Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

1
1
Fundamentals of Semiconductor
Electrochemistry and
Photoelectrochemistry
Krishnan Rajeshwar
The University of Texas at Arlington, Arlington, Texas
1.1 Introduction and Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.2 Electron Energy Levels in Semiconductors and Energy Band Model . 4
1.3 The Semiconductor–Electrolyte Interface at Equilibrium . . . . . . . . 8
1.3.1 The Equilibration Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3.2 The Depletion Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3.3 Mapping of the Semiconductor Band-edge Positions Relative to
Solution Redox Levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3.4 Surface States and Other Complications . . . . . . . . . . . . . . . . . . . 15
1.4 Charge Transfer Processes in the Dark . . . . . . . . . . . . . . . . . . . . 16
1.4.1 Current-potential Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.4.2 Dark Processes Mediated by Surface States or by Space Charge Layer
Recombination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.4.3 Rate-limiting Steps in Charge Transfer Processes in the Dark . . . . . 23
1.5 Light Absorption by the Semiconductor Electrode and Carrier
Collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.5.1 Light Absorption and Carrier Generation . . . . . . . . . . . . . . . . . . . 25
1.5.2 Carrier Collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.5.3 Photocurrent-potential Behavior . . . . . . . . . . . . . . . . . . . . . . . . . 29
1.5.4 Dynamics of Photoinduced Charge Transfer . . . . . . . . . . . . . . . . . 33
1.5.5 Hot Carrier Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
1.6 Multielectron Photoprocesses . . . . . . . . . . . . . . . . . . . . . . . . . . 34
1.7 Nanocrystalline Semiconductor Films and Size Quantization . . . . . 36
1.7.1 Introductory Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
1.7.2 The Nanocrystalline Film–Electrolyte Interface and Charge Storage
Behavior in the Dark . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
1.7.3 Photoexcitation and Carrier Collection: Steady State Behavior . . . . . 38
1.7.4 Photoexcitation and Carrier Collection: Dynamic Behavior . . . . . . . 40
1.7.5 Size Quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry
1.8 Chemically Modified Semiconductor–Electrolyte Interfaces . . . . . . 41

1.8.1 Single Crystals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
1.8.2 Nanocrystalline Semiconductor Films and Composites . . . . . . . . . 43
1.9 Types of Photoelectrochemical Devices . . . . . . . . . . . . . . . . . . . . 44
1.10 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3
1.1 and the ensuing energy crisis caused

Introduction and Scope a dramatic spurt in studies on semi-
conductor–electrolyte interfaces once the
The study of semiconductor–electrolyte energy conversion possibilities of the lat-
interfaces has both fundamental and prac- ter were realized. Subsequent progress
tical incentives. These interfaces have in- at both fundamental and applied lev-
teresting similarities and differences with els in the late eighties and nineties has
their semiconductor–metal (or metal ox- been more gradual and sustained. Much
ide) and metal–electrolyte counterparts. of this later research has been spurred
Thus, approaches to garnering a funda- by technological applicability in envi-
mental understanding of these interfaces ronmental remediation scenarios. Very
have stemmed from both electrochem- recently, however, renewed interest in
istry and solid-state physics perspectives clean energy sources that are nonfos-
and have proven to be equally fruit- sil in origin, has provided new impetus
ful. On the other hand, this knowl- to the study of semiconductor–electrolyte
edge base in turn impacts many tech- interfaces. As we also learn to un-
nologies, including microelectronics, en- derstand and manipulate these inter-
vironmental remediation, sensors, so- faces at an increasingly finer (atomic)
lar cells, and energy storage. Some of level, new microelectronics application
these are discussed elsewhere in this possibilities may emerge, thus complet-
volume. ing the cycle that first began in the
It is instructive to first examine the 1950s.
historical evolution of this field. Early The ensuing discussion of the progress
work in the fifties and sixties undoubt- that has been made in this field mainly
edly was motivated by application pos- hinges on studies that have appeared since
sibilities in electronics and came on about 1990. Several review articles and
the heels of discovery of the first tran- chapters have appeared since then that
sistor. Electron transfer theories were deal with semiconductor–electrolyte in-
also rapidly evolving during this pe- terfaces [1–10]; aspects related to electron
riod, starting from homogeneous systems transfer are featured in several of these.
to heterogeneous metal-electrolyte inter- This author is also aware of at least three
faces leading, in turn, to semiconductor- books/monographs/proceedings volumes
electrolyte junctions. The 1973 oil embargo that have appeared since 1990 [11–13]. The
reader is referred to the many authorita- electrons (e.g. Al with [Ne]3s2 3p1 ) will have
tive accounts that exist before this time a partially occupied frontier band in which
frame for a thorough coverage of details the electrons are delocalized. On the other
on semiconductor–electrolyte interfaces hand, a solid with an even number of va-
in general. Entry to this early literature lence electrons (e.g. Si having an electron
may be found in the references cited configuration of [Ne]3s2 2p2 ) will have a
earlier. In some instances, however, the fully occupied frontier band (termed a va-
discussion that follows necessarily delves lence band, (VB)). The situation for Si is
into research dating back to the 1970s and schematized in Fig. 1.
1980s. As Fig. 2 illustrates, the distinction
To facilitate a self-contained descrip- between semiconductors and insulators
tion, we will start with well-established is rather arbitrary and resides with the
aspects related to the semiconductor en- magnitude of the energy band gap (Eg )
ergy band model and the electrostat- between the filled and vacant bands.
ics at semiconductor–electrolyte inter- Semiconductors typically have Eg in the
faces in the ‘‘dark’’. We shall then 1 eV–4 eV range (Table 1). The vacant
examine the processes of light ab- frontier band is termed a conduction band,
sorption, electron-hole generation, and (CB) (Fig. 2). We shall see later that Eg
charge separation at these interfaces. has an important bearing on the optical
The steady state and dynamic aspects response of a semiconductor.
of charge transfer are then briefly con- For high density electron ensembles
sidered. Nanocrystalline semiconductor such as valence electrons in metals, Fermi
films and size quantization are then statistics is applicable. In a thermodynamic
discussed as are issues related to elec- sense, the Fermi level, EF (defined at 0 K
tron transfer across chemically modi-
fied semiconductor–electrolyte interfaces.
Finally, we shall introduce the vari- Tab. 1 Some elemental and compound
ous types of photoelectrochemical devices semiconductors for photoelectrochemical
applications
ranging from regenerative and photo-
electrolysis cells to dye-sensitized solar
Semi- Conductivity Optical
cells. conductor type(s) band gap
energy [eV]a
1.2
Si n, p 1.11
Electron Energy Levels in Semiconductors
GaAs n, p 1.42
and Energy Band Model GaP n, p 2.26
InP n, p 1.35
Unlike in molecular systems, semiconduc- CdS n 2.42
tor energy levels are so dense that they CdSe n 1.70
form, instead of discrete molecular or- CdTe n, p 1.50
bital energy levels, broad energy bands. TiO2 n 3.00(rutile)
3.2(anatase)
Consider a solid composed of N atoms. ZnO n 3.35
Its frontier band will have 2N energy
eigenstates, each with an occupancy of a The values quoted are for the bulk
two electrons of paired spin. Thus, a solid semiconductor. The gap energies increase in the
having atoms with odd number of valence size quantization regime (see Sect. 7).
1.2 Electron Energy Levels in Semiconductors and Energy Band Model 5
4 N states
0 electrons
6 N electrons
2 N states
3p
Electron energy Eg
4 N states
4 N electrons 3s
2 N states
2 N electrons
ro
Distance
Fig. 1Schematic illustration of how energy bands in
semiconductors evolve from discrete atomic states for the specific
example of silicon.
Fig. 2 Relative disposition of

the CB and VB for a semi-
conductor (a) and an insu- CB
lator (b). Eg is the optical band
gap energy.
CB
Eg
Eg
VB VB
(a) (b)
as the energy at which the probability and those above EF are likely to be
of finding an electron is 1/2) can be empty.
regarded as the electrochemical potential Electrons in semiconductors may be
of the electron in a particular phase (in regarded as low-density particle ensembles
this case, a solid). Thus, all electronic such that their occupancy in the valence
energy levels below EF are occupied and CBs may be approximated by the
Boltzmann function [14, 15]: To provide a numerical sense of the

! " situation, Nc and Nv are typically both ap-
Eo − EF
ne ≈ No exp − (1) proximately 1019 cm−3 so that the constant
kT No (Nc Nv ) in Eq. (1) is about 1038 cm−3 .
Now we come to another impor- For a semiconductor such as Si (with
tant distinction between metals and Eg = 1.11 eV, Table 1), ni will be about
semiconductors in that two types of elec- 1010 cm−3 at 300 K according to Eq. (4).
tronic carriers are possible in the latter. This rough calculation lends credence to
Consider the thermal excitation of an elec- the original rationale for the use of Boltz-
tron from VB to CB. This gives rise to a mann statistics for the electron energy
free electron in the CB and a vacancy or distribution in semiconductors (see pre-
hole in the VB. A localized chemical pic- ceding section).
ture for the case of Si shows that the hole The preceding case refers to the semi-
may be regarded as a missing electron in conductor in its intrinsic state with very
a chemical bond (Fig. 3). There is a crude low carrier concentrations under ambient
chemical analogy here with the dissocia- conditions. The Fermi level, EF , in this
tion of a solvent such as water into H3 O+ case lies approximately in the middle of
and OH− . In either case, equal numbers of the energy band gap (Fig. 4a). This sim-
oppositely charged species are produced. ply reflects the fact that the probability of
Thus, Eq. (1) becomes: electron occupancy is very high in VB and
! " very low in CB and does not imply an
EF − Ec occupiable energy level at EF itself.
ni ≈ Nc exp − (2)
kT In extrinsic semiconductors the carrier
! " concentrations are perturbed such that
Ev − EF
pi ≈ Nv exp − (3) n = p. Again the analogy with the addi-
kT
tion of an acid or base to water is quite
In Eqs. (2–3), Nc and Nv are the effective instructive here. Consider the case when
density of states (in cm−3 ) at the lower edge donor impurities are added to a neutral
and top edge of CB and VB, respectively. semiconductor. Since the intrinsic carrier
These expressions can be combined with concentrations are so low (sub-parts per
the recognition that ni = pi to yield trillion), even additions in parts per bil-
! " lion levels can have a profound electrical
Ec − Ev
n2i ≈ No exp − effect. This process is known as doping
kT of the semiconductor. In this particular
! "
Eg case, the Fermi level shifts toward the CB
≈ No exp − (4)
kT edge (Fig. 4b). When the donor level is
e−
(+) (−)
e− h+
Si Si Si
h+
Si Si Si Si Fig. 3 A localized picture of
electron-hole pair generation
(see also Fig. 2a) in silicon.
1.2 Electron Energy Levels in Semiconductors and Energy Band Model 7
Fig. 4 Relative disposition of the Fermi

level (EF ) for an intrinsic semiconductor CB
(a), for an n-type semiconductor (b),
and a p-type semiconductor (c). EF
within a few kT in energy from the CB VB

edge, appreciable electron concentrations
are generated by the donor ionization pro- (a)
cess (at ambient temperatures) such that
now n # p. This is termed n-type doping,
and the resultant (extrinsic) semiconduc-
EF
Energy
tor is termed n-type. By analogy, p-type
semiconductors have p # n. The terms
minority and majority carriers now be-
come appropriate in these cases. For a
p-doped semiconductor case, the Fermi (b)
level now lies close to the VB edge (Fig. 4c).
The movement of EF with dopant concen-
tration can also be rationalized via the
Nernst formalism [6].
Doping can be accomplished by adding
EF
altervalent impurities to the intrinsic
semiconductor. For example, P (a Group
15 or VA element) will act as a donor (c)
in Si (a Group 14 or IVA element). This
can be rationalized on chemical terms complex manner, a discussion of which is
by noting that P needs only four valence beyond the scope of this section. Excellent
electrons for tetrahedral bonding (as in the treatises are available on this topic and on
Si lattice) – the fifth electron is available the solid-state chemistry of semiconduc-
for donation by each P atom. The donor tors in general [16–22].
density, ND nominally is approximately The foregoing discussion strictly refers
1017 cm−3 . Thus, assuming that n $ ND to semiconductors in single-crystal form.
(complete ionization at 300 K), p will Amorphous and polycrystalline counter-
be only approximately 103 cm−3 [recall parts present other complications caused
that the product ni pi is ∼1020 cm−6 (see by the presence of defects, trap states,
preceding section)], bearing out the earlier grain boundaries, and the like. For this rea-
qualitative assertion that n # p. son we orient the subsequent discussion
Impurity addition, however, is not the mainly toward single crystals, although
only doping mechanism. Nonstoichiome- comparisons with less ideal cases are
try in compound semiconductors such as made where appropriate. The distinction
CdTe (Table 1) also gives rise to n- or p- between metal and semiconductor elec-
type behavior, depending on whether Cd or trodes is important when we consider
Te is in slight excess, respectively. The de- the electrostatics across the corresponding
fect chemistry in these solid chalcogenides solid–liquid interfaces; this distinction is
controls their conductivity and doping in a made in the following section.
1.3 phase. The situation before and after con-

The Semiconductor–Electrolyte Interface at tact of the two phases is illustrated in
Equilibrium Fig. 5(b) and (c) for an n-type and p-type
semiconductor, respectively. After contact,
1.3.1
the net result of equilibration is that
The Equilibration Process
EF = EF,redox and a ‘‘built-in’’ voltage, VSC
develops within the semiconductor phase,
The electrochemical potential of electrons
as illustrated in the right hand frames of
in a redox electrolyte is given by the Nernst
Fig. 5(b) and (c).
expression
! " It is instructive to further examine
o RT cox this equilibration process. Consider again
Eredox = E redox + ln (5)
nF cred an n-type semiconductor for illustrative
In Eq. (5), cox and cred are the concen- purposes (Fig. 5b). The electronic charge
trations (roughly activities) of the oxidized needed for Fermi level equilibration in the
and reduced species, respectively, in the semiconductor phase originates from the
redox couple. The parameter (Eredox = donor impurities (rather than from bond-
µe,redox ) as defined by Eq. (5) can be iden- ing electrons in the semiconductor lattice).
tified with the Fermi level (EF,redox ) in Thus, the depletion layer that arises as
the electrolyte. This was the topic of de- a consequence within the semiconduc-
bate some years back [23], although this tor contains positive charges from these
premise now appears to be well founded. ionized donors. The Fermi level in the
The task now is to relate the electron en- semiconductor (EF,n ) moves ‘‘down’’ and
ergy levels in the solid and liquid phases the process stops when the Fermi level is
on a common basis. the same on either side of the interface.
The semiconductor solid-state physics The rather substantial difference in the
community has adopted the electron density of states on either side dictates that
energy in vacuum as reference, whereas EF,n moves farther than the corresponding
electrochemists have traditionally used level, EF,redox in the electrolyte. A partic-
the standard hydrogen electrode (SHE) ularly lucid account of this initial charge
scale. While estimates vary [23–25], SHE transfer is contained in Ref. 6.
appears to lie at −4.5 eV with respect to the The band-bending phenomenon, shown
vacuum level. We are now in a position to in Fig. 5(b) and (c), is by no means
relate the redox potential Eredox (as defined unique to the semiconductor–electrolyte
with reference to SHE) with the Fermi interface. Analogous electrostatic adjust-
level EF,redox expressed versus the vacuum ments occur whenever two dissimilar
reference (Fig. 5a) phases are in contact (e.g. semiconductor-
gas, semiconductor–metal). An important
EF,redox = −4.5 eV − eo Eredox (6)
point of distinction from the correspond-
When a semiconductor is immersed in ing metal case now becomes apparent.
this redox electrolyte, the electrochemical For a metal, the charge, and thus the as-
potential (Fermi level) is disparate across sociated potential drop, is concentrated
the interface. Equilibration of this interface at the surface penetrating at most a
thus necessitates the flow of charge from few Å into the interior. Stated differ-
one phase to the other and a ‘‘band bend- ently, the high electronic conductivity of
ing’’ ensues within the semiconductor a metal cannot support an electric field
1.3 The Semiconductor–Electrolyte Interface at Equilibrium 9
Vacuum reference
χ ∅
CB ox
Energy
EF λ
Eg o
E redox
VB red
(a) Density of states
V SC E
E CB
EF E CB
E F, equil
E F, redox
E VB E VB
(b)
E CB E CB
E F, redox
E F, equil
EF
E VB V SC EVB
(c)
Fig. 5 (a) Energy levels in a semiconductor (left-hand side) and a
redox electrolyte (right-hand side) shown on a common vacuum
reference scale. χ and φ are the semiconductor electron affinity and
work function, respectively. (b) The semiconductor-electrolyte
interface before (LHS) and after (RHS) equilibration (i.e. contact of
the two phases) shown for a n-type semiconductor. (c) As in (b) but
for a p-type semiconductor.
within it. Thus, when a metal electrode rather variant behavior of a metal and a
comes into contact with an electrolyte, semiconductor at equilibrium with a redox
almost all the potential drop at the inter- electrolyte.
face occurs within the Helmholtz region
in the electrolyte phase. On the other 1.3.2
hand, the interfacial potential drop across The Depletion Layer
a semiconductor-electrolyte junction (see
following) is partitioned both as VSC and There is a characteristic region within the
as VH leading to a simple equivalent cir- semiconductor within which the charge
cuit model comprising two capacitors (CSC would have been removed by the equi-
and CH ) in series (Fig. 6). Further refine- libration process. Beyond this bound-
ments of the equivalent circuit description ary, the ionized donors (for a n-type
are given later but the point to note is the semiconductor), have their compensating
Distance coordinate
VG Gouy
region
VH
Semiconductor
V SC Electrolyte
Potential
Helmholtz
layer
C SC CH
Fig. 6 Electrostatics at a semiconductor–electrolyte interface.
A highly simplified equivalent circuit for the interface at
equilibrium is shown at the bottom. The Gouy layer is
neglected in the latter case (see text).
# $
charge (electrons), and the semiconductor eo ND
VSC = − W2 (8)
as a whole is electrically neutral. This 2εs
layer is the space charge region or the
depletion layer, so termed because the where W is the depletion layer width.
layer is depleted of the majority carriers. Further reflection shows how the mag-
The potential distribution in this interfa- nitude of W should depend on the semi-
cial region can be quantified by relating conductor parameter ND . Consider two
the charge density and the electric field cases of a semiconductor, one that is lightly
strength as embodied by the Poisson doped (say ND ∼ 1016 cm−3 ) and another
equation [14, 26]. Under restrictive con- that is heavily doped (ND ∼ 1018 cm−3 ).
ditions, more fully discussed in Refs. 6 Obviously in the former case, the charge
and 14, we obtain a particularly simple needed for Fermi level equilibration has
expression to come from deeper into the solid and
# $ so the magnitude of W will be larger.
eo ND This suggests a strategy for chemical
V (x) = − x2 (7)
2εs control of the electrostatics at the semicon-
× (o ≤ x ≤ W ) ductor–electrolyte interface [6]. Nominal
dimensions of W are in the 10–1000 nm
In Eq. (7), eo is the electronic charge and range. This may be compared with the
εs is the static dielectric constant of the corresponding Helmholtz layer width, typ-
semiconductor. The potential distribution ically 0.4–0.6 nm. With the capacitor-in-
is mapped in Fig. 6. We are now in a series model (see earlier section), we can
position to quantify the parameter VSC : see that the semiconductor space charge
layer is usually the determinant factor in In Eq. (10), Vt is the potential as measured
the total capacity of the interface. Once between an ohmic contact on the rear
again, the contrast with the correspond- surface of the semiconductor electrode
ing metal–electrolyte interface is striking. and the reference electrode (Fig. 6). The
Only when the semiconductor is degener- problematic factors in placing the semi-
ately doped (leading to rather large space conductor and solution energy levels on
charge layer charge, QSC and ‘‘thin’’ deple- a common basis involve VH and VG . In
tion layer widths) or when its surface is in other words, theoretical predictions of the
accumulation does the situation become magnitude of VSC (and how it changes as
akin to the metal–electrolyte interface (see the redox couple is varied) are hampered
following). by the lack of knowledge on the magnitude
of VH and VG . A degree of simplification
1.3.3 is afforded by employing relatively con-
Mapping of the Semiconductor Band-edge centrated electrolytes such that VG can be
Positions Relative to Solution Redox Levels ignored.
As with metals, the Helmholtz layer
Considerations of interfacial electron is developed by adsorption of ions or
transfer require knowledge of the relative molecules on the semiconductor surface,
positions of the participating energy levels by oriented dipoles, or especially in the
in the two (semiconductor and solution) case of oxides, by the formation of sur-
phases. Models for redox energy levels face bonds between the solid surface and
in solution have been exhaustively treated species in solution. Recourse to band
elsewhere [27, 28]. Besides the Fermi level edge placement can be sought through
of the redox system (Eq. 6), the thermal differential capacitance measurements on
fluctuation model [27, 28] leads to a Gaus- the semiconductor–redox electrolyte inter-
sian distribution of the energy levels for the face [29].
occupied (reduced species) and the empty In the simplest case as more fully dis-
(oxidized species) states, respectively, as cussed elsewhere [14, 15, 29], one obtains
illustrated in Fig. 5(a). The distribution the Mott-Schottky relation (for the spe-
functions for the states are given by cific instance of a n-type semiconductor) of
% & the semiconductor depletion layer capaci-
E − EF,redox − λ2 tance (CSC ), again by invoking the Poisson
Dox = exp − (9a)
4 kT λ equation
% & ! "
E − EF,redox + λ2 2 2 kT
Dred = exp − (9b) 1/CSC = (V − Vfb ) − (11)
4 kT λ ND eo εs eo
In Eqs. (9a) and (9b), λ is the solvent In Eq. (11), Vfb is the so-called flat band
reorganization energy. potential, that is the applied potential
Now consider the relative disposition of (V ) at which the semiconductor energy
these solution energy levels with respect to bands are ‘‘flat’’, leading up to the solution
the semiconductor band edge positions at junction. Several points with respect to the
the interface. The total potential difference applicability of Eq. (11) must be noted.
across this interface (Fig. 6) is given by The Mott-Schottky regime spans about
1 V in applied bias potential for most
Vt = VSC + VH + VG (10) semiconductor–electrolyte interfaces (i.e.
in the region of depletion layer formation process itself [31]. Figure 7 contains repre-
of the semiconductor space charge layer, sentative Mott-Schottky plots for both n-
see preceding section) [15]. The simple and p-type GaAs electrodes in an ambient
case considered here involves no mediator temperature molten salt electrolyte [32].
trap states or surface states at the inter- Once Vfb is known (from measure-
face such that the equivalent circuit of the ments), the Fermi level of the semicon-
interface essentially collapses to its most ductor at the surface is defined. It is
rudimentary form of CSC in series with then a simple matter to place the ener-
the bulk resistance of the semiconductor. gies corresponding to the conduction and
Further, in all the earlier discussions, it is VBs at the surface (ECB and EVB , respec-
reiterated that the redox electrolyte is suffi- tively) if the relevant doping levels are
ciently concentrated that the potential drop known. The difference between ECB and
across the Gouy (diffuse) layer in the solu- EVB should approximately correspond to
tion can be neglected. Specific adsorption the semiconductor band gap energy, Eg
and other processes at the semiconduc- (see Figs. 4 and 7). Alternatively, if Vfb is
tor–electrolyte interface will influence Vfb ; measured for one given state of doping
these are discussed elsewhere [29, 30] as of the semiconductor (n- or p-doped), the
are anomalies related to the measurement other band edge position can be fixed from
× 1013 × 1015
8.0 4.0
6.4 3.2
4.8 2.4
[farad−2]
[farad−2]
I C−2
I C−2
3.2 1.6
1.6 0.8
0 0
−0.5 −0.1 0.3 0.7 1.1 1.5
Potential
0/3 +
[V vs Al ]
Fig. 7 Mott-Schottky plots for n- and p-type GaAs electrodes in an AlCl3 /n-butylpyridinium chloride
molten-salt electrolyte. (Reproduced with permission from Ref. 32.)
knowledge of Eg . It is important to stress level of the redox system, EF,redox , on

that the semiconductor surface band edge the same diagram. Energy diagrams such
positions (as estimated from Vfb measure- as those in Fig. 8 are important in
ments) comprises all the terms in Eq. (10) considerations of charge transfer as we
and reflects the situation in situ for a shall see later. In anticipation of this
given set of conditions (solution pH, redox discussion, it is apparent that the three
concentration, etc.) of the semiconductor- situations illustrated in Fig. 9 for an n-type
redox electrolyte. The situation obviously semiconductor–electrolyte interface entail
becomes complex when the charge dis- the participation of the semiconductor CB,
tribution and mediation at the interface VB, and even states in its band gap in
changes either via surface states and illu- charge exchange with the solution species.
mination or both. These complications are Here again is a point of departure from
considered later. Figure 8 contains the rel- the metal case; viz., for a semiconductor,
ative dispositions of the surface band edges hole, electron, and surface state pathways
mapped for a number of semiconductors must all be considered.
in aqueous media. Let us return to the band bending
Having located the semiconductor band process at the interface. For a given
edge positions (relative either to the semiconductor, the expectation is that as
vacuum reference or a standard reference the redox Fermi level is moved more
electrode), we can also place the Fermi positive (‘‘down’’ on the energy diagram),
Vacuum scale
[eV] SHE
0 [V]
CdS
−3.0 −1.5
GaAs GaP
CdSe Si CdTe
−4.0 −0.5 1.4 eV TiO2 InP
0 2.4 eV
−4.5
2.3 eV 1.1 eV ∼1.3 eV 1.5 eV
−5.0 1.7 eV
Potential
Energy
+1.0
∼ 3.0 eV
−6.0
+2.0
−7.0
+3.0
−8.0
Fig. 8 Relative dispositions of various semiconductor band edge positions

shown both on the vacuum scale and with respect to the SHE reference.
These band edge positions are for an aqueous medium of pH ∼1.
CB E redox CB CB
EF EF EF
Energy
E SS o
E redox
o
E redox
VB VB VB
(a) (b) (c)
Fig. 9 Three situations for a n-type surface states (b), and with the semiconductor
semiconductor–electrolyte interface at VB (c). A discrete energy level is assumed for the
equilibrium showing overlap of the redox energy surface states as a first approximation.
levels with the semiconductor CB (a), with
VSC should increase concomitantly. This is photopotential (see following). Figure 10

the ideal (band edge ‘‘pinned’’) situation. shows that this ideal situation indeed
In other words [23] is realized for selected semiconduc-
tor–electrolyte interfaces [33]. As further
d%VSC
=1 (12) discussed later, the analogy with the corre-
dEredox
sponding metal-semiconductor junctions
Equation (12) reflects the fact that the (Schottky barriers) is direct [5, 34–36].
change in band bending faithfully tracks Complications arise when there are sur-
the redox potential change. A mea- face states that mediate charge exchange
sure of the former is the open-circuit at the interface. When their density is
1.2
I
0.8
Vph
[V]
slope = 1.0
0.4
Fig. 10 Plot of the open-circuit
photovoltage for amorphous
Si-methanol interfaces
0.0 containing a series of
−0.4 0.0 +0.4
one-electron redox couples.
Vredox (Reproduced with permission
[V vs. SCE] from Ref. 33.)
high [37], they act as a ‘‘buffer’’, in that versus Eredox (see Fig. 10). On the other
in the extreme case, carriers in the semi- hand, complete pinning results in a slope
conductor energy bands are completely of zero. Intermediate cases of Fermi level
excluded from the equilibration process. exhibit slopes between 0 and 1 [41]. As
stated earlier, there is a direct analogy
1.3.4 here with the metal/semiconductor junc-
Surface States and Other Complications tion counterparts [42, 43]:
Surface states arise because of the abrupt φB = Sφm + const (13)

termination of the crystal lattice at the In Eq. (13), φB is the so-called Schottky
surface; obviously the bonding arrange- barrier height, φm is the metal work func-
ment is different from that in the bulk. tion, and S is a dimensionless parameter.
Consider our prototype semiconductor, Si. Attempts have been made to relate S to
The tetrahedral bonding characteristic of semiconductor properties [44–48].
the bulk gives way to coordinative unsat- To further complicate matters, the
uration of the bonds for the Si surface nonideal behavior of semiconduc-
atoms. This unsaturation is relieved ei- tor–electrolyte interfaces as noted earlier
ther by surface reconstruction or bonds is exacerbated when the latter are irradi-
with extraneous (e.g. solvent) species [29, ated. Changes in the occupancy of these
38–40]. The surface bonding results in a states cause further changes in VH , so
localized electronic structure for the sur- that the semiconductor surface band edge
face that is different from the bulk. The positions are different in the dark and un-
energies of these localized surface orbitals der illumination. These complications are
nominally lie in the forbidden band gap considered later. The surface states as con-
region. The corresponding states are thus sidered earlier are shallow (with respect to
able (depending on their energy location) the band edge positions) and can essen-
to exchange charge with the conduction tially be considered as completely ionized
or VBs of the semiconductor and/or the at room temperature. However, for many
redox electrolyte [29]. oxide semiconductors, the trap states may
Unlike the case illustrated in Fig. 10, be deep and thus are only partially ionized.
changes in the solution redox potential Specifically, they may be disposed with
have been observed to cause no change respect to the semiconductor Fermi level
in the magnitude of VSC . This situa- such that they are ionized only to a depth
tion is termed Fermi level pinning; in that is small relative to W [49]. The phys-
other words, the band edge positions are ical manifestation of such deep traps as
unpinned in these cases so that the move- observed in the AC impedance behavior of
ment of Eredox is accommodated by VH semiconductor–electrolyte interfaces has
rather than by VSC . As mentioned earlier, been discussed [14, 49].
it appears [37] that surface state densities Finally, within the Mott-Schottky ap-
as low as 1013 cm−2 (∼1% of a mono- proximation (Eq. 11), large values of εs or
layer) suffice to induce complete Fermi ND can lead to the ratio VH /VSC becoming
level pinning in certain cases. Of course, significant. Figure 11 contains estimates
intermediate situations are possible. Thus, of this ratio for several values of ND for a
the ideal case manifests a slope of 1 in a semiconductor with a large εs value (TiO2 ,
plot of VSC (or an equivalent parameter) εs = 173) mapped as a function of the total
1.0
13
12
0.8
11
∆V H /∆VSC + ∆VH
10
0.6
9
0.4 8
7
6
0.2 5
3 4
2
1
0
0 0.2 0.4 0.6 0.8 1.0
∆V SC + ∆VH /V
Fig. 11 The ratio of the potential drop in the Helmholtz to the
total potential change computed as a function of the total
potential change. A static dielectric constant of 173 (typical of
TiO2 ) and a Helmholtz capacitance of 10 µF/cm2 were assumed
and the doping density was allowed to vary from 1016 cm−3
(curve 1) to 1020 cm−3 (curve 13). (Reproduced with permission
from Ref. 50.)
potential drop across the interface [50]. reverse (partial) currents exactly balance
Clearly VH can become a sizable fraction of each other and there is no net current
the total potential drop (approaching the flow across the interface. In fact, the
situation for metals) under certain con- situation here is similar to that occurring
ditions. It has been shown [51] that the for a metal–redox electrolyte interface at
Mott-Schottky plots will still be linear but the rest potential. We can write down
the intercept on the potential axis is shifted expressions for the net current using a
from the Vfb value. kinetics methodology as in Ref. 6 with
some minor changes in notation:
1.4 ic = −eo Aket cox (ns − nso ) (14)
Charge Transfer Processes in the Dark
1.4.1 In Eq. (14), ket is the rate constant for
Current-potential Behavior electron transfer, cox is the concentra-
tion of empty (acceptor) state in the redox
Let us return to the equilibrium situation electrolyte, ns and nso are the surface con-
of an n-type semiconductor in contact centrations of electrons, the subscript ‘‘o’’
with a redox electrolyte and reconsider in the latter case denoting the equilibrium
the situation in Fig. 9(a). This is shown situation. Thus, as long as the semiconduc-
again in Fig. 12(a) to underscore the fact tor–electrolyte interface is not perturbed
that the interface is in a state of dynamic by an external (bias) potential, ns ≡ nso
equilibrium. That is, the forward and and the net current is zero. The voltage
1.4 Charge Transfer Processes in the Dark 17
E E E
E CB E CB Vbias E CB
E F, redox Vbias E F, redox E F, redox
E VB E VB E VB
(a) (b) (c)

Fig. 12 Three situations for a n-type semiconductor–electrolyte interface at equilibrium (a),
under reverse bias (b), and under forward bias (c). The size of the arrows denotes the
magnitudes of the current in the two (i.e. anodic and cathodic) directions.
dependence of the current is embodied in similar to the metal case. The major dif-
the ratio, ns /nso , which can be regarded ference resides in the vastly different state
as a measure of the extent to which the densities in the solid and the existence of
interface is driven away from equilibrium. an energy gap region. The two nonequilib-
It must be noted that nso is not the bulk rium situations are shown in Figs. 12(b)
concentration of majority carriers (n) in and 12(c), respectively. Away from equilib-
the semiconductor because of the potential rium, we have the analogous Boltzmann
drop across the space charge layer [6, 35]. expression counterpart to Eq. (15)
# $ ! "
eo VSC eo (VSC + V )
nso = n exp − (15) ns = n exp − (16)
kT kT
A few words about the units of the terms leading, in turn, to
in Eq. (14) are in order at this juncture.
The term i/eo A may be regarded as a ic = −eo Aket cox nso
flux (J ) in units of number of carriers ! # $ "
eo V
crossing per unit area per second [1, 3, × exp − −1 (17)
kT
8]. The concentration terms are in cm−3 ;
thus ket has the dimensions of cm4 s−1 The assumption is inherent in the pre-
because of the second-order kinetics na- ceding discussion that all of the applied
ture stemming from the two multiplied bias (V ) drops across the space charge
concentration terms in Eq. (14) [1, 3, 8]. layer such that we are modulating only
Consider now the application of a bias the majority carrier population at the sur-
potential to the interface. Intuitively when face (and not the potential drop across the
it is such that ns > nso , a reduction cur- Helmholtz layer). In other words, the band
rent (cathodic current) should flow across edge positions are pinned or there is no
the interface such that the oxidized redox Fermi level pinning (see Sect. 1.3.4).
species are converted to reduced species Analogous expressions may be devel-
(Ox → Red). On the other hand, when oped for majority carrier flow for a p-type
nso > ns , the current flow direction is re- semiconductor in contact with a redox
versed and an anodic current should flow. electrolyte, with the important caveat that
Once again the situation here is somewhat the VB is involved in this process instead.
Equation (17) suggests that the cathodic readily rationalized. Finally, io will increase
current is exponentially dependent on po- with doping because of the ‘‘thinness’’ of
tential for V < 0. This is the so-called the resultant barrier at the surface.
forward-bias regime. On the other hand, When EF,redox is moved ‘‘down,’’ that is
when V > 0 (reverse-bias regime) the cur- more positive, the band bending increases,
rent is essentially independent of potential VSC increases and thus nso decreases. A
and, importantly, is of opposite sign. Sim- similar alteration in cox affects EF,redox
ply put, in this case, the electron flow through the Nernst expression. In both
direction (i.e. anodic) is from the occu- instances, we are influencing the Fermi
pied redox states into the semiconductor level at the interface at equilibrium. Thus,
CB (Fig. 12c). It should not, thus, be sur- in a sense, io is a quantitative measure
prising that this process is independent of of the extent of rectification of a given
potential. Both bias regimes are contained interface; that is, a smaller io value
in curve 1 in Fig. 13. translates to better rectification.
Of particular interest to this discussion The reverse bias current remains at a
is the ‘‘preexponential’’ term in Eq. (17): very low value because of the lack of
minority carriers (i.e. holes for n-type
io = eo Aket cox nso (18)
semiconductor) in the dark. Alternatively,
Analogous to the metal case, we can call injection of electrons from occupied redox
this term the exchange current; it is the levels (also an anodic current) has to ther-
current that flows at equilibrium when the mally surmount the surface barrier [5, 34,
partial cathodic and anodic components 35]. Under extreme reverse bias, however,
exactly balance one another. Of particular this barrier becomes ‘‘thin’’ and electrons
interest is the dependence of io on nso . can tunnel through it, leading to an abrupt
Also, variations in cox will affect the increase in the anodic current. This pro-
magnitude of io . Both these trends can be cess was studied even in the early days of
Light, I 2 (> I 1)
Light, I 1
Anodic
Current
0 Dark
Cathodic
Fig. 13 Current-potential curves for a

n-type semiconductor in the dark
(curve 1) and under band gap
illumination (curves 2 and 3). Two levels
of photon fluxes are shown in the
− Voltage + latter case.
semiconductor electrochemistry [52] and a such that the band bending is modest and
detailed discussion is found in a book chap- the holes recombine with electrons (either
ter [14]. Ultimately the junction ‘‘breaks via the surface states or in the space charge
down’’ (at the so-called Zener limit). This region itself), a cathodic current flows that
dark current flow is not shown in Fig. 13 is carried by the majority carriers in the
(curve 1). bulk. This recombination current pathway
Returning to the forward-bias (cathodic) is schematized in Fig. 14 and is further
current flow, Eq. (17) bears some analogy discussed in the next section. Hole injec-
to the famous Tafel expressions in electro- tion has been extensively studied especially
chemical kinetics. Thus, ignoring the unity for III–V (Group 13–15) semiconductors
term within the square brackets, Eq. (17) such as GaAs and GaP because of the
predicts a Tafel slope of 60 mV per decade relevance of this process to electroless etch-
at 298 K. In many instances [53, 54], such ing and device fabrication technology. This
a slope indeed is observed. In many cases, topic has been reviewed [61–64].
however, the slopes are higher than the The invokement of either the CB or the
‘‘ideal’’ value [14, 55–59]. VB of the semiconductor in charge ex-
The causes for this anomalous behavior change in the dark with solution redox
are still not fully understood. It appears species is not always straightforward. This
likely that many factors are involved: is particularly true for multielectron redox
surface film formation, varying potential processes to be discussed later. Movement
drop across the Helmholtz region caused, of the semiconductor band edge positions
for example, by surface state charging, and (i.e. band edge unpinning) relative to the
so on. Even crystallographic orientations redox energy levels also presents a com-
appear to be important [59]. These aspects plicating situation (see following). Some
have been discussed by other authors [14, cases (e.g. Eu2+/3+ in contact with GaAs
55, 60]. electrodes) are interesting in that the same
We have so far considered only (majority
carrier) processes involving the CB (again
assuming for illustrative purposes a n-type
E
semiconductor). Consider the interfacial
situation depicted in Fig. 9(c). The energy
Diffusion
states of the redox system now overlap with e− e− ECB
the VB of the semiconductor such that hole
injection in the ‘‘dark’’ is possible. When e−
e−
the band bending is large, the injected
holes remain at the surface and attack the
Et
semiconductor itself, causing the latter to E SS
undergo corrosion. If the bias potential is Electrolyte
e−
Fig. 14 Hole injection into the VB of a
n-type semiconductor from an oxidant E VB
(e.g. Fe3+ ) and the injection or
recombination pathway. Both surface e−
state–mediated and depletion layer e−
trap–mediated routes are shown for the
h+ h+
recombination.
couple can interact with both bands [55, of n-type and p-type samples of a given
65]. Thus, the oxidation of Eu2+ is a VB semiconductor. It may be noted that for
process (occurring at p-GaAs but not at n- a redox process occurring via one of the
GaAs in the dark), whereas the reduction bands (e.g. VB), the cathodic currents (elec-
of Eu3+ (a facile process that reportedly tron transfer from VB to Ox) are expected
occurs at rates close to the thermionic to be equal for n- and p-type materials.
emission limit, Ref. 55) is mediated by CB This idea has been pursued using the so-
electrons [65]. The [Cr(CN)6 ]3−/4− redox called quasi-Fermi level concept [55, 72,
system behaves in a similar manner with 73]. This model has been demonstrated
respect to GaAs [66]. quantitatively by studying the anodic de-
Electroluminescence, (EL), is a versatile composition of GaAs and the oxidation of
probe for studying such carrier injection redox species such as Cu+ and Fe2+ at n-
processes. Thus, hole injection into the VB and p-type GaAs electrodes [72, 73].
of a n-type semiconductor leads to cathodic
EL, whereas electron injection into the CB 1.4.2
of a p-type semiconductor leads to anodic Dark Processes Mediated by Surface States
EL [67]. Examples of studies of cathodic or by Space Charge Layer Recombination
EL are commonplace [68–70]; however,
anodic EL is not very common because Surface states were considered earlier
the energy requirement for the redox (Sect. 1.3.4) from an electrostatic perspec-
couple has a very negative redox potential. tive. Here we examine their dynamic
Nonetheless, anodic EL has been reported consequences. There are two principal
for the p-InP-[Cr(CN)6 ]4− interface [71]. charge transfer routes involving surface
Radical intermediates can also cause EL states. Consider again an n-type semicon-
as discussed later for multielectron redox ductor; the forward-bias current can either
processes. EL is treated in more depth in involve direct exchange of electrons be-
another chapter. tween the semiconductor CB and Ox states
This finally brings us to the compa- in solution (Fig. 12b) or can be mediated
rability of the current-potential behavior by surface states (Fig. 15). The second
e−
ECB
EF
ESS
Electrolyte
Fig. 15 Surface
state–mediated electron
EVB injection from the CB of a n-type
semiconductor into the
electrolyte.
route involves hole injection into the semiconductor CB also came from AC
semiconductor VB again from Ox states in impedance spectroscopy measurements
solution (Figs. 9c and 14). The recombina- on n-ZnO, n-CdS, and n-CdSe in contact
tion current is mediated either by surface with [Fe(CN)6 ]3− species [76, 77].
states or via space charge layer recombina- The electrochemical impedance for
tion. We will consider first the CB process. surface state–mediated charge transfer
Initial evidence for the intermediacy of has been computed recently [78]. The
surface states came from dark current key results are summarized in Fig. 16.
measurements on n-TiO2 and n-SrTiO3 Figure 16(a) contains the proposed equiv-
in the presence of oxidizing agents such as alent circuit for the process and features a
[Fe(CN]6 3− , Fe3+ , and [IrCl6 ]2− [74, 75]. parallel connection of the impedance for
Similar early evidence that the charge the Faradaic process [ZF (ω)] (ω = angular
transfer process was more complex than frequency, 2πf ) and the capacitance of the
direct transfer of electrons from the semiconductor depletion layer, CSC . The
C1
C2
R1
(a) R2 ∆
−1500
−1000
Xs / Ω
−500 1 Hz
100 kHz 1 kHz 100 mHz
100 Hz
10 Hz
0 10 kHz 10 mHz
1 MHz
0 500 1000 1500 2000 2500 3000 3500

(b) Rs / Ω
Fig. 16 Equivalent circuit (a) and a simulated parallel branch components represent the
Nyquist plot (b) for the charge-transfer pathway Faradaic charge-transfer process. Refer to the
illustrated in Fig. 15. The capacitance C1 original work for further details. (Reproduced
represents that of the space charge layer and the with permission from Ref. 78.)
former also involves a diffusion impedance whether recombination is mediated via

(%) of the Warburg type (see following). surface states or whether it occurs in the
The complex plane (Nyquist) plot predicted depletion layer. Thus, the parameter Rr
for the circuit is illustrated in Fig. 16(b). alone cannot afford this information and
The theoretically predicted AC impedance both the real and the imaginary parts of
response was compared with experiments this additional impedance must be consid-
on n-GaAs rotating disk electrodes in ered. Subsequent studies have addressed
sulfuric acid media [79]. The equivalent this aspect [85, 89–93]. The admittance
circuit in Fig. 16(a) was also compared corresponding to recombination at the sur-
with previous versions proposed by other face [92] and in the space charge layer [93]
authors [80–83]. These alternate versions was calculated from first principles. These
differ in their assumption of no variations computations show that the recombina-
in potential drop across the Helmholtz tion capacitance increases monotonically
region (i.e. infinite CH ) and no concentra- with decreasing band bending in the latter
tion polarization in the electrolyte phase case, whereas it shows a peak in the former
(infinite diffusion coefficient for the redox case as a function of potential.
species). Also discussed is the application Experimental studies [91] show that in
of AC impedance spectroscopy for study- the case of n-GaAs electrodes in contact
ing the kinetic reversibility of majority with Ce4+ as the hole injection agent, sur-
carrier charge transfer via the CB of a face recombination prevails. On the other
n-type semiconductor [82]. hand, with n-GaP electrodes, recombina-
AC impedance spectroscopy also has tion in the depletion layer must also be
seen extensive utility in the study of the taken into account. Other discussions of
hole injection or recombination process the use of AC impedance spectroscopy for
depicted in Fig. 14. An equivalent circuit the study of hole injection or recombina-
for this process is illustrated in Fig. 17; tion are contained in Refs. 78 and 84.
it does resemble the circuit in Fig. 16(a), The consequences of potential drop vari-
except for the Warburg component [84]. ations across the Helmholtz layer in the
Early studies [85–88] utilized the recombi- hole injection process have been exam-
nation resistance parameter, Rr , that was ined by a variety of techniques [94, 95]. For
extracted from model fits of the measured example, chemical reaction of the GaAs
AC impedance data. This parameter was surface with iodine results in a downward
seen to be inversely related to the hole in- shift of the semiconductor band edge po-
jection current, thus signifying that it is in- sitions such that the reduction of iodine
deed related to the recombination process. is mediated by CB electrons [95]. When
However, the challenge is to differentiate sufficient negative charge accumulates at
Csc
C Fig. 17 Equivalent circuit

representation of the injection
or recombination process.
R1
(Reproduced with permission
R2
from Ref. 84.)
the surface, the potential is redistributed the interface itself. However, values for ket
between the semiconductor spacecharge approximately three orders of magnitude
layer and the Helmholtz region. Now higher have also been reported for simi-
iodine is reduced by hole injection as lar interfaces, namely, n-GaAs-acetonitrile-
gauged by EL and AC impedance mea- containing cobaltocenium [Co(Cp)2 + ] ac-
surements [95]. ceptors [99, 101]. Similarly, high values
were reported for p-GaAs-acetonitrile in-
1.4.3 terfaces with ferricenium and cobaltoce-
Rate-limiting Steps in Charge Transfer nium redox species [102]. In these lat-
Processes in the Dark ter cases, alternative mechanisms (e.g.
thermionic emission, see following) must
The assumption is implicit in the discus-
be invoked in a rate-limiting role. Quartz
sion in Sect. 1.4.1 (leading to Eq. 18) that
crystal microbalance measurements have
charge transfer kinetics at the semicon- yielded evidence for adsorption of redox
ductor–electrolyte interface is the rate- species (and consequently high local sub-
limiting step. Fundamentally, we have strate concentration) in some of these
to differentiate majority carrier capture ‘‘anomalous’’ instances [101].
and minority carrier injection processes In the majority carrier capture process, if
in the dark. In the former case, tran- the interfacial charge transfer kinetics are
sit through the semiconductor itself or facile, the transport of majority carriers
charge exchange with the surface states through the space charge region can
can be potentially rate-limiting. In the lat- play a rate-limiting role. The thermionic
ter case, there are three steps involved: emission theory [34] assumes that every
hole injection into the semiconductor VB, electron that reaches the semiconductor
charge exchange between the recombina- surface, and has the appropriate energy to
tion center and the semiconductor CB, and overcome the potential barrier there, will
diffusion of majority carriers (electrons) cross the interface with a tunnel probability
from the neutral region. Finally, mass of unity. However, if the interfacial kinetics
transport processes in the electrolyte phase are sluggish, some of the electrons will be
itself can be a limiting factor in the over- reflected at the interface. In this case, the
all current flow. We shall examine carrier exchange current io is no longer described
capture and injection processes in turn. by Eq. (18) but by Eq. (19) [34]
The vast majority of outer-sphere, non- # ∗$ ' (
∗ m ns
adsorbing redox systems to date have io = AA T2 (19)
yielded values for ket in the 10−17 − me n
10−16 cm4 s−1 range [3, 8]. These include In Eq. (19), A is the electrode area, A∗ is the
n-Si-CH3 OH [96, 97], n-InP-CH3 OH [98], Richardson constant (120 A K−2 cm−2 ),
GaInP2 -coated n-GaAs-acetonitrile [99], m∗ /me is the relative effective electron
and p-GaAs-HCl [100] interfaces. The re- mass in the CB, and T is the absolute
dox couples in these studies have mostly temperature.
involved metallocenes that show low pro- In many of the reported instances [53,
clivity to adsorb on the semiconductor sur- 55, 103], the current calculated from
face. In these cases, the rate-determining Eq. (19) is much higher than that mea-
step in the overall current flow undoubt- sured experimentally, signaling that inter-
edly lies in the electron transfer event at facial charge-transfer kinetics are limiting
the overall rate. On the other hand, in the n- process was found to describe the charge
GaAs-acetonitrile-Co(Cp)2 + case [101], AC transfer at n-GaAs in 6 M HCl containing
impedance spectroscopy data appear to Cu+ as the hole injecting species [55, 73].
support the assumption that thermionic Whatever the limiting mechanism, ul-
emission is the current-limiting transport timately the current becomes limited by
mechanism. concentration polarization, that is, by the
Another factor that enters into this transport of redox species from the bulk
discussion is the mobility of the majority electrolyte to the semiconductor surface.
carriers. It has been argued [14] that in the The situation in this regard is no different
case of low mobility materials (e.g. µn ∼ from that at metal electrode–electrolyte
1 cm2 V−1 s−1 ), carrier transport from the interfaces. As in the latter case, hydro-
semiconductor bulk to the interface itself dynamic voltammetry is best suited to
can become limiting. Clearly a multitude study mass transport. AC impedance spec-
of factors are important in majority carrier troscopy can be another useful tool in this
capture: ket , acceptor concentration in the regard [105].
electrolyte and carrier mobility. In the former case, the data can be
What about the minority carrier injec- processed via Levich plots of current vs.
tion process depicted in Fig. 14? Here,
ω1/2 (ω = angular frequency). If processes
contrasting with the process considered
other than solution mass transport become
earlier, the hole injection step is usually
rate-limiting, then the plot will show a
very fast (see following). Then the current
curvature and the current will even become
is limited by diffusion or recombination
independent of the electrode rotation rate.
described by the Shockley equation [104]
In this case, inverse Levich (or Koutecky-
eo ADp n2i Levich) plots of 1/i vs. ω−1/2 can be
io = (20a) used for further analyses. Such analyses
nLp
have been done, for example, for n-GaAs-
for bulk recombination, and acetonitrile-Co(Cp)2 + interfaces [101] and
n- and p-GaAs electrodes in contact
io = 0.5 eo AW σ νth Nt ni (20b) with the Cu+/2+ redox couple in HCl
for recombination within the semiconduc- electrolyte [55, 73].
tor space charge region. In Eqs. (20a) and The diffusion impedance at semicon-
(20b), Dp is the diffusion coefficient for ductor electrodes has been considered
holes, Lp is the hole diffusion length, ni is recently [105]. This author described the
the intrinsic carrier concentration, σ is an applicability of AC impedance spec-
average capture cross-section for electrons troscopy for the study of electron capture
and holes, νth is the thermal velocity of and hole injection processes at n-GaAs-
charge carriers, and Nt is the areal den- H2 O/C2 H5 OH-methyl viologen, p-InP-
sity of recombination (trap) centers in the aq. KOH-Fe(CN)6 3 , n-GaAs-H2 SO4 -Ce4+ ,
middle of the energy gap (Fig. 14). and n-InP-aq. KOH-Fe(CN)6 3− interfaces.
The diffusion or recombination mecha- In the case of electron capture processes, a
nism results in considerable overpotential Randles-like equivalent circuit was found
for (cathodic) current flow in the dark to be applicable [105]. On the other hand,
(again assuming a n-type semiconduc- no Warburg component was present in
tor for illustration). Such a rate-limiting the hole injection case when the reverse
1.5 Light Absorption by the Semiconductor Electrode and Carrier Collection 25
reaction was negligible (Fig. 17). For a non- value (shorter wavelengths). Both types of
ideal semiconductor–electrolyte contact transitions can also be seen in the same
(see Sect. 1.3.4), a Warburg impedance ap- material, for example GaP [107].
peared in the electrochemical impedance Within the present context, the impor-
of an injection reaction as well, as exem- tant point to note is that the absorption
plified by the n-InP-Fe(CN)6 3− case [105]. depths (given by 1/α) are vastly different
for direct and indirect transitions. While in
the former case absorption depths span the
1.5
100–1000 nm range, they can be as large
Light Absorption by the Semiconductor
as 104 nm for an indirect transition [9].
Electrode and Carrier Collection
Optical transitions in semiconductors
1.5.1 can also involve localized states in the band
Light Absorption and Carrier Generation gap. These become particularly impor-
tant for semiconductors in nanocrystalline
The optical band gap of the semiconduc- form (see following). Sub–band gap transi-
tor (Sect. 2) is an important parameter tions can be probed with photons of energy
in defining its light absorption behavior. below the threshold defined by Eg .
In this quantized process, an electron-
hole pair is generated in the semicon- 1.5.2
ductor when a photon of energy hν Carrier Collection
(ν = frequency and hν > Eg ) is absorbed.
Optical excitation thus results in a delo- The number of carriers collected (in an
calized electron in the CB, leaving behind external circuit, for example) versus those
a delocalized hole in the VB; this is the optically generated defines the quantum
band-to-band transition. Such transitions yield (+) – a parameter of considerable in-
are of two types: direct and indirect. In the terest to photochemists. The difficulty here
former, momentum is conserved and the is to quantify the amount of light actually
top of VB and the bottom of CB are both absorbed by the semiconductor as the cell
located at k = 0 (k is the electron wave walls, the electrolyte, and other compo-
vector). The absorption coefficient (α) for nents of the assembly are all capable of
such transitions is given by [106] either absorbing or scattering some of the
incident light. Unfortunately, this problem
α = A+ (hν − Eg )1/2 (21)
has not been comprehensively tackled, un-
In Eq. (21), A+ is a proportionality con- like in the situation with photocatalytic
stant. Indirect transitions involve phonon reactors involving semiconductor partic-
modes; in this case Eq. (21) takes the form ulate suspensions, where such analyses
are available [108–111]. Pending these, an
α = A+ (hν − Eg )2 (22)
effective quantum yield can still be defined.
A given material can exhibit a direct or in- Returning to the carrier collection
direct band-to-band transition depending problem, consider Fig. 18 for an n-
on its crystal structure. For example, Si semiconductor–electrolyte interface. As
single crystals have an indirect transition can be seen, the electron-hole pairs are
located at 1.1 eV (Table 1). On the other optically generated, both in the field-free
hand, amorphous Si is characterized by a and in the space charge regions within the
direct optical transition with a larger Eg semiconductor. Recombination of these
1/α
LD + W
W
CB
hν
VB
Redox
Semiconductor electrolyte
Fig. 18 Photogeneration of electron-hole pairs in the field-free
region and depletion layer for a n-type semiconductor–
electrolyte interface. The characteristic regions defined by the
depletion layer (W), Debye length (LD ), and the light
penetration depth (1/α) are also compared.
carriers must be considered in the bulk, in depletion layer is dictated by the diffusion
the space charge layer, and on the semicon- process. A diffusion length, L, can be
ductor surface (the latter in contact with defined
the redox electrolyte). We are assuming
here that light is incident from the elec- ) )
Lp = Dp τp = kT µp τp
trolyte side. Rear illumination geometry
can be profitably employed and is consid- (for n-type semiconductor) (23)
ered later for the nanocrystalline film case.
The direction of the electric field at the
interface (Fig. 6, Sect. 1.3.2) is such that The subscripts in Eq. (23) remind us
the minority carriers (holes in this case) that we are dealing with minority carrier
are swept to the surface and the electrons collection; µp is the hole mobility and τp
are driven to the rear ohmic contact. How is the hole life-time. The characteristic
fast the holes are drained away (by Faradaic length Lp defines the region within
reactions involving the redox electrolyte) which electron-hole pair generation is fully
will dictate how the Fermi levels compare effective. Pairs generated at depths longer
with the equilibrium situation discussed than the Debye length, LD (LD = W + L)
earlier. The departure from equilibrium will simply recombine. Thus, the effective
has been quantified in terms of the quasi- quantum yield for a given interface will
Fermi level concept discussed later. depend on the relative magnitudes of
The extent of collection of minority LD and the light penetration depth, 1/α
carriers from the region beyond the (Fig. 18) [112, 113].
An expression for the flux of photo- and recalling that W is proportional to

generated minority carriers arriving at the VSC 1/2 (Eq. 8), and VSC = V − Vfb , a test of
surface was originally given for a solid-state the rudimentary model would lie in a plot
junction [114] and subsequently adapted to of the LHS of Eq. (25) against (Vfb − V )1/2 .
semiconductor-liquid junctions [115]. The Such plots are shown in Fig. 19 at four
major weakness of these early models selected wavelengths for the p-GaP-H2 SO4
hinges on their underlying assumption electrolyte interface [117].
for the boundary condition that the sur- While adherence to the Gärtner model
face concentration of minority carriers is is satisfactory for large values of VSC (i.e.
zero. As pointed out elsewhere [14], this is large band bending, see Fig. 6), the model
a demanding condition necessitating very fails close to the flat band situation. In-
high magnitudes for the interfacial charge terestingly, this problem is exacerbated as
transfer rate constant, ket (see previous the semiconductor excitation wavelength
section). A modicum of improvement to becomes shorter (Fig. 19). Thus, another
the basic Gärtner model was found [116] weakness of the basic Gärtner model [114,
by defining a flux rather than a concen- 115] is the neglect of surface recombina-
tration expression for the holes and a tion. At the flat band situation, this model
characteristic length where the bulk diffu- still predicts finite current flow arising
sion current transitions into a drift current. from the diffusive flow of minority carriers
However, this treatment still assumes that toward and out of the interface (Fig. 20).
every hole entering the depletion layer edge A variety of refinements have been made
exits this region and out into the electrolyte to take into account the surface recom-
phase. bination effect [117–132]. The earliest of
The Gärtner equation [114] can be writ-
these [119, 120, 123] involve some simpli-
ten in normalized form [113]
fying assumptions:
jph exp(−αW )
+≡ =1− (24)
Io 1 + αLp
In Eq. (24), + is the effective quantum

yield (see previous section), given by the 500
0.002
ratio of the photocurrent density (iph /A),
jph , to the incident light flux, Io . Recasting
Eq. (24) in the form
−ln (1 − Φ)
− ln(1 − +) = ln(1 + αLp ) + αW (25) 510
0.001
.. 520
Fig. 19 Gartner plots (see Eq. 25) for
the p-GaP −0.5 M H2 SO4 interface. The
numbers on the plots refer to the 530
excitation wavelength; Efb is the flat
band potential and E is the bias
potential. (Note that this notation is
different from that employed in the 0 0.5 1.0 1.5
text.) (Reproduced with permission
(Efb − E )1/2/V 1/2
from Ref. 117.)
1.0 Fig. 20 Comparison of calculated

current voltage profiles in the dark
0.8 (curve d) and under illumination
(a)
(curves a–c). Curve a is obtained from
..
the basic Gartner model. Curve (b)
0.6
(b) considers surface recombination and
(c)
Current density
curve (c) considers both surface

0.4
[mA cm2]
recombination and recombination in the

space charge layer. These simulations
0.2 (d) are for a n-type
semiconductor–electrolyte interface.
0
0.1 0.2 0.3 0.4 0.5 0.6 (Reproduced with permission
from Ref. 138.)
−0.2 Voltage
[V]
−0.4
1. There is no recombination in the At this juncture, it is worth noting that

depletion layer. That is, all the holes only one trap state at the surface has
optically generated in the bulk and been assumed till now; further it is as-
within the depletion layer (Fig. 18) are sumed that this surface state functions
swept to the surface without loss; both as a carrier recombination site and as
2. The steady state concentration of the op- a charge-transfer pathway (Fig. 21). Both
tically generated minority carriers does these assumptions are open to criticism.
not perturb the potential distribution in An alternative model invoking two dis-
the dark (Fig. 6); and tinct types of surface states – one active
3. There is a quasi-thermodynamic dis- in recombination and the other capable
tribution of minority carriers within of mediating charge transfer – has been
the depletion layer. This translates to considered [135]. Nonetheless, the most
a constant product term np across this serious flaw in the above treatments lie
region. in the neglect of carrier recombination in
the depletion layer itself (as distinct from
Surface recombination in the vast ma- recombination at the surface). Reexami-
jority of these treatments invoke the Hall- nation of Eqs. (24) and (26) shows that
Shockley-Read model [133, 134]. Defining the larger the Debye length, LD , and the
the Gärtner limiting expression (Eq. 24) as depletion layer width, W , the higher the
+G , we obtain [14] quantum yield. However, recombination
*# $ in the space charge layer must become sig-
Dp exp VSC
+ss = +G 1+ (26) nificant at some value of W , thus providing
LD (kt + St )
a further limit to carrier collection.
In Eq. (26), we have two new parameters, Recombination within the space charge
kt and St . These are the first-order rate con- region is a nontrivial problem to treat
stant for hole transfer (units of cm s−1 , see from a computational perspective [14].
following) and the surface recombination The methodology of Sah, Noyce, and
velocity, St . In the combined situation of Shockley [136] has been used by several
high LD , high kt , and very low (or zero) St , authors [126, 127, 128, 131, 138–140].
Eq. (26) collapses to the Gärtner limit. Figure 20 illustrates the sensitivity of the
ECB
e−
ESS
hν
hν
E VB
h+
Fig. 21 Surface state mediation of both minority carrier (i.e. hole)
transfer and recombination for a n-type semiconductor electrolyte
interface.
current-potential profiles at the semicon- the minority carrier current (iph ) is directly
ductor–electrolyte interface to this recom- proportional to the photon flux (Eq. 24).
bination mode [137]. Thus, the net current is given by:
Other models taking the above nonideal- ! # $ "
eo V
ities to varying extent have been proposed; i = iph − io exp − −1 (27)
kT
a detailed discussion of these lies be-
yond the scope of this section [141–147]. The minus sign in Eq. (27) underscores the
However, it is worth noting here that fact that the majority carrier component
in some instances involving high-quality flows opposite (or ‘‘bucks’’) the minority
semiconductor–electrolyte interfaces the carrier current flow. This photocurrent
rate-determining recombination step does component is shown as curves 2 and 3
indeed appear to lie in the bulk semi- in Fig. 13.
conductor [1, 148]. Silicon photoelectrodes Equation (27) shows that the diode
in methanolic media containing fast, one- equation is offset by the iph term; this is
electron, outer-sphere redox couples were exactly what is seen in Fig. 13. The plateau
studied in these cases. photocurrent is proportional to the photon
flux, Io , as illustrated for two different
1.5.3 values of the incident light intensity in
Photocurrent-potential Behavior Fig. 13. At the open-circuit condition of the
interface, i = 0 (and neglecting the unity
The current-voltage characteristics of an term within the square brackets relative to
illuminated semiconductor electrode in the exponential quantity), Eq. (27) leads to
contact with a redox electrolyte can be
kT iph
obtained by simply adding together the Voc $ ln (28)
majority and minority current compo- eo io
nents. The majority carrier component is Equation (28) underlines two important
given by the diode equation (Eq. 17) while trends: First, Voc increases logarithmically
with the photon flux (with a slope of curves, as illustrated in Fig. 22. Thus,
∼60 mV at 298 K). Second, Voc decreases the superposition principle [149–154] is
with an increase of io (again logarithmi- not obeyed in this instance. The dashed
cally). This underlines the importance of line in Fig. 22 is produced by translat-
ensuring that the majority carriers do not ing the photocurrent-voltage data by jSC ,
‘‘leak’’ through the interface. Because of the short-circuit current density. If the su-
the diode nature of the interface, from perposition principle is held, this dashed
a device perspective, the semiconductor curve would have overlaid the dark current-
surface must be designed to have fast mi- voltage curve. Thus, this ‘‘excess’’ (forward
nority carrier transfer kinetics (and thus bias) current embodies the APE, and the
high iph ), but must be blocking to the failure of superposition is quantified as the
flow of majority carriers (from the CB voltage difference (%V ) between the dark
for a n-type semiconductor) into the re- j − V data and the dashed line.
dox electrolyte. This challenge is similar to What are the ramifications of the cross-
what the solid-state device physicists face, over or the APE? First, mathematical
but relative to metals (with a high density modeling of carrier transport in a junction-
of acceptor states), chemical control of re- based solar photovoltaic system, accord-
dox electrolytes offers a powerful route to ing to
performance optimization of liquid-based j = jSC − jbk (V ) (27a)
interfaces as also pointed out by previous
authors [1, 6, 8, 149–154]. is not valid in the presence of this effect.
Referring back to Fig. 13, the current- (In Eq. (27a), jbk is the ‘‘bucking’’ cur-
potential curves under illumination of rent density in the forward-bias regime,
the semiconductor simply appear shifted see previous section.) That is, a fully lin-
‘‘up’’ relative to the dark i − V counter- earized system of differential equations
part. This, however, is the ideal scenario. and boundary conditions cannot be used
Anomalous photoeffects (APEs) are of- to model the interface carrier transport.
ten observed that manifest as a cross- Second, computation and modeling of the
over of light and dark current-voltage open-circuit voltage for such devices by
Light
Current density
∆V
Dark
Fig. 22 Anomalous photoeffect (APE)
showing cross-over of the dark and light
current-voltage curves again for a n-type
semiconductor-based interface. The
dashed line is obtained as described in
Potential the text.
simply equating a constant photocurrent state–mediated electron transfer to O2 (in

flux, jph against the dark (recombina- the electrolyte) as the origin of the photo-
tion) current, jo is no longer possible effect. Surface states were again invoked
(see Eq. (28) and the accompanying earlier to explain a cathodic photoeffect at n-
discussion). CdS-aqueous polysulfide interfaces [159].
Third, and perhaps practically of most This photoeffect was only observed for the
significance, the %V component repre- (0001) single-crystal face of n-CdS but not
sents a loss pathway in the photovoltage for the (0001) orientation. A subsequent
deliverable by the given device. Thus, the study of photoelectrochemical effects at
(open-circuit) photovoltage is Voc instead selenium films reported an anomalous
of Voc + %V in the ideal case in the ab- anodic photocurrent at potentials positive
sence of the APE. Therefore, it is important of the flat band location for the p-type
to quantify and understand the molecular film [160]. This effect was assigned to
and chemical origins of this effect. This has a hole injection process via a tunneling
not been done so far, at least to this writer’s mechanism. An increase of the tunnel-
knowledge, for semiconductor–electrolyte ing probability under illumination was
interfaces. accommodated by a shrinking of the space
Of course, the cross-over effect is not charge layer at the interface. Photoen-
confined to such interfaces. It is interest- hancement of the forward current flow
ing that a recent textbook [155], dealing was observed again for n-CdS, in this in-
primarily with solid-state solar cell de- stance in contact with a [Fe(CN)6 ]3−/4−
vices, makes only a fleeting reference to redox electrolyte [161]. This effect was ob-
the underlying origin of the APE. Ref- served only with a mechanically damaged
erence was made in this book to the surface, and disappeared after it had been
cross-over of experimental dark and light etched with concentrated HCl.
j − V characteristics for a Cu2 S-CdS solid- Subsequent work [162] describes sup-
state heterojunction solar cell but its origin pression of the cathodic photocurrent
was not explored. A light-induced junction for n-CdS-[Fe(CN)6 ]3−/4− interfaces me-
modification has also been reported for the chanical polishing of the electrode. As
(Cd, Zn)S-CuInSe2 solid-state system [156, with an earlier study [163], the spectral
157]. The cross-over effect appears to have dependence of the photocathodic effect
been treated in even lesser depth in some implicated sub–band gap states. The sup-
classical textbooks on semiconductor depression was explained by two alternative
vices [104, 155]. models involving a compensated insu-
Probably the first reported instance of lating layer or by Fermi-level pinning.
observation of an APE was in 1977 for Illumination was claimed to result in a dra-
a n-TiO2 -NaOH electrolyte interface [158]. matic increase of the (suppressed) cathodic
The APE was observed in the saturation current, which interestingly enough was
region of cathodic current flow and was observed only for n-CdS films but not
induced by sub–band gap irradiation of for crystals including n-CdTe, n-CdSe, n-
the photoanode. A peak in the spectrum GaAs, n-ZnO, n-TiO2 and n-ZnSe. On the
of the photoresponse at 800 nm (the cor- other hand, a subsequent paper describes
responding photon energy being lower a photocathodic effect for n-CdSe-sulfide
than the 3.0 eV band gap of TiO2 ) was interfaces in which an interfacial layer of
used by the authors to invoke a surface selenium was implicated [163].
5 × 100 Fig. 23 Experimental data embodying

