An approach to ideal semiconductor electrodes for efficient photoelectrochemical reduction of carbon dioxide by modification with small metal particles

Article abstract of DOI:10.1021/jp972663h

Photoelectrochemical reduction of carbon dioxide (CO₂) on p-type silicon (p-Si) electrodes modified with small metal (Cu, Ag, or Au) particles has been studied. The electrodes in CO₂-saturated aqueous electrolyte under illumination produce methane, ethylene, carbon monoxide, etc., similar to the metal (Cu, Ag, or Au) electrodes, but at ca. 0.5 V more positive potentials than the corresponding metal electrodes, contrary to continuous-metal-coated p-Si electrodes. The results clearly show that the metal-particle-coated p-Si electrodes not only have high catalytic activity for electrode reactions but also generate high photovoltages and thus work as an ideal type semiconductor electrode. It is discussed that the CO₂ photoreduction proceeds with an upward shift of the surface band energies of p-Si in order to get energy level matching between the semiconductor and solution reactants, though hydrogen photoevolution occurs without such an upward shift. It is also discussed that the control of surface structure on a nanometer-sized level, as well as on an atomic scale, is important for getting higher efficiencies.

Full text of DOI:10.1021/jp972663h

974
J. Phys. Chem. B 1998, 102, 974-980
An Approach to Ideal Semiconductor Electrodes for Efficient Photoelectrochemical
Reduction of Carbon Dioxide by Modification with Small Metal Particles
R. Hinogami, Y. Nakamura, S. Yae, and Y. Nakato*
Department of Chemistry, Graduate School of Engineering Science, and Research Center for Photoenergetics
of Organic Materials, Osaka UniVersity, Toyonaka, Osaka 560-8531, Japan
ReceiVed: August 14, 1997; In Final Form: NoVember 10, 1997
Photoelectrochemical reduction of carbon dioxide (CO2) on p-type silicon (p-Si) electrodes modified with
small metal (Cu, Ag, or Au) particles has been studied. The electrodes in CO2-saturated aqueous electrolyte
under illumination produce methane, ethylene, carbon monoxide, etc., similar to the metal (Cu, Ag, or Au)
electrodes, but at ca. 0.5 V more positive potentials than the corresponding metal electrodes, contrary to
continuous-metal-coated p-Si electrodes. The results clearly show that the metal-particle-coated p-Si electrodes
not only have high catalytic activity for electrode reactions but also generate high photovoltages and thus
work as an ideal type semiconductor electrode. It is discussed that the CO2 photoreduction proceeds with an
upward shift of the surface band energies of p-Si in order to get energy level matching between the
semiconductor and solution reactants, though hydrogen photoevolution occurs without such an upward shift.
It is also discussed that the control of surface structure on a nanometer-sized level, as well as on an atomic
scale, is important for getting higher efficiencies.
Introduction
The purpose of the present paper is to study the PEC reduction
of CO2 on metal (Cu, Ag, or Au) particle-modified p-Si
electrodes (cf. Figure 1) and investigate how efficiently the
electrodes of this type work for such a slow reaction as the
CO2 reduction with a high activation energy. The Cu, Ag, and
Au metals were chosen because they were reported to work as
efficient electrocatalysts for the CO2 reduction.15,16 Parts of
preliminary results were reported in a letter17 and a short paper.18
Photoelectrochemical (PEC) reduction of carbon dioxide
(CO2) with a p-type semiconductor electrode is one of the solar
energy conversion technologies and is important from a
viewpoint of the global environmental problems. The method
can be regarded as an artificial model for photosynthesis in
natural plant and is interesting in this respect.
A number of studies have been made on the PEC reduction
of CO2 on p-type semiconductor electrodes such as p-CdTe,1-3
p-InP,3,4 p-GaP,3,5,6 p-Si,3,7 p-SiC,8 etc., with or without metal
coating, and some reduction products such as formic acid and
carbon monoxide are reported to be produced with relatively
high current efficiencies. However, the photovoltage or the
solar-to-chemical energy conversion efficiency still remains
relatively low. The main reason lies in that it is very difficult
to meet all requirements for high efficiency, such as high
catalytic activity at the electrode surface, sufficient passivation
of the electrode surface (or little surface carrier recombination),
formation of a high energy barrier at the semiconductor/solution
interface, and good energy level matching between semiconduc-
tor and solution reactants for little energy losses in interfacial
reactions. The difficulty is serious because the above require-
ments are not necessarily compatible with each other for most
semiconductor electrodes. For example, the first and the second
requirements are incompatible because catalytically active
surface sites usually act as effective surface recombination
centers. They also produce surface states which may cause
Fermi level pinning leading to a low energy barrier height.
Experimental Section
Single-crystal p-Si wafers [CZ, (100), 1.3-1.75 Ω cm, or
(111), 0.85-1.05 Ω cm, Shin-Etsu Handotai Co., Ltd.] were
cut into pieces (about 1.0 × 1.0 cm2 in area), washed with
boiling acetone to remove organic contamination on the surface,
etched in CP-4A (a mixture of 47% HF, 60% HNO3, 99.7%
CH3COOH, and H2O in a volume ratio of 3:5:3:22), and finally
etched again with 12% HF. Ohmic contact was made on the
rear side of the Si piece with indium-gallium-zinc alloy. The
p-Si piece thus obtained was placed in a Teflon holder and used
as a p-Si electrode (effective area: 0.25 or 0.50 cm2).
