Electrochemically modified glassy carbon for capacitor electrodes. Characterization of thick anodic layers by cyclic voltammetry, differential electrochemical mass spectrometry, spectroscopic ellipsometry, X-ray photoelectron spectroscopy, FTIR, and AFM

Article abstract of DOI:10.1149/1.1393582

Glassy carbon (GC) electrodes were activated by electrochemical constant potential anodization in order to generate high-surface area, high-capacitance electrodes. After anodic oxidation in sulfuric acid the electrodes exhibited increased capacitance. After subsequent electrochemical reduction of the activated layer, a further significant increase in capacitance was observed. Growth, structure, and surface properties of the activated electrodes were monitored by cyclic voltammetry, differential electrochemical mass spectrometry, spectroscopic ellipsometry, X-ray photoelectron spectroscopy (XPS), and atomic force microscopy (AFM). Two different types of glassy carbon obtained by pyrolysis at 1000 °C and at 2200 °C were compared. Differential electrochemical mass spectrometry reveals that CO₂ is the main reaction product during oxidation, while CO₂ and H₂ are detected during reduction. The values of surface layer capacitance and thickness determined by spectroscopic ellipsometry increase as linear functions of oxidation time. The resulting volumetric capacitance was at least 100 F/cm³. After oxidation, the presence of functional surface groups was demonstrated by XPS. The relative contributions of the different surface functionalities depend on the pyrolysis temperature of the GC. Reduction lowered the concentration of oxygen-containing functional surface groups. The XPS results were qualitatively confirmed by Fourier transform infrared measurements carried out at the same samples. AFM measurements on glassy carbon showed that the film growth both into and out of the substrate, resulted in a raised surface after activation. A qualitative model for film growth is presented.

Full text of DOI:10.1149/1.1393582

2636
Journal of The Electrochemical Society, 147 (7) 2636-2643 (2000)
S0013-4651(99)11-045-0 CCC: $7.00 © The Electrochemical Society, Inc.
Electrochemically Modified Glassy Carbon for Capacitor Electrodes
Characterization of Thick Anodic Layers by Cyclic Voltammetry, Differential
Electrochemical Mass Spectrometry, Spectroscopic Ellipsometry, X-Ray
Photoelectron Spectroscopy, FTIR, and AFM
M. G. Sullivan,a,* B. Schnyder, M. Bärtsch, D. Alliata, C. Barbero,b R. Imhof,c and R. Kötzz
Electrochemistry Laboratory, Paul Scherrer Institute, CH-5232 Villigen, Switzerland
Glassy carbon (GC) electrodes were activated by electrochemical constant potential anodization in order to generate high-surface
area, high-capacitance electrodes. After anodic oxidation in sulfuric acid the electrodes exhibited increased capacitance. After sub-
sequent electrochemical reduction of the activated layer, a further significant increase in capacitance was observed. Growth, struc-
ture, and surface properties of the activated electrodes were monitored by cyclic voltammetry, differential electrochemical mass
spectrometry, spectroscopic ellipsometry, X-ray photoelectron spectroscopy (XPS), and atomic force microscopy (AFM). Two dif-
ferent types of glassy carbon obtained by pyrolysis at 1000ЊC and at 2200ЊC were compared. Differential electrochemical mass
spectrometry reveals that CO2 is the main reaction product during oxidation, while CO2 and H2 are detected during reduction. The
values of surface layer capacitance and thickness determined by spectroscopic ellipsometry increase as linear functions of oxida-
tion time. The resulting volumetric capacitance was at least 100 F/cm3. After oxidation, the presence of functional surface groups
was demonstrated by XPS. The relative contributions of the different surface functionalities depend on the pyrolysis temperature
of the GC. Reduction lowered the concentration of oxygen-containing functional surface groups. The XPS results were qualita-
tively confirmed by Fourier transform infrared measurements carried out at the same samples. AFM measurements on glassy car-
bon showed that the film growth both into and out of the substrate, resulted in a raised surface after activation. A qualitative model
for film growth is presented.
© 2000 The Electrochemical Society. S0013-4651(99)11-045-0. All rights reserved.
Manuscript submitted November 11, 1999; revised manuscript received March 6, 2000. This was Paper 85, presented at the San
Diego, California, Meeting of the Society, May 3-8, 1998.
Glassy carbon (GC) is a well known material frequently used in
The use of GC for electrochemical EDLCs was suggested about
20 years ago in a patent by Miklos et al.11 Activation of the GC sur-
face was attained by gas-phase oxidation at elevated temperatures.
Electrochemical activation was not considered in that patent.
analytical electrochemistry. Numerous investigations of the catalyt-
ic surface properties of activated GC have been reported.1 The GC
electrodes can be activated along different routes such as wet chem-
ical, dry chemical, or electrochemical oxidation. Laser activation has
also been suggested by Pontikos and McCreery.2 The numerous pos-
sibilities for carbon surface activation have been described in articles
and book by McCreery3 and by Kinoshita.4
Several advantages are expected from modified glassy carbon
when this is used as an electrode material in EDLCs. It is a reasonably
good electronic conductor (200 S/cm)1,12 and can therefore also be
used as the current collector. In addition, GC is impermeable to gases
and ions, so that a bipolar plate/electrode assembly (BPEA) can be
created by simply modifying a glassy carbon sheet on both sides.13
However, when working with GC one has to be aware of the fact
that depending on the precursor material and temperature used dur-
ing the pyrolysis process, rather different kinds of GC exist.1 Prop-
erties such as the conductivity, density, reactivity, number, and diam-
eter of internal closed pores, etc. are determined by the pyrolysis
temperature. GC is still considered an expensive material, because
sophisticated furnaces are needed for the slow pyrolysis process.