APE for the n-GaAs-AlCl3 /n-butylpyridi-
nium chloride molten-salt electrolyte
interface. Refer to the text for details.
5 × 10−1
(Reproduced with permission from
Ref. 167.)
5 × 10−2
[mA cm−2]
Current
5 × 10−3
5 × 10−4
5 × 10−5
−50 −150 −350 −550
Voltage vs. (Al/Al+3)
[mV]
More recent studies on n-CdS [164, 165] Figure 23 contains an example of the
and n-TiO2 photoanodes [166] implicate APE for the n-GaAs-AlCl3 /n-butylpyridi-
the formation of photoconductive layers nium chloride molten salt electrolyte inter-
in the APEs. Thus, the foregoing review face [167]. The bottom curve in Fig. 23 is
suggests the following the measured dark current-voltage profile.
The top curve was obtained from the
1. APE is a very general phenomenon that photocurrent-voltage data (under irradi-
has been observed for solid-state junc- ation of the semiconductor). Clearly, if
tions for n- and p-type semiconductors the superposition principle held, the two
alike, and for a wide variety of semicon- curves would have coincided with one
ductor materials. another.
2. The reported results and trends are APEs have also been observed for
often contradictory. It is quite possible nanocrystalline and chemically modified
that the experimental conditions in films, as discussed in a subsequent section.
these studies were quite variant, thus Light-induced changes in the electro-
precluding direct comparison of the statics at the semiconductor–electrolyte
results. interface are conveniently probed by
3. The mechanistic reasons given for capacitance-voltage measurements in the
the APE are possibly many, and gen- dark and under illumination of the semi-
eralizations may not be warranted. conductor electrode. If charge trapping
Clearly, more research is needed on at the interface plays a decisive role
this topic. (whatever be the mechanism), the voltage
drop across the illuminated interface is What are the processes important in a
altered, and consequently the semicon- dynamic interrogation of the semiconduc-
ductor band edge positions are shifted. tor–electrolyte interface?
This, of course, is the Fermi-level pin-
ning situation that was encountered earlier 1. Carrier generation within the semicon-
(Sect. 1.3.4). Examples of studies address- ductor,
ing this aspect may be found, for ex- 2. diffusion of minority carriers from the
ample on p – GaAs [124] and CdTe [168, field-free region to the space charge
169]–based aqueous electrolyte interfaces. layer edge,
3. transit through this layer,
1.5.4 4. charge transfer across the interface, and
Dynamics of Photoinduced Charge Transfer 5. carrier recombination via surface states
or via traps in the space charge layer.
So far the discussion has centered on
the steady state aspects of carrier gen- Other phenomena such as thermaliza-
eration and collection at semiconduc- tion also are important as discussed later in
tor–electrolyte interfaces. As with their the context of hot carrier effects. The time
metal electrode counterparts, a wealth constant (τcell ) of the cell and the measure-
of information can be gleaned from ment circuitry has complicated matters
perturbation-response type of measure- further and have caused some confusion
ments. An important difference, however, in the interpretation of transient data. If
lies in the vastly different timescale win- a potentiostat is not used, then this time
dows that are accessible in the two cases. constant is given by Eq. (29).
The critical RC time-constant of the cell in One can envision three types of per-
a transient experiment is given by turbation: an infinitesimally narrow light
pulse (a Dirac or δ-functional), a rectan-
τcell = C(Rm + Rel ) (29) gular pulse (characteristic of chopped or
In Eq. (29), Rm is the measurement interrupted irradiation), or periodic (usu-
resistor (across which the current or ally sinusoidal) excitation. All three types
photocurrent is measured) and Rel is the of excitation and the corresponding re-
electrolyte resistance. The term C is the sponses have been treated on a common
capacitance, which, in the metal case, is the platform using the Laplace transform ap-
Helmholtz layer capacitance, CH . (Once proach and transfer functions [170]. These
again, the Gouy region is ignored here.) perturbations refer to the temporal be-
For semiconductor–electrolyte interfaces, havior adopted for the excitation light.
we have seen that two layers are involved However, classical AC impedance spec-
in a series circuit configuration with troscopy methods employing periodic po-
corresponding capacitances of CSC and tential excitation can be combined with
CH (Fig. 6). Because CH # CSC , C $ steady state irradiation (the so-called PEIS
CSC . This assumption is usually justified experiment). In the extreme case, both
because CH $ 10−5 F cm−2 and CSC = the light intensity and potential can be
10−8 − 10−0 F cm−2 . If the composite modulated (at different frequencies) and
resistance (Rm + Rel ) is 100 ohm, then the (nonlinear) response can be mea-
τcell for metal electrodes is ∼10−3 s sured at sum and difference frequencies.
and that for the semiconductor case is The response parameters measured in
10−6 − 10−7 s. all these cases are many but include
the photocurrent, voltage, luminescence, technologically important electrochemical

or microwave conductivity. Clearly, semi- process that has also been extensively
conductor–electrolyte interfaces present a studied from a mechanistic perspective on
rich, albeit demanding landscape for prob- metallic electrodes.
ing non–steady state phenomena. Photoelectrolytic processes such as HER
The dynamics of charge transfer across can be carried out on semiconductor
semiconductor–electrolyte interfaces are electrodes. One can envision a HER
considered in more detail elsewhere in mechanism on a p-type semiconductor of
this volume. the sort:
hν
1.5.5 p − SC −−−→ e− +
CB + hVB (30)
Hot Carrier Transfer
e−
CB + S −−−→ S
−
(30a)
Short wavelength photons (of energy much − +
S + H −−−→ S + H •
(30b)
greater than Eg ) create ‘‘hot’’ carriers. If,
somehow, thermalization of these carri- e− +
CB + H −−−→ H
•
(30c)
ers can be avoided, photoelectrochemical +
H +H •
+ e−
CB −−−→ H2 (30d)
reactions that would otherwise be impos-
sible with the ‘‘cooled’’ counterparts, that H+ + H• + S −−−→ H2 + S
−
(30e)
is, at very negative potentials for n-type
h+
VB + H −−−→ H
• +
(30f)
semiconductors, would be an intriguing
possibility. The key issue here is whether h+
VB + S− −−−→ S (30g)
the rate of electron transfer across the
In this above scheme, S denotes a
interface can exceed the rate of hot elec-
surface state and both direct (Reactions
tron cooling. The observation of hot carrier
30c and 30d) and indirect (i.e. surface
effects at semiconductor–electrolyte inter-
state–mediated) (Reactions 30b and 30e)
faces is a controversial matter [3, 7, 11, 171]
radical and H2 -generating pathways are
and practical difficulties include problems
shown. Reactions 30f and 30g represent re-
with band edge movement at the interface
combination routes involving the reaction
and the like [4]. Under certain circum-
intermediates.
stances (e.g. quantum-well electrodes, ox-
Admittedly, this scheme is daunting in
ide film-covered metallic electrodes), it has
its complexity, and the kinetics implica-
been claimed that hot carrier transfer can
tions are as yet unclear. Early studies on
indeed be sustained across the semicon-
p-GaP, p-GaAs, and other Group III–V
ductor–electrolyte interface [7, 172, 173].
(13–15) semiconductors reported onset
of cathodic photocurrents (attributable
1.6 to HER) only at potentials far removed
Multielectron Photoprocesses (ca. 0.6 V) from Vfb [174]. This was at-
tributed to Steps 30b and 30g in the
This section has thus far considered redox preceding scheme. More recent work [175]
electrolytes comprising one electron oxi- has shown that the HER at illuminated
dizing or reducing agents. Multielectron p-InP-electrolyte contacts is accompanied
redox processes, however, are important by a photocorrosion reaction, leading to
in a variety of scenarios. Consider the indium formation on the semiconductor
reduction of protons to H2 (HER) – a surface.
1.6 Multielectron Photoprocesses 35
Interestingly, surface states themselves photon flux values. Even multiplication

were chemically identified with H•ads (ad- factors as high as 4 are possible as in
sorbed hydrogen atom intermediates) in the photodissolution of n-Si in NH4 F me-
the aforementioned study [166]. These dia [179–182].
species have also been implicated in ac- Photocurrent multiplication has been
cumulation layer formation and anodic observed for a variety of semiconduc-
EL at n- and p-GaAs-electrolyte inter- tors including Ge [180], Si [179–182],
faces [176–178]. ZnO [183–189], TiO2 [190–193], CdS [194,
Another interesting characteristic of 195], GaP [196], InP [197], and GaAs [198–
many multiequivalent redox systems is 200]. These studies have included both
the phenomenon of photocurrent mul- n- and p-type semiconductors, and have
tiplication. This phenomenon may be spanned a range of substrates, both or-
illustrated for two systems utilizing illu- ganic and inorganic. As in the Si case,
minated n-type and p-type semiconductors this phenomenon can also be caused by
respectively photodissolution reactions involving the
n-type semiconductor itself. The earlier studies
hν have mainly employed voltammetry, par-
n-SC −−−→ e− + h+ (31) ticularly hydrodynamic voltammetry [193].
HCOOH + h+ −−−→ COOH• + H+ As more recent examples [2, 9, 10] reveal,
intensity-modulated photocurrent spec-
(31a) troscopy, (IMPS), is also a powerful
COOH• −−−→ CO2 + H+ + e− (31b) technique for the study of photocurrent
multiplication.
p-type This leads us to another important
hν category of multielectron photoprocesses
p-SC −−−→ h+ + e− (31)
involving the semiconductor itself. While
+ − photocorrosion is a nuisance from a device
O2 + H + e −−−→ HO2 •
(31c)
operation perspective, it is an important
HO2 • + H+ −−−→ H2 O2 + h+ (31d)
component of a device fabrication se-
Thus, the key feature of photocurrent mul- quence in the microelectronics industry.
tiplication is a majority carrier injection Two types of wet etching of semiconduc-
step (Reactions 31b or 31d) from a reaction tors can be envisioned [201]. Both occur
intermediate (usually a free radical) into at open-circuit but one involves the action
the semiconductor CB or VB, respectively. of chemical agents that cause the simul-
In the preceding examples, each photon taneous rupture and formation of bonds.
generates two carriers in the external cir- Several aspects of photoetching have been
cuit, affording a quantum yield (in the ideal reviewed [62–64, 202, 203] including re-
case) of 2. This is the ‘‘current-doubling’’ action mechanisms, morphology of the
process. etched surfaces, and etching kinetics in
Practically, however, quantum yields the dark and under illumination. Gen-
somewhat lower than 2 are usually eral rules for the design of anisotropic
measured because Steps 31b or 31d photoetching systems have also been
compete with the further photooxidation formulated [204]. Photoelectrochemical et-
or photoreduction of these intermediates, ching is considered in more detail else-
respectively. This is true especially at high where in this volume.
1.7 in this chapter. By way of contrast, the

Nanocrystalline Semiconductor Films and amount of species that can be confined
Size Quantization in a monolayer on a corresponding flat
1.7.1 surface would be negligibly small.
Introductory Remarks As we shall see later, electron trans-
port in nanocrystalline films is necessar-
From a materials perspective, the field of ily accompanied by charge-compensating
semiconductor electrochemistry and pho- cations because the holes are rapidly in-
toelectrochemistry has evolved from the jected into the flooded electrolyte phase.
use of semiconductor single crystals to This provides opportunities for studying
polycrystalline thin films and, more re- ion transport processes in mesoporous
cently, to nanocrystalline films. The latter media that are coupled to electron mo-
have been variously termed membranes, tion. Ion insertion also has practical con-
nanoporous or nanophase films, meso- sequences as in energy storage device
porous films, nanostructured films, and applications [206].
so on; they are distinguished from their Surface state densities on the order of
polycrystalline electrode predecessors by ∼1012 cm2 are commonplace for semi-
the crystallite size (nm versus µm in conductor electrodes of the sort consid-
the former) and by their permeability ered in previous sections of this chapter.
to the electrolyte phase. These films are These translate to equivalent volume den-
referred to as ‘‘nanocrystalline in the fol- sities of ∼1018 cm−3 for nanocrystalline
lowing sections. These features render films. Such high densities enhance light
three-dimensional geometry to nanocrys- absorption by trapped electrons in sur-
talline films as opposed to the ‘‘flat’’ or face states, giving rise to photochromic
two-dimensional (planar) nature of single and electrochromic effects [198–200] (see
crystal or polycrystalline counterparts. following). Unusually high photocurrent
What are the virtues of these emerging quantum yields are also observed with
photoelectrode materials? The first is sub–band gap light with these photoelec-
related to their enormous surface area. trode materials. Corresponding sub–band
Consider that the 3D structure is built up gap phenomena are rather weak and
of close-packed spheres of radius, r. Then difficult to detect with single-crystal coun-
ignoring the void space, the specific area, terparts.
As (area/volume) is given by 3/r [205]. 1.7.2
For r = 10 nm, As is on the order of The Nanocrystalline Film–Electrolyte
106 cm−1 , and for a 1 cm2 film of 1 µm Interface and Charge Storage Behavior in
thickness, this value corresponds to an the Dark
internal surface area of ∼100 cm2 (i.e.
a ‘‘surface roughness factor’’ of 100). Understanding of the electrostatics across
Clearly, this becomes important if we nanocrystalline semiconductor film-elec-
want the electrolyte redox species to be trolyte junctions presents interesting chal-
adsorbed on the electrode surface (see lenges, particularly from a theoretical per-
following). Alternatively, a large amount of spective. Concepts related to space charge
sensitization dye can be adsorbed onto the layers, band bending, flat band potential,
support semiconductor although this dye and the like (Sect. 3) are not applicable
sensitization approach is not considered here because the crystallite dimensions
1.7 Nanocrystalline Semiconductor Films and Size Quantization 37
comprising these layers are comparable to detected, for example, by electron param-
(or even smaller than) nominal depletion agnetic resonance spectroscopy [215, 216].
layer widths. In either case, the resultant negative
The rather complete interpenetration of charge generated by electron accumula-
the electrode and electrolyte phases must tion at the internal surfaces has to be
mean that the Helmholtz double layer ex- balanced by cations (from the electrolyte
tends throughout the interior surface of the phase) for charge compensation. Such
nanoporous network, much like a superca- ion insertion reactions have been stud-
pacitor [9, 210] situation. Finally, unlike in ied using techniques such as voltammetry,
the cases treated earlier, the semiconduc- reflectance or absorption spectroscopy,
tor (especially if it is a metal oxide) is not chronoamperometry, and electrochemi-
heavily doped such that free majority carri- cal quartz-crystal microgravimetry [213,
ers are not present in appreciable amounts. 217–22]. Both aqueous and aprotic elec-
This is indicated, for example, by the sensi- trolytes have been deployed for these
tivity of the conductivity of nanocrystalline studies.
TiO2 layers to UV light – the conductivity Unlike in the single-crystal cases treated
is strongly enhanced on UV exposure, sim- earlier, placement of the semiconductor
ilar to a photoconductive effect. This effect energy band positions at the interface
has been interpreted, in terms of trap fill- via Mott-Schottky analyses is not straight-
ing with recombination times considerably forward for nanocrystalline films. Abrupt
slower than the trapping processes under changes in slope and other nonideali-
reverse bias [211, 212]. The light sensitivity ties [215, 227, 223] have been observed,
also is diagnostic of the fact that the low for example, in the Mott-Schottky plots for
electronic conductivity in the dark is not TiO2 films and attributed to the influence
due to high interparticle resistances (i.e. in of the conductive glass that is normally
the ‘‘neck’’ regions), but rather is indicative employed to support these films. This be-
of the low electron concentrations. havior is especially prevalent under reverse
The electron concentration can be in- bias. The onset of majority carrier op-
creased by forward-biasing the nanocrys- tical absorption (in the visible and near
talline electrode–electrolyte interface po- IR range) under forward-bias instead has
tentiostatically. The interface is driven thus been profitably employed to place the CB
into the accumulation regime for the ma- positions of TiO2 in aqueous media [208].
jority carriers, and if a transparent rear Impedance spectroscopy and electro-
contact (e.g. F-doped, SnO2 or Sn-doped chemical dye desorption experiments have
indium oxide) is used, the resultant blue been employed [224] to study the electri-
(or bluish-black) coloration of the film cal characteristics of TiO2 nanocrystalline
can be spectroscopically monitored [208, films in the dark. This study as well as the
209, 213]. Whether the optical response others cited earlier demonstrate how the
arises from CB electrons or from electrons conductivity changes (as a result of elec-
trapped in surface states is not entirely tron injection from the support electrode)
clear. It has been claimed [214] that the can cause the porous or nanocrystalline
absorption spectrum of the latter differs layer to manifest itself electrically, such
significantly from CB electrons. Electrons that the active region moves away (i.e.
in surface states can be chemically iden- outward) from the support as the forward-
tified with Ti3+ defect sites that can be bias voltage is increased. The potential
distribution has also been analyzed de- by the electrolyte redox (specifically Red)
pending on whether the depletion layer species, collection of the photogenerated
width exceeds or is smaller than the typi- electrons at the rear contact becomes
cal dimension of the structural units in the the determinant factor in the quantum
nanocrystalline network [223]. yield. Thus, the quasi-Fermi level for holes
remains close to EF,redox and that for
1.7.3
electrons, EF,e moves ‘‘up’’, as depicted
Photoexcitation and Carrier Collection:
in Fig. 24b. Illumination thus induces an
Steady State Behavior
electron flux, Jn (x) through the nanocrys-
Figure 24 contains a schematic representa- talline phase. Under steady state condi-
tion of the nanocrystalline semiconductor tions, the photocurrent density (jph ) is
film–electrolyte interface at equilibrium equal to eo Jn (x = d). The driving force for
(Fig. 24a) and the corresponding situation electron diffusion through the network of
under band gap irradiation of the semicon- nanocrystallites has been calculated from
ductor (Fig. 24b) [9]. Because the diffusion first principles [225]. It has been found that
length of the photogenerated carriers is the driving force is approximately kT /eo
usually larger than the physical dimen- divided by the thickness of the network.
sions of the structural units, holes and Importantly, this free-energy gradient is
electrons can reach the impregnated elec- found to be independent of the incident
trolyte phase before they are lost via bulk photon flux.
recombination. This contrasts the situa- It is important to reiterate that the
tion with the single-crystal cases discussed charge separation in a nanocrystalline
earlier. semiconductor–electrolyte interface does
If, as is the case with TiO2 nanocrystal- not depend on a built-in electric field
line films, the holes are rapidly scavenged at the junction as in the single-crystal
E
EC
E F,redox
EV
Substrate
d 0
(a) x
E
− − −
je(d ) E F,e
E F,redox Fig. 24 Schematic representation of a
nanocrystalline semiconductor–
+
electrolyte interface in the dark (a) and
under illumination from the electrolyte
Substrate Red side (b). Ec and Ev correspond to ECB
Ox
and EVB in our notation. (Reproduced
d 0
with permission from the authors of
(b) x Ref. 9.)
1.7 Nanocrystalline Semiconductor Films and Size Quantization 39
case. Instead, the differential kinetics for sensitization experiments) although other
the reactions of photogenerated electrons redox electrolytes [e.g. SCN− /(SCN)2 − ;
and holes with electrolyte redox species 228] have been employed as well. For the
account for the charge separation (and the chalcogenide films, sodium selenosulfite
generated photovoltage). The molecular was employed [227]. It must be noted that,
factors underlying the sluggish scavenging aside from losses caused by the surface
of electrons at the nanocrystalline film- recombination and back-reactions, an ad-
electrolyte boundary (by the redox species) ditional loss component from the increase
are as yet unclear. Clearly, the competition in film resistance must also be recog-
between surface recombination of these nized, especially as the film thickness is
electrons (with the photogenerated holes) increased. The resistance loss manifests
and collection at the rear contact dictates as a deterioration in the photovoltage and
the magnitude of the quantum yield that is fill factor.
experimentally measured for a particular In the discussion to this point, we have
junction. not considered trapping or release of the
Photocurrent losses have been recorded photogenerated electrons as they undergo
for electrolytes dosed with electron accep- transit to the rear contact. However, elec-
tors such as O2 and iodine [226]. Nanocrys- trons trapped in localized interfacial states
talline TiO2 electrodes with thicknesses induce a countercharge in the Helmholtz
ranging from 2 µm to 38 µm were included double-layer, as discussed in the preced-
in this study. In the presence of these ing discussion. The resultant voltage drop
electron-capture agents, electron collec- can introduce a nonnegligible field com-
tion (i.e. photocurrent) at the rear contact ponent into the diffusional process. The
was seriously compromised. On the other time-dependence of the electron density,
hand, as high as 10% of the photons were n(x, t) is given by [9]
converted to current for a 38 µm thick film
in a N2 -purged solution [226]. ∂n(x, t) ∂ 2 n(x, t)
= ηαIo e−αx + Dn
The result was obtained with front-side ∂t ∂x 2
illumination geometry. As one would in- n(x, t) − no
tuitively expect, carrier collection is most − (32)
τ
efficient close to the rear contact. In-
deed, marked differences have been ob- In Eq. (32), η is the electron injection
served for photoaction spectra with the efficiency, Dn is the diffusion coefficient
two irradiation (i.e. through the elec- of electrons, and τ is the pseudo first-
trolyte side vs. through the transparent order lifetime of electrons determined by
rear contact) geometries for TiO2 , CdS, back-reaction with Ox.
and CdSe nanocrystalline films [227, 228]. Even if the migration component is
Obviously, the relative magnitudes of negligible (but see following), solution of
the excitation wavelength and the film Eq. (32) presents difficulties because of the
thickness critically enter into this variant possible dependence of Dn on n and x.
behavior. Similarly, τ may depend on these two vari-
In the vast majority of cases, the ables also. Nonetheless, the steady state
iodide/triiodide redox couple has been em- solution of Eq. (32) has been obtained [229]
ployed (presumably because of its success by assuming that D and τ are constant
in shuttling the photooxidized dye in the and that η = 1. Under these conditions,
the photocurrent is predicted to be in- insulating and (2) the electrolyte perme-
dependent of voltage – a rather physically ates into the network and possibly contacts
implausible situation. In the forward-bias the rear support electrode. The transition
regime, η is expected to decrease and the in the profiles from spiked at potentials
back-reaction of the photogenerated elec- near Von (photocurrent onset) to more
trons (with Ox) can no longer be neglected. rectangular at positive bias potentials (not
This brings us to the rear support-film shown in Fig. 25) must mean that the volt-
interface. What sort of barrier exists at this age does exert an effect on carrier transit
junction? Are the electron exchange ki- through the network. No satisfactory expla-
netics voltage-dependent at this interface? nation appears to exist at present to resolve
The effect of changing the work function of this apparent anomaly.
the substrate on the current-voltage curves
(in the dark and under illumination) has 1.7.4
been investigated for TiO2 nanocrystalline Photoexcitation and Carrier Collection:
films [230]. The onset potential for the Dynamic Behavior
photocurrent is found to be the same re-
gardless of whether SnO2 , Au, or Pt is In this section, we briefly consider the
used to support the film. A Fermi level response of nanocrystalline semiconduc-
pinned rear interface was used to explain tor–electrolyte interfaces to either pulsed
the results. or periodic photoexcitation. Several points
In general, the voltammograms for are noteworthy in this respect: (1) The
nanocrystalline electrodes are similar to photocurrent rise-time in response to an il-
what is observed for their single-crystal lumination step is nonlinear. Further, the
counterparts. An example of a pho- response is faster when the light intensity
tovoltammogram for CdS is contained is higher. (2) The decay profiles exhibit fea-
in Fig. 25. The fact that a S-type pro- tures on rather slow timescales extending
file is observed culminating in a pho- up to several seconds. (3) The photocur-
ton flux-limited plateau regime is rather rent decay transients exhibit a peaking
surprising given that (1) the film is rather behavior. The time at which this peak
2 µA
Fig. 25 Photovoltammogram
−0.8 −0.6 −0.4 −0.2 0 0.2 under interrupted illumination
of a nanocrystalline CdS-sodium
Potential sulfite electrolyte interface in the
[V vs. Ag/AgCl] reverse-bias regime.
1.8 Chemically Modified Semiconductor–Electrolyte Interfaces 41
occurs varies with the square of the film for semiconductors is that the energy band
thickness, d. (4) A similar pattern is also gap (Eg ) increases, or equivalently, the ab-
seen in IMPS data where the transit time, sorption threshold exhibits a blue shift.
τ is seen to be proportional to d 2 . The critical dimension for size quantiza-
These observations have been inter- tion effects to appear in semiconductors
preted within the framework of two dis- depends on the effective mass (m∗ ) of the
tinct models, one involving trapping or electronic charge carriers. For m∗ ∼ 0.05,
detrapping of the photogenerated elec- the critical dimension is about 300 Å; it
trons [231, 232] and the other based on decreases approximately linearly with in-
electron diffusion (or field-assisted dif- creasing m∗ [7].
fusion) not attenuated by electron local- Size quantization effects and quantum
ization [233, 234]. The millisecond transit dot photoelectrochemistry are discussed in
times also mean that the transit times more detail elsewhere in this volume.
are very long compared to equilibration
of majority carriers in a bulk semiconduc-
tor or electron-hole pair separation within 1.8
the depletion layer of a flat electrode. The Chemically Modified
slow transport is rationalized by a weak Semiconductor–Electrolyte Interfaces
driving force and by invoking percolation 1.8.1
effects [223]. Single Crystals
It is interesting that the response pat-
terns differ for different nanocrystalline Much of the research in the early
electrodes [223]. For example, while trap- 1980s on chemically modified semicon-
ping or detrapping effects appear to be ductor–electrolyte interfaces was directed
relatively unimportant for GaP, the re- toward protecting them from photocor-
sponse for TiO2 , especially at low photon rosion; this body of work has been re-
fluxes, is governed by electron trapping viewed [226]. Parallel efforts also went
or detrapping kinetics. This accounts for into improving minority carrier transfer
the faster response at higher photon fluxes at the interface by chemisorbing metal
(see preceding section). ions such as Ru3+ on the semiconductor
surface. Chemical agents such as sulfide
1.7.5
Size Quantization ions are known to passivate the semicon-
ductor against surface recombination [6].
When electronic particles such as electrons A study [22a] on electron transfer dynam-
and holes are constrained by potential ics at p-GaAs-acetonitrile interfaces where
barriers to regions of space that are com- the semiconductor surface was sulfide-
parable to or smaller than their de Broglie passivated exemplifies this fact. In general,
wavelength, the corresponding allowed en- the mechanistic issue of whether these
ergy states become discrete (i.e. quantized) chemical agents improve minority carrier
rather than continuous. This manifests charge transfer by minimizing surface re-
in the absorption (or emission) spectra combination or by a true catalytic action
as discrete lines that are reminiscent of has not been completely resolved [1].
atomic (line) transitions; these sharper Yet another tactic involves perturbing
features often appear superimposed on a the electrostatics at the semiconduc-
broader envelope. Another manifestation tor–electrolyte interface by adsorbing
Tab. 2 A summary of approachesa to chemical modification of

semiconductor–electrolyte interfaces
Modification Semiconductor(s) Modification Sample

agent(s) objectiveb Reference(s)
Ru3+ n-GaAs A 239, 240

Co3+ n-GaAs A 241
Os3+ n-GaAs A 241
Ag+ p-InP A 242
S2− p-GaAs A 236
HS− CdS, CdSe B 243
Thiolates CdS, CdSe B 244
Dithiocarbamate CdS, CdSe B 245, 246
Lewis acids CdS, CdSe B 247, 248
Lewis bases CdS, CdSe B 249
Cl− n-GaAs B 238
Benzoic acid derivatives n+ -GaAs B 250
Noble metals n-TiO2 C 251
Noble metals p-InP C 239, 252–254
RuO2 c n-CdS C 255, 256
RuO2 c n-Si C 257
Ptc p-Si C 257, 258
Ptc n-CdS C 259
Noble metals n-CdS C 254
Noble metalsd p-InP C 260
a Approaches to photoanode stabilization based on polymer films containing

redox functionalities have been reviewed elsewhere, e.g. Refs. 6 and 226.
b A: minority carrier transfer catalysis and or surface state passivation;
B: electrostatic modification; C: catalysis of multielectron photoprocesses

(refer to text).
c In these cases, the semiconductor electrode also contained a coating, either
polymeric or indium tin oxide.

d The chemically modified photocathode was used in conjunction with
n-MoSe2 (or n-WSe2 ) in a two-photoelectrode cell configuration.
(or even chemically attaching) electron course, such ‘‘fixed-charge’’ effects have
donors or acceptors on the semiconductor long been known to the solid-state device
surface [237]. In favorable cases, this physics and gas phase catalysis communi-
increases the band bending at the interface ties. Other agents that have been deployed
by thus introducing a fixed counter- for chemical tuning of the interfacial en-
charge of opposite polarity (negative for ergetics at the semiconductor–electrolyte
a n-type semiconductor) at the junc- interface are listed in Table 2.
tion. Chloride ion adsorption on the Native semiconductor surfaces are fairly
n-GaAs surface from ambient temper- inactive from a catalysis perspective.
ature AlCl3 /n-butylpyridinium chloride Thus, noble metal or metal oxide islands
melts [30, 238] is a case in point; this have been implanted on photoelectrode
process serves to improve the junction surfaces as electron storage centers to
and the photovoltage that it delivers. Of drive multielectron redox processes such
1.8 Chemically Modified Semiconductor–Electrolyte Interfaces 43
as HER, photooxidation of H2 O, and 1.8.2

photooxidation of HCl, HBr, or HI. Nanocrystalline Semiconductor Films and
Examples of this sort of chemical modifica- Composites
tion strategy are also contained in Table 2.
The advent of self-assembled mono- Dye sensitization of nanocrystalline semi-
layer (SAM) films on electrode surfaces conductor films certainly represents one
has rendered a high degree of molecular popular approach to chemical modifica-
order and predictability to the chemi- tion of the interface. However, this topic
cal modification approach. In particular, is covered in detail elsewhere in this vol-
the use of these insulating, molecular ume. Other examples, from a non–dye
spacers enables interrogation of critical sensitization perspective, are less com-
issues in electron transfer such as the mon but two recent studies are noted [268,
influence of chemical bonding and dis- 269]. One utilizes the surface affinity
tance between the support electrode and of TiO2 toward suitably derivatized vio-
the redox moieties on the rate constant logens to construct chemically modified
for electron transfer. Many such studies nanocrystalline films suitable for displays,
on gold-confined SAMs have appeared electrochromic (smart) windows, sensors,
recently [261–263]. Corresponding stud- and the like [268]. In the other study [269],
ies on semiconductor surfaces (particularly the TiO2 film surface was modified with
Group II–V compounds such as GaAs and phosphotungstic acid (PWA). This com-
InP [264–266] and elemental semiconduc- pound belongs to a family of polyoxomet-
tors such as Si [267]) have also begun to allates that exhibit interesting electron-
appear. and proton-transfer or storage properties
Alkanethiol-based or alkylsiloxane-based and also high thermal stability [269]. Thus,
SAMs have been profitably employed in all these modified films would be applicable
these instances to probe the distance effect in areas such as catalysis, sensors, elec-
in electron transfer dynamics. The thiol- tronics, and even medicine.
based SAMs have the virtue that the spacer These TiO2 -PWA films represent a
length can be predictably altered simply by logical bridge connecting single-phase
varying the number of methylene units semiconductor films and multicompo-
in the chain. The distance dependence nent composite systems. Of course, highly
of ket is embodied in the parameter β, evolved multicomponent assemblies occur
the decay coefficient. For a critical dis- in nature and there is no better exam-
cussion of the subtleties involved in the ple than the plant photosynthetic system.
extraction and interpretation of this pa- The plant photosynthetic architecture con-
rameter, we refer to Ref. 262. A value of tains synergistic components (e.g. light-
0.49 ± 0.07 has been reported for this pa- harvesting antennae, membranes) each
rameter for n-InP-alkanethiol-ferrocyanide with a well-defined and complementary
interfaces [266]. This value is smaller than function, to convert the incident pho-
its counterpart for corresponding films on ton energy, to move electrons vectorially,
gold surfaces, which range from ∼0.6 to and to store the reaction products. The
1.1 per methylene unit. The reason for this design and implementation of artificial
difference is not entirely clear, although analogs have proved to be a daunting task,
several hypotheses were advanced by the both from a synthetic and characterization
authors [266]. perspective. While this topic is covered
elsewhere in this series of volumes, we of the composite because of the high local
briefly discuss in what follows, some concentration of the substrate resulting
simple multicomponent assemblies based from the matrix adsorption process. In
on semiconductors. principle, high surface area supports of the
Early examples in the 1980s were aimed sort that are normally used in the gas-phase
at the design of composite systems for catalysis community can also be used in
photoelectrolytic generation of H2 . Thus, conjunction with TiO2 [286, 287]. These
Nafion and SiO2 were used as supports would include materials such as Al2 O3 ,
for coprecipitated ZnS and CdS for pho- SiO2 , or diatomaceous earth. The resul-
toassisted HER from aqueous sulfide me- tant composite films, however, cannot be
dia [270]. Subsequent work has addressed used as electrodes because of their poor
the mechanistic role of the support in the electronic conductivity.
photoassisted HER [271]. Vectorial elec- The second important feature of a
tron transfer was demonstrated in bipolar metal-semiconductor composite approach
TiO2 /Pt or CdSe/CoS photoelectrode pan- is that the metal can function as a
els arranged in series arrays for the template for chemical or electrochemical
photodecomposition of water to H2 and derivatization to afford a film compris-
O2 [272, 273]. ing molecular redox-semiconductor (or
More recently, matrix-semiconductor even semiconductor-semiconductor) con-
composites, that is, films comprising tacts. Figure 26 generically illustrates the
of semiconductor particles that are dis- occlusion electrosynthesis approach for
persed in a nonphotoactive continuous preparing M/TiO2 composite films and
matrix have been developed. Examples of a subsequent derivatization with ferri/
matrix candidates are metals and poly- ferrocyanide to afford the correspond-
mers [274–279]. Occlusion electrosynthe- ing metal hexacyanoferrate (MHCF)/TiO2
sis is a versatile method for preparing counterparts [288]. These chemically mod-
such composite films as exemplified by the ified films exhibit interesting ‘‘bipolar’’
Ni/TiO2 and Ni/CdS family [280–282]. photoelectrochemical behavior [289] and
Matrix-semiconductor composite films
photoelectrochromic properties [290].
have two virtues from a photoelectro-
chemical perspective. First, their com-
ponents can be separately chosen and 1.9
optimized for a specific function. Thus, Types of Photoelectrochemical Devices
the matrix component can be chosen to
have good adsorption tendency toward As Fig. 27 illustrates, there are basi-
a targeted substrate. The semiconductor cally three types of photoelectrochemi-
component then functions in the role cal devices for solar energy conversion.
of photogenerating charge carriers either The first type is regenerative in nature
for reducing or oxidizing this sequestered and the species that are photooxidized
substrate. This photocatalytic strategy has at the n-type semiconductor electrode
been recently demonstrated both for or- are simply re-reduced at the counter-
ganic substrates (methanol and formate electrode (Fig. 27a). Instead of an elec-
ion) [283, 284] and an inorganic substrate trocatalytic electrode [291, 292] where the
(sulfite) [285]. The net result in either case counterelectrode reaction occurs in the
is an enhanced photocatalytic performance dark (this is the situation schematized
1.9 Types of Photoelectrochemical Devices 45
Au
TiO2
Mn+
TiO2 occlusion
Au
M-TiO2
composite
Derivatization
Au
MHCF-TiO2
composite
Fig. 26 Schematic illustration of the occlusion electrosynthesis approach for the

preparation of M/TiO2 (M = metal) composite films and subsequent chemical
derivatization to yield the MHCF/TiO2 counterparts. Refer to the text for further
details.
in Fig. 27a), a p-type semiconductor for alternative energy sources to fossil-

photo-cathode may also be deployed in derived fuels. Thus in a landmark paper,
a tandem regenerative cell. In all these Fujishima and Honda [293] demonstrated
cases, the cells operate in the photovoltaic that sunlight could be used to drive the
mode where the input photon energy is photoelectrolysis of water using an n-TiO2
converted into electricity. photoanode and a Pt counterelectrode.
Interesting enough, it is the second type Unfortunately, the requirements for effi-
of device, namely a photoelectrolytic cell ciently splitting water are rather stringent,
(Fig. 27b), that first caught the attention as discussed elsewhere in this volume.
of a scientific and technological com- In the third type of energy conversion
munity in the 1970s that was searching device, the initial photoexcitation does not
e− Fig. 27 Types of photoelectrochemical

devices for solar energy conversion. (a),
(b), and (c) depict regenerative,
e−
e− photoelectrolytic, and dye-sensitized
hν Ox configurations, respectively. As in the
remainder of this chapter, an n-type
semiconductor is assumed in these
Red cases for specificity.
h+
(a) Semiconductor Electrolyte Metal Other variants of the three types of
device operation may be envisioned for
e− semiconductor-liquid junctions. Thus, in
the photoelectrolytic mode, the cell reac-
e− tion clearly is driven (by light) in the contra-
e−
Ox′ thermodynamic direction, that is, %G > 0.
hν
However, there are many instances, involv-
Red Red′ ing, for example, the photooxidation of
organic compounds in which light merely
h+ Ox serves to accelerate the reaction rate. Thus
(b)
these cells operate in the photocatalytic
e− mode. In fact, aqueous suspensions com-
prising irradiated semiconductor particles
D ∗/D +
e− e−
may be considered to be an assemblage of
short-circuited microelectrochemical cells
Ox
operating in the photocatalytic mode.
Finally, a storage electrode may be incor-
D +/D Red porated even in a regenerative photoelec-
D Ox trochemical cell of the sort schematized in
Fig. 27(a). Thus, when the sun is shining,
(c) D+ Red
this storage electrode is ‘‘charged’’; in the
dark, energy may be tapped (as from a bat-
occur in the semiconductor (unlike in the tery) from this storage electrode [295–298].
device counterparts in Figs. 27a and b) but Further details of these device types
occurs instead in a visible light-absorbing as well as nonenergy-related applications
dye (Fig. 27c). Subsequent injection of of photoelectrochemical cells (such as in
an electron from the photoexcited dye environmental remediation) may be found
into the semiconductor CB results in in the chapters that follow in this volume.
the flow of a current in the external
circuit. Sustained conversion of light 1.10
energy is facilitated by regeneration of the Conclusion
reduced form of the dye via a reversible
redox couple (e.g. iodide/triiodide) [294]. In this introductory chapter, we have
Therefore, this device, as its counterpart in discussed the electrostatics of the semi-
Fig. 27(a), also operates in a photovoltaic conductor–liquid interface considering
mode, or perhaps more appropriately, in a both single crystals as well as their
photogalvanic mode. nanocrystalline counterparts. The charge
1.10 Conclusion 47
transfer dynamics across both these types 9. L. M. Peter, D. Vanmaekelbergh, Adv. Elec-
of interfaces have been described in the trochem. Sci. Eng. 1999, 6, 77.
10. L. M. Peter in Comprehensive Chemical Ki-
dark and under photoexcitation of the netics, (Eds.: R. G. Compton, G. Hancock),
semiconductor. Finally, the various types Elsevier, Amsterdam, 1999, pp. 223–280.
of photoelectrochemical devices for solar 11. R. J. D. Miller, G. McLendon, A. J. Nozik
energy conversion are introduced. Sub- et al., in Surface Electron-transfer Processes,
VCH Publishers, New York, 1995.
sequent chapters in this volume provide
12. K. Rajeshwar, L. M. Peter, A. Fujishima
further elaboration of some of these topics et al., Eds., in Photoelectrochemistry Proc.,
considered herein. The Electrochemical Society, Pennington,
NJ, 1997, Vol. 20.
Acknowledgments 13. N. Sato in Electrochemistry at Metal and
Semiconductor Electrodes, Elsevier, Amster-
dam, 1998.
Research in the author’s laboratory on 14. A. Hamnett in Comprehensive Chemical
semiconductor electrochemistry and pho- Kinetics (Eds.: R. G. Compton), Elsevier,
toelectrochemistry since 1995 is funded, Amsterdam, 1987, p. 61, Vol. 27.
in part, by the Office of Basic Energy Sci- 15. J. S. Newman in Electrochemical Systems,
2nd edition, Prentice-Hall, Englewood
ences, the US Department of Energy. A
Cliffs, NJ, 1991, p. 496.
number of talented and dedicated cowork- 16. R. A. Smith in Semiconductors, Cambridge
ers and colleagues have been involved in University Press, Cambridge, 1964.
collaborative research with the author over 17. F. A. Kröger, H. J. Vink in Solid State Physics
the past twenty years; their names appear (Eds.: F. Seitz, D. Turnbull), Academic
Press, New York, 1956, Vol. 3.
in the publications cited from this labora- 18. F. A. Kröger in The Chemistry of Imperfect
tory. I also thank the University of Texas Crystals, North Holland, Amsterdam, 1964.
at Arlington for providing the facilities 19. N. B. Hannay, Ed., in Semiconductors, Rein-
and infrastructure. I am grateful to Prof. hold, New York, 1959.
20. A. R. West in Solid State Chemistry and Its
S. Licht for comments on an earlier ver-
Applications, John Wiley, New York, 1984.
sion of the manuscript. Last but not least, 21. P. A. Cox in The Electronic Structure and
I thank Ms. Gloria Madden and Ms. Rita Chemistry of Solids, Oxford University Press,
Anderson for assistance in the preparation Oxford, England, 1987.
of this chapter. 22. R. Hoffmann in Solids and Surfaces, VCH,
New York and Weinheim, 1988.
23. K. Uosaki, H. Kita in Modern Aspects of
References Electrochemistry (Eds.: R. E. White, J. O’M.
Bockris, B. E. Conway), Plenum, New York,
1. N. S. Lewis, Acc. Chem. Res. 1990, 23, 176. 1986, pp. 1–60, Vol. 18.
2. L. M. Peter, Chem. Rev. 1990, 90, 753. 24. S. Trasatti in The Absolute Electrode Potential:
3. N. S. Lewis, Annu. Rev. Phys. Chem. 1991, An Explanatory Note, IUPAC Commission
42, 543. I.3 (Electrochemistry), 1984.
4. C. A. Koval, J. N. Howard, Chem. Rev. 1992, 25. H. Reiss, J. Electrochem. Soc. 1978, 125, 937.
92, 411. 26. Yu. V. Pleskov, Yu. Ya. Gurevich in Semi-
5. A. Kumar, W. C. Á. Wilisch, N. S. Lewis, conductor Photoelectrochemistry, Consultants
Crit. Rev. Solid State Mater. Sci. 1993, 18, Bureau, New York, 1986.
327. 27. H. Gerischer, Adv. Electrochem. Eng. 1961,
6. M. X. Tan, P. E. Laibinis, S. T. Nguyen et al., 1, 139.
Prog. Inorg. Chem. 1994, 41, 21. 28. H. Gerischer in Physical Chemistry: An Ad-
7. A. J. Nozik, R. Memming, J. Phys. Chem. vanced Treatise (Eds.: H. Eyring, D. Hender-
1996, 100, 13061. son, W. Jost), Academic Press, New York,
8. N. S. Lewis, J. Phys. Chem. 1998, 102, 4843. 1970, Vol. 9A.
29. S. R. Morrison in The Chemical Physics of 54. R. A. L. Vanden Berghe, F. Cardon, W. P.

Surfaces, Plenum, New York and London, Gomes, Surf. Sci. 1973, 39, 368.
1977. 55. D. Meissner, R. Memming, Electrochim.
30. P. Singh, R. Singh, R. Gale et al., J. Appl. Acta 1992, 37, 799.
Phys. 1980, 51, 6286. 56. H. Gerischer, Ber. Bunsen-Ges. Phys. Chem.
31. K. Rajeshwar in Molten Salt Techniques 1965, 69, 578.
(Eds.: R. J. Gale, D. G. Lovering), Plenum, 57. R. Memming, G. Schwandt, Electrochim.
New York and London, 1984, pp. 221–252, Acta 1968, 13, 1299.
Vol. 2. 58. M. J. Madou, F. Cardon, W. P. Gomes, Ber.
32. R. Thapar, K. Rajeshwar, Electrochim. Acta Bunsen-Ges. Phys. Chem. 1977, 81, 1186.
1983, 28, 198. 59. H. H. Goossens, W. P. Gomes, J. Electro-
33. C. M. Gronet, N. S. Lewis, G. W. Cogan chem. Soc. 1991, 138, 1696.
et al., J. Electrochem. Soc. 1984, 131, 2873. 60. D. Vanmaekelbergh, L. S. Yun, W. P.
34. E. H. Rhoderick in Metal-Semiconductor Gomes et al., J. Electroanal. Chem. 1987,
Contacts, Clarendon Press, Oxford, 1980. 221, 187.
35. S. Kar, K. Rajeshwar, P. Singh et al., Sol. 61. J. J. Kelly, J. E. A. M. van den Meerakker,
Energy 1979, 23, 129. P. H. L. Notten et al., Philips Tech. Rev. 1988,
36. G. Horowitz, P. Allongue, H. Cachet, 44, 61.
J. Electrochem. Soc. 1984, 131, 2563. 62. P. H. L. Notten, J. E. A. M. van den Meer-
37. A. J. Bard, A. B. Bocarsly, F. R. F. Fan et al., akker, J. J. Kelly in Etching of III-V Semicon-
J. Am. Chem. Soc. 1980, 102, 3671. ductors: An Electrochemical Approach, Else-
38. A. Many, Y. Goldstein, N. B. Grover in vier, Oxford, 1991.
Semiconductor Surfaces, North Holland, Am- 63. H. H. Goossens, W. P. Gomes, Electrochim.
sterdam, 1965. Acta 1992, 37, 811.
39. G. A. Somorjai in Introduction to Surface 64. W. P. Gomes, H. H. Goossens in Adv. Elec-
Chemistry and Catalysis, John Wiley, New trochem. Sci. Eng. (Eds.: H. Gerischer,
York, 1994. C. W. Tobias), VCH, Weinheim, 1994,
40. A. W. Adamson, A. P. Gast in Physical Vol. 3.
Chemistry of Surfaces, John Wiley, New York, 65. D. Meissner, C. Sinn, R. Memming
1997. et al., in Homogeneous Photocatalysis (Eds.:
41. R. L. Van Meirhaeghe, F. Cardon, W. P. E. Pelizetti, N. Serpone, D. Reidel), Dor-
Gomes, J. Electroanal. Chem. 1985, 188, 287. drecht, 1986.
42. J. Bardeen, Phys. Rev. 1947, 71, 717. 66. J. E. A. M. van den Meerakker, J. Electro-
43. A. M. Cowley, S. M. Sze, J. Appl. Phys. 1965, anal. Chem. 1988, 243, 161.
36, 3212. 67. B. Pettinger, H-R. Schöppel, H. Gerischer,
44. S. Kurtin, T. C. McGill, C. A. Mead, Phys. Ber. Bunsen-Ges. Phys. Chem. 1976, 80, 849.
Rev. Lett. 1969, 22, 1433. 68. K. Uosaki, Trends Anal. Chem. 1990, 9, 98.
45. T. C. McGill, J. Vac. Sci. Technol. 1974, 11, 69. K. Uosaki, H. Kita, Ber. Bunsen-Ges. Phys.
935. Chem. 1984, 91, 447.
46. L. J. Brillson, Phys. Rev. Lett. 1978, 40, 260. 70. A. Manivannan, A. Fujishima, J. Lumin.
47. L. J. Brillson, J. Vac. Sci. Technol. 1979, 16, 1988, 42, 43.
1137. 71. G. Oskam, E. A. Meulenkamp, J. Elec-
48. W. E. Spicer, P. W. Chye, P. R. Skeath et al., troanal. Chem. 1992, 326, 213.
J. Vac. Sci. Technol. 1979, 16, 1422. 72. R. Reincke, R. Memming, J. Phys. Chem.
49. G. Nogami, J. Electrochem. Soc. 1982, 129, 1992, 96, 1310.
2219. 73. R. Reineke, R. Memming, J. Phys. Chem.
50. K. Uosaki, H. Kita, J. Electrochem. Soc. 1983, 1992, 96, 1317.
130, 895. 74. J. Vandermolen, W. P. Gomes, F. Cardon,
51. R. de Gruyse, W. P. Gomes, F. Cardon J. Electrochem. Soc. 1980, 127, 324.
et al., J. Electrochem. Soc. 1975, 125, 711. 75. P. Salvador, C. Gutierrez, J. Electrochem.
52. J. C. Tranchart, L. Hollan, R. Memming, Soc. 1984, 131, 326.
J. Electrochem. Soc. 1978, 125, 1185. 76. V. A. Tyagai, G. Ya. Kolbasov, Surf. Sci.
53. S. R. Morrison, Surf. Sci. 1969, 15, 363. 1971, 28, 423.
1.10 Conclusion 49
77. W. P. Gomes, F. Cardon, Ber. Bunsen-Ges. 102. Y. Rosenwaks, B. R. Thacker, R. K. Ahren-

Phys. Chem. 1970,74, 431. kiel et al., J. Phys. Chem. 1992, 96, 10096.
78. Z. Hens, J. Phys. Chem. B 1999, 103, 122. 103. H. Gerischer, J. Phys. Chem. 1991, 95, 1356.
79. Z. Hens, W. P. Gomes, J. Phys. Chem. B 104. S. M. Sze in Physics of Semiconductor De-
1999, 103, 130. vices, John Wiley, New York, 1981.
80. F. Cardon, Physica 1972, 57, 390. 105. Z. Hens, W. P. Gomes, J. Phys. Chem. B
81. G. Horowitz, J. Electroanal. Chem. 1983, 1997, 101, 5814.
159, 421. 106. J. L. Pankove in Optical Processes in Semi-
82. D. Vanmaekelbergh, Electrochim. Acta 1997, conductors, Prentice Hall, Englewood Cliffs,
42, 1135. NJ, 1971.
83. Z. Hens, W. P. Gomes, Electrochim. Acta 107. D. Vanmaekelbergh, B. H. Erne, C. W. Che-
1998, 43, 2577. ung et al., Electrochim. Acta 1995, 40, 686.
84. W. P. Gomes, D. Vanmaekelbergh, Elec- 108. I. Rosenberg, J. R. Brock, A. Heller, J. Mol.
trochim. Acta 1996, 41, 967. Catal. 1992, 96, 3423.
85. P. A. Allongue, H. Cachet, J. Electrochem. 109. J. Valladares, J. R. Bolton in Photocatalytic
Soc. 1985, 132, 45. Purification and Treatment of Water and
86. J. E. A. M. Van der Meerakker, J. J. Kelly, Air (Eds.: D. F Ollis, H. Al-Ekabi), Elsevier,
P. H. L. Notten, J. Electrochem. Soc. 1985, Amsterdam, 1993, p. 111.
132, 638. 110. M. I. Cabrera, O. M. Alfano, A. E. Cassano,
87. K. Schröder, R. Memming, Ber. Bunsen-Ges. J. Phys. Chem. 1996, 100, 20043.
Phys. Chem. 1985, 89, 385. 111. C. A. Martin, M. A. Baltanas, A. E. Casano,
88. D. Vanmaekelbergh, W. P. Gomes, Environ. Sci. Technol. 1996, 30, 2355.
F. Cardon, Ber. Bunsen-Ges. Phys. Chem. 112. K. Rajeshwar, Spectroscopy 1993, 8, 16.
113. L. M. Peter in Comprehensive Chemical Ki-
1986, 90, 431.
netics (Eds.: R. G. Compton), Elsevier, Am-
89. D. Vanmaekelbergh, W. P. Gomes,
sterdam, 1989, pp. 353–383, Vol. 29.
F. Cardon, Ber. Bunsen-Ges. Phys. Chem.
114. W. W. Gärtner, Phys. Rev. 1959, 116, 84.
1985, 89, 994.
115. M. A. Butler, J. Appl. Phys. 1977, 48, 1914.
90. D. Vanmaekelbergh, W. P. Gomes,
116. V. A. Tyagai, Russ. J. Phys. Chem. 1965, 38,
F. Cardon, J. Electrochem. Soc. 1987, 134,
1335.
891.
117. J. Li, L. M. Peter, J. Electroanal. Chem. 1984,
91. D. Vanmaekelbergh, R. P. ter Heide, 165, 41.
W. Kruijt, Ber. Bunsen-Ges. Phys. Chem. 118. L. M. Peter, J. Li, R. Peat, J. Electroanal.
1989, 93, 1103. Chem. 1984, 165, 29–41.
92. D. Vanmaekelbergh, F. Cardon, J. Phys. D: 119. R. H. Wilson, CRC Crit. Rev. Solid State
Appl. Phys. 1986, 19, 643. Mater. Sci. 1980, 10, 1.
93. D. Vanmaekelbergh, F. Cardon, Semicond. 120. R. H. Wilson, J. Appl. Phys. 1977, 48, 4297.
Sci. Technol. 1988, 3, 124. 121. K. Rajeshwar, J. Electrochem. Soc. 1980, 129,
94. J. J. Kelly, P. H. L. Notten, Electrochim. Acta 1003.
1984, 29, 589. 122. J. F. McCann, D. Haneman, J. Electrochem.
95. J. E. A. M. van den Meerakker, Electrochim. Soc. 1982, 129, 1134.
Acta 1985, 30, 435. 123. R. Memming, Surf. Sci. 1964, 1, 88.
96. A. M. Fajardo, N. S. Lewis, Science 1996, 124. J. J. Kelly, R. Memming, J. Electrochem. Soc.
274, 969. 1982, 129, 730.
97. A. M. Fajardo, N. S. Lewis, J. Phys. Chem. B 125. S. U. M. Khan, J. O’M. Bockris, J. Phys.
1997, 101, 11136. Chem. 1984, 88, 2504.
98. K. E. Pomykal, N. S. Lewis, J. Phys. Chem. B 126. F. E. Guibaly, K. Colbow, B. L. Funt, J. Appl.
1997, 101, 2476. Phys. 1981, 52, 3480.
99. A. Meier, S. S. Kocha, M. C. Hanna et al., 127. F. E. Guibaly, K. Colbow, J. Appl. Phys.
J. Phys. Chem. B 1997, 101, 7038. 1982, 53, 1737.
100. I. Uhlendorf, R. Reincke-Koch, R. Mem- 128. F. E. Guibaly, K. Colbow, J. Appl. Phys.
ming, J. Phys. Chem. 1996, 100, 4930. 1983, 54, 6488.
101. A. Meier, D. C. Selmarten, K. Siemoneit 129. J-N. Chazalviel, J. Electrochem. Soc. 1982,
et al., J. Phys. Chem. B 1999, 103, 2122. 129, 963.
130. M. Nishida, J. Appl. Phys. 1980, 51, 1669. 158. H. Morisaki, M. Hariya, K. Yazawa, Appl.
131. F. E. Guibaly, K. Colbow, Can. J. Phys. 1981, Phys. Lett. 1977, 30, 7.
59, 1682. 159. H. Minoura, M. Tsuiki, Chem. Lett. 1978,
132. D. L. Ullman, J. Electrochem. Soc. 1981, 128, 205.
1269. 160. W. Gissler, J. Electrochem. Soc. 1980, 127,
133. S. Ramakrishna, S. K. Rangarajan, 1713.
J. Electroanal. Chem. 1991, 308, 49. 161. B. Vainas, G. Hodes, J. DuBow, J. Elec-
134. W. Shockley, W. T. Read, Phys. Rev. 1952, troanal. Chem. 1981, 130, 391.
87, 835. 162. N. Müller, G. Hodes, B. Vainas, J. Elec-
135. R. N. Hall, Phys. Rev. 1982, 87, 387. troanal. Chem. 1984, 172, 155.
136. S. J. Fonash in Solar Cell Device Physics, Aca- 163. R. Tenne, W. Giriat, J. Electroanal. Chem.
demic Press, San Diego, 1981, Chapter 6, 1985, 186, 127.
pp. 262–326. 164. D. P. Amalnerkar, S. Radhakrishnan,
137. C. T. Sah, R. N. Noyce, W. Shockley, Proc. H. Minoura et al., Sol. Energy Mater. 1988,
IRE 1957, 45, 1228. 18, 37.
138. J. Reichman, Appl. Phys. Lett. 1980, 36, 574. 165. D. P. Amalnerkar, S. Radhakrishnan,
139. W. I. Albery, P. N. Bartlett, A. Hamnett H. Minoura et al., J. Electroanal. Chem.
et al., J. Electrochem. Soc. 1981, 128, 1492. 1989, 260, 433.
140. W. J. Albery, P. N. Bartlett, J. Electrochem. 166. S.-E. Lindquist, H. Vidarsson, J. Mol. Catal.
Soc. 1983, 130, 1699. 1986, 38, 131.
141. H. Reiss, J. Electrochem. Soc. 1978, 125, 937. 167. P. Singh, R. Singh, K. Rajeshwar et al.,
142. J. Thomchick, A. M. Buoncristiani, J. Appl. J. Electrochem. Soc. 1981, 128, 1145.
Phys. 1980, 51, 6265. 168. D. Lincot, J. Vedel, J. Electroanal. Chem.
1987, 220, 179.
143. W. Lorenz, M. Handschuh, J. Electroanal.
169. D. Lincot, J. Vedel, J. Phys. Chem. 1988, 92,
Chem. 1980, 111, 181.
4103.
144. W. Lorenz, C. Aegerter, M. Handschuh,
170. L. M. Peter in Photocatalysis and the Envi-
J. Electroanal. Chem. 1987, 221, 33.
ronment (Ed.: M. Schiavello), Kluwer Aca-
145. M. Handschuh, W. Lorenz, J. Electroanal.
demic Publishers, Dordrecht, 1988, p. 243,
Chem. 1988, 251, 1.
Vol. 237.
146. P. Lemasson, A. Etcheberry, J. Gautron,
171. B. Ba, H. Cachet, B. Fotouhi et al., Elec-
Electrochim. Acta 1982, 27, 607. trochim. Acta 1992, 37, 309.
147. S. Ramakrishna, S. K. Rangarajan, 172. Y-E. Sung, F. Galliard, A. J. Bard, J. Phys.
J. Electroanal. Chem. 1994, 369, 289. Chem. B 1998, 102, 9797.
148. M. L. Rosenbluth, C. M. Lieber, N. S. Lewis, 173. Y-E. Sung, A. J. Bard, J. Phys. Chem. B 1998,
Appl. Phys. Lett. 1984, 45, 423. 102, 9806.
149. S. Licht, Nature 1987, 330, 148. 174. M. P. Dare-Edwards, A. Hamnett, J. B.
150. S. Licht, D. Peramunage, Nature 1990, 345, Goodenough, J. Electroanal. Chem. 1981,
330. 119, 109.
151. S. Licht, N. Myung, J. Electrochem. Soc. 175. J. Schefold, J. Phys. Chem. 1992, 96, 8692.
1995, 142, 840. 176. K. Uosaki, H. Kita, J. Am. Chem. Soc. 1986,
152. S. Licht, N. Myung, J. Electrochem. Soc. 108, 4294.
1995, 142, 845. 177. S. Kaneko, K. Uosaki, H. Kita, J. Phys. Chem.
153. S. Licht, F. Forouzan, J. Electrochem. Soc. 1986, 90, 6654.
1995, 192, 1539. 178. K. Uosaki, Y. Shigematsu, S. Kaneko et al.,
154. S. Licht, D. Peramunage, J. Phys. Chem. B J. Phys. Chem. 1989, 93, 6521.
1996, 100, 9082. 179. M. Matsumura, S. R. Morrison, J. Electro-
155. H. J. Hovel in Semiconductors and Semimet- anal. Chem. 1983, 144, 113.
als, Solar Cells, Academic Press, New York, 180. M. Matsumura, S. R. Morrison, J. Elec-
1975, Vol. 11. troanal. Chem. 1983, 147, 157.
156. R. R. Potter, J. R. Sites, IEEE Trans. Electron 181. H. J. Lewerenz, J. Stumper, L. M. Peter,
Devices 1984, ED-31, 571. Phys. Rev. Lett. 1988, 61, 1989.
157. R. R. Potter, J. R. Sites, Appl. Phys. Lett. 182. L. M. Peter, A. N. Borazio, H. J. Lewerenz
1983, 43, 843. et al., J. Electroanal. Chem. 1990, 290, 229.
1.10 Conclusion 51
183. S. R. Morrison, T. Freund, J. Chem. Phys. 206. S. Y. Huang, L. Kavan, I. Exnar et al., J. Elec-
1967, 47, 1543. trochem. Soc. 1995, 142, L142.
184. W. P. Gomes, T. Freund, S. R. Morrison, 207. J. G. Highfield, M. Grätzel, J. Phys. Chem.
Surf. Sci. 1968, 13, 201. 1988, 92, 464.
185. W. P. Gomes, T. Freund, S. R. Morrison, 208. B. O’Regan, M. Grätzel, D. Fitzmaurice,
J. Electrochem. Soc. 1968, 115, 818. Chem. Phys. Lett. 1991, 183, 89.
186. K. Micka, H. Gerischer, J. Electroanal. 209. B. O’Regan, M. Grätzel, D. Fitzmaurice,
Chem. 1972, 38, 397. J. Phys. Chem. 1991, 95, 10525.
187. J.-S. Lee, T. Kato, A. Fujishima et al., Bull. 210. B. E. Conway, J. Electrochem. Soc. 1991, 138,
Chem. Soc. Jpn. 1984, 57, 1179. 1539.
188. G. H. Schoenmakers, D. Vanmaekelbergh, 211. R. Könenkamp, R. Henninger, P. Hoyer,
J. J. Kelly, J. Phys. Chem. 1996, 100, 3215. J. Phys. Chem. 1993, 97, 7328.
189. G. H. Schoenmakers, D. Vanmaekelbergh, 212. R. Könenkamp, R. Henninger, Appl. Phys.
J. J. Kelly, J. Chem. Soc., Faraday Trans. A 1994, 58, 87.
1997, 93, 1127. 213. I. Bedja, S. Hotchandani, P. V. Kamat,
190. M. Miyake, H. Yoneyama, H. Tamura, J. Phys. Chem. 1996, 100, 19489.
Chem. Lett. 1976, 635. 214. G. Boschloo, D. Fitzmaurice, J. Phys. Chem.
191. N. Hykaway, W. M. Sears, H. Morisaki B 1999, 103, 2228.
et al., J. Phys. Chem. 1986, 90, 6663. 215. F. Cao, G. Oskam, P. C. Searson et al.,
192. D. J. Fermin, E. A. Ponomarev, L. M. Peter, J. Phys. Chem. 1995, 99, 11974.
in Photoelectrochemistry, Proc., (Eds.: K. 216. R. F. Howe, M. Grätzel, J. Phys. Chem.
Rajeshwar, L. M. Peter, A. Fujishima et al.), 1985, 89, 4495.
The Electrochemical Society, Pennington, 217. L. A. Lyon, J. T. Hupp, J. Phys. Chem. 1995,
99, 15718.
NJ, 1997, Vol. 97–20.
218. H. Lindström, S. Södergren, A. Solbrand
193. G. Nogami, J. H. Kennedy, J. Electrochem.
et al., J. Phys. Chem. B 1997, 101, 7710.
Soc. 1989, 136, 2583.
219. H. Lindström, S. Södergren, A. Solbrand
194. W. J. Albery, N. L. Dias, C. P. Wilde, J. Elec-
et al., J. Phys. Chem. B 1997, 101, 7717.
trochem. Soc. 1987, 134, 601.
220. S. Lunell, A. Stashans, L. Ojamäe et al.,
195. P. Herrasti, L. M. Peter, J. Electroanal.
J. Am. Chem. Soc. 1997, 119, 7374.
Chem. 1991, 305, 241.
221. L. A. Lyon, J. T. Hupp, J. Phys. Chem. B
196. J. Li, L. M. Peter, J. Electroanal. Chem. 1985, 1999, 103, 4623.
182, 399. 222. R. van de Krol, A. Goossens, J. Schoonman,
197. B. H. Erné, D. Vanmaekelbergh, I. E. Ver- J. Electrochem. Soc. 1997, 144, 1723.
meir, Electrochim. Acta 1993, 38, 2559. 223. J. J. Kelly, D. Vanmaekelbergh, Electrochim.
198. J. Li, R. Peat, L. M. Peter, J. Electroanal. Acta 1998, 43, 2773.
Chem. 1986, 200, 333. 224. A. Zaban, A. Meier, B. A. Gregg, J. Phys.
199. R. Peat, L. M. Peter, Electrochim. Acta 1986, Chem. B 1997, 101, 7985.
31, 731. 225. D. Vanmaekelbergh, P. E. de Jongh, J. Phys.
200. R. Peat, L. M. Peter, J. Electroanal. Chem. Chem. B 1999, 103, 747.
1986, 209, 307. 226. H. Rensmo, H. Lindström, S. Södergren
201. D. Vanmaekelbergh, J. J. Kelly, J. Phys. et al., J. Electrochem. Soc. 1996, 143, 3173.
Chem. 1990, 94, 5406. 227. G. Hodes, I. D. J. Howell, L. M. Peter,
202. W. P. Gomes, S. Lingier, D. Vanmaekel- J. Electrochem. Soc. 1992, 139, 3136.
bergh, J. Electroanal. Chem. 1989, 269, 237. 228. A. Hagfeldt, U. Björkstén, S.-E. Lindquist,
203. R. Tenne in Semiconductor Micromachining: Sol. Energy Mater. Sol. Cells 1992, 27, 293.
Fundamentals Electrochemistry and Physics 229. S. Södergren, A. Hagfeldt, J. Olsson et al.,
(Eds.: S. A. Campbell, H.-J. Lewerenz), John J. Phys. Chem. 1994, 98, 5552.
Wiley, New York, 1994, pp. 139–175, Vol. 1. 230. A. Shiga, A. Tsujiko, T. Ide et al., J. Phys.
204. J. van de Ven, H. J. P. Nabben, J. Elec- Chem. B 1998, 102, 6049.
trochem. Soc. 1990, 137, 1603. 231. K. Schwarzburg, F. Willig, Appl. Phys. Lett.
205. G. A. Somorjai in Chemistry in Two Dimen- 1991, 58, 2520.
sions: Surfaces, Cornell University Press, 232. P. E. de Jongh, D. Vanmaekelbergh, Phys.
Ithaca, NY, 1981. Rev. Lett. 1996, 77, 3427.
233. A. Solbrand, A. Henningson, S. Södergren 260. C. Levy-Clement, A. Heller, W. A. Bonner

et al., J. Phys. Chem. B 1999, 103, 1078. et al., J. Electrochem. Soc. 1982, 129, 1701.
234. A. Solbrand, H. Lindström, H. Rensmo 261. C. E. D. Chidsey, Science 1991, 251, 919.
et al., J. Phys. Chem. B 1997, 101, 2514. 262. J. F. Smalley, S. W. Feldberg, C. E. D. Chid-
235. K. Rajeshwar, J. Appl. Electrochem. 1985, 15, sey et al., J. Phys. Chem. 1995, 99, 13141.
1. 263. S. B. Sachs, S. P. Dudek, R. P. Hsung et al.,
236. Y. Rosenwaks, B. R. Thacker, R. K. Ahren- J. Am. Chem. Soc. 1997, 119, 10563.
kiel et al., J. Phys. Chem. 1992, 96, 10096. 264. C. W. Sheen, J. X. Shi, J. Martensson et al.,
237. S. Licht, V. Marcu, J. Electroanal Chem. J. Am. Chem. Soc. 1992, 114, 1514.
1986, 210, 197. 265. Y. Gu, D. H. Waldeck, J. Phys. Chem. 1996,
238. P. Singh, K. Rajeshwar, J. Electrochem. Soc. 100, 9573.
1981, 128, 1724. 266. Y. Gu, D. H. Waldeck, J. Phys. Chem. B
239. A. Heller, Acc. Chem. Res. 1981, 14, 154. 1998, 102, 9015.
240. B. A. Parkinson, A. Heller, B. Miller, Appl. 267. A. Haran, D. H. Waldeck, R. Naaman et al.,
Phys. Lett. 1978, 33, 521. Science 1994, 263, 948.
241. B. J. Tufts, I. L. Abrahams, L. G. Casa- 268. A. Hagfeldt, L. Walder, M. Grätzel, Proc.
grande et al., J. Phys. Chem. 1989, 93, 3260. Soc. Photo-Opt. Instrum. Eng. 1995, 2531, 60.
242. A. Heller, H. J. Leamy, B. Miller et al., 269. H. Lindstrom, H. Rensmo, S.-E. Lindquist
J. Phys. Chem. 1983, 87, 3239. et al., Thin Solid Films 1998, 323, 141.
243. K. Uosaki, Y. Shigematsu, H. Kita et al., 270. N. Kakuta, K. H. Park, M. F. Finlayson et al.,
Anal. Chem. 1989, 61, 1980. J. Phys. Chem. 1985, 89, 732.
244. M. J. Natan, J. W. Thackeray, M. S. 271. A. Ueno, N. Kakuta, K. H. Park et al.,
Wrighton, J. Phys. Chem. 1986, 90, 4089. J. Phys. Chem. 1985, 89, 3828.
245. J. W. Thackeray, M. J. Natan, P. Ng et al., 272. E. S. Smotkin, A. J. Bard, A. Campion et al.,
J. Am. Chem. Soc. 1986, 108, 3570. J. Phys. Chem. 1986, 90, 4604.
246. J. J. Hickman, M. S. Wrighton, J. Am. 273. E. S. Smotkin, S. Cervera-March, A. J. Bard
Chem. Soc. 1991, 113, 4440. et al., J. Phys. Chem. 1987, 91, 6.
247. C. J. Murphy, A. B. Ellis, J. Phys. Chem. 274. S. Kuwabata, N. Takahashi, S. Hirao et al.,
1990, 94, 3082. Chem. Mater. 1993, 5, 437.
248. J. Z. Zhang, A. B. Ellis, J. Phys. Chem. 1992, 275. C. S. C. Bose, K. Rajeshwar, J. Electroanal.
96, 2700. Chem. 1992, 333, 235.
249. C. J. Murphy, G. C. Lisensky, L. K. Leung 276. Y. Son, N. R. de Tacconi, K. Rajeshwar,
et al., J. Am. Chem. Soc. 1990, 112, 8344. J. Electroanal. Chem. 1993, 345, 135.
250. S. Bastide, R. Butruille, D. Cahen et al., 277. N. R. de Tacconi, Y. Son, K. Rajeshwar,
J. Phys. Chem. B 1997, 101, 2678. J. Phys. Chem. 1993, 97, 1042.
251. C -M. Wang, A. Heller, H. Gerischer, J. Am. 278. N. R. Avvaru, N. R. de Tacconi, K. Rajesh-
Chem. Soc. 1992, 114, 5230. war, Analyst 1998, 123, 113.
252. E. Aharon-Shalom, A. Heller, J. Electrochem. 279. S. Ito, T. Deguchi, K. Imai et al., Elec-
Soc. 1982, 129, 2865. trochem. Solid State Lett. 1999, 2, 440.
253. A. Heller, E. Aharon-Shalom, W. A. Bonner 280. M. Zhou, W-Y. Lin, N. R. de Tacconi et al.,
et al., J. Am. Chem. Soc. 1982, 104, 6942. J. Electroanal. Chem. 1996, 402, 221.
254. D. E. Aspnes, A. Heller, J. Phys. Chem. 281. M. Zhou, N. R. de Tacconi, K. Rajeshwar,
1983, 87, 4919. J. Electroanal. Chem. 1997, 421, 111.
255. K. Rajeshwar, M. Kaneko, A. Yamada et al., 282. N. R. de Tacconi, H. Wenren, K. Rajeshwar,
J. Phys. Chem. 1985, 89, 806. J. Electrochem. Soc. 1997, 144, 3159.
256. K. Rajeshwar, M. Kaneko, J. Phys. Chem. 283. N. R. de Tacconi, H. Wenren, D. McChes-
1985, 89, 3587. ney et al., Langmuir 1998, 14, 2933.
257. L. Thompson, J. DuBow, K. Rajeshwar, 284. N. R. de Tacconi, J. Carmona, K. Rajeshwar,
J. Electrochem. Soc. 1982, 129, 1934. to be published.
258. G. Hodes, L. Thompson, J. DuBow et al., 285. N. R. de Tacconi, M. Mrkic, K. Rajeshwar,
J. Am. Chem. Soc. 1983, 105, 324. Langmuir 2000, 16, 8426.
259. A. W.-H. Mau, C-B. Huang, N. Kakuta et al., 286. C. Anderson, A. J. Bard, J. Phys. Chem. B
J. Am. Chem. Soc. 1984, 106, 6537. 1997, 101, 2611.
1.10 Conclusion 53
287. N. Takeda, M. Ohtani, T. Torimoto et al., 293. A. Fujishima, K. Honda, Nature 1972, 238,
J. Phys. Chem. B 1997, 101, 2644. 37.
288. N. R. de Tacconi, K. Rajeshwar, R. O. Lezna, 294. B. O’Regan, M. Grätzel, Nature 1991, 353,
Electrochim. Acta 2000, 45, 3403. 737.
289. N. R. de Tacconi, J. Carmona, K. Rajeshwar, 295. J. Manassen, G. Hodes, D. Cahen,
J. Phys. Chem. B 1998, 101, 10151. J. Electrochem. Soc. 1977, 124, 532.
290. N. R. de Tacconi, J. Carmona, W. L. Balsam 296. S. Licht, G. Hodes, R. Tenne et al., Nature
et al., Chem. Mater. 1998, 10, 25. 1987, 326, 863.
291. G. Hodes, J. Manassen, D. Cahen, J. Appl. 297. S. Licht, B. Wang, T. Soga et al., Appl. Phys.
Electrochem. 1977, 7, 182. Lett. 1999, 74, 4055.
292. G. Hodes, J. Manassen, D. Cahen, J. Elec- 298. S. Licht, J. Phys. Chem. B 2001, 105,
trochem. Soc. 1977, 127, 544. 6281.
59
2.1 However, two developments in the 1970s