Copper (Cu) particles were deposited on p-Si in 0.01 M
CuSO4 + 0.18 M H2SO4 (M ) mol/dm3) under illumination
with a tungsten-halogen lamp, with the potential kept at -0.1
V vs SCE. The current density decreased initially from ca. 5.0
to ca. 0.8 mA/cm2 but became constant in ca. 30 s. Ag and Au
particles were deposited similarly at 0.0 V vs SCE in 0.01 M
AgNO3 + 0.66 M HNO3 and in 0.01 M HAuCl4, respectively.
The electricity passing across the electrode surface was mea-
sured with a Coulomb meter for all the cases and regulated such
that the volume of the deposited metal was nearly the same
((2.7-3.0) × 10-12 m3) among the Cu, Ag, and Au metals. In
other words, the electricity was regulated to be 80, 28, and 80
mC/cm2 for the Cu, Ag, and Au deposition, respectively, by
taking account of the differences in the lattice constant of the
metals and the number of electrons needed to reduce one metal
ion. For comparison, continuous-metal layers were deposited
We have found, through studies of n-Si-based PEC solar
cells,9,10 that a semiconductor electrode modified with small
metal particles (e.g., n-Si modified with a Langmuir-Blodgett
layer of colloidal Pt particles11,12) can meet all the requirements
mentioned above and thus becomes an ideal type semiconductor
electrode. This principle was applied successfully to hydrogen
photoevolution on platinum-particle-modified p-Si13 and p-InP.14
S1089-5647(97)02663-1 CCC: $15.00 © 1998 American Chemical Society
Published on Web 01/22/1998

Photoelectrochemical Reduction of CO2
J. Phys. Chem. B, Vol. 102, No. 6, 1998 975
Figure 2. Scanning electron micrographs of p-Si with photoelectro-
chemically deposited metals: (A) Cu, (B) Ag, and (C) Au.
The j-U curves for the CO2 photoreduction were measured
under continuous bubbling of CO2 (99.99% in purity) through
the catholyte under magnetic stirring, but photoelectrolyses for
product analyses were performed under a sealed condition;
namely, CO2 was first bubbled into the stirred catholyte until
the air in the cathode compartment was completely substituted
for CO2, then the cathode compartment was sealed, and the
photoelectrolysis was carried out potentiostatically under stirring
for a certain period of time with recording of the current density.
The temperature of the catholyte was kept at 20 ( 1 °C. To
remove heavy metal ions contained as impurities in the
catholyte, it was beforehand purified by preelectrolysis with a
Pt-black electrode at -0.7 V vs SCE for more than 15 h under
a N2 atmosphere. Special-grade chemicals and water purified
with a Milli-Q water purification system (Nihon Millipore
Kogyo) were used for all experiments.
Reduction products in the gas phase of the cathode compart-
ment were analyzed by gas chromatography. An active carbon
column was used for analysis of CO and hydrocarbons, and a
molecular sieve 13X column was used for H2, N2, and O2.
Organic acids in the catholyte were analyzed by high perfor-
mance liquid chromatography, using a TSK-gel SCX(H+)
column (Tosoh). Alcohols in the catholyte were analyzed with
gas chromatography using a TSG-1 15% column with a support
of Shincarbon A 60-80 (Shinwa Kakou).
Figure 1. Schematic cross sections of p-Si electrodes coated with small
metal particles, together with some electronic processes at the surface:
(A) model electrode and (B) actual electrode. See the Discussion section
for details.
on p-Si by the vacuum evaporation method under a pressure
less than 1 × 10-5 Torr (1 Torr ) 133.3 Pa). Chips of Cu
(99.994% in purity), Ag (>99.95%), and Au (99.99%) were
used as a target. The thickness of the metal layer was monitored
with a quartz oscillator placed perpendicular to the direction of
the evaporated metal vapor. The thickness was regulated to be
10-30 nm.
Copper metal electrodes were prepared using Cu rod (99.999%)
or Cu sheet (99.994%). They were in some cases electrochemi-
cally polished at 2.0 V vs a Cu-plate counter electrode in 85.0%
phosphoric acid for ca. 1 min. Silver electrodes, made of Ag
plate (99.99%), were etched with a mixture of 60% HNO3 and
methanol (volume ratio 1:9). Gold electrodes, made of Au plate
(99.99%), were etched with aqua regia.
