b present address: Institute for Applied Photovoltaics, Haydnstr.19,. Dâ45884 Gelsenkirchen, Germany c. Department of Physics, Carl-von Ossietzky-University, ...
J. Electrochem. Soc., revised version submitted Feb. 13, 1996
Electrochemical
Iver
Properties
Lauermanna,b , Rüdiger
of
Silicon
Carbide
Memming* , a,c and Dieter Meissner * ,a,d
a Institute for Solar Energy Research (ISFH), Sokelantstr.5, 30165 Hanno-
ver, Germany address: Institute for Applied Photovoltaics, Haydnstr.19, D–45884 Gelsenkirchen, Germany c Department of Physics, Carl-von Ossietzky-University, P.O.Box 2503, 26111 Oldenburg, Germany d present address: Forschungszentrum Juelich GmbH (KFA), Institute of Energy Process Engineering, D–52425 Juelich, Germany b present
Abstract The electrochemical properties of single crystalline n- and p-type electrodes of two different modifications of SiC in aqueous solutions were investigated in the dark and under illumination. The main emphasis was laid on investigations of the problem of water splitting in photoelectrolysis cells. No oxygen was detected at SiC electrodes. Instead anodic oxidation of SiC was found leading to the formation of CO2 , CO and silicon oxides. Since the oxide film is very porous the anodic corrosion current remained on a high level. In the presence of concentrated hydrofluoric acid a different surface layer was formed possibly due to disproportionation processes and the formation of amorphous SiC. Flatband potentials at different
pH-values
were
measured
and
the
charge
transfer
processes
with various redox systems were examined. A quantitative analysis was very difficult or even impossible, probably due to the complicated surface chemistry and layer formation.
* Electrochemical Society Active Member
- 2 -
Introduction Since 1974 the photoelectrochemical properties of SiC have been investigated by various authors.1-9 Because of the large bandgap and the excellent chemical stability, it was assumed that SiC should not be subject to anodic corrosion under conditions at which photoelectrolysis of H2 O was expected although there are contradictory reports in the literature. Values for the flatband potential of n-SiC given in the literature vary in a range of -0.7 to -1.8 V vs. SCE even for the same pH and there is yet no agreement about a possible corrosion of SiC photoanodes nor about the corrosion products. Some of these results can be attributed to the inferior quality of SiC crystals that were available at the time those experiments were performed. More
recently
new
interest
in
SiC
as
a
semiconductor
material
for
devices like blue light emitting diodes has led to the investigation of (photo)electrochemical ested
in
the
etching
methods.10,11
photoelectrochemistry
of
We
large
were
mainly
bandgap
inter-
semiconduc-
tors for the use in photoelectrochemical solar energy devices and examined two SiC modifications with bandgaps of 2.4 and 3.1 eV, respectively. The use of large bandgap semiconductors ( Eg > 2 eV) allows band
a
much
processes
bandgap those
better in
distinction
fundamental
semiconductors
with
smaller
tend
bandgaps.
to The
between
valence
and
conduction
investigations.
Furthermore,
be
more
chemically
cubic
modification
stable of
SiC
large than (3C-
SiC) with a bandgap of 2.4 eV would in theory be especially suitable for
water
cathodic
splitting
since
reactions,
and
overpotentials thermodynamic
for
both,
the
restrictions
anodic
and
demand
that
the energy separation between the excited carriers must be consid-
- 3 -
erably higher than the theoretical value of 1.23 eV. In addition, the absorption of the cubic SiC is still in the visible range of the solar spectrum. At present single crystalline SiC is only available in the hexagonal
modification
Eg = 3.1 eV) or as cubic material
(6H-SiC,
(3C-SiC), the latter being deposited as a thin layer on a Si substrate in a CVD process. Another advantage of SiC over many other large bandgap materials is the availability of p-doped crystals. This fact makes it interesting to be used as a photocathode in a photoelectrolysis cell. In such a case corrosion problems may be avoided because in this configuration H 2 would be formed at the p-type SiC photocathode and O 2 at the metal counter electrode. Other semiconductors which are available as n- and p-type material such as GaAs and InP have a too small bandgap for a water splitting reaction.
Experimental Electrochemical
experiments
were
Procedures performed
in
an
electrically
shielded
dark box using the common three electrode configuration. The reference electrode chloride
was
a
saturated
electrode
in
a
calomel
saturated
electrode potassium
(SCE) chloride
or
a
silver/silver
solution.
In
this
paper all potentials are given with respect to SCE. The potentiostat was a Solartron 1255, operated through a Macintosh II via a GPIB interface using
a
home
made
LabVIEW
program
(National
Instruments,
Austin,
Texas, USA). A Tacussel bipotentiostat BI-PAD was used for the rotating ring disc experiments. In this case the data acquisition was performed through a HP 7090 digital plotter. Illumination sources used
- 4 -
for the photoelectrochemical experiments were 1000 or 450 W high pressure Osram XBO xenon lamps with a high percentage of UV light (Müller lamp or Oriel 140). The n- and p-type crystals used in our experiments, had a doping level ranging from 1016 to 101 9 cm- 3 . Crystal sources and a few
physical
properties are given in Table 1. Exact doping levels were provided only for samples obtained from source 4; for all other samples they were not known to the manufacturer. Although the doping density can be calculated in theory from the slope of a straight Mott-Schottky plot of impedance data, in many cases no reliable values were obtained because the curves were not straight over a sufficiently large potential
range.
