi, A. Nikitin, Y. Haihui, W. Zhou, J. E. Fischer, B. Yi, H. C. Foley, and M.
W. Barsoum, Nat. Mater., 2, 591 ͑2003͒.
25. V. Kustnetsov, S. Gordeev, and T. Ekstrom, Proceedings 6th International Seminar
¨
on Double Layer Capacitors and Similar Energy Storage Devices, Deerfield Beach,
FL, Dec, 1996.
26. R. G. Avarbz, A. V. Vartanova, S. K. Gordeev, S. G. Zjukov, B. A. Selenov, A. E.
Kravtjik, V. P. Kuznesov, J. A. Kukusjkina, T. V. Mazaeva, O. S. Pankina, and V. V.
Sokolov, U.S. Pat. N5,876,787 ͑1999͒.
27. H. Ura, T. Nishina, and I. Uchida, J. Electroanal. Chem., 396, 169 ͑1995͒.
28. J. K. Floess and Y. Van Lishout, Carbon, 30, 967 ͑1992͒.
29. T. Nishina, H. Ura, and I. Uchida, J. Electrochem. Soc., 144, 1273 ͑1997͒.
30. R. M. Wightman, Anal. Chem., 53, 1125A ͑1981͒.
31. R. Greef, R. Peat, L. M. Peter, D. Pletcher, and J. Robinson, Ellis Horwood Series
in Physical Chemistry, Instrumental Methods in Electrochemistry, Chap. 6, John
Wiley & Sons, Chichester, U.K. ͑1985͒.
4. The diffusion coefficients decrease as the positive potential is
increased, thus giving an indication of interaction at the double layer
between the chemisorbed ions on the carbon surface and similar ions
in the bulk solution. Hence, an increase of the time constant for
linear diffusion, ␶, is expected when the NC is used as a positive
electrode.
32. H.-S. Kim, T. Itoh, M. Nishizawa, M. Mohamedi, M. Umeda, and I. Uchida, Int. J.
Hydrogen Energy, 27, 295 ͑2002͒.
33. K. Dokko, Y. Fujita, M. Mohamedi, M. Umeda, I. Uchida, and J. R. Selman,
Electrochim. Acta, 47, 933 ͑2001͒.
34. J. S. Gnanaraj, M. D. Levi, E. Levi, G. Salitra, D. Aubarch, J. E. Fisher, and A.
Claye, J. Electrochem. Soc., 148, A525 ͑2001͒.
35. F. Tuinstra and J. L. Koenig, J. Chem. Phys., 53, 1126 ͑1970͒.
36. R. Escribano, J. J. Sloan, N. Siddique, N. Sze, and T. Dudev, Vib. Spectrosc., 26,
179 ͑2001͒.
Acknowledgments
The authors thank Anders Strandroth and Professor Jan Lindgren
from the Angstrom laboratory, Uppsala University, for help with the
¨
37. J. Zheng, T. C. Ekstrom, S. K. Gordeev, and M. Jacob, J. Mater. Chem., 10, 1039
˚
¨
͑2000͒.
Downloaded on 2015-03-08 to IP 132.174.255.116 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).