The PEC reduction of CO2 was performed using an H-shaped
Pyrex cell, with a saturated calomel electrode (SCE) as the
reference electrode and a Pt plate as the counter electrode. The
cathode and anode compartments were separated with a cation
exchange membrane (Nafion 117) in order to avoid mixing of
products at the cathode and the anode. The electrolyte was
aqueous 0.1 M KHCO3. Current-potential (j-U) curves were
measured with a potentiostat (Nikko-Keisoku NPOT-2501) and
a function generator (Hokutodenko HB-III). The ohmic (iR)
drop in the electrolyte (0.1 M KHCO3) between the cathode
and SCE, reaching ca. 0.5 V at 10 mA/cm2, was corrected by
measuring the solution resistance by an iR compensation
instrument (Hokutodenko HI-203). A tungsten-halogen lamp
was used as the light source. The illumination intensity was
adjusted such that it gave nearly the same photocurrent as
illumination with a solar simulator (AM 1.5 G 100 mW/cm2,
Wacom WXS-85H) for easy comparison with reported solar-
cell characteristics.9-12
The electrode surface was inspected with a high-resolution
scanning electron microscope (Hitachi S-5000). The surface
composition was analyzed by an X-ray photoelectron spectrom-
eter (XPS, Shimadzu ESCA-1000).
Results
1. Surface Structure of Metal Deposited p-Si. Figure 2
shows scanning electron micrographs (SEMs) of p-Si electrodes
coated with photoelectrochemically deposited Cu, Ag, and Au.
All metals are deposited in the form of small particles of 20-
200 nm in size. The particle density was about 2 × 109 cm-2
for Cu and Ag and about 2 × 1010 cm-2 for Au, as counted
from the SEMs. The Au particles are so dense that they form

976 J. Phys. Chem. B, Vol. 102, No. 6, 1998
Hinogami et al.
aggregates. The inclined SEMs (the right-hand side of Figure
2) show that the particles are not spherical but have a rocky
shape. Because the metals were deposited such that their
volumes were nearly the same as each other, as mentioned in
the preceding section, the difference in the particle density
among Cu, Ag, and Au may partly be due to the difference in
the particle shape. Another possible reason is efficient hole
injection by AuCl4- (and Ag+) ions compared with Cu2+ (i.e.,
efficient reduction of the ions (or efficient deposition of Au
and Ag)) accompanied by formation of Si oxide layer as
described later, which occurs with no flow of Faradaic current.
SEM inspection showed that the increase of the electricity
passing across the electrode surface did not affect the particle
density, only increasing its size. On the other hand, vacuum-
deposited metals existed in the form of nearly continuous layers,
contrary to the photoelectrochemically deposited metals.
XPS analysis showed that the p-Si surfaces after the photo-
electrochemical deposition of metals were covered with Si
oxide layers, the thickness being calculated to be 0.3 ( 0.1 nm
for Cu, 0.5 ( 0.2 nm for Ag, and 1.5 ( 0.3 nm for Au according
to a reported method.19 The formation of such Si oxide layers
was also observed for p-Si simply immersed in the solutions of
Cu2+, Ag+, and AuCl4-. This indicates that the metal ions can
inject holes into the valence band of p-Si, leading to the Si oxide
formation and metal deposition.20-22 The difference in the oxide
thickness is probably due to the difference in the reduction
potential of Cu2+ (0.10 V vs SCE), Ag+ (0.56 V vs SCE), and
-
AuCl4 (0.76 V vs SCE) ions.
2. Hydrogen Photoevolution on Particulate-Metal/p-Si.
Hydrogen evolution on Cu, Ag, and Au occurs more efficiently
(i.e., with much less overvoltages) than the CO2 reduction.
Therefore, to investigate how well the metal particles were
deposited, we first measured the j-U curves for hydrogen
photoevolution on the p-Si electrodes coated with photoelec-
trochemically deposited metal particles (hereafter denoted as
particulate-metal/p-Si electrodes) in 1.0 M HCl (pH 0.2). Figure
3 summarizes the j-U curves for various related electrodes.
No correction for the iR drop was made in this case, because
the 1.0 M HCl solution has a high conductivity and the iR drop
is negligibly small, contrary to the case of 0.1 M KHCO3
described later.
From a glance at Figure 3, it is clear that the onset potentials
of photocurrents for the particulate-metal (Cu, Ag, and Au)/p-
Si electrodes (after HF etching for the Ag and Au-coated ones)
are much more positive than that for a naked p-Si electrode.
Decreased photocurrent densities for the particulate-metal/p-Si
electrodes are due to decreases in incident light intensity by
the deposited metal particles. Moreover, the onset potentials
of photocurrents for the particulate-metal/p-Si electrodes lie
much more positive than those for the corresponding metal (Cu,
Ag, and Au) electrodes, indicating that high photovoltages are
generated in p-Si, and also are 0.20-0.25 V more positive than
the equilibrium hydrogen-evolution potential (-0.25 V vs SCE)
in 1.0 M HCl (pH 0.2). Poor j-U curves before the HF etching
for the particulate-Ag and Au/p-Si (Figure 3) are due to the
presence of Si oxide layers for the as-deposited electrodes as
already mentioned. A simple immersion in 3% HF for 3 min
gives effective j-U curves such as shown in Figure 3.