Ohmic contacts to n- doped SiC crystals were achieved by rubbing liquid indium on the back of the samples followed by annealing under hydrogen atmosphere at 450° C. To obtain an ohmic contact on p-doped samples aluminum was evaporated onto the sample with subsequent annealing in hydrogen at 900° C. Although annealing of evaporated aluminum on p-SiC has been reported by some authors to produce a Schottky contact, we found that annealing led to the formation of a reliable
ohmic
contact,
in
accordance
with.12 The contacted samples
were glued onto a brass conductor mounted in a Teflon sample holder using silver epoxy and subsequently the sample was covered with epoxy so that only the polished [0001- ]- or [0001]-surface was exposed. Prior to each electrochemical experiment the electrodes were polished using diamond powder (0.25 µm) to give a mirror like surface, and then etched in 48% aqueous HF for 15 to 60 s to remove any native oxide surface. In the case of the 3C-SiC samples different approaches had to be used. These
- 5 -
samples consisted of thin SiC films (6 to 12 µm) grown on a highly ndoped and
silicon the
wafer.
Si/SiC
To
contact
eliminate on
the
possible
effects
electrochemical
of
the
Si-substrate
experiments,
samples
were prepared with ohmic contacts on both sides, one on the SiC front surface and the other on the back of the Si substrate. Another set of samples was prepared by forming at first an ohmic contact to the SiC front surface and protecting it by embedding the entire surface in an epoxy resin. The Si substrate was then etched away by concentrated HF/HNO3
solutions. The SiC surface appearing after etching was then
used for the electrochemical measurements. Because of the mechanical instability of these thin layers polishing was not possible and etching in HF was used as the only surface pretreatment. Fig. 1 illustrates the different
electrode
discs
of
6H-SiC
preparation (5 mm
methods.
diameter)
For
were
ring cut
disc from
electrodes crystalline
circular platelets
using an ultrasonic saw (KLN) and then pressed into thin machined Teflon tubes.
These
were
then
squeezed
into
exactly
fitting
platinum
tubes
which formed the ring electrodes of the RRDE assembly so that SiC, Teflon and platinum were sealed against each other. Gases generated in the electrochemical cell were analyzed using a gas chromatograph (Shimadzu GC 8A) placed inside the same glove box in which the experiments were conducted, thus avoiding contamination of gas samples with outside air. In order to separate O2 , N2 , Ar, H2 and CO, a 50 m
long porous-layer-open-tube
(PLOT)
quartz
capillary
column,
the
inner surface of which was covered by a molecular sieve was used at 0 ° C. To separate CO2 and the remaining gases, a 2 m stainless steel column filled with alumina powder was employed. All chemicals were of pro analysi grade and were not further purified.
- 6 -
Cobalt sepulchrate was prepared according to the procedure given by Creaser 13. The deposition of metals and metal oxides on SiC was performed by using different methods. Copper (up to 10 nm) were deposited in vacuum by evaporation at an argon pressure of 5.10- 2 mbar. Platinum was deposited on SiC by photoelectrochemical and photochemical reduction from a solution of hexachloroplatinic acid in H2 O by UV-radiation.
Results Capacity
measurements: For any photoelectrochemical system the knowl-
edge of the energetic position of the band edges at the semiconductor is crucial because a charge transfer takes place only if there is a sufficient overlap between the electronic states in the semiconductor and in the electrolyte.
Corresponding
informations
can
be
obtained
by
measure-
ments of the space charge capacity. In Fig. 2 Mott-Schottky curves are presented as obtained with the hexagonal n- and p-type (6H-SiC) electrodes and in Fig. 3 for the cubic n-type crystals (n-3C-SiC). We found that in many cases the plot of 1/ C2 vs. UE does not yield straight lines as
shown
in
extrapolating yield
very
Fig. 2. the
plots
determination
Mott-Schottky
reproducible
Mott-Schottky
The
results.
vary
curves
of
considerably
flatband
potential
by
1/C2 vs. UE does also not
to
Especially
the
with
with
(6H-SiC)
time.
More
electrodes
the
reliable
data
were obtained with the cubic crystals (as shown in Fig. 3). In this case only a week drift of the Mott-Schottky have been found. The time effects observed with the (6H-SiC)-electrodes have been found in all electrolytes including 48% HF - solutions and for all electrodes. No surface pretreatment such as etching, polishing or soaking in the solution for a certain period of time, led to more reliable capacity measurements. The
- 7 -
Mott-Schottky curves given in Fig.2 have been obtained with (6H-SiC)electrodes during the first potential sweep after having kept the electrodes in the solution for at least one hour. They are given here in order to get a reasonable basis for comparing the flatband potentials and consequently the position of energy bands of the 6H- and 3C-SiC electrodes (see discussion section). Electrochemical
behavior: In order to examine the feasibility of splitting
water at SiC- electrodes, current-voltage measurements were conducted in aqueous electrolytes without any additional redox system. Some of these experiments are similar to those we have reported earlier.1,2 Since now also cubic SiC (3C-SiC) is available which has a smaller bandgap, it is interesting to compare the electrochemical behavior of both modifications.