Photoelectrochemical Systems led to a huge revival of interest and
Characterization to the ‘‘birth’’ of photoelectrochemistry.
John J. Kelly, Zeger Hens and In 1971, Fujishima and Honda showed
Daniel Vanmaekelbergh that it was possible to photolyze water
Utrecht University, Utrecht, The Netherlands in an electrochemical cell with a TiO2
photoanode and a Pt cathode without an
Zeger Hensalso external source [4]. Subsequently, a num-
Laboratorium voor Fysische Chemie, Gent, ber of groups [5–8] described regenerative
Belgium photoelectrochemical solar cells based on
2.1.1
narrow band gap semiconductors and a
Introduction redox couple. They showed that with the
proper choice of reducing species it was
This chapter focuses on the characteri- possible to achieve stability of the pho-
zation of photoelectrochemical systems. toanode during operation of the cell over
Before the essential features of such sys- longer periods. In further studies, Licht
tems are described in Sect. 2.1.2, the scope and coworkers developed efficient systems
of the chapter is first defined on the basis that couple solar energy conversion with
of a brief history of the field. energy storage and show improved en-
ergy conversion and stability [9–14]. At the
2.1.1.1 Scope of the Chapter same time, the growth of the optoelec-
Semiconductor electrochemistry devel- tronics industry, based largely on III–V
oped as a discipline with the development materials, revealed a need for new process-
of semiconductor device technology [1–3]. ing methods [15], including etching and
Wet chemical processing including etch- metallization. Wet chemical and electro-
ing was important for the fabrication chemical methods proved very successful
of early silicon and germanium devices. in this field. In addition, electrochemical
With the increasing sophistication and methods could be used in a simple way
miniaturization in silicon technology, wet to make and characterize materials for a
processes were replaced by dry, mainly whole range of applications. The work de-
physical, methods. As a result, inter- scribed so far relates to single crystalline
est in semiconductor electrochemistry and, to a lesser extent, polycrystalline ma-
declined. terials.
60 2 Experimental Techniques
Interest in porous photoelectrochemical obvious way to characterize a photoelec-

systems was stimulated by the report in trochemical system is by electrochem-
1991 of O’Regan and Grätzel [16] of a ical methods, and this approach will
novel solar cell based on a dye-sensitized constitute the focus of the chapter. A
nanoparticulate TiO2 photoanode. This brief and very general introduction to
work raised a whole range of interesting (photo)electrochemical characterization is
fundamental issues with regard to charge given in Sect. 2.1.2. This is not intended to
carrier dynamics, transport and interfacial be a conclusive review; for this a whole vol-
transfer in porous semiconductor matrices ume would be required. The application of
permeated by an electrolyte solution. The these methods to the three systems is de-
discovery in 1992 [17] of the striking scribed in the subsequent sections. Where
optical properties of nanoporous silicon, relevant, other nonelectrochemical meth-
obtained by (photo)anodic etching, led to ods of characterization are mentioned.
a reappraisal of the photoelectrochemistry
of this material and to studies of other 2.1.1.2 Special Features of
Photoelectrochemical Systems
porous crystalline semiconductors [18].
Photoelectrochemical systems rely on the
Research on size quantization in col-
properties of a semiconductor electrode
loidal systems was responsible for the
or of an electrode provided with a ‘‘sensi-
development by Bard and coworkers of a
tizer’’ layer consisting of, for example, dye
new field in photoelectrochemistry, that of
molecules or quantum dots. This working
nanodot electrodes [19, 20]. An ordered or
electrode (WE) forms part of an electro-
disordered monolayer or sub-monolayer
chemical cell that also contains a counter
of nanometer-sized semiconductor parti-
electrode (CE) and, in many cases, a refer-
cles is attached to a conducting substrate
ence electrode (RE). Absorption of light by
either directly or via a self-assembled or- the semiconductor or the sensitizer layer
ganic monolayer. The monolayer acts as gives rise to a photocurrent (PC) and/or a
a spacer, allowing the distance between photovoltage, which can be measured in
the dot and the substrate to be varied. the external circuit. Conversely, the pas-
Absorption of light by the semiconduc- sage of current through the interface of
tor dots gives rise to processes similar the working electrode with the electrolyte
in many ways to those observed in bulk solution can lead to light emission. In
electrodes. However, because of the size this section, some important aspects of
quantization and the distinctive electrode such photoelectrochemical systems are re-
geometry, striking new effects are found. viewed. These topics are dealt with later in
In this chapter, the characterization the chapter.
of the three types of systems described When a semiconductor is brought into
earlier have been considered: single crys- electrical contact with an electrolyte solu-
tal and polycrystalline bulk electrodes tion containing a redox couple (Red/Ox+ ),
(Sect. 2.1.3), dye-sensitized and quantum equilibrium is established by exchange of
dot electrodes (Sect. 2.1.4), and macro- charge between the two phases [21–23].
porous and nanoporous semiconductors Figure 1(a) shows an example of an n-type
(Sect. 2.1.5). Some essential features, com- semiconductor with a redox couple whose
mon to all these systems, are intro- Fermi energy (which is related to the re-
duced in the following subsection. The dox potential VRed/Ox ) is located in the
2.1 Photoelectrochemical Systems Characterization 61
e φSC EF,n
Energy
EF,n ERed / Ox ERed / Ox
Distance
(a) (b)
Fig. 1 A scheme of the energetics at an n-type semiconductor
electrode in contact with a redox system in an electrolyte solution.
(a) The situation under conditions of electronic equilibrium. The
electrochemical potential of the electrons is the same in both
phases, i.e. the electron Fermi-level in the semiconductor EF,n has
the same value as the Fermi-level of the electrons in the redox
system ERed/Ox . (b) Case in which the energy bands in the
semiconductor are flat; this situation, corresponding to maximum
photovoltage, is reached under strong illumination at open circuit.
WSC is the width of the depletion layer and eφSC is the
band-bending.
band gap of the solid. Electrons have been oxidizing the reduced species
transferred from the conduction band (CB)
of the semiconductor to solution creating Red + hVB + −−−→ Ox+ (1)
a space charge layer (a depletion layer
of width WSC ) within the semiconductor The electrons are detected as photocurrent
with a band bending eφSC . In equilibrium, in the external circuit. This ‘‘short-circuit
the Fermi level is constant throughout the photocurrent’’ depends on the competition
system. There are various ways in which between hole transfer across the interface
equilibrium can be disturbed, for example, (the Faraday reaction) and electron-hole
by illumination, by an externally applied recombination in the bulk semiconductor
potential, or by charge carrier injection and at the surface of the electrode. Under
from solution. open-circuit conditions no photocurrent
A photovoltaic (two-electrode) cell oper- can flow in the external circuit. The
ates without an external voltage source [22, Fermi level in the semiconductor is
23]. The working electrode is illuminated. raised with respect to that in solution
Two limiting cases can be distinguished. and a photovoltage is established. In the
If the electrodes are short-circuited, then limiting case at high light intensity, the
the electrons and holes, generated by energy bands become flat (Fig. 1b). The
supra–band gap light, are separated by mi- maximum photovoltage is determined by
gration within the depletion layer and by the difference in Fermi level under flat
diffusion. The holes react at the interface, band and equilibrium conditions. Clearly,
such a system can supply electrical power into a p-type semiconductor can give rise
(Sect. 2.1.2.2). to light emission [25].
Unlike photovoltaic cells, most photo- Macroporous semiconductor electrodes
electrochemical systems do not operate resemble in many respects the bulk elec-
without an external source. In a more trodes described earlier. However, there
general approach, the potential of the are clear differences between porous and
working electrode can be varied with re- nonporous systems that become even
spect to the equilibrium potential Veq (or more pronounced in nanoporous elec-
to the potential of a RE) by means of an trodes. Porous systems are considered in
external voltage source (e.g. a potentiostat) Sect. 2.1.5.
connected between WE and CE [21, 24]. Recently, insulating nanocrystals have
The current density (indicated here by j ) been attached to conducting substrates
is measured both in the dark and under (metal or indium tin oxide) by van der
illumination as a function of the applied Waals interactions or covalent bonding.
potential V . This is described in more When such a system is used as a working
detail in Sect. 2.1.2.2. electrode in a photoelectrochemical cell,
A strong oxidizing agent (OxS + ), that a photocurrent (in the µA range) can be
is one with electron acceptor levels cor- measured [19, 20, 26]. Clearly, a monolayer
responding to the valence band of the of nanocrystals can absorb a sufficient
semiconductor, can extract electrons from fraction of the incident light to give to a
the band thus creating holes measurable photocurrent. The absorption
of a photon in a nanocrystal leads to a
OxS + −−−→ RedS + hVB + (2)
photoexcited state; an electron is promoted
In an n-type semiconductor, these holes from a lower energy level (HOMO or
can recombine with majority carriers (elec- valence band) to a higher energy level
trons) from the CB [25]. Electrolumines- (LUMO or CB) (Fig. 2). A photocurrent
cence (EL) is expected from a semiconduc- will be observed if the electron and
tor with a direct band gap for conditions in hole are separated effectively before the
which the surface electron concentration system relaxes to the ground state. In
is significant. Similarly, electron injection Fig. 2, the electron is transferred to the
µe Fig. 2 Scheme representing the general

principle of a photoelectrochemical
system. Electrons are photoexcited in
Energy
the absorber. The electron and hole are

selectively transferred to an electron
conductor (usually a metal or a
µh semiconductor) and to a hole conductor
(a redox system in a liquid electrolyte).
The photovoltage is the difference
between the electrochemical potentials
Electron Light Hole in the electron and hole conducting
conductor absorber conductor phases; thus µe − µh .
metal, the hole to an oxidizable species in mechanisms two complex systems are
solution. This leads to an upward shift considered. Finally, charge carrier trans-
of the electrochemical potential of the port in porous semiconductor systems is a
electrons µe and a downward shift of topic deserving special attention.
the electrochemical potential of the holes
µh (and thus to a photovoltage Vphoto = 2.1.2
(µe − µh /e). Such a photoelectrochemical Photoelectrochemical Characterization
Methods
system shows a strong analogy with
molecular (e.g. dye) systems adsorbed on
For scientific research on photoelectro-
metal or semiconductor electrodes. chemical systems, the photoelectrode is
The photoelectrochemical solar cell of generally the working electrode in a three-
O’Regan and Grätzel [16] combines the electrode electrochemical cell. Using this
special features of semiconductor poros- set up, the system can be investigated by
ity and sensitization described earlier. perturbing it and recording the system’s
This system consists of nanometer-sized response. In general, such perturbation-
TiO2 crystallites interconnected to form a response methods may be classified ac-
three-dimensional porous assembly. Dye cording to the time dependence of the
molecules acting as the light absorber are perturbation (steady state versus time-
chemisorbed on the internal surface. The resolved) or to the physical nature of
efficiency of light absorbance is close to the perturbation and the response. In
unity as a result of the fact that photons this section, the principles of steady state
encounter an adsorbed dye molecule many and time-resolved methods in general are
times. The electrolyte solution, containing first discussed. Later on, two different
an I − /I3 − redox system, permeates the perturbation-response techniques used to
pores of the system. The TiO2 assembly study semiconductor-based photoelectro-
acts as an electron conductor and the elec- chemical systems are discussed. Finally,
trolyte as the ‘‘hole conductor’’. Because methods in which illumination or minority
of the interpenetration of the electron and carrier injection gives rise to light emission
hole conducting phases on a nanometer from the electrode are considered.
scale, back transfer of the electron to I3 −
is the main source of recombination. 2.1.2.1 Steady State and Time-resolved
The three systems described earlier have Methods
a number of features in common. To ob- Steady state methods are basically simple:
serve photocurrent, light obviously has one applies a time-independent pertur-
to be absorbed by the system. The spa- bation x to a system and records – if
tial separation of photogenerated electrons it exists – the time-independent response
and holes must be more effective than y(x) of the system. This may be repeated
their recombination. This requires effi- for different levels of the perturbation. In
cient kinetics for electron transfer across this way, a functional relation between
the solid/solution interface. These are perturbation and response is determined,
topics that are addressed in the remain- which should provide information on the
der of the chapter. Electrode reactions system studied. Consider, for example, an
in photoelectrochemical systems may be electrical resistor through which a current
quite complicated. To illustrate the gen- is passed (perturbation). Instantaneously,
eral approach to the elucidation of reaction the system responds by maintaining a
potential difference across the resistor. The Time-resolved measurements on linear

linear relation between both quantities al- systems may be represented in many ways.
lows the characterization of the system. For example, in the Nyquist representa-
A major drawback of steady state tech- tion, the transfer function H (ω) is plotted
niques is that no information is obtained as a point in a two-dimensional plane
on the dynamic properties of a system. having coordinates [Re(H ) and Im(H )],
Take for instance an electronic circuit, for each frequency measured. In specific
which simulates a simple resistor except cases, the transfer function may be rep-
that the potential difference is established resented also by an equivalent electrical
between its poles after some time delay. circuit. This is a combination of lumped
Clearly, this circuit would emerge from a circuit elements (resistor, capacitor, etc.)
steady state analysis as a simple resistor: having the same perturbation-response be-
the system dynamics, which is a basic char- havior as the system studied.
acteristic of any system, can be revealed In the case of a nonlinear system, a
only by recording the response from the similar approach using harmonic pertur-
moment the perturbation is applied (time- bations is possible if a ‘‘small-signal’’ per-
resolved measurements). turbation x(t) = Re[#X(ω)! exp(iωt)], su-
Many time-resolved methods do not perimposed on a time-independent ‘‘bias’’
record the transient response as outlined perturbation, is applied to the system. If
in the earlier example. In the case of the signal level of the perturbation is suf-
linear systems, all information on the ficiently small, a linear dependence of
dynamics may be obtained by using the response on the perturbation can be
sinusoidally varying perturbations x(t) achieved (i.e. y(t) = Re[#Y !(ω) exp(iωt)]).
(harmonic modulation techniques) [27], a Clearly, the transfer function defined in
method far less sensitive to noise. In Eq. (3a) becomes a differential quantity:
this section, the complex representation !(ω)
#Y
of sinusoidally varying signals is used, H (ω) = (3b)
!
#X(ω)
!
that √is, x(t) = Re[X(ω) exp(iωt)], where
i = −1. The quantity X(ω) ! contains the At low frequencies, H (ω) is equal to the
amplitude and the phase information of slope ∂y/∂x of the steady state response
the sinusoidal signal, whereas the complex y(x). The time-resolved electrochemical
exponential exp(iωt) expresses the time techniques discussed in Sects. 2.1.2.2 and
dependence. A harmonically perturbed 2.1.2.3 pertain to this class of small-signal
linear system has a response that is – after modulation techniques.
a certain transition time – also harmonic, In general, measuring the transfer
differing from the perturbation only by function of a system under study us-
its amplitude and phase (i.e. y(t) = ing harmonic modulation techniques is
Re[Y!(ω) exp(iωt)]). In this case, all the straightforward. Interpreting the experi-
information on the dynamics of the system mental data, however, is not. As will be
is contained in its transfer function H (ω), demonstrated by experimental examples
which is a complex function of the angular in Sects. 2.1.3 and 2.1.5, time-resolved
frequency, defined as [27, 28] methods become most powerful if the
experimental impedance can be analyzed
!(ω)
Y using a dynamic model for the system
H (ω) = (3a)
!
X(ω) studied.
2.1.2.2 Current Density Versus Potential electron concentration at the surface. As a

Techniques result, a net cathodic current flows across
Because most applications of (photo)elec- the interface due to the reduction of the
trochemical systems involve the transfer of oxidized species
electrons across an interface (Sect. 2.1.1),
current density-potential techniques are Ox+ + eCB − −−−→ Red (4)
commonly used in (photo)electrochemis-
try. In this case, the difference in elec- Here, we assume that electron transfer
trochemical potential of electrons across only occurs via the CB and not via sur-
the interface of interest (accessible via face states. As in a Schottky diode, j
the working electrode – reference elec- generally increases exponentially with (de-
trode potential difference) and the current creasing) potential (Fig. 3a). The form of
density through this interface are used the dark current-potential curve, however,
as the perturbation and the response (or depends on the mechanism and kinetics
vice versa). Two approaches can be distin- of the charge-transfer reaction. At high
guished. When (quasi) steady state signals overpotential, corresponding to a large
are used, one speaks of current density deviation from equilibrium, the reaction
versus potential measurements whereas expressed by Eq. (4) may become limited
harmonically modulated signals, superim- by mass transport in solution, that is,
posed on a bias, are involved in electro- the cathodic current becomes potential-
chemical impedance spectroscopy (EIS). independent (this is not shown in Fig. 3).
We introduce these two approaches on the If the potential of the electrode of Fig. 1
basis of the kinetics of the simple system is made more positive than Veq , the band-
shown in Fig. 1. bending increases with respect to that at
If the potential of the n-type electrode,
whose energy band diagram is shown in
j
Fig. 1(a), is made negative with respect to
the equilibrium potential Veq , then, the b
band-bending eφSC decreases until finally
flat band condition (V = Vfb ) is reached a
(Fig. 1b). If the potential is made more
V vs Veq
negative than the flat band value (V <
Vfb ), then majority carrier ‘‘accumulation’’
occurs at the surface. The decrease in
band-bending on going from Figs. 1(a) to
(a)
1(b) is accompanied by an increase in the
sc
Fig. 3 (a) Schematic representation of j PH
the current-potential curves for an
n-type semiconductor (see Fig. 1) in the m
j PH
dark (a) and under illumination with
supra-band gap light (b). (b) The part of
curve (b) relevant to photovoltaic V oc
PH
applications. The open-circuit
photovoltage VPHoc and the short-circuit
sc
photocurrent jPH are indicated, as is the m
V PH
m and V m .
rectangle defining jPH (b)
PH
equilibrium and strong depletion, or even (Lmin ) and on α (which is a function

inversion, may result. This corresponds to of the wavelength). Light absorption just
the blocking current range of the diode above the band edge depends strongly on
(V > Veq ). A small potential-independent whether the semiconductor has a direct
anodic current results from electron injec- band gap (for which α is large and the
tion from the reduced species across the penetration depth is small) or an indirect
barrier into the conduction band. band gap (for which α is small and 1/α is
Under illumination, a photocurrent is large). The importance of such factors is
observed under depletion conditions if illustrated in Sect. 2.1.5.2.
the photogenerated electron and hole are Recombination of electrons and holes
spatially separated before recombination via surface states competes with hole
can occur. The hole reacts at the surface transfer to solution and thus reduces, and
(Eq. 1) and the electron is collected at the may even suppress, the photocurrent. To
counter electrode. For the simplest case obtain a limiting photocurrent, a stronger
in which no recombination occurs at the band-bending is required. Obviously, more
surface, Gärtner [29] derived an expression extended models are needed to describe
for the photocurrent density jPH taking the photocurrent-potential characteristics
into account the absorbed photon flux in this ‘‘less favorable’’ case [24].
%, the absorption coefficient α, and the A schematic curve for the dependence
minority carrier diffusion length Lmin of the total current density (dark current +
" # PC) on the potential is shown in Fig. 3(a).
exp(−αWSC ) It is clear that the form of this curve on the
jPH = e% 1 − (5a)
1 + αLmin anodic side will depend on the parameters
as discussed earlier. The part of the
The potential dependence of the photocur- curve relevant to photovoltaic applications
rent is determined in the model by the is shown in Fig. 3(b). The short-circuit
dependence of the depletion layer thick- photocurrent density jPH sc and the open-
ness on the band-bending oc
circuit photovoltage VPH are indicated. It
$ %1/2 should be noted that this part of the curve
2εε0 φSC
WSC = (5b) can be mapped without an external voltage
eND source by measuring the photocurrent
where ND is the donor density, ε is the through and the voltage across a load
dielectric constant of the semiconductor, resistor (between WE and CE) whose
and φSC equals V − Vfb (see Sect. 2.1.3.1). resistance is varied from zero to a very
In essence, this simple model states that large value. The conversion efficiency η of
an electron/hole pair will contribute to a photovoltaic cell is defined as η = Pm /Po
the photocurrent if it is generated within where Po is the power of the incident
a distance from the electrolyte interface radiation and Pm , the maximum power
within which the electron and hole can be output of the cell, is given by
separated by migration and diffusion, that m
Pm = jPH m
× VPH (6a)
is, the penetration depth of the light 1/α
is less than Lmin + WSC . Apart from the The values of jPHm and V m depend
PH
applied potential (which determines WSC ), on the shape of the current-potential
the efficiency of charge separation depends curve under illumination (i.e. on both the
on the quality of the semiconductor dark current and photocurrent-potential
characteristics). For a given temperature defined as the transfer function relating a

m and V m are chosen
and light intensity jPH small signal variation of the working elec-
PH
to give a rectangle of maximum area in the trode potential and of the current density
current-potential plot (see shaded area in through the working electrode/electrolyte
Fig. 3b). An important parameter for the interface:
characterization of a solar cell is the fill
#j˜(ω)
factor (FF) defined as YEC (ω) = (7)
!(ω)
#V
sc V oc
jPH PH
FF = m Vm (6b) In the case of a semiconductor-based pho-
jPH PH toelectrochemical system, the measure-
In steady state measurements, one gener- ment of the electrochemical admittance
ally applies a fixed potential to the working serves two purposes. As is explained in
electrode and measures the steady state Sect. 2.1.3.1, it allows on the one hand
current through the cell. Alternatively, the the in situ determination of the energetics
potential is scanned at a fixed rate and the of the (bulk) semiconductor surface. On
current is measured continuously. As in the other hand, it makes the dynamics of
metal electrochemistry, information about various (photo)electrochemical processes
the kinetics of surface reactions and the experimentally accessible. Clearly, EIS is
hydrodynamics of the system can be ob- also possible using an illuminated semi-
tained by varying the potential scan rate. conductor, an experimental method some-
The rotating disk electrode is a useful tool times referred to as PEIS. Finally, it should
for studying the role of mass transport. be noted that although the electrochemical
The rotating ring-disk electrode (RRDE) admittance is determined experimentally
has two important applications in semi- (the applied electrode potential is used
conductor electrochemistry. In studies of as the perturbation), the electrochemical
competitive reactions at the semiconductor impedance is generally plotted as the result
disk, the reaction products can be analyzed of an EIS measurement.
at the ring of the RRDE (see Sect. 2.1.3.3).
Alternatively, reactive species can be gener- 2.1.2.3 PC versus Light Intensity
ated at the metal disk of a RRDE and their Techniques
electrochemistry studied at the semicon- Typical to the study of photoelectrochem-
ducting ring [25]. Important parameters ical systems are measurement techniques
in photoelectrochemistry are the intensity that use the light flux % incident on the
and the spectral distribution of the light working electrode as a perturbation and
used for photocurrent and photovoltage the resulting PC jPH as the system’s re-
studies. In photovoltaics, steady state mea- sponse [30]. In this case, the electrode
surements can be used to obtain FF and potential can be used as an additional ex-
solar energy conversion efficiencies [24]. perimental variable. Clearly, the incident
Such results may give insight into reaction light flux cannot be quantified directly.
mechanisms. Therefore, a reference signal proportional
From the earlier discussion, it is clear to this light flux is generally used, for exam-
that the j versus the V relation of an elec- ple, the voltage generated by the light flux
trochemical system is nonlinear. There- on a reference photodiode. This technique,
fore, the electrochemical admittance (or its which for steady state conditions has no
inverse, the electrochemical impedance) is particular name, has been applied several
times with success, for example, to identify by illumination are effectively separated
(light-intensity dependent) photocurrent by migration and diffusion (Fig. 4). PL,
multiplication processes (see Sect. 2.1.3.3). on the other hand, is expected when the
The optoelectrical transfer function relates photogenerated carriers recombine radia-
an incident sinusoidally modulated light tively, that is, at potentials approaching the
flux and the resulting modulated pho- flat band potential Vfb (see Sect. 2.1.3.1).
tocurrent density. This quantity is defined Emission may result either from direct
as [30] band-band recombination or from indirect
recombination via a band gap state [25]. In
#j˜PH (ω) the simplest case in which surface recom-
YOE (ω) = (8)
!
e#%(ω) bination can be disregarded, the potential
dependence of the emission intensity IPL
Again, it should be stressed that the opto- is described by the Gärtner equation:
electrical transfer function is defined using " #
small-signal perturbations, superimposed exp(−αWSC )
IPL = κ% (9)
on time-independent bias signals. Mea- 1 + αLmin
surement of the optoelectrical impedance
where % is the incident photon flux and κ
is often referred to as intensity modulated
is the ratio of the rate of radiative recombi-
photocurrent spectroscopy (IMPS).
nation to the total recombination rate. In
principle, it is possible to obtain values for
2.1.2.4 Luminescence-based Techniques Lmin , WSC , and α from the potential depen-
Photoluminescence (PL), like photocur- dence of IPL . The Gärtner model has been
rent, is a technique in which a light flux extended by Gerischer and coworkers [31,
incident on the working electrode acts as a 32] to account for surface recombination.
perturbation. In this case, the response of In this case, surface recombination rates
the system is followed by measuring the in- can be obtained from photoluminescence
tensity of the emitted light. As in photocur- measurements.
rent measurements, the electrode potential While photoluminescence has been
can be used as an additional variable. mainly studied under steady state con-
As explained in Sect. 2.1.2.2, photocur- ditions or with transient techniques,
rent is obtained when, under depletion harmonic-modulation measurements are
conditions, electrons and holes created possible. Beckmann and Memming [33]
PL PC
Intensity /Current
Potential Fig. 4 Schematic representation of the

− potential dependence of the PC and the
− PL intensity of an n-type semiconductor
in an indifferent electrolyte solution.
Energy band diagrams are shown for the
+ illuminated semiconductor under flat
+ band and depletion conditions.
studied the photoluminescence of n-type complicated potential-dependence points

GaP by perturbing the system with a small to changes in surface chemistry that influ-
sinusoidal modulation of the potential dur- ence nonradiative surface recombination.
ing a potential scan and measuring the The observation of electroluminescence in
response with lock-in techniques. an electrochemical system clearly shows
EL is observed when minority carriers, the involvement of minority carriers in the
injected into a semiconductor from solu- charge-transfer processes.
tion, recombine radiatively with majority As far as we are aware, experiments
carriers [25]. This is illustrated schemati- on electroluminescence involving a sinu-
cally in Fig. 5 for an n-type semiconductor. soidal perturbation of the applied potential
At positive potentials, the surface electron have not been performed but large-signal
concentration is very low. Holes injected potential step and pulse measurements
by the oxidizing agent in solution are held have been reported. We shall return briefly
at the electrode surface by the electric to these aspects in Sect. 2.1.3.2.5.
field of the depletion layer. The surface
holes generally cause oxidation and dis- 2.1.3
solution of the semiconductor. Clearly, in Bulk Systems
this potential range neither current nor
light emission is observed. As the potential In this section, we consider some im-
is made negative and the band-bending de- portant aspects of photoelectrochemical
creases, the injected holes recombine with systems based on single-crystal and poly-
electrons supplied via the external circuit. crystalline electrodes. In Sect. 2.1.3.1, the
This results in a cathodic current and, if re- use of EIS is described for the determina-
combination is radiative, in light emission. tion of the energetics of the semiconduc-
In general, the cathodic current becomes tor/electrolyte interface. Section 2.1.3.2
potential-independent at negative poten- deals with the dynamics of electron/hole
tials as a result of mass-transport limita- recombination at the semiconductor sur-
tions in solution. The emission intensity face and the importance of electrochemical
should, therefore, become constant. How- and optoelectrical impedance techniques
ever, this is often not the case. A more for such studies. Finally, a somewhat
Intensity/Current
EL
Potential
Fig. 5 Schematic representation of the Current

potential dependence of the cathodic
current and the EL intensity of an n-type −
semiconductor in a solution containing
an oxidizing agent that injects holes into
the valence band of the solid. Schematic
energy band diagrams are shown for flat +
band and depletion conditions. +
arbitrary choice of complex reactions is be calculated from the electrochemical ad-

used to illustrate the general approach to mittance (YEC = iωC).
the study of reaction mechanisms and re- Most simply, the system is modeled by
action kinetics (see Sect. 2.1.3.3). assuming depletion or accumulation of
free charge carriers at the semiconductor
2.1.3.1 In Situ Energetics of side of the interface. This charge is neu-
Semiconductor/Electrolyte Interfaces tralized by an ionic counter-charge at the
An aspect of great importance to the electrolyte side of the interface (Helmholtz
electrochemical properties of a (bulk) layer) [36, 38]. From this picture, two (dif-
semiconductor/electrolyte (s/e) interface ferential) capacitances may be defined. The
is the energetic position of the upper first, CSC , relates a change of the charge
edge of the valence band and the lower QSC accumulated in the semiconductor to
edge of the CB of the semiconductor at a change of the potential drop φSC across
the interface. According to the Gerischer- the semiconductor (CSC = ∂QSC /∂φSC ).
model for charge transfer at bulk semi- The second, CEL , relates analogously the
conductor electrodes, the energetics of charge QEL in and the potential difference
the s/e interface plays an essential role φEL across the electrolyte side of the inter-
in determining the rate and the mech- face (CEL = −∂QEL /∂φEL ). Assuming that
anism of an electrode process [34–36]. the total potential drop φ across the in-
Capacitance measurements – hence, elec- terface equals the sum φSC + φEL , it may
trochemical impedance measurements – be easily shown that the interfacial capaci-
constitute the most widely used in situ tance C corresponds to the capacitance of
method to determine the energetics of s/e the series connection of the capacitors CSC
interfaces [37]. and CEL [36, 37, 39].
As described earlier, the s/e interface In Fig. 6, the inverse square of the ca-
is electrified, that is, charge is separated pacitance of an n-InP|1.2 M HCl solution
at the interface. According to Gauss’ law, interface is plotted as a function of the po-
the charge separation is accompanied by tential applied to the n-InP electrode. The
an inner-potential difference between the capacitance has been calculated by model-
semiconductor and the electrolyte. Con- ing the interface as a parallel connection
sequently, in the absence of any electro- of a capacitor and a resistor in series with
chemical process, the s/e interface acts as the cell resistor. Clearly, both quantities
a capacitor, the capacitance of which may are linearly related. This result can be
C −2 / 1012 F−2 cm4
10
Vfb − kT/e
Fig. 6 Mott-Schottky curve determined
5 at an n-InP electrode in a 1.2 M HCl
aqueous solution. The interfacial
capacitance is determined from the
0 electrochemical impedance measured at
8.2 kHz using a parallel connection of a
−500 0 500 resistor and a capacitor in series with
V / mV vs. SCE the cell resistor.
understood if majority charge carriers are Obviously, such so-called frequency dis-
depleted from the semiconductor surface. persion hampers a proper determination
In that case, the capacitance CSC is given of Vfb although reliable values are generally
by the Mott-Schottky equation, which for obtained at high measuring frequencies
an n-type semiconductor reads [36, 37, 39]: (>10 kHz). The origin of this nonideal
$ % $ % behavior is not well understood [37, 41].
eND εε0 1/2 kB T −1/2 However, to check the reliability of Mott-
CSC = φSC −
2 e Schottky measurements, the capacitance
(10) should be measured in a broad frequency
where T is the absolute temperature range [42–44].
and the constants e and kB indicate
the elementary electrical charge and the
2.1.3.2Electron-hole Recombination
Boltzmann constant, respectively. The Dynamics at the s/e Interface
capacitance CSC can be identified directly
with the interfacial capacitance if CSC % 2.1.3.2.1 Introduction The dynamics of
CEL . Although exact values of CEL are electron-hole recombination at the semi-
scarce in the literature, this inequality conductor surface has been extensively
is generally assumed to be fulfilled for studied both at illuminated and at dark
moderately doped semiconductors under s/e interfaces [45–53]. For recombination,
depletion conditions [36, 37, 39, 40]. As minority charge carriers should be present
a consequence, a change of the potential at the interface. For n-type semiconduc-
drop φ across the interface results mainly tors, holes may be supplied to the surface
in a change of φSC (i.e. φEL is approximately by illumination under depletion condi-
constant). Hence, φSC may be written tions using supra-band gap light (see
as the difference between the applied Sect. 2.1.2.2). Alternatively, holes may be
electrode potential V and the so-called flat injected into the valence band by a strong
band potential Vfb [37]: oxidizing agent in the electrolyte solution.
If a depletion layer exists, the injected holes
φSC = V − Vfb (11)
accumulate at the semiconductor surface.
At the flat band potential, there is no As it gives a nice and relatively simple
potential drop across the semiconductor illustration of the use of various charac-
and, hence, the semiconductor energy terization methods, we will discuss the
bands are flat from the bulk up to subject of electron-hole recombination dy-
the semiconductor surface (see Fig. 1b). namics in the next sections, taking the
Moreover, because φEL is approximately n-GaAs photoanode as the main example.
constant, the energy bands are fixed at More complicated topics – related to inter-
the semiconductor surface. Hence, the facial transfer of photogenerated charge
position of the band edges at the surface carriers – are discussed in Section 2.1.3.3.
may be calculated once the flat band
potential is known. 2.1.3.2.2 Current Density Versus Potential
A problem often encountered when Measurements Figure 7 shows a current
calculating the flat band potential from density versus potential curve for an il-
Mott-Schottky data is the frequency de- luminated n-GaAs electrode (λ = 480 nm)
pendence of both the slope and the extrap- in a 0.1 M H2 SO4 solution. As reported
olation point of the C −2 versus V curve. for various photoanodes in the literature,
Fig. 7 Current density versus potential

500 A B C curve, recorded at an illuminated
n-GaAs electrode in a 0.1 M H2 SO4
aqueous solution. Indicated are the flat
[µA cm−2]
band potential (as determined in the

250
dark) and the different potential ranges
j
(see text).
Vfb
0
−1000 −750 −500 −250 0
V / mV vs. SCE
the j versus V plot shows – at potentials For various illumination intensities, the
more positive than the flat band poten- diameter of the semicircle fitting the data
tial – three different regions, indicated as at high frequencies equals approximately
A, B, and C in Fig. 7 [45, 47, 49, 50, 54]. In kT /e|jPH | [45–47, 49]. In addition, it was
region A, that is closest to Vfb , no current shown that upon illumination, a capacitive
passes through the s/e interface, despite peak appears in the C −2 versus V plot of
the minority charge carrier flux towards the n-GaAs|0.1 M H2 SO4 interface [45, 46,
the surface. On the other hand, a pho- 51]. The peak value proved to be a func-
tocurrent plateau appears in region C – at tion of the frequency and the photocurrent
the most positive potentials. The interme- density as measured in region C [51]. This
diate region B shows a transition between behavior is markedly different from the
these two extreme situations. Clearly, the purely capacitive impedance (vertical line
occurrence of a photocurrent plateau is in the Nyquist plane and straight Mott-
in accordance with the Gärtner-equation: Schottky plot) expected for a blocking s/e
in this potential region, all photogenerated interface (see Sect. 2.1.3.1).
holes reach the semiconductor surface and The appearance of both the semicircle
participate in a charge-transfer reaction and the capacitive peak were accounted
(the oxidation of the semiconductor). On for by Vanmaekelbergh and coworkers,
the other hand, since α −1 % WSC in the by considering recombination of photo-
case of n-GaAs, the absence of photocur- generated holes with CB electrons at the
rent in region A is not accounted for by semiconductor surface [51, 55–57]. The
the Gärtner-equation. This indicates a loss recombination mechanism assumed by
of photogenerated holes, which can be at- these authors consists of the successive
tributed to electron-hole recombination at capture of an electron in an empty surface
the semiconductor surface. state and of a hole in an occupied sur-
face state. Taking the rates of the electron
2.1.3.2.3 The Electrochemical Impedance (hole) capture steps to be first order in
of Surface Recombination Figure 8 shows the CB electron density nS (valence band
the impedance spectrum of the illumi- hole density pS ) and the density of empty
nated n-GaAs|0.1 M H2 SO4 interface, as (filled) surface states, an electrochemical
recorded in the potential region A (no impedance corresponding to the equiva-
steady state photocurrent). One can see lent circuit shown in Fig. 9 was calculated
a small semicircle at high frequencies. for this recombination mechanism. The
−100
R1
[Ω cm2]
Im(Z )
−50 512 Hz
32 kHz
4 kHz
0
0 50 100 150
Re(Z )
[Ω cm2]
Fig. 8 High-frequency electrochemical impedance spectrum,
obtained at an illuminated n-GaAs electrode in a 0.1 M H2 SO4
aqueous solution. Bias potential: – 600 mV versus SCE (i.e.
potential region A); limiting photocurrent density: 380 µA cm−2 .
The resistance R1 equals 71 + cm2 , that is, 1.06 × kT/ejPH .
Fig. 9 Equivalent circuit C1

obtained for surface
recombination at a
semiconductor/electrolyte
interface. Explicit expressions of
the different circuit elements are
given in the text.
R1 C2
impedance of the three different circuit space charge layer of the semiconductor,
elements read [51]: the overall impedance of the s/e interface
consists of the parallel connection of the
kT
R1 = (12) surface-recombination impedance and the
e|jREC | space charge layer capacitor. At high fre-
kT βn nS 1 quencies, the recombination impedance
R2 = (13)
e|jREC | βp pS 1 − |jREC /jPH | reduces to the resistor R1 . Hence, the
e|jREC | 1 high-frequency impedance of the inter-
C1 = (14) face corresponds to the parallel connection
kT βn nS
of R1 and CSC . This parallel combination
where jREC equals the current density accounts for the experimentally observed
associated with recombination and βn (βp ) high-frequency semicircle (note the cor-
denote the rate constant for electron (hole) respondence between the experimental
capture. diameter and R1 ). In addition, the features
Because the photocurrent density adds of the capacitive peak in the Mott-Schottky
to the current density that (dis)charges the plot could be qualitatively explained by
considering the parallel connection of the addition, it was shown that the typical
recombination impedance and the space semicircle resulting from the parallel con-
charge layer capacitor [51]. nection of the space charge layer capacitor
Clearly, electron-hole recombination is and a resistor with resistance kT /e|j | is not
not limited to illuminated semiconductors. uniquely related to surface-recombination
Minority charge carriers also may be in- processes. Any reaction step, the rate of
jected by an oxidizing or reducing agent in which is proportional to the density of
the electrolyte solution (see Sects. 2.1.1.2 majority charge carriers at the interface,
and 2.1.2.4). Also in this case, the parallel may contribute this feature to the over-
connection of the surface recombination all impedance [59–61]. Hence, for direct
impedance and the space charge capac- transfer of majority charge carriers or
itor provides an accurate description of for surface-state mediated transfer, this
the experimental impedance [45]. For in- semicircle also may appear in the over-
stance, the impedance spectrum of the n- all impedance spectrum [59]. These latter
GaAs|Ce4+ system – Ce4+ is a well-known examples demonstrate the need for a
hole-injecting agent for n-GaAs – shows reliable dynamic model of the charge-
a capacitive semicircle with a diameter transfer reaction for the interpretation of
equal to kT /e|j | at high frequencies [45, impedance data.
58]. It was also demonstrated that a capac-
itive peak, exhibiting the same functional 2.1.3.2.4 The Optoelectrical Transfer Func-
dependence on frequency and current tion of Surface Recombination If the n-
density as obtained with the n-GaAs pho- GaAs|0.1 M H2 SO4 system, polarized in
toanode, is present in a Mott-Schottky potential region A, is suddenly exposed
plot measured with the n-GaAs|Ce4+ sys- to a constant illumination, a photocur-
tem [45]. rent transient decaying from the value as
Both examples attribute convincingly the measured in potential region C to zero is
loss of photogenerated holes at the n-GaAs recorded [30]. This photocurrent decay is
photoanode polarized in potential region related to the increase from zero recom-
A to electron-hole recombination at the bination at the moment of the exposure
electrode surface. Because similar results (maximum photocurrent) to complete re-
have been obtained at the n-CdTe photoan- combination (no photocurrent). The same
ode [50], one could think of the surface- picture arises from the optoelectrical trans-
recombination impedance as a general fer function (Fig. 10), which corresponds
fingerprint of surface-recombination steps to a semicircle ranging from YOE = 1
in an overall reaction mechanism. This at high frequencies to YOE = 0 at low
is, however, not the case. For the n-InP frequencies [62]. As for the transient at
photoanode, the features of the surface short times, recombination is ruled out
recombination are absent although re- at high modulation frequencies leading
combination was shown to occur at the to a maximum value of the modulated
semiconductor surface [54]. In this partic- photocurrent, whereas at low frequencies
ular case, this discrepancy was resolved recombination is fully operative, causing
by assuming that recombination does not the disappearance of the photocurrent.
occur at fixed recombination centers but This impedance could be analyzed using
rather at intermediates of the anodic de- the same dynamic model that describes the
composition of the semiconductor [53]. In surface-recombination impedance. If the
0.5
104 Hz
i˜R
i˜h
)
ω max = 2π (4000) s−1
Im(1 −
103 Hz
0.0
0.0 0.5 1.0
i˜
Re(1 − R )
i˜h
Fig. 10 Plot of the optoelectrical transfer function measured at an illuminated
n-GaAs electrode in a 1 M H2 SO4 aqueous solution. Bias potential: – 600 mV
versus SCE (i.e. potential region A); limiting photocurrent density:
800 µA cm−2 (from Ref. 62).
response time of the electrochemical cell determination of reaction mechanisms,

is sufficiently fast, the resulting optoelec- especially if they are combined with a
trical impedance, valid for potential region mathematical model of the system studied.
A, reads according to Vanmaekelbergh and Clearly, EIS and IMPS yield partly the
coworkers [62]: same information. However, they are
also complementary: IMPS enables the
βn nS
YOE = 1 − (15) determination of the rate constant of
iω + βn nS majority charge carrier capture whereas
Apart from the flattening of the semicir- the density of majority charge carriers at
cle, this transfer function accounts well the interface is accessible only via EIS.
for the experimental impedance. More- Moreover, EIS allows one to investigate
over, from the characteristic frequency analogous processes at dark electrodes.
of the semicircle, the rate constant of
electron capture may be calculated if the 2.1.3.2.5 Recombination Studied by Lumi-
electron density at the semiconductor sur- nescence On normalizing the photolu-
face is known. From the example given minescence intensity IPL in Eq. (9) with
in Fig. 10, a value of βn = 10−6 cm3 s−1 respect to the maximum intensity Imax
is obtained [62]. Analogous to the case (for WSC = 0) one obtains
of the n-GaAs photoanode, loss of pho-
IPL
togenerated holes could be attributed to = exp(−αWSC ) (16)
Imax
surface recombination by IMPS in the
case of the illuminated n-CdS and n-InP Because the thickness of the depletion
electrodes [62, 54]. layer can be obtained as a function of
Summarizing, we can conclude that potential from impedance measurements
both examples demonstrate that impe- (see Sect. 2.1.3.1) Eq. (16) can be used
dance techniques are powerful tools for the to check the validity of the Gärtner
1.0
3
−In(IPL/Imax)
0.8 2
1
0.6
[a.u.]
IPL
0
0 10 20 30 40 50 60 70
0.4
WSC
[nm]
0.2
0.0
−1.0 −0.5 0.0 0.5 1.0 1.5 2.0
V vs SCE
[V ]
Fig. 11 The intensity of the PL measured at 2.2 eV (the ‘‘yellow’’ emission) from an n-type
GaN electrode as a function of potential. The excitation energy was 4 eV. In the inset the
results are plotted according to Eq. (16) (from Ref. 25).
approach and to determine the absorption cathodic reduction. These processes intro-
coefficient [63] (see Fig. 11). Alternatively, duce surface or near-surface states that
if α is known accurately then Eq. (16) provide pathways for nonradiative recom-
gives information about the potential bination. Gerischer and coworkers [31, 32]
distribution within the semiconductor, as have studied such systems using an ex-
shown by Ellis and coworkers for GaAs [63, tended Gärtner model.
64]. If α values are either available or There are numerous reports on the po-
determined for different wavelengths, then tential dependence of the electrolumines-
the minority carrier diffusion length [Lmin cence for n-type electrodes under (quasi)
in Eq. 9] can be obtained. steady state conditions and a limited num-
In many cases, the photoluminescence ber for p-type electrodes [25]. Such studies
intensity of n-type electrodes at nega- were mainly used for diagnostic purposes,
tive potential does not attain a constant that is, to investigate the role of minority
value as would be expected from Eq. (16) carriers in electrode processes.
[25]. Hysteresis is also frequently encoun- While there have been reports of time-
tered in cyclic scanning experiments [25]. resolved photoluminescence [65–68] and
Such effects are due to potential-dependent electroluminescence experiments [69–71]
changes in the surface chemistry. Under for the study of recombination, trapping
accumulation conditions in aqueous so- and detrapping of photogenerated charge
lution, hydrogen is evolved and may be carriers, potential modulation has scarcely
incorporated into the electrode. In ad- been used. An illustration of the possi-
dition, the semiconductor may undergo bilities of this approach is provided by
work of Decker and coworkers [72] who that is capable of injecting an electron
used a frequency-dependent train of po- into the CB. This electron also contributes
tential pulses to study the effect of surface to the photocurrent. As a result, for each
chemistry on the radiative recombination photon absorbed in the system two charge
of holes injected into n-type GaAs from carriers are registered as current. This cor-
Fe(CN)6 3− in solution. An unexpected responds to a ‘‘quantum efficiency’’ of two.
potential-dependence of the emission was For formate oxidation at n-type ZnO, Mor-
attributed to a change in the surface termi- rison and coworkers [74, 75] suggested a
nation from hydroxide to hydride coverage mechanism of the type
on going from positive to negative po-
HCOO− + hVB + −−−→ H+ + COO −
•
tentials. A time constant of 0.1 ms was
(17)
estimated for this transition. This trans-
COO − −−−→ CO2 + eCB − (18)
•
formation was subsequently observed by
Erné and coworkers using in situ in-
frared spectroscopy [73]. It is clear that Results obtained by Honda and cowork-
potential-modulated luminescence tech- ers [76] showed that the reaction is more
niques deserve wider attention. complicated. The ZnO electrode is, in fact,
dissolved during the current-doubling re-
action. The stoichiometry found in their
2.1.3.3 Complex Charge-transfer
work is given by
Processes
In this section we use two types of com- ZnO + 2HCOO− + 2hVB + −−−→
plex charge-transfer reaction to illustrate
the general approach to the elucidation of Zn2+ + H2 O + CO2 + 2eCB − (19)
reaction mechanisms at single crystal elec- This result calls into question mecha-
trodes. These reactions are photocurrent nisms in which the reducing agent reacts
doubling at n-type and p-type semicon- directly with valence-band holes. Honda
ductors and the (photo)anodic oxidation of and coworkers proposed that formate is
the semiconductor itself. oxidized by atomic oxygen formed in the
photoanodic oxidation of ZnO.
2.1.3.3.1 PC-doubling Reactions PC doub-
ZnO + 2hVB + −−−→ Zn2+ + Oads (20)
ling refers to a type of charge-transfer
reaction in which both bands of the semi- A subsequent study of the dependence of
conductor are involved, thus emphasizing the kinetics of the photocurrent-doubling
the distinctive features of semiconduc- reaction on the light intensity supports
tor electrochemistry. The first examples this idea [77]. The following mechanism
of such reactions relate to the photoan- was suggested
odic oxidation of species such as formate
Oads + HCOO− −−−→ OH + COO −
• •
and tartrate at wide band gap n-type elec-
(21)
trodes [74, 75]. A photon generates an
OH + HCOO− −−−→ H2 O + COO −
• •
electron-hole pair in the semiconductor.
The electron and hole are separated by the (22)
electric field of the depletion layer. The The stoichiometry of this scheme agrees
electron is detected as photocurrent in the with that reported by Honda and cowork-
external circuit. The hole oxidizes a species ers. This result could be checked by study-
from solution producing an intermediate ing the kinetics of two types of competing
process. Iodide ions present in solution In 1969, Memming reported photocur-

compete for reactive intermediates and rent doubling for the reduction of H2 O2
thus influence the quantum efficiency. and S2 O8 2− at p-type GaP. In this case,
In the potential range between the onset two holes are detected as photocurrent for
and saturation of photocurrent, electron- each photon absorbed. This result was ex-
hole recombination competes with the plained by a two-step mechanism [82]. The
charge-transfer reactions. The study of first step involves reduction of the oxidiz-
these competing processes by EIS and ing agent (e.g. H2 O2 ) by a photogenerated
IMPS supports the mechanism indicated electron
by Eqs. (21) and (22) [77]. One of the prob-
H2 O2 + eCB − −−−→ OH− + OH
•
(23)
lems encountered in the investigation of
such complex mechanisms is the identifi- and the hydroxyl radical intermediate
cation of (reactive) intermediates. For the injects a hole into the valence band
ZnO/formate system Harbor and Hair [78]
OH −−−→ OH− + hVB +
•
used ESR spin-trapping experiments to (24)
show that the COO− radical anion is in-
•
deed formed. Since this first report, photocurrent dou-

Quantum efficiencies higher than one bling has been found for a whole range
have been observed for a range of reducing of two-electron oxidizing agents at various
semiconductors including Si, SiC, CdTe,
agents at ZnO [74, 75]. The mechanisms
and III–V materials [83]. A striking exam-
have been studied less thoroughly. In the
ple of photocurrent ‘‘multiplication’’ is the
case of the oxidation of methanol, it is
reduction of iodate at p-GaAs [83]. In a
clear that the mechanism differs from
wide range of light intensity, a quantum
that of formate (and tartrate) [77]. Pho-
efficiency of three is observed, whereas at
toanodic current doubling also has been
low light intensity there is evidence for an
observed for other n-type semiconductors
efficiency of six. This would mean that five
such as TiO2 [79] and CdS [80]. Bogdanoff
of the six intermediates formed on reduc-
and Alonso-Vante [81] have described an ing IO3 − to I− could inject a hole into
interesting study of the competitive pho- the valence band of GaAs. In a number of
toanodic oxidation of formic acid and water these systems, the oxidizing agent causes
at TiO2 using differential electrochemical chemical dissolution of the solid. In the
mass spectroscopy. In contrast to ZnO, case of GaAs, studies of etching and cur-
TiO2 is a stable photoanode. On the ba- rent doubling have led to the conclusion
sis of on-line mass detection, the authors that the various processes involved (elec-
conclude that formic acid is oxidized by tron capture, hole injection, and chemical
hydroxyl radicals produced by the pho- etching) are linked via a common interme-
toanodic oxidation of water, a reaction diate formed by the chemisorption of the
somewhat similar to that of oxygen radicals oxidizing agent on the surface [15, 84, 85].
at ZnO (see Eq. 21). In the case of formate Obviously, information about the mech-
oxidation on CdS, there is no evidence for anisms of such reactions can be obtained
corrosion of the semiconductor. A direct by studying the photocurrent as a func-
reaction of the current-doubling agent with tion of system parameters (potential, light
a valence-band hole was suggested for the intensity, concentration, hydrodynamics,
first reaction step [80]. etc.). However, such measurements do not
yield information about the reaction kinet- the p-type semiconductor dissolves anod-
ics. Peter and coworkers [86, 87] were the ically in the dark showing an exponential
first to show that rate constants for major- increase of current with increasing poten-
ity carrier injection could be determined tial, whereas the n-type semiconductor can
for such systems by IMPS. They studied only be oxidized if minority carriers are
current doubling for oxygen reduction at generated by light under depletions condi-
p-type GaAs and GaP. From a complex- tions (see Fig. 7). The anodic oxidation of
plane representation of the optoelectrical most semiconductors is a complicated pro-
impedance, rate constants in the range cess. The reaction of GaAs, for example,
104 –105 s−1 were calculated. This result which can be represented schematically by
suggests that hole injection is a thermally
•
activated process. HO2 , postulated as the
GaAs + 6hVB + −−−→ Ga(III) + As(III)
injecting species, gives rise to a surface
(25)
energy level; in the case of GaAs, this is
requires six valence-band holes to form
located about 0.4 eV above the valence-
trivalent gallium and arsenic species.
band edge.
Chemical reactions are obviously also
involved. The final products depend on
2.1.3.3.2 Dissolution of Semiconductors the nature of the electrolyte solution. The
Another class of complex reactions that information that can be obtained from
has been widely studied is the oxidative cyclic voltammetry is clearly rather limited.
dissolution of elemental and compound Kinetic studies in which oxidation of a re-
semiconductors. There are a number of ducing agent present in solution competes
reasons for the interest in these systems.
with oxidation of the semiconductor have
The possible application of semiconduc-
yielded a wealth of information, allowing
tors in regenerative photoelectrochemical
quite detailed dissolution mechanisms to
solar cells required the complete suppres-
be proposed [88]. Generally, in these stud-
sion of corrosion of the photoelectrode [5,
ies an RRDE is used; the products formed
6, 22, 23]. On the other hand, with the de-
at the semiconductor disk are detected
velopment of the optoelectronics industry
electrochemically at a noble metal ring.
based on III–V materials there was a need
for a more fundamental understanding of Semiconductors can be etched under
etching processes [15]. The revival of inter- open-circuit conditions with an oxidiz-
est in porous semiconductors, triggered by ing agent that is capable of being re-
the discovery of the unusual optical prop- duced by extracting electrons from the
erties of porous Si, led to a general revival valence band, that is, creating holes in
of interest in the mechanisms of porous the band (see Sect. 2.1.3.2.3) [15]. Infor-
etching [18]. mation on hole injection can be obtained
Because the bonding states correspond from electroluminescence studies on n-
to the valence band of a semiconductor, type electrodes. Electrochemical studies of
one expects holes to be important for such ‘‘electroless’’ etching systems have
the oxidation reaction [15, 24]. This is the shown that, during dissolution, interme-
case with most etching systems. The im- diates are formed with energy levels in
portance of holes is immediately clear the band gap [89]. From such states, elec-
from simple current-potential or cyclic trons can be thermally excited into the CB.
voltammetric measurements. Generally, If this occurs in an n-type semiconductor
under depletion conditions then the in- InP systems, providing information about
jected electron can be detected as current the sequence of the reaction steps and
in the external circuit. By measuring this the magnitude of the rate constants for
anodic current as a function of injection majority carrier injection [54, 92].
rate and temperature, one gets extensive Because electron injection is detected
information about both chemical and elec- during dissolution of n-type semiconduc-
trochemical processes [88, 89]. Gomes and tors under illumination, it seems likely that
coworkers have shown that electron injec- it should also occur during dissolution of
tion and RRDE competition studies are p-type semiconductors in the dark. There
complementary [90]. The contribution of are two ways in which this can be checked.
electron injection to the total anodic oxi- If the injected electrons recombine radia-
dation rate is, for most etching systems, tively with majority carriers (the holes),
small compared to that of holes. light is emitted. EL corresponding to band-
Electron injection from reaction interband recombination has been observed
mediates of the oxidation of n-type semi- during anodic dissolution of p-type InP in
conductors can be observed as quantum ef- HCl solution (Fig. 12) [93]. The quantum
ficiency larger than unity in photocurrent- efficiency for light emission, that is, the
potential measurements. There are two number of photons emitted per electron
striking examples in the literature: the passed through the external circuit, was
photoanodic dissolution of n-type silicon very low (approximately 10−6 ). This proba-
in HF solution [91, 92] and of n-type InP bly is due to a low rate of electron injection
in HCl solution [54]. In these cases the and a high rate of nonradiative recom-
quantum efficiency at low light intensity bination at the etching surface. Another
is exceptionally high, four for silicon and approach that allows one to distinguish
two for InP. In the case of Si, this means quantitatively between electron and hole
that only one photon (and thus one hole) contributions to the current is the p-n junc-
is required to dissolve each silicon atom; tion configuration [94, 95]. This technique
three electrons are injected into the con- has been used for the study of the anodic
duction band dissolution of silicon in alkaline solution.
This is an unusual system showing a novel
Si + hVB + −−−→ Si(IV) + 3eCB − (26)
coupling of chemical and electrochemical
At higher light intensity, the quantum effi- steps [95]. Both p-type and n-type silicon
ciency drops indicating that reaction steps can be oxidized electrochemically in the
involving valence-band holes take over dark. In both cases, passivation occurs if
from electron injection steps. A quantum a limiting current is exceeded. The anodic
efficiency of two for InP means that three peak current observed as n-type silicon
of the six oxidation steps require minority passivates is comparable to that of p-type
carriers and thus photons. As for silicon, silicon. Anodic current from an n-type
the quantum efficiency decreases (from 2 electrode in the dark can, under normal
to 1) as the light intensity is increased. As circumstances, only result from electron
described in the previous section, IMPS injection into the CB. Because there is no
can be used very effectively to study the electron donor present in solution, elec-
mechanisms of ‘‘photocurrent multiplica- tron injection must occur from a surface
tion’’ reactions. The method has proved species. This species very likely arises as
particularly successful for the silicon and a result of the chemical-etching reaction
1.0
1.0
0.8
0.8
0.6
0.4
0.6 0.2
[a.u.]
IEL
0.0
800 900 1000
0.4 Wavelength
[nm]
0.2
0.0
3.0
[mA/cm2]
2.0
j
1.0
0.0
0.0 0.2 0.4 0.6
V vs SCE
[V]
Fig. 12 The potential dependence of the EL intensity IEL and of
the current density j for a p-type InP electrode, dissolving
anodically in a 1 M HCl solution. The inset shows the EL
spectrum (from Ref. 93).
of silicon by water. Anodic current with p- Mott-Schottky theory can be used to

type silicon can be due to a hole reaction or, determine the flat band potential (see
as in the case of n-type silicon, to electron Sect. 2.1.3.1). This then allows one to cal-
injection. Measurements with a p-n junc- culate the band-bending for any value of
tion electrode showed the latter to be the the applied potential, if the Helmholtz
case [95]; most of the anodic current can potential remains constant (i.e. the band
be attributed to minority carrier injection. edges are pinned). The band-bending de-
Electrical impedance measurements are termines the concentration of majority
essential in semiconductor dissolution carriers at the surface (see Eqs. 25 and 26).
studies. Under depletion conditions, a In a p-type electrode these are holes, which
change in applied potential will usu- are essential for the dissolution reaction.
ally give rise to a change in the po- In an n-type electrode, the band-bending
tential drop within the semiconductor. determines the surface electron concentra-
Impedance measurements combined with tion and thus, the rate of recombination
with photogenerated or injected holes; this also can be used to get information about
process competes with the oxidation of the electron injection processes [97].
semiconductor by holes. Electrochemical measurements of the
The assumption of pinned band edges is type described earlier give indirect evi-
very often not valid. Lincot and Vedel [50] dence about dissolution processes. More
in an early study of the photoanodic dis- direct chemical information can be ob-
solution of n-type CdTe used EIS to show tained from in-situ spectroscopies, in
that the Fermi level becomes pinned in particular from IR and Raman methods.
a wide potential range positive with re- Chazalviel and coworkers have showed the
spect to Vfb . This means that in this range power of this approach in studies on sili-
the band-bending within the semiconduc- con and GaAs [73, 98, 99]. Electrochemical
tor remains constant while the potential and spectroscopic techniques are macro-
changes across the Helmholtz layer. In scopic methods giving a view of the whole
their analysis, Lincot and Vedel consider electrode surface. To study semiconduc-
the rate constants for change transfer to be tor dissolution at the microscopic (atomic)
exponentially dependent on the Helmholtz level, one needs techniques such as scan-
potential. ning tunneling microscopy (STM) and
In many early studies of anodic oxida- atomic force microscopy (AFM). The an-
tion of p-type semiconductors it was tacitly odic and chemical dissolution of silicon
assumed that the reaction was controlled has been studied in very elegant work by
solely by the surface hole concentration, Allongue and coworkers [100–102].
that is, by the potential drop across the
space charge layer of the solid. The role 2.1.4
of the Helmholtz potential was neglected. Sensitizer-based Photoelectrochemical
For p-type electrodes, anodic oxidation oc- Systems
curs at potentials close to the flat band
value and under accumulation conditions. Dye molecules adsorbed on an electrode
In addition, there will be a high den- form, in principle, the simplest class
sity of surface states as a result of the of photoelectrochemical system. The ba-
breaking of surface bonds. In this case, sic scheme presented at the end of
a change in the applied potential is likely Sect. 2.1.1.2 is directly applicable for un-
to be distributed partly or even completely derstanding the mechanism of photocur-
over the Helmholtz layer. The potential rent generation. The dye molecules act
dependence of the anodic current then is as the light-absorbing species. If injec-
due to (in part) the changes in the rate tion of a photogenerated charge carrier
constants for hole capture resulting from into the electrolyte or electrode com-
changes in the Helmholtz potential. Elec- petes effectively with relaxation of the
trical impedance spectroscopy is necessary excited state, a photocurrent might be
to decide whether etching is under space observed in the external circuit. Singlet-
charge layer or Helmholtz-layer control. to-triplet crossing often leads to a relatively
The latter was shown to be the case for long-lived excited state, which allows
the dissolution of p-GaAs in acidic elec- electron-hole separation to occur by an
trolytes [96]. Vanmaekelbergh and Searson electron transfer process. The photochem-
in a study of the dissolution of p-type istry and photoelectrochemistry of dye-
silicon in HF solution showed that EIS sensitized electrodes has been studied
extensively, and the research is still contin- and bottom of the CB and to an increase
uing. Dye-sensitization of photochemical of the band gap energy with respect to that
reactions is a key topic in photographic of a macroscopic crystal. The energy-level
research [103, 104]. At present, there is spectrum shifts from that of a classical
a considerable effort being devoted to insulator to that of a molecule with a re-
dye-sensitized porous photoelectrochem- duction of the dimensions in the 20–1-nm
ical solar cells [16, 105]. The fundamental range. Second, a considerable fraction of
research is focused on the chemical bond- the atoms of a nanocrystal lies at the sur-
ing between the light-absorbing molecules face. This leads to surface-electron states.
and the electrode surface, the mechanism States of energy in the band gap can have
of light absorption, and the dynamics of a strong influence on the optical proper-
charge injection and recombination in the ties of nanocrystals. It is clear that the
dyes. In this research, scanning probe surface chemistry is extremely important.
methods [106, 107] and time-resolved op- Organic and inorganic molecules can pas-
tical spectroscopy [108, 109] are used to- sivate surface states, thus removing them
gether with photoelectrochemical charac- from the optical gap. Functionalized cap-
terization [24]. The reader is referred to a ping molecules play an important role in
more specialized review for details [24]. providing stability against coagulation and
Electrodes to which insulating light-
in allowing the attachment of nanocrys-
absorbing nanocrystals, instead of dyes,
tals to solid (electrode) surfaces (see the
are attached form a relatively new class of
following text).
system. With colloidal solution chemistry,
Early reports have shown photoelec-
a large variety of insulating nanocrystals
trochemical activity when, for example,
can be prepared [110–113]. Well-known
CdS and PbS nanocrystals are attached
examples are II–VI compounds (CdS,
to a metal electrode in a sub-monolayer
CdSe, and ZnO), III–V compounds (InP,
array [19, 20, 116–120]. Clearly, photoex-
InAs, and GaN), and transition-metal
oxides. In addition, there are various citation of the nanocrystals can lead to a
methods in solid-state chemistry and long-lived state, which allows one of the
electrochemistry for preparing nanocrys- charge carriers to be transferred from the
tals directly on surfaces. For instance, nanocrystal before recombination occurs
CdS, CdSe, and PbS nanocrystals can be (see Sect. 2.1.1.2).
electrochemically deposited on gold elec- A macroscopic PbS crystal has a band
trodes [114]. Electrochemical oxidation of gap of 0.41 eV. Because of the small ef-
Si leads to porous silicon that may contain fective mass of the electrons and holes
a large number of Si nanocrystals [115]. (me,eff = mh,eff = 0.09 × me ), strong size-
Nanocrystalline colloidal systems have quantization occurs. The absorption spec-
been extensively characterized with opti- trum of an aqueous suspension (of
cal spectroscopy [110–113]. There are two polyvinyl alcohol-capped) PbS nanocrys-
effects that are essential for understand- tals, 6.5 nm in diameter (see TEM picture)
ing the electronic and optical properties is shown in Fig. 13. The HOMO-LUMO
of insulating nanocrystals. First, quan- optical transition occurs at 2.1 eV, and two
tum confinement of the electron waves other absorption peaks are seen at 3.2 and
in the nanocrystal leads to discrete elec- 4.3 eV. When a gold electrode is immersed
tron states at the top of the valence band in this colloidal solution, PbS nanocrystals
Fig. 13 Absorption spectrum of an

1.0 aqueous suspension of nanocrystalline,
size-quantized PbS particles (capped
0.8
with polyvinylalcohol). A HR-TEM image
of a typical PbS nanocrystal is shown in
Absorbance
a the insert; the diameter of the

0.6
[a.u.]
nanocrystals is about 6.5 nm (from

Ref. 122).
0.4
0.2 10 0
[nm]
0
1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Energy
[eV]
are attached and form a monolayer array, state contains an electron located at 0.4 eV
see Fig. 14 [121, 122]. versus SCE, thus considerably below the
The photoelectrochemical activity of LUMO of Q-PbS. The much faster kinetics
a Au/Q-PbS electrode, illuminated with of the photoexcited hole transfer indicates
chopped green light, is shown in Fig. 14 that the hole in the long-lived excited state
for three different aqueous electrolyte is still delocalized, that is, it occupies the
solutions. With a 1 M KCl solution (case a), HOMO. It was thus concluded that the
photocurrent transients are observed when excited state in Q-PbS leading to pho-
the electrode potential is more positive toinduced electron transfer consists of a
then −0.4 V (versus SCE). Illumination trapped electron and a delocalized hole in
gives a cathodic current that decreases to the HOMO (denoted as Q(etrap , hHOMO )).
zero with increasing time; turning off the The following scheme describes the decay
light induces an anodic current. Clearly, il- dynamics of the long-lived state in Q-PbS:
lumination leads first to the transfer of
an electron from gold to the photoex- hν relaxation
Q −−−→ Q∗ −−−−−−→
cited hole in the PbS nanocrystals (a
‘‘cathodic’’ transfer), followed by trans- relaxation
fer of the photoexcited electron to empty Q(etrap , hHOMO ) −−−−−→ Q + heat
states in the gold (an ‘‘anodic’’ transfer). (27)
Energy relaxation in the gold due to elec- transfer from Au
tron transfer competes with electron-hole Q(etrap , hHOMO ) −−−−−−−−−→ Q− (etrap )
recombination in Q-PbS. For this to oc- (28)
cur, the electrochemical potential in the transfer to Au
gold electrode must be located between Q− (etrap ) −−−−−−−−→ Q (29)
the energy levels corresponding to the
photoexcited electron and hole. Recombi- Interestingly, when tartrate is added as an
nation transients are suppressed when the electron donor to the solution, an anodic
Fermi level is above 0.4 eV (versus SCE), photocurrent is observed in the poten-
which indicates that the long-lived excited tial range positive with respect to −0.4 V
1.0
0.0
−1.0 Light on
−2.0
−3.0 100
Light off
(a)
Light off 50
1.0
[µA /cm2]
0
J
0.5 0 50 100 150

Light on
[nm]
0.0
(b)
1.0
0.0
Light on
−1.0
Light off
(c)
−0.4 −0.3 −0.2 −0.1 0.0
V/V (SCE)
Fig. 14 Current-potential curves, under chopped light (514 nm), measured with a Q-PbS
covered gold electrode in a photoelectrochemical cell with three aqueous electrolytes. The
insert gives an STM picture of a part of the Q-PbS covered electrode showing a disordered
monolayer coverage. (a) 1 M KCl; (b) 1 M KCl + 0.2 M tartrate (electron donor);
(c) 1 M KCl + 0.01 M K3 Fe(CN)6 (electron acceptor) (Ref. 122).
(Fig. 14, case b). This means that electron above 0.4 eV. This is logical, because
donation from tartrate can compete effec- the cathodic photocurrent corresponds to
tively with electron donation from the gold electron transfer to the oxidized species,
electrode [i.e. Eq. (28)]: not to the electrode.
Photoelectrodes, consisting of CdS
transfer from tartrate
Q(etrap , hHOMO )+Red −−−−−−−−−−−−→ nanocrystals that are directly attached
to an electrode (gold or conducting ox-
Q− (etrap ) + Ox+ (30) ide) or indirectly via a dithiol molecular
linker have been studied extensively [19,
In contrast, a cathodic photocurrent is 20, 116–120]. Here, we will focus on
observed when Fe(CN)6 3− is added to the work that demonstrates the power of
solution (Fig. 14, case c). Thus, electron small-amplitude methods in photoelectro-
donation to the oxidizing species in chemical characterization [26, 123, 124].
solution competes effectively with electron By measuring the photoelectrochemical
donation to gold; it can be clearly seen activity of gold/Q-CdS electrodes, as the
in Fig. 14, case c that the electrochemical potential of the gold electrode was varied,
activity in the latter solution does not stop it could be concluded that photoinduced
when the electrode Fermi level is increased electron transfer between the gold surface
and the quantum dots is due to a long- photoinduced electron transfer from the
lived state in Q-CdS. This state, denoted as LUMO in Q-CdS to empty states in the
[Q(eLUMO , htrap )], consists of an electron gold electrode:
in the LUMO and a hole trapped in a level
ωb(
about 0.6 eV below the LUMO:
Q(eLUMO , htrap ) −−−→ Q+ (htrap ) (32)
hν relaxation
Q −−−→ Q∗ −−−−−−−→ ωc
Q+ (htrap ) + Red −−−→ Q + Ox+ (33)
relaxation
Q(eLUMO , htrap ) −−−−−→ Q + heat With a 1 M KCl solution, the low-fre-
(31) quency limit ∂jPH /e∂% of the optoelectri-
cal transfer function is zero; this agrees
Time-resolved photobleaching experi- with the fact that there is no steady state
ments on Q-CdS colloidal solutions photocurrent. The high-frequency limit of
showed the existence of a long-lived the optoelectrical transfer function is also
state (lifetime 50 ms) that, very probably, zero. Starting from 105 s−1 , an anodic vec-
corresponds to Q(eLUMO , htrap ) [125]. tor develops with decreasing modulation
Figure 15 shows the optoelectrical trans- frequency. This corresponds to electron
fer function (see Sect. 2.1.2.4), measured transfer from the Q-CdS LUMO to gold.
with near UV-light from an argon laser The rate ωb of this process cannot be
for a 1 M KCl solution (a, b) and a KOH- clearly distinguished from the discharge
tartrate solution (c, d). In a and c, the frequency of the photoelectrochemical cell.
transfer function is plotted in the complex Electron transfer from the LUMO to the
plane with the frequency as a parameter, gold is probably faster than 3 × 104 s−1 .
in b and d, the modulus of the transfer At lower frequencies, the modulus of the
function is plotted versus the modulation optoelectrical transfer function shrinks to
frequency ω. In the KOH-tartrate solu- zero. This is because of the electron back-
tion (c, d), an anodic vector is observed transfer from gold to the empty level in
at sufficiently low frequency. This means Q-CdS following the faster LUMO-to-gold
that ∂jPH /e∂% > 0 (see Sect. 2.1.2.2). This electron transfer:
agrees with the observation of a steady ωa
state anodic photocurrent, which increases Q+ (htrap ) −−−→ Q (34)
with increasing light intensity. The anodic
vector shrinks with increasing modula- The rate of this transfer (ωa ) is 6 × 103 s−1
tion frequency ω and, eventually, becomes and is independent of temperature. Thus,
zero for ω > 5 × 104 s−1 . Two semicircles at sufficiently low modulation frequency,
with characteristic frequencies ωc = 6 s−1 both the processes represented by Eqs. (32)
and ωb = 700 s−1 are observed (measure- and (34) are in phase with the modulated
ments at 25 ◦ C). Measurements in the light intensity, and the resulting photocur-
temperature range 0–60 ◦ C showed that rent is zero. When no redox system is
ωb is temperature-independent, while ωc present, relaxation of the long-lived excited
increases with increasing temperature. state in Q-CdS, Q(eLUMO , htrap ), occurs via
Hence, ωc corresponds to the rate of elec- consecutive steps involving electron trans-
tron donation from tartrate to the hole fer from the LUMO to the gold, and from
trapped in Q-CdS, while ωb is the rate of the gold to the trapped hole. This relaxation
103 × 2 π rad/s
1.0 ωa
Imaginary [ j(ω)/e Φ(ω)]
0.5
Modulus [ j(ω)/e Φ(ω)]

102 × 2 π rad/s
[10−3]
0 1.5
[10−3]
10 × 2 π rad/s
−0.5 1.0
0.5
−1.0 104 × 2 π
rad/s ωb
0 0.5 1.0 1.5 2.0 2.5 102 103 104