Considering the rather inexpensive polymeric starting material, how-
ever, there may exist a potential for significant cost reduction.
In the present work, we used potentiostatic electrochemical acti-
vation of the GC surface. The activated surfaces were characterized
in view of electrochemical double-layer capacitor applications. Ana-
lytical methods included cyclic voltammetry (CV) for capacitance
determination, ellipsometry for thickness measurements, Fourier
transform infrared (FTIR) and X-ray photoelectron spectroscopy
(XPS) for chemical characterization, and atomic force microscopy
(AFM) and scanning electron microscopy (SEM) for structure eval-
uation. An extensive impedance analysis of the activated GC sur-
faces is presented in a separate paper.14 The results reported primar-
ily refer to the high-temperature (HT) GC if not mentioned other-
wise. Results for low-temperature (LT) GC are only mentioned
where significant deviations from those for HTGC were observed.
Differences between HTGC and LTGC are discussed in greater
detail in another publication.14
In electrochemical activation, a number of choices exist as to how
to modify the surface, such as galvanostatic, potentiostatic, or cyclic
polarization in various electrolytes. Film growth by cycling and the
corresponding optical properties of the active layer were investigat-
ed earlier in our laboratory using spectroscopic ellipsometry.5
Electrochemical double-layer capacitors (EDLC), also called
supercaps or ultracaps, utilize high-surface area electrodes in order to
achieve a high-double-layer capacitance. Three main categories of
electrode materials typically are used in these EDLCs, viz., carbons,
polymers, and metal oxides.6 For noble metal oxides such as RuO2
specific capacitance of more than 700 F/g was reported,7 but these
materials are generally considered as being too costly. Redox active
polymer films are also considered to be potential electrode materials
for EDLCs,8 but most of them are rather slow, and their long-term sta-
bility and cycle life are still uncertain. High-surface area carbons are
relatively inexpensive EDLC electrode materials with a relatively high
specific capacitance of up to 100 F/g.9 Therefore, in most of the capac-
itors available today, carbon materials are used for the electrodes.
Problems still arise from the contact resistance between carbon pow-
der particles and from that between the active layer and the current
collector sheet. Metal particles or fibers have been added to the carbon
powder in order to overcome the grain-to-grain resistance.10
* Electrochemical Society Active Member.
Present address: Roche Diagnostics Corporation, Roche Patient Care, Indianapolis,
a
Indiana 46250-0457.
b Present address: Departamento de Química y Física, Universidad Nacional de Río
Cuarto, 5800-Río Cuarto, Argentina.
Experimental
The GC was purchased from HTW (Hochtemperatur Werkstoffe
GmbH, Thierhaupten, Germany). Two types of GC are available
c Present address: Renata AG, CH-4452 Itingen, Switzerland.
z
E-mail: ruedigerkoetz@psi.ch
Downloaded on 2014-09-06 to IP 141.219.44.39 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, 147 (7) 2636-2643 (2000)
2637
S0013-4651(99)11-045-0 CCC: $7.00 © The Electrochemical Society, Inc.
open-circuit potential, blown dry with nitrogen, and finally intro-
duced into the ultrahigh vacuum chamber.
from this company, an LTGC which is pyrolyzed at 1000ЊC, and an
HTGC pyrolyzed at 2200ЊC. The material was purchased in the form
of 3 mm diam rods or as disks 1 mm thick and 50 mm in diam.
Electrodes were constructed from the GC rods by press fitting
into an 11 mm diam Kel-F insulating shroud (exposed electrode disk
area 0.0707 cm2). For each experiment, a fresh surface was produced
by sanding with wet 320 and 600 grit paper, then with diamond paste
of 9, 6, 3, and 1 ␮m. The electrodes were sonicated in ULTRAMET
sonic cleaning solution (Buehler Ltd.), then in distilled water.
Pieces of GC were also cut from the disks with a diamond saw.
These pieces were used as such after sonication in ULTRAMET
cleaning solution, then distilled water. Such electrodes were only
used once.
Infrared.—FTIR analysis was performed in a Perkin-Elmer Sys-
tem 2000 spectrometer. The samples were measured in the reflection
mode using a variable angle reflection (VAR) accessory (Harrick
Scientific Corporation). The angle of incidence was typically 60Њ.
The resolution was 4 cmϪ1, and typically 16 scans were recorded at
an optical phase difference (OPD) velocity of 5 cm/s.