In
Fig. 4 a
the
j/UE curve of a n-type 6H-SiC electrode in 0.1 M
solution of Na2 SO4 under
chopped
illumination
(white
light,
0.4 Wcm- 2 )
is shown. The photocurrent curve measured under illumination by chopped light shows a very slow rise between - 1 and +1.5 Volts with pronounced current transients which occur over the whole potential range when the light is turned on. Although the j/UE curve presented in Fig. 4a is typical for n-6H-SiC, the onset of photocurrent and the quantum efficiency varied considerably from one sample to the other. This is probably due to the heterogeneous nature of most rather crude SiC crystals from which we cut our individual samples. In Fig. 4 b a current-voltage dependence is shown as obtained with the cubic material (n-3C-SiC). This electrode consists of a thin n-3C-SiC layer on a highly doped silicon substrate and the
ohmic
contact
is
on
the
back
side
of
the
silicon
substrate
as
described in the experimental section. The j/UE -curve is very similar to that of the n-(6H-SiC) electrode (compare Figs. 4 a and b). The main difference lies in the more positive onset of the photocurrent and in the
- 8 -
current transients which are less pronounced and occur in a smaller potential range in the case of 3C-SiC. In the case of these two modifications we also measured the excitation spectra with n- type electrodes., as given in Fig. 5. These spectra were measured keeping the electrode at a potential where the photocurrent is about saturated, i.e. at 1 V (SCE). Note that two different current scales are used because the quantum efficiency is much smaller for 3C-SiC. The cathodic dark currents increase more or less exponentially with the electrode potential. According to our chemical analysis the corresponding
cathodic
current
is
entirely
due
to
hydrogen
formation,
i.e.
no
cathodic corrosion was found. In the case of the n-type 3C-SiC electrode the current - potential curves were quite well reproducible (Fig 6a). A cathodic shift of the curve with increasing pH was observed as quantitatively shown in Fig. 6b, similarly as in the Mott-Schottky measurements (Fig. 3).
The
shift
of
the
current-potential
curve
which
represents
the
hydrogen formation, amounts to about 60 mV (Fig. 6b). This value agrees well with the shift of the capacity curves measured with 3C-SiC electrodes. In both cases the shift agrees with what have been expected theoretically. The electrochemical behavior of p-type electrodes was examined in order to evaluate their possible use as photocathodes in a water splitting system. Since no p-doped cubic material was available only the hexagonal material was investigated in more detail. A current potential curve as obtained with a p-6H-SiC electrode is given in Fig. 7. In most cases the cathodic dark current is relatively high. Also the photocurrent- potential curve is far from an ideal shape. Although a photocurrent was detected positive of +1 V, which is close to the reported flatband potential, the
- 9 -
full photocurrent is only reached at rather negative potentials, whereas in the range between +1.2 and 0 V a smaller photocurrent plateau is observed . The curves given in Fig. 7 were obtained if an electrolyte of 40% HF in water was used in order to remove oxide layers from the electrode. In a further set of experiments we investigated the oxidation products formed during anodic polarization of p-type electrodes in the dark and of n-type samples upon illumination. We especially checked the head space of the electrolysis cell by GC-analysis for gaseous oxidation products such as oxygen. According to our experimental results performed with both types of electrodes in a sealed cell, however, no oxygen was found even after passing more than 1017 Coulomb across the interface. We also found that no oxygen was formed with electrodes, on which small Ptislands have been deposited. Since no H2 O2 could be detected either, the anodic current must be entirely due to anodic corrosion. With n-type electrodes we observed in addition that the photocurrent decreased with time in acidic solutions without HF as well as in concentrated hydrofluoric acid. In both cases the photocurrent-time curves showed a fast initial and later a slow degradation of the photocurrent over a period of several hours (Fig. 8). This behavior is in contrast to that of silicon electrodes
insofar
as
with
silicon
in
solutions
without
HF
the
anodic
current drops always to a very low value within a few seconds due to the formation of a thin impermeable insulating oxide layer. After
prolonged
current
flow
in
sulfuric
acid
solutions
the
electrodes
are covered with a thick white layer of silicon oxides which can be readily dissolved in HF. The head space analysis of gaseous products reveals CO and CO2 in a ratio close to 1:1 as products of the oxidation. On the
- 10 -
other hand, if the electrolysis was carried out in 48% hydrofluoric acid, then
the
surface
of
the
electrode
was
covered
with
a
yellow-brown
layer, which was insoluble in HF. These layers show surface features similar to those on the original surface and they were identified as silicon carbide by XPS and by chemical analysis. Current flow over periods of days have led to a complete destruction of the electrode. Redox
reactions: In a previous paper we have presented some current-
potential curves in the presence of a redox system at hexagonal n-type SiC
electrodes
(n-6H-SiC).2 In the present paper we also want to report
on redox processes at cubic electrodes. In Fig. 9 a-c current- potential curves for the electrochemical reduction of various redox systems at n3C-SiC
electrodes
are
given.