(a) Real [ j(ω) /e Φ(ω)] (b) Frequency
[10−3] [s−1]
1
10−1 × 2π rad/s 4
Modulus [ j(ω)/e Φ(ω)]
0
3
[10−2]
[10−2]
ωc 2
−1
10 × 2 π
rad/s 1
103 × 2 π rad/s 102 × 2 π rad/s
ω b′
−2
0 1 2 3 4 100 101 102 103 104
(c) Real [ j(ω) /e Φ(ω)] (d) Frequency
[10−2] [s−1]
Fig. 15 The optoelectrical transfer function discussed in the text. In (b), the modulus of the
measured for a Q-CdS covered gold electrode in transfer function is plotted vs. the modulation
an electrochemical cell. The attached quantum frequency. (c,d) Results obtained in an aqueous
dots have an average diameter of 5 nm. (a,b). solution (pH = 14) with 0.5 M tartrate as an
Results obtained in an aqueous solution electron donor. (c) Plot of the transfer function
(pH = 7) with 1 M KCl (297 K). In (a), a plot of in the complex plane with the modulation
the transfer function is shown in the complex frequency as a parameter. (d) Plot of the
plane with the modulation frequency as a modulus of the transfer function versus the
parameter. The meaning of the transfer function modulation frequency (from Ref. 123).
and of the characteristic frequencies (ωa , ωb ) is
competes effectively with relaxation of the to determine the rates of photoinduced

excited state in the particle. electron transfer in gold/spacer/Q-CdSe
It can be concluded that measurement assemblies [121, 126, 127]. Cyclohexyli-
of the optoelectrical transfer function in dene disulfides form well-ordered and
a photoelectrochemical cell is a powerful rigid molecular monolayers on gold (111)
technique for studying the mechanism and surfaces because of strong Au-S bonding
kinetics of photoinduced electron transfer. and intermolecular van der Waals inter-
The technique has been further exploited actions [121]. Q-CdSe quantum dots were
linked covalently at the other end of the heterogeneous catalysis, storage batter-
molecule via the S-termination. The rates ies, fuel cells, and super-capacitors [41].
of photoinduced electron transfer from Porous insulating or semiconducting
the LUMO of Q-CdSe to gold (kL,Au ), phases have been considered as photo-
and from gold to the trapped hole in chemical devices for the light-stimulated
Q-CdSe (kAu,T ) were obtained from the oxidation of organic waste components
optoelectrical transfer function. An expo- [heterogeneous photocatalysis] [110–113,
nential decrease of the electron transfer 129]. Porous photoelectrochemical sys-
rate with increasing length of the spacer tems have been studied extensively only in
molecule is found, with a decay parameter the last ten years [16, 105, 130–137]. Their
β = 0.5 Å−1 . This low value (the value in fabrication is more demanding than that
vacuum is typically four times larger) in- of photochemical systems, because electri-
dicates a strong through-bond electronic cal work is delivered in an external circuit
coupling in the cyclohexylidene spacer upon illumination (see Sect. 2.1.2.2). In a
molecules. This agrees with the result of porous photoelectrochemical system, the
quantum chemical calculations and with solid phase is insulating or semiconduct-
ing, whereas the permeated phase is an
the strong S–S coupling shown by photo-
electrolyte solution. The current in the
electron spectroscopy [126, 127].
solid phase is due to the motion of elec-
The measurement of changes in the op-
trons or holes; the current in the solution
tical properties of nanocrystals attached
is due to the motion of an oxidized or
to an electrode caused by variations in the
reduced species.
electrode potential forms a challenging but
A porous photoelectrochemical system
interesting characterization method. Re-
can be prepared by the deposition of
cently, a bleaching of the HOMO-LUMO colloidal particles on a conducting sub-
transition of CdS nanocrystals has been ob- strate. It is essential that the particles are
served when the electrode Fermi level was electrically connected and that there is elec-
in resonance with the LUMO [128]. This trical contact between the particles and
led the authors to conclude that the pho- the conducting substrate. Furthermore,
toexcited state in CdS contains an electron the pore system should form a single-
in the LUMO (and a trapped hole). This permeated phase. A well-known example
is in agreement with the results obtained of such a system is the particulate TiO2
by photoelectrochemical characterization photoelectrode, which forms the basis of
of gold/Q-CdS electrodes (see earlier). the photoelectrochemical solar cells pro-
posed by O’Regan and Grätzel and other
2.1.5 groups [16, 130–137]. A SEM picture of
Porous Photoelectrochemical Systems a TiO2 network, consisting of intercon-
nected spheres with a diameter of 30 nm,
2.1.5.1 Introduction is shown in Fig. 16(a); electrical contact
Here, we define a porous solid as a phase between the particles and between the con-
that contains empty spaces that are in- ducting substrate and the particles was
terconnected. Thus, there is a single-solid achieved by slight sintering at 450 ◦ C.
phase that can be permeated with a second An alternative route exists for the prepa-
phase. Porous metals have been exten- ration of porous semiconductors. Many
sively studied, and are widely applied in n-type single crystals, such as GaP, GaAs,
Acc.V. Magn.
10.0 kV 108639 200 nm
(a)
(b) 00018 3 µm
Fig. 16 (a) A particulate network consisting of 25 nm TiO2 particles prepared by
deposition from a colloidal solution followed by slight sintering; (b) A porous GaP
crystalline network prepared by anodic etching of an n-type GaP crystal at positive
potential. The structural units have dimensions in the 100–300 nm range.
Si, SiC, and TiO2 can be transformed into would explain why anodic dissolution is
a porous network by anodic etching un- so strongly nonuniform over the surface
der conditions of severe band-bending [18, and eventually leads to the formation of
115, 138–144]. Under such conditions, a porous semiconductor. However, many
surface electrons located at the top of the questions remain regarding porous pho-
valence band or in band gap states can toelectrochemical etching. A well-studied
tunnel through the gap into the CB; the example is macroporous GaP [139, 143].
surface localized holes generated in such When n-type GaP is subjected to a poten-
a way are consumed in anodic dissolu- tial of 5 V or more (versus SCE) a highly
tion of the material. The rate of inter- porous network is formed (Fig. 16b) with
band tunneling (Zener breakdown [145]) dimensions of the structural units and
is strongly dependent on the electric field the pores in the 100-nm range. The GaP
and the presence of surface defects. This network retains its crystallinity [146, 147].
A ‘‘single-crystal GaP sponge’’ is different much smaller but also larger than the di-
from a particulate network in that there mensions of the solid structures and the
are no grain boundaries and the connec- pores. Therefore, porous semiconductors
tion between the network and the bulk can show remarkable charge-storage prop-
GaP matrix is ideal. erties. Other relevant length scales are the
wavelength of visible light (400–700 nm)
2.1.5.2 Special Properties of Porous and the diffusion length of charge carri-
Photoelectrochemical Systems ers before recombination. These length
Porous photoelectrochemical systems con- scales also can be in the same range
sist of an insulating or semiconducting as the dimensions in the porous struc-
solid network permeated with a conduct- ture; this leads to striking optical and
ing electrolyte solution; the dimensions electrodynamic properties. Extensive re-
of the solid structures and pores are in search is being performed in these fields,
the 1–500-nm range. A typical semicon- and a comprehensive review is beyond
ductor/electrolyte interface has a width the scope of this section. Instead, the
of between 0.5 nm (the Helmholtz layer electrostatic, optical, and electrodynamic
in a concentrated electrolyte solution) properties of porous semiconductor (elec-
and 100 nm (typical depletion layer in a trodes) are briefly discussed and the
semiconductor). Thus, the width of the reader is referred to more detailed pub-
solid/electrolyte interfacial layer can be lications.
EF,n EF,n
Depletion
n-type
Semiconductor Depletion
Solution Solution
200 nm 200 nm
(a) (b)
EF,n
Depletion
Solution
20 nm
(c)
Fig. 17 Schemes of the semiconductor/electrolyte interface for a macroporous and a
nanoporous electrode. (a) An n-type macroporous electrode under moderate depletion:
structural units contain a depleted region and a bulk region (free electrons in the
nondepleted region). (b) A macroporous n-type electrode at a strongly positive potential:
the entire porous electrode is depleted of free electrons. (c) A nanoporous electrode in
which depletion occurs without band bending.
2.1.5.2.1 Penetration of the Interfacial units in macroporous GaP (doping density

Layer in a Porous Semiconductor Electrode 1017 cm−3 ) have typical dimensions of
In Fig. 17(a), the energetics of a typical about 150 nm [138, 139, 146, 147]. This
interfacial region between an n-type semi- means that, if the band bending of the
conductor and an electrolyte solution is electrode is smaller than one 1 eV, WSC
shown (see also Sect. 2.1.2.2). Electronic (50 nm, see Sect. 2.1.2.2) is smaller than
equilibrium exists between the semicon- half the width of a structure: the inside
ductor and a redox system present in the edge of the depletion layer is in the porous
solution: the electrochemical potential of network; there is still a semiconducting
electrons µe in the solid is equal to that in bulk region in the porous solid (see
the liquid phase, and does not change with Fig. 17a). Therefore, the total surface area
the spatial coordinate x, perpendicular of the inside edge of the depletion layer
to the solid/liquid interface. The electro- is very large and so is the interfacial
chemical potential of the electrons is also capacitance that is approximately equal
equal to the electron Fermi level, denoted to the internal surface area times CSC .
as EF,n and can be written as However, if the band-bending is more than
3 eV, WSC is larger than half the width of
EF,n (x) − EFref = −e[ϕ(x) − ϕ ref ]
typical structures in macroporous GaP; the
+ kT ln[n(x)/nref ] (35) entire porous GaP structure is depleted of
electrons, the inside edge of the depletion
where n(x) is the electron concentration layer is located in the bulk substrate,
in the CB of the semiconductor (i.e. the outside the porous film (Fig. 17b). The
free-electron concentration), and −eϕ(x) surface area of the inside edge of the
the potential energy of an electron. depletion layer then corresponds to the
It is clear from Fig. 17(a) that a depletion macroscopic (geometrical) area of the
layer for free electrons is present near the electrode.
solid/solution interface. From Eq. (35) it Erné and coworkers measured the in-
follows that terfacial capacitance of macroporous GaP
electrodes as a function of the electrode
n(x) = nbulk e−[Ec (x)−Ec,bulk ]/kT (36)
potential [138, 139]. It was found that
Here, EC (x) and EC,bulk is the energy of the capacitance is large for sufficiently
the CB edge at position x in the depletion small band-bending (interfacial layer in
layer and in the bulk, respectively; EC (x = the porous solid) and decreases to the
0) − EC,bulk is the band bending. The value of a nonporous interface at larger
concentration of free electrons in the band-bending. Similar effects have been
bulk, nbulk , is determined by the density found with macroporous SiC and Si elec-
of dopant atoms. The depletion layer of trodes [18, 141]. In fact, the interfacial
an n-type electrode is positively charged, capacitance is a measure for the surface
because the ionized dopant atoms are not area of the macroporous network, with the
fully compensated by free electrons. The width of the depletion layer, WSC , as a
counter charge is located on the electrolyte measuring stick.
side, very close to the interface (Helmholtz Finally, in nanoporous networks, the
layer, width 0.5 nm). structural units have dimensions in the
Consider now electronic equilibrium for range 25–1 nm. For instance, the TiO2
a macroporous electrode. The structural particulate electrode forming the basis
of a dye-sensitized solar cell consists of in microwave absorption induced by a

nanocrystals of 10–25 nm (Fig. 16b). It given change in the Fermi level observed
is clear that, under conditions of free- at macroporous GaP [150]. The large dy-
electron depletion, band bending is nearly namic range in the total number of free
absent. (Fig. 17c). When electrons are electrons in a semiconducting network can
supplied to the nanoporous system (by be used as a tool for photoelectrochemical
injection from a photoexcited dye or from characterization. For instance, the optical
a conducting substrate), the difference absorbance by free carriers in nanoporous
between EF and EC becomes smaller in TiO2 electrodes has been used to detect
the entire nanocrystal and the conductivity changes in the free carrier concentration
of the system increases. in this system due to voltage modula-
Under conditions of electron accumu- tion and modulation of the photoexcitation
lation, the interfacial capacitance C of rate [151].
a semiconductor/electrolyte contact tends Clearly, these considerations are of
to that of the Helmholtz layer (see importance for systems in which porous
Sect. 2.1.3.1 with CSC ≥ CH ). The width semiconductors or insulators are used
of the interfacial double layer is reduced [super-capacitors, (chemical) sensors, and
to about 0.5 nm; hence, it follows the in- electrochromic devices] [152, 153].
ternal surface of a porous electrode. As a
result, the overall interfacial capacitance of 2.1.5.2.3 Charge Storage in a Quantum Dot
a nanoporous system can be huge, being System Hoyer and coworkers [154, 155]
determined by the product of the total in- reported that the electrochemical poten-
ternal surface area of the system and the tial of a porous particulate ZnO electrode
Helmholtz-capacitance per unit geometric (with ZnO dots of 5-nm diameter) shifts
surface area [148, 149]. to higher energy with increasing electron
density n in a much more pronounced
2.1.5.2.2 Charge Storage in a Porous Semi- way than predicted from Eq. (35). This is
conductor Electrode In a bulk single- caused by two physical phenomena that
crystal electrode with a flat semiconduc- become important with very small par-
tor/electrolyte interface, the electrochemi- ticles (quantum dots). First, as a result
cal potential can be changed considerably of size-quantization, the energy levels of
by a relatively small change in the number the CB (and valence band) become dis-
of electrons present in the semiconductor. crete and separated by considerable energy
This is due to the relatively small inter- gaps (typically in the range of 0.1 eV).
facial capacitance per unit of geometric Even in assemblies in which the nanocrys-
surface area. In contrast, the entire three- tals are covalently linked, size-quantization
dimensional structure is interfacial in the may persist. Second, the charging en-
case of macroporous and nanoporous sys- ergy per particle, e2 /Cparticle , (typically
tems interpenetrated with a conducting 0.01–0.1 eV) can form an important con-
electrolyte. Consequently, the capacitance tribution to the electrochemical potential
per geometric area can be very large, of nanometer-sized particles. Investigating
which means that a relatively large num- nanoparticulate ZnO electrodes similar to
ber of electrons are needed for a change those used by Hoyer, Meulenkamp [156]
in the electrochemical potential. This is reported that the relationship between the
exemplified by the considerable variation electrochemical potential and the electron
density depends on the nature of the coefficient of the light can be estimated
electrolyte solution; this clearly shows from effective medium theory.
the importance of the charging energy.
Size-quantization and single-dot charging 2.1.5.2.5 Electron-hole Photoexcitation by
energy play an important role in electron Sub-band Gap Light In a dye-sensitized
transport in metal–nanodot–metal double porous photoelectrode, an electron from
barrier tunnel junctions [157–161]. Study the dye is photoexcited into the CB by a
of electron transport in two-dimensional photon of energy considerably below the
or three-dimensional assemblies consist- band gap of the semiconductor. The dye
ing of nanometer-sized particles is still in molecules are anchored on the internal
its infancy; one may expect that single-dot surface of the porous semiconductor. Light
charging (leading to Coulomb-blockade) absorption is very effective because of
and size-quantization will result in inter- multiple interactions of a single photon
esting and novel transport phenomena. with the dye molecules. Similarly, a porous
semiconductor without dye molecules may
2.1.5.2.4 Light Scattering in Macroporous absorb sub-band gap light, and this may
Semiconductors The dimensions of the lead to photogeneration of free electrons
structural units and pores in macro- and holes. The mechanisms of free carrier
porous semiconducting and insulating generation with sub-band gap light in
networks are often in the 100-nm range. macroporous GaP photoelectrodes have
This is the same range as the wave- been investigated in detail [166, 167].
length of visible and UV light. Because Surface-localized electrons involved in
of the structural variation of the refrac- two-photon transitions and in a coupled
tive index on the wavelength scale, visible optical-thermal transition were found to
light can be strongly scattered in macro- give rise to significant sub-band gap
porous networks [162–165]. In macrop- photocurrent in this system.
orous GaP, for instance, the propagation
of red (sub–band gap) light is strongly 2.1.5.2.6 Effective Electron-hole Separation
attenuated [146, 147]. The importance for The ability of porous photoelectrochemi-
photoelectrochemical systems lies in the cal systems to separate effectively electrons
fact that the effective absorption length of and holes is widely known since the pre-
supra–band gap light in macroporous sys- sentation of the dye-sensitized particulate
tems is reduced considerably with respect TiO2 solar cell [16, 105, 130–137]. In this
to that in single crystals. For example, the system, the photocurrent quantum yield
penetration depth 1/α of green light in a (the number of electrons counted in the
bulk GaP single crystal is about 10 µm, external circuit as photocurrent divided by
whereas macroporous GaP networks with the number of absorbed photons) is close
a thickness of only 2 µm completely absorb to unity. This means that electron-hole pair
green light [138, 139]. Visible and near- recombination is essentially absent. Effi-
UV light is not scattered in nanoporous cient separation of photogenerated elec-
systems because the structural variation trons and holes was demonstrated with
in the refractive index occurs on a scale several other photoelectrochemical sys-
much smaller than the wavelength of light. tems [105, 130–137]. Photovoltaic devices
In such a case, the effective absorption based on permeated hole-conducting and
electron-conducting polymer phases also reach the surface without recombining.

show an enhanced photocurrent quantum If surface recombination is slower than
yield [168–170]. The origin of this desir- transfer of the hole to the liquid electrolyte
able feature can be demonstrated by com- phase, the quantum yield will be close to
paring the photocurrent quantum yield of one, in agreement with the experimen-
a GaP bulk single crystal with a macro- tal result. A considerable enhancement
porous GaP photoelectrode (see Fig. 18). of the photocurrent quantum yield has
GaP absorbs light of energy between 2.2 been observed in several porous photo-
and 2.7 eV by an indirect transition (hence, electrochemical [16, 18, 105, 130–139] and
weakly, the absorption depth of green light photovoltaic systems [168–170].
is 10 µm). Because of the diffusion length
Lmin of minority carriers in n-GaP is rela- 2.1.5.2.7 Luminescence from Porous Elec-
tively small (about 50 nm), the penetration trodes As in bulk systems, photogenera-
depth 1/α is much larger than the width tion of charge carriers in a porous electrode
of the retrieval region (Lmin + WSC ). This or injection of minority carriers from so-
results in a photocurrent quantum yield lution can lead to light emission [25]. In a
of about 0.01 [∼ = α(Lmin + WSC )]; that is, macroporous system in which the deple-
99% of the absorbed photons are con- tion layer can follow the contours of the
verted into heat by recombination in the porous matrix, one does not expect signifi-
bulk. A macroporous GaP electrode, on the cant differences between bulk and porous
other hand, shows a photocurrent quan- electrodes with regard to the potential de-
tum yield of unity in a large potential pendence of the emission. If, however, the
range. The reason for this spectacular en- porosity is high and the dimensions of the
hancement is illustrated in Fig. 17(b): in structures become very small (e.g. <5 nm)
a macroporous GaP electrode, all minor- then special effects may be expected. These
ity carriers (holes in the valence band) are are indeed found as, for example, with
photogenerated within a distance from the nanoporous silicon.
interface that is smaller than the diffusion One of the most striking proper-
length Lmin . Minority carriers can thus ties of nanoporous silicon is its strong
1.0
(b)
0.8 Fig. 18 Comparison of the
photocurrent quantum yield
0.6 [Q = jPH /e(1 − R)%] versus the
(a) wavelength of the incident light,
Q
measured with a nonporous n-type

0.4
single-crystal electrode of GaP (a), and
with a macroporous GaP electrode (b).
0.2 With a nonporous electrode the
quantum yield is very low for light
0.0 absorbed in the indirect transition
2.0 2.5 3.0 3.5 4.0 (hν < 2.7 eV). In contrast, for a porous
electrode the quantum yield is unity for
hν light of energy above the gap
[eV] (hν > 2.2 eV) (from Ref. 139).
photoluminescence in the visible spectral with respect to the flat band potential
range [171]. Of the various explanations (Fig. 19) [172, 173]. Because of the absence
given for this phenomenon, the most of an electric field in the porous layer and
widely accepted is that of size quanti- the strong confinement of the carriers, the
zation. As a result of the confinement electron and hole are not separated in the
of charge carriers within nanometer-sized potential range corresponding to depletion
structures in the porous matrix, the ef- in a bulk electrode. As at a single-crystal
fective band gap is widened while the electrode, hole injection from a strongly
oxidizing species such as SO4 − (gener-
•
ratio of radiative to nonradiative recom-
bination is considerably enhanced [171]. ated electrochemically by the reduction of
In situ luminescence measurements pro- S2 O8 2− at the electrode) gives rise to vis-
vide information about the physical and ible electroluminescence in porous n-type
chemical properties of porous silicon and silicon [172, 173]. The emission increases
about charge-transfer reactions at the sili- in the range in which the photolumines-
con/solution interface [25]. cence decreases [172, 173] (see Fig. 19). An
In contrast to the photoluminescence interesting aspect of the electrolumines-
from a single-crystal electrode (see Fig. 11), cence is the voltage tunability of the colour.
the emission from a porous n-type silicon The emission maximum shifts to shorter
electrode is constant at positive potentials wavelength as the potential is scanned to
and decreases only in the range negative negative values, until finally the emission
1.0
EL
PL
0.8
Intensity [a.u.]
0.6
0.4 700 nm
750 nm
800 nm
0.2
0.0
−0.8 −1.0 −1.2 −1.4

V [V vs SCE]
Fig. 19 The potential dependence of the emission intensity from a porous
n-type silicon electrode in H2 SO4 solution. Three emission wavelengths are
shown. For the PL measurements an argon-ion laser was used as excitation
source. EL was excited by reduction of peroxydisulfate, added to the H2 SO4
solution. Note, the potential scale is reversed in this figure (from Ref. 173).
is quenched. EL can be expected from a A wide range of luminescence effects

quantized structure if it is populated by have been reported for porous silicon;
an electron supplied from the bulk silicon. these have been reviewed by Kelly and
Because of the larger band gap of porous coworkers [25]. During anodic oxidation,
silicon there is a mismatch of the CB edges. porous p-type silicon shows electrolumi-
As a result the Fermi energy of the bulk nescence [174, 175], similar to that de-
silicon must be raised to a level close to scribed for p-type InP in Sect. 2.1.3.3.2. In
the conduction band edge of the quantized the case of porous silicon, however, a very
structure. This will occur for larger parti- strong light emission is observed in the vis-
cles at less negative potential because the ible spectral range (because of size quanti-
mismatch of the CB edges is small in this zation). On excitation of the bulk substrate
case. As a result, long wavelength emis- with near-infrared light, porous n-type sil-
sion will be turned on first. Gradually as icon can be photoanodically oxidized; this
the Fermi level is raised further, that is, as process is also accompanied by strong
the potential is made more negative, the emission of visible light [176]. These re-
smaller particles will participate and the sults provide insight into the mechanism
emission maximum will shift to shorter of anodic oxidation of the porous semicon-
wavelength. Quenching of EL and PL has ductor.
been attributed to Auger recombination.
This can explain why the rise in EL on 2.1.5.3 Electron Transport
going to negative potentials is coupled to 2.1.5.3.1 Electron Diffusion, Collection,
the quenching of PL. To give EL, an elec- and Recombination A unique feature of
tron is required in a particle to create a a porous photoelectrochemical system is
hole via S2 O8 2− reduction (compare with the permeation of the solid semiconduct-
Eqs. 23 and 24). On the other hand, pho- ing network by an electrolyte solution on
toexcitation of a particle already occupied a scale smaller than Lmin (Fig. 17b). As a
by an electron leads to Auger recombina- result, one of the photogenerated charge
tion; that is, the PL is quenched. At more carriers can be transferred to the solu-
negative potentials, the supply of an elec- tion. In the following, we will assume that
tron to a particle in which an electron-hole the hole is transferred to the electrolyte
pair is present, leads to Auger quenching solution and oxidizes a reduced species.
of the EL [172, 173]. Photogenerated electrons are left in the
EC e e e
Fig. 20 Energy scheme for
photogeneration and diffusion of
electrons in a porous photoelectrode
Energy
under steady state conditions. The light

EF h ν is incident from the electrolyte side
(x = 0) and is absorbed in a region of
width 1/α. Photogenerated electrons
diffuse toward the collecting back
contact (x = d) caused by a gradient in
EV the electrochemical potential EF
h
x (dashed line). Transport is attenuated
d 1/ α 0 by multiple trapping/detrapping.
solid network and diffuse, because of a time that electrons need to travel through
gradient of the electrochemical potential, the system before collection, that is, the
through the network over a considerable transit time τtran (d) is in the millisec-
distance toward the metal contact, where ond to second range [179–183, 187, 188].
they are collected (Fig. 20). The thickness, As a result, photogenerated electrons can
d, of a porous electrode and, hence, the be lost before collection, by transfer to
length of the electron pathway is between the oxidized species in the solution, a
1 and 100 µm. process characterized by a time con-
The transport of electrons in porous stant τrec . Electron back-transfer forms
semiconductors is of interest to a wide an important recombination process in
audience, not only to electrochemists. On dye-sensitized photoelectrochemical sys-
one hand, porous semiconducting and in- tems [16, 179–183].
sulating networks can be considered as
disordered systems, which show a strong 2.1.5.3.2 Characterization of Electron Dif-
resemblance to amorphous semiconduc- fusion and Back-transfer by Light Intensity
tors [177, 178]. In porous semiconductors, Modulated Techniques In Sect. 2.1.2, it
diffusing electrons can be scattered not was shown that time-resolved methods
only by the lattice but also by the surface of are required to obtain information on
the matrix and by grain boundaries. Scat- the kinetics and dynamics of a system.
tering is a friction phenomenon reducing The measurement of the photocurrent re-
the free-electron mobility µ. In addition, sponse upon a small modulation of the
porous networks have a large interfacial light intensity is a very effective method to
area. Therefore, a huge volume density of study electron dynamics in a porous sys-
interfacial electronic states, distributed in tem [184–186, 187, 188]. Neglecting trap-
the band gap, can be expected. The volume ping/detrapping of diffusing electrons and
density of band gap states can be much assuming the electrochemical potential
larger than that of a macroscopic crystal gradient in the porous system to be in-
and comparable to that of an amorphous dependent of the spatial coordinate x (see
semiconductor [177, 178]. A diffusing elec- Fig. 20), Vanmaekelbergh and coworkers
tron can be trapped in a state in the gap and demonstrated that the optoelectrical trans-
hence, become temporarily localized. The fer function may be written as a function
electron is promoted back into the CB by of τtran en τrec [189, 190]:
thermal excitation (trapping/detrapping).
It is clear that electron scattering and #j˜PH (ω)
multiple trapping/detrapping are different e(1 − R)#%̃(ω)
physical processes; scattering reduces the 1 − e−iωτtran (d) e−τtran (d)/τrec
electron flux in the system by reducing = (37)
iωτtran (d) + [τtran (d)/τrec ]
the mobility of the diffusing free electrons,
whereas trapping decreases the electron Plots of the optoelectrical transfer func-
flux by reducing the density of free elec- tion in the complex plane are presented in
trons. Fig. 21. Attention is drawn to two limiting
Many recent experimental results show cases. If the transit time of photogener-
that electron transport through a porous ated electrons τtran (d) through the porous
semiconducting network is a slow pro- network is much smaller than the electron
cess [179–183]. For instance, the average lifetime τrec , photogenerated electrons will
0.0
τrec / τd = 0.1
1/ τ rec
−0.2
Im [ j(ω)/e Φ(ω )]
−0.4 τrec / τd = 1
−0.6 1/ τ d
τrec / τd = 100
−0.2 0.0 0.2 0.4 0.6 0.8 1.0

Re [ j(ω)/e Φ(ω)]
Fig. 21 Plots of the calculated optoelectrical transfer function (Eq. 37) in the
complex plane for three ratios of the electron lifetime τrec to transit time τd .
For τrec /τd = 100, all electrons are collected, the transfer function is typical
for electron transport; the characteristic frequency gives the transit time. For
τrec /τd = 0.1, only 10% of the photogenerated electrons are collected: the
characteristic frequency gives the recombination lifetime.
reach the collecting contact without being recombination is dominant (τtran (d) +
lost; the collection efficiency is nearly one, τrec ), the optoelectrical transfer function
and thus also the low-frequency limit of the reduces to
transfer function. In this case (Fig. 21), the #j˜(ω) τrec /τtran (d)
optoelectrical transfer function describes = (39)
e(1 − R)#%̃(ω) iωτrec + 1
electron propagation. The shape of the
function is determined by electron diffu- This transfer function corresponds to a
sion only and enters, surprisingly, the third semicircle in the complex plane (Fig. 21).
quadrant of the complex plane at high fre- The recombination lifetime follows from
quencies. The average transit time through the characteristic frequency of the transfer
the porous network follows from the angu- function: 1/ωm = τrec ; the transit time can
lar frequency ωm at which the imaginary then be extracted from the low-frequency
part of the optoelectrical transfer function limit given by
shows its first minimum (characteristic ∂jphoto τrec
frequency, starting from zero frequency): = (40)
(1 − R)e ∂φ τtran (d)
3
τtran (d) = (38) Including trapping/detrapping in the
ωm model that leads to the transfer func-
It is easily inferred from Eq. (38) that the tion expressed by Eq. (37) does not alter
factor e−τtran (d)/τrec damps the oscillating the optoelectrical transfer function itself.
function e−iωτtran (d) . As a result, the However, the transit time that is ob-
optoelectrical transfer function becomes tained from ωm is increased by a factor
increasingly semicircular if τtran (d)/τrec [1 + 0(EF,n )]. The value of the trapping
increases toward and above unity. When parameter 0(EF,n ) (with respect to 1)
determines to what extent multiple trap- in the electrolyte) in nanoporous TiO2

ping or detrapping attenuates electron photoelectrodes has been studied ex-
diffusion: electron transport is trap-limited tensively [184–186, 187, 188, 191–194,
if 0(EF,n ) + 1. It was found that the trap- 195]. It has been found that the transit
ping parameter is given by the ratio of time through such an electrode depends
the overall volume density of states near strongly on the background light inten-
the Fermi level and the average density of sity; several research groups found that
free electrons in the porous system for a τtran (d) ∝ %−0.7 . In other words, the effec-
given position of the electron Fermi level tive diffusion coefficient Dn (EF,n ) is pro-
[n(EF,n )]: portional to %0.7 . Values between 10−8 and
& 10−4 cm−2 s−1 were found for Dn (EF,n )
kT sv,i (EF,n )
for typical light intensities between 1010
i
0= (41) and 1015 cm−2 s−1 . These very low val-
n(EF,n )
' ues and their dependence on the light
In Eq. (41), i sv,i (E) denotes the volume intensity (hence, on the position of the
density of band gap states per unit energy, electron Fermi level in the band gap)
at a given energy E. In a porous sys- show that electron diffusion is strongly at-
tem, this quantity can be easily between tenuated by trapping. In the framework
1018 and 1020 cm−3 eV−1 ; it is expected presented earlier (trap-limited electron
that 0 is considerably above unity. This transport), the effective diffusion coeffi-
means that transport will be trap-limited. cient reads
In a porous photoelectrochemical system,
µ n(EF,n )
n(EF,n ) increases linearly with increasing Dn (EF,n ) = & (42)
background light intensity %. Hence, we e sv,i (EF,n )
predict that the trapping parameter 0 and i
thus, the attenuation of electron transport where µ is free-electron mobility. A
caused by trapping will strongly decrease density-of-state function, which increases
with increasing light intensity. This has exponentially with increasing energy in
been observed with several porous pho- the band gap has been derived from
toelectrochemical systems [184–186, 187, the earlier relationship. The photovoltage
188] (see following text). The fact that response upon modulation of the light
the transit time decreases with increasing intensity (with a given background light
background light intensity also shows that intensity) has been used to study the
attenuation of electron transport is due to recombination kinetics in dye-sensitized
trapping, not to scattering. Furthermore, TiO2 solar cells [191–195]. It was found
the experimental relationship between the that the back-transfer of the electrons from
transit time τtran (d) and the background the TiO2 network to I3 − is extremely slow.
light intensity % can be used' to map the This contrasts with the fast reduction of
density of state function kT i sv,i (E) in a this species at Pt, where chemisorbed I3 −
certain energy range of the band gap. plays an essential role. A detailed study of
the modulated photovoltage as a function
2.1.5.3.3 Experimental Results on Elec- of the background light intensity revealed
tron Transport Electron transport and that τrec ∝ 1/n(EF,n ) (this means that the
recombination (corresponding to electron rate of back-transfer is second order in
back-transfer to the oxidized species I3 − the concentration of electrons) [196]. Peter
and coworkers proposed a mechanism for Macroporous GaP is an ideal model

this multistep electron transfer process system for several reasons: (1) the porous
in which the radical anion I2 − plays an
•
network is prepared by anodic etching
essential role. of a single crystal; this gives a very
Electron diffusion through macroporous good reproducibility in the preparation of
GaP electrodes has also been studied. the samples, (2) the porous network is a
0.0
−0.1
−0.2
−0.3
ωm
−0.4
−0.5
−0.6
−0.7
−0.2 0.0 0.2 0.4 0.6 0.8 1.0
(a) Real [ j(ω)/e Φ(ω)]

Imaginary [ j(ω )/e Φ(ω)]
0.0
−0.1
−0.2
−0.3
ωm
−0.4
−0.5
−0.6
−0.7
10−2 10−1 100 101 102 103
Modulation frequency
(b) [Hz]
Electrolyte solution is 0.5M H2SO4 Fig. 22 (a) Complex-plane
200 Porosity is 25% representation of the measured
Trapping level density: 2kBTΣS(EF)
optoelectrical transfer function (points)

for a porous GaP electrode (40-µm
150 thick) permeated with an aqueous
10 µm H2 SO4 solution. The background light
20 µm intensity is 2 × 1014 cm−2 s−1 . (b) Plot
[1015/cm3]
40 µm of the imaginary part (points) as a

100 100 µm function of the modulation frequency.
10 µm
20 µm Full and dashed lines in (a) and (b)
40 µm represent calculated plots: the dashed
plot corresponds to Eq. (37); the full
50
plot takes into account a Gaussian
distribution around the average transit
time. (c) Density of state function
0 versus the energy in the band gap for
300 400 500 600 700 800 macroporous GaP. The data are
obtained from the trap-limited transit
EF-EC time (3/ωm ) measured with different
(c) [meV] background light intensities, (see text).
single crystal without grain boundaries, 6. A. B. Ellis, J. M. Bolts, M. S. Wrighton,

and (3) the thickness of the porous film J. Electrochem. Soc. 1977, 124, 1603.
7. G. Hodes, J. Manassen, D. Cahen, Nature
and hence, the length of the electron
1976, 261, 403.
pathway can be varied at will between 20 8. B. Miller, A. Heller, Nature 1976, 262, 680.
and 200 µm. This enables an additional 9. S. Licht, G. Hodes, R. Tenne et al., Nature
check of the theoretical understanding 1987, 330, 148.
of electron transport. The optoelectrical 10. B. Wang, S. Licht, T. Soga et al., Solar En-
ergy Mater. Sol. Cells 2000, 64, 311.
transfer function measured with a 40- 11. S. Licht, H. Tributsch, B. Wang et al.,
µm porous GaP photoelectrode is shown J. Phys. Chem. B 2000, 104, 8920.
in Fig. 22 [188]. The transfer function 12. S. Licht, J. Phys. Chem. 1986, 90, 1096.
is presented in the complex plane; the 13. S. Licht, D. Peramunage, Nature 1990, 345,
evolution of the imaginary component as 330.
14. S. Licht, O. Khaselev, T. Soga et al., Elec-
a function of the modulation frequency is trochem. Sol. State Lett. 1998, 1, 20.
also shown. It is clear that the results are 15. P. H. L. Notten, J. E. A. M. van den Meer-
in agreement with the model presented akker, J. J. Kelly, in Etching of III–V Semi-
earlier (Eq. 37) if a Gaussian distribution conductors, An Electrochemical Approach, El-
sevier Advanced Technology, Oxford, 1991.
around the average transit time of the
16. B. O’Regan, M. Grätzel, Nature 1991, 353,
electrons is assumed. At a background 737.
light intensity % of 2 × 1014 cm−2 s−1 , the 17. L. T. Canham, Appl. Phys. Lett. 1990, 57,
transit time τtran (d) = 3/ωm is 1 s, that is, 1046.
a factor 105 larger than that expected if one 18. J. J. Kelly, D. Vanmaekelbergh, Porous-
etched Semiconductors: Formation and
assumes that the mobility of the diffusing Characterization, in The Electrochemistry
electrons is the same as in bulk GaP. Thus, of Nanostructures: Preparation and Proper-
electron transport is strongly attenuated by ties, Chap. 4 (Eds.: G. Hodes), Wiley-VCH,
the effect of multiple trapping. The fact that Weinheim, 2001.
τtran (d) decreases strongly with increasing 19. S. Ogawa, F. F. Ran, A. J. Bard, J. Phys.
Chem. B 1995, 99, 11182.
background light intensity shows that 20. S. Ogawa, K. Hu, F. F. Ran et al., J. Phys.
trapping/detrapping in band gap states Chem. B 1997, 101, 5707.
close to the electron Fermi level attenuates 21. S. R. Morrison, in Electrochemistry at Semi-
diffusion. A detailed analysis of the results conductor and Oxidized Metal Electrodes,
Plenum Press, New York, 1980.
obtained with porous GaP electrodes of
22. J. Gobrecht, H. Gerischer, Solar Energy
different thickness led to a mapping of the Mater. 1979, 2, 131.
density-of-state function in the upper half 23. R. Memming, Electrochim. Acta 1980, 25,
of the band gap (Fig. 22) [188]. 77.
24. R. Memming, in Semiconductor Electrochem-
istry, Wiley-VHC, Weinheim, 2000.
References 25. J. J. Kelly, E. S. Kooij, E. A. Meulenkamp,
Electrochim. Acta 1999, 45, 561.
1. A. Uhlir, Bell Systems Techn. J. 1956, 35, 333. 26. E. P. A. M. Bakkers, E. Reitsma, J. J. Kelly
2. H. Gerischer, F. Beck, Z. Phys. Chem. N.F. et al., J.Phys. Chem. B 1999, 103, 2781.
1957, 13, 389. 27. J. R. Macdonald in Impedance Spectroscopy,
3. D. R. Turner, J. Electrochem. Soc. 1958, 105, Emphasizing Solid Materials and Systems,
402. John Wiley & Sons, New York, 1987.
4. A. Fujishima, K. Honda, Bull. Chem. Soc. 28. R. Unbehauen in Systemtheorie 1, Allgemeine
Jpn. 1971, 44, 1148. Grundlagen, Signale und lineare Systeme im
5. A. B. Ellis, S. W. Kaiser, M. S. Wrighton, Zeit- und Frequenzbereich R. Oldenbourg
J. Am. Chem. Soc. 1976, 98, 1635. Verlag, München, 1997.
29. W. Gärtner, Phys. Rev. 1959, 116, 84. 54. B. H. Erné, D. Vanmaekelbergh, I. E. Ver-
30. L. M. Peter, Chem. Rev. 1990, 90, 753. meir, Electrochim. Acta 1993, 38, 2559.
31. B. Smandek, G. Chmiel, H. Gerischer, Ber. 55. D. Vanmaekelbergh, F. Cardon, J. Phys. D:
Bunsen-Ges. Phys. Chem. 1989, 93, 1094. Appl. Phys. 1986, 19, 643.
32. G. Chmiel, H. Gerischer, J. Phys. Chem. 56. D. Vanmaekelbergh, F. Cardon, Semicond.
1990, 94, 1612. Sci. Technol. 1988, 3, 124.
33. K. H. Beckmann, R. Memming, J. Elec- 57. D. Vanmaekelbergh, F. Cardon, Electrochim.
trochem. Soc. 1969, 116, 368. Acta 1992, 37, 837.
34. H. Gerischer, Z. Phys. Chem. 1960, 26, 325. 58. Z. Hens, W. P. Gomes, J. Phys. Chem. B
35. H. Gerischer, Z. Phys. Chem. 1961, 27, 48. 1997, 101, 5814.
36. H. Gerischer, Semiconductor Electrochem- 59. Z. Hens, W. P. Gomes, J. Electroanal. Chem.
istry, in Physical Chemistry, an Advanced 1997, 437, 77.
Treatise, Vol. IXA/Electrochemistry, Aca- 60. Z. Hens, J. Phys. Chem. B 1999, 103, 122.
demic Press, New York, 1970. 61. Z. Hens, W. P. Gomes, J. Phys. Chem. B
37. W. P. Gomes, F. Cardon, in Progress in Sur- 1999, 103, 130.
face Science (Eds.: S. G. Davison), Pergamon 62. D. Vanmaekelbergh, A. R. de Wit, F. Car-
Press, 1982, p. 155, Vol. 12. don, J. Appl. Phys. 1993, 73, 5049.
38. J. O’ M. Bockris, A. K. N. Reddy in Modern 63. A. A. Burk, P. B. Johnson, W. S. Hobson
Electrochemistry, Plenum Press, New York, et al., J. Appl. Phys. 1986, 59, 1621.
1977, Vol. 2. 64. A. B. Ellis, Luminescent Properties of Semi-
39. Y. V. Pleskov, Y. Y. Gurevich in Semicon- conductor Electrodes, in Chemistry and
ductor Photoelectrochemistry, Consultants Structure of Interfaces, New Laser and Op-
Bureau, New York, 1986. tical Techniques, Chapter 6 (Eds.: R. B. Hall,
40. R. Degryse, W. P. Gomes, F. Cardon et al., A. B. Ellis), VCH, Deerfield Beach (FL),
J. Electrochem. Soc. 1975, 122, 711. 1986.
41. T. Pajkossy, J. Electroanal. Chem. 1994, 364, 65. B. A. Balko, G. L. Richmond, J. Phys. Chem.
111. 1993, 97, 9002.
42. J. F. McCann, S. P. S. Baldwall, J. Elec- 66. O. Kruger, Ch. Jung, Ber. Bunsen-Ges. Phys.
trochem. Soc. 1982, 129, 551. Chem. 1994, 98, 1022.
43. Z. Hens, W. P. Gomes, Phys. Chem. Chem. 67. J. F. Kauffman, C. S. Liu, M. W. Karl,
Phys. 1999, 1, 3607. J. Phys. Chem. 1998, 102, 6766.
44. Z. Hens, W. P. Gomes, Phys. Chem. Chem. 68. G. N. Ryba, C. N. Kenyon, N. S. Lewis,
Phys. 1999, 1, 3617. J. Phys. Chem. 1993, 97, 13814.
45. K. Schröder, R. Memming, Ber. Bunsen-Ges. 69. A. Manivannan, K. Hashimoto, A. Fuji-
Phys. Chem. 1985, 89, 385. shima, J. Phys. Chem. 1992, 96, 3766.
46. D. Vanmaekelbergh, W. P. Gomes, Ber. 70. A. Manivannan, K. Itoh, K. Hashimoto
Bunsen-Ges. Phys. Chem. 1985, 89, 994. et al., J. Electrochem. Soc. 1990, 137, 3121.
47. J. E. A. M. van den Meerakker, J. J. Kelly, 71. J. Ouyang, F.-R. Fan, A. J. Bard, J. Elec-
P. H. L. Notten, J. Electrochem. Soc. 1985, trochem. Soc. 1989, 136, 1033.
132, 638. 72. F. Decker, F. Prince, P. Motisuke, J. Appl.
48. J. E. A. M. van den Meerakker, Electrochim. Phys. 1985, 57, 2900.
Acta 1985, 30, 435. 73. B. H. Erné, F. Ozanam, J.-N. Chazalviel,
49. D. Vanmaekelbergh, W. P. Gomes, F. Phys. Rev. Lett. 1998, 80, 4337.
Cardon, Ber. Bunsen-Ges. Phys. Chem. 1986, 74. S. R. Morrison, T. Freund, J. Chem. Phys.
90, 431. 1967, 47, 1543.
50. D. Lincot, J. Vedel, J. Electroanal. Chem. 75. W. P. Gomes, T. Freund, S. R. Morrison,
1987, 220, 179. J. Electrochem. Soc. 1968, 115, 818.
51. D. Vanmaekelbergh, W. P. Gomes, F. 76. A. Fujishima, T. Kato, E. Maekawa et al.,
Cardon, J. Electrochem. Soc. 1987, 134, 891. Bull. Chem. Soc. Jpn. 1981, 54, 1671.
52. D. Vanmaekelbergh, R. P. ter Heide, W. 77. G. H. Schoenmakers, D. Vanmaekelbergh,
Kruijt, Ber. Bunsen-Ges. Phys. Chem. 1989, J. J. Kelly, J. Chem. Soc., Faraday Trans.
93, 1103. 1997, 93, 1127.
53. Z. Hens, W. P. Gomes, J. Phys. Chem. B 78. J. R. Harbor, M. L. Hair, J. Phys. Chem.
2000, 104, 7725. 1979, 83, 652.
79. G. Nogami, J. H. Kennedy, J. Electrochem. 104. R. W. Berriman, P. B. Gilman, Photogr. Sci.

Soc. 1989, 136, 2583. Eng. 1973, 17, 235.
80. P. Herrasti, L. Peter, J. Electroanal. Chem. 105. J. J. Kelly, D. Vanmaekelbergh, Electrochim.
1991, 305, 241. Acta 1998, 43, 2773.
81. P. Bogdanoff, N. Alonso-Vante, J. Elec- 106. J. W. Gerritsen, E. J. G. Boon, G. Janssens
troanal. Chem. 1994, 379, 415. et al., Appl. Phys. A 1998, 66, S79.
82. R. Memming, J. Electrochem. Soc. 1969, 116, 107 M. Kawasaki, H. Ishii, J. Imaging Sci. Tech-
785. nol. 1995, 39, 210.
83. J. J. Kelly, B. P. Minks, N. A. M. Verhaegh 108. T. Hannappel, B. Burfeindt, W. Storck
et al., Electrochim. Acta 1992, 37, 909. et al., J. Phys. Chem. B 1997, 101, 6799.
84. B. P. Minks, G. Oskam, D. Vanmaekelbergh 109. Y. Tachibana, J. E. Moser, M. Graetzel
et al., J. Electroanal. Chem. 1989, 273, 119. et al., J. Phys. Chem. 1996, 100, 20056.
85. B. P. Minks, D. Vanmaekelbergh, J. J. Kelly, 110. A. Henglein, Top. Curr. Chem. 1988, 143,
J. Electroanal. Chem. 1989, 273, 133. 115.
86. R. Peat, L. M. Peter, Electrochim. Acta 1986, 111. H. Weller, Adv. Mater. 1993, 5, 88.
31, 731. 112. A. P. Alivisatos, J. Phys. Chem. 1996, 100,
87. R. Peat, L. M. Peter, J. Electroanal. Chem. 13226.
1986, 209, 307. 113. A. N. Shipway, E. Katz, I. Willner, Chem.
88. W. P. Gomes, H. H. Goossens, Elec- Phys. Chem. 2000, 1, 18.
trochemistry of III–V Compound Semicon- 114. B. Alperson, H. Demange, I. Rubinstein
ductors: Dissolution Kinetics and Etching in et al., J. Phys. Chem. B 1999, 103, 4943.
Advances in Electrochemical Science and Engi- 115. R. L. Smith, S. D. Collins, J. Appl. Phys.
neering, (Eds.: H. Gerischer, C. W. Tobias), 1992, 71, R1.
VCH, Weinheim, 1994, Vol. 3. 116. V. L. Colvin, A. N. Goldstein, A. P.
89. D. Vanmaekelbergh, J. J. Kelly, J. Elec- Alivisatos, J. Am. Chem. Soc. 1992, 114,
trochem. Soc. 1989, 136, 108. 5221.
90. W. P. Gomes, S. Lingier, D. Vanmaekel- 117. M. Miyake, H. Matsumoto, M. Nishizawa
bergh, J. Electroanal. Chem. 1989, 269, 237. et al., Langmuir 1997, 13, 742.
91. M. Matsumura, S. R. Morrison, J. Elec- 118. T. Nakanishi, B. Ohtani, K. Uosaki, J. Phys.
troanal. Chem. 1983, 147, 159. Chem. B 1998, 102, 1571.
92. L. M.Peter, A. M. Borazio, H. J. Lewerenz 119. K. Hu, M. Brust, A. J. Bard, Chem. Mater.
et al., J. Electroanal. Chem. 1990, 290, 229. 1998, 10, 1160.
93. G. H. Schoenmakers, R. Waagenaar, J. J. 120. S. Drouard, S. G. Hickey, J. D. Riley, Chem.
Kelly, J. Electrochem. Soc. 1995, 142, L60. Commun. 1999, 1, 67.
94. S. Cattarin, L. M. Peter, D. J. Riley, J. Phys. 121. E. P. A. M. Bakkers in Charge transfer be-
Chem. B 1997, 101, 4071. tween semiconductor nanocrystals and a metal,
95. X. Xia, C. M. A. Ashruf, P. J. French et al., Ph. D. Thesis, University of Utrecht, 2000.
J. Phys. Chem. B 2001, 105, 5722. 122. E. P. A. M. Bakkers, J. J. Kelly, D. Vanmae-
96. B. H. Erné, D. Vanmaekelbergh, J. Elec- kelbergh, J. Electroanal. Chem. 2000, 482,
trochem. Soc. 1997, 144, 3385. 48.
97. D. Vanmaekelbergh, P. C. Searson, J. Elec- 123. E. P. A. M. Bakkers, E. Reitsma, J. J. Kelly,
trochem. Soc. 1994, 141, 697. D. Vanmaekelbergh, J. Phys. Chem. B 1999,
98. B. H. Erné, F. Ozanam, M. Stchakovsky 103, 2781.
et al., J. Phys. Chem. B 2000, 104, 5961. 124. S. G. Hickey, D. J. Riley, Electrochim. Acta
99. B. H. Erné, F. Ozanam, M. Stchakovsky 2000, 45, 3277.
et al., J. Phys. Chem. B 2000, 104, 5974. 125. W. J. Albery, G. T. Brown, J. R. Darwent
100. P. Allongue, H. Brune, H. Gerischer, Surf. et al., J. Chem. Soc. Faraday Trans. 1 1985,
Sci. 1992, 275, 414. 81, 1999.
101. P. Allongue, V. Costa Gieling, H. Gerischer, 126. E. P. A. M. Bakkers, A. W. Marsman, L. W.
J. Electrochem. Soc. 1993, 140, 1009. Jenneskens et al., Angew. Chem. Int. Ed.
102. P. Allongue, Phys. Rev. Lett. 1996, 77, 1986. 2000, 39, 2297.
103. W. West, B. H. Carrol in The Theory of 127. E. P. A. M. Bakkers, A. L. Roest, A. W.
Photographic Processes, (Eds.: C. E. K. Mees, Marsman et al., J. Phys. Chem. B 2000, 104,
T. H. James), McMillan, New York, 1966. 7266.
128. S. G. Hickey, J. D. Riley, E. Tull, J. Phys. 153. D. Cummins, G. Boschloo, M. Ryan et al.,
Chem. B 2000, 104, 7623. J. Phys. Chem. B 2000, 104, 11449.
129. M. R. Hoffmann, S. T. Martin, W. Choi 154. P. Hoyer, R. Eichberger, H. Weller, Ber.
et al., Chem. Rev. 1995, 95, 69. Bunsen-Ges. Phys. Chem. 1993, 97, 630.
130. G. Hodes, J. Electrochem. Soc. 1992, 139, 155. P. Hoyer, H. Weller, J. Phys. Chem. 1995,
3136. 99, 14096.
131. R. Grünwald, H. Tributsch, J. Phys. Chem. 156. E. A. Meulenkamp, J. Phys. Chem. B 1999,
B 1997, 101, 2564. 103, 7831.
132. S. Y. Huang, J. Phys. Chem. B 1997, 101, 157. D. L. Klein, P. L. McEuen, J. E. B. Katari
2576. et al., Appl. Phys. Lett. 1996, 68, 2574.
133. D. Liu, R. W. Fessenden, G. L. Hug et al., 158. D. L. Klein, R. Roth, A. K. L. Lim et al., Na-
J. Phys. Chem. B 1997, 101, 2583. ture 1997, 389, 699.
134. R. Argazzi, C. A. Bignozzi, T. A. Heimer 159. U. Banin, Y. Cao, D. Katz et al., Nature
et al., J. Phys. Chem. B 1997, 101, 2591. 1999, 400, 542.
135. H. Rensmo, K. Keis, H. Lindstroem et al., 160. O. Millo, D. Katz, Y. Cao et al., Phys. Rev. B
J. Phys. Chem. B 1997, 101, 2598. 2000, 61, 16773.
136. A. Wahl, J. Augustynski, J. Phys. Chem. B 161. E. P. A. M. Bakkers, D. Vanmaekelbergh,
1998, 102, 7820. Phys. Rev. B 2000, 62, R7743.
137. W. Kubo, K. Murakoshi, T. Kitamura et al., 162. J. D. Joannopoulos, P. R. Villeneuve, S. Fan,
Chem. Lett. 1998, 12, 1241. Nature 1997, 386, 143.
138. B. H. Erné, D. Vanmaekelbergh, J. J. Kelly, 163. U. Grüning, V. Lehmann, C. M. Engelhardt,
Adv. Mater. 1995, 7, 739. Appl. Phys. Lett. 1995, 66, 3254.
139. B. H. Erné, D. Vanmaekelbergh, J. J. Kelly, 164. J. E. G. J. Wijnhoven, W. L. Vos, Science
1998, 281, 802.
J. Electrochem. Soc. 1996, 143, 305.
165. Y. A. Vlasov, N. Yao, D. J. Norris, Adv.
140. G. Oskam, A. Natarajan, P. C. Searson,
Mater. 1999, 11, 165.
F. M. Ross, Appl. Surf. Sci. 1997, 119, 160.
166. D. Vanmaekelbergh, L. van Pieterson, Phys.
141. J. van de Lagemaat, M. Plakman, D.
Rev. Lett. 1998, 80, 82.
Vanmaekelbergh et al., Appl. Phys. Lett.
167. D. Vanmaekelbergh, M. A. Hamstra, L. van
1996, 69, 2801.
Pieterson, J. Phys. Chem. B 1998, 102, 7997.
142. T. Sugira, T. Yoshida, H. Minoura, Elec-
168. J. J. M. Halls, C. A. Walsh, N. C. Greenham
trochem. Solid-State Lett. 1998, 1, 175. et al., Nature 1995, 376, 498.
143. A. O. Konstantinov, C. I. Harris, E. Janzén, 169. G. Yu, J. Gao, J. C. Hummelen et al., Sci-
Appl. Phys. Lett. 1994, 65, 2699. ence 1995, 270, 1789.
144. A. O. Konstantinov, A. Henry, C. I. Harris 170. U. Bach, D. Lupo, P. Comte et al., Nature
et al., Appl. Phys. Lett. 1995, 66, 2250. 1998, 395, 583.
145. C. Zener, Proc. R. Soc. London 1934, 145, 171. A. G. Cullis, L. T. Canham, P. D. J. Calcott,
523. J. Appl. Phys. 1997, 82, 909.
146. F. J. P. Schuurmans, D. Vanmaekelbergh, 172. E. A. Meulenkamp, L. M. Peter, D. J. Riley
J. Van de Lagemaat et al., Science 1999, 284, et al., J. Electroanal. Chem. 1995, 392, 97.
141. 173. L. M. Peter, D. J. Riley, R. I. Wielgosz et al.,
147. F. J. P. Schuurmans, M. Megens, D. Thin Solid Films 1996, 276, 123.
Vanmaekelbergh et al., Phys. Rev. Lett. 1999, 174. A. Bsiesy, F. Gaspard, R. Herino et al.,
83, 2183. J. Electrochem. Soc. 1991, 138, 3450.
148. L. M. Peter, D. J. Riley, R. I. Wielgosz, Appl. 175. M.Ligeon, F. Muller, R. Herino et al.,
Phys. Lett. 1995, 66, 2355. J. Appl. Phys. 1993, 74, 1265.
149. L. M. Peter, R. I. Wielgosz, Appl. Phys. Lett. 176. E. S. Kooij, A. R. Rama, J. J. Kelly, Surf. Sci.
1996, 69, 806. 1997, 370, 125.
150. W. H. Lubberhuizen, D. Vanmaekelbergh, 177. R. A. Street, Cambridge Solid State Science
E. Van Faassen, Journal of Porous Materials Series, in Hydrogenated Amorphous Silicon,
2000, 7, 147. (Eds.: R. W. Cahn, E. A. Davis, I. M. Ward),
151. B. Enright, B. G. Redmond, D. Fitzmaurice, 1991.
J. Phys. Chem. 1994, 98, 6195. 178. H. Scher, M. F. Shlesinger, J. T. Bendler,
152. M. Gratzel, Nature 2001, 409, 575. Phys. Today 1991, 44, 26.
179. L. M. Peter, D. Vanmaekelbergh in Advances 188. A. L. Roest, D. Vanmaekelbergh, Phys. Rev.

in Electrochemical Science and Engineering, B 2000, 62, 16926.
(Eds.: R. C. Alkire, D. M. Kolb), Wiley VCH, 189. D. Vanmaekelbergh, P. E. de Jongh, Phys.
Weinheim, 1999, Vol. 6. Rev. B 2000, 61, 4699.
180. K. Schwarzburg, F. Willig, Appl. Phys. Lett. 190. P. E. de Jongh in Photoelectrochemistry of
1991, 58, 2520. nanoporous semiconductor electrodes, Ph.D.
181. P. Hoyer, H. Weller, J. Phys. Chem. 1995, Thesis,. University of Utrecht, 1999.
99, 14096. 191. A. C. Fisher, L. M. Peter, E. A. Ponomarev
182. A. Solbrand, H. Lindstroem, H. Rensmo et al., J. Phys. Chem. B 2000, 104, 949.
et al., J. Phys. Chem. B 1997, 101, 2514. 192. J. van de Lagemaat, N. G. Park, A. J. Frank,
183. F. Cao, G. Oskam, G. J. Meyer et al., J. Phys. J. Phys. Chem. B 2000, 104, 2044.
Chem. 1996, 100, 17021. 193. N. Kopidakis, E. A. Schiff, N. G. Park et al.,
184. P. E. de Jongh, D. Vanmaekelbergh, Phys. J. Phys. Chem. B 2000, 104, 3930.
Rev. Lett. 1996, 77, 3427. 194. J. van de Lagemaat, A. Frank., J. Phys. Chem.
185. L. Dloczik, O. Ileperuma, I. Lauermann B 2000, 104, 4292.
et al., J. Phys. Chem. B 1997, 101, 10281. 195. N. W. Duffy, L. M. Peter, R. M. G. Raja-
186. P. E. de Jongh, D. Vanmaekelbergh, J. Phys. pakse et al., J. Phys. Chem. B 2000, 104,
Chem. B 1997, 101, 2716. 8916.
187. D. Vanmaekelbergh, F. Iranzo Marı́n, J. van 196. G. Schlichthörl, S. Y. Huang, J. Sprague
de Lagemaat, Ber. Bunsen-Ges. Phys. Chem. et al., J. Phys. Chem. B 1997, 101, 8141.
1996, 100, 616.
2.2 Si substrate, the method to reduce disloca-

Deposition of (Multiple Junction) tion density, the method to passivate the
Semiconductor Surfaces defects in GaAs on Si, and the application
to photovoltaic solar cell are presented. Al-
Tetsuo Soga
though highly mismatched systems such
Nagoya Institute of Technology, Nagoya,
as GaAs-on-Si, AlGaAs-on-Si, and GaN-on-
Japan
Si are interesting for device applications,
2.2.1 the study on the crystal growth of GaP on Si
Introduction substrate is interesting for the fundamen-
tal understanding of the III–V compounds
A limited fraction of incident solar pho- on Si substrate. It is expected that the
tons have sufficient (greater than band understanding of the growth mechanism
gap) energy to initiate charge excitation will lead us to achieve the growth of low-
and separation within a semiconductor. dislocation-density III–V compounds on
Wide band gap semiconductor solar cells Si substrate.
are capable of generating a high photovolt- The deposition of III–V compounds has
age but have a low limiting photocurrent recently been actively performed. Espe-
caused by the low fraction of short wave- cially, the crystal growth on Si substrate
length light in the solar spectrum. Smaller has attracted attention since high-quality
band gap cells can utilize a larger fraction GaAs layers were successfully grown on Si
of the incident photons but generate lower substrates in 1984 [3–7]. Although various
photovoltage. Multijunction devices, also devices such as lasers [8] and solar cells [9]
referred to as tandem, multiple window, have been fabricated on Si substrate, the
split spectrum, and cascade solar cells, device characteristics are not satisfactory
can overcome this limitation [1]. A vari- because the existence of a high density
ety of multiple semiconductor layers have of the threading dislocations in the epi-
been fabricated and explored for photo- taxial layer. The reduction of dislocation
voltaic applications. These include GaInP density is an important issue to obtain
on GaAs, GaAs on Si, GaAs on GaSb, a high-performance compound semicon-
InP on GaInAs, GaAs on GaInAsP, etc [2]. ductor device on an Si substrate. The
Through a detailed specific example of one dislocations are generated due to lattice
of these important multiple semiconduc- mismatch, thermal expansion mismatch,
tor systems consisting of various GaAs the crystal structure difference, the gener-
layers on various Si layers, this paper de- ation of APD, surface contamination, and
scribes preparation and optimization of so on.
multijunction systems that are capable of The threading dislocations in the epi-
efficient photon absorption, charge sepa- taxial layer on Si substrate are classi-
ration, and charge transfer. fied into (1) dislocations that originate
This section reviews the growth process from the dislocations in the Si substrate,
of III–V compound semiconductor on Si (2) dislocations generated by the coales-
substrate grown by metalorganic chemical cence of the islands at the initial stage,
vapor deposition (MOCVD). The nucle- (3) dislocations generated by the lattice
ation of GaP and GaAs on Si substrate, mismatch, and (4) dislocations generated
the dislocation generation mechanism, an- by the thermal stress during the cooling
tiphase domain (APD) structure of GaP on stage from the growth temperature.
2.2 Deposition of (Multiple Junction) Semiconductor Surfaces 107
Among these four types of dislocations, diethylzinc (DEZ), AsH3 and H2 Se, re-
we do not need to take into account spectively. The Si substrate orientation was
item (1) because the dislocation density (001) 2–4◦ tilted toward [110] direction. Si
of Si is generally very low. In order to substrate was rinsed in organic solvents,
reduce the dislocation density according followed by the repetition of the oxidation
to items (2), (3), and (4), many efforts by H2 SO4 : H2 O = 4 : 1 and the removal
have been made. The efforts to change of the oxides by 25% HF solution. After
the initial growth mode from three- loading the substrate into the reactor, the
dimensional (3D) to two-dimensional (2D) substrate was heated at 1000 ◦ C for 10 min
have been made to avoid the dislocations at hydrogen atmosphere. The V/III ratio
generated by the coalescence of the islands was varied from 100 to 6400 with chang-
at the initial stage of the growth [10]. ing the PH3 flow rate, keeping the TMG
Although many methods using strained flow rate constant for the growth of GaP
layer superlattice (SLS) buffer layer [3, on Si substrate. The gas pressure was
11, 12], rapid thermal annealing [13], varied from 76 to 380 torr. The epitaxial
thermal cycle annealing (TCA) [14, 15], layer thickness was varied from 20 nm to
and so on have been adopted to reduce 3.7 µm. The growth temperature was kept
the dislocation density of item (3), the constant at 900 ◦ C. 3D growth was not ob-
dislocation density is still on the order of served under these growth conditions for
106 cm−2 . The low temperature growth the case of GaP growth on GaP substrate.
has been investigated to decrease the The samples were examined using No-
generation of the dislocation by the marski optical microscopy, cross-sectional
thermal stress [16, 17]. It is expected that transmission electron microscopy (TEM),
the number of dislocations generated by and so on. GaAs was grown on Si sub-
the thermal stress is reduced when the strate by the two-step growth method with
growth temperature is low. Although the 10-nm-thick GaAs buffer layer grown at
dislocation density of GaAs-on-Si on the 400 ◦ C.
order of 104 cm−2 has been obtained at
the growth temperature of 350 ◦ C, thermal 2.2.3
stability is the problem. Deposition of Gallium Phosphide on Silicon
Substrate
2.2.2
MOCVD 2.2.3.1 Nucleation of GaP on Si Substrate
In general, the growth mode is divided
The epitaxial growth was performed us- into three categories, namely, 2D type,
ing low-pressure or atmospheric pressure Volmer-Weber (3D) type, and Stranski-
MOCVD. The former consists of lamp- Krastanov (2D + 3D) type. In the case of
heated and the latter consists of rf-heated the 2D mode, the dislocations are gener-
horizontal reactor with load lock chamber. ated when the layer thickness exceeds the
The substrate was put on the SiC-coated critical thickness [18]. On the other hand,
carbon susceptor and the temperature was the dislocations and stacking faults are
controlled by the thermocouple inserted generated at the coalescence of the islands
into the susceptor. Source gases for Al, formed at the initial stage of the growth
Ga, Zn, As and Se are trimethylallu- in the case of the 3D mode [19]. There-
minum (TMA), trimethylgallium (TMG), fore, the defect density will be significantly
Tab. 1 Material parameters of III–V compound

semiconductors and Si
GaP GaAs InP Si
Lattice constant (Å) 5.45 5.65 5.87 5.43

Linear thermal expansion 5.9 6.8 4.6 2.6
coefficient (×10−6 K−1 )
Crystal structure ZB ZB ZB Diamond
Note: ZB: Zinc Blend.
reduced if a 2D growth can be realized on the growth mode is expected to be

from the beginning of the growth. minimized and other effects such as the
The 3D growth mode has origins not surface migration effect, surface contam-
only in the basic material property differ- ination, polar or nonpolar structure are
ences between the epitaxial layer and the emphasized.
substrate (e.g. lattice constant mismatch, Figure 1 shows the growth mode of GaP-
polar or nonpolar effect, etc.) but also in on-Si for various gas pressure and the V/III
the growth conditions. ratio [20]. All the surface morphologies of
The material properties for various samples were classified into three-types,
III–V compounds and Si are shown in namely, island-type, mixture-type (mixture
Table 1. As shown in this table, the lattice of island and layer-type), and layer-type.
constant of GaP is closest to that of Si. It is indicated that the growth mode of
Therefore, the effect of lattice mismatch GaP changes from island-type to layer-
type with increasing V/III ratio. The V/III
ratio at which the growth mode changes
from island-type to layer-type decreases
with increasing gas pressure. A very high
V/III ratio of 3200 is necessary to obtain
a GaP layer without island-type crystal at
3200
the gas pressure of 76 torr [8]. Island-type
Layer-type
growth is clearly demonstrated, and it is
1600
observed that the islands are not connected
V/III ratio
by a GaP layer between themselves at the

800
edge of the island. This means that the
Mixture-type
growth mode of GaP on Si substrate under
400
these growth conditions is not Stranski-
200 Island-type Krastanov type but Volmer-Weber type.
Faceting was observed at the boundary
100 either on (111) or (211) type planes.
The island formation is interesting
because these islands are not formed in
0 76 152 380
Fig. 1 Growth mode of GaP on Si
Gas pressure substrate for various V/III ratio and gas
[Torr] pressure.
the case of homoepitaxy; it is particular Therefore, in the case of the deposition

for the heteroepitaxy. In the case of of Ga on Si, the migrating species are
the homoepitaxial growth, the source supposed to be Gax -type clusters. It is
gases are usually incorporated into the expected that the molecular mass of a
step edge or terrace of the misoriented cluster increases during the surface migra-
substrate. However, the island spacing tion. Because the island density increases
is several orders of magnitude larger gradually with increasing V/III ratio, the
than that of the average step distance migration species would be a Gax Py -type
of the misoriented substrate. Therefore, cluster [21]. When the PH3 flow rate is
the nucleation site is not governed by the increased, the number of decomposed P
substrate steps. Furthermore, the residual atoms is increased, and a high density of
oxide or impurity is not the nucleation site Gax Py -type cluster with large mass is eas-
because the GaP island density changes ily formed. Clusters with large molecular
drastically with the growth conditions [21]. mass are expected to migrate more slowly
In the nucleation process for GaP on than those with smaller masses. When the
Si substrate, the diffusion of the growth cluster size exceeds the critical size, the
species through the boundary layer, the clusters are deposited on the Si surface
surface migration, and the nucleation at and the islands are formed.
the nucleation sites should be taken into Moreover, when the concentration of P
account. Comparing the heteroepitaxy of atoms on the Si surface is high, P atoms
GaP-on-Si and the homoepitaxy, there absorbed on the Si surface are increased.
should be no difference in the diffusion It results in the formation of the flat
process if the growth conditions are the layer because the P atoms absorbed on
same. Therefore, it is deduced that the Si capture the migrating species.
difference between the homoepitaxy and To explain the gas pressure dependence
the heteroepitaxy on Si is caused by the of the growth mode, other factors in
difference in the migration on the Si the reactor must be considered. A pos-
substrate. The migration length of the sible factor affected by the gas pressure
migrating species on Si is considered to is the flow velocity in the reactor. TMG
be longer than for homoepitaxy. This is is almost completely decomposed at the
due to the weak interaction between Si growth temperature [25]. In contrast, the
and Ga or P atoms. An example of strong decomposition of PH3 varies with the
atomic interaction is the growth of a III–V flow velocity because the decomposition
compound semiconductor containing Al rate for PH3 is not as fast as TMG [26].
on Si substrate. It has been reported These results are qualitatively explained
that AlGaP [22], AlGaAs [23], and AlAs [24] as follows. It is evident that the flow
layers grown on Si substrate are flat from velocity is increased when the pressure
the beginning of growth. is decreased. The pyrolysis of PH3 takes
Before considering the migrating species place in the heated region in the reactor.
for GaP on Si, the surface migration of Therefore, the decrease of pressure makes
Ga is discussed. Usually, the migration the PH3 decomposition difficult because
species for the deposition of Ga films on the resident time for the source gases in
Si substrate is the cluster of Ga atoms. the heated region becomes short, result-
TMG is probably perfectly decomposed ing in the atomic arrangement as shown
to Ga and a metal-radical at 900 ◦ C [25]. in Fig. 2(a). So the higher V/III ratio is
Ga
P
GaP Si
Si
(a)
GaP
Si
(b)
Fig. 2Schematic atomic arrangement of 3D GaP on Si (a) and 2D
GaP on Si (b).
necessary for lower growth pressure to velocity v is proportional to the total flow
produce a layer-type growth as shown in rate (F ) and inverse of the growth pressure
Fig. 2(b). Pg , that is,
In the growth of GaP, a simple reaction F
v=b (4)
involving PH3 is considered to be the Pg
decomposition of PH3 and the production
and Eq. (3) is changed into
of phosphorus molecules, that is, # $
Pg
PH3 = P + 3H (1) pPH3 = pPH30 exp −c (5)
F
Using simple reaction kinetics and where c is a modified constant including
solving the differential equations, that is, a, b, and x. On the other hand, using these
d[PH3 ] notations, V/III ratio is given by
= a[PH3 ] (2)
dt pPH30
the partial pressures of PH3 in these cases [V/III] = (6)
pTMGa
are given by
! x" The ratio of the phosphorus concen-
pPH3 = pPH30 exp −a (3) tration([P]) to the gallium concentration
v
([Ga]), which is available to the growth on
where pPH3 is the partial pressure of PH3
the substrate surface is expressed as,
at the distance x from the entrance of
the heated zone in the flow direction with [P/Ga] = [P]/[Ga] (7)
the flow velocity v, pPH30 is the initial
partial pressure of PH3 or the supplied where [P] and [Ga] are proportional to
partial pressure into the reactor, and a the concentration of the decomposed PH3
is a constant including the reaction rate and TMGa, respectively. [Ga] is supposed
constant, the cross-sectional area of the to be constant on the growing surface
reactor and other parameters. The flow for various growth pressures because the
TMGa is almost decomposed at the growth In the TEM picture with the thickness
temperature. Thus, substituting the partial of 3.7 µm, two kinds of dislocations with
pressures and the relation for the V/III an extra-half plane in the Si substrate
ratio into Eq. (6), the above ratio is given by (A-type) and the GaP layer (B-type) are
observed [28]. The schematic models of
(pPH30 − pPH3 )
[P/Ga] = a = a[V/III] atomic arrangements for these structures
pTMGa are shown in Fig. 3. In general, the
% # $&
Pg dislocation generation should take place
× 1 − exp −c (8)
F at random in the isotropic crystal [29], that
is, four kinds of A-type dislocations with
where a is the proportionality constant. Burgers’ vector of 1/2[1 0 1], 1/2[0−11],
Therefore, the initial growth mode of GaP, 1/2[0 1 1] and 1/2[−101] can be generated.
which is governed by [P]/[Ga] ratio is However, one salient feature came to be
expressed using the V/III ratio, Pg and seen; instead of these four directional
F . From these results, it can be said that A-type dislocation, only two directional A-
at lower growth pressure, a higher V/III type dislocations with Burgers vector of
ratio is required for the layer growth. 1/2[1 0 1] or 1/2[0−11] are observed. A
possible interpretation for this generation
2.2.3.2 Generation of Dislocation of A-type dislocation with only two-kinds
In the heteroepitaxial growth, the gener- of Burgers’ vectors instead of the four
ation of misfit dislocation and the stress previously reported types of dislocations
relaxation are related to each other. The is the difference of the situation of the site,
misfit dislocation generation and the stress that is, dislocations are created at the step
relaxation of GaP layer on Si substrate
grown under high V/III ratio (layer-type
growth mode) are described. Because the
defects associated with the coalescence of GaP
the islands are not generated, the observed
dislocations are generated after the layer
thickness exceeds the critical thickness.
From the cross-sectional TEM micro- Si
graph, it can be seen that the GaP surface Si
is very flat, and defects such as disloca- P
A-type
tions or structural defects are not observed Ga
at all when the thickness is thinner than
90 nm [27]. This means that GaP grows on
Si coherently with compressive stress at
the initial stage. The TEM measurement GaP
for GaP on Si with various layer thickness
shows that the dislocations at the interface
are observed when the GaP layer thickness
exceeds 90 nm. Si
Fig. 3 Atomic arrangement for GaP on