Results
Electrochemistry.—CVs of HTGC are reproduced in Fig. 1 for
nonactivated GC and for various stages of constant potential activa-
tion at a potential of 1.95 V vs. SCE and subsequent reduction at
Ϫ0.3 V vs. SCE. The CV of the nonactivated sample coincides with
the horizontal x axis for the current scale used in Fig. 1. The charge
under the voltammogram increases considerably with activation
time and is a direct measure for the increased double-layer capaci-
tance. The current maxima evolving around 0.4 V vs. SCE can be
attributed to active surface groups of the quinone type.4,16 The CVs
of Fig. 1 appear to be asymmetric around the horizontal axis, with
higher anodic currents at the positive end and higher cathodic cur-
rents at the negative end. This tilt of the CV can be attributed to slow
surface reactions associated with the redox couple at 0.4 V vs. SCE.
Figure 2 compares the capacitance values determined after 10 h
of oxidation from the cyclic voltammogram and from impedance
spectroscopy at 0.1 Hz. The capacitance maximum at 0.4 V appears
to be rather symmetric for the values determined from the imped-
ance measurements at steady state (open circles), while the capaci-
tance of the CV recorded with a scan rate of 100 mV/s exhibits high-
er values anodic to the maximum in the anodic scan. Obviously, slow
surface oxidation processes lead to a higher current and capacitance
anodic to the current maximum at 0.4 V in the anodic scan and
cathodic to 0.4 V in the cathodic scan. The values determined from
the CV agree with those obtained in the impedance measurements,
anodic to 0.4 V for the cathodic scans and cathodic to 0.4 V for the
anodic scans. This becomes evident from the dashed line in Fig. 2,
where the cathodic scan was mirrored through the x axis.
CV and surface modification.—A standard three-electrode cell
was used for the anodic modification procedure, where a 2 cm2 Pt foil
or a 20 cm2 glassy carbon disk was the auxiliary, and standard calomel
electrode (SCE) the reference electrode. All reagents were used as
received. The anodic modification was performed in 3 M H2SO4
(Fluka), by maintaining the glassy carbon electrode at ϩ1.95 V vs.
SCE for periods of up to 30 h. During long anodization tests (more
than 1 h), a small stream of Ar bubbles was directed at the working
electrode surface. This serves to dislodge any gas bubbles which may
begin to form on the electrode surface. The current generally remained
constant at about 0.8 mA/cm2. The electrode was then reduced at
Ϫ0.3 V vs. SCE for 10 min and cycled several times in fresh 3 M
H2SO4 between potentials of Ϫ0.2 and ϩ1.0 V. No attempt was made
to eliminate O2 from the fresh electrolyte solution.
Differential electrochemical mass spectrometry (DEMS).—The
experimental setup was described elsewhere.15 A typical working
electrode was prepared from GC powder (also available from HTW,
Germany) slurried up in a solution of polyvinylidene fluoride
(PVDF, Aldrich; 10 wt %) dissolved in 1-methyl-2-pyrrolidinone
(Fluka). This slurry was sprayed with an air brush onto a GORE-
TEX epoly(tetrafluoroethylene) photovoltaic 9008 membrane
(60 ␮m thick, pore size 0.02 ␮m). The electrode was dried in a vac-
uum at 80ЊC. About 0.2 mg/cm2 of the graphite/binder composite
electrode material was found to be deposited. The measurements
were made with a three-electrode setup, using a platinum wire as the
counterelectrode and Pd/H2 as a reference electrode. The anodic
modification was performed in 3 M H2SO4 (Fluka), by maintaining
the GC electrode for 16 min at ϩ2.2 V vs. Pd/H2. The electrode was
then reduced for 16 min at 0 V vs. Pd/H2. No attempt was made to
eliminate O2 from the fresh electrolyte solution. The currents and
mass signals were recorded simultaneously as functions of potential.
The electrochemical activation of GC involves oxidation and
subsequent reduction of the electrode. The effect of the reduction
step is demonstrated in Fig. 3. The evolution of the capacitance with
activation time is shown in Fig. 3 on a logarithmic scale for the GC
sample, both directly after anodic oxidation and after subsequent
reduction at Ϫ0.3 V vs. SCE. Starting from a capacitance of several
microfarads per centimeter squared the capacitance increases by sev-
eral orders of magnitude, provided the surface is reduced after oxi-
dation. It is obvious from Fig. 3 that an electrochemical reduction of
Spectroscopic ellipsometry.—Ellipsometry was performed using a
SOPRA model ES4G MOSS spectroscopic ellipsometer at a 70Њ
angle of incidence. A spectral range of 1.5 to 4 eV (827 to 310 nm)
was used. The sample was a 1 ϫ 2 ϫ 0.1 cm piece of GC whose back
and sides were coated with Lacomit (AGAR Scientific Ltd.) varnish.
The samples were kept submerged in 3 M H2SO4 at all times. No
potential control was used during the measurements, but the samples
typically came to a rest potential of about ϩ0.2 V vs. SCE after an
anodic treatment at ϩ1.95 V and subsequent reduction at Ϫ0.3 V.
Ellipsometry was also performed on unreduced samples, which were
anodically modified at ϩ1.95 V vs. SCE and then held for several
minutes at ϩ0.4 V vs. SCE. Left without potential control, these unre-
duced samples typically came to a rest potential of about ϩ0.65 V vs.