Figure 3. Current-potential (j-U) curves for hydrogen evolution in
1.0 M HCl (pH 0.2) at 20 °C: (A) Cu-related electrodes, (B) Ag-related
electrodes, and (C) Au-related electrodes. Eeq(H+/H2) is the equilibrium
hydrogen-evolution potential at pH 0.2.
photoeffect and nearly agrees with that for an Au electrode,
indicating that the Au/p-Si contact is nearly ohmic. For the
continuous-Cu and Ag/p-Si, the dark currents have a tendency
of saturation at largely negative potentials, and the currents are
increased to some extent by illumination. These results indicate
that the Cu/p-Si and Ag/p-Si contacts form only low Schottky
barriers. The behavior of the continuous-Cu, Ag, and Au/p-Si
electrodes described here is in good agreement with reported
barrier heights for contacts between p-Si (or n-Si) and Cu, Ag,
and Au.23-25
3. CO2 Photoreduction on Particulate-Metal/p-Si. Figure
4 summarizes the j-U curves for the CO2 reduction on the
particulate-metal/p-Si and other related electrodes in CO2-
saturated aqueous 0.1 M KHCO3 (pH 6.8). It is reported that
the solubility of CO2 in aqueous solution is 38 mM at 25 °C
The p-Si electrodes coated with vacuum-deposited Cu, Ag,
and Au layers (denoted as continuous-metal/p-Si in Figure 3)
show cathodic currents even in the dark, and the onset potentials
are similar to those for the corresponding metal electrodes, quite
contrary to the case of the particulate-metal/p-Si electrodes. In
particular, the j-U curve for the continuous-Au/p-Si shows no

Photoelectrochemical Reduction of CO2
J. Phys. Chem. B, Vol. 102, No. 6, 1998 977
Figure 5. Current efficiencies vs potential for various CO2-reduction
products after potentiostatic electrolyses on particulate-Cu/p-Si and Cu-
metal electrodes at 20 °C.
To get accurate photovoltages for the particulate-metal/p-Si
electrodes, detailed studies have been made for the Cu-related
electrodes. Figure 4A includes the j-U curve for a Cu
(99.994%) electrode on which a Cu layer was electrodeposited
in the same way as for particulate-Cu/p-Si (denoted by ED-Cu
in Figure 4A) and that for a Cu electrode on which a Cu layer
was vacuum-deposited in the same way as for continuous-Cu/
p-Si (denoted by VD-Cu), together with the curve for a highly
pure (99.994%) Cu electrode whose surface was electrochemi-
cally polished in 85.0% H3PO4 (denoted by EP-Cu). The
difference in the j-U curves among ED-Cu, VD-Cu, and EP-
Cu suggests that the catalytic activity of Cu is sensitive to the
surface structure and impurities.26-28 The accurate photovoltage
for the particulate-Cu/p-Si is thus obtained by comparing the
j-U curve for this electrode with that for the ED-Cu electrode.
From Figure 4A, we can conclude that the photovoltage is about
0.50 V.
Figure 5 shows the results of product analyses for the CO2
reduction on the particulate-Cu/p-Si and Cu (99.999%) elec-
trodes in CO2-saturated 0.1 M KHCO3. The current efficiency
for each reduction product is plotted as a function of the
electrode potential at which the electrolysis was performed. The
iR drop between the working and reference electrodes is
corrected in Figure 5, similar to the j-U curves in Figure 4,
though no correction was made in our previously reported
results.17 The results for the Cu-metal electrode agree well with
those reported by Hori et al.26 and Ito et al.16 A parallel shift
of the current efficiency vs potential curve for each product
between the particulate-Cu/p-Si and Cu electrodes by ca. 0.50
V indicates that the Cu particles on p-Si act as an effective
catalyst for the CO2 reduction (cf. Figure 1) with the generation
of a high photovoltage of ca. 0.50 V in harmony with the
photoshift in the j-U curves of Figure 4A.
Table 1 shows the current efficiencies of reduction products
for the particulate-Ag and Au/p-Si electrodes compared with
those for Ag, Au, and naked p-Si electrodes. A naked p-Si
electrode gave mainly H2 together with a small amount of CO
and HCOOH. The particulate-Ag and Au/p-Si electrodes
produced mainly CO, similar to the Ag and Au electrodes,15,16
but at more positive potentials. Somewhat smaller current
efficiencies of CO for the particulate-Ag and Au/p-Si than those
for Ag and Au, together with increased current efficiencies of
H2 on the particulate-Ag and Au/p-Si, may suggest that the
reduction reactions on the (HF- etched) particulate-Ag and Au/
p-Si proceed partly on the naked p-Si surface. The total current
efficiency for all the reduction products at each potential does
Figure 4. Current-potential (j-U) curves for CO2 reduction in CO2-
saturated aqueous 0.1 M KHCO3 (pH 6.8) at 20 °C: (A) Cu-related
electrodes, (B) Ag-related electrodes, and (C) Au-related electrodes.