Si with A-type and B-type dislocation. B-type
edge of the misoriented Si substrate rather lattice strain of GaP is relaxed completely
than randomly isotropic generation. The at the growth temperature by introducing
direction of Burgers’ vector for the B-type A-type misfit dislocations, the tensile stress
dislocations is the reverse of that of the is produced in the GaP layer during the
A-type dislocation. cooling process because of the difference of
Assuming that the strain is completely the thermal expansion coefficients of GaP
accommodated at the growth temperature and Si. The thermal expansion coefficient
by 60◦ dislocations, the spacing between of GaP is about 2.5 times larger than that
dislocations corresponds to 110 nm. How- of Si. In order to relax the tensile stress
ever, the spacing between A-type disloca- in the GaP layer, the dislocations with the
tions is much smaller than the calculated extra-half plane in the GaP layer should
value. This difference might be due to the be introduced. Accordingly, it is proved
uniformity of the Si substrate steps. that the dislocations with the extra-half
If the dislocations are generated by plane in the Si substrate are formed by
the lattice mismatch, an extra-half plane the lattice mismatch and that those in the
should occur in the Si substrate, which GaP layer are formed during the cooling
has a small lattice constant [29]. Because process to relax the thermal stress. This
both the room temperature lattice constant experiment supports the report that proved
and the thermal expansion coefficient of the generation of threading dislocations
GaP are larger than those of Si, GaP in GaP on Si during the cooling down
should have the larger lattice constant process [30].
than Si at the growth temperature. Hence, The stress applied to the GaP layer as
the lattice mismatch relaxation at the a function of the thickness measured by
growth temperature is responsible for the X-ray diffraction is shown in Fig. 4 [20].
generation of the A-type dislocation. The The dotted line shows the stress value for
B-type dislocations cannot be explained the thermal stress calculated by the bimetal
merely on the grounds of lattice mismatch. model. A thin GaP layer has a compres-
Because no dislocation is generated at sive stress, and the stress changes to tensile
the initial stage of the growth, the B-type with increasing thickness. Considering the
dislocations should be generated during thermal stress between the growth temper-
the growth or cooling down process. If the ature and room temperature, the GaP layer
Compressive
1
[× 109 dyn cm−2]
0
Stress
1
2
Tensile
0 1 2 3 4
Fig. 4 Stress of GaP layer on Si
Thickness substrate as a function of the
[µm] thickness.
Fig. 5 APD structure and single Si [001]

domain structure at different step P
height. Ga −
[110] [110]
must have compressive stress at growth

temperature when the layer thickness is GaP
thinner than the critical thickness. Because
the strain energy, which is caused by the
misfit, increases with the thickness, the Si
misfit dislocations are generated with in- Antiphase domain
creasing thickness. The change of stress
in GaP on Si from compressive to tensile
with the thickness is due to the increase of
the misfit dislocation density at the growth GaP
temperature. The strain is smaller than
the calculated thermal stress because the
thermal stress is partly relaxed by the gen- Si
eration of misfit dislocations with the extra
Single domain
half plane in the GaP layer.
Although the dislocation is not gener-
ated at the initial stage of the growth, the conditions using (002) (a) and (002̄) (b)
dislocations with the extra-half plane in the reflections. In these figures, the contrasts
Si substrate are generated at the growth of the domain region and matrix region are
temperature to relax the lattice mismatch. inverted by changing the reflection vector
During the cooling down process, the ther- g from 002 to 002̄. The amplitudes of 002
mal stress is relaxed by generating the and 002̄ reflections have been calculated to
misfit dislocation with the extra-half plane unequal for most thicknesses in the case
in the GaP layer. Below the dislocation of zincblende structure [31, 32]. Therefore,
frozen temperature, the thermal expansion it is proved that the antiphase boundary
mismatch produces the tensile stress in is normal to the (001) plane near the Si
substrate. In most cases, the APD is anni-
the epitaxial layer without generating new
hilated on changing the orientation of the
dislocations, resulting in the large tensile
boundary from the (001) normal to higher
stress at room temperature.
index planes so as to minimize the total
energy.
2.2.3.3 Annihilation of APD Structure The mechanisms for generation and
The problem of the APD occurs at the Si annihilation of APDs are discussed. The
surface step as a result of the polar or Si surface steps are usually composed of
nonpolar structure. As shown in Fig. 5, single and double atomic steps. The initial
the APDs are generated when the Si has growth mode of GaP on Si is 2D as it
a single (or odd) atomic step, whereas the is grown under a high V/III ratio. In the
single domains are formed when the Si initial stage of growth, the Si substrate
surface is double (or even) atomic steps. is covered with P under a high PH3
The typical APD is shown in Fig. 6. rate. Therefore, APD is introduced at the
These were taken under dark field single atomic step position. In the case of
Fig. 6 Dark field TEM micrograph of APD

002 annihilated during the growth. The
reflection vector is (002) (a) and (00-2) (b).
GaP
If the size of APD introduced in

the initial stage is small, the APD is
Si
annihilated in the early stage of growth.
On the other hand, if the APD is large, a
thick layer is necessary for all the APD to
(a) be annihilated. Therefore, APD remains
at the surface. The observation that the
− APD is observed in GaP on 2◦ off (001)
002 Si indicates the existence of single atomic
GaP
steps. However, the density of single steps
is lower than that of the exactly (001) Si
substrate. On the other hand, all the steps
Si change to double atomic steps in the cases
of 4◦ -off and 6◦ -off substrates after the
annealing process. This is inferred from
the fact that the APD is not detected in
(b) 0.1 µm these samples.
the growth on only (001) Si, the spacing 2.2.4

between the steps is assumed to be large Deposition of Gallium Arsenide on Silicon
compared with the misoriented substrate. Substrate
Therefore, the size of the APD is large.
On the other hand, the size of the APD 2.2.4.1 Nucleation
is small in the case of the misoriented For lattice mismatched III–V semiconduc-
substrate. tors on Si, two kinds of misfit dislocations
Calculation shows that the antiphase are observed; one is the pure-edge Lomer
boundary for the (211) and (110) antiphase misfit dislocation, whose Burgers vector is
boundaries are energetically more likely parallel to the interface (type-I dislocation),
to form than those for the (111) or (100) and the other is the misfit dislocation,
planes [33]. Therefore, the appearance of whose Burgers vector is 60◦ from the
the (110) antiphase boundary is in good dislocation line (60◦ dislocation or type-
agreement with the calculations. The total II dislocation) [34]. Schematic illustrations
energy is increased with increasing thick- of type-I and type-II dislocations are shown
ness. In order to reduce the total energy, in Fig. 7(a) and 7(b), respectively.
the antiphase boundary changes its orien- A periodic array of misfit dislocations
tation to the low-energy index plane. The with an average spacing of about 8.1 nm
higher index plane of the APD is estimated is observed and the ratio of number of
to be the (211) plane from the angle [31]. type-I to type-II dislocations is about 3 : 1
This result also supports the calculation for GaAs on Si substrate. This means that
that the energy of the antiphase boundary the lattice mismatch is completely relaxed
for (211) is smaller than that of (110). by the misfit dislocations. The majority of
Fig. 7 Schematic illustration of type-I

dislocation (a) and type-II dislocation (b).
GaAs
the dislocations are type-I dislocation, and

Si
this is also the case for materials grown by
molecular beam epitaxy [35]. Si
(a) Type I As
On the other hand, in the case of GaP
Ga
on Si, most dislocations are also type-
II dislocations. The HRTEM micrographs
show that most dislocations in GaAs on
GaP substrates are type II.
GaAs
This section discusses the generation
mechanisms for type-I and type-II dis-
locations. Although the lattice mismatch
for GaAs/Si and GaAs/GaP are almost Si
the same, the types of dislocations are
different. Therefore, the type of dislocation
cannot be explained alone by lattice mis- (b) Type II
match. Furthermore, the thermal stress
in the epitaxial layer cannot explain the
difference of the dislocation type because mismatch is large. Therefore, another dis-
the thermal stress for GaAs/Si and GaP/Si location generation mechanism should be
is almost the same although the types of considered because the dislocation gener-
dislocations are different. ation is modified by island formation. For
Some mechanisms have been proposed example, misfit dislocations are generated
to explain the generation of type-I in GaAs/Si heterostructures at the begin-
and type-II dislocations. The reported ning of growth and before the coverage
mechanisms for the generation of type-II of Si with a GaAs layer has been com-
dislocation are (1) the bending of the pleted [38, 39]. Therefore, it is deduced that
threading dislocation in the substrate par- the misfit dislocation generation is greatly
allel to the interface [36] and (2) the glide of affected by the initial growth mode.
dislocation from the surface, which forms The type of dominant dislocation, the
a half-loop [29]. The reported mechanisms lattice mismatch and the growth mode for
for the generation of type-I dislocation GaAs on Si, GaP on Si, GaAs on GaP,
are (1) the reaction of two type-II dislo- and AlGaAs on Si are summarized in
cations [37] and (2) the dislocation climb Table 2. The table shows that the type of
of the pure edge dislocation from the sur- dislocation is affected by the growth mode
face [29]. rather than the lattice mismatch; type-I
Misfit dislocation generation can be dislocations are dominant for material
explained by the earlier-mentioned mech- systems with a 3D growth mode, and
anisms when the growth mode is 2D and the type-II dislocations are dominant
the epitaxial layer is flat. On the other hand, for the material systems with a 2D
no dislocation generation mechanism has growth mode.
been reported when 3D islands are formed In the cases of GaP/Si and GaAs/GaP,
at the initial stage of growth and the lattice the type-II dislocation generation probably
Tab. 2Dominant type of dislocation, lattice The generation of type-I dislocations by

mismatch, and growth modes the climb process is enhanced by increas-
ing the point defect density. Therefore, the
Dislocation Lattice Growth type-I dislocation should be observed in
mismatch mode
(%)
GaP on Si as the density of point defects
is high in GaP grown on Si because of
GaAs on Si Type-I 4.1 3D the high growth temperature. But, type-I
AlGaAs on Si Type-I 4.2 3D dislocations are rarely observed in GaP on
GaP on Si Type-II 0.37 2D Si. Therefore, another mechanism should
GaAs on GaP Type-II 3.7 2D be considered to explain the generation of
type-I dislocations.
If the 3D islands are formed at the initial
is due to glide from the surface because
stage of the growth, the size of the island
2D growth occurs at the beginning of
increases with growth time, although the
the growth, and the dislocation density
island density is constant. This means that
of the substrate is extremely low compared
the misfit dislocations are generated while
with that of the epitaxial layer. On the
the island size is increasing because the
other hand, in the case of GaAs/Si and
spacing between islands is much larger
AlGaAs/Si, the dislocation generation can-
than that of the misfit dislocations. The
not be explained solely by the mechanisms
type-I dislocation generation mechanism
reported until now. If the type-I disloca-
for 3D growth is shown in Fig. 8 in
tions are generated after the reaction of
the case of GaAs on Si. When the
two 60◦ dislocations, the Burgers vectors
island size is smaller than the critical
should satisfy the condition:
size, the misfit strain is accommodated
a/2[011] + a/2[101̄] −−−→ a/2[110] by elastic strain. The strain energy of
(a : lattice constant) the island is raised when the island
(type II) (type II) (type I) size is increased. In order to relieve the
lattice mismatch, misfit dislocations are
However, the probability that all the generated at the edge of the island when
type-II dislocations are changed to type-I the critical island size has been exceeded
dislocations is very low. Moreover, it is im- as shown in Fig. 8. Therefore, type-I
possible to explain the generation of type-I dislocation generation is preferentially
dislocations by climb from the surface. enhanced when the initial growth mode is
GaAs
Fig. 8 Dislocation generation

Si
mechanism for GaAs on Si
substrate.
3D, and type-II dislocations are dominant Figure 10 shows the dark spot defect
when the initial growth mode is 2D. (DSD) density as a function of the number
of TCA1 for various number of TCA2. The
2.2.4.2 Effect of TCA DSD density of GaAs on Si without SLS
In0.1 Ga0.9 As/GaAs SLS was inserted in the is also shown. The DSD density decreases
GaAs layer to reduce the dislocation den- with increasing the number of TCA1 and
sity. The individual layer thickness and the TCA2. The lowest DSD density obtained in
total layer numbers of SLS are 20 nm and this study is 3.8 × 106 cm−2 . It is indicated
10, respectively. The TCA was performed that the TCA both before and after the
just before the SLS growth (TCA1) and SLS growth is more effective in reducing
after the SLS growth (TCA2). The upper the DSD density. Even if the dislocation
and lower temperatures of TCA are 900 density is reduced by TCA1, dislocations
and 300 ◦ C, respectively. Numbers of TCA are generated in the SLS by the lattice
were changed in this study. The intentional mismatch of SLS and GaAs. Although
doping was not performed in GaAs on Si. the individual layer thickness of SLS is
The sample structure is shown in Fig. 9. thinner than the critical thickness, the total
GaAs (1.8 µm)
TCA 2
InGaAs (20 nm)/GaAs (20 nm) SLS
TCA 1
GaAs (1.0 µm)
Si
Fig. 9 GaAs on Si substrate

using InGaAs/GaAs SLS buffer
layer.
108
Without SLS
DSD Density
With SLS (TCA2 = 0)

[cm−2]
107
With SLS (TCA2 = 1)
Fig. 10 DSD density of GaAs 106

0 1 2 3 4 5 6
on Si as a function of TCA1 for
various number of TCA2. Number of TCA1
SLS thickness is thicker than the critical is due to the enhancement of dislocation
thickness [12]. It is suggested from the movement, large compressive stress, and
experimental results that the dislocations the generation of point defects at higher
generated at SLS are bended by TCA2, temperature, which in turn reduce the dis-
resulting in the low dislocation density. location density effectively.
Until now, the low etch pit density on the Figure 12 shows the minority carrier
order of 106 cm−2 has been obtained using lifetime of GaAs, Al0.15 Ga0.85 As and
SLS and TCA for the total epitaxial layer Al0.22 Ga0.78 As grown on Si for various
thickness of more than 3.5 µm [40–43]. TCA temperatures. The minority carrier
Few papers have been reported on the lifetime is also improved with increasing
growth of GaAs on Si, with the dislocation the TCA temperature, which is supported
density of 106 cm−2 at the epitaxial layer by the decrease of DSD at high temper-
thickness of less than 3 µm. ature. The minority carrier lifetime of
The crystal quality of GaAs with 900 ◦ C GaAs grown on Si with 1000 ◦ C TCA is
TCA and SLS is inferior to that on a GaAs 3.36 ns. Those for GaAs and AlGaAs for
substrate because a high density of dislo- various Al compositions, are shown in
cations is generated in the epitaxial layer, Fig. 13. The lifetime of GaAs grown on
which degrades the minority carrier life- GaAs substrate is also plotted for compar-
time. The TCA temperature was optimized ison [44–48]. Although it is impossible to
to improve the minority carrier lifetime. compare the lifetime because the carrier
In this experiment, SLS buffer later was concentration is not the same for all the
not used. samples, it is estimated that the lifetime of
Figure 11 shows the DSD density of GaAs and AlGaAs on Si is approximately
GaAs grown on Si substrate revealed by one order shorter than those grown on
electron beam–induced current (EBIC) GaAs substrate.
measurement for various TCA temper-
ature. The DSD density decreases with 2.2.4.3 Effect of Hydrogenation
increasing the TCA temperature gradu- Hydrogenation was carried out in a
ally and is on the order of 106 cm−2 at quartz tube, where a hydrogen plasma
1000 ◦ C. The crystal quality improvement was excited by rf power via a copper
by the relatively high TCA temperature coil encircling the quartz tube. The
108
Dark spot density
[cm−2]
107
106
900 950 1000
Fig. 11 DSD density of GaAs
TCA Temperature on Si as a function of TCA
[°C] temperature.
Fig. 12 Minority carrier lifetime 4

of GaAs and AlGaAs for various
TCA temperature.
x=0
3
Minority carrier lifetime

[ns]
2
x = 0.15
1
x = 0.22
0
850 900 950 1000
TCA Temperature
[°C]
100
Ahrenkiel et al. (1988)
'tHooft et al. (1981)
Timmons et al. (1990) on GaAs
Zarem et al. (1989)
Minority carrier lifetime τ
Ahrenkiel et al. (1991)

GaAs and AlGaAs on Si with 1000 °C TCA
10
[ns]
0 0.1 0.2 0.3 0.4

x in Alx Ga1 − xAs
Fig. 13 Minority carrier lifetime of GaAs on Si with 1000 ◦ C TCA.
plasma power, the treatment time, and the at various temperatures ranging from
substrate temperature during the plasma 350 ◦ C to 450 ◦ C for 10 minutes. The TCA
treatment were 90 W, 2 hours, and 250 ◦ C, temperature was 900 ◦ C and the SLS buffer
respectively. In order to recover shallow layer was not used.
level passivation and the damage induced Carrier concentration profiles of unin-
by the plasma treatment, post annealing tentionally doped GaAs grown on Si sub-
was performed in an AsH3 + H2 ambient strates, before and after hydrogenation, are
1018
As-grown
Carrier concentration
1017
[cm−3 ]
Hydrogenated
1016
1015
0 1 2 3 4
Depth
[µm]
Fig. 14 Carrier concentration profile of GaAs on Si with and without
hydrogenation.
shown in Fig. 14. Undoped GaAs-on-Si is Figure 15 shows the 4.2 K PL spectra
n-type (1 × 1017 cm−3 ) because of Si auto- of GaAs on Si for as-grown sample and
doping during the growth [49]. For the hydrogenated sample, hydrogenated sam-
hydrogenated sample, the carrier concen- ple annealed at 450 ◦ C. The major peaks
tration is reduced to about 3 × 1016 cm−3 are peak B corresponding to the heavy
at depth exceeding 1 µm. This is due to the hole-associated free exciton and peak C
electrical passivation of the shallow levels. corresponding to the carbon impurity-
Because the major donor in GaAs grown bound exciton. After the sample is treated
on Si substrates is Si via auto-doping from by the hydrogen plasma, the full width
the substrate, passivation will occur by the at a half maximum (FWHM) of peak B
formation of SiH0 complexes via the reac- narrows from 4.49 meV to 3.83 meV. This
tion narrowing is due to the passivation of local-
ized states. With 450 ◦ C annealing where
Si+ + H0 + e− −−−→ SiH0
the shallow level is completely recovered,
The SiH0 complex will then be dissociated the FWHM is a little narrower than that of
by heat or applied electric fields. There is a the as-grown sample.
kink in the concentration profile curve at a Figure 16 shows the minority carrier life-
depth of nearly 0.8 µm, which corresponds time derived from the time resolved pho-
to the plasma-induced damage [50], as the toluminescence decay curve. The minority
knee goes deeper with increased plasma carrier lifetime increases from 1.66 ns
treatment time. After a 10-minute anneal- (as-grown) to 4.66 ns after the plasma
ing at 450 ◦ C in AsH3 + H2 ambient, the treatment, and gradually decreases with
donor electrical activities were completely increasing annealing temperature. It is dif-
restored to their initial levels. ficult to judge the crystal quality only by the
4.2 K
A B C D E
4.19 meV
Hydrogenated +
PL Intensity
annealed (450 °C)

[a.u.]
3.83 meV
Hydrogenated
4.49 meV
As-grown
800 820 840 860 880

Wavelength
[nm]
Fig. 15 4.2 K PL spectra of GaAs on Si with and without hydrogenation.
Fig. 16 Minority carrier lifetime of 5

hydrogenated GaAs on Si with various
postannealing temperature.
4
minority carrier lifetime because the life-
Minority carrier lifetime
time is also affected by the shallow carrier

concentration. In Fig. 14, the shallow car- 3
rier concentration of the 450 ◦ C annealed
[ns]
sample is the same as that of the as-

2
grown sample because with passivation the
As-grown
shallow level is completely restored. The
minority carrier lifetime after hydrogen 1
plasma treatment followed by annealing at
450 ◦ C (2.27 ns) is longer than that of the
as-grown sample (1.66 ns), suggesting that 0
Hydrogenated 350 450
the defects generated by lattice mismatch Annealing temperature
and thermal expansion mismatch are [°C]
electrically and optically passivated. The
longer minority carrier lifetime of the hy- passivated by hydrogenation. The passi-
drogenated sample (before annealing) is vation effect remains even after 450 ◦ C
due to defect passivation and shallow level annealing, where the shallow level is com-
passivation. Species in the plasma include pletely restored.
free radicals, ions, and electrons. Among From these experiments it can be
these species, free radicals can effectively concluded that (1) the shallow level that
passivate the defects. Furthermore, it is has been passivated by hydrogenation
well known that ions can damage the semi- is completely recovered by annealing at
conductor surface. Therefore, during the 450 ◦ C in AsH3 + H2 ambient, (2) the
hydrogenation process, defect passivation deep levels are still passivated by hydrogen
and damage formation take place at the after annealing at 450 ◦ C in AsH3 + H2
same time. The minority carrier lifetime ambient, and (3) hydrogenation followed
of the hydrogenated sample annealed at by 450 ◦ C annealing produces a longer
450 ◦ C is longer than that of the as-grown minority carrier lifetime at the same
sample because the defects generated dur- shallow carrier concentration.
ing the plasma treatment are passivated.
The DLTS spectra show that the peak of 2.2.4.4 Application to Photovoltaic Device
the Si-defect related level becomes smaller Figure 17 shows the schematic cross-
after hydrogen plasma treatment. It sug- sectional view of a GaAs/Si tandem solar
gests that the Si-related defect level is cell. It consists of n+ -GaAs buffer layer,
Au/AuZn
ZnS/MgF2
p + -GaAs
p +-Al0.8Ga0.2As 0.05 µm
p +-Alx Ga1−x As 0.2 µm (x : 0 → 0.29)
p +-GaAs 0.25 µm Top cell
n -GaAs 0.6 µm
n +-GaAs 1.5 µm Au/AuSb
n +-Si 1.1 µm
p -Si (350 µm) Bottom cell
p +-Si 0.5 µm
Au
Fig. 17Schematic cross-sectional view of three-terminal GaAs/Si
tandem solar cell.
n-GaAs layer, p + -GaAs, p + -Alx Ga1−x . Tab. 3 Photovoltaic properties of GaAs/Si

As graded band emitter layer (x: 0 → tandem solar cell with 1000 ◦ C TCA
0.29), p + -Al0.8 Ga0.2 . As window layer
and p-GaAs cap layer. After the epi- (mA/cm2 ) (V) (%) (%)
Jsc Voc FF η
taxial growth, electrodes of Au-Zn/Au,
Au-Sb/Au, and Au were formed on the p + - Top cell 34.8 0.90 76.8 17.7
GaAs layer, n+ -Si and p + -Si, respectively. (32.6) (0.88) (75.7) (16.1)
The surface of the cell was coated with a Bottom cell 15.0 0.52 77.2 4.4
double-layer antireflection coating (ARC) (14.8) (0.52) (78.3) (4.5)
using ZnS (49 nm) and MgF2 (71 nm). The Total 22.1
total area and active area of the solar cell (20.6)
are 25 mm2 and 22.05 mm2 , respectively.
The photovoltaic measurements were per- Note: Values in ( ) show those with 900 ◦ C TCA.
formed under AM0 and 1sun conditions
at 27 ◦ C. GaAs top cell (η = 17.7%) and Si bottom
The open-circuit voltage (Voc ) of GaAs cell (η = 4.4%) in a three-terminal config-
solar cell fabricated on Si substrate as a uration. This is the highest efficiency for
function of TCA temperature is shown the GaAs/Si monolithic tandem solar cell
in Fig. 18 [51]. Voc of GaAs solar cell ever reported.
fabricated on GaAs substrate is also Although the conversion efficiency of
indicated. It is known that Voc is very the top cell is improved with increasing the
sensitive to the minority carrier lifetime. TCA temperature from 900 ◦ C to 1000 ◦ C,
Voc of solar cell grown on Si is improved the conversion efficiency of the Si bottom
with increasing the TCA temperature, but cell is reduced. This would be due to the
0.1–0.12 V smaller than that fabricated on degradation of minority carrier lifetime
GaAs substrate. and the formation of the deep junction by
Table 3 shows short-circuit current (Jsc ), the high-temperature heat treatment.
Voc , fill factor (FF) and conversion effi- A two-terminal Al0.15 Ga0.85 As/Si tan-
ciency (η) of the solar cell with 1000 ◦ C dem solar cell was fabricated in order
TCA and 900 ◦ C TCA (shown in the paren- to attain the photocurrent matching be-
theses). A total conversion efficiency of tween the top cell and the bottom cell.
22.1% has been achieved by combining the The structure and the current-voltage
1.1
Open-circuit voltage
1.0 on GaAs
[V]
on Si
0.9
0.8
900 950 1000
Fig. 18 Open-circuit voltage of
GaAs solar cell on Si as a TCA Temperature
function of TCA temperature. [°C]
Au-Zn/Au
p+-GaAs MgF2 /ZnS
p+-Al0.8Ga0.2As 1 × 1018 cm−3 0.05 µm
p+-Alx Ga1−x As 1× 1018 cm−3 0.3 µm
(x : 0.15 → 0.30)
n -Al0.15Ga0.85As 2 × 1017 cm−3 1.0 µm

Top cell
n+-Al0.15Ga0.85As 1× 1018 cm−3 2.5 µm
n+-GaAs 20 nm
p+-Si 1.0 µm
n -Si 350 µm Bottom cell
n+-Si 0.8 µm
Au-Sb/Au
Fig. 19 Schematic cross-sectional view of two-terminal

Al0.15 Ga0.85 As/Si tandem solar cell.
30 Fig. 20 Current-voltage
characteristics of two-terminal
25 Al0.15 Ga0.85 As/Si tandem solar
cell.
Current density
20
[mA cm−2 ]
15
Jsc (mA cm−2) Voc (V) FF (%) η (%)
10
23.6 1.57 77.2 21.2
5
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Voltage
[V]
characteristics are shown in Figs. 19 and the two-terminal configuration has been
20, respectively. It is observed that the obtained [52].
short-circuit current of the top cell is per- It has been pointed out that a 2-terminal
fectly matched to that of the bottom cell. tandem solar cell with an efficiency higher
A conversion efficiency of 21.2% utilizing than 30% can be obtained by using a
top cell material with a band gap energy pressure. When the V/III ratio or the gas
of 1.7–1.8 eV over an Si bottom cell. pressure is increased, the growth mode
This structure resulted in photocurrent changes from island-type to layer-type for
matching between the top cell and the thin GaP layer thickness. The two-type of
bottom cell. However, in our study, the misfit dislocations, which are generated
current matching was obtained by using by the lattice mismatch and the thermal
the Al0.15 Ga0.85 As top cell, of which the expansion mismatch, are observed. A high
band gap energy is 1.61 eV. This is because density of APDs that propagate to the
the short-circuit current of the top cell is surface and are annihilated during growth
inferior to the ideal one. The main reason has been observed. The nucleation of GaAs
for the degradation is the short minority on Si and the effects of SLS buffer layer,
carrier lifetime caused by a high density of TCA, and hydrogenation are described.
dislocation in the AlGaAs layer on Si. If it The crystal quality has been improved by
becomes possible to grow an AlGaAs layer using SLS buffer layer, increasing the TCA
on an Si substrate with a long minority temperature and using hydrogen plasma
carrier lifetime, comparable to that grown treatment. The GaAs layer grown on Si
on GaAs substrate, a higher efficiency substrate has been applied to photovoltaic
tandem solar cells can be obtained by devices.
increasing the Al composition so that the
photocurrent matching between the top References
cell and the bottom cell is retained.
The improvement of the Si bottom cell is 1. J. P. Benner, J. M. Olson, T. J. Coutts, in
also important for the increase of the total Advances in Solar Energy, (Ed.: K. W. Boer),
American Solar Energy Society, Inc., 1992,
conversion efficiency. The main problem is
p. 125, Vol. 7, Chapter 4.
that the conversion efficiency is degraded 2. M. A. Green, K. Emery, K. Bucher et al.,
after the crystal growth process [53]. The Progr. Photovolt. 1999, 11, 31.
junction depth becomes deeper and the As 3. T. Soga, S. Hattori, S. Sakai et al., Electron.
atoms diffuse into the Si substrate during Lett. 1984, 20, 916–918.
4. M. Akiyama, Y. Kawarada, K. Kaminishi,
the growth. A low temperature growth
Jpn. J. Appl. Phys. 1984, 23, L843–L845.
process is necessary for the improvement 5. W. I. Wang, Appl. Phys. Lett. 1984, 44,
of the bottom cell. Therefore, in the future, 1149–1150.
the technology to grow an AlGaAs layer 6. B.-Y. Tsaur, G. M. Metze, Appl. Phys. Lett.
with long minority carrier lifetime at 1984, 45, 535–537.
7. W. T. Masselink, T. Henderson, J. Klem
low temperature should be investigated to
et al., Appl. Phys. Lett. 1984, 45, 1309–1311.
increase the efficiency of the tandem cell. 8. T. Egawa, H. Tada, Y. Kobayashi et al., Appl.
Phys. Lett. 1990, 57, 1179–1181.
2.2.5 9. T. Soga, T. Kato, M. Yang et al., J. Appl. Phys.
Summary 1995, 78, 4196–4199.
10. T. Soga, T. George, T. Suzuki et al., Appl.
The nucleation, dislocation generation, Phys. Lett. 1991, 58, 2108–2110.
stress relaxation, and the annihilation of 11. R. Fischer, D. Neuman, H. Zabel et al., Appl.
APD in the GaP/Si heteroepitaxial growth Phys. Lett. 1986, 48, 1223–1225.
12. Y. Watanabe, Y. Kadota, H. Okamoto et al.,
have been reviewed. The growth mode J. Cryst. Growth 1998, 93, 459–463.
appears to be island-type for the low 13. N. Chand, R. People, F. A. Baicocchi et al.,
values of V/III ratio and the low gas Appl. Phys. Lett. 1986, 49, 815–817.
14. M. Yamaguchi, A. Yamamoto, M. Tachikawa 35. N. Otsuka, C. Choi, Y. Nakamura et al.,

et al., Appl. Phys. Lett. 1988, 53, 2293–2295. Mater. Res. Soc. Symp. Proc. 1985, 67,
15. M. Yamaguchi, J. Mater. Res. 1991, 6, 85–90.
376–384. 36. J. W. Matthews, A. E. Blakeslee, J. Cryst.
16. K. Nozawa, Y. Horikoshi, Jpn. J. Appl. Phys. Growth 1974, 27, 118–123.
1991, 30, L668–L670. 37. K. Rajan, M. Denhoff, J. Appl. Phys. 1987, 62,
17. H. Shimomura, Y. Okada, M. Kawabe, Jpn. 1710–1715.
J. Appl. Phys. 1992, 31, L628–L630. 38. D. Gerthsen, D. K. Biegelsen, F. A. Ponce
18. F. R. N. Nabarro (Ed.), in Dislocations in et al., J. Cryst. Growth 1990, 106, 157–163.
Solids, North-Holland, Amsterdam, 1979, 39. T. Yao, H. Kakao, H. Kawanami et al.,
pp. 461–545. J. Cryst. Growth 1989, 95, 107–111.
19. K. Tamamura, K. Akimoto, Y. Mori, J. Cryst. 40. T. Soga, T. Jimbo, M. Umeno, Appl. Phys.
Growth 1998, 94, 821–825. Lett. 1990, 56, 1443–1445.
20. T. Soga, T. Suzuki, M. Mori et al., J. Cryst. 41. T. Soga, S. Nozaki, N. Noto et al., Jpn. J. Appl.
Growth 1993, 132, 134–140. Phys. 1989, 28, 2441–2445.
21. T. Soga, T. George, T. Jimbo et al., Jpn. 42. N. Hayafuji, S. Ochi, M. Miyashita et al.,
J. Appl. Phys. 1991, 30, 3471–3474. J. Cryst. Growth 1989, 28, 2441–2445.
22. N. Noto, S. Nozaki, T. Egawa et al., Mater. 43. T. Soga, H. Nishikawa, T. Jimbo et al.,
Res. Soc. Symp. Proc. 1989, 48, 247–252. J. Cryst. Growth 1991, 107, 479–483.
23. T. Soga, T. George, T. Jimbo et al., Appl. 44. R. K. Ahrenkiel, D. J. Dunlavy, T. Hanak,
Phys. Lett. 1991, 58, 1170–1172. Sol. Cells 1998, 24, 339–347.
24. O. Ueda, K. Kitahara, N. Ohtsuka et al., 45. R. K. Ahrenkiel, B. M. Keyes, T. C. Shen
Mater. Res. Soc. Symp. Proc. 1991, 221, et al., J. Appl. Phys. 1991, 69, 3094–3099.
393–398. 46. H. A. Zarem, J. A. Lebens, K. B. Nordstrom
25. J. Nishizawa, T. Kurabayashi, J. Cryst. Growth et al., Appl. Phys. Lett. 1989, 55, 2622–2624.
1988, 93, 98–102. 47. M. L. Timmons, T. S. Colpitts, R. Venkaa-
26. G. B. Stringfellow, J. Cryst. Growth 1991, 115, subramanian et al., Appl. Phys. Lett. 1990, 56,
418–420. 1850–1852.
27. T. Soga, T. Jimbo, M. Umeno, Jpn. J. Appl. 48. G. W. ’tHooft, C. Van Opdorp, H. Veenvliet
Phys. 1993, 32, L767–L769. et al., J. Cryst. Growth 1981, 55, 173–179.
28. T. Soga, T. Jimbo, M. Umeno, Appl. Phys. 49. S. Nozaki, J. J. Murray, A. T. Wu et al., Appl.
Lett. 1993, 63, 2543–2545. Phys. Lett. 1989, 55, 1674–1676.
29. J. W. Matthews, in Epitaxial Growth Part B, 50. N. Das Gupta, R. Riemenschneider, H. L.
Academic, New York, 1975, Chapter 8. Hartnagel, J. Electrochem. Soc. 1993, 140,
30. M. Tachikawa, H. Mori, Appl. Phys. Lett. 2038–2043.
1990, 56, 2225–2227. 51. T. Soga, M. Kawai, K. Otsuka et al., 2nd
31. T. Soga, H. Nishikawa, T. Jimbo et al., Jpn. J. World Conference and Exhibition on Photo-
Appl. Phys. 1993, 32, 4912–4915. voltaic Solar Energy Conversion, European
32. O. Ueda, T. Soga, T. Jimbo et al., Commission, Joint Research Centre, Ispra,
Defect Control in Semiconductor, (Ed.: Italy, 1998, pp. 3737–3740.
K. Sumino), Elsevier Science, Amsterdam, 52. T. Soga, K. Baskar, T. Kato et al., J. Cryst.
1990, pp. 1141–1146. Growth 1997, 174, 579–584.
33. P. M. Petroff, J. Vac. Sci. Technol. 1986, B4, 53. M. Yang, T. Soga, T. Egawa et al., Sol. Energy
874–876. Mater. Sol. Cells 1994, 35, 45–51.
34. T. Soga, T. Jimbo, M. Umeno, J. Cryst.
Growth 1994, 145, 358–362.
2.3 Grafting Molecular Properties onto Semiconductor Surfaces 127
2.3 world and the junction between two differ-

Grafting Molecular Properties onto ent semiconductors involve their surfaces.
Semiconductor Surfaces Even one of the simplest of electronic de-
vices, the homojunction diode, requires
Rami Cohen, Gonen Ashkenasy,
outside metal–semiconductor contacts.
Abraham Shanzer, David Cahen
Therefore, understanding and control over
Weizmann Institute of Science, Rehovot, Israel
the electronic properties of semiconduc-
2.3.1 tor surfaces are essential for construct-
Surface Electronic Properties ing devices and for fine-tuning their
performance.
The chemistry of the crystal surface is very Figure 1 shows the energy-band diagram
different from that of the bulk. Although of an n-type semiconductor and the
the atoms are arranged in a well-ordered electrical properties associated with the
structure in the bulk, on the surface they surface. This chapter mostly deals with
may adopt a different position from that n-type semiconductors and definitions are
corresponding to the perfect crystal lattice given only for this type. The difference
because of the absence of part of the neigh- between the energetics of the bulk and
boring atoms. In that case, the surface that of the surface is demonstrated by
is said to be relaxed or reconstructed [1]. the relatively high density of energy levels
A second difference between nearly every inside the semiconductor band gap. These
levels are referred to as surface states and
surface and the bulk is associated with
their importance stems from the fact that
the chemical environment of the bond-
they are involved in most electronic loss
ing atoms. Under ambient conditions, the
mechanisms such as charge-trapping and
surface is covered by one or more atomic
recombination.
layers of adsorbates such as oxygen (mostly
It was shown [4], using quantum me-
as oxide or hydroxide), carbon or hydrocar-
chanical calculations, that surface states
bons and is not atomically clean (i.e. the
can localize electrical charges in contrast to
atomic composition is not the same as
bulk states where electrons are delocalized.
that in the bulk). Thus, it is not surprising Because of this localization the surface be-
that nearly always the ‘‘real’’ surfaces are comes charged with respect to the bulk
disordered and lack the periodicity of the and an electric potential difference is cre-
bulk. Correspondingly, the positions and ated between the surface and the bulk,
distribution of energy levels and their occu- the so-called built-in potential (Vs ). This is
pation by electrons at the surface generally shown in the band diagrams as bending
differ from those associated with the bulk. of the valence and conduction bands, a
The importance of this situation be- feature called ‘‘surface band-bending’’. In
comes obvious when we try to use these relation to the bulk, the surface states of
crystals, especially semiconductor crystals, an n-type semiconductor localize negative
in a device. In this case, surface defects and charges and the opposite is true for p-type
imperfections play a dominant role in the material. Because it is easier to measure
electron transport in and out of the device the band-bending than the surface states,
and in this way influence its performance. in many cases the surface band-bending is
It is to be remembered that all con- used as an indicator for changes in the den-
tacts of semiconductors to the outside sity of surface states. Because the capture
Vacuum level
Electron affinity
Φ
Work χ
function CB
Empty Band bending
surface Vs
states
Occupied EF
surface
states Eg
Surface
resonances
VB
Energy
Distance
Density Space charge
of states region
(SCR)
Distance
Distance
Distance
Fig. 1 (a) One-electron energy level diagram of an n-type doped semiconductor. VB
represents the valence band, comprising largely filled, closely spaced energy levels. CB
is the conduction band, which consists of largely empty, closely spaced energy levels.
EF is the semiconductor Fermi level and Eg is the band gap energy. The surface states
are occupied until sufficient space charge is created corresponding to an electrostatic
field, which prevents further electrons from going to the surface. It is this field that
causes the shift in the band energies, i.e. Vs . (b) Side view of the crystal. The surface
localizes electrons at the surface states, and thus is negatively charged. Figure 1 also
shows the presence of surface energy levels that lie outside the band gap. These states
are referred to as surface resonances and unlike surface states they are degenerate
with the bulk states and can mix with them [2, 3]. A surface resonance has a varying
degree of localization in the surface region, although it should be noted that there is
no absolute definition for how strong a surface localization should be in order for it to
be defined as a surface resonance.
cross-sections of electrons and holes are determines the surface electronic prop-
higher on the surface than in the bulk, sur- erties. Therefore, to control the surface
face states also serve as fast recombination electrical properties and to make the semi-
centers of electrons and holes. This phe- conductor useful for electronic devices,
nomenon can critically affect the electron- we look for treatments that will interact
transfer efficiency in electronic devices and with the surface and modify it chemically.
thus their performance. The recombina- The requirements of molecular surface
tion rate of electrons and holes at the treatments, which are also named passi-
surface is expressed in units of velocity vation treatments, are the following:
(cm s−1 ) and is called surface recombina-
tion velocity (SRV or s). It reflects the rate 1. Remove surface and interface states or
at which minority carriers (holes in n-type at least eliminate them in the energy
materials) are consumed at the surface. In interval of the band gap. Because the
many cases, SRV is determined by the sur- presence of surface states is conducive
face states close to the middle of the band for charge-trapping and recombination
gap where the chances for electron-hole effects, moving them out of the band
recombination are the highest [5]. gap is necessary for the construction
Other surface electrical properties in- of high-speed semiconductor devices
dicated in Fig. 1 are the work function, like the ones that are based on GaAs
!, and the electron affinity, χ. In semi- and InP.
conductors, electron affinity is defined as 2. Tuning the surface electron affinity.
the difference in energy between the lo- Control over the electron affinity is
cal vacuum level [6] and the bottom of the essential for band edge engineering
conduction band (CB) at the surface. The of interfaces such as those in photo-
(local) work function is the minimum en- voltaic solar cells and for controlling
ergy required for an electron to escape the barrier height for electron loss to
into vacuum (just outside the range of the the surroundings. The second effect is
crystal potential) from the Fermi level. Al- demonstrated for the GaAs/(Al,Ga)As
though this definition is straightforward system where a band offset of 0.4 eV [7]
for metals, for semiconductors it should did not block electron loss from the
be borne in mind that there are mostly no GaAs into the passivating (Al,Ga)As
real energy levels at the Fermi level. The layer.
work function is determined by the elec- 3. Strong bonding that can withstand
tron affinity, the band-bending, and the device-processing. Ideally, the molecu-
energy difference between the Fermi level les will chemisorb on the surface.
and the conduction/valence band in the It should be noted that the binding
bulk. groups themselves, while interacting
with the surface, could modify the
2.3.2 surface chemistry and thus the surface
Requirements of Molecular Surface energetics on binding.
Treatments 4. Chemical protection from adsorbates,
especially atmospheric adsorbates. In
Because the surface charge and dipole are those cases where surface oxidation is
determined by the surface chemistry, it is detrimental, there is the daunting task
the surface chemistry that to a large extent of protection from O2 .
2.3.3 Figure 2 indicates a method to modify

Strategies to Control Surface Electronic the band-bending by shifting surface-state
Properties energies with respect to the band edges. If
the surface states are shifted out of the
Strategies for tuning surface electronic band gap, above the CB minimum or
properties, work function, electron affinity, below the valence band (VB) maximum
band-bending, and SRV have to take (i.e. converted into surface resonances),
into account the origin of chemical and they will couple to the band continuum
physical properties. Use of the versatility of and their charge will be delocalized in the
molecular chemistry can help control these band. As a result, Vs will be modified.
properties in a predetermined fashion. Another way to change the Vs is by
changing the surface-state energy inside
2.3.3.1 Controlling the Band-bending (Vs ) the band gap. The states are occupied
and Surface Recombination Velocity (SRV) at energies below the Fermi level and
As noted earlier, band-bending near the are empty at those above it (this holds,
surface results from charge localization on strictly speaking, only at 0 ◦ K). Therefore
the surface states. Thus, control over the by modifying the surface states so that their
band-bending requires a mechanism that energies will change from above (below) to
will control the density and occupation of below (above) the Fermi level, the surface
the surface states inside the band gap. states can be occupied (emptied) and the
Vacuum level Vacuum level
CB Surface CB
treatment EF
EF
Occupied Vs′
surface Vs
states VB
VB
SCR SCR
Fig. 2 Left to right: Reduction of the density of surface states of n-type semiconductor,
as a result of surface treatment, which leads to decrease in width of the space charge
region (SCR) and a decrease in the surface band-bending from Vs to Vs# .
surface charge changed. Change in the from a change in the chemical environ-
distribution of energies of the surface ment caused by adsorbed extrinsic atoms.
states can also affect the SRV. Because
the SRV depends on the density of surface 2.3.3.2 Controlling the Electron Affinity, χ
states with energies around the midgap, By definition, χ depends on the energy
shifting of the surface-state energies away difference between the vacuum level and
from the midgap can lead to reduction in the bottom of the CB at the surface.
the probability of charge recombination. Therefore, any treatment that influences
The strategy to modify surface states, the surface potential will modify χ.
and thus the Vs and SRV, is based on in- Modification of the surface potential can
teraction of chemically grafted molecules be achieved by utilizing polar molecules
with these states. The key is to find that will bind to the surface and change
molecules that will modify the semicon- its potential. Figure 3 shows schematically
ductor surface chemistry in a way that in- the manner in which a polar molecule can
volves the surface states. In this respect, the modify the surface χ. The surface poten-
origin of the surface states should be con- tial will be reduced if the molecular dipole
sidered. Intrinsic surface states originate is pointing toward the surface and will
from the termination of the crystal bulk increase if the dipole is directed in the
and the breaking of chemical bonds at the opposite direction. This approach of us-
surface, whereas extrinsic surface states ing dipolar molecules was applied to tune
originate from crystal imperfections, such the χ of metals [8–11] and semiconduc-
as missing surface atoms, line defects, or tors [8, 12–14]. It should be mentioned
(a) A
+ q
A d
m
Decrease of c
Binding −
d
group
(b) B
q
B +d
Increase of c
−d
Fig. 3 Strategy to tune the electron affinity, χ, of the

semiconductors by adsorption of polar molecules. All the
molecules have a similar binding group that allows strong
binding to the surface and a different end group (A, B) to tune the
molecular polarity (i.e. dipole moment). A molecular dipole
moment pointing toward the surface (a) leads to decrease in χ,
whereas a dipole pointing away from the surface (b) increases the
electron affinity. µ and θ are the molecular dipole moment and
tilt angle, respectively.
that because short-range atomic forces or band-bending, we can tune the work
determine the energy positions of the function
surface states inside the band gap, they
need not be affected by the presence of $! = $χ ± $Vs (2)
the macropotential. Therefore, looking at
with + for n-type and − for p-type.
Fig. 1, a polar molecule modifies the en-
ergy of the bands with respect to the
2.3.4
vacuum level, but not necessarily those
Strategy for Molecule Selection
of the surface states with respect to the
band edges or the surface-state (and band)
The chemical strategy to control the sur-
energy level densities.
face energetics using organic molecules is
The change in the electron affinity
illustrated schematically in Fig. 4. The idea
caused by the molecule’s dipole moment
can be described in terms of a parallel is to incorporate several molecular proper-
plate capacitor, using the well-known ties simultaneously in one molecule and to
Helmholtz equation allow systematic modification of one spe-
cific property, independent of others. This
cos θ approach provides
$V = N × µ × (1)
εεo
1. a simple tool to investigate the re-
where $V is the potential drop caused
lation of the macroscopic proper-
by the dipole layer, µ is the dipole
ties of the semiconductors and the
moment, N the dipole density per unit
molecular properties of the adsorbed
area, θ the angle between the dipole
molecules;
and the surface normal, ε is the relative
2. a simple tool to enable the development
dielectric permittivity of the film, and εo
of models for surface engineering.
the permittivity of free space [15]. If values
for N , cos θ, and ε are known, $V (and A different approach, which is more suit-
thus $χ) can be calculated. The equation able for nonmolecular, extended bonded,
also demonstrates that controlling χ can electrically conducting inorganic materi-
be achieved by tuning the dipole direction als, is to use different molecular layers,
(toward/from the surface), its magnitude each of which has a different function.
and its orientation with respect to the For example, the first layer can electri-
surface normal. It has to be noted that cally passivate the surface and a second
the concept of a dielectric constant for a layer can give long-term chemical protec-
monolayer is problematic. It can be related tion [7].
to the molecular polarizability using the
Clausius-Mosotti relation (cf. discussion 2.3.5
in [14]). Organic and Inorganic Molecular Surface
Treatments
2.3.3.3 Controlling the Work Function, "
As shown in Fig. 1, the work function is In general, we can classify the chemi-
the energy difference between the Fermi cal surface treatments into two classes,
and vacuum levels and depends on the organic and inorganic. Table 1 summa-
electron affinity and band-bending. Thus, rizes the advantages and disadvantages of
by controlling either the electron affinity each group. Although, as noted in the
Auxiliary groups Binding groups

Polar group Molecular
covalent bond
Light-sensitive group slightly polar bond properties
Alkyl chain/ ionic bond
Hydrophobic group π system [benzene]
Synthesis
A A A
Adsorption
A A A
Macroscopic
properties
Φ, χ, Vs
Fig. 4 Chemical strategy to control the surface electrical properties using

multifunctional molecules.
table, the chemical stability of organic 2.3.5.1 Inorganic Surface Treatments

treatments is generally lower than that Table 2 presents several examples of
of inorganic ones, Dorsten and cowork- inorganic surface treatments used to
ers [16] demonstrated that sulfidization of modify the surface electronic properties.
GaAs using octadecylthiol gave rise to a The first is oxidation of silicon (Si), a
sulfide layer that was as stable as that which cornerstone of today’s electronic device
could be obtained with corresponding in- industry [17]. The treatment moves most
organic treatments. Using the flexibility of of the surface states out of the Si band
organic synthesis, one can incorporate sev- gap [2, 18] and enables the worldwide use
eral functional groups in a single molecule, of silicon for device manufacturing. On
each performing a different role and ideally III–V and II–VI group semiconductors,
independent of each other. The versatility the chemistry of the semiconductor’s
of the organic treatments can also be used native oxides is more complicated than on
to study the relationship between modifi- silicon, and plain oxidation of the surface
cation of the surface chemistry and surface cannot lead to chemically stable, nearly
electronic effects. defect-free surfaces [7]. For example, on
Tab. 1 Advantages and disadvantages of organic and inorganic treatments
Organic treatments Inorganic treatments
Advantages * Structural versatility and * Chemical stability

flexibility
* Can incorporate several * Thermal stability
properties
simultaneously (polarity,
hydrophobicity, light
sensitivity)
* Allow systematic * Strong binding/interaction
modification of one
specific property, often
without effect on other
properties
Disadvantages * Limited chemical and * Limited chemical flexibility
thermal stability
Tab. 2 Examples of substrates for which inorganic chemical surface modifications have been
developed
Semiconductor Treatment References
Silicon Oxygen 2, 18
Hydrogen 18, 28
Bromine 29
As 30
GaAs Sulfide and Selenide 20, 24, 27, 31, 32, 33–37
Ruthenium 38
Phosphine (PH3 ) 39
P2 O5 /NH4 OH 40, 41
Chlorine 42
Cesium 43
Sb 44, 24
H2 S 26
Iodine 45
InP CdS 46
Ruthenium 47, 48
CdTe (polycrystalline) Hydrogen plasma 49
Ruthenium 50, 51
Reactive metal interlayer 52
Oxygen anneal 53
(Hg, Cd)Te Sulfide 54
Hydrogen 55, 56
CuInSe2 Oxygen anneal 53, 57
(polycrystalline)
GaAs(100) and (110) [2] surfaces in the support this hypothesis and reveal a
ambient, high surface state densities pin strong reduction in the band-bending of
the Fermi level and generate a high SRV. GaAs on treatment with Se-containing
Covering GaAs(100) with (Al,Ga)As was reagents. The stable phase was found to
found to be effective in terms of reducing be Ga2 Se3 [32], which has a close lattice
surface electron-hole recombination, but match to GaAs, and therefore creates an
could not block loss of electrons to almost strain-free layer.
the passivating (Al,Ga)As layer. Another Another class of surface treatments is
approach to modify the surface chemistry based on halogens. Halogens are mostly
of GaAs is the use of inorganic sulfides. used for surface etching and were found
These treatments were found to remove to dissociate upon adsorption [19]. The ef-
part of the surface oxide and form ficacy of halogen treatment for modifying
sulfides of Ga and As [19]. Studies of the surface electronic properties was found
metal-insulator semiconductor (MIS) and to depend on the morphology and compo-
Schottky diode structures of GaAs, treated sition of the surface. In the case of GaAs,
with (NH4 )2 S and Na2 S [20, 21], revealed halogens showed high reactivity toward the
that aqueous sulfide treatments induce Ga-rich surface [61, 62].
only minor changes in the net surface Parkinson and coworkers demonstrated
state density. Their main effect is reduction improvement in the open-circuit voltage
of the density of the trap states, which (Voc ) and fill factor of n-GaAs/K2 Se-K2 Se2 -
are farther in energy from the band KOH/C photoelectrochemical solar cells
edges than the shallow, doping levels. on treatment with Ru (III) [38]. The im-
This means that these treatments modify provement in the cell performance was
the positions of the surface-state energy ascribed to the shift of the surface-state
levels [19, 22]. This observation agrees with energies, which was thought to lead
results from other studies where sulfide to a reduction in the main power-loss
treatment was claimed to repin the surface mechanism, electron-tunneling through
Fermi level at a different energy [22–25]. the surface states. Treatment with Ru
A different approach was introduced by was also found to improve the perfor-
Shen and coworkers [26] who used plasma mance, specially the stability, of InP-based
H2 S treatment. In contrast to the aqueous photoelectrochemical solar cells [47]. This
sulfide treatments, this treatment leads to improvement was related to an increased
significant reduction of the surface-state barrier height [48].
density (up to three orders of magnitude)
on sulfidization. The main drawback 2.3.5.2 Organic Surface Treatments
of the inorganic sulfide treatments is Table 3 summarizes examples of organic
the instability of the sulfur-GaAs bond surface treatments used for changing the
in the ambient (vs oxygen attack, in surface electronic properties. As a first
particular) [19, 27]. example we consider self-assembly of
According to Lunt [58], the efficacy of organo-silanes on oxidized silicon. These
a selenide treatment applied to GaAs self-assembled monolayers are known to
is expected to be stronger than that of be stable and well ordered [63, 64]. Mod-
a sulfide one because of the electron- ification of the monolayer properties can
deficient nature of the GaAs binding be achieved in a predetermined fashion
site. Indeed, several studies [59, 60, 31] by changing the chemical structure of the
Tab. 3 Several organic chemical treatments used for surface modifications
Semiconductor Treatments References
Silicon, single Self assembly of organosilanes 65

crystal and porous Acids and bases 69
Amines 66, 70
Common organic solvents 71
Other organic adsorbates 67, 128
GaAs Organic disulfides 72
Porphyrins for NO Detection 73, 129
Dicarboxylic acids 72, 74, 75–77
Sulfides and thiols 58, 68, 78, 79, 80
Dithiocarbamate 81
CH3 CSNH2 81, 82
Benzoic acids 14, 83, 130
Alcohols 84
Ethyl iodide 85
Dimethylcadmium 86
Dimethylzinc 86
Trimethylgallium 86
Diethylzinc 87
HS(CH2 )Si(OCH3 ) 88
InP Dicarboxylic acids 76
Alkanethiol 89
Various other molecules 90
CdTe Benzoic acids 12
Dicarboxylic acids 75
CdSe Dicarboxylic acids 76
Dithiocarbamate 91
Thiols 92
Anionic sulfur donor 93
Ethylenediaminetetraacetic acid 94
Dialkyl chalcogenides 95
Olefins 96
Aniline derivatives 97
Amines 98, 99
Boranes 100
Fullerenes 101
Carbonyl compounds 102
TCNQ derivatives 103
Benzoic acids 13
Silapentanes and chlorinated Silanes 104
Cyanide 105
TOPO 106
Porphyrins for O2 Detection 131, 132
TiO2 Benzoic acids 107
ITO Benzoic and other carboxylic acids 108–110
CuInSe2 Dicarboxylic acids 111
Benzoic and hydroxamic acids 12
Disulfides 8
self-assembling molecules. On the basis of data that were obtained after adsorption
that, a systematic modification of the sur- of several series of organic molecules (in-
face electron affinity and band-bending of cluding aniline and carbonyl compounds)
n-type silicon was demonstrated by using were interpreted in terms of changes in the
self-assembly of substituted quinoline band-bending [96–99, 102], whereas CPD
chromophores [65]. The chemical scheme changes induced by adsorption of benzoic
included two steps: acids and aniline derivatives were rational-
ized in terms of changes in electron affin-
• self-assembly of organo-silanes that add ity [13]. The different conclusions from lu-
organic functionality to the surface, and minescence intensity measurements and
• grafting of a series of substituted CPD measurements were ascribed to the
quinoline chromophores that add the fact that the PL measurements were done
variable (in this case polar) functionality. in solution, whereas the CPD measure-
ments were performed in air. This differ-
On porous silicon, which has a high ence also reflects the effects of interaction
surface-to-volume ratio and thus a large with the surrounding medium.
number of accessible surface states, a To provide information on molecule-
substantial change in PL was recorded surface interaction and on the feasibility of
on exposure to amines and different sol- chemical treatments for use in electronic
vents [66, 67]. On GaAs, organic sulfides devices, like sensors, attempts have been
and thiols were reported to induce PL made to establish a correlation between
changes that were attributed to changes molecular properties and changes in sur-
in both the SRV and the Vs [58, 68]. face properties. Table 4 summarizes works
Adsorbing a series of benzoic acids in which such a correlation was found.
with systematically varying dipole mo-
ment led to a different response. Changes 2.3.5.3.1 Correlation of Molecular Param-
in contact potential differences (CPD) eter with Changes in the Surface Electron
were correlated to changes in the elec- Affinity As mentioned earlier, control of
tron affinity without observable effect on the electron affinity or surface potential
the surface band-bending [14]. Substituted can be achieved by modifying the molec-
dicarboxylic acids, which were found to ular dipole moments of the adsorbed
bind by a two-site mechanism, rather molecules. Figure 5 shows the change
than a one-site one, such as the ben- in the electron affinity of CdTe, CdSe,
zoic acids [112], modified both the band- and GaAs, which was deduced from CPD
bending and the electron affinity of the measured by a Kelvin probe, upon graft-
GaAs [72, 74]. This difference in the elec- ing a series of substituted benzoic acids
trical effect on treatment with carboxylic onto the semiconductor surface. The lin-
and dicarboxylic acids can be explained ear relation between the change in the
by the different binding group and the electron affinity and the dipole moment
change in the molecular energy levels [75]. of the substituted benzoic acid is clearly
On CdTe, CPD measurements showed observed. The dipole moments of the sub-
that organic benzoic and dicarboxylic acids stituted phenyl groups reflect the electron
could modify the electron affinity and/or withdrawing or donating power of the
the surface band-bending [12, 75]. On the substituents. Bruening et al. used elec-
wider band gap semiconductor, CdSe, PL trochemical and CPD measurements in
Tab. 4 Transfer of molecular properties to the macroscopic properties of semiconductors
Type of correlation Sample References
Change in electron affinity, χ, with dipole Silicon (single crystal) 65

moment of the benzoyl substituent and GaAs 14, 77, 130
cammett substituent constants CdTe 12, 75
CdSe 13
CuInSe2 12, 8
TiO2 107
ITO 109
Change in built-in potential, Vs , with ionization Silicon 65, 113
potential, proton affinity, electron-accepting CdTe 75
power; LUMO energy of the molecules CdSe 96, 97, 114, 99
Change in photoluminescence, PL, with CdSe 103
electrochemical redox potential
Change in surface recombination velocity CdSe 115
(SRV), s, with ionization potential
Change in PL with Hammett parameters CdSe, CdS 102
Change in barrier height with number of carbon GaAs (thiols) 80
atoms in chain
Change in barrier height with molecular dipole GaAs, Si 77, 128, 130
moment
combination with ellipsometry and FTIR can be understood from the increase in
data to study the effects of binding a series effective dielectric constant of the molecu-
of cyclic disulfides with systematically vary- lar layer [130].
ing dipole moment and different degrees
of hydrophobicity to Au and CuInSe2 [8]. 2.3.5.3.2 Correlation of Molecular Parame-
They found that the magnitude and di- ters with Changes in the Band-Bending In
rection of the change in electron affinity general, the change in Vs can be viewed by
depend on the surface coverage, the ori- two mechanisms,
entation of the molecular dipole relative
• molecule-induced surface oxidation/
to the surface normal, and the mode of
reduction (generalized acid–base reac-
binding. Knowing the molecular surface
tion) where the molecules either ac-
coverage and tilt angle and thus the resid-
cept or donate electrons [117] accord-
ual molecular dipole moment, as well as
ing to the difference in the oxida-
the atomic electronegativity difference at
tion/reduction potentials, or alterna-
the molecule-surface interface allowed the-
tively by
oretical estimates of the experimentally
• frontier orbital interaction mechanism,
observed change in the electron affinity,
which involves the energy levels of the
that is, it gave predictive power to Eq. (1).
molecular frontier orbitals and of the
In further work Wu et al. tested Eqn. (1)
semiconductor surface states [118, 119].
and used a series of molecules similar to
that shown in Fig. 7, but completely con- In the second mechanism, we con-
jugated. The main effect was found to be sider the interaction between the high-
a smaller influence of the dipoles which est occupied molecular orbital (HOMO)
600
R = CH3O, H, F, Br, NO2
R
400
200
COOH
[mV]
Nitro
∆χ
Bromo
Hydro Fluoro
0
Methoxy CdSe
−200 GaAs
CdTe
−400
−2 0 2 4
Dipole moment
[Debye]
Fig. 5 Linear correlation of the change in the electron affinity of n CdTe,
CdSe, and GaAs, as function of the dipole moment of benzoic acids [116],
adsorbed on the semiconductors [12–14].
level of the surface or the molecule and in Fig. 6(b), the molecule’s LUMO level
the lowest unoccupied molecular orbital interacts with occupied surface states,
(LUMO) level of the other. It is an ex- which are below the Fermi level and there-
tension of the well-known frontier orbital fore assume the role of the HOMO level.
interaction between the energy levels of On interaction, the surface-state energies
two molecules forming a complex, as increase (with respect to vacuum), that is,
shown in Fig. 6(a). On interaction, the they are ‘‘pushed’’ down in energy toward
HOMO level is stabilized to lower energy, the VB, whereas the molecular LUMO level
whereas the LUMO level is destabilized is ‘‘pushed’’ up in energy toward the CB.
and pushed up in energy. As a result, electrons that were formerly
Table 4 summarizes studies in which localized on the surface states may now oc-
the molecular ionization potential, which cupy surface resonances with only partial
is related to the HOMO level, and the localization at the surface. Therefore, they
molecular electron affinity, related to the do not anymore (or little) contribute to the
molecular LUMO level, are correlated to net surface charge. Reduction of Vs is thus
the changes in Vs . This correlation fits the expected. A different result is expected if
orbital interaction mechanism. the surface states are close to the semicon-
Figure 6 shows several scenarios for ductor midgap, as shown in the second
the interaction of a given molecular case (Fig. 6c). Because in this case the
LUMO level with surface states at dif- surface states are well removed from the
ferent energies. In the first case, shown band edge, only part of them will turn into
(a) LUMO
HOMO
CB
(b) LUMO EF
e- VB
CB
(c) LUMO
EF
VB
(d) h+ CB
EF
LUMO VB
Fig. 6 Molecular orbital (HOMO–LUMO) interaction of two

molecules (a) and of a molecule with semiconductor surface states
(b–d). Different results are obtained after interaction with shallow
acceptor states (occupied surface states close to the VB) (b), deep
acceptor states (occupied states close to midgap) (c), and shallow
donor states (close to the CB) (d). In general, the donor HOMO level is
slightly stabilized by the interaction, whereas the acceptor LUMO level
is slightly destabilized.
surface resonances and the net effect on on the density of the midgap (deep) sur-
the total surface state density, and thus face states [5].
on Vs , is expected to be moderate, com- In the third scenario, shown in Fig. 6(d),
pared to the first case (Fig. 6b). The SRV, we consider the interaction of a molecular
on the other hand, is expected to change LUMO level with empty surface states of a
significantly because it depends critically p-type semiconductor. Because the LUMO
level of the molecule is well below the • a constant change of 170 mV because of
Fermi level, electron transfer from the sur- the binding group and
face to the molecule is expected and the • a second contribution that could be
molecule plays the role of the HOMO level correlated with the molecules’ LUMO
in the interaction. As in the other cases, energy level and that increased with
the surface states that were pushed in en- decreasing energy separation between
ergies to more than the CB minimum will the molecule’s LUMO state- and the
turn into surface resonances that will lead surface state-energy level [75].
to a reduction in Vs .
The energy levels of the molecule after This second contribution is denoted
interaction should also be considered. schematically in Fig. 7 and can be at-
These levels are modified (i.e. changed tributed to extended coupling of the
in energy, relative to the band edges) on molecules’ energy level and the semicon-
interaction with the surface states, as was ductor surface states, as the molecules’
demonstrated [120, 133] for the interaction LUMO energies become closer in energy
of thiophenol derivatives with Cu(111). to those of the surface states. The predic-
If, on interaction, the molecular LUMO tive power of the frontier orbital interaction
level is at an energy that corresponds to scheme, noted earlier, was demonstrated
somewhere within the band gap and below for n-CdSe, n-GaAs, n-InP, and p-GaAs
EF , the state will act as an occupied surface crystals when changes in the Vs and in the
state and will increase (decrease) the net SRV of the crystals on interaction with a
surface charge of n-(p-)type surfaces. On given dicarboxylic acid molecule (DCDC
the other hand, if the molecular LUMO and DHDC) could be explained [112]. On
level is pushed below the VB maximum, the basis of the studies mentioned earlier,
then the state acts as a surface resonance it was found that the ability of the chemical
and only a moderate effect on the band- treatment to modify the electronic proper-
bending is expected. It should be noted ties of semiconductor surfaces depends on
that although our discussion in this the following parameters:
article is limited to the molecular LUMO
level, analogous interactions of molecular 1. The molecule’s frontier orbital energy
HOMO levels with surface states can also level and the difference in energy
be energetically favorable [119] and lead to between that level and those of the
modification of surface states and/or SRV. interacting surface states. The smaller
Figure 7 shows an example of the the energy distance, the stronger is
change in Vs of an etched n-CdTe sin- the molecule-surface coupling and the
gle crystal as a function of the LUMO larger can be the induced change in the
energy level of a series of dicarboxylic surface electronic properties.
acid derivatives [75]. The systematic mod- 2. The surface-states’ energy levels and
ification of the molecular LUMO level densities. In surfaces where the dom-
was achieved by changing the molecule’s inant surface states are close to the
substituents from electron-donating to band edges, the main effect on interac-
electron-accepting groups. The change tion is a change in band-bending. On
in Vs was suggested to stem from two surfaces with dominant states close to
contributions: midgap, the main effect on interaction
(a)
X X X
( )B
600 ∞ ∞ ∞∞
o o O O
O O O O
500 HO HO OH VI
OH
I. X = OMe; DMDC II. X = OMe; PMDC
III. X = H; DHDC IV. X = CN; PCDC V
400 V. X = CF3; DFDC
[mV]
∆V S
VI. X = CN; DCDC

IV
300
200
I II III
100
−3.5 −4 −4.5 −5 −5.5
LUMO energy
[eV]
(b) I
CdTe electron energy levels
−4 II −4
CdTe
LUMO energy
III
[eV]
[eV]
−5 −5
IV EF
VI
−6 −6
−7 −7
Fig. 7 (a) Change in Vs of n-CdTe upon adsorption of dicarboxylic
acid derivatives as a function of the benzoyl substituents’ LUMO
energy [75]. $Vs is the change in Vs relative to the etched surface.
(b) Energy diagram of bare CdTe and of LUMO of isolated molecule
before adsorption. The CdTe energy levels were experimentally
measured by CPD, and the Fermi level was calculated from the doping
density.
is a change in surface recombina- technique to prepare a Schottky diode of

tion velocity. Au/Cd-stearate/CdTe on both n- and p-
CdTe. Such a molecular film increased
According to this approach, designing both the barrier height and the open-
molecular surface treatments should start circuit voltage of this MIS (metal-insulator
by mapping the surface-state energies and semiconductor) solar cell structure. The fill
then choosing a molecule with energy factor was found to depend on the thick-
levels that are close to the target surface ness of the LB chain and was the highest
states and that can successfully interact for one monolayer (2.5 nm). These find-
with them. In practical cases, semiconduc- ings are in accordance with the optimum
tor surfaces may possess several popula- insulator thickness found for silicon MIS
tions of surface states and thus molecular solar cells of 1 to 2 nm. Allara and cowork-
selectivity will be required. ers [80] studied the effect of self-assembled
Further progress in this direction de- monolayer (SAM) alkane thiols with differ-
pends on ent chain lengths on the Schottky barrier
height of the Au/n-GaAs system. They
• better knowledge of molecular energy
found that while the barrier height in-
levels on an absolute scale (to allow
creased only very slightly, the increase
comparison with semiconductor en-
correlated linearly with the number of the
ergy levels);
carbon atoms in the alkane chain. They
• consideration of other factors that might
attributed this change to an increase of
affect molecule-surface coupling, such
the negative charge at the binding group
as, for example, the shape of the orbitals,
(i.e. sulfur atom) as the thickness of the
that is, symmetry considerations.
SAM increased. However, an opposite ef-
2.3.6
fect on the barrier height was noted for the
Examples of Molecular Control over thiol-passivated Cu/n-GaAs system, which
Optoelectronic Devices suggests that other mechanisms should
also be considered.
The progress in device-processing tech- Although tuning the Schottky barrier
niques and the ability to design new mul- properties by way of changing the length
tifunctional molecules, whose properties of the molecules is essentially similar
can be modified systematically, opens new to the effect of changing the thickness
directions for construction of molecular of any dielectric film between the metal
devices. Given below are several examples and semiconductor, recent examples [77,
of the application of organic molecules 107–109, 123, 128, 130] show that true
for controlling the performance of op- molecular effects can be obtained.
toelectronic devices based on inorganic, Vilan et al. demonstrated that molecules
nonmolecular materials. adsorbed at the Au/n-GaAs interface can
The first system we consider is a Schot- tune the energetics of this metal–semi-
tky diode and its derivatives. The electrical conductor diode, because of their chemical
characteristics of this diode depend on character, namely, by systematic sub-
the interface states and dipoles and can stitution on aromatic rings [77]. Using
thus reflect changes induced by adsorbed a series of dicarboxylic acid derivatives
molecules. Petty, Roberts, and coworkers whose dipole was varied systematically,
[121, 122] used the Langmuir-Blodgett (LB) they produced diodes in which the effective
barrier height is tuned by the molecular current–voltage characteristics of the

dipole moment. Qualitatively, compared inorganic–organic TiO2 /spiro-MeOTAD
to the unmodified junction, molecules (an amorphous organic hole conductor)
with a dipole pointing toward the surface heterojunction [107]. These changes were
increased the forward current in the cur- correlated with the changes in the work
rent–voltage (I –V ) curve, whereas those function of the modified TiO2 and were
with a dipole pointing away from the sur- attributed to variations in the built-in volt-
face decreased it. The interesting thing age at the TiO2 /spiro-MeOTAD interface.
here is that the molecules do not form A rough correlation between the change of
a perfect monolayer and there is high the work function and the current–voltage
probability of pinholes in the layer. De- characteristics was demonstrated also by
spite this, the molecules strongly affect the Cambell and colleagues [123], who used
diode characteristics. These results reveal a self-assembled layer of conjugated thiol
that it is not necessary to use molecules molecules to manipulate the Schottky bar-
that conduct current through them to af- rier height between a Cu electrode and
fect device electrical characteristics and an organic electronic material used for
that molecules that electrostatically inter- light-emitting diode (LED) systems.
act with the surface can also be applied. Zuppiroli, Nuesch, and coworkers
The results of Wu et al. [130], who demonstrated [108, 109, 134] that pre-
prepared and measured Au/molecule/n- treatment of the transparent conductor
GaAs diodes with the earlier mentioned indium tin oxide (ITO) with a few deriva-
series of conjugated molecules agree with tized organic molecules can significantly
these conclusions. improve (decrease) the turn-on char-
Selzer and Cahen [128] showed that a acteristic (field) that denotes the field
complementary configuration can be used, above which light emission is observed
as well. They adsorbed ligands on the in the ITO/poly(paraphenylene)/Al LED
metal rather than on the semiconduc- system. The turn-on field was reduced
tor side of the junction. In that case the from 200 MV m−1 to 100 MV m−1 af-
metal work function, rather than the semi- ter molecule adsorption. The decrease in
conductor’s electron affinity is changed. the turn-on field was correlated with the
They used the molecules of Ref. [8], which dipole moment of the carboxylic group.
are similar to those shown in Fig. 7(a), The molecules modified (increased) the
but with cyclic disulfide, instead of dicar- work function of the ITO. In this way they
boxylic acid binding groups, so as to allow increased the hole injection efficiency in
chemisorption onto Au. The molecularly the device. Moreover, the device durability
modified metal was then used to prepare was increased. The stability under DC op-
Au/molecule/SiOx /Si diodes. Results are eration was increased from 15 minutes to
naturally opposite to those obtained with 2.5 hours after the molecular treatment.
GaAS/Au diodes [77], because the sub- Friend and coworkers [124] showed that
stituted phenyl groups point in opposite molecular surface treatments could also be
directions in the two cases. applied in the new field of organic-based
Molecular layers will probably also optoelectronic circuits. Improvement in
find their way in optoelectronic de- charge mobility of a field effect transis-
vices. Krueger et al. used a series of tor (FET)-LED device was demonstrated
benzoic acids [12–14] to modify the by using hexamethyldisilazane before
Window layer
CdS
microcrystals
Functional
molecules
CuIn(Ga)Se2
Back contact
Fig. 8 Schematic view (not to scale!) of molecular control over