SCE. Data acquisition and analysis software was supplied by
SOPRA, version ELLI 4.52.
XPS.—XPS analysis was performed on a VG ESCALAB 220i
XL using monochromated Al K␣ radiation. The spectra were record-
ed with a pass energy of 20 eV in the CAE mode. The base pressure
of the system was typically 5 ϫ 10Ϫ9 mbar. The electron emission
angle was 90Њ. The samples were investigated as received. Even
slight sputtering of the surfaces turned out to alter the results signif-
icantly. After electrochemical preparation, the samples were re-
moved from the electrolyte, either under potential control or at the
Figure 1. CVs of HTGC electrodes recorded in 3 M H2SO4 after different
times of anodic activation at 1.95 V. The electrode was oxidized at ϩ1.95 V
vs. SCE for the time shown, then reduced.
Downloaded on 2014-09-06 to IP 141.219.44.39 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

2638
Journal of The Electrochemical Society, 147 (7) 2636-2643 (2000)
S0013-4651(99)11-045-0 CCC: $7.00 © The Electrochemical Society, Inc.
Figure 2. Comparison of the ac and CV capacitance for single electrodes
(HTGC). CV scan rate, 100 mV/s. AC capacitance measured at 0.1 Hz (cir-
cles). The dotted line is a mirror image of the cathodic CV current through
the E axis. Activation time, 10 h at ϩ1.95 V vs. SCE.
the modified surface layer is very important for the attainment of a
high double-layer capacitance. Without reduction, the capacitance
saturates at about 0.5 mF/cm2.
The observed increase in capacitance with activation time at
1.95 V vs. SCE was about three times larger for LTGC than for
HTGC. This faster growth rate at constant potential for LTGC is
associated with an increase in current by the same magnitude. Thus,
when related to the charge consumed, the capacitance growth rate is
independent of the GC pyrolysis temperature.
The effect of the electrolyte on the electrochemical activation of
the GC surface was tested by using various anions such as SO24Ϫ and
ClOϪ4 in different concentrations and at different values of pH. No
significant effect on the growth rate was discerned when referring to
the charge consumed. During long-term potentiostatic activation,
constant currents are attained. The highest current during film
growth was observed for the ClOϪ4 anion.
Figure 4. Potential (top) and mass intensity (bottom) of an HTGC powder
electrode in 1 M H2SO4 as functions of time. m/z ϭ 2 (H2) and m/z ϭ 44
(CO2) are shown.
trolyte, generation of a significant amount of oxygen during oxida-
tion can be ruled out.
During reduction at 0.0 V, the evolution of hydrogen is clearly
indicated by the respective mass signal. However, before H2 evolu-
tion sets in, a peak of mass 44 is evident. Most probably, the oxygen-
containing groups generated anodically are reduced first, yielding
CO2, before the layer will sustain H2 evolution. The latter only starts
20 s after the time when the lower potential limit of 0.0 V vs. Pd/H2
is reached.
Ellipsometry.—The thickness of the activation layer on GC was
measured with the sample immersed in the electrolyte. The model
assumptions used for the thickness determination are the same as in
our previous study of much thinner activation layers on glassy car-
bon.5 We assume that one mixed layer of GC particles and voids
(consisting of H2O) exists on top of a bulk GC substrate when H2O
is the ambient. The optical properties of the surface layer are calcu-
lated using the Bruggeman effective medium approach.17 Therefore,
the layer thickness and the void concentration are the free variables.
The ellipsometric parameters psi and delta are shown in Fig. 5 for
the sample after oxidation and after subsequent reduction. Cos(delta)
and tan(psi) are also shown for the untreated GC surface. The spec-
tra of the ellipsometric parameters of the sample after oxidation (full
line) show distinct maxima and minima, but reduction of the sample
results in rather featureless spectra (dashed lines). The modulation of
the spectra seen after oxidation can be attributed to interference pat-
terns produced in the dielectric film.
Differential electrochemical mass spectrometry (DEMS).—
DEMS was applied to determine the volatile reaction products dur-
ing anodic oxidation and subsequent reduction. Figure 4 shows the
potential profile together with the evolution of masses m/z ϭ 44
(CO2) and m/z ϭ 2 (H2). The main reaction product during oxidation
at 2.2 V vs. Pd/H2 is CO2. While there was a permanent background
signal for oxygen which originates from dissolved air in the elec-
The ellipsometrically determined thickness after oxidation of an
HTGC is shown in Fig. 6. A growth rate of 1.4 ␮m/h is obtained for
a potential of 1.95 V. The thickness of the film existing after the oxi-
dation step is relatively easy to determine, because the film has
dielectric character, and for thick films interference structures evolve
in the tan(psi) and cos(delta) data, which simplify the fitting proce-
dure. At an anodization potential of 2.1 V the film grows much
faster, the growth rate being about 7.5 ␮m/h. The model used in the
thickness determination is a stratified layer model with a single layer
on bulk GC. This layer is composed of a mixture of GC and voids.