Eeq(CO2/CH4), Eeq(CO2/C2H4), and Eeq(CO2/CO) are the equilibrium
potentials for the CO2 reduction to CH4, C2H4, and CO at pH 6.8,
respectively.
and the CO2 reduction current cannot exceed ∼10 mA/cm2 due
to the diffusion limitation.3 Therefore, only the curves in the
region of j less than 10 mA/cm2 are shown in Figure 4. All
the j-U curves in Figure 4 are corrected for the iR drop between
the working and reference electrodes. The j-U curves for the
particulate-Cu/p-Si and other Cu-related electrodes, reported in
our previous letter,17 were not corrected for the iR drop and are
largely different from the present curves in Figure 4A. A part
of the results for particulate-Au/p-Si was reported in a short
paper.18
Similar to the case of hydrogen evolution (Figure 3), the
photocurrent onsets for the particulate-metal (Cu, Ag, and Au)/
p-Si electrodes (after the HF etching for the particulate-Ag and
Au/p-Si) are much more positive than those for the correspond-
ing metal electrodes. Also the photocurrent onsets for the
particulate-Cu and Au/p-Si lie more positive than the equilibrium
potentials for the CO2 reduction to C2H4 and CO at pH 6.8,
respectively (Figure 4A,C). On the other hand, the j-U curves
for the continuous-metal (Cu, Ag, and Au)/p-Si show almost
no photoeffect and the current onsets lie in the same potential
region as those for the corresponding metal electrodes.

978 J. Phys. Chem. B, Vol. 102, No. 6, 1998
Hinogami et al.
TABLE 1: Current Efficiencies of Various Reduction Products for Particulate-Ag and Au/p-Si Compared with Those for Ag,
Au, and Naked p-Si
current efficiency [%]
HCOOH
potential
[V vs SCE]
ja
Qb
electrode
naked p-Si
Ag
particulate-Ag/p-Si
Au
particulate-Au/p-Si
H2
CO
CH4
C2H4
[mA/cm2]
[C/cm2]
-1.27
-1.49
-1.05
-1.21
-0.74
73.5
14.1
38.4
3.7
12.0
75.9
50.9
82.2
62.2
4.3
2.1
0.8
0.24
0.0
0.44
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
2.40
1.69
2.29
2.13
1.87
1.2
10.0
10.0
1.6
9.0
10.0
a Average current density during the electrolysis. b Electricity which passed during the electrolysis.
not reach unity. This is most probably because the electrolysis
experiments were performed in a sealed cell for long periods
of time of 1 h and thus parts of the reduction products escaped
from the cell (cathode compartment) through the O-ring seals,
the cation exchange membrane, etc. However, the essential
features of the results shown in Table 1 and Figure 5 were well-
reproduced by several-time experiments. Preelectrolysis of the
electrolyte did not affect the current efficiencies.
Discussion
As described in the preceding section, a p-Si electrode
modified with small metal (Cu, Ag, or Au) particles generates
a high photovoltage of ca. 0.5 V and works as an effective
electrode for the PEC reduction of CO2, though a p-Si electrode
coated with a continuous metal layer shows almost no photo-
effect. Such a sharp contrast clearly shows the prominent
effectiveness of particulate-metal coating. Similar sharp con-
trasts between particulate- and continuous-metal coating were
obtained for Pt coated n-Si,9-12 p-Si,13 and p-InP,14 independent
of methods of metal deposition and quantities of deposited metal
down to a sub-monatomic layer amount,9 indicating the general-
ity of the result. The mechanism of the efficient CO2 photore-
duction on particulate-metal/p-Si can be explained on the basis
of our previously proposed theoretical model for metal-particle-
coated n-Si-based solar cells.9-12
Figure 6. Schematic two-dimensional energy band diagram for a
particulate-metal/p-Si electrode under a slight reverse (negative) bias
under illumination. ECS(metal) and EVS(metal) are the bottom of the
conduction band and the top of the valence band at the p-Si surface in
the area of the direct metal/Si contact, respectively, and ECS(soln) and
EVS(soln) are those in the area of naked p-Si surface, respectively. EF-
(metal) and EF(p-Si) are the Fermi level of metal particles and p-Si,
S
respectively. eφBm ) EF(metal) - EVS(metal), eφBe ) EF(metal) - EV
-
(soln), and eφB′ (effective barrier height) is the energy difference
between EF(metal) and the saddle point of the modulated valence band
(point A in the figure).
Let us first consider a model electrode such as shown in
Figure 1A, the difference from actual electrodes (Figure 2) being
discussed later. Metal is deposited sparsely in the form of small
particles on p-Si. The ideal size of the areas of direct Si/metal
contacts and their separation are estimated theoretically to be
about 5 and 20 nm, respectively.9,10 Photogenerated holes in
p-Si enter into the silicon electrode, whereas photogenerated
electrons come out to the surface due to the presence of band
bending near the p-Si surface. The electrons then transfer into
the metal particles and reduce CO2 at the surface with an aid
of catalytic activity of the metal particles. The HF-etched naked
p-Si surface is covered with Si-H and Si-OH bonds29 and
passivated.30 Because the metal particles are small and sparsely
distributed, the major part of the p-Si surface is thus kept nearly
free from surface states (recombination centers). Moreover, the
electrode reactions occur mainly on the metal particles, not on
the Si surface, and thus no surface reaction intermediates
(surface states and recombination centers) are formed at the Si
surface during the electrochemical reactions. Thus, the present
type electrode can have high catalytic activity for the electrode
reactions and a low surface recombination rate.
to a region of the same size as the area of the direct metal/Si
contact.9,10 Thus, the effect of the modulation on the barrier
height is negligible if the area of the direct metal/Si contact is
much smaller than the width of the space charge layer of p-Si.