Cu(In,Ga)Se2 /CdS heterojunction thin film, polycrystalline solar cell.
The molecules are at the interface between the CuInSe2 film and the CdS
microcrystals.
deposition of the conjugated polymer. n- and p-type surface. Gal et al. [111] used
The molecular treatment was suggested an experimental scheme similar to Fig. 8
to promote phase segregation, that is, to show the effect of organic acids on the
structural order at the interface. The electrical characteristics of polycrystalline
earlier-mentioned studies reveal the po- CuIn(Ga)Se2 (CIGSe)/CdS solar cells (see
tential of molecular treatments for tuning Fig. 9). The changes in the I –V char-
the charge transfer and charge transport acteristics could be correlated with the
across (semi)conductor/(semi)conductor molecules’ dipole moments that modify
interfaces. the band line-up at the interface, rather
A different potential application of a than as a direct effect on the surface Vs .
molecular treatment is their use as a Specially in view of the later work [77, 107,
blocking layer against charge leakage. For 128, 130], these results indicate that it is not
example, Langmuir-Blodgett (LB) films of essential to form an ordered layer in order
22-tricosanoic, CH2 =CH(CH2 )20 −COOH to modify device characteristics. However,
[67] and hexadecanol [68] were shown to what is required is a molecule that will
have charge-blocking properties on GaAs- modify the energetics at the interface so
based devices that are superior to those that any charge carrier passing from one
shown with them on (Al,Ga)As ones. side of the junction to the other will be
Using capacitance measurements of the influenced by it.
metal-(molecular insulator)-semiconduc-
tor (MIS) structure indicated reduction of 2.3.7
the interface trap density [83] and of the Summary
barrier characteristics [84].
Solar cells are another field of applica- Fine-tuning of semiconductor surface
tions for functional molecules. This idea is electrical properties can be achieved by
schematically presented in Fig. 8 for a het- grafting multifunctional organic mole-
erojunction solar cell, where the molecules cules onto the surface. In such molecules,
are located at the interface between the one function takes care of the binding
4 a
c
2
Current
[mA]
−2
−0.8 −0.4 0 0.4 0.8

Voltage
[V]
Fig. 9 Current–voltage characteristics of molecular, treated (a and b) and
untreated (c) Au/CdS/Cu(In,Ga)Se2 /Mo solar cells in dark and under
illumination (light intensity: ∼1.5 AM). The molecules used, shown
schematically in Fig. 7, are DMDC and DCDC for a and b, respectively. The
molecules were deposited on the polycrystalline Cu(In,Ga)Se2 surface before
the wet chemical deposition of the CdS [111].
to the semiconductor surface, whereas and help design molecular tools for
another one conveys a desired property controlling semiconductor surfaces. Re-
to that surface and thus to the semicon- cent work shows that this can be carried
ductor. The way in which this can happen over to interfaces of semiconductors in
is conveniently studied by using a series of devices where molecules will directly or in-
molecules in which the desired property is directly influence active device characteris-
varied in a systematic fashion. tics. In this way, even small modifications
Implementation of molecular tools pro- in the molecular structure can induce sig-
vides new opportunities for hybrid or- nificant and predictable changes in device
ganic–inorganic systems. The increasing behavior.
interest in, and importance of, nanoscale
technologies and the use of nanocrys-
tals, where surface-related effects often Acknowledgments
dominate bulk properties [106, 125–127],
adds to the importance of controlling For our own contributions mentioned
and engineering surface properties. Cor- in this chapter we thank Leeor Kro-
relating the molecular properties of the nik, Ellen Moons, Merlin Bruening,
grafted molecules to the changes in Stephane Bastide, Dori Gal, Ayelet Vi-
the macroscopic semiconductor proper- lan, Tamar Moav and Rahel Lazar, as
ties can extend our understanding of well Ron Naaman and his group, for
molecule-surface interaction mechanisms fruitful collaborations. We acknowledge
the pivotal contributions of the late 10. S. D. Evans, A. Ulman, Surface Potential
Dr. Jacqueline Libman to the early parts of Studies of Alkyl-Thiol Monolayers Ad-
sorbed on Gold, Chem. Phys. Lett. 1990,
our work and Art Ellis (Univ. of Wisconsin,
170(5, 6), 462–466.
Madison) for collaboration and inspira- 11. I. H. Campbell et al., Controlling Schottky
tion for the molecule-surface interaction Energy Barriers in Organic Electronic
mechanism. Partial support from the US- Devices using Self-assembled Monolayers,
Israel Binational- and the Israel Science Phys. Rev. B 1996, 54, R1432–R14 324.
Foundations, the Minerva Foundation 12. M. Bruening et al., Polar Ligand Adsorption
Controls Semiconductor Surface Potentials,
(Munich), the Fussfeld Fund and the Ger-
J. Am. Chem. Soc. 1994, 116, 2972–2977.
man BMBF, through its bilateral Energy 13. M. Bruening et al., Controlling The Work
Research Program with the Israel Ministry Function of CdSe by Chemisorption of Ben-
of Science is gratefully acknowledged. zoic Acid Derivatives and Chemical Etching,
J. Phys. Chem. 1995, 99, 8368–8373.
14. S. Bastide et al., Controlling the Work
References Function of GaAs by Chemisorption of
Benzoic Acid Derivatives, J. Phys. Chem.,
1. E. H. Rhoderick, R. H. Williams, Metal- 1997, 101, 2678–2684.
Semiconductor Contacts, 2nd edition, Ox- 15. D. M. Taylor, O. N. Oliveira, H. Morgan,
ford, Clarendon, 1988. Models for Interpreting Surface Potential
2. H. Lüth, Surfaces and Interfaces of Solid Measurements and their Application to
Materials, 3rd edition, Springer, Germany, Phospholipid Monolayers, J. Coll. Int. Sci.
1998. 1990, 139(2), 508–518.
3. G. V. Hansson, R. I. G. Uhrberg, Photo- 16. J. F. Dorsten, J. E. Maslar, P. W. Bohn, Near-
electron Spectroscopy of Surface States
Surface Electronic Structure in GaAs (100)
on Semiconductor Surfaces, Sur. Sci. Rep.
Modified with Self-assembled Monolayers
1988, 9, 197–292.
of Octadecylthiol, Appl. Phys. Lett. 1995, 66,
4. A. Many, Y. Goldstein, N. B. Grover, Semi-
1755–1757.
conductor Surfaces, Amsterdam, North-
17. S. M. Sze, Physics of Semiconductor Devices,
Holland, 1965.
7th edition, 2nd Chapter, John Wiley &
5. D. V. Lang, C. H. Henry, Nonradiative Re-
Sons, USA, 1981.
combination at Deep Levels in GaAs
18. W. Mönch, Semiconductor Surfaces and In-
and GaP by Lattice-Relaxation Multi-
phonon Emission, Phys. Rev. Lett. 1975, 35, terfaces, 2nd edition, in Surface Science
1525–1528. (Ed.: E Gerhard), Springer, Berlin, 1995,
6. H. Ishii et al., Energy Level Alignment pp. 274–284.
and Interfacial Electronic Structures at 19. F. Seker et al., Surface Chemistry of Proto-
Organic/Metal and Organic/Organic Inter- typical Bulk II–VI and III–V Semiconduc-
faces, Adv. Mater. 1999, 11(8), 605–625. tors and Implications for Chemical Sensing,
7. A. M. Green, W. E. Spicer, Do We Need a Chem. Rev. 2000, 100, 2505–2536.
New Methodology for GaAS Passivation? 20. J.-F. Fan, H. Oigawa, Y. Nannichi, The
J.Vac. Sci. Technol. A 1993, 11, 1061–1069. Effect of (NH4 )2 S Treatment on the
8. M. Bruening et al., Simultaneous Con- Interface Characteristics of GaAs MIS
trol of Surface Potential and Wetting of Structures, Jap. J. Appl. Phys. 1988, 27,
Solids with Chemisorbed Multifunctional L1331–L1333.
Ligands, J. Am. Chem. Soc. 1997, 119, 21. J.-F. Fan, H. Oigawa, Y. Nannichi, Metal-
5720–5728. Dependent Schottky Height with the
9. S. D. Evans et al., Self-Assembled Mono- (NH4 )2 Sx Treated GaAs. Jap. J. Appl. Phys.
layers of Alkanethiols Containing a Polar 1988, 27(11), L2125–L2127.
Aromatic Group: Effect of the Dipole Posi- 22. D. Liu et al., Deep level transient spec-
tion on Molecular Packing, Orientation, and troscopy study of GaAs surface states treated
Surface Wetting Properties, J. Am. Chem. with inorganic sulfides, Appl. Phys. Lett.
Soc. 1991, 113, 4121–4131. 1988, 53, 1059–1061.
23. H. Hasegawa et al., Control of Fermi Level 36. C. J. Sandroff et al., Electronic Passivation
Pinning and Recombination Processes at of GaAs Surfaces Through the Formation
GaAs Surfaces by Chemical and Photo- of Arsenic – Sulfur Bonds, Appl. Phys. Lett.
chemical Treatments, J. Vac. Sci. Technol. 1989, 54(4), 362–364.
B 1988, 6, 1184–1192. 37. J. S. Herman, F. L. Terry, Hydrogen Sulfide
24. S. Hildebrandt et al., Optical Study of Band- Plasma Passivation of Gallium Arsenide,
Bending and Interface Recombination Appl. Phys. Lett. 1992, 60, 716–717.
at Sb, S, and Se Covered Gallium 38. B. A. Parkinson, A. Heller, B. Miller, Ef-
Arsenide Surfaces, Appl. Surf. Sci. 1993, 63, fects of Cations on the Performance of
153–157. the Photoanode in the N-GaAs/K2 Se-K2 Se2 -
25. R. S. Besser, C. R. Helms, Effect of Sodium KOH/C Semiconductor Liquid Junction
Sulfide Treatment on Band Bending in Solar Cell, J. Electrochem. Soc. 1979, 126(6),
GaAs, Appl. Phys. Lett. 1988, 52, 1707–1709. 954–960.
26. H. Shen, W. Zhou, J. Pamulapati, Photore- 39. P. Viktorovitch et al., Improved Electronic
flectance Study of H2 S Plasma-Passivated Properties of GaAs Surfaces Stabilized with
GaAs Surfaces, Appl. Phys. Lett. 1999, 74, Phosphorous, Appl. Phys. Lett. 1991, 58(21),
1430–1432. 2387–2389.
27. V. L. Berkovits et al., Chemical and Photo- 40. H. H. Lee, R. J. Racicot, S. H. Lee, Surface
chemical Processes in Sulfide Passivation of Passivation of GaAs, Appl. Phys. Lett. 1989,
GaAs(100), In Situ Optical Study and Pho- 54(8), 724–726.
toemission Analysis, J. Vac. Sci. Technol. A 41. L. Ferrari et al., Optical and Spectroscopic
1998, 16, 2528–2538. Characterization of GaAs Passivated Sur-
28. M. Stutzmann, C. P. Herrero, Hydrogen faces, Surf. Sci. 1995, 331–333, 447–452.
Passivation of Shallow Acceptors in Silicon, 42. D. Troost et al., Pinning of the Fermi level
Phys. Scripta. 1989, T25, 276–282. Close to the Valence-Band Top by Chlorine
29. M. J. Bedzyk, W. M. Gibson, J. A. Golov- Adsorbed on Cleaved GaAs (110) Surfaces,
chenko, X-ray Standing Wave Analysis for J. Vac. Sci. Technol. 1987, B5(4), 1119–1124.
Bromine Chemisorbed on Silicon, J. Vac. 43. H. J. Clemens, J. Von Wienskowski,
Sci. Technol. 1982, 20, 634–637. W. Mönch, On the Interaction of Cesium
30. R. I. G. Uhrberg et al., Electronic Structure, with GaAs(110) and Ge(111) Surfaces: Work
Atomic Structure, and the Passivated Na- Function Measurements and Adsorption
ture of the Arsenic-Terminated Si (111) Site Model, Surf. Sci. 1978, 78, 648–666.
Surface, Phys. Rev. B 1987, 35, 3945–3951. 44. P. Skeath et al., Bonding of Antimony on
31. B. A. Kuruvilla et al., Passivation of GaAs GaAs (110): A Prototypical System for
(100) Using Selenium Sulfide, J. Appl. Phys. Adsorption of Column-V Elements on
1993, 73, 4384–4387. III–V Compounds, Phys. Rev. B 1983, 27,
32. J. Sun et al., Chemical Bonding and 6246.
Electronic Properties of SeS2 -Treated 45. K. Jacobi, G. Steinert, W. Ranke, Iodine
GaAs(100), J. Appl. Phys. 1999, 85, Etching of the GaAs(-1-1-1) Surface Studied
969–977. by LEED, AES, and Mass Spectroscopy, Surf.
33. E. Yablonovitch et al., Band-Bending, Fermi Sci. 1976, 57, 571–579.
Level Pinning and Surface Fixed Charge on 46. K. Vaccaro et al., Indium Phosphide Passi-
Chemically Prepared GaAs Surfaces, Appl. vation Using Thin Layers of Cadmium Sul-
Phys. Lett. 1989, 54, 555–557. fide, Appl. Phys. Lett. 1995, 67(4), 527–529.
34. E. Yablonovitch et al., Nearly Ideal Elec- 47. D. N. Bose, Y. Ramprakash, S. Basu, Char-
tronic Properties of Sulfide-Coated GaAs acterization of n-InP Surfaces Before and
Surfaces, Appl. Phys. Lett. 1987, 51(6), After Surface Modification, Mater. Lett.
439–441. 1989, 88(9), 364–368.
35. C. J. Sandroff et al., Dramatic Enhance- 48. D. N. Bose, J. N. Roy, S. Basu, Improved
ment in the Gain of a GaAs/AlGaAs Schottky Barrier on n-InP by Surface Modi-
Heterostructure Bipolar Transistor by Sur- fication, Mater. Lett. 1984, 2(5B), 455–457.
face Chemical Passivation, Appl. Phys. Lett. 49. A. J. Nelson, S. P. Frigo, R. A. Rosenberg,
1987, 51(1), 33–35. Chemistry and Electronic Structure of the
H2 Plasma Passivated Surface of CdTe, 63. R. Maoz, J. Sagiv, On the Formation and
J. Appl. Phys. 1994, 73(3), 1632–1637. Structure of Self-Assembling Monolayers,
50. D. N. Bose, S. Basu, K. C. Mandal, Char- J. Coll. Int. Sci. 1984, 100(2), 465–496.
acterization of Chemically Modified CdTe 64. A. Ulman, An Introduction to Ultrathin
Surfaces, Thin Solid Films 1988, 164, 13–19. Organic Films, Academic Press, New York,
51. D. N. Bose et al., XPS Investigation of 1991.
CdTe Surfaces: Effect of Ru Modification, 65. R. Cohen et al., Molecular Electronic Tun-
Semicon. Technol. 1989, 4, 866–870. ing of Si Surfaces, Chem. Phys. Lett. 1997,
52. J. L. Shaw et al., Chemically Controlled 279, 270–274.
Deep Level Formation and Band-Bending 66. B. Sweryda et al., A Comparison of Porous
at Metal-CdTe Interfaces, Appl. Phys. Lett. Silicon and Silicon Nanocrystallite Pho-
toluminescence Quenching with Amines,
1988, 53(18), 1723–1725.
J. Phys. Chem. 1996, 100(32), 13 776–13 780,
53. D. Cahen, R. Noufi, Surface Passivation of
and reference therein.
Polycrystalline, Chalcogenide Based Pho-
67. J. M. Lauerhaas, M. J. Sailor, Chemical
tovoltaic Cells, Solar Cells 1991, 30(1–4),
Modification of the Photoluminescence
53–59. Quenching of Porous Silicon, Science 1993,
54. Y. Nemirovsky, D. Rosenfeld, Surface Pas- 261, 1567–1568.
sivation and 1/f Noise Phenomena in 68. S. R. Lunt, P. G. Santangelo, N. S. Lewis,
HgCdTe Photodiodes, J. Vac. Sci. Technol. Passivation of GaAs Surface Recombination
1990, A8(2), 1159–1167. with Organic Thiols, J. Vac. Sci. Technol. B
55. Y. F. Chen et al., Hydrogen Passivation in 1991, 9(4), 2333–2336.
Cd1−x Znx Te Studied by Photolumines- 69. J. K. M. Chun et al., Proton Gated Emission
cence, Appl. Phys. Lett. 1991, 58(5), 493–495. from Porous Silicon, J. Am. Chem. Soc.
56. Y. F. Chen, W. S. Chen, Influence of Hy- 1993, 115, 3024–3025.
drogen Passivation on the Infrared Spectra 70. R. R. Chandler et al., Steric Considerations
of Hg0.8 Cd0.2 Te, Appl. Phys. Lett. 1991, in the Amine-Induced Quenching of Lumi-
59(6), 703–705. nescent Porous Silicon, J. Phys. Chem. 1995,
57. D. Cahen, R. Noufi, Defect Chemical Ex- 99(21), 8851–8855.
planation for the Effect of Air Anneal on 71. J. M. Lauerhaas et al., Reversible Lumines-
CdS/CuInSe2 Solar Cell Performance, Appl. cence Quenching of Porous Si by Sol-
Phys. Lett. 1989, 54(6), 558–560. vents, J. Amer. Chem. Soc. 1992, 114(5),
58. S. R. Lunt et al., Chemical Studies of the 1911–1912.
Passivation of GaAs Surface Recombination 72. A. Vilan, Chemical Modification of the
Using Sulfides and Thiols, J. Appl. Phys. electronic Properties of GaAs Surface, M.Sc
1991, 70(12), 7449–7467. thesis, Weizmann Institute of Science,
59. C. J. Sandorff et al., Enhanced Electronic Rehovot, 1996.
Properties of GaAs Surfaces Chemically 73. D. G. Wu et al., Novel NO biosensor based
on surface derivatization of GaAs by
Passivated by Selenium Reactions, J. Appl.
‘‘hinged’’ Fe porphyrin, Angew. Chem. Int.
Phys. 1990, 67(1), 586–588.
Ed. 2000, 39, 4496–4500.
60. S. A. Chambers, V. S. Sundaram, Struc-
74. R. Cohen et al., Frontier Orbital Model of
ture, Chemistry, and Band-Bending at Se-
Semiconductor Surface Passivation: Dicar-
Passivated GaAs(001) Surfaces, Appl. Phys. boxylic Acids on n- and p-GaAs, Adv. Mater.
Lett. 1990, 57, 2342–2344. 2000, 12, 33–37.
61. W. C. Simpson, J. A. Yarmoff, Fundamen- 75. R. Cohen et al., Controlling Electronic Prop-
tal studies of halogen reactions with III–V erties of CdTe by Adsorption of Dicarboxylic
semiconductor surfaces, Ann. Rev. Phys. Acid Derivatives: Relating Molecular Pa-
Chem. 1996, 47, 527–554. rameters to Band-Bending and Electron
62. Y. Liu, A. J. Komrowski, A. C. Kummel, Affinity Changes, Adv. Mater. 1997, 9,
Side–selective reaction of Br2 with second 746–749.
layer Ga atoms on the As-rich GaAs(001) 76. R. Cohen et al., Molecular Control over
2 × 4 surface, Phys. Rev. Lett. 1998, 81, Semiconductor Surface Electronic Proper-
413–416. ties: Dicarboxylic Acids on CdSe, CdTe,
GaAs and InP, J. Am. Chem. Soc. 1999, 121, 90. M. Sturzenegger, N. S. Lewis, An X-Ray
10 545–10 553. Photoelectron Spectroscopic and Chemical
77. A. Vilan, A. Shanzer, D. Cahen, Molecular Reactivity Study of Routes to Functionaliza-
Control over Au/GaAs Diodes, Nature 2000, tion of Etched InP Surfaces, J. Am. Chem.
404, 166–168. Soc. 1996, 118, 3045–3046.
78. C. D. Bain, A New Class of Self-Assembled 91. J. W. Thackeray et al., Interactions of Di-
Monolayers: Organic Thiols on Gallium ethyldithiocarbamate with n-type Cadmium
Arsenide, Adv. Mater. 1992, 4(9), 591–594. Sulfide and Cadmium Selenide: Effi-
79. W. C. Sheen et al., A New Class of Orga- cient Photoelectrochemical Oxidation to the
nized Self-Assembled Monolayers: Alkane Disulfide and Flat-Band Potential of the
Thiols on GaAs (100), J. Am. Chem. Soc. Semiconductor as a Function of Adsorbate
1992, 114, 1514–1515. Concentration, J. Am. Chem. Soc. 1986, 108,
80. O. S. Nakagawa et al., GaAs Interfaces with 3570–3577.
Octadecyl Thiol Self-Assembled Monolayer: 92. M. J. Natan, J. W. Thackeray, M. S. Wrigh-
Structural and Electrical Properties, Jap. ton, Interaction of Thiols with n-Type
J. Appl. Phys. 1991, 30, 3759–376. Cadmium Sulfide and n-Type Cadmium
81. K. Asai et al., Electronic Passivation of Selenide in Aqueous Solutions: Adsorption
GaAs Surfaces by Electrodeposition of of Thiolate Anion and Efficient Photoelec-
Organic Molecules Containing Reactive trochemical Oxidation to Disulfides, J. Phys.
Sulfur, J. Appl. Phys. 1995, 77, 1582–1586. Chem. 1986, 90, 4089–4098.
82. E. D. Lu et al., A Sulfur Passivation of 93. J. J. Hickman, M. S. Wrighton, Face-Speci-
GaAs Surface by an Organic Molecular, fic Interactions of Anionic Sulfur Donors
CH3 CSNH2 Treatment, Appl. Phys. Lett. with Oriented Crystals of (0001) CdX (X =
1996, 69, 2282–2284. Se, S) and Correlation with Electrochemical
83. M. Tabib-Azar, A. S. Dewa, W. H. Ko, Properties, J. Am. Chem. Soc. 1991, 113,
Langmuir-Blodgett Film Passivation of 4440–4448.
Unpinned n-Type Gallium Arsenide 94. T. Uchihara et al., Effect of Ethylenedi-
Surfaces, Appl. Phys. Lett. 1988, 52, aminetetraacetic Acid on the Photocatalytic
206–208. Activities and Flat-Band Potentials of Cad-
84. V. J. Rao, V. S. Kulkarni, Interface Barrier mium Sulfide and Cadmium Selenide,
Height in n-GaAs(100)/Langmuir-Blodgett J. Phys. Chem. 1990, 94, 415–418.
Film Structures, Thin Solid Films 1991, 198, 95. J. K. Lorenz, T. F. Kuech, A. B. Ellis, Cad-
357–362. mium Selenide Photoluminescence as a
85. N. K. Singh et al., Etching Reactions of Probe for the Surface Adsorption of Di-
C2 H5 I on GaAs(100), Sur. Sci. 1998, 409, alkyl Chalcogenides, Langmuir 1998, 14,
272–282. 1680–1683.
86. P. J. Lasky et al., The Adsorption and 96. G. J. Meyer et al., Semiconductor-Olefin
Thermal Reaction of Dimethylcadmim, Adducts. Photoluminescent Properties of
Dimethylzinc and Trimethylgallium on Cadmium Sulfide and Cadmium Selenide
GaAs(110), Surf. Sci. 1996, 364, 312–324. in the Presence of Butenes, J. Am. Chem.
87. H. T. Lam, N. Venkateswaran, J. M. Vohs, Soc. 1989, 111, 5146–5148.
The Reactions of Diethylzinc on Gallium- 97. C. J. Murphy et al., Photoluminescence-
Rich and Arsenic-Rich Reconstructions of Based Correlation of Semiconductor Elec-
GaAs(100), Surf. Sci. 1998, 401, 34–46. tric Field Thickness with Adsorbate Ham-
88. T. Hou et al., Passivation of GaAs (100) mett Substituent Constants. Adsorption
with an Adhesion Promoting Self-Assemb- of Aniline Derivatives onto Cadmium
led Monolayer, Chem. Mater. 1997, 9, Selenide, J. Am. Chem. Soc. 1990, 112,
3181–3186. 8344–8348.
89. Y. Gu, D. H. Waldeck, Electron Tunneling 98. G. C. Lisensky et al., Electro-Optical Evi-
at the Semiconductor-Insulator-Electrolyte dence for Chelate Effect at Semiconductor
Interface. Photocurrent Studies of the Surfaces, Science 1990, 248, 840–843.
n-InP-Alkanethiol-Ferrocyanide System, 99. G. J. Meyer, G. C. Lisensky, A. B. Ellis, Ev-
J. Phys. Chem. B 1998, 102, 9015–9028. idence for Adduct Formation at the
Semiconductor-Gas Interface. Photolumi- 109. L. Zuppiroli et al., Self-Assembled Mono-

nescent Properties of Cadmium Selenide in layers as Interfaces for Organic Optoelec-
the Presence of Amines, J. Am. Chem. Soc. tronic Devices, Eur. Phys. J. B 1999, 11,
1988, 110, 4914–4918. 505–512.
100. D. R. Neu, J. A. Olson, A. B. Ellis, Photolu- 110. F. Nuesch et al., Chemical Potential Shifts
minescence as a probe of the Adsorption at Organic Device Electrodes Induced by
of Gaseous Boranes onto the Surface of Grafted Monolayers, Chem. Phys. Lett. 1998,
Cadmium Selenide Crystals, J. Phys. Chem. 288, 861–867.
1993, 97, 5713–5716. 111. D. Gal et al., Engineering the Interface
101. J. Z. Zhang, M. J. Geselbracht, A. B. Ellis, Energetics of Solar Cells by Grafting
Binding of Fullerenes to Cadmium Sulfide Molecular Properties onto Semiconductors,
Proc. Indian Acad. Sci. 1997, 109, 487–496.
and Cadmium Selenide Surfaces, Photolu-
112. A. Vilan et al., Real-Time Electronic Moni-
minescence as a Probe of Strong, Lewis
toring of Adsorption Kinetics: Evidence for
Acidity-Driven, Surface Adduct Formation,
Two-site Adsorption Mechanism of Dicar-
J. Am. Chem. Soc. 1993, 115, 7789–7793.
boxylic Acids on GaAs(100), J. Phys. Chem.
102. K. D. Kepler et al., Surface-Bound Carbonyl B 1998, 102, 3307–3309.
Compounds as Lewis Acids. Photolumines- 113. Y. Kanemitsu, Slow Decay Dynamics of
cence as a Probe for the Binding of Ketones Visible Luminescence in Porous Silicon:
and Aldehydes to Cadmium Sulfide and Hopping of Carriers Confined on a Shell
Cadmium Selenide Surfaces, J. Phys. Chem. Region in Nanometer-size Si Crystallities,
1995, 99(43), 16 011–16 017. Phys. Rev. 1993, B48, 12 357.
103. J. Z. Zhang, A. B. Ellis, Adsorption of TCNQ 114. J. K. Lorenz, T. F. Kuech, A. B. Ellis, Cad-
Derivatives onto the Surface of Cadmium mium Selenide Photoluminescence as a
Selenide Single Crystals. Quenching of Probe for the Surface Adsorption of Di-
Semiconductor Photoluminescence by a alkyl Chalcogenides, Langmuir 1998, 14,
Family of Strong-Acids, J. Phys. Chem. 1992, 1680–1683, and reference therein.
96, 2700–2704. 115. W. J. Dollard, M. L. Shumaker, D. H. Wal-
104. R. J. Brainard et al., Modulation of Cad- deck, Time-Resolved Studies of Charge
mium Selenide Photoluminescence Inten- Carrier Relaxation in Chemically Modified
sity by Adsorption of Silapentanes and Semiconductor Electrodes: n-CdSe/Silane
Chlorinated Silanes, J. Phys. Chem. B 1997, Interfaces, J. Phys. Chem. 1993, 97,
101, 11 180–11 184. 4141–4148.
105. S. Licht, D. Peramunage, Rational Electro- 116. A. L. McClellan, Tables of Experimental
lyte Modification of n-CdSe/([KFe(CN)6 ]3- Dipole Moments, Freeman, San Francisco,
/2-) Photoelectrochemistry, J. Electrochem. 1963.
Soc. 1992, 139, L23–L26. 117. J. E. Huheey, Inorganic Chemistry, 3rd edi-
106. J. K. Lorenz, A. B. Ellis, Surfactant-Semi- tion, Harper & Row, New-York, 1983,
pp. 312–325.
conductor Interfaces: Perturbation of the
118. G. L. Miessler, D. L. Tarr, Inorganic Chem-
Photoluminescence of Bulk Cadmium Se-
istry, Prentice Hall, Englewood Cliffs, 1991.
lenide by Adsorption of Tri-n-octylphos-
119. R. Hoffman, Solids and Surfaces: A Chemist’s
phine Oxide as a Probe of Solution Ag-
View of Bonding in Extended Structures, VCH
gregation with Relevance to Nanocrystal Publishers, New York, 1988, pp. 68–71.
Stabilization, J. Am. Chem. Soc. 1998, 120, 120. T. Vondrak, C. J. Cramer, X.-Y. Zhu, The
10 970–10 975. Nature of Electronic Contact in Self-
107. J. Krüger, U. Bach, M. Grätzel, Modifica- Assembled Monolayers for Molecular Elec-
tion of TiO2 Heterojunctions with Ben- tronics: Evidence for Strong Coupling,
zoic Acid Derivatives in Hybrid Molecular J. Phys. Chem. B 1999, 103, 8915–8919.
Solid-State Devices, Adv. Mater. 2000, 12, 121. M. C. Petty, G. G. Roberts, CdTe/Langmuir-
447–451. Film MIS Structures, Electron. Lett. 1979,
108. F. Nuesch et al., Derivatized Electrodes in 15(12), 335–336.
the Construction of Organic Light Emitting 122. I. M. Dharmadasa, G. G. Roberts, M. C.
Diodes, Adv. Mater. 1997, 9, 222–225. Petty, Cadmium Telluride/Langmuir Film
Photovoltaic Structures, Electron. Lett. 1980, tions by a New GaAs-Based Sensor, Chem-
16(6), 201–202. istry Eur. J. 2001, 7, 1743–1749.
123. I. H. Campbell et al., Controlling Charge 130. D. G. Wu et al., Tuning of Au/n-GaAs
Injection in Organic Electronic Devices Diodes with Highly Conjugated Molecules,
Using Self-Assembled Monolayers, Appl. J. Phys. Chem. B 2001, in press.
Phys. Lett. 1997, 71, 3528–3530. 131. G. Ashkenasy et al., Assemblies of
124. H. Sirringhaus, N. Tessler, R. H. Friend, ‘‘Hinged’’ Iron-Porphyrins as Potential
Integrated Optoelectronic Devices Based on Oxygen Sensors, J. Amer. Chem. Soc. 2000,
Conjugated Polymers, Science 1998, 280, 122, 1116–1122.
1741–1744. 132. A. Ivanisevic et al., Linker-enhanced Bind-
125. L. Kronik et al., Surface States and Photo- ing of Metalloporphyrins to CdSe and
voltaic Effects in CdSe Quantum Dot Films, Implications for O2 detection, Langmuir
J. Electrochem. Soc. 1998, 145, 1748–1755. 2000, 16, 7852–7858.
126. A. Hagfeldt, M. Grätzel, Molecular Photo- 133. T. Vondrak et al., Interfacial Electronic
voltaics, Acc. Chem. Res. 2000, 33, 269–277. Structure in Thiolate Self-assembled Mono-
127. D. Cahen et al., Nature of Photovoltaic Ac- layers; Implications for Molecular Elec-
tion in Dye-Sensitized Solar Cells, J. Phys. tronics, J. Amer. Chem. Soc. 2000, 122,
Chem. B 2000, 104, 2053–2059. 4700–4707.
128. Y. Selzer, D. Cahen, Fine-tuning of 134. M. Carrara, F. Nüesch, L. Zuppiroli, Car-
Au/SiO2 /Si diodes by varying interfacial boxylic acid anchoring groups for the
dipoles using molecular monolayers, Adv. construction of self-assembled monolayers
Mater. 2001, 13, 508–511. on organic device electrodes, Synth. Metals
129. D. G. Wu et al., Direct Detection of Low 2001, 121, 1633–1634.
Concentration NO in Physiological Solu-
2.4 Capacitance, Luminescence, and Related Optical Techniques 153
2.4 whereas the p-type semiconductor has an

Capacitance, Luminescence, and Related electron acceptor such as boron atoms in
Optical Techniques p-Si. The space charge in the semicon-
ductor is composed of an ionized electron
Yoshihiro Nakato
donor or acceptor. The space charge layer
Osaka University, Osaka, Japan
of a finite thickness is formed in the semi-
2.4.1 conductor when the density of the electron
Capacitance donor, ND , or the density of the electron
acceptor, NA , is low, in a range from 1014 to
2.4.1.1 Space Charge Layer at a 1018 cm−3 . The principle of electrical neu-
Semiconductor–Electrolyte Interface trality at the electrical double layer leads to
An electrical double layer is formed at a an extended distribution of charges deep
semiconductor–electrolyte interface, sim- toward the interior of the semiconductor.
ilar to a metal–electrolyte interface [1–3]. The situation is rather similar to the for-
The difference, however, is in the charge mation of an extended charge distribution,
distribution between the two interfaces. At called a Gouy layer, in the electrolyte in
a metal–electrolyte interface, the charge cases where the electrolyte concentration
on the metal side is localized just at the is low. The difference is that the space
metal surface, whereas, at a semiconduc- charges in the Gouy layer are composed
tor–electrolyte interface, the charge on the of electrolyte ions and are mobile, whereas
semiconductor side is distributed deep in those in the semiconductor are composed
the interior of the semiconductor, forming of ionized donors or acceptors and there-
a wide space charge layer. In concentrated fore are spatially fixed and immobile.
electrolyte solutions (∼0.5 M and higher),
the charge on the electrolyte side is lo- 2.4.1.2 Differential Capacitance of a
calized at the (outer) Helmholtz layer for Semiconductor–Electrolyte Interface
interfaces with either metals or semicon- The potential distribution in the space
ductors. charge layer of a semiconductor can be
Figure 1 schematically illustrates (1) the given by solving the Poisson equation for
charge distribution, (2) the charge-density a given charge distribution [1–3]. For a
distribution, (3) the electric field distri- semiconductor–electrolyte interface such
bution, (4) the potential distribution, and as the one shown in Fig. 1, if the
(5) the band bending at the semiconduc- density of an electron donor, ND , is
tor–electrolyte interface, under an ideal constant throughout the semiconductor,
condition that no surface charge nor sur- the potential, φ(x), at a distance, x, from
face dipole is present at the semiconductor. the surface is given as follows [1, 2]:
The semiconductor is assumed to be n- ! "
type, but a similar diagram can be drawn qND 1 2
φ(x) = W x − x + φ(0)
for a p-type semiconductor with the sign of ε0 εs 2
electric charges inverted. The band bend- (0 ≤ x ≤ W ) (1)
ing in the semiconductor is simply caused
by the potential distribution in it. where q is the elementary charge, ε0
The n-type semiconductor has an elec- the permittivity of vacuum, εs the dielec-
tron donor in the crystal, doped as an tric constant of semiconductor, φ(0) the
impurity, such as phosphor atoms in n-Si, potential at the surface (x = 0), and W the
− Fig. 1 Schematic illustration of

− + + (a) charge distribution,
− + (b) charge-density distribution,
−
Ions at the − + + (c) electric field distribution,
Helmholtz layer −
−
+ Ionized donor (d) potential distribution, and (e) band
+ +
− bending at an n-type
n -type semiconductor
(a) Electrolyte semiconductor–electrolyte interface
under an ideal condition that neither
Depletion surface charge nor surface dipole is
region n
present at the semiconductor.
Region
Charge density
0 X
0
(b)
Electric field
0 W
0 X
(c)
Potential
X
0
(d)
− eUfb
Energy
EC
EF = − eU
Eredox
EV
(e)
width of the space charge layer. The φ(0) given by

and W are given by
Qs = qND W = [(2qε0 εs ND )
qND W × (U − Ufb − kT /q)]1/2 (4)
φ(0) = δ (2)
ε0 εs
# ! "$1/2 Thus, the differential capacitance of the
2ε0 εs kT space charge layer, Cd , per unit area is
W = U − Ufb −
qND q given as follows:
(3)
#
∂Qs 2
where δ is the thickness of the (outer) Cd = =
∂U qε0 εs ND
Helmholtz layer, U is the electrode
! "$−1/2
potential, and Ufb the flat band potential. kT
× U − Ufb − (5)
The space charge, Qs , per unit area is q
This equation can be rewritten as follows: charges and surface dipoles cause a shift
! " of the semiconductor band positions at the
1 2 kT surface, as shown schematically in Fig. 2.
= U − U fb − (6)
Cd2 qε0 εs ND q The shift is expressed as a change in φ(0) or
Ufb in the foregoing equations. However,
The plot of 1/Cd2 measured against U is the preceding equations themselves can
called the Mott–Schottky plot, which can be be applied to such cases with the changed
used to determine Ufb and ND (or NA ), as φ(0) or Ufb .
discussed later.
It has been assumed thus far that no sur- 2.4.1.3 Measurement of Differential
face charge nor surface dipole is present at Capacitance
the semiconductor. In general cases, both The capacitance of the semiconduc-
surface charges and surface dipoles are tor–electrolyte interface can be measured
present in the semiconductor owing to ad- by use of a semiconductor electrode, in
sorption equilibria of various ions between which the front side of the semiconductor
the electrolyte and the semiconductor sur- is in contact with the electrolyte and the
face as well as formation of polar bonds at rear side is electrically connected with a
the semiconductor surface. Such surface metallic leading wire via an ohmic contact.
−
− +
Ions at the + − +
− +
Helmholtz layer + − + +
− + Ionized donor
Surface charge −
+ − + +
n -type semiconductor
Electrolyte
(a)
Depletion n
region Region
Charge density
0 X
0
(b)
Potential
0
X
(c)
Fig. 2 Schematic illustration of
(a) charge distribution, − eUfb
(b) charge-density distribution,
Energy
(c) potential distribution, and (d) band EC

bending at the EF = − eU
semiconductor–electrolyte interface Eredox
under a condition that negative surface
(d) EV
charges are present.
Fig. 3 Experimental setup for

Recorder measurements of capacitance at
Impedance analyzer or computer
the semiconductor–electrolyte
interface. W.E.: working
Function electrode (semiconductor
Potentiostat
generator electrode), C.E.: counter
electrode, and R.E.: reference
electrode.
R.E.
C.E. W.E.
The experimental setup for capacitance hydrogen iodide [5]. It should be noted that
measurements is schematically shown in the slope of the straight line depends not
Fig. 3. The electrode potential U is regu- only on ND (or NA ) and εs but also on the
lated with a potentiostat. The differential true surface area (or surface roughness) of
capacitance is measured by superimposing the semiconductor electrode.
an AC voltage with a small amplitude of The Ufb is one of the most important
about 10 mV and a frequency of a few Hz quantities for semiconductor electrodes
to 1 MHz on the electrode potential. One because it determines the band edge
can use a commercial impedance analyzer positions at the semiconductor–electrolyte
to measure the differential capacitance, to- interface, which in turn, determine the
gether with a personal computer to analyze energies of conduction-band electrons
obtained data automatically. and valence-band holes reacting with the
Recently, scanning capacitance micro- electrolyte solution. It is known that Ufb
scopy (SCM) using a fine tip has been for most semiconductors, such as n- and
developed to investigate two-dimensional p-GaAs, GaP, InP, n-ZnO, n-TiO2 , and
dopant profiling of semiconductor sur- n-SnO2 , in aqueous electrolytes is solely
faces [4]. determined by the solution pH and shifts
in proportion to pH with a ratio of
2.4.1.4 Mott–Schottky Plots and Flat Band −0.059 V/pH [1, 2]. This is explained by an
Potentials adsorption equilibrium for H+ or OH− at
Equation 6 indicates that a plot of 1/Cd2 the semiconductor–electrolyte interface,
against U gives a straight line with a for example,
slope of (2/qε0 εs ND ), which is termed the
Ss -OH + Haq + &
−−
−−−
%− Ss -OH2
+
(7)
Mott–Schottky plot, as mentioned earlier.
The extrapolation of the straight line to where Ss -OH refers to surface OH group
1/Cd2 = 0 gives (Ufb + kT /q). Therefore, at the semiconductor.
the plot can be used to determine the The Ufb for n- and p-Si [6] and metal
flat band potential Ufb . The donor density calcogenide semiconductors such as n-
ND (or the acceptor density NA ) can CdS, n-CdSe, and CdTe [2, 7] does not
also be determined from the slopes obey the foregoing rule, remaining nearly
of the plots. Figure 4 shows examples constant in a range of pH lower than about
of Mott–Schottky plots, obtained for 6 for Si and about 10 for n-CdS. This is
n-Si(111) and n-Si(100) electrodes in 7.1 M most probably because the semiconductor
Fig. 4 Examples of 6
Mott–Schottky plots obtained n- Si (111)
for n-Si(111) and n-Si(100) 5 10.0 ∼ 15.0 Ωcm
(1/C2) × 1017
electrodes in aqueous 7.1 M HI. 4
[F−2cm4]
3
2 n- Si (100)
0.8 ∼ 1.16 Ωcm
1
0
−0.8 −0.4 0 0.4 0.8 1.2 1.6
Potential versus SCE
[V]
surface has no OH group in this pH range the radiative recombination may occur ei-
and no adsorption equilibrium for H+ or ther in the semiconductor bulk or at the
OH− is attained. For metal calcogenide semiconductor surface (Fig. 5a).
semiconductors such as n-CdS, the Ufb On the other hand, EL is emitted when
shifts by adsorption of HS− and Cd2+ an n-type semiconductor electrode, for ex-
ions. It is also known that Ufb for ample, is negatively biased in an electrolyte
some semiconductor electrodes shifts by a solution containing a strong oxidant. Un-
change in the surface termination bond [8, der this condition, holes are injected into
9], as well as electrode illumination [10, the valence band by the oxidant, some
11] and the presence of a redox couple [10, of which recombine with electrons in the
12]. The Ufb in nonaqueous electrolytes conduction band existing as the majority
has been reported for some semiconductor carrier in the n-type semiconductor, result-
electrodes [13, 14]. ing in emission of luminescence called EL,
as illustrated in Fig. 5(b). Similarly, EL is
2.4.2 emitted when a p-type semiconductor elec-
Luminescence from Semiconductor trode is positively biased in an electrolyte
Electrodes
solution containing a strong reductant be-
2.4.2.1 Photoluminescence and cause some of the electrons, injected into
Electroluminescence the conduction band by the reductant, re-
Illumination of a semiconductor electrode combine with holes existing as the majority
generates excited electrons in the con- carrier in the p-type semiconductor. In
duction band and holes in the valence both cases, EL may be emitted either via
band. Some of them recombine with each direct recombination between the conduc-
other radiatively, resulting in emission of tion and valence bands or via an impurity
luminescence called photoluminescence level(s), and either via recombination in
(PL). The radiative recombination may the semiconductor bulk or at the semi-
occur directly between the conduction conductor surface. The situation is quite
and valence bands (inter-band transition) the same as that in PL emission (Fig. 5).
or via certain impurity or defect levels Figure 6 shows examples of PL and
within the band gap at which either elec- EL spectra from some semiconductor
trons or holes, or both are trapped, as electrodes [15]. It is to be noted, however,
schematically illustrated in Fig. 5(a). Also, that such spectra strongly depend on
Conduction band
−
− − − − −
Surface
luminescence
Illumination
Bulk luminescence
Surface
states
+ +
+
Valence band
(a)
Conduction band
− − − −
Surface luminescence
Bulk luminescence
Surface
states
+ +
e Oxidant
+ (Fe3+, HO , or SO4 −)
• •
Valence band
(b)
Fig. 5 Mechanisms of emission of (a) PL and (b) electroluminescence
(EL) from a semiconductor electrode in an electrolyte solution.
semiconductor materials used (that is, DC method as shown in Fig. 7, although

the kind and the amount of impurities no illumination equipment is necessary for
and/or defects) and the kind of surface EL measurements. The spectra in Fig. 6
treatments made. are measured by this conventional DC
method.
2.4.2.2Measurements of Luminescence In the DC method, the luminescence
from Semiconductor Electrodes from a semiconductor electrode is usually
collected in a normal direction by use of a
2.4.2.2.1 DC Methods The luminescence lens and led to a slit of a monochromator,
(PL or EL) from a semiconductor electrode followed by detection with an appropriate
can be measured by a simple conventional photomultiplier. The electrode potential of
Fig. 6 Examples of PL and EL λ

spectra of semiconductor [nm]
electrodes. Eg : the band gap 400 500 600 700 800 1000
energy, hν: photon energy, λ:
wavelength. n -GaP
Eg
Luminescence intensity n -CdS

Eg
n -ZnO
Eg
n -TiO2
Eg
25 20 ∼
v 15 10
[103 cm−1]
3.0 2.5 2.0 1.5

hν
[eV]
Monochromater
Mirror
Hg lamp
Filter
Monochromater
Photomultiplier
Pt
Semiconductor V
SCE
Potentiostat Potential sweeper

X-Y recorder
Fig. 7 Experimental setup for simple DC measurements of PL and

EL from a semiconductor electrode.
the semiconductor electrode is regulated may change. Recently, space-resolved two-

with a potentiostat. By this way, one can dimensional detection of EL has been
measure PL or EL spectra at constant developed by use of a CCD camera [18].
electrode potentials, and the lumines-
cence intensity versus potential curves as 2.4.2.2.2 Pulsed Techniques Time-resol-
compared with simultaneously measured ved PL (and EL) measurements give much
current versus potential curves. The use of information on carrier dynamics in the
a personal computer is convenient for data semiconductor bulk and at the semi-
processing and storage. conductor surface. Figure 8 schematically
For PL measurements, the semiconduc- illustrates a convenient method, called
tor electrode is illuminated by monochro- a time-correlated single photon counting
matic light, using a high-pressure mercury method or a picosecond single photon tim-
lamp, a nitrogen laser, an argon laser, and ing method [19, 20, 21], by which weak PL
so forth, as the light source. Much care can be measured at a high sensitivity with
should be taken to minimize a contri- a time resolution on the order of picosec-
bution of scattered stray light from the onds. A picosecond or femtosecond pulse
illumination light to luminescence spec- laser with a pulse width of a few ps is
tra. The detection limit of weak PL is used for excitation of the sample (semi-
in most cases determined by how much conductor electrode). The luminescence
the intensity of the scattered stray light is from the sample is usually detected
reduced. High-intensity illumination with with a microchannel plate photomultiplier
high-quality monochromatic light by use (MCP-PMT) having a small transient time
of a laser is effective to detect weak PL, spread. A part of the excitation laser pulse
but very high intensity illumination gener- is split from the main beam and detected
ates very high density electrons and holes with a photodiode for use as the starting
at the semiconductor surface, which may pulse. The luminescence signal, detected
alter the essential mechanism of carrier with the MCP-PMT, is used as the stop-
dynamics in the semiconductor. Very high ping pulse. This means that the sample
intensity illumination may also generate was excited at the time of the starting
much heat (temperature increase) at the pulse and a luminescent photon was emit-
semiconductor surface. ted at the time of the stopping pulse. The
For EL measurements, the detection starting and stopping pulses are led to a
limit of weak EL is mainly determined by time-to-amplifier converter (TAC) to mea-
the detection sensitivity of an experimental sure the time separation between them.
apparatus used because no illumination Repeated measurements give a histogram
is needed in this case. One can use of the number of luminescent photons
various high-sensitivity detection methods versus time, which should agree with a
such as a lock-in amplifier [16, 17], boxcar luminescence decay curve.
integration [17], and photon counting [17]. PL measurements with a higher time
Care should, however, be taken when the resolution of several tens of femtoseconds
time-averaging method is used because the can be achieved by means of femto-
luminescence intensity often changes with second up-conversion spectroscopy [22].
time even at a constant electrode potential, The experimental setup is illustrated in
probably because the surface chemical Fig. 9. In this method, the luminescence
structure of the semiconductor electrode intensity is measured by making use
Photodiode Multichannel
pulse height
Laser analyzer
Start
Monochromater TAC A/D Conv.
Stop
Time difference Voltage change

Sample
MCP-PMT
Counts
Time
Picosecond single-photon timing spectroscopy
Fig. 8 Experimental setup for a time-correlated single-photon counting
method or picosecond single-photon timing method, used for
time-resolved luminescence measurements in a range of picoseconds.
PM
T
Pulses from 800 nm,

Ti:sapphire laser. f = 150 mm
82 MHz, 50 fs Spectrometer
10 nJ/pulse. BBO
SHG beam: SFG
∼ 400 nm
Sample
Scanning optical delay,
400 nm,
BBO < 0.5 nJ 850 ps time window
Fundamental beam: ∼ 800 nm
SHG
λ /2 plate
Dichroic
beamsplitter
Fig. 9 Experimental setup for femtosecond up-conversion spectroscopy used for time-resolved
luminescence measurements in a range of several tens femtoseconds.
of ‘‘gating’’ by another laser pulse. A up-conversion spectroscopy [23]. An aque-

femtosecond laser, such as a Ti-sapphire ous or methanol suspension of nanocrys-
laser, is normally used as the light source talline TiO2 particles, on which a cumarin
for excitation. The emitted luminescence dye is adsorbed, is used as the sample.
with a frequency ω1 is mixed with a The luminescence is emitted from the ad-
gating laser pulse having a frequency ω2 sorbed dye, and its decay represents the
in a nonlinear crystal such as BBO (β- rate of electron transfer from excited ad-
barium borate) to give an up-conversion sorbed dye to the conduction band of TiO2 .
light pulse with a frequency ω3 = ω1 + ω2 ,
which is measured with a photomultiplier. 2.4.2.3 Bulk and Surface Luminescence
The up-conversion light pulse with ω3 The radiative recombination of electrons
is generated only when the emitted and holes can occur either in the semi-
luminescence and the gating laser pulse conductor bulk (bulk PL or EL) or at
come to BBO simultaneously. Also, the the semiconductor surface (surface PL or
intensity of the generated up-conversion EL), as mentioned earlier. In semicon-
light is in proportion to the product of ductor electrochemistry, it is important
the intensity of the emitted luminescence to distinguish whether observed PL or
and that of the gating laser pulse. Thus, EL is emitted from the semiconductor
the luminescence decay curve in relative bulk or surface because surface PL and
intensity is measured by giving various EL give more information on surface
delays to the gating laser pulse. structures and processes at semiconductor
Figure 10 shows examples of lumines- electrodes, such as surface states, surface
cence decays obtained by the femtosecond reaction intermediates, and so forth.
1000 C343/MeOH
( τ ET > 1.5 psec)
Counts
D-1421/MeOH
D-1421/water ( τ ET = 30 fsec)
C343/water ( τ ET < 20 fsec)
0
−0.5 0 0.5 1.0 1.5
Time
[ps]
Fig. 10 Examples of luminescence decays obtained by femtosecond up-conversion
spectroscopy. The luminescence is emitted from a cumarin dye adsorbed on nanocrystalline
TiO2 particles suspended in water or methanol (MeOH). C343: cumarin 343 and D-1421:
7-diethylaminocumarin-3-carboxylic acid.
Naturally, the intensity of surface PL is dependence of the luminescence intensity.

strongly influenced by the chemical com- The surface luminescence is observed only
position of the electrolyte, in particular, under a forward bias with weak band bend-
by whether oxidants or reductants are in- ing or in the presence of an accumulation
cluded in it because electrons or holes layer, namely, the surface luminescence is
trapped at surface states can easily react observed only when the surface densities
with oxidants or reductants in the elec- of both electrons and holes are sufficiently
trolyte, resulting in luminescence quench- high. On the other hand, the bulk lumines-
ing. The intensity of surface PL is also cence is observed even in the presence of
strongly dependent on surface pretreat- large band bending under which the sur-
ments. However, it is to be noted that face density of the majority carrier is very
bulk PL is also influenced by these factors low, if the penetration depth of incident
because photo-generated electrons and/or illumination light is considerably larger
holes in the semiconductor bulk can dif- than the width of the space charge layer.
fuse to the surface and react with an The arguments made thus far show that
oxidant or reductant in the electrolyte and both bulk and surface PL can be used
also surface species (surface states) before to investigate surface processes at semi-
they recombine. This leads to a decrease conductor electrodes. The measurements
in the densities of electrons and holes in of bulk PL indicate that most of semi-
the semiconductor bulk and hence to a conductors emit bulk PL. They can be
decrease in the intensity of bulk PL. used to detect some surface processes and
One effective way to distinguish exper- reactions [24]. The measurement of sur-
imentally the bulk and surface lumines- face PL, if measured, is very powerful be-
cence is to measure the electrode-potential cause its behavior gives direct information
−1.0 Conduction band

− Luminescence
(840 nm)
−
− −
0.0 ε0 I−/I3−
Ef O2
U versus SCE/V
H2Q/Q
X1.47 e
Br − / Br2
1.0 hν
(Illumination) CI− / CI2
2.0
OH−
3.0 +
Valence band
n-TiO2 Electrolyte (pH 1.0)
Fig. 11 Energy band diagram for explaining a PL band at 840 nm, emitted via a
surface state connected with a surface reaction intermediate (X1.47 ) of water
photooxidation on n-TiO2 .
on surface structures and processes. The reaction intermediate (X1.47 , probably va-
most important is to measure surface PL cant 2p-level of surface Ti-O· radical) of
that is emitted via surface states connected water photooxidation, as shown in Fig. 11.
with surface intermediates of electrode re-
actions. An example of such luminescence 2.4.3
is a luminescence band peaked at 840 nm Other Optical Techniques
(Fig. 6), observed for a rutile-type n-TiO2
electrode, which was activated beforehand 2.4.3.1 Time-resolved Laser Spectroscopy
by photoetching in aqueous H2 SO4 un-
der anodic bias. In situ PL measurements, 2.4.3.1.1 Transient Absorption Spectro-
combined with in situ photocurrent mea- scopy Measurements of transient ab-
surements, can lead to detailed studies on sorption spectra are much more difficult
molecular mechanisms of electrode reac- than the measurements of luminescence
tions. It is concluded [25] that the 840-nm (PL) described thus far, especially for solid
band for rutile n-TiO2 is emitted via a samples. However, time-resolved tran-
transition of an electron in the conduc- sient absorption spectroscopy has been
tion band to a vacant level of a surface used to study carrier dynamics in the
Femtosecond Ti:sapphire laser system

780 nm, 3 mJ/pulse, 170 fs (10 Hz)
SHG
2ω
ω
H2O
Optical delay
(in 1 cm cell)
fs white-light
continuum
Pump pulse
(390 nm)
Monochro- MCPD2
mator
Powder sample
(in 2 mm cell)
Iris
Monochro-
mator
; Mirror
MCPD1
; Dichroic mirror
; Filter
Fig. 12 Experimental setup for femtosecond (pump probe)

transient absorption spectroscopy.
semiconductor bulk and at the semicon- conduction band of TiO2 particles. The rate
ductor surface. Figure 12 shows an exam- of photoelectron transfer from adsorbed
ple of experimental setup for femtosecond dyes to the conduction band of TiO2 in
(pump probe) transient absorption spec- vacuum and solution is also determined by
troscopy [26]. The principle is to generate use of measurements of transient absorp-
an excited state of the sample (semiconduc- tion of injected electrons. It is reported
tor electrode) by a laser pulse (pump pulse) that the photoelectron transfer occurs in
and then measure an absorption spectrum a time range of 150–25 femtoseconds
of the excited state by another laser pulse for chemically bound, strongly interacting
(probe pulse), with varied delays in time dyes [28–31].
from the pump pulse. A femtosecond
laser, such as a Ti-sapphire laser, is used 2.4.3.1.2 Transient Grating Spectroscopy
as the light source. In Fig. 12, a laser pulse Transient grating spectroscopy is rela-
with a frequency ω from the light source tively easily handled compared with the
is converted with BBO to a light pulse with transient absorption spectroscopy, and is
2ω for use as the pump pulse. The original often used to study carrier dynamics at
pulse with ω is separated with a mirror semiconductor electrodes [32]. Figure 14
and used for the probe pulse. The latter schematically shows the principle of tran-
pulse is led through an optical delay cir- sient grating spectroscopy. A femtosecond
cuit to an H2 O cell to be converted into laser pulse for sample excitation is split
white light continuum for measurements into two beams, which are crossed again
of absorption spectra. The spectral inten- at the semiconductor surface to produce
sity of the probe pulse is measured with a an optical striped interference pattern. The
multichannel photodiode (MCPD). interference pattern produces a striped pat-
Figure 13 shows an example of tran- tern of the densities of photo-generated
sient absorption spectra observed by direct electrons and holes near the semicon-
excitation of nanocrystalline TiO2 parti- ductor surface. The latter striped pattern
cles suspended in vacuum [27]. The broad gives rise to a striped pattern of optical
absorption band in the red to near in- refractive index near the semiconductor
frared region is assigned to light absorp- surface, which is monitored by measuring
tion of photo-generated electrons in the a diffraction pattern of a second probe laser
20 0.2 ps
2 ps
10 ps
15 100 ps
% Absorption
1 ns
5 ns
[%]
10
5
Fig. 13 Transient absorption spectra of
nanocrystalline TiO2 particles in
vacuum after excitation by a 0
femtosecond laser pulse (390 nm, 400 450 500 550 600 650 700 750
170 fs, 2 mJ cm−2 ). The delay time after Wavelength
excitation is indicated in the figure. [nm]
Sample Λ = λ ex / 2sin(θ /2) ∼

∆n
Pump
White light
continuum Λ Diffraction
spectrum
θ
Wavelength / nm
Probe
Transient grating
Reflection
Transmission
Fig. 14 A schematic view of a sample region in transient grating spectroscopy to
explain its principle.
Fig. 15 Examples of transient grating

(100) signal decays, observed for an n-TiO2
Diffraction intensity
e (100) electrode by excitation at 360 nm

and probing at 670 nm. The excitation
d pulse intensity in a µW unit is (a) 1.4,
(b) 4.1, (c) 6.1, (d) 9.7, and (e) 42.
c
b
a
0 10 20 30 40 50
Time
[ps]
pulse (white light continuum). The carrier surface structures at semiconductor elec-
dynamics near the semiconductor surface trodes. In situ multiple internal reflectance
is observed as rise and decay curves for Fourier Transform Infrared (FTIR) spec-
the diffraction pattern of the second laser troscopy is widely used to investigate
probe pulse. surface termination bonds and their re-
Figure 15 shows examples of decays of actions. Figure 16 shows a semiconductor
transient grating signal (the intensity of the (Si) wafer and an electrochemical cell
diffraction pattern) observed for an n-TiO2 for in situ multiple internal reflectance
(100) electrode by excitation at 360 nm spectroscopy. By this method, various
and probing at 670 nm [33]. The decays hydrogen termination bonds (Si−H, SiH2 ,
are related with the rate of electron-hole and SiH3 ) at terraces and steps for hy-
recombination near the n-TiO2 surface. drogen fluoride–etched or ammonium
fluoride–etched Si (111) and (100) are
2.4.3.2 In Situ Spectroscopic Investigation clearly detected with good spectral reso-
of Semiconductor Surfaces lution [34, 35].
Various spectroscopic techniques have In situ investigations of semiconduc-
been used for in situ investigations of tor surfaces are also done by means
Fig. 16 A silicon wafer and an IR light

electrochemical cell for in situ Si
multiple internal reflectance
FTIR spectroscopy.
Solution
In situ cell
IR
Si prism
of second harmonic generation (SHG) related with the density of free carriers,
spectroscopy, ultraviolet and infrared light and modulated illumination enables us to
mixing spectroscopy, (surface enhanced) investigate minority carrier dynamics.
laser Raman spectroscopy, X-ray diffrac- Surface photovoltage (SPV) measure-
tion (XRD) method, electron paramag- ments have been used to investigate
netic resonance (EPR) spectroscopy, and electronic structures at the semiconduc-
so forth. Space-resolved investigations are tor surfaces for semiconductor–vacuum
also done by means of surface near-field and semiconductor–gas interfaces. This
optical microscopy (SNOM), photoconduc- method is applied to semiconductor–
tive atomic force microscopy, and so forth. liquid interfaces in case of insulating liq-
uids [39].
2.4.3.3 Other Miscellaneous Techniques
2.4.4
The electrolyte electroreflectance (EER) Summary
method is successfully used [36, 37] to
determine the flat band potential (Ufb ). Various methods to measure the capac-
In this method, the optical reflectance itance and luminescence (PL and EL)
at the semiconductor electrode surface at the semiconductor–electrolyte inter-
is measured under modulation of the face have been reviewed together with
electrode potential by superimposition of related optical techniques to investigate
a small AC voltage. The modulation of surface structures, optical properties, and
the electrode potential causes modulation carrier dynamics at the interfaces. Re-
of the density of majority carriers near cent rapid progress in laser spectroscopy
the semiconductor surface, which in turn, has enabled us to investigate very fast
causes modulation in light reflectance. interfacial processes with a high time
In situ microwave photoconductivity resolution of a few ten femtoseconds.
measurements, combined with photo- Moreover, rapid progress in scanning
current measurements, are also success- probe microscopy, combined with optical
fully used [38] to investigate dynamics techniques, has enabled us to investigate
of photo-generated minority carriers and interfacial structures and processes with
interfacial kinetics at semiconductor elec- a high space resolution of an atomic
trodes. The microwave signal is closely (subnanometer) scale. Progress in these
fields continues and it is thus highly prob- 18. M. Oyama, M. Mitani, M. Washida et al., J.
able that other new powerful methods will Electroanal. Chem. 1999, 473, 166–172.
19. D. V. O’Connor, D. Phillips in Time-correla-
appear in the near future.
ted Single Photon Counting, Academic Press,
New York, 1984.
References 20. I. Yamazaki, N. Tamai, H. Kume et al., Rev.
Sci. Instrum. 1985, 56, 1187–1194.
1. S. R. Morrison in Electrochemistry at Semicon- 21. N. Tamai, M. Ishikawa, N. Kitamura et al.,
ductor and Oxidized Metal Electrodes, Plenum Chem. Phys. Lett. 1991, 184, 398–403.
Press, New York, 1980. 22. G. S. Beddard, T. Dout, G. Porter, Chem.
2. Y. V. Pleskow, Y. Y. Gurevich, in Semicon- Phys. 1981, 61, 17.
ductor Photoelectrochemistry (Translated by 23. K. Murakoshi, S. Yanagida, M. Capel et al.,
P. N. Bartlett), Consultants Bureau, New ACS Symp. Series 679 (Nanostructured
York, 1986. Materials-Clusters, Thin Films, and Compos-
3. S. M. Sze in Physics of Semiconductor Devices, ites), Amer. Chem. Soc. 1997, Chap. 17,
2nd edition, John Wiley & Sons, New York, pp. 221–238.
1981. 24. K. Mecker, A. B. Ellis, J. Phys. Chem. B 1999,
4. V. V. Zavyalov, J. S. McMurray, C. C. 103, 995–1001.
Williams, Rev. Sci. Instrum. 1999, 70, 25. Y. Nakato, H. Akanuma, Y. Magari et al., J.
158–164. Phys. Chem. B 1997, 101, 4934–4939.
5. Unpublished result in a laboratory of 26. T. Asahi, A. Furube, M. Ichikawa et al., Rev.
Prof. Y. Nakato, Osaka University. Sci. Instrum. 1998, 69, 361.
6. Y. Nakato, T. Ueda, Y. Egi et al., J. Electro- 27. A. Furube, T. Asahi, H. Masuhara, J. Phys.
chem. Soc. 1987, 134, 353–358. Chem. B 1999, 103, 3120.
7. H. Minoura, T. Watanabe, T. Oki et al., Jpn. 28. Y. Tachibana, J. E. Moser, M. Graetzel et al.,
J. Appl. Phys. 1977, 16, 865. J. Phys. Chem. 1996, 100, 20056–20062.
8. J. N. Chazalviel, J. Electroanal. Chem. 1987, 29. T. Hannappel, B. Burfeindt, W. Storck et al.,
233, 37. J. Phys. Chem. B 1997, 101, 6799–6802.
9. M. Fujitani, R. Hinogami, J. G. Jia et al., 30. R. J. Ellingson, J. B. Asbury, S. Ferrere et al.,
Chem. Lett. 1997, 1041–1042. J. Phys. Chem. B 1998, 102, 6455–6458.
10. A. J. Nozik, R. Memming, J. Phys. Chem. 31. J. B. Asbury, R. J. Ellingson, H. N. Ghosh
1996, 100, 13061–13078. et al., J. Phys. Chem. B 1999, 103, 3110–3119.
11. A. J. McEvoy, M. Etman, R. Memming, J. 32. J. J. Kasinski, L. A. Comez-Jahn, K. J. Faran
Electroanal. Chem. 1985, 190, 225. et al., J. Chem. Phys. 1989, 90, 1253–1269.
12. Y. Nakato, A. Tsumura, H. Tsubomura, J. 33. Unpublished result in a laboratory of
Electrochem. Soc. 1981, 128, 1300–1304. Prof. N. Tamai, Kwansei-Gakuin University.
13. B. L. Wheeler, A. J. Bard, J. Electrochem. Soc. 34. P. Jakob, Y. J. Chabal, J. Chem. Phys. 1991,
1983, 130, 1680. 95, 2897–2909.
14. W. Jaegermann in Modern Aspects of Electro- 35. M. Niwano, T. Miura, N. Miyamoto, J. Elec-
chemistry (Eds.: R. E. White, B. E. Conway, trochem. Soc. 1998, 145, 659–661.
J. O’M. Bockris), Plenum, New York, 1996, 36. M. Pujadas, P. Salvador, J. Electrochem. Soc.
pp. 136. 1989, 136, 716–723.
15. Y. Nakato, A. Tsumura, H. Tsubomura, 37. L. J. Ferrer, H. Muraki, P. Salvador, J. Phys.
Chem. Phys. Lett. 1982, 85, 387–390. Chem. 1986, 90, 2805–2807.
16. ftp://ftp.batnet.com/pub/wombats/srsys/ 38. A. M. Chaparro, H. Tributsch, J. Phys. Chem.
ftp/lia.pdf B 1997, 101, 7428–7434.
17. ftp://ftp.batnet.com/pub/wombats/srsys/ 39. S. Bastide, D. Gal, D. Cahen, Rev. Sci. In-
ftp/sr.pdf strum. 1999, 70, 4032–4036.
173
3.1 Different strategies are used to prepare

Preparation of Nanocrystalline and keep the particles small and involve
Semiconductor Materials control of nucleation and growth of the
crystallites. This can be achieved in a
Gary Hodes, Yitzhak Mastai
number of different ways and is described
Weizmann Institute of Science, Rehovot, Israel
in the following section.
3.1.1 Electrodeposition normally leads to
Introduction small particle size, largely because it is
a low-temperature technique (molten salt
The field of nanophase materials is one electrolysis being an exception), thereby
that covers a wide and active area of minimizing grain growth. Other factors
research. The various properties of these also contribute to limitations of grain size
materials, including mechanical, optical, and are discussed in this chapter. The low
electrical, and structural, are often very temperatures also minimize interdiffusion
different from the same materials in between electrodeposit and substrate or be-
the ‘‘bulk’’ phase. An example of this tween different electrodeposited layers. In
difference is the case of quantum dots addition, electrodeposition allows a very
(QDs) – nanoparticles that are sufficiently high degree of control over the amount
small that their electronic energy structure of material deposited through Faraday’s
is changed from that of the bulk material. law, which relates the amount of material
The size regime of semiconductor QDs deposited to the deposition charge. This
varies from about 1 nm up to tens of feature is particularly desirable when iso-
nm in size, depending on the material lated nanocrystals are to be deposited on a
properties. substrate.
Considerable effort has been made in Several different methods of electrode-
recent years to synthesize nanocrystalline position, such as reduction of (chalcogen)
semiconductors with control over the size oxide or zero valent (chalcogen) com-
and size distribution and to characterize pounds in aqueous electrolytes, deposition
their properties. As a result of these from nonaqueous solvents, anodization,
studies, it is now possible to synthesize a deposition from molten salts, and depo-
variety of nanocrystalline semiconductors sition by a multistage process, have been
with varying degrees of control over the used. Electrodeposition offers a wide range
size, shape, and a narrow size distribution. of control of the nanocrystalline film owing
174 3 Semiconductor Nanostructure
to the many parameters involved (potential Step I – Electrodeposition of metal nano-

or current density, pulsed electrodepo- particles.
sition, solution composition, convection, In the E/C synthesis, the first step in
and temperature). which metal nanocrystals are deposited on
Although this chapter is limited to elec- the substrate is critical. The semiconduc-
trodeposition of semiconductors, it is only tor particles grow from the metal particles
fair to mention, even if briefly, some on a particle-by-particle basis (i.e. each
examples of electrodeposition of metal metal particle is chemically transformed
nanostructures. This is important because to the corresponding semiconductor). The
the principles and techniques used in size and size distribution, therefore, in the
electrodepositing metals are essentially first step determine the final size of the
the same as those used for deposit- semiconductor particles. For that reason,
ing semiconductors – the main difference it is important to achieve an understand-
is that almost all studies on electrode- ing of the growth mechanism of the metal
position of nanocrystalline semiconduc- nanocrystal deposition. Penner and as-
tors involve compound semiconductors, sociates studied the electrodeposition of
with the added complications this entails. various metal nanoparticles (Ag, Pt, Zn,
Examples include pulsed electrodeposi- Cu, Cd) mainly onto basal plane–oriented
tion of metal multilayers [1, 2], porous graphite and also onto Si electrodes [6–11].
membrane-templated electrodeposition of The depositions were carried out from di-
gold nanotubes [3], and Ni nanowires [4]. lute aqueous solutions of metal ions using
a potentiostatic pulse regime. A short (typ-
3.1.2 ically tens of ms) potential pulse was ap-
Electrochemical/Chemical Hybrid plied followed by open-circuit conditions.
Deposition It was demonstrated that the deposition
of metal nanocrystals occurs via a Volmer-
A method for the synthesis of semi- Weber (three-dimensional growth) mode.
conductor QDs, called the Electrochemi- For example, in the deposition of platinum
cal/Chemical method (E/C), has been de- on graphite, platinum particles with a
veloped by Penner and his associates [5]. mean diameter of 5.2 nm were formed fol-
The E/C synthesis of semiconductors lowing the deposition of quantities equal
nanocrystals involves three steps: to only 0.039 monolayers of platinum [10].
Volmer-Weber growth is favored at sur-
1. In the first step, metal nanocrystals M0 faces that are characterized by low inter-
(e.g. Cd, Cu, Zn) are electrochemically facial free energy (like the graphite basal
deposited onto basal plane–oriented plane). Transmission electron microscopy
graphite electrode surfaces from aque- (TEM) and selected area electron diffrac-
ous solution of metal ions Mn+ . tion (ED) of the graphite surfaces revealed
2. Next, the metal nanocrystals are oxi- that the deposited Pt particles are not epi-
dized, either chemically or electrochem- taxially oriented with the graphite surfaces.
ically, to the metal oxide or hydroxide. On the basis of electron microscopy and
3. Finally, the oxide (hydroxide) is dis- electrochemical studies, a two-step mech-
placed with gaseous hydrogen chalco- anism for the growth of Pt nanoparticles
genide or chalcogenide anion (e.g. S−2 ) was proposed: ‘‘instantaneous’’ nucleation
in solution to form nanocrystals of MX. that takes part within a few ms, followed by
3.1 Preparation of Nanocrystalline Semiconductor Materials 175
diffusion-controlled growth from solution. Steps II and III – The conversion of