The voids are assumed to be H2O, the ambient is also H2O.5
However, electrodes which have not been reduced after the anod-
ic treatment at ϩ1.95 V vs. SCE do not exhibit the large capacitance
Figure 3. The capacitance of activated HTGC as a function of activation
time. Electrodes were activated in 3 M H2SO4 at 1.95 V (full squares) and
subsequently reduced at Ϫ0.3 V (open squares). The capacitance was deter-
mined from impedance measurements.
Downloaded on 2014-09-06 to IP 141.219.44.39 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, 147 (7) 2636-2643 (2000)
2639
S0013-4651(99)11-045-0 CCC: $7.00 © The Electrochemical Society, Inc.
Figure 7. Thickness (᭹) and void fraction (᭺) of the activated layer on
HTGC electrodes after anodic oxidation at 1.95 V and subsequent reduction
at Ϫ0.3 V in 3 M H2SO4, determined by spectroscopic ellipsometry as func-
tions of activation time.
Ϫ0.3 V. The thickness growth rate is linear with activation time at
0.65 ␮m/h while the void fraction reaches a constant value of 75%
after about 20 min. As the reduced films are absorbing, it is not pos-
sible to measure films thicker than about 0.75 ␮m. It is very likely,
however, that since the thickness of the oxidized films increases lin-
early with activation time for several hours, the thickness of the films
does likewise. Assuming this linearity to be preserved, the reduced
active-layer thickness would be 18 ␮m after 30 h of activation at ϩ
1.95 V vs. SCE, which is the longest activation attempted.
Using spectroscopic ellipsometry on the as-received GC samples,
we determined the bulk optical properties of HT and LT type glassy
carbon in the photon energy range 1.5 eV to 4.5 eV. The results for
n and k are shown in Fig. 8 for HTGC and are similar to data report-
ed previously5 from our laboratory for GC obtained from a different
supplier. Really important differences do not exist between the two
Figure 5. Ellipsometric spectra of an HTGC electrode activated at 1.95 V for
30 min (dashed line) and of the same electrode after reduction at Ϫ0.3 V for
3 min (full line). The thin line represents the spectrum of the as-received GC
electrode. All measurements were performed in situ.
needed for an electrochemical capacitor. Performing the necessary
reduction of the surface layer at Ϫ0.3 V vs. SCE results in a film
with significantly stronger optical absorption. Figure 7 shows the
thickness and void fraction of the activation layer as functions of
activation time after oxidation at 1.95 V and subsequent reduction at
Figure 8. Optical properties, n and k, of the anodic layer formed on HTGC
before (dashed line) and after reduction (full line). The thin lines represent
the optical constants for untreated GC as determined in our laboratory, and
those for films composed of 20 and 50% voids in GC.
Figure 6. Thicknesses of the porous surface layer on HTGC after oxidation
at 1.95 and 2.1 V in 3 M H2SO4 as determined by ellipsometry.
Downloaded on 2014-09-06 to IP 141.219.44.39 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

2640
Journal of The Electrochemical Society, 147 (7) 2636-2643 (2000)
S0013-4651(99)11-045-0 CCC: $7.00 © The Electrochemical Society, Inc.
types of GC. The somewhat lower value of n found for the HT type
of GC is in harmony with its lower gravimetric density. Figure 8
shows the optical properties of the activation layers after oxidation
and after oxidation and reduction. The optical properties of bulk GC
are also shown, together with the optical properties calculated for
layers with 20 and 50% voids (thin lines). After oxidation, a dielec-
tric layer with k close to 0 is formed, but the layer formed after oxi-
dation and reduction is absorbing with k > 0.2. The optical proper-
ties were determined by growing thick films and calculating n and k
from the ellipsometric spectra assuming a bulk sample.
XPS.—The formation of functional surface groups during the
anodic oxidation of carbon is well known, and has been demonstrat-
ed by several studies.18,19 The XPS spectra of HTGC samples as re-
ceived, after 10 min of oxidation at 1.95 V vs. SCE, and after subse-
quent reduction at Ϫ0.3 V vs. SCE are shown in Fig. 9 for HTGC
and in Fig. 10 for LTGC. For the as-received sample, only one C 1s
peak at 284.3 eV can be detected, but the oxidized samples yield two
distinct chemically shifted C 1s peaks. The peak shifted by 2.2 eV
can be assigned to carbonyl and/or quinone, the peak shifted by
4.1 eV can be assigned to carboxyl.20,21 The relative concentrations
of the two functional surface species are different for the two types
of GC. After reduction the amount of chemically shifted C 1s species
is significantly reduced for both types of GC, as is evident from
Fig. 9 and 10. On HTGC (Fig. 9) the reduction of functional surface
entities appears to be more effective than on LTGC.
Figure 10. XPS spectra of the C 1s level of LTGC electrodes before oxida-
tion, after oxidation, and after activation (oxidation and reduction). Oxidation
potential, 1.95 V; reduction potential, Ϫ0.3 V.