The effective barrier height (eφB′ in Figure 6) is then nearly
e
equal to the barrier height for naked p-Si (eφB ) and much higher
than the intrinsic barrier height at the direct p-Si/metal contact
(eφBm). On the other hand, the barrier height for continuous-
metal/p-Si is given by eφBm. A semiconductor electrode coated
with homogeneously distributed metal atoms of a sub-mon-
atomic layer amount behaves similar to the continuous-metal-
coated electrode.9
For the CO2 photoreduction on the particulate-metal/p-Si
electrodes, there remains another notable point. To make it
clearer, Figure 7 compares the j-U curves for hydrogen
evolution and CO2 reduction on the particulate-metal/p-Si and
metal electrodes. For the hydrogen evolution (Figure 7A), the
onset potentials of photocurrents for particulate-Cu, Ag, and
Au/p-Si (denoted as Cu, Ag, and Au/p-Si in Figure 7 for
simplicity) agree with each other, though the onset potentials
for the Cu, Ag, and Au metal electrodes are different from each
other. On the contrary, for the CO2 reduction in Figure 7B,
the onset potentials of photocurrents for particulate-Cu, Ag, and
Au/p-Si are different from each other, similar to those for the
corresponding metal electrodes, yielding a nearly constant
Another important point for the present type electrode is that
a high (effective) energy barrier is formed at the p-Si/solution
contact, irrespective of the intrinsic barrier height for the metal/
p-Si contact. The surface band energies of p-Si are more or
less modulated by the deposited metal particles and the
modulation extends toward the interior of p-Si, as schematically
shown in Figure 6, but the modulation is limited spatially only

Photoelectrochemical Reduction of CO2
J. Phys. Chem. B, Vol. 102, No. 6, 1998 979
Figure 7. Current-potential (j-U) curves of particulate-metal/p-Si
(denoted as Cu, Ag, and Au/p-Si in the figures) and metal electrodes;
(A) for H2 evolution in 1.0 M HCl (pH 0.2) and (B) for CO2 reduction
in CO2-saturated 0.1 M KHCO3 (pH 6.8).
Figure 8. Energy band diagrams explaining the photoeffect of
particulate-metal/p-Si: (A) for hydrogen evolution in 1.0 M HCl (pH
0.2) and (B) for CO2 reduction in CO2-saturated 0.1 M KHCO3 (pH
6.8). Horizontal lines in solution phase, denoted by H2(Au), CO2(Au),
etc., represent the observed onset potentials of the H2 evolution and
CO2 reduction on the metal electrodes indicated in the parentheses.
Eeq(H+/H2) is the equilibrium potential for hydrogen evolution at pH
0.2 and Eeq(CO2/CO) and Eeq(CO2/CH4) are those for the CO2 reduction
to CO and CH4 at pH 6.8, respectively. ECS(soln), EF(metal), and
EF(p-Si) have the same meaning as those in Figure 6. See text for details.
photovoltage of 0.5 V. Such an apparent contradiction between
the hydrogen evolution and CO2 reduction can be explained by
taking into account the difference in the reduction potentials
for the reactions.
The conduction and valence band edges at the naked p-Si
surface in 1.0 M HCl are estimated to be -0.78 V and +0.34
V vs SCE, from the flat-band potential of n-Si (-0.62 V)13,31,32
in 1.0 M HCl together with the band gap of Si (1.12 eV) and
the energy difference (0.16 eV) between the Fermi level and
the bottom of the conduction band for n-Si (ND ) 6.4 × 1014
cm-3). Thus, the onset potentials for the H2 evolution on the
Cu, Ag, and Au metal electrodes (-0.60, -0.57, and -0.35 V,
shift at a ratio of 0.059 V/pH but shifts at 0.023 V/pH in the
pH range below 731,32), though the equilibrium potentials for
the CO2 reduction to CO, CH4, and C2H4 are -0.75, -0.48,
and -0.56 V vs SCE at pH 6.8, respectively, and below the
conduction band edge of the naked p-Si (Figure 8B). Thus,
the CO2 photoreduction on the particulate-metal/p-Si can occur
only when the surface band energies of p-Si as well as the Fermi
level of the metal particles are shifted upward to some extent
as shown in Figure 8B. Such an upward shift is achieved by
accumulation of electrons at the p-Si surface as well as in the
metal particles, accompanied by changes in potential drops in
the Helmholtz layers at the p-Si and metal-particle surfaces.