The
cathodic
saturation
currents
are
entirely due to mass transport limitation in the electrolyte. The further current increase beyond the saturation current corresponds to hydrogen evolution. In the case of cobalt (II/III) sepulchrate we have also measured the current-potential dependence at various pH-values (not shown). There not
it
shift
was
interesting
although
to
according
note to
that
the
capacity
current-potential measurements
curves
(Fig.
3)
did the
energy bands are shifted upwards upon increasing the pH of the solution. Another peculiar result is that no reoxidation of cobalt (II) sepulchrate could be found although its standard potential (-0.5 V vs.SCE) is rather negative. In addition we studied the reduction of CO2 at SiC electrodes which has been reported in the literature.5,14 When bare SiC- electrodes were used, no difference can be detected between j/UE -curves obtained in CO2 containing solutions and those in CO2
-free
electrolytes.
For
electrodes
that have been plated electrochemically with thin layers or islands of
- 11 -
copper, however, a clear difference is visible (Fig. 10). Copper was chosen as a catalyst because other authors had used copper electrodes to reduce CO2 .15 The reduction wave caused by CO2 ,however, is often small and ill defined. In order to obtain more evidence for the reduction of CO2 the following experiment was conducted: The electrode was held at a fixed potential of -1.35 V and the current was measured over a period of 600 s. In one case the solution was purged with nitrogen only, in the other case CO2 was bubbled into the solution for the first 300 s and then it was switched back to nitrogen (see insert in Fig. 10). It is clearly seen that CO2 is reduced at the copper plated SiC-electrode. This reduction was only observed at more or less neutral solutions, it did not occur in strongly acidic nor in strongly alkaline electrolytes. There is also no reduction of sodium carbonate solutions without purging it with CO2 . Since these solutions were buffered, a simple pH-effect due to the presence of CO2 in water according to the reaction CO2 + H2 O -----> CO3 2- + 2 H+
( 1 )
can be ruled out. In that case the reduction of protons might have been the observed electrochemical process. The reduction products were analyzed by gas chromatography. The only volatile product to be found was methane. In a further set of experiments we investigated the oxidation of some redox systems. These investigations were mainly focussed on the competition between the oxidation of a redox system and the anodic decomposition reaction. Since corresponding measurements had to be performed with RRDE-assemblies they could only be done with the hexagonal electrodes. It was not possible to make a rotating ring-disc electrode from the very thin cubic samples because the silicon substrate could not be
- 12 -
sufficiently isolated from the electrolyte. In solutions without any redox system we did not observe any ring current when the disc potential was scanned to positive values. This also proves that water is not oxidized at 6H-SiC. Upon addition of certain redox systems, however, the oxidation of the SiC electrode is clearly reduced in favor of the oxidation of the redox system itself. The best stabilization was obtained with redox systems, the oxidation of which occurs at lower anodic potentials than the anodic decomposition of SiC. This is illustrated in Fig. 11 a and b for the redox couple Cu+ / Cu2 + in chloride solutions. In the same range also the ring current rises but decreases to a very low value at a potential around this
+2
V
(Fig. 11 b).
hysteresis
(Fig. 11 a).
The
corresponding
This
result
disc
indicates
current that
also
the
reflects
oxidation
of
the redox system is blocked once a corrosion current is reached exceeding that caused by the redox reaction. The original situation can not be restored
by
cathodic
polarization
but
only
by
polishing
the
electrode
with a diamond paste. Similar observations have been made for the oxidation of bromide, i.e. the latter process only took place if a large corrosion process was avoided (not shown). Only in the case of the I- / I3 - couple the oxidation continued even in the range of higher corrosion currents probably due to the adsorption of the iodide.
Discussion Potential obtain
distribution:
straight
As
Mott-Schottky
already plots
mentioned and
it
is
above
it
is
difficult
to
even
more
difficult
to
derive a reliable flatband potential because of changes with time. Our measurements have shown that, even with identical electrodes and under identical conditions, values of the flatband potential can vary by more
- 13 -
than one volt. This result indicates that the surface chemistry and especially the formation of a surface layer plays a dominant role. A comparison of papers published by different authors also shows a wide range of values for the flatband potential of both, n- and p-type electrodes (see Table 2). This scattering of data has probably the same origin as in our case. On the other hand, the quality of crystals plays an important role. This becomes quite obvious by comparing the data obtained with epitaxially grown cubic crystals and the hexagonal SiC produced in high temperature
process.
Since the Mott-Schottky curves were measured under the same conditions, the flatband potentials obtained with n-6H (Uf b = -1.25 V at pH0) and n-3C-SiC electrodes (Uf b =-0.35 V at pH0) can be compared. According to these results the difference of the flatband potentials amount to ∆ U =0.9 V, i.e. a value which is slightly above the difference of the fb bandgaps of the two modifications (∆ Eg=0.7 eV). Taking into account the uncertainty in the determination of the flatband potential it can be con-
cluded from this result that the conduction band of 3C-SiC is by 0.7-0.9 eV lower than that of 6H-SiC whereas the position of the valence band is nearly identical for both types of electrodes. There also seems to be no agreement on the pH- dependence. Whereas some authors report a pH- dependence of about 59 mV per unit pH, others have seen no pH dependence at all. We found with the n-3C-SiC electrodes clearly a pH dependence of 52 mV per unit pH, which is slightly smaller than the expected 59 mV (Fig. 3). The same shift by pH was found in the onset of photocurrent. The pH dependence does certainly also exist for 6H-SiC but it was less pronounced and could not be quantified because of poor reproducibility. It can be concluded at least from these
- 14 -
results that the surface should be OH-terminated. In general, the pH dependence of the flatband potential is attributed to the dissociation of surface groups which depends on the pH as it has been found with various semiconductors
(see
e.g.19). The surface of SiC is probably terminated
with Si-OH groups or perhaps also with C-OH groups. It should also be mentioned that the Mott-Schottky curves are shifted upon illumination as observed with many semiconductors (see e.g.20). This result
indicates
that
minority
carriers
are
not
transferred
across
the
interface at a very high rate but that they are trapped in some surface states. This has not yet been further investigated. Hydrogen evolution and anodic decomposition: As proved analytically, SiC is stable under cathodic polarization and the corresponding current is entirely due H2
-formation.