The formation and reduction of functional surface groups is also
reflected in the amount of oxygen that can be detected in the XPS
spectra. The relative oxygen content was determined by integration
of the O 1s emission peak. The as-received LTGC and HTGC sam-
ples have relative oxygen contents of 3.2 and 7.8%, respectively,
which increase to 28.4 and 30.2% after oxidation. After subsequent
reduction the oxygen contents are reduced to 11.2 and 22.6 % for
HTGC and LTGC, respectively.
duced at Ϫ0.3 V. The spectra were obtained in the reflection mode
at 60Њ with p-polarized light. The formation of the surface layer is
evident from the differences between the reference spectrum and the
spectra of the oxidized or oxidized and reduced sample. The struc-
tures at 1720 and 1620 cmϪ1 can be assigned to the CϭO double
bonds of carboxyl groups and of the quinone structure, respectively,
the broad feature at 1225 cmϪ1 can be attributed to the C–O stretch-
ing and O–H bending modes of COOH groups.22,23 After reduction,
the absorption features are reduced significantly and shifted slightly,
but still visible. The origin of the strong band at 600 cmϪ1 in the
spectrum after oxidation and reduction is not clear. In the near IR
range (frequencies above 2500 cmϪ1) the absorption observed for
the oxidized sample decreases after reduction. Qualitatively similar
observations were made for low-temperature GC.
FTIR.—Photoelectron spectroscopy (see above) is a very sensi-
tive technique for the detection of functional surface species, as the
sampling depth is only about 50 Å. However, the corresponding XPS
results are not relevant for the bulk of the surface layers on the GC
samples. Therefore, FTIR was applied to obtain this information.
Figure 11 shows the FTIR spectra of as-received HT type GC, of a
sample oxidized at 1.95 V, and of a sample oxidized at 1.95 and re-
SEM and transmission electron microscopy (TEM).—Samples
were investigated with SEM (Fig. 12) and TEM in order to deter-
mine the microstructure and thickness of the films generated. The
films were visible in SEM in terms of contrast differences, but no
structural differences between bulk and film could be identified.
Similarly, no indications for porosity could be detected. The film
thickness of the sample reproduced in Fig. 12 is 1.0 ␮m, which is in
good agreement with the ellipsometric results for a similar sample.
Figure 9. XPS spectra of the C 1s level of high-temperature (HT) GC elec-
trodes before oxidation, after oxidation, and after activation (oxidation and
reduction). Oxidation potential, 1.95 V; reduction potential; Ϫ0.3 V.
Figure 11. Ex situ FTIR spectra of the anodic film formed on HTGC before
and after reduction. Reference spectrum was the untreated surface. The spec-
tra were obtained in the reflection mode at an angle of 70Њ.
Downloaded on 2014-09-06 to IP 141.219.44.39 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, 147 (7) 2636-2643 (2000)
2641
S0013-4651(99)11-045-0 CCC: $7.00 © The Electrochemical Society, Inc.
As described in a separate publication,24 the film thickness decreases
during reduction. An increase in surface roughness could not be
observed, and no indications of surface porosity could be detected.
Discussion
It was demonstrated in our laboratory that on GC, the anodic lay-
ers can be grown to thicknesses of up to 100 ␮m at constant rate.
These layers are obviously porous, i.e., have a high internal surface
area resulting in a high volumetric capacitance. The Warburg resis-
tance of the layer on HTGC is rather low, as indicated by impedance
measurements.14 How and why does this porous layer grow?
First, one has to keep in mind that GC is a porous material. The
density of 1.5 is significantly lower than that of graphite. However,
GC is a gastight material with a rather low permeability.12 The pores
in GC are closed and have diameters of the order of nanometers de-
pending slightly on the pyrolysis temperature.25 We assume, there-
fore, that the anodic treatment does not generate a new porosity but
merely opens closed pores already present. This step is sketched in
Fig. 14 a and b. The pore dimensions of the activated layers are in the
nano range and not visible in the SEM pictures reported here.
The results of the DEMS analysis clearly show that during anod-
ic potentiostatic oxidation of GC, carbon is burned off to form
gaseous CO2. This process would usually lead to strong corrosion
and or dissolution of the electrode without formation of a porous
layer. Growth of a thick layer can only occur if the potentiostatic oxi-
dation leads to an oxide-like layer with low electronic conductivity.
For such systems, the electrochemical reaction front moves into the
bulk until the potential drop across the layer lowers the electrode
potential and eventually stops the growth process.
According to our ellipsometry, XPS, and FTIR analyses, a dielec-
tric layer is formed on GC during anodic oxidation. The imaginary
part of the dielectric constant is close to zero and oxygen-containing
species are clearly evident from XPS and FTIR. Such a layer, which
is sketched in Fig. 14c, has insulating properties, is electrochemical-
ly inactive, and does not contribute to the double-layer capacitance.
Therefore, during oxidation, the capacitance only increases up to a
limiting value of about 300 ␮F/cm2, and does not increase further
with continued film growth. This value reflects the capacitance of
the moving reaction layer between GC substrate and growing oxide
layer. Because of its open porosity, the surface oxide layer neither
passivates the underlying substrate, nor contributes to the overall
capacitance.
Figure 12. SEM picture of the fracture plane of an HTGC sample activated
for 1 h at 1.95 V vs. SCE.
High resolution TEM pictures showed the small graphitic units of
the GC, but again, we were not able to identify reproducible differ-
ences between bulk material and the film material.