A similar mechanism was observed in the metal nanoparticles to semiconductor
deposition of other metal nanoparticles. nanoparticles.
For short pulse durations (approximately The conversion procedure of metal
10 ms), particles of a few nm with a nanoparticles to the corresponding semi-
narrow size distribution were obtained. conductor particles involves one or two
The particle size and size distribution both steps. If the desired semiconductor is
increased for increasingly greater pulse an oxide, such as ZnO [8], the conver-
durations. In general for pulse plating, sion procedure involves one step, namely,
a high overpotential (or high current the oxidation of the metal nanoparticles.
density) and short pulse duration favor Other anions require an additional step
reduced grain size caused by increased in which the oxide (or hydroxide) is dis-
nucleation density and less time for placed by other ions such as in the case
growth. of CdS [14–16] and CuI [6]. (Although the
This increasing size distribution with formation of an oxide is not essential to
growth time contrasted with three-dimen- obtain the final semiconductor, oxide for-
sional nucleation and growth of colloidal mation will normally occur unless steps
particles, where instantaneous nucleation are taken to prevent it, and it is more
logical to deliberately oxidize the metal
followed by diffusion-limited growth leads
electrochemically.)
to a narrow size distribution. This differ-
For CdS, the Cd nanoparticles are
ence was explained by a nonhomogeneous
spontaneously oxidized in the plating
distribution of initially nucleated particles
solution (at pH = 6 and open circuit) to
on the substrate [12, 13]. Where many
Cd(OH)2 . The displacement of OH by S
nuclei were formed close together, their
to yield CdS can be carried out in two
diffusion layers interacted, resulting in
different ways: immersion of the Cd(OH)2
a reduced rate of diffusion and slower
in an aqueous sulfide solution (Na2 S
growth. On the other hand, particles that at pH = 10) [15], or by exposure of the
nucleated relatively far from other nuclei Cd(OH)2 to gas phase H2 S at 300 ◦ C [16].
retained independent and thinner diffu- In the synthesis from aqueous S2− , CdS
sion layers and therefore grew faster. In particles with a mean crystal diameter
other words, nanoparticles that nucleate between 2 and 8 nm were prepared. The
close together end up being smaller than nanocrystals were oriented with the c-axis
particles that nucleate far from neighbors. of the CdS-wurtzite unit perpendicular to
The size distribution could be narrowed the electrode surface (i.e. graphite basal
for greater amounts of deposit by em- plane). In the gas-phase conversion, CdS/S
ploying a train of short deposition pulses core-shell structures were obtained by
(≤10 ms) followed by a much longer decomposition of the H2 S to S on the
off-time (at open circuit), rather than a CdS surface.
single longer deposition pulse [14]. These β-CuI was formed by first converting
shorter deposition pulses diffusionally de- the Cu nanocrystals to Cu2 O followed by
couple the growth of densely nucleated treatment in aqueous KI [6].
regions from sparsely nucleated regions, The various nanocrystalline semicon-
and by this means, they narrow the size ductors exhibited strong, size-dependent
distribution of the nanocrystallites. room temperature luminescence, although
they were deposited directly on graphite [6, ‘‘wires’’, was used to improve thin film
8, 14]. The graphite, in contrast to metals, CdS/CuInS2 photovoltaic cells [20].
apparently does not efficiently quench the The nature of the anion of the metal
luminescence. Of particular importance is salt is important in this technique. Salts of
the observation that, under suitable prepa- many anions are either insoluble or unsta-
ration conditions, all the semiconductors ble in hot dimethyl sulfoxide (DMSO) con-
emitted essentially band gap luminescence taining dissolved chalcogen. Perchlorate
at room temperature with little or no and chloride are the most commonly used
deep sub–band gap band response that anions although other halides, methylsul-
is normally characteristic of these materi- fonate, and borofluoride have also been
als. This implies that the nanocrystals were used. However, another reason that the
of very high quality. anion is important, particularly relevant
to the present discussion, is that it af-
3.1.3 fects the crystal size and therefore the
Electrodeposition of Nanocrystals from band gap. Films deposited from Cl− so-
Nonaqueous Solution lutions exhibit band gaps between 0.1 and
0.2 eV higher than those of the same com-
One of the simplest techniques used to pound deposited from ClO4 − [21]. More
electrodeposit semiconductors is cathodic recent and detailed studies indicate that
deposition from nonaqueous solutions that this difference is due to adsorption
containing elemental chalcogen (S, Se) and of the more strongly adsorbed Cl− on the
a metal salt, first described by Baranski and growing CdX surface, preventing further
Fawcett [17]. Two main mechanisms have crystal growth [22].
been considered: deposition of metal (e.g. Although both CdSe and CdS can be
Cd) followed by chemical reaction with deposited in nanocrystalline form by this
elemental chalcogen in solution and re- nonaqueous deposition, the essentially
duction of chalcogen to (poly)chalcogenide total insolubility of Te in DMSO prevents
followed by ionic reaction between chalco- the use of the method for deposition of
genide and metal cations. Which mech- nanocrystalline tellurides (a small amount
anism will dominate depends on the of Te can be codissolved with Se and mixed
specific system (substrate, semiconductor, selenide-tellurides with small amounts of
deposition conditions) and may change in Te can be deposited; see following section).
the same deposition. However, a related method to deposit
This technique was found to give CdTe has been described by Cocivera and
nanocrystalline semiconductors which ex- associates [23, 24]. They reacted elemental
hibited quantum size effects with typical Te with tri-n-butyl phosphine (TBP), which
lateral crystal size of 5 nm [18, 19]. The reacts with Te to form TBP telluride.
height of the crystals [measured by X-ray This compound, together with a Cd salt
diffraction (XRD)] was often up to several dissolved in propylene carbonate, allowed
times larger than the lateral dimensions cathodic electrodeposition of CdTe. The
(measured by TEM). In fact, the increased as-deposited films were reported to be X-
transparency to shorter wavelengths be- ray amorphous, a fact that suggested that
cause of size quantization, together with they might in fact be nanocrystalline [26].
good (photo)conductivity expected from (Cd,Hg)Te films grown by the same
technique exhibited a crystal size of about and {110}Au ||{11.0}Cdse orientation rela-
5 nm [25]. tionship. The epitaxy arises because of
This technique was modified and sim- the good lattice match between the CdSe
plified, in particular, by using a one- and Au in a 2 : 3 relationship (−0.6% mis-
step technique to prepare the solution in match). Beyond the first layer of crystals,
DMSO [26]. The CdTe was indeed found to the epitaxy is gradually lost and the crystal
be nanocrystalline with a wide size distri- size grows for the perchlorate bath; for a
bution varying from several nm up to tens CdCl2 bath, the crystal size does not grow
of nm. In addition, the films were gen- much because of capping, as discussed ear-
erally nonstoichiometric with excess Te or lier. The mismatch strain, which gradually
Cd, depending on the deposition potential. increases as the crystal grows, eventually
It was difficult to deposit close to stoichio- leads to termination of growth. This can ex-
metric film. To improve the stoichiometry, plain the relatively narrow size distribution
reverse pulse deposition was used to strip obtained.
excess Cd or Te during the anodic pulse. This hypothesis of strain-determined
The pulse regime also decreased the crystal growth termination suggested that crys-
size to an extent depending on the pulse tal size should be controllable by choice
parameters (crystal size typically several of the semiconductor and substrate lat-
nm) and improved the size distribution. tice parameters: the larger the mismatch,
The main factors limiting crystal growth the smaller the crystal size, and vice
were short pulse on-times and capping of versa. A range of different electrode-
the crystals with strongly adsorbing phos- posited semiconductor-substrate combi-
phine, mainly during the pulse off-time. nations has been investigated to test this.
The lattice parameter of the CdSe was
3.1.4 varied with (assumed) little change in
Size Control Using Semiconductor- chemical interaction with the substrate
substrate Lattice Mismatch by depositing an alloy of CdSex Te1−x [29].
From Vegard’s law, a value of x = 0.88
An important factor that can influence (12% Te) should result in an increased
crystal size, particularly for the first layer lattice with a perfect match to Au. Although
of crystals, is the nature of the substrate. Te could not be dissolved in DMSO,
Deposition of CdSe from the above DMSO small amounts of Te could be dissolved
electrolyte using Cd(ClO4 )2 onto films of in the presence of dissolved Se. The
evaporated Au on glass or mica resulted concentration of Te in the electrolyte
in crystals of about 4–5 nm in size. was not known (it was almost certainly
The distribution of the nanocrystals on <1 mM), but the amount of Te in
the Au depended on current density and the deposit was quantitatively varied by
deposition temperature: high currents and varying the deposition current density
low temperatures favored isolated crystals, and temperature. Relatively high current
whereas increasing aggregation occurred densities resulted in severe depletion of the
with higher temperature and decreased Te near the cathode, thereby favoring the
current [27]. more concentrated Se in the deposit. In the
ED [27] and HTEM [28] showed that same way, higher deposition temperatures
the CdSe crystals were epitaxially de- increased the concentration of Te at the
posited on the Au in a {111}Au ||{00.2}Cdse cathode and therefore in the deposit.
The maximum Te concentration in the forms fairly large particles (10–20 nm),
deposits was measured to be only about which become larger with increasing de-
5%. However, this was sufficient to see position time. Even more important, the
a major increase (by up to several times) CdSe deposit on Pd is both XRD and ED
in nanocrystal size with increase in Te amorphous. However, careful examina-
concentration, whereas the epitaxy was, as tion of a HTEM image reveals the presence
expected, retained. of many regions showing short-range or-
The lattice parameter was decreased der, often with hexagonal symmetry. An
(with expectation of reduced crystal size) by in-depth analysis of this short-range order
alloying the Cd with Zn [30]. Cdx Zn1−x Se showed the presence of irregular struc-
was epitaxially deposited using very low tures of ordered CdSe, on the order of 1 nm
concentrations of Cd in the deposition in size, surrounded by disordered CdSe.
solution. The crystal size for x = 0.78 was Furthermore, Fourier analysis of HTEM
between 2 and 3 nm – half the size of the images showed some degree of preferen-
corresponding CdSe deposits. The crystal tial orientation of the ordered structures
height, measured by XRD, was 5 nm. relative to the Pd lattice, but rotated 30◦
The crystals are therefore short quantum to the Pd [32]. Subsequent modeling of
columns rather than QDs. the superimposed CdSe-Pd lattices showed
CdS has a 4.5% mismatch with Au (2 : 3 that this 30◦ rotation indeed resulted in
ratio). If epitaxy was to occur for such a an improved lattice match compared with
large mismatch, the crystal size should the aligned lattices [33]. The very small
be very small. In this case, the proper- ordered structures are in qualitative agree-
ties of the deposit were dependent on the ment with the mismatch strain argument
amount deposited [31]. For small amounts that predicts very small crystals. However,
of deposit, the crystals are predominantly it is likely that differences in chemical re-
epitaxial (although not perfectly so as for activity between CdSe and Au or Pd also
CdSe on Au). The average crystal size play a role.
grows, and the degree of orientation de- CdS has only a small mismatch with
creases as the amount of deposit increases. Pd – +0.17% (2 : 3 ratio). However, at-
The crystal height for a nominally 5-nm tempts to obtain thin electrodeposits of
thick (in reality, ≤3 nm) deposit was found CdS on Pd were not successful. Small par-
to be 3 nm – similar to the lateral ones. ticles (approximately 2 nm) were obtained;
The mismatch may also be varied by however, the identity of these particles was
changing the material of the substrate. Pd not known – XPS analysis showed the al-
has the same fcc cubic structure as Au; most total absence of Cd. In this case,
the lattice spacing, however, is almost 5% chemical interaction between S and Pd
smaller than that of Au, with the result probably dominates the surface.
that the mismatch between CdSe and Pd In an attempt to separate chemical
is +4.1%. This is close to the mismatch effects from lattice mismatch, we studied
between CdS and Au but in the opposite an alloy of 3% Cd with Au. Cd was chosen
direction, that is, the CdSe should be com- because of its strong interaction with Se.
pressed instead of stretched. In contrast The change in lattice-spacing owing to
to CdSe on Au, which grows as isolated the 3% of Cd is very small (+0.07%) and
crystals on Pd, the CdSe covers the en- therefore any change in CdSe properties
tire surface and, with further deposition, can be attributed to differences in chemical
interaction. The lateral crystal size is current or potential [35]. The deposition is
approximately 5 nm – similar to that on based on oxidation of Tl(I) to Tl(III) and
pure Au. However, the crystal height is Pb(II) to Pb(IV); the higher oxidation states
considerably smaller than this, as seen of these ions are readily hydrolyzed to the
qualitatively by TEM contrast and surface oxides. Because Tl2 O3 is deposited prefer-
coverage. Most surprising, however, is entially (owing to a less anodic potential)
that the deposit is not only a Wurtzite compared to PbO2 , low current densities
structure but also a mixture of Wurtzite, (or low anodic potentials) form essentially
Zincblende, and Rocksalt. Rocksalt CdSe only Tl2 O3 from mixed Tl/Pb solutions.
is normally stable only at high pressures. It If a solution is made with a higher Pb
was suggested that surface tension forces, than Tl concentration, then as the cur-
together with (or maybe because of) the rent density (anodic potential) is increased
strong interaction between the CdSe and the concentration of Pb in the resulting
the substrate, stabilized this high-pressure oxide increases as the low concentration
structure [34]. Tl becomes diffusion-limited. The inter-
play of relative cation concentrations, mass
3.1.5 transport conditions, and pulse character-
Electrodeposition of Layered istics (current/potential and time scales)
Nanostructures
imparts control of the stoichiometry, as
well as thickness of the individual layers.
Many modern semiconductor devices
The sharpness of the interface between
comprise alternating layers of different
layers can also be controlled: deposition
materials forming superlattices and mul-
where one of the constituent ions is in
tiple quantum wells. One well-known
the diffusion-controlled regime results in a
example of such structures is the diode
graded composition, whereas if deposition
laser, a mass-produced device. This device
depends on confinement of charges in the takes place in the kinetic control regime
two-dimensional structures for enhanced for both metal ions, a sharp compositional
laser output at lowered current thresholds. difference between layers is obtained [36].
Such alternating semiconductor layers are STM was used to identify the individ-
usually manufactured either by chemical ual layers [37, 38]. Both Tl2 O3 and PbO2
vapor deposition or by molecular beam are degenerate semiconductors with band
epitaxy. The thickness of the layers can be gaps of about 1.4 and 1.8 eV, respectively.
closely controlled in both techniques. As Therefore, such layers could be used as
mentioned earlier, electrodeposition also quantum wells with electrons confined
allows good control of thickness. preferentially in the lower band gap Tl-rich
The group of Switzer has been active layers.
in electrodeposition of alternating layers Layers of Tl2 O3 or Pb-Tl-O could be de-
of different metal oxides or metal/metal posited on steel with strong (100) or (210)
oxides. Ceramic superlattices of Pba -Tlb - texturing, respectively, and superlattices
Oc /Pbd -Tle -Of with individual thickness of Pb-Tl-O could then be electrodeposited
of 5–10 nm were electrodeposited onto epitaxially on these prelayers with these
stainless steel from an aqueous solution same orientations [39].
of TlNO3 and Pb(NO3 )2 in 5 M (NaOH) Tl2 O3 can be electrodeposited with a
at room temperature using either pulsed defect chemistry controllable through the
deposition overpotential. Oxygen vacan- At low current densities (≤1 mA cm−2 ),

cies are produced at high values of no potential oscillations occur, and pure,
overpotential, whereas Tl interstitials are strongly oriented (100) Cu2 O is formed
favored at low overpotential. This was with a crystal size > 1 µm. With increasing
used to deposit alternating ‘‘defect chem- current density, the Cu content increases
istry superlattices’’ of several nm period along with the appearance of the oscilla-
by pulsing the potential during depositions and the color changes from the red
tion [40]. Electrical and optical studies of of the pure Cu2 O through green to gold.
Tl2 O3 deposited at different overpotentials These layers are nanocomposites that com-
were carried out [41]. The resistivity of the prise Cu2 O and metallic Cu nanocrystals.
degenerate oxide could be varied from sev- The nanocrystal size decreases as the cur-
eral hundred µ"-cm at low overpotential rent density is increased and varies from 5
to 75 µ"-cm over a range of <0.1 V. This to 10 nm for the Cu and from 4 to 20 nm
change in resistivity was due to changes in for the Cu2 O in the composite. The poten-
the electron mobility (because the carrier tial oscillations suggest that the materials
density was constant at about 1021 cm−3 )
are modulated and this was observed in
caused by an increase in the concentra-
SEM micrographs [43].
tion of charged interstitials. The optical
Cu2 O is a p-type semiconductor with a
band gaps of the high overpotential films
band gap of about 2.1 eV, hence its red
were blue-shifted by as much as 1.1 eV
color. Absorption spectra of the composite
from the normal value of 1.4 eV because
films show a blueshift of the Cu2 O
of conduction band-filling (Burstein-Moss
absorption edge by up to 0.4 eV (for 8 nm
effect).
Cu2 O crystallites). On the basis of strong
Another system studied by the same
confinement, this shift is much larger than
group comprises alternating Cu and Cu2 O
layers formed by constant current electrol- expected, and it may be because of the
ysis of aqueous Cu(II) + lactate solutions confinement in the nanocrystals beyond
at a pH of 8–10 [42, 43]. Although a that due to simple physical size.
constant current is applied, spontaneous These structures are highly anisotropic
potential oscillations occur, leading to vari- with respect to their electrical conductivity.
ations in the stoichiometry of the deposit There is a difference of about 12 orders of
from Cu to Cu2 O. The cause of these os- magnitude in the conductivities of metallic
cillations is not clear but appears to be Cu and Cu2 O. This difference is mani-
related to formation and breakdown of fested in both the lateral (parallel to the
surface compounds. The oscillation period layers) and perpendicular conductivities of
(typically tens of seconds to minutes), and the alternating layers. The lateral resis-
therefore the thickness of the different lay- tivity varied from 10−4 "-cm for very Cu
ers and their compositions, depends on the rich samples to >106 "-cm for Cu-poor
current density and, in particular, on the ones. In particular, at a Cu concentration
solution pH. Quartz crystal monitor stud- >9.8 vol%, the resistivity decreased sud-
ies suggested that Cu2 O is formed during denly, typical of a percolation process. The
positive potential spikes in the oscillations, resistivity in the perpendicular direction
whereas a composite of Cu and Cu2 O is de- was much larger than that in the lateral
posited during the more negative plateau one for reasonably Cu-rich samples (ratios
region of the oscillation [44]. of >109 ). No less important is the fact that
although the transport in the lateral di- forming a rectifying contact. AC deposi-
rection was always ohmic, the transport in tion then resulted in formation of CdS
the vertical direction exhibited pronounced only in the pores and not in cracks or de-
negative differential resistance, attributed fects in the film. The CdS wire dimensions
to resonant tunneling from Cu into hole closely followed those of the pores. For the
states in Cu2 O [45, 46]. smaller pores, the wires were apparently
made up of single crystals joined in the ax-
3.1.6 ial direction, whereas for the larger pores,
Template-directed Electrodeposition the coherence length, measured by XRD,
was less than the pore diameter, suggest-
There are a number of studies on electrode- ing that the wires were composed of several
position of semiconductors in the pores of crystals also in the radial direction (or pos-
various membranes forming nanowires of sibly, defected single crystals). It should be
the semiconductor. Klein and associates noted that even for films of CdS deposited
electrodeposited CdSe and CdTe into the by this technique the crystal size is very
pores of Anopore membranes [47]. These small (see preceding section). The CdS
membranes are 50-µm thick with closely was hexagonal (wurtzite) with its c-axis
spaced 200-nm-sized pores. One side of aligned predominantly along the mem-
the membrane was sputter-coated with Au brane thickness. HTEM studies of these
followed by Ni electrodeposition, which nanowires confirmed the well-ordered and
partially filled the pores, as a substrate. basal plane-textured CdS crystallinity [50].
CdSe and CdTe were electrodeposited Small quantum size effects were seen for
from acidic aqueous solutions containing the smallest wires, with band gaps varying
CdSO4 and SeO2 or TeO2 . The semicon- from 2.36 to 2.42 eV [51].
ductors grow as compact wires that are The same technique to form CdS
many µm long and 200–300 nm in di- nanowires, only employing dc electrode-
ameter. By electrodepositing first CdSe position, has been used by Xu and asso-
and then CdTe, nanowires were formed, ciates [52]. In this case, the Al was etched
which were compositionally graded over away and a silver film evaporated on the
their length. membrane. The wires in this case were
Chakarvarti and Vetter electrodeposited textured with the (10.1) plane parallel
Se in the pores of nuclear track mem- to the membrane surface [in contrast to
branes [48]. These pores were relatively the ac deposition where the (00.1) plane
large (2.5 µm). grew parallel to the surface]. This tech-
Using porous anodic aluminium ox- nique was extended to CdSe and CdTe
ide films, Routkevitchand associates elec- nanowires [53].
trodeposited very thin CdS nanowires [49]. This group also deposited CdS nano-
The alumina membranes were typically wires in porous alumina templates using
1–3-µm thick with pore sizes ranging from an aqueous solution containing CdCl2 and
9 to 35 nm. The CdS was electrodeposited, thioacetamide at pH = 4.6 [54]. This de-
using ac electrodeposition, from a DMSO position is actually an electrochemically
solution containing CdCl2 and elemental induced chemical solution deposition ini-
S. The porous alumina, as anodized, was tiated by local pH changes caused by
separated from the Al by a dense oxide electrolysis [55]. In this case, the wires
grew with the (00.1) plane parallel to the photoelectrochemical response) of TiO2
membrane surface. was dependent on the electrodeposited
metal matrix: deposition of Ni, Cu, Ag, and
3.1.7 In matrices resulted in different disper-
Occlusion Electrodeposition of Composites sion geometries of the TiO2 particles [62].
In the case of Cu, the photoelectrochem-
Apart from the layered Cu/Cu2 O com- ical response at a high pH showed both
posites described earlier, composites of n-type and p-type polarities because of the
nanocrystalline semiconductors with non- formation of p-Cu2 O. The incorporation
semiconductors (metals or polymers) have of alumina into electrodeposited Ni films
been electrodeposited by incorporation was shown to be controlled by diffusion
of the semiconductor phase from so- of the alumina particles to the Ni surface
lution into the electrodepositing metal and their residence time at the surface [63].
(also known as occlusion). WO3 particles CdS, as well as TiO2 particles occluded into
suspended in solution can be incorpo- electrodeposited Ni, were shown to exhibit
rated in electrodeposited polypyrrole [56] increased photoelectrochemical activity af-
or polyaniline [57] films, and the result- ter potential cycling owing to passivation
ing films exhibit electrochromism, both of of the Ni matrix [64].
the polymer and of the WO3 . Although this
technique worked only for oxides with very
3.1.8
low isoelectric points, the isoelectric point Sonoelectrochemical Formation of
of oxides with relatively high points of zero Nanocrystalline Semiconductors
charge could be increased by adsorption of
anions, such as sulfate or iodide, from
Sonoelectrochemical synthesis has re-
solution. In that manner TiO2 was incor-
cently been used for the preparation of
porated into electrodeposited polypyrrole
films, and the resulting films showed a semiconductor nanocrystalline powders.
photoanodic photocurrent response be- In the sonoelectrochemical method, the
cause of the incorporated TiO2 [58]. Very ultrasound horn acts as both cathode and
high concentrations of WO3 – up to 53 ultrasound emitter. This technique was
wt% – were incorporated into electrode- used for preparing metal powders [65]
posited polypyrrole [59]. High stirring and was extended to CdTe, although de-
rates and low current densities of plating tails of the CdTe particle size were not
favored high rates of incorporation. In- given [66]. CdSe nanocrystalline powders
corporation of TiO2 into polypyrrole was have been prepared by pulsed sonoelec-
attributed to mechanical entrapment of the trochemical reduction from an aqueous
TiO2 particles in the rough, soft polypyr- selenosulfate solution. The crystal size
role. In addition, the electric field at the could be varied from X-ray amorphous up
surface of the growing film was also as- to 9 nm (sphalerite phase) by controlling
cribed a role [60]. the various electrodeposition and sonic pa-
In a similar manner, TiO2 particles in rameters [67]. Crystal size was smaller for
suspension were incorporated into grow- lower preparation temperatures, higher ul-
ing electrodeposited Ni films and the trasound intensity, and shorter current
resulting films were photoelectrochem- pulse width. These dependencies could
ically active [61]. The morphology (and be explained based on a pulse of electric
current producing a high density of fine 6. G. S. Hsiao, M. G. Anderson, S. Gorer et al.,

particles on the sonic tip, followed by a J. Am. Chem. Soc. 1997, 119, 1439.
7. S. Gorer, G. S. Hsiao, M. G. Anderson et al.,
burst of ultrasonic energy, which removes
Electrochim. Acta 1998, 43, 2799.
the particles from the cathode into the so- 8. R. M. Nyffenegger, B. Craft, M. Shaaban
lution and prevents them from growing. et al., Chem. Mat. 1998, 10, 1120.
Similarly, PbSe nanocrystalline powders 9. J. V. Zoval, R. M. Stiger, P. R. Biernacki et al.,
with a crystal size of 10 to 16 nm were J. Phys. Chem. 1996, 100, 837.
10. J. V. Zoval, J. Lee, S. Gorer et al., J. Phys.
prepared in the same manner from a sim-
Chem. B 1998, 102, 1166.
ilar solution using salts of Pb instead of 11. R. M. Stiger, S. Gorer, B. Craft et al., Lang-
Cd [68]. muir 1999, 15, 790.
MoS2 can be electrodeposited from 12. J. L. Fransaer, R. M. Penner, J. Phys. Chem.
a thiomolybdate solution [69, 70]. How- B 1999, 103, 7643.
13. H. Liu, R. M. Penner, J. Phys. Chem. B 2000,
ever, the deposit is apparently amorphous
104, 9131.
and requires annealing to crystallize. We 14. S. Gorer, R. M. Penner, J. Phys. Chem. B
reasoned that the high temperatures in 1999, 103, 5750.
and around ultrasonic cavitation bub- 15. M. A. Anderson, S. Gorer, R. M. Penner, J.
bles might crystallize the electrodeposited Phys. Chem. B 1997, 101, 5895.
16. S. Gorer, J. A. Ganske, J. C. Hemminger
MoS2 . Sonoelectrochemical formation of
et al., J. Am. Chem. Soc. 1998, 120, 9584.
MoS2 powders using the above solution 17. A. S. Baranski, W. R. Fawcett, J. Electrochem.
indeed resulted in partial crystallization of Soc. 1980, 127, 766.
the as-deposited MoS2 , but in the form of 18. G. Hodes, A. Albu-Yaron, Optical, Struc-
multiple closed shell fullerene-like structural and Photoelectrochemical Properties of
Quantum Box Semiconductor Films, Proc.
tures [71]. The structures were polyhedra
Electrochem. Soc. 1988, 88–14, Vol. 298.
of typically several tens of nm in size 19. G. Hodes, Isr. J. Chem. 1993, 33, 95.
comprising layers (typically about 10) of 20. D. Gal, G. Hodes, D. Hariskos et al., Appl.
crystallized MoS2 surrounding what was Phys. Lett. 1998, 73, 3135.
probably amorphous MoS2 . In some cases, 21. G. Hodes, T. Engelhard, A. Albu-Yaron et al.,
Mat. Res. Soc. Symp. 1990, 164, 81.
nanotubes of MoS2 were also obtained. Al-
22. Y. Mastai, D. Gal, G. Hodes, J. Electrochem.
though the cause of formation of these Soc. 2000, 147, 1435.
structures is not known, it is likely to be 23. A. Darkowski, M. Cocivera, J. Electrochem.
related to the buildup and collapse of the Soc. 1985, 132, 2768.
cavitation bubbles. 24. J. von Windheim, M. Cocivera, J. Elec-
trochem. Soc. 1987, 134, 440.
25. C. L. Colyer, M. Cocivera, J. Electrochem. Soc.
References 1992, 139, 406.
26. Y. Mastai, G. Hodes, J. Phys. Chem. B 1997,
1. J. Yahalom, Surf. Coat. Technol. 1998, 105, 101, 2685.
VII–VIII. 27. Y. Golan, L. Margulis, I. Rubinstein et al.,
2. T. Cohen, J. Yahalom, W. D. Kaplan, Rev. Langmuir 1992, 8, 749.
Anal. Chem. 1999, 18, 279. 28. Y. Golan, L. Margulis, G. Hodes et al., Surf.
3. C. R. Martin, D. T. Mitchell in Electroanalyti- Sci. 1994, 311, L633.
cal Chemistry (Eds.: A. J. Bard and I. Rubin- 29. Y. Golan, J. L. Hutchison, I. Rubinstein et al.,
stein), Marcel Dekker, Inc., New York, Basel Adv. Mater. 1996, 8, 631.
1999, p. 1, Vol. 21. 30. G. Hodes, Y. Golan, D. Behar et al., in
4. L. Sun, P. C. Searson, C. L. Chien, Appl. Nanoparticles and Nanostructured Films,
Phys. Lett. 1999, 74, 2803. (Ed.: J. H. Fendler), Wiley-VCH, Germany
5. R. M. Penner, Acc. Chem. Res. 2000, 33, 78. 1998, 1.
31. D. Behar, I. Rubinstein, G. Hodes et al., Su- 52. D. S. Xu, Y. J. Xu, D. P. Chen et al., Chem.
perlattices Microstruct. 1999, 25, 601. Phys. Lett. 2000, 325, 340.
32. Y. Golan, E. Ter-Ovanesyan, Y. Manassen 53. D. S. Xu, D. P. Chen, Y. J. Xu et al., Pure
et al., Surf. Sci. 1996, 350, 277. Appl. Chem. 2000, 72, 127.
33. Y. Golan, G. Hodes, I. Rubinstein, J. Phys. 54. D. S. Xu, Y. J. Xu, D. P. Chen et al., Adv.
Chem. 1996, 100, 2220. Mater. 2000, 12, 520.
34. Y. Zhang, G. Hodes, I. Rubinstein et al., Adv. 55. K. Yamaguchi, T. Yoshida, T. Sugiura et al.,
Mater. 1999, 11, 1437. J. Phys. Chem. B 1998, 102, 9677.
35. J. A. Switzer, M. J. Shane, R. J. Phillips, Sci- 56. H. Yoneyama, Y. Shoji, J. Electrochem. Soc.
ence 1990, 247, 444. 1990, 137, 3826.
36. J. A. Switzer, R. J. Phillips, T. D. Golden, 57. H. Yoneyama, S. Hirao, S. Kuwabata, J. Elec-
Appl. Phys. Lett. 1995, 66, 819. trochem. Soc. 1992, 139, 3141.
37. J. A. Switzer, R. P. Raffaelle, R. J. Phillips 58. K. Kawai, N. Mihara, S. Kuwabata et al., J.
et al., Science 1992, 258, 1918. Electrochem. Soc. 1990, 137, 1793.
38. T. D. Golden, R. P. Raffaelle, J. A. Switzer, 59. F. Beck, M. Dahlhaus, J. Appl. Electrochem.
Appl. Phys. Lett. 1993, 63, 1501. 1993, 23, 781.
39. R. J. Phillips, T. D. Golden, M. G. Shumsky 60. F. Beck, M. Dahlhaus, N. Zahedi, Electro-
et al., Chem. Mat. 1997, 9, 1670. chim. Acta 1992, 37, 1265.
40. J. A. Switzer, C. J. Hung, B. E. Breyfogle 61. M. Zhou, N. R. deTacconi, K. Rajeshwar, J.
et al., Science 1994, 264, 1573. Electroanal. Chem. 1997, 421, 111.
41. R. A. Vanleeuwen, C. J. Hung, D. R. Kamm- 62. N. R. de Tacconi, C. A. Boyles, K. Rajeshwar,
ler et al., J. Phys. Chem. 1995, 99, 15 247. Langmuir 2000, 16, 5665.
42. J. A. Switzer, C. J. Hung, E. W. Bohannan 63. P. M. Vereecken, I. Shao, P. C. Searson, J.
et al., Adv. Mat. 1997, 9, 334. Electrochem. Soc. 2000, 147, 2572.
43. J. A. Switzer, C. J. Hung, L. Y. Huang et al., 64. N. R. de Tacconi, H. Wenren, K. Raje-
J. Mat. Res. 1998, 13, 909. shwar, J. Electrochem. Soc. 1997, 144,
44. E. W. Bohannan, L. Y. Huang, F. S. Miller 3159.
et al., Langmuir 1999, 15, 813. 65. J. Reisse, H. Francois, J. Vandercam-
45. J. A. Switzer, C. J. Hung, L. Y. Huang et al., men et al., Electrochim. Acta 1994, 39,
J. Am. Chem. Soc. 1998, 120, 3530. 37.
46. J. A. Switzer, B. M. Maune, E. R. Raub et al., 66. J. Reisse, T. Caulier, C. Deckerheer et al.,
J. Phys. Chem. B 1999, 103, 395. Ultrason. Sonochem. 1996, 3, S147.
47. J. D. Klein, R. D. Herrick, D. Palmer et al., 67. Y. Mastai, R. Polsky, Y. Koltypin et al., J. Am.
Chem. Mat. 1993, 5, 902. Chem. Soc. 1999, 121, 10 047.
48. S. K. Chakarvarti, J. Vetter, J. Micromech. 68. J. J. Zhu, S. T. Aruna, Y. Koltypin et al.,
Microeng. 1993, 3, 57. Chem. Mat. 2000, 12, 143.
49. D. Routkevitch, T. Bigioni, M. Moskovits 69. H. J. Byker, A. E. Austin, J. Electrochem. Soc.
et al., J. Phys. Chem. 1996, 100, 14 037. 1981, 128, 381 (CA no. 536).
50. J. L. Hutchison, D. Routkevitch, A. Albu- 70. E. A. Ponomarev, M. NeumannSpallart,
Yaron et al., Microscopy of Semiconducting G. Hodes et al., Thin Solid Films 1996, 280,
Materials 1997, 1997, 389. 86.
51. D. Routkevitch, T. L. Haslett, L. Ryan et al., 71. Y. Mastai, M. Homyonfer, A. Gedanken
Chem. Phys. 1996, 210, 343. et al., Adv. Mat. 1999, 11, 1010.
3.2 Macroporous Microstructures Including Silicon 185
3.2 Si structures that can form in hydrofluo-

Macroporous Microstructures Including ric acid (HF) – based solutions [5, 7–11],
Silicon models evoked for explaining them [5, 12,
3], elucidation of the structure of lumines-
C. Lévy-Clément
cent material and the origin of its efficient
LCMTR, Thiais, France
luminescence [11, 16], and numerous tech-
3.2.1 nological applications of porous Si [17,
Introduction 18]. Tremendous quantity of results (sev-
eral thousand publications) have been
Porous materials are generally obtained by analyzed all along the 1990s in various
localized electrochemical corrosion under conference proceedings [19–26] and re-
anodic polarization (anodization). Some view articles [5, 8, 9, 11, 14–18, 27–30],
metals such as aluminum or stainless as well as edited books [32–36]. Following
steel can be made porous. Comparatively, the work on luminescent porous Si, a num-
a far larger number of semiconductors ber of studies have been undertaken to
can be made porous, leading to important render other semiconductors porous. This
changes in the physical properties. The chapter will focus on the electrochemical
best and first known example is that of aspects of porous semiconductors includ-
porous silicon (Si), which exhibits strong ing porous Si.
photoluminescence in the visible range in-
stead of the weak luminescence of bulk Si. 3.2.2
Since its discovery more than forty years Basic Principles
ago [1–3] and recognition of its porous
nature ten years later [4], it has attracted 3.2.2.1 Thermodynamic Considerations
considerable research interest because of The general concept of the anodic disso-
its morphological, electrical, and optical lution process and reaction pathway was
properties, which are different from those first explored with group-IV semiconduc-
of bulk silicon [5]. Interest in porous Si tors (i.e. silicon and germanium) (1–3).
during the 1970s and 1980s was because of
the fact that it could easily be converted into Si + 6F− + (4 − n)h+ −−−→
thick oxide films and be used for dielec-
tric (trench) isolation of active Si devices SiF6 2− + ne− (1)
and in the full isolation by porous oxidized
Si (FIPOS) process, resulting in less than h+ and e− are a valence band hole
200 published papers, spanning a period and a conduction band electron, respec-
of 35 years [5]. During that period of time, tively. Quadrivalent dissolution to Si(IV)
investigation of the localized corrosion of is assumed and n is a number between
other semiconductors (groups IV, III–V, 0 and 4. This work led Gerischer to
and II–VI semiconductors) raised limited develop a general model of semiconduc-
interest, until the discovery in 1990 of the tor/electrolyte interfaces, particularly for
strong luminescence of porous silicon aris- III–V materials [37]. This model com-
ing from quantum-confinement effects [6]. bines semiconductor band theory and
Since then, research work was mainly fo- electrochemistry. It assumes isoenergetic
cussed on various aspects of porous Si, electron transfer across the semiconduc-
particularly on the rich variety of porous tor/solution interface, with the rate of
186 3 Semiconductor Nanostructures
reaction dependent on the overlap of the decomposition potential is below the con-
density of states. duction band minimum (Ec < n Ed ) i.e.
When photoelectrochemical solar cells more positive than the conduction band.
became popular in the 1970s, many reports A similar statement can be given for the
appeared concerning the stability, dissolu- holes. The semiconductor electrode is un-
tion, and flat-band potential of semicon- stable against anodic (photo)corrosion if its
ductors in solutions. These papers investi- decomposition potential is higher (more
gated parameters such as the energy level negative) than the valence band maximum
of the band edges, which is critical for the (Ev > p Ed ) of the semiconductor. In this
thermodynamic stability of the semicon- case, the hole arriving at the interface car-
ductor and how to determine the potential ries out an oxidation reaction with the
for the onset of the (photo)electrochemical semiconductor or undergoes fast recombi-
etching [38–40]. The criterion for ther- nation with the electrons. The conditions
modynamic stability of a semiconductor of corrosion in an electrolyte are illustrated
electrode in an electrolyte solution is deter- in Fig. 1.
mined by the position of the Fermi level EF Generally, the semiconductors are pro-
with respect to the decomposition potential tected against cathodic (photo)corrosion in
of the electrode with either the conduction indifferent electrolytes because of the for-
band electrons n Ed or valence band holes mation of a thin metallic film produced
p E . Under illumination, the quasi-Fermi at the first instant of (photo)corrosion by
d
level replaces the Fermi level. The Fermi the (photo)electrons, which for GaAs, for
example, can be written as
level is usually found within the band gap
of the semiconductor and its position is GaAs + 3e− −−−→ Ga0 + As3− (2)
not easily evaluated (especially the quasi-
Fermi level of minority carriers). Therefore The situation is different in the an-
it was found more practical to use the con- odic regime. Gerischer [38] and Heller [42]
duction band minimum (Ec ) and valence pointed out that the majority of semi-
band maximum (Ev ) as criteria for elec- conductors (in the dark for p-type or
trode corrosion. Thus, a semiconductor under illumination for n-type) are unsta-
will be corroded in a certain electrolyte ble against corrosion with valence band
by the conduction band electrons if its (photo)holes, as it is illustrated in Figs. 1(c)
nE n
E d E Ed
Ec Ec Ec Ec
n
nE
d
Ed
p
pE
d
Ed
Ev Ev Ev Ev
pE pE
d d
(a) (b) (c) (d)

Fig. 1 Typical correlation between energy positions of band edges
and decomposition potentials, controlling thermodynamic stability
against photodecomposition. (a) stable, (b) unstable, (c) unstable
against anodic decomposition, (d) unstable against cathodic
decomposition (from Ref. 39).
Fig. 2 Schematic drawings of the band

diagram representing decomposition of
semiconductor under anodization. Ec Ec
(a) photogenerated holes in an n-type
semiconductor, (b) bias-generated
e−
majority carriers in a p-type
semiconductor (from Ref. 41). EF
p pE
Ed d
h+ h+
hv EF
Ev Ev
h+
h+
(a) (b)
and 2. The oxidation reaction induced by which case a rough or porous morphol-
the holes in the valence band is the oxida- ogy is developed at the semiconductor
tion of the semiconductor itself, which for surface. Under high current densities, a
GaAs, for example, can be written as: solid oxide is formed at the semiconduc-
tor surface because of mass transport of
GaAs + 6h+ −−−→ Ga3+ + As3+ (3) a limited supply of reactants or prod-
ucts. The reaction is diffusion-controlled
Depending on the composition of the elec- and a smooth surface is consequently ob-
trolyte, the ions from the semiconductor tained. This is best demonstrated for Si in
are either soluble in the electrolyte or are the presence of a solution of hydrofluoric
insoluble, and a new phase can form, acid (HF) [44, 45]. Current–voltage studies
generally, an oxide. By choosing an ap- (Fig. 3) performed on Si-HF electrolyte
propriate electrolyte, potential bias, and junctions show that the main requirement
anodization time, the corrosion can be for electrochemical porous Si formation
controlled. is that anodic current densities below a
critical value, icrit , must be used. For a low-
3.2.2.2 Electrochemical Reactions doped n-type semiconductor, light must be
As for many other electrochemical reac- supplied.
tions, two kinetic regimes are distinguish- The exact dissolution chemistry of sil-
able in the anodic dissolution reaction [43]: icon is still a matter of debate; however,
(1) reaction kinetics is rate-limiting (ki- ignoring the intermediate steps involved in
netic control) and (2) the rate-limiting step the dissolution mechanism, the overall dis-
is the diffusion of one of the reactants to (or solution reaction for porous Si formation
products from) the electrode surface. The is as follows:
morphology and properties of the semi-
Si + 6HF + (2 − λ)h+ ===⇒ SiF6 2− + H2
conductor surface are totally different in
the two regimes. Generally, if the elec- + 4H+ + λe− , with λ ≤ 1 (4)
trochemical corrosion reaction is carried
out at sufficiently low current densities, Porous Si formation is characterized by
the electrode kinetics is rate-limiting, in the fact that two charges are required to
14
Electropolishing
12
10 PS formation
icrit
[mA cm–2 ]
8
I
6
Anodic dissolution
4
0
−1 0 1 2
E
[V/SCE]
Fig. 3 Typical current–potential curve for p-type Si in the dark in HF.
dissolve one Si atom from the electrode, 3.2.2.3 Influence of Crystallography

(i.e. the dissolution valence number equals Different crystallographic planes of a semi-
2) [1, 2, 46a,b] and H2 is evolved. When conductor electrode usually exhibit differ-
current density is larger than icrit , mass ent reaction kinetics. It was found in III–V
transfer to the solution limits the reaction compounds in indifferent electrolytes that
and electropolishing occurs with a uniform the (111)B face terminated with the an-
dissolution of all the Si atoms at the ion plane (P, As) etched faster than the
surface of the wafer (dissolution valence (111)A face containing the cations (Ga,
number is 4). It is assumed that an anodic In) [47]. The planes composed entirely of
oxide is formed and is in turn chemically metal atoms react more slowly than any
dissolved in HF. The overall dissolution other crystal plane because of the stable
reaction (Eq. (1) can then be better written metal oxide layer, which can be formed
as [46a,b]) on such planes. Consequently on these
planes termed ‘‘etch stop’’ planes, provi-
Si + 2H2 O + 4h+ ===⇒ SiO2 + 4H+ sion of reactants (diffusion control) is not
(5a) rate-limiting. In Si, the (100) planes are
known to etch faster than (111) planes
SiO2 + 6HF ===⇒ SiF6 2− + 3H2 O in alkaline solutions. This property is at
(5b) the origin of various applications, such as
texturization of silicon surface [formation
icrit value depends on HF concentration. of pyramids on (100) planes], which al-
For low HF concentration (<5%) kinetic lows reduction of reflectivity of the front
control of the electrochemical corrosion surface of solar cells and Si micromachin-
reaction is obtained at current densities no ing [48]. The semiconductor surface may
larger than 50–100 mA cm−2 . be shaped during the anodic dissolution
as a result of the crystal reactivity at the word hole always refers to a defect electron,
different planes. whereas a pore is understood as an etchpit
whose depth exceeds its width.
3.2.2.4 Terminology Anodization under galvanostatic condi-
Beside the diffusion-rate-limited elec- tions is generally the preferred approach
tropolishing resulting in flat surfaces, two for reproducibility attaining a wide range
types of corrosion with kinetically con- of porosity or thickness, but studies under
trolled etching rate can be distinguished. potentiostatic conditions have been per-
When the corrosion slightly modifies the formed. Porous Si films have also been
surface properties leading to the for- generated by stain-etching without apply-
mation of surface microrelief structures, ing light or electrical bias. Strong oxidation
improvement of the surface chemistry and agents added to the electrolyte are used
optoelectronic properties is observed. This to remove electrons from the conduction
corrosion process is named electrochem- band and lead to chemical-etching by the
ical etching (EC etching) or photoelectro- resulting holes.
chemical etching (PEC etching) when it is
additionally light-controlled. On the other 3.2.3
hand, the corrosion type that transforms Porous Silicon
the semiconductor into a porous material
exhibiting different properties than those 3.2.3.1 Anodic Dissolution
of the bulk material is referred to as porous Several remarkable features characterize
etching or porous photoetching when it is the anodic oxidation of Si in fluoride-
light-assisted. containing solutions. First of all, the
Following the IUPAC guidelines, porous anodization is a multistep process that
media in general should be classified as involves charge-transfer reaction both via
the following: microporous, mesoporous, the valence band (hole capture) and via the
and macroporous, when the dominant conduction band (electron injection) [8,
pore sizes are <2 nm, in the 2–50- 50–58]. The dissolution valence number
nm range and >50 nm, respectively [49]. n = 2. Hydrogen evolution is observed
In the porous Si literature, the term at potentials far more positive than the
microporous has often been too loosely Nernst potential of the H/H2 couple [5, 45,
applied and bears no resemblance to 55, 59]. SiF6 2− is the final dissolution prod-
the criteria accepted by the community uct and it is well established (from Fourier-
studying porous carbons, glasses, and so transform infrared measurements) that
on. Part of the problem originated from during pore formation, the Si surface
the fact that the optoelectronic properties is mainly hydride-terminated (Si−H) [60,
of porous Si are largely determined by 61]. The two models proposed by Lehmann
the skeleton size and not pore size. New and Gösele [62] and Gerischer and cowork-
terminology such as nanoporous Si has ers [57, 58] take into account all the facts
been frequently used to indicate a solid mentioned earlier. Both models include
skeleton with dimensions small enough two electrochemical reactions (hole cap-
to constitute a nanostructure. The term ture and electron injection) followed by a
nanoporous generally comprises those of chemical oxidation reaction. It is generally
microporous and mesoporous Si. To use accepted that the formation or decompo-
consistent terminology in this paper, the sition of a crystal occurs preferentially at
weakly coordinated sites, such as kink sites by injection of an electron into the con-
(present in steps on the surface or at the induction band. The reaction steps at a
tersection of dislocations), where Si atoms kink site (in which the Si atom has two
have two or three ‘‘dangling’’ bonds ex- bonds linked to the lattice and two other
posed to the solution. Si atoms on kink site bonds saturated by H-ligands) for the dis-
atoms have a binding energy, that is, on an solution in aqueous HF are described in
average, half of that which they would have Fig. 4, at a molecular level with a two-
inside the crystal. In both models it is as- dimensional representation, following the
sumed that a valence band hole is captured Gerischer and coworkers’ model. The elec-
in a Si−H surface species, resulting in an trochemical reaction is initialized by a hole
electron-deficient Si−H bond. The main capture at a Si−H bond, which is thus
difference between the two descriptions of weakened. The H atom is captured by an
the surface chemistry is the position of the F− ion as a proton H+ , leaving its electron
•
positive charge in the electron-deficient at the Si atom (≡Si radical). By interac-
Si−H bond. Lehmann and Gösele as- tion with F− ions, the unpaired electron
sume that the positive charge is located can be excited up to the conduction band
on the silicon side of the bond, resulting edge. The excited unpaired electron is in-
in a hydrogen radical [62]. Gerischer and jected into the conduction band, whereas
coworkers assume that hole-trapping in the electrons of the F− ion form a new
the Si−H bond results in a silicon sur- very stable Si−F bond. The large elec-
face radical species and a proton [57]. In tronegativity difference between Si and F
both cases the hole-capture step is followed is responsible for the polarization of the
A B C
Si Si Si
H H
Si Si H Si OH
H
+h+ + +H2O +H+
(I) Si Si Si Si +H Si Si
H −e− H
Si Si H Si
H H H
C′ A′
Si Si
Si H Si H OH H
F +2H2O
+F− + Si
(II) Si Si Si H
+OH− H F
Si H OH
Si
H H
+ H2O, HF X X
(III) Si + H2 (X = OH, F)
(In solution) X F
Fig. 4 Two-dimensional description of the reaction steps at a kink site on a silicon
surface for the oxidation initiated by a hole (from Ref. 58).
two Si−Si back bonds underneath. This intensity, Q is almost 4 but it drops to 2
increases their chemical reactivity and fa- as the photon flux is increased. An expla-
vors the chemical disruption of one back nation for the current-quadrupling effect
bond. The splitting of the back bond may has been sought in terms of a model in-
well occur by HF or by H2 O. Two new volving four consecutive electrochemical
bonds are formed: Si−H at a back Si, and dissolution steps [53, 55, 56, 59]. In the
Si−F at the kink site (or Si−OH with fur- initial step, a photogenerated hole is cap-
ther exchange of the OH ligand by F− ). tured in a surface bond and a surface
The remaining Si−Si back bond is further intermediate is formed. This intermediate
weakened (and even more polarized) and is further oxidized to a Si(IV) species by
will be quickly split by another HF (or three consecutive electrochemical steps,
H2 O) molecule. The Si atom at the kink in which electrons are thermally excited
site leaves the surface with a divalent oxi- to the conduction band. The variation of
dation state as a HSiF3 molecule and the Q with the light intensity is explained by
kink site situation is restituted at the next assuming that electron injection from sur-
atom of the step (situation A& ). The HSiF3 face intermediates and hole capture by
molecule reacts chemically with water to surface intermediates are competing pro-
give H2 gas, as shown in the following cesses in two of the three charge-transfer
reaction: steps.
HSiF3 + H2 O ===⇒ SiF3 OH + H2 (6a) 3.2.3.2 Influence of the Si Conduction

SiF3 OH + HF ===⇒ SiF4 + H2 O (6b) Type on Current–Voltage Curves
Although the chemical reactions at the sili-
However, attempts to detect the hypo- con/HF interface are the same for n- and
thetical intermediate HSiF3 species dur- p-type silicon, there is a basic asymmetry
ing porous Si formation by in situ in their electronic properties. For the
Fourier-transform electrochemically mod- dissolution of silicon under an ‘‘anodic
ulated infrared spectroscopy have been bias’’, the p-type Si is forward-biased
unsuccessful [61]. Calculations show that and the current is caused by thermally
if such a species or a similar one ex- generated majority carriers. The n-type Si
ists, their lifetime must be shorter than is reverse-biased and undergoes charge
0.3 ms. depletion. The current is characteristic of
Undoubtedly, the most convincing ev- a minority carrier flow. For low-doped n-
idence in support of the hypothesis of Si, anodic dissolution uses photogenerated
the electron-injection process is the obser- minority carriers. The i-V curve for p-
vation of the photocurrent multiplication type Si has been discussed earlier and
observed during anodization of n-type in comparison with the i-V curves for n-Si
HF solutions. [8, 53–56]. More than one are presented here. For n-Si under reverse
charge carrier per absorbed photon con- bias condition, a dark current and/or a
tributes to the photocurrent measured in photocurrent are observed, depending on
the external circuit. Photocurrent quan- the doping density and illumination level.
tum efficiency, Q, (i.e. number of charge In the dark: The i-V curve is character-
carriers measured in the external cur- ized by a breakdown mechanism (Fig. 5,
rent by absorbed photon) is dependent curve d). It occurs in Si for electrical field
on the incident photon flux. At low light strength in excess of 3 × 105 V cm−1 [63].
60 Fig. 5 Current–potential curves vs SCE

a
for n-type lightly doped (Nd = 1015 cm−3 )
50 Si in 5% HF in the dark and under
illumination. p-Type Si (curve a) is
Current density j
40 indicated for comparison.

[mA cm−2 ]
j max
30
b
20
j psi
10 c
d
0
−2 0 2 4 6 8 10 12
Applied voltage
[V]
When a potential that is large compared in a doubling or quadrupling of the

with kT is applied to the interface, a rel- photocurrent, respectively [8, 53–56]. For
atively large band-bending occurs. The lightly doped Si, the total current corre-
thickness of the space charge layer, also sponds to the superposition of two cur-
termed surface charge region (SCR), be- rents: the light-generated hole current and
comes very small and the field becomes the electron-injection current. The latter is
very strong. Electrons can tunnel directly not constant but depends on the absolute
from the surface Si atoms into the conduc- current density [68, 69]. For highly doped
tion band of the semiconductor, resulting Si, the dark current caused by the electrical
in the dissolution of silicon [64–67]. The breakdown adds a third contribution to the
onset of the dark tunnelling current de- total current (Fig. 6) [70].
pends on the doping density. For flat In short, for high illumination intensi-
surfaces, it occurs at high anodic overpo- ties, the i-V curve of n-type Si is identical
tential (V = 10 to 15 V) for lightly doped to that of p-type Si, with the exception of
n-type Si (Nd = 1015 cm−3 ) and at low volt- a cathodic shift of a few hundred milli-
age (V = −0.2 V vs SCE) for highly doped volts [44, 45, 69, 71] (Fig. 5, curve b). The
n-type Si (Nd = 1018 cm−3 ) [67]. photocurrent varies with the light inten-
Under illumination: The photocurrent sity up to 50 mW cm−2 [70]. At higher
induced by the light is not directly given intensities, the current becomes almost
by the number of photogenerated holes independent of the light intensity, indi-
generated [68, 69] and as a result does cating that the rate-limiting process is no
not increase linearly with the light flux longer the supply of holes to the surface
because of the electron-injection process. but rather the dissolution of the corrosion
For example, as already mentioned, at products. If the light flux is limited such
very low illumination intensities (less that the photocurrent never reaches the
than 5 µW cm−2 ) for every one and up icrit value, the photocurrent levels off to
to three photogenerated holes, electrons a specific constant value independent of
may be injected into the conduction the applied voltage (Fig. 5, curve c) [44, 45,
band by intermediate complexes, resulting 69]. At potentials to the positive side of
Fig. 6 Current–potential curves for 35

n-type highly doped (Nd = 1018 cm−3 )
Si in the dark and under illumination in 30
5% HF; scan rate = 100 mV s−1 (from
Ref. 70). Under illumination
25
[mA cm−2 ]
20
j
15
10
In the dark
5
0
−1.0 −0.5 0 0.5 1.0
E vs SCE
[V]
icrit , the electropolishing regime is never 3.2.3.4 Interfacial Potential Distribution

reached because of the lack of carriers. In and Pore-Formation Conditions
this case it is possible to adjust the voltage An anodically biased n-type semiconductor
and current independently by controlling is divided into two regions: a space charge
the light flux. In particular, the depth of the region (SCR), which is a depletion region
SCR can be changed without changing the and a field-free region, which extends
current density. In Sect. 3.5, a discussion after the SCR into the bulk region. The
on how this influences the morphology of thickness of the SCR, W , at a planar
the porous Si is presented [72–75]. surface under depletion conditions is given
in Eq. (7). It is a function of the voltage drop
3.2.3.3 Influence of the Solvent on in the SCR, VSCR , and the doping level Nd :
Current–Voltage Curves ! "0.5
2ε0 ε1 VSCR
Experiments have been carried out in HF- W = (7)
contained acetonitrile (MeCN), dimethyl- eNd
formamide (DMF), and dimethylsulfoxide For example, W can vary more than one or-
(DMSO) with tetrabutylammonium per- der of magnitude (from 0.5 to 0.02 µm for a
chlorate (TBAP) as the supporting elec- 1 Volt polarization) between lightly (Nd =
trolyte [76–80]. The oxidation current for 1015 cm−3 dopant) and highly doped Si
p-type (in the dark, ρ = 100 − 0.2 # cm) (Nd = 1018 cm−3 ), respectively.
and n-type Si (under illumination) in- On the electrolyte side, two regions can
creases linearly because of the combined be distinguished: the Helmholtz layer,
solution and semiconductor resistance. which is a region of specifically absorbed
Neither a critical current peak (icrit ) nor an ions, water molecules, and acceptor ions,
electropolishing region is observed. There and the bulk region.
was no H2 evolution when water-free HF It is important to consider the relative
was used. magnitude of the potential drop in each
region because the less rate-limiting re- ranging from nanoporous silicon with
action creates the greatest potential drop pores and crystallites in the nanometer
and therefore dictates the specific disso- scale up to macroporous silicon with pore
lution step. Impedance spectroscopy and and pillar dimensions in the microme-
capacity measurements of the silicon/HF ter scale. Each of these variations has its
system have helped to clarify part of the origin in different physical effects. The
controversy in the potential distribution main pores grow in (100) directions. Most
at the silicon–electrolyte interface [81–84]. pores extend from the surface down to
The applied potential is partitioned across the porous Si/bulk Si interface. Although
the electrochemical double-layer and the side pore formations and branching may
SCR in the silicon. The fraction of the be extensive, pores are never seen to inter-
potential dropped across the SCR is depen- connect. Porous Si is macroscopically and
dent on dopant concentration and etching roughly characterized by its porosity, fvoid ,
conditions although n-type Si represents a which is the fraction of voids in the porous
limiting case wherein the applied poten- layer.
tial is in the major part dropped across the
SCR when no pores are present (flat sur-
3.2.3.5.1 Nanoporous Silicon Including Mi-
face). This means that the potential drop is
cro and Mesoporous Si It consists of a
equal to the difference between the applied
potential, V , and the flat band potential, spongelike randomly distributed array of
Vfb (VSCR = V − Vfb ). For p-type and de- nanometer-sized silicon features. When
generately doped Si, the process is more the size of the features is less than 5 nm,
difficult to analyze quantitatively because quantum size effect caused by the confine-
of the partitioning of the potential and the ment of the excitons is assumed, which
nonuniform current distribution at the in- gives rise to room temperature photolumi-
terface between the Si substrate and the nescence and electroluminescence in the
porous layer. Under actual pore-formation visible range (1.6–1.8 eV). These proper-
conditions for all types of Si samples, one ties have been intensively investigated for
emerging opinion is that both the SCR and applications of light-emitting devices [6,
the Helmholtz regions control the anodic 88], although porous Si is highly resistive
behavior of silicon. The Helmholtz region in air (ρ = 105 –107 # cm) [89].
controls the charge-transfer reaction rate,
whereas the semiconductor SCR controls Microporous Si It originates from low-
the specific morphology [83, 84]. doped p-type silicon (1 # cm) [5–7, 85,
89, 90]. By employing modified transmis-
3.2.3.5 Many Different Morphologies sion electron microscopy (TEM) specimen
The formation of porous Si critically de- preparation such as direct specimen cleav-
pends on the type of conductivity and age, it has been observed that the coral-like
doping level of the Si electrode, HF concen- structure of anodized p-type Si is com-
tration in the electrolyte, nature of solvent posed of highly interconnected micropores
in which HF is diluted, and applied current in the size range ≤2 nm (Fig. 7). High res-
density (or voltage) [5, 7, 10, 11, 71, 85–87]. olution transmission electron microscopy
Porous Si exhibits distinctive appear- (HRTEM) images showed that strongly
ances (Table 1). The size of the structures luminescent samples (>70% porosity) ex-
varies by three orders of magnitude, hibit narrow undulating Si columns with
Tab. 1 Different varieties of porous Si.
Type Doping density HF Current density/ Morphology

concentration voltage
p 1015 cm−3 25% aqueous 50 mA cm−2 Microporous [6]

HF
p+ and n+ 1017 –1019 cm−3 25% aqueous 3–300 mA cm−2 Mesoporous, [86, 89]
HF breakdown
tunneling
p 1014 –1015 cm−3 30% aqueous 40 mA cm−2 Macroporous [102]
HF
1016 –1017 cm−3 8% HF in DMF 10 mA cm−2 Idem [80]
or DMSO
n 1015 –1018 cm−3 5% aqueous 15 mA cm−2 Duplex layer
front-side (nanoporous +
illumination macroporous) [70]
n, (100) 2.1015 cm−3 6% HF aqueous 10 mA cm−2 Array of vertical
oriented back-side macropores developed
illumination on prestructured
nuclei [72]
n 2.1015 cm−3 5% HF aqueous 10 V 20 µm deep
channels [67]
3.1018 cm−3 2–15 mA cm−2 Duplex layer
(nanoporous +
macroporous) [93]
Fig. 7 TEM image and diffraction pattern from microporous Si

formed on low-doped p-type Si (from Ref. 11).
Fig. 8 TEM image (bright-field,

under focus) of thin,
high-porosity luminescent
microporous. Columnar Si
structures are arrowed (from
Ref. 7).
diameter size down to <3 nm (Fig. 8). The As a result, the photoluminescence output
porous material is completely crystalline. rises dramatically and the peak position
This material also presents a very large is blue-shifted toward 780 nm (Fig. 9).
internal surface area (1000 m2 cm−3 ) and Because the removal of Si constitutes
is therefore suitable for catalytic surface an overall decrease in Si skeleton, this
reactions. Freshly prepared microporous experience was considered as an indirect
Si luminesces at room temperature. The evidence that visible room temperature
position and intensity of the maximum photoluminescence could be attributed to
of the luminescence peak depends on the quantum size effect [6].
porosity. It varies from 1000 nm to 880 nm
for 64% and 77% porosity, respectively. Mesoporous Silicon This kind of porous
Porosity of microporous Si, further sub- silicon forms on heavily (degenerated)
jected to a continuous chemical etching doped silicon substrates, either n+ -
in 40% HF, increases from 70 to 80%. type or p + -type (0.01–0.001 # cm), under
1.4 1.6 1.8 2.0
300K
Photoluminescence intensity
[arbitrary units]
1hr
2hr
6hr
×1 Fig. 9 Room temperature

×3
×40 photoluminescence spectra of a
freshly prepared microporous Si
as a result of a partial chemical
dissolution in 40% aqueous HF
1.0 0.9 0.8 0.7 0.6 for the times indicated (from
Photon wavelength Ref. 6).
[µm]
Fig. 10 TEM cross-sectional (b)

image of mesoporous structure
formed by anodizing low-doped
p-type Si (64% porosity) (from
Ref. 89).
conditions that would produce microp- in aqueous HF electrolyte [101, 103]. Low-
ores in low and moderately doped Si. and medium-doped p-type (5–0.2 # cm) is
The pore and pillar sizes are in the made macroporous in HF-containing or-
10–50 nm range [5, 11, 85, 86, 89, 91, ganic electrolyte [79, 80, 03]. Macroporous
92] (Fig. 10). The macroscopic surface is Si consists of micrometer-sized pores and
slightly darker-looking in comparison to silicon pillars. The internal surface area is
bulk silicon, and of excellent smoothness. only 1 m2 cm−3 .
The internal surface area is important
(100 m2 cm−3 ). Mesoporous Si (like mi- Macropore on n and n+ -Type Si Un-
croporous Si) can be easily etched in der Illumination Macropore formation
low-concentration alkaline solutions. The on n-type in aqueous HF under il-
mesoporous Si has been exploited for lumination was studied in detail in
the fabrication of silicon on insulator the 1990s [70–72, 99, 100]. Macropore
(SOI) structures [94], using variations in formation is anisotropic and occurs
porosity in buried layers with altered dop- preferentially along the (100), (010), and
ing. It is also used as a sacrificial layer (001) directions. Figure 11 shows that
in microelectronic and thin film micro- anisotropy should result in the formation
crystalline solar cells [95]. Mesoporous Si of square-shaped tubular pores on a (100)-
layers, depending on the porosity and layer oriented substrate and triangular pores on
width, present remarkable antireflective a (111)-oriented substrate [71]. This has in-
properties and as such can be used as an deed been experimentally confirmed and
antireflective coating in crystalline Si so- macropores formed on (100) substrates are
lar cells [96–98]. Mesoporous Si exhibits tubular with a square-sectional form and a
a weaker luminescence than microporous round bottom (Fig. 12a) [70]. In the case of
Si. This result suggests that quantum size (111) substrates, they have an almost trian-
Si particles are present within the meso- gular section (Fig. 12b). In both cases the
porous structure. macropores delineate a regular network.
The pore-to-pore distance equals twice the
3.2.3.5.2 Macroporous Silicon n-Type value of the SCR, which can be adjusted
silicon (low, medium, and high) [67, by doping level and applied voltage. The
69–71, 99, 100] and very low-doped p- pore size can be controlled by doping, HF
type (<5 # cm) are made macroporous concentration, and applied current density.
Fig. 11 Schema showing that the

porous photoetching of Si occurs
(100) plane (111) plane preferentially along the (100), (010), and
(001) directions. Pores with a square
shape and triangular shape are expected
to be formed for (100)- and
(111)-oriented substrates, respectively
(from Ref. 71).
SC lk
R
bu
SCR
bulk
(a) (b)
Fig. 12 Plan view SEM images of macroporous Si formed on highly doped

(Nd = 1018 cm−3 ) n-type Si with a charge of 10 C cm−2 and after dissolution of the
nanoporous layer. (a) (100)-oriented substrate; (b) (111)-oriented substrate (from
Ref. 70).
The active state of the pore tips is due to the tubular and triangular macropores formed
collection efficiency required for the holes on (100)- and (111)-oriented substrates
initiating the dissolution process. The pas- (Fig. 14). The nanoporous Si is charac-
sivation of the pore walls is ascribed to terized through a network of tangled
a depletion of holes in the walls between filaments of nanometer width, some
the pores. This has been confirmed by of them exhibiting quantum size effect
spectroscopy impedance measurements (Fig. 14) [10, 91]. This nanoporous Si that
on macroporous n+ -type Si [104]. appears fibrous was shown to be the
When the silicon–electrolyte interface source of the visible luminescence [92].
is front-side illuminated during the an- This is not the case when the interface
odization, the macropore Si walls and the is illuminated from the backside of the
top of the Si pillars are covered by a silicon substrate leading to macroporous
nanoporous silicon layer (Fig. 13) [70, 91, Si formation only [72]. The possibility of
92, 99, 105, 106]. Cross-sections of elec- controlling pore location with a standard
trochemically detached nanoporous layers photolithography technique and produc-
well illustrate the morphology of the ing (with rear-light-controlled corrosion)
Fig. 13 Cross-sectional SEM image of porous Si formed by

porous photoetching on (100) low-doped Si (Q = 4 C cm−2 ),
revealing the distribution of the nanoporous (top layer) and
the macroporous layer (bottom layer) (from Ref. 70).
(a) (b)
500 nm
500 nm
Fig. 14 TEM micrographs (cross-section) revealing the typical internal structure through
the thickness of the nanoporous Si formed on highly doped n-type Si (Nd = 1018 cm−3 )
with a charge of Q = 10 C cm−2 : (a) (100) oriented; (b) (111) oriented (from Refs. 91).
well-defined macroporous structures in Si, low-doped (Nd = 2.1015 cm−3 ) n-type Si in

with a pore aspect ratio of 250, has led 5% HF under breakdown conditions (10 V
to new technological applications such as bias) [67]. Recently, it has been shown that
two-dimensional photonic crystals with a highly doped (Nd > 2.1018 cm−3 ) porous
band gap in the near and mid-infrared re- Si obtained under breakdown conditions
gions [75, 107–111], microcapacitors [73, (0 V/SCE) is constituted of a duplex layer
74], etc. of mesoporous and macroporous Si. Pore
diameter (15–70 nm) and shape (round or
Macropore on n+ -Type Si in the Dark square section) of mesopores depends on
Macropore formation was first observed in the amount of charge and current density
1972 by Theunissen who found formation used for its formation. Pore diameter of
of 20 µm deep channels that occurred on macroporous Si is constant and of the
order of 0.3 µm [93]. The mesoporous Si and pore wall thicknesses are two times
layer can be dissolved in 0.1 M NaOH, smaller (Fig. 15). When porous-etching
whereas the macroporous Si was only is done in an organic electrolyte, the
dissolved in HF (48%): HNO3 (69%) (1 : 9 macropore shape is better defined and the
ratio volume). This permits measuring the diameter is larger (0.3–1 µm) [113].
thickness of each layer by profilometry.
Depending on the current density and Macropore on p-Type Si Macropore for-
amount of charge used during porous- mation in p-type Si was recently evi-
etching, thicknesses up to 5 µm and 4 µm denced [77–80, 101–103, 108]. Macrop-
for nanoporous and macroporous layers orous p-type Si was first observed on
respectively, have been measured [112]. low-doped (>10 # cm) substrates. It
The morphology resembles that shown in can be formed in HF contained aque-
Fig. 13 for macroporous Si obtained under ous and organic solutions (Fig. 16a).
illumination, except that pore diameter Using specific HF/organic electrolytes
(a) (b)
20 nm 35 nm
Fig. 15 HRTEM micrographs showing details of the nanometer-sized

fibers (plan-view): (a) (100) oriented, Nd = 1015 cm−3 ; (b) (111)
oriented, Nd = 1018 cm−3 (from Refs. 91, 92).
a b
Fig. 16 Cross-sectional SEM images of macroporous Si formed on (100)

p-type Si. (a) Na = 1015 cm−3 in 10% HF/dimethylsulfoxide;
(b) Na = 1017 cm−3 in 8% HF/dimethylformamide.
[dimethylformamide (DMF) and dimethyl- Kinetics of pore nucleation on (100)

sulfoxide (DMSO)] permitted macropore p-type (0.05 # cm) and n-type (1–4 # cm)
formation on (0.2 # cm) medium-doped was observed as small steps in fast-
Si (Fig. 16b) [80]. The macropores diam- current and potential pulse transients.
eter and depth depend on the doping It was shown that the formation of
density. The diameter varies from a few porous Si could be better explained and
microns to 0.2 µ. The macropore depth correlated with experimental results if
can reach several hundred microns when a silicon oxide intermediate was consid-
organic electrolytes are used [114]. ered [116].
From the following observations that
3.2.3.6 Models of Porous Si Formation (1) Si dissolves with formation of molecu-
Several qualitative models have been pro- lar H2 and protons, (2) only Si forms mi-
posed to explain porous Si formation croporous layers with a single crystalline
but none of them allow full explana- skeleton, and (3) hydrogen (Deuterium)
tion of the rich variety of morphology species penetrate into the Si substrate
exhibited by porous Si and, in particu- upon porous Si formation, a model was
lar, the formation of the duplex layers proposed to explain pore initiation. It was
(nano + macroporous). In addition, they assumed that hydrogen incorporation in-
possess very little predictive power. A ma- duces structural defects in Si, which may
jority of the models focussed on the pore act as active sites for the localized Si disso-
propagation, whereas the mechanism of lution. The model relates pore initiation to
pore initiation received very little attention. the selective dissolution of the hydrogen-
A comprehensive review of the various induced structural defects at the surface of
models proposed to explain pore forma- bulk silicon [117, 118].
tion is found in excellent review articles by
Smith and Collins [5], Parkhutik [12], and Microetchpits One model developed to
Chazalviel and coworkers [13]. Two main explain etchpit formation on II–VI semi-
categories of models have been proposed. conductors [119–121] has been extended
The first one is basically electrostatic in to porous Si. This model attempts to
nature, based on the consideration that address the initiation of the n-type macrop-
physical effects associated with the SCR orous morphology through nonuniformi-
play a major role in the pore-formation ties in the photoetching current caused by
mechanism. The second category is based nonuniform microfields existing around
on computer simulations. the dopant atoms nearest to the semicon-
ductor–electrolyte interface [100], or alter-
3.2.3.6.1 Pore Initiation The few pro- natively near the semiconductor surface
posed models are related to the surface because of a trapped (localized) positive
chemistry of Si during pore initiation. In charge formed in the first instant of the
general, defects are invoked as active sites photocorrosion at the interface. A calcula-
for pore initiation. However, scanning tion of the relevant parameters as a func-
tunneling microscopy (STM) observations tion of the doping density of the silicon
have clearly shown that in initial stages wafer shows that up to Nd = 1017 cm−3 ,
the dissolution is isotropic and uniform the Debye screening length is larger than
on the nanometer scale, meaning that it is the mean distance of the first donor atoms.
independent of defects [115]. This means that up to this doping value
the donor charges are not shielded and at the etchpit initiation sites or at the bot-
that the donors are expected to have an in- tom of the pores). Electrochemical kinetics
fluence on the etchpit pattern. The first determines how fast holes react with the
layer of surface at sites when they are available.
# dopant
1/3 $
(with a mean distance
a = 1/ 2Nd from the semiconductor Specific surface chemistries can make
surface) contributes to a nonuniform pho- the reactivity for holes strongly differ-
tocurrent, which is high near the dopant ent from one semiconductor–electrolyte
atoms and can be ascribed to a photoin- system to another. The disadvantages of
duced avalanche effect with a multiplica- the most popular models proposed to ex-
tion factor M(Ea ) where Ea is the electric plain porous Si formation are that they
field near the dopant atom [100, 122]. An do not take into account the chemical as-
important consequence of this model is pects that may influence the kinetics at the
that the surface layer becomes dopant-poor Si/electrolyte interface.
after PEC-etching. This assumption was For pores to propagate, the pore walls
supported by the reduced surface recombi- have to be passivated and the pore tips
nation velocity, as indicated by photovoltaic to be active in the dissolution reaction.
measurements, which show substantial Consequently, a surface, which is depleted
gain in the short wavelength part of the of holes, is passivated to porous-etching,
photocurrent spectrum and increased fill which means that anodization is self-
factor, following PEC etching [123]. Once limiting.
the initiation process sets in, other mecha- One must emphasize the distinct role
nisms, such as photocurrent focussing in of the SCR for n- and p-type Si. In the
the tip, may take place. A redistribution case of p-Si, it acts as a barrier for the
holes, and only near the pore tips is the
process starts. Some pores stop growing
field sufficiently large for opening gaps
and terminate, whereas other pores con-
at the top of the barrier. The reaction
tinue to grow with increasing diameters.
may proceed entirely over the valence
Lehmann and Föll [71, 72] showed that
band (n = 2). In the case of n-type Si,
pore initiation at the surface is determined
the SCR rather acts as a sink channeling
by the doping density of the Si substrate.
the holes to the pore tips. The porous Si
From the initial 1010 cm−2 PEC-etched
may be formed on n-type Si under front
pits formed, only one in a hundred or
and back illumination or under strong
one in a thousand survive and become
polarization (avalanche breakdown). On
pore tips. The density of the resulting pore
heavily doped n- and p-type Si, porous
tips is determined by the SCR width and
Si may be obtained by hole-tunneling.
therefore by the doping density.
Eventual presence of intermediate Si
oxide may favor the electron-tunneling
3.2.3.6.2 Pore Propagation process through the conduction band.
Models Based on the Role of the Space In summary, the effects responsible for
Charge Region Two fundamental aspects pore wall passivation (top row) and for
of porous-etching are the electrostatics and passivation breakdown at the pore tip
electrochemical kinetics. Electrostatics is (middle row) as well as the resulting kind of
crucial in determining the sites wherein porous Si structure together with substrate
semiconductor valence band holes are doping type (bottom row) are presented in
available for the dissolution reactions (e.g. Fig. 17 [86].
Field effect
(a)
(SCR)
(b) Avalanche Minority carrier