The
shift
of
the
current-potential
curves
with pH (Fig. 6b) amounts to about 60 mV per pH unit which is in accordance with the shift of the Mott-Schottky plots (Fig. 3). The cathodic photocurrent
observed
with
p-type
electrodes
exhibits
a
considerable
overvoltage (Fig. 7) because the onset only occurs at around +1.2 V whereas the flatband potential can be estimated to be at around +2 V (Fig.
2 b).
The
nature
of
the
shoulder
in
the
photocurrent-potential
dependence at p-6H-SiC is not yet understood. The current-potential curves for the n-type 6H-SiC and the 3C-SiC materials are rather similar although with the latter electrode higher photocurrent densities and a steeper photocurrent onset were found (compare Figs. 4 a and b). This may be partly due to the better quality of the cubic crystal. On the other hand the cubic crystal absorbs already at higher wavelengths as shown by the excitation spectrum (Fig. 5). However, the quantum efficiency found with the cubic electrodes in the long wave-
- 15 -
length region is considerably smaller than that found for hexagonal electrodes. Since the energy gap of the cubic crystal corresponds to an indirect transition, the absorption coefficient is expected to be small, which would lead to a relatively large penetration depth of light. Obviously the diffusion length of holes in the epitaxially grown films on highly doped silicon substrates is smaller than the penetration depth of light which causes the low quantum yield. Concerning the onset of the anodic photocurrents for both electrodes, n-(6H-SiC) and n-(3C-SiC) a considerable overvoltage was found (compare Fig. 4 with Figs.2 and 3). We will return to this effect below. The analysis of the oxidation products during anodic current flow in the dark or upon illumination has clearly shown that no oxygen was formed but
only
corrosion
takes
place.
In
solutions
without
HF
white
silicon
oxide layers were formed. It has been suggested by other authors 2 1 an eight hole process for the dissolution as given by: SiC + 8 h+ + 4 H2 O ----> SiO2 + CO2 + 8 H+ .
(2)
This reaction describes the formation of CO2 . The production of CO can only be explained by a 4-hole process such as: SiC + 4 h+ + 2 H2 O ----> SiO + CO + 4 H+ .
(3)
If both reactions proceed at the same rate the resulting net reaction is given by: 2 SiC + 12 h+ + 6 H2 O ----> SiO2 + SiO + CO2 + CO + 12 H+ (4) i.e. a 6-hole reaction per SiC molecule is assumed. Recently, J. Shor found experimental evidence for a 6-hole corrosion reaction by comparing the consumed charge with the amount of material etched away from
- 16 -
the surface 10. These experiments were conducted with cubic 3C-SiC electrodes
whereas
we
used
the
hexagonal
6H-modification.
Thus
far,
however, the chemistry of both modification must be assumed to be very similar
18
.
Another interesting finding is the formation of an amorphous layer of SiC on the surface of the SiC electrodes when anodically etched in concentrated HF solutions. This behavior is very similar to the formation of porous Si on the surface of silicon electrodes etched under similar conditions
22
by
similar
a
. A paper by Brander and Boughey suggests the formation of SiC mechanism.23 Such a reaction, however,
disproportionation
would be very unlikely because it involves the disproportionation of the free species such as CF2 and SiF2 to form SiC, SiF4 and CF4 . More likely is the gradual braking and forming of bonds as illustrated in Fig. 12. Recent SEM investigations by J. Shor have shown that these layers have a porous texture 24 and that their formation can indeed be compared with the formation of porous silicon. All these investigations have proven that no oxygen is formed although SiC fulfills all energetic conditions, such as width of bandgap and positions of the conduction and valence bands with respect to the reduction potential of the H2 O/H2 couple and the oxidation potential of the H2 O/O2 system,
respectively,
splitting.
These
towards
oxidation
as
findings
required reflect
for
the
the
photoelectrochemical
thermodynamic
instability
water of
SiC
(∆ G= -1180 kJ/mole for the reaction SiC + 2O2 -->
CO2 + SiO2 ) and show that the well known chemical stability has only kinetic reasons. On the other hand, this instability toward anodic corrosion
offers
conditions
the rather
possibility than
of
using
etching
SiC
electrochemically
molten
salts
at
very
high
under
mild
temperatures
- 17 -
which for device fabrication will be of great importance. This has been demonstrated by Shor and Kurtz who produced nice etch patterns by photoelectrochemistry
11
.