The accumulation of electrons continues until the conduction
band edge of the p-Si surface and the Fermi level of the metal
particles reach the onset potentials of the CO2 reduction on the
metal particles. Therefore, the onset potentials of photocurrents
for the CO2 reduction on the particulate-metal/p-Si depend on
the kind of the deposited metals. This argument implies that
the present type semiconductor electrode has an ability to shift
its surface band energies so as to get good energy-level matching
between the semiconductor and solution reactants, similar to
naked semiconductor electrodes. This is in sharp contrast to
respectively, as taken from the potentials of j ) 1.0 mAcm-2
)
are all below the conduction band edge of the naked p-Si
surface, as shown in Figure 8A. This implies that photogener-
ated electrons can move to the metal particles and cause
hydrogen evolution without any shift of the p-Si band edges
for all the particulate-metal/p-Si electrodes. In other words, the
onset potentials of hydrogen-evolution photocurrent for the
particulate-metal/p-Si are solely determined by the flatband
potential of the naked p-Si, independent of the difference in
the onset potentials for hydrogen evolution on the deposited
metals, and agree with each other as really observed (Figure
7A).
On the contrary, the onset potentials of the CO2 reduction
on the Cu, Ag, and Au electrodes (-1.30, -1.45, and -1.15
V, respectively) are far above the conduction band edge of the
naked p-Si surface in CO2-saturated 0.1 M KHCO3 of pH 6.8
(ca. -0.93 V vs SCE, the flat-band potential of n-Si does not

980 J. Phys. Chem. B, Vol. 102, No. 6, 1998
Hinogami et al.
behavior of molecular (and particulate semiconductor) photo-
catalysts.33,34
Acknowledgment. The present work was partly supported
by a Grant-in-Aid for Scientific Research on Priority Areas of
the Ministry of Education, Science, Sports and Culture (Grant
09237105).
From the arguments made so far, we can conclude that the
electrode of the present type (Figure 1) can in principle meet
all the requirements for high efficiency, mentioned in the
Introduction section and become an ideal type electrode for the
CO2 photoreduction. Now let us consider the surface structure
of actual electrodes as compared with the model electrode of
Figure 1A. The size of deposited metal particles seen in the
SEMs (20-200 nm, Figure 2) is considerably larger than the
theoretically expected optimum value (ca. 5 nm)9,10 and the
particle density is high in the case of Au (Figure 2C),
nevertheless a high photovoltage of 0.5 V is generated. This
is most probably due to the fact that the areas of direct metal/
Si contacts are much smaller than the particle size seen in the
SEMs, the electrolyte partly penetrating into the metal-particle/
Si interface, such as shown in Figure 1B. This expectation is
supported by the aforementioned fact that the increase in the
electricity of metal particle deposition only increased the particle
size, not affecting the particle density. This fact implies that
the metal deposition occurs only on metal nuclei formed on
certain active site of the p-Si surface at the initial stage of the
electrodeposition, and thus the metal particles are in direct
contact with p-Si only in the small areas of the initial metal
nuclei. The small areas of the direct Si/metal contact may also
be caused by geometrically rough structure of concentrated-
HF-etched p-Si surfaces on a few nanometer scale as revealed
by recent STM studies.35 Thus we can say that the actual
electrodes (Figure 1B) have essentially the same metal-contact
structure as the model electrode of Figure 1A.
References and Notes
(1) Yoneyama, H.; Sugimura, K.; Kuwabata, S. J. Electroanal. Chem.
1988, 249, 143.
(2) Bockris, J. O.; Wass, J. C. J. Electrochem. Soc. 1989, 136, 2521.
(3) Taniguchi, I. In Modern Aspects of Electrochemistry; Bockris, J.
O., White, R. E., Conway, B. E., Eds.; Plenum: New York, 1989; Vol. 20,
Chapter 5.
(4) Noda, H.; Ikeda, S.; Saito, Y.; Nakamura, T.; Maeda, M.; Ito, K.
Denki Kagaku 1989, 57, 1117.
(5) Ikeda, S.; Saito, Y.; Yoshida, M.; Noda, H.; Maeda, M.; Ito, K. J.
Electroanal. Chem. 1989, 260, 335.
(6) Flaisher, H.; Tenne, R.; Halmann, M. J. Electroanal. Chem. 1996,
402, 97.
(7) Junfu, L.; Baozhu, C. J. Electroanal. Chem. 1992, 324, 191.
(8) Hashimoto, K.; Fujishima, A. Spec. Publ. R. Soc. Chem. 1994, 153,
388 (Carbon Dioxide Chemistry: EnVironmental Issues).
(9) Nakato, Y.; Ueda, K.; Yano, H.; Tsubomura, H. J. Phys. Chem.
1988, 92, 2316.
(10) Nakato, Y.; Tsubomura, H. Electrochim. Acta 1992, 37, 897.
(11) Yae, S.; Nakanishi, I.; Nakato, Y.; Toshima, N.; Mori, H. J.
Electrochem. Soc. 1994, 141, 3077.
(12) Jia, J.-G.; Fujitani, M.; Yae, S.; Nakato, Y. Electrochim. Acta 1996,
42, 431.
(13) Nakato, Y.; Yano, H.; Nishiura, S.; Ueda, T.; Tsubomura, H. J.
Electroanal. Chem. 1987, 228, 97.