AFM.—Surface modifications of the activated GC samples were
investigated using in situ AFM. A GC sample partly coated with var-
nish was activated in the electrochemical cell of the scanning probe
microscope, then the varnish was removed. The resulting topographic
image of the edge of the activated area is shown in Fig. 13 together
with a line scan across the edge. The surface of the activated region
(right) is higher than the nonactivated surface, both before and after
reduction, indicating a significant volume increase during film growth.
It is only after reduction of the electrode (see Fig. 14d) that the
whole layer becomes electrochemically active again, resulting in a
significant increase in double-layer capacitance. The reduction step
removes oxygen from the oxide, as is evident from XPS, and leaves
an electronically conducting carbon skeleton behind. The transition
to increased conductivity is also evident in the optical properties of
the reduced layer. The imaginary part of the refractive index, k,
increases from near zero to >0.2.
The charge consumed during reduction of the oxide layer in-
creases with the time of anodic oxidation. The evolution of hydrogen
only starts after complete reduction of the oxide layer, as seen in the
DEMS transient. The origin of the CO2 (mass 44) detected during
the reduction step is not quite clear yet. Part of the CO2 generated
during oxidation may be trapped in the pores of the anodic layer and
is released during reduction. A similar observation was made by
Novak and Vielstich on porous platinum electrodes in the context of
lithium cells.26 Alternatively, CO2 may be generated during reduc-
tion by decomposition of surface oxides,3 most likely of the car-
boxylic type. Such a process can be envisaged by analogy to CO2
generation during reduction of organic electrolytes such as
chloroethylene carbonate observed in lithium cells by Winter et al.27
The formation of a surface oxide on carbon has been postulated
in the past by several authors,3,28,29 and it was even suggested to use
the nonactive anodic layer for surface structuring by local chemical
or electrochemical reduction of the anodic oxide.30 Besenhard and
Fritz28 described an electrochemical graphite oxide (EGO), which
can be generated by anodic overoxidation of graphite in aqueous
Figure 13. AFM surface topographic image of a HTGC sample partially acti-
vated for 3 h at 1.95 vs. SCE. The line indicates the edge between activated
(right) and nonactivated regions.
Downloaded on 2014-09-06 to IP 141.219.44.39 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

2642
Journal of The Electrochemical Society, 147 (7) 2636-2643 (2000)
S0013-4651(99)11-045-0 CCC: $7.00 © The Electrochemical Society, Inc.
Figure 14. Sketch of the anodic film
growth process on a GC sample. The
untreated electrode (a) is progressively
oxidized (b and c) and subsequently
reduced (d).
acidic solutions. Overoxidation is limited by a significantly decreas-
ing electric conductivity of the EGO. An empirical formula of EGO
was given by Besenhard31 to be C4.6Oи0.75H2O for the case of EGO
generation in 70% HClO4 from graphite foil. The authors also men-
tioned that EGO can be reduced back to graphite with a current effi-
ciency of nearly 100%. Reduction, however, leads to a distortion of
the graphite. The structure of glassy carbon may be described as
consisting of graphitic ribbons1 and therefore the concept of EGO
may be well applicable in a qualitative manner.
sequence, the electrochemical reaction front moves into the bulk
during anodic constant-potential polarization, enabling the growth of
rather thick film.
After reduction at potentials in the region of hydrogen evolution,
the film regains its activity and conductivity and can be utilized as a
porous structure with high volumetric capacitance. Due to signifi-
cant volume increase, the anodic film on GC grows into the bulk as
well as out of the bulk into the electrolyte, as demonstrated by in situ
AFM measurements.
The model presented in Fig. 14 is qualitatively valid for GC
regardless of the pyrolysis temperature. However, there are quantita-
tive differences. The layer growth at constant potential proceeds
faster on the LTGC, because the currents at identical values of poten-
tial are higher. This is in agreement with the activation energies
measured on a number of carbons1,32 which were found to be inde-
pendent of the pyrolysis temperature, while the reaction rate is sig-
nificantly higher for LTGC than for HTGC. The reaction rate during
gas phase oxidation increases with decreasing pyrolysis temperature,
probably due to an increased permeability to the attacking gas.1
Additionally, the reduced LTGC has significantly more oxygen
than the reduced layer formed on HTGC. Therefore, the electro-
chemical properties of the active layers on LTGC and HTGC are
expected to be different. Such differences were monitored by elec-
trochemical impedance spectroscopy in terms of a much higher War-
bug resistance of the layer formed on LTGC.14
Acknowledgment
The Board of the Swiss Federal Institutes of Technology (BSIT)
is gratefully acknowledged for financial support of this project. We
thank V. Shklover (ETHZ) for SEM and TEM pictures and O. Haas
and P. Novák for numerous discussions and invaluable ideas.
The Paul Scherrer Institute assisted in meeting the publication costs of
this article.
References
1. G. M. Jenkins and K. Kawamura, Polymeric Carbons, Carbon Fiber, Glass and
Char., University Press, Cambridge (1976).
2. N. M. Pontikos and R. L. McCreery, J. Electroanal. Chem. Interfacial Elec-
trochem., 324, 229 (1992).
3. R. L. McCreery, in Electroanalytical Chemistry, Vol. 17, A. J. Bard, Editor, p. 221,
Marcel Dekker, New York (1991).