The question arises whether a photoelectrolysis cell, consisting of a ptype SiC and a Pt counter electrode could be made, in which H2 is formed at the p-type electrode and oxygen at the Pt- electrode during illumination. Although the thermodynamic requirements are fulfilled (Fig. 7), not a sufficient photovoltage can be developed because of the rather large overvoltage for the onset of the photocurrent at the p-type electrode. Accordingly, such a cell could only be operated by applying an external bias. Charge transfer processes with redox systems: According to our experimental results there are a number of redox systems which can be reduced by an electron transfer from the conduction band to the oxidized species of the redox couple. This can already be concluded from the currentpotential (Fig. 9 a-c)
curves since
which the
were
obtained
corresponding
with
currents
n-type occur
SiCat
less
electrodes cathodic
potentials than the H2 -evolution. From this result it can be concluded that the rate constants for these redox reactions are considerably higher than that for the reduction of water. Since this finding is in agreement with common charge transfer models it may not be worth mentioning. However,
with
several
other
semiconductor
electrodes,
such
as
for
instance n-GaAs which is much better defined than SiC, this result is not obtained 25,29. From this point of view we expected SiC to be a very suitable semiconductor for studying basic charge transfer processes in more detail. For instance, according to the usual charge transfer model it is required that the cathodic current should be proportional to the concen-
- 18 -
tration of the electron acceptor in the solution and to the electron density at the surface, i.e. j c-
= - kc- cox ns = - kc- cox nso exp
( eUE -
eUredox
)
/ kT .
(5)
in which j-c is the cathodic exchange current density via the conduction band, k-c the corresponding charge transfer rate constant, cox the concentration of the oxidized species at the electrode surface, nso the electron density
at
the
semiconductor
surface
at
the
redox
potential,
UE and
Uredox the electrode potential and the electrochemical potential of the redox couple, respectively. The current is essentially determined by ns and the rate constant kc- , the latter depending on the reorientation energy. Since a relatively large overvoltage with respect to the redox potential was found for the reduction of Fe3+ we assume at present that the rate constant is rather small for this process. Accordingly, the current should increase by one order of magnitude when the electrode potential is varied by 60 mV. Unfortunately, an evaluation of the current-voltage curves in Fig. 9 yielded for all examples a much weaker current increase. It is not sure whether this is due to an oxide layer across which some of the externally applied potential occurs, or whether the flatband potential shifts in the range of the cathodic current flow as it has been found for GaAs.27 In principle, this can be analyzed by using impedance spectroscopy. Since, however, the capacity drifts considerably, it was impossible to obtain reliable impedance data. Similar
observations
were
made
for
the
oxidation
of
redox
systems.
However, a p-type SiC electrode changes irreversibly if polarized so far anodically that the corrosion current considerably exceeds the diffusion current of the redox system. Here, the oxide formed in the corrosion
- 19 -
reaction,
inhibits
the
redox
reaction
(Fig. 11).
Although
the
spongy
nature of the oxide layer still permits a current flow, an electron donor such as Cu+ or Br- obviously cannot diffuse through this oxide layer. An exception is the oxidation of I– with which no hysteresis occurred even after strong anodic polarization. This is probably caused by the strong adsorption of the I2 / I -
on the SiC surface which seems to prevent effi-
ciently the oxidation of these adsorption sites.
Conclusion We have shown that in spite of the relatively large bandgap and the favorable position of the band edges at the surface neither with n- nor with
p-type
should
be
investigations
SiC
electrodes
mentioned with
that
SiC
H2 O we
-photoelectrolysis
have
particles
also
loaded
performed with
was
detected.
some
metals
It
preliminary
and/or
metal
oxides as catalysts. Also in this case no photoelectrolysis was observed. Interestingly, however, fuels such as methane can be produced in ethanol containing solutions. The mechanism is not yet understood. Further studies with SiC monograin membranes in which the particles are geometrically fixed and on which a catalyst can be deposited in a defined way, may help to solve the reaction mechanism.
- 20 -
Acknowledgments: The authors are indebted to Prof. Dr. B. Kastening, Hamburg, for many valuable discussions and support. We also thank Marion Bahnemann and Ursula Neff for performing some of the electrochemical measurements. The supply of SiC-samples by various companies and institutions listed in Table 1 is acknowledged. This work was financially supported by the German Federal Minister for Research and Technology (BMFT) under contracts No. 329081A and 329580 as well as by the Volkswagen Foundation under contract number I/72 365.
- 21 -
References 1.
R. Memming, Photochem.
Photobiol., 1 6 , 325 (1972).
2.
M. Gleria and R. Memming, J. Electroanal. Chem., 6 5 , 163 (1975).
3.
Y.V. Pleskov and Y.Y. Gurevich, in Modern Aspects of Electrochem-
i s t r y , Vol.16, B.E. Conway, R. E. White and J. O. Bockris, eds., p. 189, Plenum Press, New York 1985. 4.
M. D. Krotova and Y. V. Pleskov, Soviet
Electrochemistry, 1126
(1981). 5.
R. L. Cook, R. C. MacDuff and A. F. Sammells, J. Electroanal. Chem., 1 3 5 , 3069 (1988).
6.