(14) Kobayashi, H.; Mizuno, F.; Nakato, Y.; Tsubomura, H. J. Phys.
Chem. 1991, 95, 819.
(15) Hori, Y.; Kikuchi, K.; Suzuki, S. Chem. Lett. 1985, 1695.
(16) Noda, H.; Ikeda, S.; Oda, Y.; Imai, K.; Maeda, M.; Ito, K. Bull.
Chem. Soc. Jpn. 1990, 63, 2459.
(17) Hinogami, R.; Mori, T.; Yae, S.; Nakato, Y. Chem. Lett. 1994,
1725.
The photovoltage of 0.5 V obtained in the present work is a
little smaller than the open-circuit photovoltages of 0.62 to 0.64
V obtained for metal-particle-coated n-Si-based PEC solar
cells.9-12 A possible reason may lie in that the areas of direct
metal/Si contacts in the present work are still too large for some
metal particles, leading to a decrease in the photovoltage.
Another possible reason is that H+ ions penetrate into the interior
of p-Si, resulting in interstitial H atoms,36 which act as a
recombination center. These discussions indicate that the control
of surface structure not only on an atomic scale but also on a
nanometer-sized level is important for getting a higher ef-
ficiency.
(18) Hinogami, R.; Nakamura, Y.; Yae, S.; Nakato, Y. Appl. Surf. Sci.
1997, 121/122, 301.
(19) Kobayashi, H.; Ishida, T.; Nakato, Y.; Tsubomura, H. J. Appl. Phys.
1991, 69, 1736.
(20) Morinaga, H.; Futatsuki, T.; Ohmi, T.; Fuchita, E.; Oda, M.;
Hayashi, C. J. Electrochem. Soc. 1995, 142, 966.
(21) Morinaga, H.; Suyama, M.; Ohmi, T. J. Electrochem. Soc. 1994,
141, 2834.
(22) Chyan, O. M. R.; Chen, J.-J.; Chien, H. Y.; Sees, J.; Hall, L. J.
Electrochem. Soc. 1996, 143, 92.
(23) Sze, S. M. Physics of Semiconductor DeVices, 2nd ed.; John Wiley
and Sons: New York, 1981; p 291.
(24) Chen, C.; Frese, K. W., Jr. J. Electrochem. Soc. 1992, 139, 3243.
(25) Chen, C.; Frese, K. W., Jr. J. Electrochem. Soc. 1993, 140, 1355.
(26) Hori, Y.; Murata, A.; Takahashi, R. J. Chem. Soc., Faraday Trans.
1 1989, 85, 2309.
The Cu, Ag, and Au metals used in the present work are the
most active electrocatalysts for the CO2 reduction for the present,
but the onset potentials of the CO2-reduction are still consider-
ably more negative than the equilibrium potentials and hence
above the conduction band edge of p-Si (Figure 8B). For this
reason, however, the Fermi level of metal particles can coincide
with the conduction band edge in a stationary state under
illumination (Figure 8B), leading to the formation of the highest
energy barrier height. Though it does not seem to be easy to
find more active metallic electrocatalysts for the CO2 reduction,
if they are found, they will shift the onset potentials of the CO2
reduction close to the equilibrium potentials and below the
conduction band edge of p-Si. For such a case, the high energy
barrier can be obtained either by a downward shift of the
conduction band edge of p-Si (e.g., through the formation of
Si-F termination bonds37) or by using another semiconductor
having an appropriate conduction band edge.
(27) Kyriacou, G.; Anagnostopoulos, A. J. Electroanal. Chem. 1992,
322, 233.
(28) Hori, Y.; Wakebe, H.; Tsukamoto, T.; Koga, O. Surf. Sci. 1995,
335, 258.
(29) Takahagi, T.; Ishitani, A.; Kuroda, H. J. Appl. Phys. 1991, 69, 803.
(30) Yablonovitch, E.; Allara, D. L.; Chang, C. C.; Gmitter, T.; Bright,
T. B. Phys. ReV. Lett. 1986, 57, 249.
(31) Nakato, Y.; Ueda, T.; Egi, Y.; Tsubomura, H. J. Electrochem. Soc.
1987, 134, 353.
(32) Madou, M. J.; Loo, B. H.; Frese, K. W.; Morrison, S. R. Surf. Sci.
1981, 108, 135.
(33) Ishida, H.; Terada, T.; Tanaka, K.; Tanaka, T. Inorg. Chem. 1990,
29, 905.
(34) Kanemoto, M.; Shiragami, T.; Pac, C.; Yanagida, S. J. Phys. Chem.
1992, 96, 3521.
(35) Itaya, K.; Sugawara, R.; Morita, Y.; Tokumoto, H. Appl. Phys. Lett.
1992, 60, 2534.
(36) Mierry, P. de; Etcheberry, A.; Aucouturier, M. J. Appl. Phys. 1991,
69, 1099.
(37) Fujitani, M.; Hinogami, R.; Jia, J.-G.; Ishida, M.; Morisawa, K.;
Yae, S.; Nakato, Y. Chem. Lett. 1997, 1041.