4. K. Kinoshita, Carbon: Electrochemical and Physicochemical Properties, Wiley,
New York (1988).
The above model is quite simple, and does not explain the
swelling of the layer observed by AFM. One explanation of the sig-
nificant volume increase may be intercalation. The anodic potential
existing during layer formation is sufficiently positive for anion
intercalation.33 This idea is also supported by the fact that the acti-
vation proceeds faster in HClO4 than in H2SO4. The ClOϪ4 ions inter-
calate with higher efficiency than HSOϪ4 ions.33
5. C. Barbero and R. Kötz, J. Electrochem. Soc., 140, 1 (1993).
6. S. Sarangapani, B. V. Tilak, and C.-P. Chen, J. Electrochem. Soc., 143, 3791 (1996).
7. J. P. Zheng, P. J. Cygan, and T. R. Jow, J. Electrochem. Soc., 142, 2699 (1995).
8. X. Ren, S. Gottesfeld, and J. P. Ferraris, in Electrochemical Capacitors, F. M. Del-
nick and M. Tomkiewicz, Editors, PV 95-29, p. 138, The Electrochemical Society
Proceedings Series, Pennington, NJ (1996).
9. D. Qu and H. Shi, J. Power Sources, 74, 99 (1998).
10. A. Yoshida, A. Nishino, I. Tanahashi, and Y. Takeuchi, U.S. Pat. 4,597,028 (1986).
11. J. Miklos, K. Mund, and W. Naschwitz, Offenlegungsschrift DE 30 11 701 A1,
(1981).
Conclusions
12. R. Dübgen and G. Popp, Z. Werkstofftech., 15, 31 (1984).
13. M. Bärtsch, M. G. Sullivan, R. Kötz, and O. Haas, J. Power Sources, 65, 276
(1997).
14. M. G. Sullivan, R. Kötz, and O. Haas, J. Electrochem. Soc., 147, 308 (2000).
15. R. Imhof and P. Novák, J. Electrochem. Soc., 145, 1081 (1998).
16. C. Barbero, J. J. Silber, and L. Sereno, J. Electroanal. Chem., 248, 321 (1988).
17. D. E. Aspnes, J. B. Theeten and F. Hottier, Phys. Rev. B, 20, 3292 (1979).
18. K. M. Sundberg, Lj. Atanasoska, R. Atanasoski, and W. H. Smyrl, J. Electroanal.
Chem., 220, 161 (1987).
Anodic film growth on GC electrodes in acidic electrolytes has
similarities with the growth of oxide films on metals. The formation
of a dielectric film with low electronic conductivity and increased
oxygen content could be verified by various in situ and ex situ ana-
lytical techniques.
In contrast to passive film formation on metals, the anodic film
formed on GC is porous and has a high ionic conductivity. As a con-
Downloaded on 2014-09-06 to IP 141.219.44.39 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, 147 (7) 2636-2643 (2000)
2643
S0013-4651(99)11-045-0 CCC: $7.00 © The Electrochemical Society, Inc.
19. A. D. Jannakoudakis, P. D. Jannakoudakis, E. Theodoridou, and J. O. Besenhard,
Goerigk, J. Non-Cryst. Solid, 260, 1 (1999).
J. Appl. Electrochem., 20, 619 (1990).
26. P. Novák and W. Vielstich, J. Electrochem. Soc., 137, 1681 (1990).
20. C. Kozlowski and P. M. A. Sherwood, J. Chem. Soc., Faraday Trans. 1, 80, 2099
(1984).
27. M. Winter, R. Imhof, F. Joho, and P. Novak, J. Power Sources, 81-82, 818
(1999).
21. C. Kozlowski and P. M. A. Sherwood, Carbon, 24, 357 (1986).
22. C. Selliti, J. L. Koenig, and H. Ishida, Mater. Sci. Eng., A126, 235 (1990).
23. F. Rositani, P. L. Antonucci, M. Minutoli, N. Giordano, and A. Villari, Carbon, 25,
325 (1987).
24. D. Alliata, P. Häring, O. Haas, R. Kötz, and H. Siegenthaler, Electrochem. Solid-
State Lett., 2, 33 (1999).
28. J. O. Besenhard and H. P. Fritz, Angew. Chem., 95, 954 (1983).
29. L. J. Kepley and A. J. Bard, Anal. Chem., 60, 1459 (1988).
30. R. C. Engstrom, B. B. Ratcliff, and J. C. O’Brien, Abstract 886, p. 1396, The Elec-
trochemical Society Meeting Abstracts, Vol. 95-2, Chicago, IL, Oct 8-13, 1995.
31. J. O. Besenhard, Z. Naturforsch., B32, 1210 (1977).
32. F. C. Cowlard and J. C. Lewis, J. Mater. Sci., 2, 507 (1967).
33. F. Beck, H. Junge, and H. Krohn, Electrochim. Acta, 26, 799 (1981).
25. A. Braun, M. Bärtsch, B. Schnyder, R. Kötz, O. Haas, H. G. Haubold, and G.
Downloaded on 2014-09-06 to IP 141.219.44.39 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).