S. Yamamura, H. Kojima, J. Ioda and W. Kawai, J.
Electroanal.
Chem., 2 2 5 , 287 (1987); J. Electroanal. Chem., 2 4 7 , 333 (1988). 7.
S. Mirsagatow and M. Duisenaev, Geliotechnika , 3 , 8 (1977).
8.
T. Inoue and T. Yamase, in 4th Int. Conf. of Photochemical Conver-
sion and Storage of Solar Energy, Book of Abstracts, J. Rabani, ed., p. 37, Jerusalem (1982). 9.
T. Inoue and T. Yamase, Chem. Lett., 7 , 869 (1985).
10.
J. S. Shor, I. Grimberg, B.-Z. Weiss and A. D. Kurtz, Appl.
Phys.Lett.,
6 2 , 2836. (1993). 11.
J. S. Shor and A. D. Kurtz, J. Electrochem. Soc., 1 4 1 , 778 (1994).
12.
H. Daimon, M. Yamanaka, E. Sakuma, S. Misawa and S. Yoshida, Jap.
J. Appl. Phys., 2 5 , L 592 (1986). 13.
I. I. Creaser, R. Geue, J. M. Harrowfield, A. Herrit, A. M. Sargeson, M. R. Snow and J. Springborg, J. Am. Chem. Soc., 1 0 4 , 6016 (1982).
14.
J. O. Bockris and K. Uosaki, J. Electrochem. Soc., 1 2 4 , 1348 (1977).
15.
Y. Hori, K. Kikuchi, A. Murata and S. Susuki, Chem. Lett., 897 (1986)
16.
H. Morisaki, H. Ono and K. Yazawa, J. Electrochem. Soc.,1 3 1 , 2081 (1984)
- 22 -
17.
A. Mannivannan, A. Fujishima and G. V. S. Rao, Ber. Bunsenges. Phys.
Chem., 9 2 , 1522 (1988). 18.
I. Lauermann, D. Meissner, R. Memming, R. Reineke and B. Kastening in Dechema-Monographie, chaft,
19.
Frankfurt
Vol.
124,
p.617,
VCH
Verlagsgesells-
(1991).
R. Memming in Comprehensive
Treatise
of
Electrochemistry, Vol.
7, B. E.Conway and J. O. Bockris eds., p. 529, Plenum Press, New York (1983). 20.
R. Memming in Topics
of
Current
Chemistry, Vol.169, M. Mattey,
ed., p.105, Springer-Verlag, Berlin (1994). 21.
H. Hirayama, T. Kawabuko, A. Goto and T. Kaneko, J. Am. Ceram.
Soc., 7 2 , 2049 (1989). 22.
R. Memming and G. Schwandt, Surf. Sci., 5 , 109 (1966).
23.
R. W. Brander and A. L. Boughey, Brit. J. Appl. Phys., 1 8 ,
905
(1967). 24.
J. S. Shor, J. Grimberg, B. Z. Weiss and A. D. Kurtz in Meeting of the
Electrochem. Soc., Abstract No.634, Honolulu (1993). 25.
D. Meissner, Ch. Sinn, R. Memming, P. H. L. Notten and J. J. Kelly in
Homogeneous and Heterogeneous Photocatalysis, E. Pelizzetti and N. Serpone, eds., p.343, D. Reidel Publ., Dordrecht (1986). 26.
R. Memming, Ber. Bunsenges. Phys. Chem., 9 1 , 353 (1987).
27.
I. Uhlendorf, R. Reineke-Koch and R. Memming, J.Phys.Chem.,in press
28.
R. Memming in Topics in Current Chemistry, Vol.143, E. Steckhan, ed., p. 79, Springer Verlag, Berlin (1988).
- 23 -
Table 1: Crystal sources
source
crystal
doping
doping
type
density
type
color
_________________________________________________________ 1
3
4
5
6H
high
n
green
6H
high
p
blue
low
p
blue
6H
high
n
green
6H
high
p
blue
3C
10 1 7 - 1 0 1 8 cm- 3
n
yellow
3C
10 1 6 - 1 0 17 cm - 3
n
yellow
3C
high
n
dark
6H
high
n
green
_______________________________________________________ 1
Elektroschmelze Kempten GmbH, München.
2
Siemens AG, Zentrale Forschung und Entwicklung, München.
3
Showa Denko, Ceramics Div., Tokyo.
4
Dr. R. F. Davis, North Carolina State University, Dpt. of Material Science and Engineering, Raleigh, NC.
5
Sharp Corporation, Tokyo.
- 24 -
Table 2:
Values of flatband potentials in Volt (SCE) for n- and p-(6HSiC) as reported by different author
n-SiC
pH
p-SiC
pH
Reference
________________________________________________________ -1,6
1.3
2
+1.44
0
14
+1.14
14
14
13.0
16
-1.45 to -1.65 -0.7
1
+1.6
0.3
9
-0.96
0.3
+1.52
0.3
17
-1.66
13.0
+1.3
13.0
17
- 25 -
Figure Fig. 1:
Captions
Contacting methods for 3C-SiC on Si substrates, a) via SiC; b) via Si; c) via SiC with subsequent removal of the substrate by etching
Fig. 2:
Mott-Schottky plots for 6H-SiC electrodes in acid and alkaline solutions. a): n-type b) p-type Measuring frequency; 20 kHz; scan rate 0.05 V/s
Fig. 3:
Mott-Schottky plot for an n-type 3C-SiC electrode at various pH-values
Fig. 4:
Current-voltage dependence for a n-6H (a) and a n-3C-SiC (b) electrode
under
intermittent
light
excitation
at
pH 7;
N2 -
purged; scan rate: 50 mVs- 1 Fig. 5:
Photocurrent excitation spectra of 6H- and 3C-SiC electrodes in 1 M NaOH at UE= + 1 V
Fig. 6:
a)Current-voltage curves at an n-type 3C-SiC electrode with respect to the reduction of H2 O at different pH-values. b) Potentials at which a current of 50 µ A was measured vs. pH
Fig. 7:
Current-voltage dependence for a p-type 6H-SiC electrode in a 40% HF solution
- 26 -
Fig. 8:
Anodic photocurrent vs. time at an n-type 6H-SiC electrode at pH 4,
electrode
potential:
+1
V,
illumination:
white
light
(0.2 Wcm - 2 ) Fig. 9:
Reduction of various redox systems at an n-type 3C-SiC electrode in 1 M HCl; scan rate: 10 mVs- 1
Fig. 10:
Reduction of CO2 at an n-type 6H-SiC electrode loaded with Cu in the dark in 0.1 M Na2 SO4 solution. Scan rate: 0.1 Vs- 1 Insert: reduction current vs. time at UE= -1.35 V
Fig. 11:
Current-voltage curves for a p-type 6H-SiC electrode at pH 1 in the presence of 25 µM Cu1 + , 1000 rpm; scan rate: 0.1 Vs- 1
Fig. 12:
Model for the rearrangement of the surface layer during anodic polarization
in
HF-solutions
- 27 -
a
a
In In
epoxy
epoxy
ccontact ontact
SiC SiC
SSii
b
In
contact
SiC
Si
c
In
epoxy contact
SiC
Figure 1
- 28 -
[10 - 1 2cm 4 / F 2 ]
8
- 2
a
1 0
2
pH 14 6
pH 0
C
4
0
- 2 -1.5 - 1 -0.5 potential
0 [V
0.5 vs
1
1.5
2
SCE]
b
Figure 2
[10 - 1 2cm 4 / F 2 ]
- 29 -
pH 14
C
- 2
pH 10
pH 4,7 pH 0
- 2
- 1
0 1 2 potential [V vs SCE]
3
Figure 3
b [ µ A / cm 2]
current
[ µ A / cm 2]
a
current
- 30 -
250
500 1 2
1
2
Figure 4
0.24
1
0.2
0.8
0.16 n-6H-SiC
n-3C-SiC
0.6
0.12
0.4
0.08
0.2
0.04
0
0 -0.04
-0.2 300
[ µ A / cm 2]
current
1.2
current
[ µ A / cm 2]
- 31 -
380
460 wavelength [nm]
540
620
Figure 5
- 32 -
a
potential [V vs SCE]
current
pH 14
[ µ A / cm 2]
pH 4.7 pH 6.8
pH 2
pH 0
pH 10 pH 11.5
b
Figure 6
- 33 -
current
[ mA / cm 2 ]
0.25 0.2 0.15 0.1 0.05 dark
0 -0.05
illuminated
-0.1 -0.15 - 1
-0.5
0
0.5
1
1.5
2
Figure 7
- 34 -
3.8
current
[ µ A / cm 2]
3.4 3 2.6 2.2 1.8 1.4 1 0
2000
4000 time [s]
6000
Figure 8
- 35 -
a
0.5
-0.5 [V vs SCE]
3 -
16.7 mM [Fe(CN) 6
]
current [ µ A/cm 2]
potential
-500
b -0.5
potential
[V vs SCE]
Fe 3 +
-1000
[ µA / cm 2]
4.7 mM
current
- 1
c potential
current
- 1
[V vs SCE] -0.5
0.4 mM
1 mM
Co(sep.) 3 +
[ µA / cm 2]
-100
Figure 9
- 36 -
N2 CO2
-0.4 -0.6 -0.8 -1.6
[ µ A / cm 2 ]
-0.2 -0.05
-0.1
current
current
[ µ A / cm 2 ]
0
-0.15
stop purging CO 2
CO -0.2
0
100
200
300
time
400
500
600
[s]
-1.2 -0.8 -0.4 potential [V vs SCE]
0
Figure 10
- 37 -
a) disc current
0 0 0
without Cu
current
[ µA ]
-0.8 -1.6
with Cu -2.4 -3.2
b) ring current - 4
-0.8
0 0.8 1.6 potential [V vs SCE]
2.4
Figure 11
- 38 -
F C C
F
C
Si
C
C
Si
C
C
Si
C
C
Si
C
F F
F
Si
C
C
Si
C
C
Si
C
F OH
C
Si
C
C
OH
C
F
C
F F Si
F
F
F F
F C
F
F
F F F F F C
C Si
F F C F
C
F
Si F C
F OH
Si
C
C
OH
F
F
F
F
F
F C
F
F
F Si
C
F OH
F
Si
C
C
OH
F
Figure 12