Tunneling Diffusion
breakdown collection
Mesoporous Si Macroporous Si Macroetchpits Macroporous Si

(c)
on p + and n on p − on n − on n −
Fig. 17 Field effect is proposed to be responsible for pore wall passivation

(a). Effects responsible for passivation breakdown at the pore tip (b) and the
resulting porous silicon structure together with substrate doping type (c)
(adapted from Ref. 86).
Models Relevant to Micropore Formation field. As a pore grows, the radius of curva-
The mechanism of microporous Si for- ture increases and the local field decreases
mation from p-type is controversial. A and the pore growth decreases. To main-
puzzling question remains as to the un- tain the total current density, new pores
usually high stability of thin fibers or start to grow. As soon as the fiber/particle
nanoparticles of Si to electrolytic attack or, diameter in microporous Si gets thin, its
in other words, what prevents nanometer- resistivity is sharply increased by the Fermi
thick fibers/particles from being dissolved level pinning to surface states. Because the
altogether as the etching front moves mi- electrochemical dissolution is essentially a
crons into the crystal bulk. current-controlled process, further etching
The first model, proposed by Beale and is impossible and a thin fiber/nanoparticle
coworkers [85, 89], assumes that at po- is stable. This model is often disregarded
tentials corresponding to pore growth the because Fermi level pinning by surface
Fermi level of Si is pinned near the middle states must be assumed. Unfortunately
of the band gap due to of a high number electron paramagnetic resonance (EPR)
of surface states. This results in a potential measurements indicate that there is a low
drop at the interface, which restricts the density of surface states in the porous
hole transfer from the substrate. Anodic layer [124, 125]. This has been confirmed
currents may flow as a result of thermionic by Searson and coworkers, using in situ
emission across the potential barrier. The impedance spectroscopy to measure inter-
local current flow is then determined by face states at [(111), Nd = 1013 cm−3 ] Si
the potential height, which in turn is de- surfaces in fluoride solutions [126, 127].
pendent on the local electric field. High They found, at pH between 3 and 6, a very
electric field favors hole transfer. Hence, low density of electrically active surface
the electric field strength being greatest at states equal to 2 × 1010 cm−2 correspond-
the base pore, this results in dissolution ing to about one in every 105 surface states.
only at the pore tips. Therefore, following Nevertheless, the idea of barrier-lowe-
pore initiation, the pores propagate as a re- ring may remain relevant. For example,
sult of the focussing effect of the electrical if part of the applied potential appears in
the Helmholtz layer [83, 84], the barrier current [128]. In addition, the ‘‘quantum
will be decreased by the corresponding wire model’’ has nothing to say about the
amount. If the Helmholtz potential drop formation of all other types of pores.
is proportional to the surface electric field,
the barrier lowering will be inversely Models Relevant to Mesopore Formation
proportional to the SCR thickness. The Beale and coworkers’ model has been
A very often cited model is the so-called extended to explain mesopore formation
‘‘quantum wire model’’ [62]. The aim of on degenerately doped p- and n-type Si (in
the model developed by Lehmann and the dark) for which hole-tunneling is the
Gosële was to explain the morphology ob- major charge-transfer mechanism [85, 89].
served for porous Si, which luminesces as The probability of a hole-tunneling across
a result of quantum confinement. As the the Si-electrolyte interface depends on both
dimensions of the particle decrease, the the potential barrier height and SCR width.
band gap increases (from 1.1 to 1.5 eV). The total electric field at the pore tip results
Because of quantum confinement, the in both a lowering of the potential barrier
energy of valence band states is low- and a decrease of the SCR width. Hence,
ered in the walls between the micropores. as mentioned earlier, the propagation of
The holes are transferred to the interface pores is favored. As the pores grow, the
only at the pore bottoms. The quantum depletion layers of adjacent pores may
confinement model may be regarded as overlap. This results in the probability of
chemical passivation of pore wall disso- tunneling becoming negligible at the pore
lution, where the nanocrystalline silicon walls; that is, the reaction is passivated
particles are the passivating species. How- at the pore walls. Overlap of the depletion
ever, this model seems to be contradictory layers prevents branching and explains the
to the well-established properties of iso- observed columnar structure.
type heterojunctions. A barrier height of a Zhang [66] and Searson [9] confirmed
few hundred meV is known to be largely that model and evoked a ‘‘field-enhance-
insufficient for blocking majority carrier ment effect’’ at the tips of the pores
transport in a heterojunction at room tem- wherein the current flow is controlled
perature for both the directions of electric by a tunneling mechanism (Fig. 18). Their
Field strength (a.u.)
Fig. 18 Field strength distribution around an elliptically shaped

pore tip (from Ref. 8).
model predicts the formation of highly on n+ -Si as well as on low- and medium-
oriented parallel pores for (100)-oriented doped p-type Si.
substrates and gives a reasonable expla- 1. n-type porous Si under illumination:
nation for the regular pore distribution There is a reasonable agreement on the
and spacing by considering the depletion formation of macroporous Si from n-
layer width. The calculation of the current type Si [71, 72]. The formation mechanism
distribution around the pore front using of n-type macroporous Si is ruled by
a two-dimensional solution to Poisson’s the reversed-biased SCR at the solution
equation shows that the current is consid- interface.
erably greater at the pore tip as compared Front-side illumination: The condition
with the pore wall [9], accounting for the at the pore tip of an illuminated n-type
unidirectional growth of the pores. In the electrode is different from that in the
case of multiple adjacent pores, the equipo- dark, because the presence of a breakdown
tential lines in the region between the pore field is not necessary to generate charge
walls are shifted further away from the carriers. Every depression or pit in the
surface and into the bulk, effectively de- surface of the n-type silicon anode bends
pleting the pore wall regions. The pore the electric field in the SCR in a way
spacing is determined by the distance at that the concave surface regions become
which the region between the pore walls more efficient in collecting holes than the
becomes completely depleted and is com- convex ones. Concave regions are etched
parable to the calculated SCR width at a preferentially and the pores start to grow,
planar surface. consequently enhancing the local current
In 1972, Theunissen [67] observed for- density [68].
mation of wide-etched channels for donor After initiation of microetchpits, macro-
concentrations of less than Nd = 2 × pore formation occurs. The density of the
1017 cm−3 under high voltage (>10 V), in resulting pore tips is determined by the
2.5–5% HF, when anodization was per- SCR width and therefore by the doping
formed in the dark. He proposed that at density. The thickness of the remaining sil-
some pore tips the electrical field strength icon walls between the pores is two times
was sufficient to allow avalanche break- the SCR width and as a result the pore walls
down at the depletion layer, which then are charge carrier–depleted (Fig. 19). This
generated the necessary carriers for fur- finding has been confirmed by impedance
ther pore growth. This was corroborated spectroscopy studies [104].
recently by Lehmann et al. [86] show- Until now, no model has been proposed
ing experimentally and by simulations to explain the simultaneous formation
based on the electrical field distribution of the duplex nanoporous–macroporous
present at the pore tip and pore walls structure, when the n-Si and n+ -type Si
that the mesopore formation is domi- are illuminated from the front side.
nated by charge carrier tunneling, whereas The striking property of the orienta-
avalanche breakdown is responsible for the tion dependence of the nanoporous and
formation of large etchpits. macroporous structures may be correlated
with the observation that the critical cur-
Models Relevant to Macropore Formation rent icrit is greatest for the (100)-oriented
Macropores can be formed under illumi- substrate [72]. One possible explanation
nation on n- and n+ -Si and in the dark considers the process in which one Si−H
Fig. 19 Schema showing the effective

width of the SCR. (a) when the wall
thickness, d, is smaller than twice the
SCR thickness, W, (b) when d is larger
than twice W (from Ref. 66).
bond is substituted by a Si−F bond and suitable wavelength from the rear of the
may well be an indication that the surface Si wafer favors the collection of the pho-
chemistry is an extremely sensitive func- togenerated holes in the bulk and their
tion of the silicon crystallographic orienta- migration toward the tip of the initiated
tion. For instance, the unreconstructed Si pits. Such an experiment is only possible
(111) surface is ideally terminated by one for lightly doped n-type Si, because the dif-
H atom per Si, forming a monohydride fusion length of the holes is comparable
(≡SiH) [129], and the Si (100) surface by to the sample thickness. Stable macropore
two H atoms per Si, forming a dihydride growth occurs when the current density
(=SiH2 ) [130]. It has been proposed that is limited by hole generation and not by
because of the presence of the two H atoms the applied bias. In addition, if the local
per Si the dihydride-terminated (100) sur- current density at the pore tip, itip , exceeds
face is sterically hindered [77, 78, 31]. The the critical current density (itip ≥ icrit ), no
steric hindrance causes bond strain, en- nanoporous silicon will form at the pore
hances the chemical reactivity, and the tip because of the presence of the elec-
dissolution will occur faster along the (100) tropolishing regime. For a given initiation
planes [5, 72]. pattern, the pore diameter, dp , and wall
Rear illumination: When anodization is thickness, W , are determined by the ratio
performed under illumination from the between the applied current, i, and the
backside, holes move to the interface by critical current, icrit [Eqs. (8) and (9)].
diffusion until they are captured by the % &
i 0.5
SCR, which accelerates them toward the dp = p (8)
icrit
pore bottoms. Lehmann [72] has clearly
demonstrated the validity of the macro- and
' % &0.5 (
pore propagation model by studying, in i
aqueous HF solution, the formation of w =p 1− (9)
icrit
a regular orthogonal pattern of cylin-
drical macropores on a rear-illuminated The diameter of the macropores etched at
substrate. He initiated the pores by us- different current densities, i, is a linear
ing standard lithography to produce a function of the square root of i/icrit
predetermined homogeneous pattern of following Eq. (8). It was found that the wall
pits and subsequently developed the pores thickness could be up to a factor 10 smaller
by alkaline etching. Illumination with a or wider than the SCR width. The rate of
pore growth is a function only of the critical macropore arrays (Fig. 20). It has been
current density icrit (which is a function of shown recently that 100-µm wide pores
the HF concentration and temperature). with 2-µm wall thickness and 200-µm
The stable macropore formation ob- depth (Fig. 21) can be formed on highly re-
tained for an applied bias is sufficient sistive (2000–5000 # cm) n-type Si [132].
to generate the critical current density The demonstration that there is no restric-
(about 1 V). This understanding of the tion concerning the wall spacing opens
formation mechanism has allowed good the route to form vertical structure by
control of the geometrical parameters of (photo)electrochemistry.
Fig. 20 Cross-sectional SEM images and 45◦ array. Pore growth was induced by regular
bevel of n-type Si samples (Nd = 1015 cm−3 ) patterns of etchpits produced by standard
showing the predetermined patterns of lithography and subsequent alkaline etching
macropores. (a) orthogonal array, (b) hexagonal (inset upper right) (from Ref. 72).
Fig. 21 Cross-sectional SEM

image of macropores obtained
from a prepatterned n-type Si
(2000–5000 # cm) in 3% HF
electrolyte with backside
10 µm
illumination (from Ref. 132).
A recent systematic study of macropore They suggest that the contrast between
formation performed on various doped n- aqueous and nonaqueous etching can be
type Si substrates with rear illumination, attributed to two factors, the competition
by Föll and coworkers [106] showed that a of OH− with F− for complexing Si, and
strong influence of the SCR on the average the kinetically slow dissolution of oxide
macropore density is indeed observed in (or hydroxide) species formed in aqueous
accordance with the Lehmann model [72] solutions.
(i.e. an increased anodic bias decreases the
density of pores), except for highly doped Computer Simulations These have been
Si. It was observed that an increasing attempted to obtain morphologies similar
anodic bias increases the pore density, in to those observed in porous Si. The models
contrast to the prediction. The pore growth proposed to explain porous Si formation
seems to be dominated by the chemical- are similar in spirit to those previously
transfer rate and most likely calls for a used to understand the complex crystal-
chemical passivation mechanism of the growth phenomena. Although porous Si
macropore walls. formation is a dissolution process, similar-
2. p-type Si: Formation of macroporous ity is found with the growth phenomenon.
Si is surprising at first sight because in The models fall in two categories. The first
contrast to n-type Si, p-type electrodes type corresponds to the popular diffusion-
are under forward bias conditions. An ex- limited aggregation (DLA) model, which is
tended SCR is therefore not expected for a based on the diffusion of an electrostatic
p-type electrode in the anodic regime. To species such as hole (electron) to (from)
explain macropore formation process on the interface [133]. The second type is the
p-type Si, Lehmann and Rönnebeck [102] Mullins-Sekerta instability model [134]. It
put forward the key role of diffusion consists of analyzing the linear instability
across the space charge region. Especially, of the Si/electrolyte interface by taking into
the field enhancement and the associated account both the transport of holes in the
narrowing of the SCR at the pore tips semiconductor and ions in the electrolyte
are assumed to be responsible for the together with the surface tension of Si. The
pore tip dissolution and pore wall pas- DLA model has first been used by Smith
sivation. This model predicts macropore and coworkers to explain pore formation
formation up to Na = 1017 cm−3 doping as a result of the diffusion of holes to the
level, but fails to explain why macroporous interface [135, 136]. Yan and Hu adopted a
formation on moderately doped p-type different approach by modeling the inter-
Si (1–0.2 # cm) is only observed when facial dynamics governing the formation
HF-contained organic protophilic solvents of porous Si using a two-dimensional two-
(DMF and DMSO) are used [80]. component resistor network model [137].
Another model proposed by Kohl The ratio of the resistances of the two
and coworkers [77, 78] is based upon networks is used as a control parameter
the strain – induced preferential etching to simulate the advance of the interface
described earlier. The model accounts for according to stochastic dynamics, which
the formation of macropores and highly involves local current. John and Singh
branched micropores when the silicon developed a diffusion-induced nucleation
is rendered porous in either nonaque- model for the formation of porous Si
ous or aqueous HF solutions, respectively. based on two primary processes [138]. The
diffusion of holes from the bulk to the distribution of structure sizes (from mi-
surface is controlled mainly by (1) the SCR croporous to macroporous) as the layer
width (w) and (2) the drift-diffusion length thickened, and the dependence of the
(l) of holes inside the lattice. The theoreti- pore sizes on the resistivity of the starting
cal models permit obtaining the porous Si material.
morphology, which looks very similar to
what is observed in practice but generally 3.2.4
lack physical substantiation and serve as Porous Semiconductors: A Review
an illustrative facility rather than an ana-
3.2.4.1 PEC and EC Etching (Table 2)
lytical tool relevant for both scientific and
The first observation of intentional local-
practical applications.
ized corrosion on semiconductors, other
The concept of the instability model
than Si, was purposely performed to corru-
used to explain porous Si formation was
gate the semiconductor surface, decrease
first introduced by Kang and Jorné [139,
the reflectivity, improve the optoelectronic
140]. The model considers the pore nucle-
properties, and consequently increase the
ation at the Si surface as a mathematical
performance of photoelectrochemical so-
problem of the instability of a planar in-
lar cells.
terface toward small perturbations. The
interface can be destabilized for an opti-
mal deformation wavelength as a result 3.2.4.1.1 Group II–VI Semiconductors
of a competition between the destabilizing Beside the porous Si, II–VI semicon-
effect of hole diffusion and the stabiliz- ductors are those for which the origin
ing ones because of ion diffusion in the of deep microrelief (micro-etchpit) forma-
electrolyte and surface tension of Si. The tion was studied first. Long ago, n-type
optimal wavelength is expected to give an CdS [144], CdSe [145, 146, 47], n- and
order of magnitude of the interpore spac- p-type CdTe [119, 120, 148, 149], n-type
ing. The analytical and numerical stability CdSe0.65 Te0.35 [150], ZnSe [121], and Cd-
HgTe [151] have been found to undergo
analysis of the Si/electrolyte interface for
an extreme surface roughening under EC
PEC-etching of n-type Si was performed
and PEC etching. The II–VI electrodes
by Valance [141, 142]. This model allowed
were exposed in aqua regia to a reverse
expression of the dissolution speed and
bias of over 1 V/SCE accompanied by high-
derivation of the scaling laws for interpore
intensity illumination. The duration of the
spacing as a function of the doping level
PEC-etching generally did not exceed 5 s.
of Si and applied potential. Another im-
The anodic dissolution valency is two,
portant result obtained by Kang and Jorné
which implies the formation of elemental
was the relationship between the intensity
tellurium forming an insulating layer that
of the rear illumination and the pore di-
hampers further transfer of photogener-
ameter: the higher the intensity, the larger ated holes.
the photocurrent per pore and the larger
the pore diameter. Chazalviel and cowork- CdTe + 2h+ −−−→ Cd(aq) 2+ + Te0 (10)
ers recently addressed the linear stability
analysis of the interface to the case of After removal of the insulating (photo)-
p-type Si [13, 143]. It allowed an under- corrosion products obtained by dissolution
standing of the observed changes in the in a suitable medium, the surface contains
210
Tab. 2 PEC-etched semiconductors.
Semi- Type Band Conduction Electrolyte Potential Morphology

conductor gap type (V/SCE)/current
size (eV) density
CdS [144] d 2.42 n, 2 # cm HNO3 , HCl, +1 V, light, Etchpits λ laser (nm)

(4.5 × 1010 cm−2 H2 O (1/4/20) 3–4.5 s, 12 mA 2 × 109 cm−2 457.9
donors) 2 × 108 cm−2 514
CdSe [146, d 1.72 n, 10 # cm HNO3 , HCl, 1-µm etchpits
3 Semiconductor Nanostructures
+1 V, light,
147] H2 O (1/4/20) 3–4.5 s
1 M Na2 SO3
0.2 M FeSO4 +
0.1 M H2 SO4
1016 –1017 cm 0.2 µm etchpits
(1 # cm)
CdTe (111) d 1.44 n, 1000 # cm HNO3 , HCl, 1 µm triangular etchpits
[119, 120, H2 O (1/4/20)
148, 149]
1016 –1017 cm−3 Idem 0.15 µm
(5–0.05 # cm)
ZnSe [121] d 2.7 n, 4 × 1017 cm−3 HNO3 , HCl, +1 V, light Etchpits λ laser (nm)
H2 O (1/4/20) 1010 cm−2 457.9
5 × 108 cm−2 472.7
0 488
InP (149) i 1.35 n, 1 M HCl 10 mA cm−2 , Elongated etchpits µm. Groove
1–2 × 1018 cm−3 light, structure along (011) axis
1–120 min.
WSe2 [158] i 1.22 n 0.4–1 M +1 V, light Microetchpits on van der Waals
d 1.37 HClO4 planes on initiated punctures
(a) (b)
Fig. 22 Plan view SEM images of photoetched (111) n-type CdSe. Influence of
the doping concentration on the density of the micro-etchpits.
(a) Nd = 1016 cm−3 ; (b) Nd = 1017 cm−3 (from Refs. 145, 146).
a pattern of highly dense microetchpits difference with the II–VI compounds is

(>109 cm−2 ) whose density and shape that the corrosion products do not lead to a
vary with the doping density and crys- passivation layer and a long duration PEC
tallographic orientation of the material etching renders the InP surface porous.
(Figs. 22a and 22b). The size of the mi- This will be discussed in the next section.
croetchpits observed under scanning elec-
tron microscopy (SEM) is on the order of
3.2.4.1.3 Layered Semiconductors Laye-
a few hundred nanometers. On (110) and
red semiconductors such as InSe [155],
(111) surfaces, the pits have parallelogram
WX2 (X = S, Se) [156–158] and SnS2
and triangular shapes, respectively. The
PEC-etched (texturized) surfaces have not (X = S, Se) [158] have been photoelectro-
been described in terms of the formation chemically etched in a manner similar to
of a porous layer, especially as the depth the II–VI semiconductors. A review of
of the pores is not known. However, there this work is given in Ref. 158. The shape
is a striking similarity with some of the of the etchpits depended strongly of the
macroporous silicon morphologies. It has crystallographic orientation of the etched
been suggested that the texturized surface surface. Once again, the purpose of this
arises from the preferential etching of sur- work was to improve the photoelectro-
face defects in the vicinity of dopants [121]. chemical characteristics of electrodes. In
The characteristic size (200 nm–1 µm) of the case of the van der Waals surface,
the observed structures and its decrease known to be free of surface states (dan-
with increasing sample doping were given gling bonds), no photoetchpits are formed
as arguments in favor of this model. upon PEC etching. On the contrary, the
surface perpendicular to the van der Waals
3.2.4.1.2 Group III–V Semiconductors surface that has a significant number of
General trends observed with II–VI dangling bonds is strongly corroded. Mi-
semiconductors are also observed on crometer photo-etchpits can be initiated
III–V semiconductors. Formation of on WSe2 van der Waals planes after cre-
etchpits was observed on n-InP in ating defects by microscopic punctures at
the presence of HCl solution, under selected places [159] (Fig. 23). No porous
illumination [47, 152–154]. The main layer has been observed.
Fig. 23 Plan view SEM image of photo-etchpits initiated at the van der
Waals surface of WS2 , after creating defects by microscopic puncture at
selected places and upon PEC etching (adapted from Ref. 41).
3.2.4.2 Porous-etching (Table 3) illumination. Depending on the experi-

Following the work on luminescent porous mental conditions (n- or p-type, in the dark
Si, a number of studies have been un- or under illumination), the reported struc-
dertaken to render other semiconduc- ture sizes generally range from 10 nm to
tors nanoporous. The motivation of such 1 µm.
studies was based on the fact that if they In the case of a-SiH and Si1−x Cx ,
can exhibit tunable luminescence in a sim- the thickness of the nanoporous layer is
ilar way to nanoporous Si, then common limited by the formation of macropores,
features or differences might reveal the an instability of the growth front attributed
mechanisms involved. Studies concern in- to the high resistivity of the starting
direct band gap semiconductors such as material [160].
those of column IV (Ge, Si1−x Gex , SiC)
1. p-type Ge: Porous Ge [161–166]
and GaP, as well as direct band gap III–V
formed in 10% ethanolic HF on
alloys (GaAs and InP) and II–VI com-
6–10 # cm (100) p-type Ge under
pounds (CdTe, Cd0.95 Zn0.05 Te, CdTe, and
galvanostatic conditions (50 mA cm−2 )
ZnTe).
most probably has a similar nanostruc-
ture (∼2 nm) to porous Si obtained on
3.2.4.2.1 Porous Group-IV Semiconduc- low-doped Si as suggested by extended
tors, Other Than Si Various column-IV X-ray absorption fine structure (EX-
semiconductors have been made mi- AFS) measurements [164]. One main
croporous using recipes inspired from difference is that the porous Ge sur-
those used in porous silicon: p-type face is much rougher than the porous
Ge (0.66 eV), Six−1 Gex (x = 0.04–0.4) al- Si surface. A 10-min. porous-etching
loys, n- and p-type SiC (2.8–3.2 eV), p-type produces a roughness greater than
hydrogenated amorphous Si (a-SiH), and 5 µm, which is similar to that ob-
Si1−x Cx alloys have been made porous in tained from porous-etching Si for an
ethanolic HF, either in the dark or under hour under similar conditions. The
Tab. 3 Porous semiconductors other than Si
Semi- Type Band gap Conduction type Electrolyte Potential (V/SCE) or Morphology
conductor size (eV) current density
Ge [161–166] i 0.66 p, 6–10 # cm (10%) HF 50 mA cm−2 , Porous Ge, features

10 min less than 50 nm
diameter
Si0.8 Ge0.2 i p, 1019 cm−3 HF (12.5–24%)/ 10–40 mA cm−2 Mesoporous pore
[166–175] 1017 cm−3 C2 H5 OH/H2 O diameter 10 nm,
SiGe skeleton 5 nm
diameter, porosity
70%
6H-SiC i 3 p, Diluted HF 50 mA cm−2 Porous +
[176–195] 2.2 × 1018 cm−3 roughened
morphology and
shallow depression
(<150 nm).
n, 2.5 M NH4 F 2.5 mA cm−2 1–10 nm Interpore
1–3 × 1018 cm−3 0.5–1 h + light spacing, Crystallite
morphology
undetermined i.e.
spongelike
crystallite network,
or wire-like
(continued overleaf)
3.2 Macroporous Microstructures Including Silicon
213
214
Tab. 3 (continued)
Semi- Type Band gap Conduction type Electrolyte Potential (V/SCE) or Morphology
3 Semiconductor Nanostructures
conductor size (eV) current density
20–25 V dark Porous 50 nm

current mode
2.5 mA cm−2 , Platelike 0.5–1 µm
0.5–1 h, light fiber thickness
365 nm or
+1.5–2 V
GaAs [197–212] d 1.424 n, 4 × 1016 cm−3 0.1 M HCl 3V, 5 min Inhomogeneous:
6 V, 5 min Pitting including
GaAs needles sub
µm, top covered by
AsCl3 crystallite
layer and 20 nm fine
features
p, 9 × 1017 cm−3 0.1 M HCl + 1V Pitting
NH4 H2 PO4
GaP [211–227] i 2.26 n, 5 × 1017 cm−3 (1/1) 40% 20 mA cm−2 , Porous layer 3 µm
HF/MeOH 10 min thick
0.5 M H2 SO4 −10 V, 2 h 100 nm, 50 µm
thick, distance
between pits
10–30 µm
InP [152–154, i 1.35 n, 1 M HCl 10 mA cm−2 , light, Macropores
228–235] 2.5 × 1018 cm−3 1–120 min. 1–6 µm, Groove
structure along
(011) axis
1 M HCl or 2–5 V
1 M HBr
n, (111) 15% HCl 15 mA cm−2 50 nm size
2–3 × 1018 cm−3 2–5 min nanostructure
GaN [237] 3.36 5 × 1017 cm−3 0.5 M H2 SO4 Cycling between −1 40–120 nm size
+10% ethanol and −2 V structure
CdTe d 1.44 p, 1015 cm−3 0.5 M H2 SO4 3 mA cm−2 1.2 V 0.4–0.8 µm
Cd0.95 Zn0.05 Te macropores
[239, 240]
ZnTe [241–242] d 2.23 p, HNO3 : HCl : H2 O 300 mA cm−2 (0.1–1 µm)
1014 –1015 cm−3 (1 : 4 : 20) 5–25 sec. macropores.
Morphology
depends on
crystallographic
orientation
+1 V 25 nm length
(100 mA cm−2 ), columnar structures
30 s
TiO2 (001) d 3 M 0.5 M H2 SO4 1–1.5 V + light Macropores 0.2 µm
rutile [246–249] (16 C) diameter
1 M HClO4 idem (45 C) Macropores 0.5 µm
diameter
3.2 Macroporous Microstructures Including Silicon
215
room-temperature photoluminescence was increased by a factor of 3

centered at 1.17 eV is very similar to for porous Si0.95 Ge0.05 and re-
the photoluminescence from porous Si, duced by an order of magnitude
but weaker in magnitude by a factor of for Si0.8 Ge0.2 compared to porous
about 10 [163, 164]. Porous Ge made Si [167, 169–174]. The photolumi-
on 0.2–0.5 # cm p-type Ge does not nescence was interpreted in terms
exhibit visible luminescence [166]. of quantum size effects
2. p-type Si-Ge alloys [166–175]: Study – Low-doped Porous Si0.85 Ge0.15
of these compounds was undertaken (1–15 # Cm): The morphology of
because it was expected that the Si porous SiGe that originated from
luminescence wavelength could be con- low-doped substrate could unfor-
trolled by changing the Ge content tunately not be analyzed by TEM
(0.9 eV band gap) in the material. because of the low density of the
porous layers produced. Photolumi-
– Heavily doped (Na = 1017 − nescence spectra were similar to
19 −3
10 cm ): When porous Si−Ge those observed for highly doped
was made by stain-etching, no porous samples [175].
significant spectral shift was ob-
served in the photoluminescence 3. n- and p-type SiC: Porous SiC, a promis-
spectrum of samples with the ing candidate for blue-light emission
nominal 0.04 < x < 0.41 compo- devices, is, beside porous Si, the most-
sition, whereas the signal inten- studied porous semiconductor. The
sity became weaker with increased band gap energy of SiC is considerably
x [166]. When porous Si0.95 Ge0.05 higher than that of Si and varies with
and Si0.8 Ge0.2 were made electro- the polytypes modification from 2.3,
chemically in 12.5–24% HF ethano- 3.0, and 3.2 eV for the 3C, 6H, and 4H
lic solution and current density main polytypes, respectively. The three
between 10 and 40 mA cm−2 , a polytypes can be made porous in HF
column-like pore structure was ob- and NH4 F electrolytes [176–195]. For
tained, closely analogous to that n-type 6H-SiC (Nd = 3 × 1018 cm−3 )
observed for mesoporous p + -type porous photoetching was done galvano-
porous Si of comparable 50% poro- statically (2–40 mA cm−2 ) under ultra
sity [169]. Typical pore dimensions violet (UV) photoassistance (365 nm,
were 10 nm, and the remaining 0.1–0.2 W) [177–178, 181] or in the
SiGe skeleton had a diameter of dark under tunnelling breakdown
5 nm. The existence of a local sub- (20–25 V) [179]. For Al-doped p-type
structure with lower dimensions (Na = 2–5 × 1018 cm−3 ), it was per-
might be evoked to explain the formed galvanostatically (5–150 mA
visible photoluminescence within cm−2 ) in the dark [178, 184]. p-Type
the quantum confinement model. porous SiC showed great similarity to
The maximum photoluminescence p + -porous Si sponge-like crystalline
band exhibited by porous Si-Ge is network morphology (Fig. 24) [178,
shifted by 0.2 eV to a lower energy 183]. The interpore spacing was found
compared to porous Si whereas the to be in the range of 1–10 nm with
intensity of the photoluminescence an average size of 6 nm. Low-doped
Fig. 24 SEM images of porous

(a) (b)
n-type 6H-SiC. (a) after
porous-etching at +20–25 V
(plan view); (b) after porous
photoetching (cross-sectional
view) (from Ref. 179).
porous p-type SiC (1.5 # cm) was very different morphology were observed [186].
fragile and looked very dendritic. SiC The 3C, 6H, and 4H n-type porous SiC
structures down to 10 nm could be exhibited identical luminescent emission
seen [186]. Dramatic difference in pore spectra with a broad peak around 2.6 eV
wall size was observed between the (except for an additional broad shoulder
structure of porous n-type materi- at 1.9–2.1 eV for the 3C polytype) demon-
als produced under light (10–30 nm strating a clear impedance of the band gap
pore size, 10–100 nm pore walls) and energy of a particular SiC polytype [182].
in the dark (0.1–1 µm pore walls) The photoluminescence of porous SiC
(Fig. 24) [184]. with different microstructures was found
to become enhanced as the microporous
Pores in 4H-SiC first propagate nearly
SiC changed from fibrous to porous den-
parallel to the substrate surface and gradu-
dritic. A weak blue electroluminescence
ally change direction toward the direction
was also observed on porous n-type 6H-SiC
of the c-axis [195]. Similar anisotropic
porous-etching was found in some 6H- (0.1 # cm) [180, 183].
porous SiC [189]. The electrical properties measured with
Intense blue–green luminescence (two the use of a metal-point probe contact
order of magnitude stronger than that ex- showed for porous SiC obtained in the
hibited by crystalline SiC) were observed dark, current–voltage characteristics qual-
for porous n-type 6H-SiC produced by itatively similar to those of bulk SiC,
porous photoetching in HF-based elec- whereas porous SiC obtained under light
trolyte [176, 177, 181, 182]. The photolumi- was insulating. This was interpreted in
nescence peaked at wavelengths (2.6 eV) terms of Fermi level pinning at the surface
lower than those corresponding to inter- of the 50 nm thick fibers corresponding to
band transition in crystalline SiC, thus the SCR thickness for highly doped SiC.
showing the possibility of a related de- In the case of SiC, the reaction with HF is
fect rather than quantum confinement known to produce a carbon-enriched phase
mechanism of light emission from porous at the surface, which may provoke a high
SiC [181, 182, 186, 190]. Similarity be- density of surface states. A model based
tween photoluminescence properties of n- on the Fermi level pinning was evoked to
and p-type porous SiC layers with highly explain porous SiC formation [182].
3.2.4.2.2 Porous III–V Semiconductors separate submicron pits whose number

Porous III–V semiconductors are of great roughly scales with dislocation density of
interest because of possible applications the original substrate. Pore formation can
in optoelectronics and photonics. Study be triggered at surface defects created on
of the luminescence from porous direct purpose.
band gap semiconductors such as GaAs n-Type GaAs(1016 cm−3 ) was made me-
and InP presents an additional interest soporous in HCl- and Cl− -containing so-
because quantum confinement effects are lutions [200, 201], when the electrode was
well established in two-, one-, and zero polarized at a potential larger than 3 V,
dimensional electron systems prepared on which is the critical potential [pore forma-
crystalline III–V materials. Thus it would tion potential, (PFP)] at which breakdown
be possible to compare the light-emission of the Schottky barrier occurs. Pore forma-
properties between structures obtained by tion occurs in two steps. In the first step,
porous-etching and those produced in the crystallographic etching occurs at the
well-characterized lithographically defined (111) GaAs exposed planes. Then during
structures directly. GaP was studied for a second phase, the electrochemical at-
comparison because it is an indirect tack spreads out from the initial etchpits
band gap. and leaves distinct patches behind. Within
Variations in the conditions of forma- the patches a porous structure was found
tion of the III–V porous semiconduc- (Fig. 25). It has been suggested that the
tors and in the obtained morphologies initial pits develop at the defects or dis-
have been observed [197–238]. The III–V locations and radial pore growth occurs
semiconductors were rendered porous in through enhancement of the electric field
HF and HCl solutions but other acidic at the space charge boundary. Polarization
solutions such as H2 SO4 permitted ob- for a long time at a potential above PFP
taining porous GaAs and GaP. n- and leads to a complete coverage of the sur-
p-types GaAs were also made porous face by a black porous layer of 60–80%
in KOH and NH4 OH alkaline solutions. porosity. Typical structure size is about
Generally the porous-etching was per- 100 nm, but generally few structural data
formed in the dark, under breakdown are available to reveal the dimensional-
conditions (either by avalanche or tunnel- ity of the porous skeleton. For p-type,
ing) of the semiconductor-depletion layer. porous GaAs is only obtained in phosphate
Porous photoetching was also used. The buffer solution because a stable oxide film
majority of these III–V substrates uti- can be formed. The presence of Cl− in
lized were n-type heavily doped (Nd = the solution was also necessary [201]. The
5.1017 − 1.1018 cm−3 ). pore formation mechanism on n- and
1. Mesoporous: Porous layers on GaAs p-type GaAs illustrates the asymmetry
and GaP surfaces with (100) orientation that exists in the conduction properties
obtained under breakdown conditions of the semiconductor/electrolyte junction
after porous-etching in HCl and H2 SO4 under anodic polarization. For n-type, the
solutions show a network shaped like current-limiting force is the SCR and
cellular structures. It is noteworthy that porous GaAs occurs in Cl− -contained elec-
for GaP this additional feature is actually trolyte, whereas for the p-type, formation
buried under the etched surface; the of a passivation oxide layer is necessary to
top layer still consisting of distinct and establish the current limitation.
Fig. 25 Plan view SEM image

of one of the etchpit formed on
(100) n-type GaAs after
porous-etching in 0.1 M HCl
(from Ref. 200).
20 µm
Fig. 26 Plan view SEM image of porous n-type GaP after

porous-etching in the dark (+10 V vs SCE, 2 h, 0.5 M H2 SO4 )
(from Ref. 218).
Porous-etching of (001)-oriented Nd = electrode. The porosity of the layer was

5.1017 cm−3 n-type GaP crystals was per- 25%. At an early stage of anodization, the
formed in 0.5 M H2 SO4 at +10 V [214, porous domains are circular and have al-
215]. After porous-etching, the surface most the same diameter. The domains
looked flat though slightly spitted, and re- expand until the pore fronts from neigh-
flected light as well as a polished wafer. boring domains meet, at which time the
To reveal the porous structure below the lateral etching stops [215, 220] (Fig. 27).
flat surface, the electrode was dipped in When the applied voltage was temporally
an aqueous bromide solution to remove increased from 5 to 15 V during anodiza-
the top 0.5 µm of the GaP electrode. A tion, formation of annular traceries around
pattern of porous domains was revealed each etchpit was observed (Fig. 28). This
(Fig. 26). The pore diameter is similar strongly suggests that the etching proceeds
to the pore wall thickness and equals radially at the same rate in all directions
90–120 nm. It was found that the nucleus away from the primary pores. Following
(from which pores radiate to the thin walls a preliminary high-energy particle treat-
of the domain) of each porous domain is ment of GaP wafer (Kr+ preimplanted),
located exactly below the site at which a pit prior to porous-etching, to create a large
emerges at the flat surface of the anodized number of surface states, the observed
a Fig. 27 Schematic illustration of

various stages of GaP porous-etching:
(a) initiation of pores at defect sites;
b (b) continued growth of the initial
pores; (c) appearance of the first
branches or secondary pores; (d) further
c branching, separate porous domains are
formed defined by pore fronts which
have a more or less hemispherical
geometry. The pore walls are 100-nm
100 nm thick; and (e) porous domains meet,
d
creating 100-nm thick nanoporous
walls. A porous layer now covers the
entire sample.
100 nm
Fig. 28 SEM image of the top surface of GaP sample anodized

under temporary variations of the applied voltage from +5 to
+15 V (from Ref. 220).
porous morphology resembled that which thickness of 50–100 nm, sufficient to leave
was previously obtained after a long etch- a GaP skeleton completely free of carriers,
ing time and removal of the surface as a consequence of the SCR effect.
layer [220]. The high number of surface Porous-etching of InP done in the dark
defects created by ion implantation pro- under tunneling breakdown conditions
vided a high density of etchpits, and thus a leads to the formation of small pores
high density of primary pores with a wall with a diameter around 50 nm on (111)-
and (100)-oriented heavily doped n-type in chlorine-contained methanol. Tunnels

substrates. The pore morphology depends were formed when the Ga (111)A face was
strongly on the nature of the halogen acid in contact with the electrolyte and the elec-
present in the electrolyte. Blue, colorless, trode polarized at 120 V (20 mA cm−2 ).
and gray layers were observed when the an- When the P (111)B face was in contact
odization was performed in 1 M HCl, HBr, with the electrolyte, complete dissolution
and HF, respectively [233]. The pore diam- of the surface occurred by chemical dis-
eter was larger in HBr than in HCl. In HF, solution [213]. The results were compared
microscopic inhomogeneity was present, with the channels observed by Theunis-
with the presence of opaque regions sug- sen on Si. Formation of the tunnels
gesting a certain amount of occluded was interpreted in terms of a localized
porosity. An enrichment of phosphorous electrochemical attack resulting from the
within the porous structure obtained in avalanche breakdown of the SCR. It was
HCl solution indicates a selective disso- shown that in GaP, the tunnels were not
lution of indium. Porous InP was also related to dislocations.
obtained when the n-type electrode was Porous-etching in an HF/ethanol solu-
continuously biased between 0.1 and 2 (3, tion of a (111)A GaP surface leads to a
4) V. The pore diameter increased from 40 structure consisting of triangular prism-
to 120 nm when the final bias varied from shaped columns. In H2 SO4 , freestanding
2 to 4 V [233]. membranes of porous GaP (111) were ob-
2. Substrate Orientation: Pore morpho- tained. The 2 µm thick membranes com-
logy depends on the doping density and prised arrays of isolated pores stretching
crystallographic orientation of the surface. perpendicular to the surface of the sample
For example, it was shown for porous GaAs with average pore and skeleton thickness
obtained in 25 M HF that the primary of about 50 nm (30 min, 50 mA cm−2 in
pores having triangular or hexagonal cross- the dark) [223].
section grow in the (111)A direction Porous-etching of a (111)A InP surface
(corresponding to the direction from Ga in HCl/H2 O solution leads to a top layer
plane to As plane) [204, 205], which is with a pillar structure characterized by
in contrast to porous Si where the pores isolated columns stretching perpendicu-
grow in the (100) direction. In both GaAs larly to the initial surface [228]. The shape
and Si, the pore tip is composed of the and size of pores can be automatically
slowest reacting planes in the structure. fixed when porous-etching was combined
This trend is general for all porous III–V with conventional electron-beam lithog-
semiconductors. raphy with an appropriate mask pattern.
(111)-oriented surfaces: First, n-type Triangular pores with fairly good size and
GaAs was made porous in the dark in KOH uniform and ordered InP triangular ver-
and H2 SO4 solutions. The porous layer tical pillar structures were obtained [229].
contained ‘‘tunnels’’ of diameters ranging The average size of the pillar structures
from 0.5 to 5 µm and a pore density of was reduced from 60–50 to 30–40 nm
about 108 cm−2 [197, 198], which propa- when the doping density of the sub-
gate only in the (111)A direction [199]. strate increased from Nd = 1018 cm−3 to
The first report on porous-etching of 1019 cm−3 (Fig. 29) [230]. Quantum-wire
GaP was related to the formation of and quantum-box structures with low-size
high-density networks of tunnels formed fluctuations were obtained.
Fig. 29 Plan view SEM image

<211>
of the porous (111)A n-type InP
<111> substrate with SiO2 mask (from
Ref. 230).
<112> <121>
60 nm
(001)-oriented surfaces: The direction of in Fig. 26. (001)-oriented nanometer-sized

the pores depends strongly on the ap- straight pores, free of any branching with
plied bias. On (001) InP (1 M HCl, Nd = 250 nm pore diameter, 16 nm wall thick-
1018 cm−3 ), the pore growth changed the ness, and 80 µm pore length were obtained
direction from (111) when the applied on a 3 min treatment (Fig. 30) [231]. The
bias was below 3 V to (100) when it formation of (001)-oriented pores was
was over 4 V [231]. Below 3 V, the pores interpreted by the combination of an en-
were aligned along the direction having hanced electric field at the pore tip and a
a tilt angle of 63◦ with respect to the short hole lifetime of 0.1 µs in InP. The
(100) plane resulting in a tree-like cross- pores exhibited square shapes defined by
sectional structure. At potentials above four crystalline (001) and (011) facets when
4 V the branching disappeared and fairly a new solution (1 M HCl + HNO3 ) was
regular straight pores basically running to- used [232].
ward the (100) direction were observed. 3. Macroporous obtained under illumina-
When the porous-etching is done at high tion: Ordered honeycomb hollow arrays
over-potential, above 7 V, the surface mor- are formed on n-type GaAs (001) sub-
phology of porous InP looks similar to strates highly doped (2 × 10−3 # cm) in
that of the porous n-type GaP shown 29 wt% NH4 OH by porous photoetching
(a) Vs = 2 V, ta = 3 min
Fig. 30 Cross-sectional SEM

image of porous n-type InP after
2 µm porous-etching at +5 V during
1 min, in 1 M HCl (from
(b) Vs = 5 V, ta = 1 min
Ref. 231).
with a bias voltage of 12 V. The hexagonal ratio structures (grooves) of average pe-
hollows exhibited 19 nm average diameter riod of 1 µm are developed (Fig. 31).
with a standard distribution of 200 nm. The reflectance of such highly corru-
Self-organization formation of the hollows gated surfaces drops from 40 to 4% after
in porous GaAs was explained by repulsive photoetching at 20 mA cm−2 current den-
interaction between the hollows [210]. sity during 60 min. in 1.2 M HCl [154].
The first work on porous InP, [(001)- This effect can, in principle, be very useful
oriented, Nd = 1018 cm−3 ], was accom- in the development of new antireflective
plished with anodization in HCl elec- surfaces for solar cells.
trolyte under illumination, for producing Porous photoetching of n-type GaP in
surface relief features in optoelectronic H2 SO4 /H2 O/H2 O2 leads to the formation
of macroporous GaP (random networks
devices [152]. The pore location was con-
of single crystalline GaP with pore sizes
trolled with standard photolithography,
of 150 nm and porosity of 35 to 50% for
and well-defined macroporous structures
porous etched and porous photoetched, re-
with a pore aspect ratio up to 250 were
spectively), which appears to be the most
produced by photoelectrochemistry in an
strongly scattering material in visible light
analogous manner to the macroporous pat- reported to date and localization effects are
terned n-Si. When the initial surface was anticipated [224, 225].
mirror polished, the surface microstruc- 4. Photoluminescence: Porous GaAs
ture varied from 50 nm to several µm obtained in HCl exhibits visible region
depending on the porous-etching duration luminescence (around 540–570 and
and was strongly affected by the crystallog- 850–870 nm) (Fig. 32). The green and
raphy [47]. Using a homogeneous white infrared bands were interpreted by
light beam, Decker and coworkers [153] Schmuki and coworkers as being caused
found that microetchpits develop prefer- by a quantum confinement effect in GaAs
entially along the (011) direction. Upon microcrystallites and nanocrystallites,
longer etching time, high depth/width respectively [201, 207–209]; whereas
Fig. 31 SEM image of n-type InP after PEC-etching in 1.2 M HCl

during 9.2 min (from Ref. 154).
120000 80000
(a) Porous GaAS (b)
500× 8× 70000
100000
1 clean 60000
3 GaAs
80000
50000
Intensity
Intensity
[a.u.]
[a.u.]
60000 40000
1
2 2 30000
40000
3 20000
20000
10000
0 0
500 600 700 800 900 500 600 700 800 900
Wavelength Wavelength
[nm] [nm]
Fig. 32 Photoluminescence spectra (room −0.5 to 6 V (SCE); (b) highest ‘‘green’’
temperature) of porous GaAs obtained in 0.1 M photoluminescence intensity observed from a
HCl: (a) for three different surface locations on specific location on a porous GaAs sample
porous GaAs formed by a potential sweep from formed by a potential step (8 V/SCE, 5 min).
Finnie and coworkers identified the Contradictory results have been pub-
source of the green luminescence to be lished on the photoluminescence of
arsenic oxide microcrystals formed during porous InP. Some authors reported blue-
porous-etching [211, 212] and that of the shifted photoluminescence peaks lying
infrared band to the scattered excitation above the band gap energy interpreted
radiation, exciting luminescence from the in terms of quantum confinement [230].
relatively unperturbed outer regions of the Hamamatsu and coworkers reported that
etchpit [211]. contrary to expectation of blue-shifted
Photoluminescence spectra of porous emission caused by quantum confinement
GaP obtained in HF solution exhibits three at pore walls, porous (001) InP samples
strong sharp emission lines in the orange exhibited an intense red-shifted photolu-
spectral region (2.1–2.3 eV, 580 nm) and minescence peak [231]. This was explained
a weak emission between 300 and 700 nm by the formation of a set of well-defined
with maxima in the blue (2.4 eV) and new surface state levels on anodized pore
UV (3.2 eV) spectral regions [216]. The wall surfaces.
sharp lines are characteristic of bulk GaP. 5. GaN: This semiconductor was made
The blue and UV emission has been porous by continuous cycling of potential
explained by Anedda and coworkers [216] between −1 and 2 V in 0.5 M H2 SO4 +
on the basis of charge-carrier confinement 10% ethanol [237, 238].
in crystalline quantum wires of about
25-nm diameter. However, similar results 3.2.4.2.3 Porous II–VI Semiconductors:
observed by Kelly and coworkers when Low-doped (1015 cm−3 ) p-type CdTe [239],
porous GaP is obtained in H2 SO4 solution Cd0.95 Zn0.05 Te [239, 240], and ZnTe [241–
are interpreted, not in terms of quantum 243] were made porous by anodization in
confinement effect, but as caused by the dark.
gallium oxide (Ga2 O3 or GaOx ) formed p-Type CdTe and Cd0.95 Zn0.05 Te: Por-
on the surface of porous GaP [217]. ous-etching was performed in 0.5 M
Fig. 33 Cross-sectional SEM images of p-type Cd0.95 Zn0.05 Te

after porous-etching (Q = 100 C cm−2 ). (a) low magnification,
(b) high magnification (from Ref. 240).
H2 SO4 at +1.2 V (3 mA cm−2 ) for several and p-Cd0.95 Zn0.05 Te, which implies that
hours. A 100 µm thick layer consisting no blocking insulating layer is formed.
of highly porous network (0.5–0.8 µm Indeed when anodic dissolution is slow,
macropore diameter) was observed be- further oxidation of Te to HTeO2(aq) + is
neath a polished surface, which was left in- not rate-limiting [Eq. (12)].
tact [Fig. 33]. The reaction mechanism for p-Type ZnTe: Porous-etching of p-
the anodic dissolution of Cd0.95 Zn0.05 Te type polycrystalline ZnTe has been un-
is expected to be similar to that of n- dertaken to texturize the surface for
CdTe [244, 245]. The reaction proceeds in photovoltaic applications [241, 243]. The
two steps, the formation of elemental tel- goal was to increase the surface of
lurium followed by its oxidation: ZnTe while decreasing the reflectivity.
Low-doped (Na = 1014 − 1015 cm−3 ) large
CdTe + 2h+ −−−→ Cd(aq) 2+ + Te0 (11) grains polycrystalline ZnTe is rendered
porous in acidic solution composed of
Te0 + 4h+ + 2H2 O −−−→
HNO3 : HCl : H2 O (1 : 4 : 20) either poten-
HTeO2(aq) + + 3H(aq) + (12) tiostatically at 1 V vs SCE or galvanostat-
ically at 300 mA cm−2 constant current
Because elemental Te0 is an intermediate density. After a few seconds and subse-
species in the dissolution process, the sur- quent dissolution of the corrosion layer, a
face of the porous layer is at least partially layer of macroporous ZnTe with a colum-
covered with Te0 . This was confirmed by nar structure is formed. Diameter of the
XPS, a 14 nm thin Te porous layer lining pores (3–5 µm deep) varies from 0.1 to
the pore walls [239]. The results obtained 1 µm upon increase of porous-etching
under PEC etching with the formation of duration from 5 to 25 secs. The pore
an insulating Te0 layer are not necessary in wall thickness remains constant around
conflict with the obtaining of porous CdTe 100 nm. The effect of grain orientation
results in different pore shapes, for ex- The TiO2 crystals or sintered pellets were
ample, tubular or flat isosceles triangles first treated in hydrogen at 650 ◦ C for
or inverted pyramids with a rectangular 30 min. to obtain n-type semiconductiv-
base [243]. The grain boundaries remain ity. Porous photoetching of 0.2–2 # cm
largely undissolved. Enhanced photore- (100) rutile type TiO2 done potentio-
sponse (30%) of a ZnTe/aqueous poly- statically (1–1.5 V vs SCE) renders the
sulfide photoelectrochemical cell is ob- surface velvety black due to light con-
served for the porous surface because finement arising from the presence of
of decrease in reflectivity [242]. Further square shaped macropores (0.3–1-µm di-
porous-etching results in the formation ameter), whose morphology is similar to
of a needle-like structure with a triangular that of highly doped porous Si obtained
cross-section, the smallest structure reach- by porous photoetching (Fig. 12a) [246,
ing 25 nm typical size. Such porous layer 247]. Oxygen gas is evolved at the TiO2
exhibits blue-shifted luminescence com- surface. Increasing the SO4 2− concen-
pared to bulk ZnTe, which is attributed to tration and pH of the solution (from
quantum-confinement effect [241]. 0.45 to 1.65) results in the formation of
spare macropores on a much finer scale
3.2.4.2.4 Porous Oxides Among the ox- (<0.1 µm). In concentrated 9 M H2 SO4 ,
ides, only titanium dioxide, TiO2 , has been corrosion of TiO2 is inhibited. Addition
made porous. TiO2 is one of the few semi- of Co2+ in 0.5 M H2 SO4 quenches the
conducting materials highly stable against porous-etching, the photogenerated holes
photocorrosion, with the conduction band- oxidizing preferentially SO4 2− ions to
edge located so that photogenerated holes S2 O8 2− ions, whereas addition of Co2+
have sufficient energy to oxidize water in 1 M HClO4 does not prevent porous-
or the reduced member of redox species etching [248].
involved. It plays a leading role in the Porous photoetching was also done on
photoenergy conversion and acts as an effi- sintered pellets of TiO2 . The influence of
cient photocatalyst to decompose harmful applied potential on corrosion mechanism
organic compounds as in the TiO2 dye- is illustrated in Fig. 34. When anodization
sensitized photoelectrochemical cells. A is done at 0.2 V vs SCE, the grains are
highly developed surface morphology of selectively dissolved, whereas the grain
TiO2 is a crucial factor in these applica- boundaries are not dissolved creating a
tions and porous-etching is one technique ‘‘skeleton structure’’. By contrast under
that allows surface enlargement. Porous 1 V bias, the grain boundaries of the
photoetching of n-type TiO2 , under xenon pellet are selectively dissolved and a
lamp illumination, occurs in 0.05–0.5 M characteristic porous pattern appears on
H2 SO4 and 1 M HClO4 (perchloric acid). each grain surface [249].
Porous photoetching reaction was as-
sumed to be a two-hole process following 3.2.5
reaction (13) in H2 SO4 [246]: Photoelectrochemical Properties of Porous
Semiconductors
TiO2 + SO4 2− + 2h+ −−−→
Porous-etching has been applied in the de-
TiO·SO4 + 12 O2 (13) sign of photoelectrochemical devices. The
Fig. 34 SEM images of

macroporous TiO2 after porous (a)
photoetching of a sintered pellet
of TiO2 . (a) +0.2 V bias,
(b) +1 V bias (from Ref. 249).
1 µm
(b)
1 µm
porosity and pore wall size of porous semi- intensity close to 40% is also observed. In
conductors have different effects on the addition, the external quantum efficiency
photocurrent depending on the band gap vs light wavelength of the porous elec-
type (indirect or direct) of the crystalline trode is greatly enhanced compared to the
material. flat electrode, without any shift in the on-
Photoelectrochemical cells with indirect set of the photocurrent [27]. Similar trends
band gap porous semiconductors such as are observed for n-type porous GaP and
n-type Si (obtained under illumination), porous SiC [215, 250]. Photocurrent in-
SiC, and GaP (made porous in the reverse creases by a factor 100 after porous-etching
bias range) show spectacular results. Pres- of GaP, whereas the internal quantum ef-
ence of a porous layer on lightly and highly ficiency (corrected by the photon flux) of
doped Si strongly modifies the photocur- a 30 µm thick porous layer electrode is
rent potential curve of porous Si/HI (57%) spectacularly enhanced with respect to a
junctions, whose rising part is shifted flat electrode for light absorbed in the in-
to more negative potentials by −0.25 V direct band gap and for sub–band gap
compared to a flat (nonporous) electrode light [214, 219] (Fig. 36). Porous n-type
(Fig. 35). In the case of lightly doped Si, 6H-SiC electrode exhibits a similar be-
an increase in the plateau photocurrent havior. The enhancement of the quantum
4 Fig. 35 (Photo)current potential curves

b′
for n-type Si (Nd = 1015 cm−3 ) in 57%
b
3 HI, irradiance 50 mW cm−2 (tungsten
halogen lamp), surface area = 0.21 cm2 .
2 a) In the dark, b) under illumination,
before porous photoetching, a& ) in the
[mA]
1 dark, and b& ) under illumination after

i
a porous photoetching (Q = 4 C cm−2 )

0 a′ (from Ref. 27).
b′ a′
a,b
−1
−2
−1.0 −0.5 0.0 0.5 1.0
E vs SCE
[V]
1.0
II
0.8
0.6
jph/ef
I
0.4
0.2
0.0
2.0 2.5 3.0 3.5 4.0
hν /eV
Fig. 36 Quantum efficiency (jph /eφ) at 1 V (SCE) as function
of photon energy for a polished n-type (100) GaP electrode
(0 C cm−2 ) and for the same electrode after porous-etching
at +10 V (SCE) in the dark for the time needed to pass
16 C cm−2 . The indirect optical transition is at 2.24 eV and
the direct transition at 2.76 eV (from Ref. 218).
efficiency can be attributed to two ef- the band gap may be very large. Such states
fects related to the nature of the porous can give rise to sub–band gap photocur-
layer. First, light scattering gives rise to a rent as it has been observed for porous
more effective absorption of photons in the GaP and SiC. In addition, charge carriers
mesoporous layer. Second, the size of the in bulk band gap states, which normally
pore walls corresponds to the SCR width do not contribute, can also give rise to
in the photocurrent range, where the band photocurrent in a porous system.
bending is sufficient for effective compe- For materials with a direct optical tran-
tition of charge separation with surface sition (short absorption length) and a
recombination. Because of the large inter- low dopant density (wide SCR) or large
nal surface area of porous semiconductors, diffusion for minority carriers, porous-
the volume density of the interface states in etching is less interesting. This was
demonstrated on II–VI porous semicon- favors recombination, especially at sites

ductors with direct band gap. Anodization where the pore walls are the thinnest, re-
of p-type CdTe and Cd1−x Znx Te con- sulting in a decrease of the photocurrent.
tributes to changes in the photocurrent
spectra. At short anodization time (EC 3.2.6
etching) the photocurrent can increase, Potential Applications of Porous Si and
when surface roughening leads to a de- Other Porous Semiconductors
crease in reflectivity. The decrease in
photocurrent at longer anodization times Although more work is needed to
(porous-etching) is due in part to the for- characterize and hence understand
mation of the Te0 layer produced during pore-nucleation processes, the fabrication
methods, although not described in detail
the porous-etching. But this is not the only
here are now well understood and
effect [239]. Chemical removal of Te0 using
controlled. Porous Si has been extensively
potassium cyanide (KCN) solution causes
studied using spectroscopic techniques,
only partial recovery of the initial photocur-
and our understanding of the interesting
rent before pore formation. The second
and unusual optical properties of this
effect is recognized as being caused by the material is now well developed. Its recently
porous morphology, which appears to have established nanoscale structure and light-
an effect of its own, leading to an increase emitting properties have tremendously
in recombination. With p-type semicon- increased the interest of the scientific
ductors rendered porous in the forward community, and new applications in
bias range, the size of the porous features different fields of technology, summarized
corresponds to a SCR thickness and band in Table 4 together with the key properties
bending that are markedly lower (0.2 V) of porous Si, have been investigated.
than in the potential range, where surface The efficient visible room temperature
recombination is negligible for a flat sur- photoluminescence and electrolumines-
face (nonporous electrode) (>1 eV). The cence of porous Si [6, 263] provide ap-
weak band bending at the pore surfaces plication possibilities within, for example,
Tab. 4 Porous Si as a multidisciplinary material (from Ref. 273)
Scientific area Key properties Application example References
Optoelectronics Light emission Displays 251

Analytical chemistry Porosity Sensors 252
Optics Tunable refractive index Optical filter 253
Ultrasonics Low thermal conductivity Transducer 254
Surface science High surface area Catalyst 255
Microbiology Biocompatibility Bioreactor 256
Energy conversion Low reflectivity Solar cell 96–98
Astrophysics Cosmic abundance Dust chromophore 257
Microengineering Lithographic patterning Microsystems 258
Nuclear science Conversion to oxide Radiation-hard circuits 259
Electronics Dielectric properties Microwave circuits 260
Signal processing Dynamic process Process control 261
Medicine Resorbability Drug delivery 262
optoelectronics [251]. Porous Si is also an wavelength suitable for communication

interesting candidate for application in systems. In addition, Raman scattering
the fields of gas and biosensors tech- spectroscopy has revealed the existence
nologies [252, 266, 267]. This is due to of surface related vibrational mode in
the possibility of a wide range variation both GaAs and GaP. These modes of
of characteristics such as pore size, vol- the Frölich character exhibit a downward
ume porosity, specific surface area, and frequency shift with increasing porosity
thickness through different choices of pro- and/or dielectric constant of the surround-
cessing conditions like etching solution ing medium that hints at new possibilities
composition, current density, crystal orien- for optical phonon engineering in these
tation of the Si substrate, or temperature. materials [206, 221]. Porous GaN may be
Furthermore because porous-etching oc- used as a buffer layer for growth of
curs at the pore bottoms, that is, at the bulk and epitaxial GaN films with low
interface between the etching solution stress [238].
and the silicon substrate, it is possible
to fabricate multilayer structures, each 3.2.7
layer having a different porosity (com- Conclusion
monly called porous Si superlattices) that
can be designed, for example, to func- A large variety of porous semiconductors
tion as Bragg reflectors [17, 268, 269], can be produced chemically by stain-
Fabry-Pérot filters [18, 270–272], or optical etching or electrochemically, whose struc-
waveguides [253, 274]. The low reflectivity
tural, optical, and electrical properties are
exhibited by porous Si has been used to
different from those of bulk semiconduc-
design high-efficiency Si solar cells with a
tors, opening the route to new potential
reduced number of steps in the process-
technological applications. In view of its
ing of industrial cells [96–98]. Porous Si
room temperature efficient visible light-
matrices are very efficient templates for
emitting properties, porous Si was re-
implementation in medical therapeutics
garded as a prime candidate material for a
and biologic sensors [262].
Over the last ten years, porosity new generation of Si chips, which might
has emerged as a promising tool for extend the functionality of Si technology
controlling optical and photoelectrochem- from microelectronics into optoelectron-
ical properties in semiconductors other ics. Recent progress in realizing electrolu-
than Si, especially III–V compounds. In minescence efficiencies comparable to the
GaP, for instance, porosity leads to an photoluminescence is promising. How-
enhanced photoresponse of solid/liquid ever, it is not yet clear if porous Si will
junctions, whereas macroporous GaP ap- finally enable optical and electronic tech-
pears to be the most strongly scattering nologies to be integrated on a single silicon
material for visible light to date, and local- chip. The major interests in porous Si
ization effects are anticipated [224, 225]. It might well be foreseen in other domains
has recently been shown that the nanos- of applications described earlier, such as
tructured membrane of GaP produced by sensors, dielectric multilayer, Bragg reflec-
porous-etching are suitable for nonlinear tors, Microcavities, Waveguides, Photonic
optical applications [223]. In InP, poros- band gap systems, microcapacitors, photo-
ity induces a sizeable birefringence at voltaic, and biomedical applications.
Although the number of publications de- 5. R. L. Smith, S. D. Collins, J. Appl. Phys.

voted to porous semiconductors is much 1992, 71, R1–R22.
6. L. T. Canham, Appl. Phys. Lett. 1990, 57,
less than that devoted to porous Si, some
1046–1048.
important experimental findings have re- 7. A. G. Cullis, L. T. Canham, Nature 1991,
cently been reported, in particular with 353, 335–338.
porous III–V compounds. Electrochem- 8. P. C. Searson, J. M. Macaulay, S. M. Prokes,
istry offers an accessible and cost-effective J. Electrochem. Soc. 1992, 139, 3373–3378.
9. P. C. Searson, J. M. Macaulay, F. M. Ross,
approach for ‘‘drilling’’ vertical arrays of
J. Appl. Phys. 1992, 72, 253.
macropores in semiconductor materials. 10. C. Lévy-Clément, Porous Silicon Science
Similar to porous Si, quasi-uniformly dis- and Technology (Eds.: J. C. Vial, J. Derrien)
tributed parallel pores can be introduced Les Editions de Physique, Paris, 1995,
into III–V semiconductors, by photon- P. 329–344.
assisted or MeV-ion implantation-assisted 11. A. G. Cullis, L. T. Canham, P. D. J. Calcott,
J. Appl. Phys. 1997, 82, 919–965.
anodic etching. Freestanding GaP mem- 12. V. Parkhutik, Electrochem. Soc. Proc. 1995,
branes represent promising materials for PV 95-25, 105–114.
advanced nonlinear optical applications or 13. J.-N. Chazalviel, R. B. Werhspohn, F. Oza-
can make possible new applications such nam, Mater. Sci. Eng., B, Solid State Mater.
as gas-sensing devices. Advanced Technology 2000, B69–70, 1–10.
14. R. T. Collins, P. M. Fauchet, M. A. Tischler,
Nevertheless, the field of porous semi- Phys. Today 1997, 50, 244.
conductors is still in its infancy and lot of 15. K. H. Jung, S. Shih, D. K. Kwong, J. Elec-
work has to be done before foreseeing in- trochem. Soc. 1993, 140, 3046–3064.
dustrial devices using porous Si and other 16. P. M. Fauchet, J. Lumin. 1996, 70, 294–309.
porous semiconductors, whose properties 17. W. Theiss, Surf. Sci. Rep. 1997, 29, 91.
18. L. Pavesi, La Rivista del Nuovo Cimento 1997,
have so far been studied rather scantily. 20, 1–126.
19. Special issue of Phys. Status Solidi (a) 2000,
Acknowledgments 182 No. 1.
20. (a) MRS Symposium Proceedings, Materi-
als Research Society, Pittsburgh. Vol. 256,
My collaborators Stéphane Bastide and 1992, 20; (b) Vol. 283, 1993, 20; (c) Vol. 358,
Christian Vard are warmly acknowledged 1995, 20; (d) Vol. 452, 1997, 20; (e) Vol. 486,
for their help in preparing the manuscript. 1998.
I am grateful to Prof. Leigh Canham for 21. Advanced Luminescent Materials, The Elec-
trochemistry Society Proceedings Series,
giving me an exhaustive list of references
Pennington, NJ PV 95-25 1996.
on porous semiconductors. This work was 22. The Electrochemistry Society Proceedings
partially funded by CNRS-ECODEV and Series Pits and Pores: Formation, Properties
ADEME within the French Photovoltaic and Significance for Advanced Luminescent
Program. materials, Corrosion/Luminescence and Dis-
play Materials, PV 97-7 1997.
23. Special issue of J. Lumin. 57, 1993; 80, 1998.
References 24. (a) Special issue of Thin Solid Films 255,
1995, 23; (b) 276, 1996, 23; (c) 297, 1997.
1. A. Uhlir, Bell Syst. Tech. J. 1956, 35, 333. 25. Special issue of J. Porous Mater. 2000, 7,
2. D. R. Turner, J. Electrochem. Soc. 1958, 105, 1–414.
402–408. 26. Special issue of Mater. Sci. Eng. B, Solid
3. D. R. Turner, J. Electrochem. Soc. 1960, 107, State Mater. Adv. Technol. 2000, B69–70,
810. 1–215.
4. Y. Watanabe, Y Arita, T. Yokoyama et al., J. 27. C. Lévy-Clément, A. Lagoubi, R. Tennee
Electrochem. Soc. 1975, 122, 1351. et al., Electrochim. Acta. 1992, 37, 877–888.
28. C. Lévy-Clément, Semiconductor Micropro- 46. (a) R. Memming, G. Schwandt, Surf. Sci.
cessing (Eds.: H. J. Lewerenz, S. A. 1966, 4, 109; (b) 1966, 5, 97–110.
Campbell), Wiley and Sons, Chapter 5, 47. P. A. Kohl, C. Wolowodiuk, F. W. Oster-
1998, pp. 177–215. mayer Jr., J. Electrochem. Soc. 1983, 130,
29. C. Lévy-Clément, S. Bastide, Zeitschrift für 2288–2293.
Physikalische Chemie 1999, 212, 123–143. 48. N. J. Hovel, Semiconductors and Semimetals,
30. O. Bisi, Surface Science Reports 1999, 38, Vol. 11, Solar Cells (Eds.: R. K. Willardson,
1–126. A. C. Beer), Academic Press, New York,
31. Optical properties of Low dimensional Sili- 1975, pp. 225.
con Structures, (Eds.: D. C. Benshael, L. T. 49. J. Rouquerol, D. Avnir, C. W. Fairbridge
Canham, S. Ossicini) NATO ASI Series, et al., Pure Appl. Chem. 1994, 66, 1739.
Vol. 244, Kluwer Academic Publishers, Dor- 50. M. Matsumura, S. R. Morrison, J. Elec-
drecht, 1993. troanal. Chem. 1983, 144, 113–120; (b) Ibid
32. Porous Silicon Science and Technology (Eds.: 1983, 147, 157–166.
J. C. Vial, J. Derrien) Les Editions de 51. H. Gerischer, M. Lübke, Ber. Bunsenges.
Physique, Paris, 1995. Phys. Chem. 1987, 91, 394–398.
33. Porous Silicon (Eds.: Z. Chuan Feng, 52. H. Gerischer, M. Lübke, J. Electrochim. Soc.
R. Tsu), World Scientific, New York, 1995. 1988, 135, 2782.
34. Properties of Porous Silicon (Ed.: L. T. Can- 53. H. J. Lewerenz, J. Stumper, L. M. Peter,
ham) IEE INSPEEC, The Institution of Phys. Rev. Lett. 1988, 61, 1989–1992.
Electrical Engineers, London, 1997. 54. J. Stumper, H. J. Lewerenz, C. Pettenkofer,
35. Structural and Optical Properties of Porous Phys. Rev. B. 1990, 41, 1592.
Silicon Nanostructures, (Eds.: G. Amato, 55. L. M. Peter, A. M. Borazio, H. J. Lewerenz,
C. Delerue H. J. von Bardeleben), Gordon J. Stumper, J. Electroanal Chem. 1990, 290,
and Breach, Amsterdam, 1997. 229–248.
36. Light emission in Silicon: from Physics to 56. J. Stumper, L. M. Peter, J. Electroanal.
Devices. Semiconductors and Semimetals, Chem. 1991, 309, 325.
Vol. 49, (Ed.: D. J. Lockwood), Academic 57. H. Gerischer, P. Allongue, V. Costa Kieling,
Press, London, 1998. Ber. Bunsenges. Phys. Chem. 1993, 97,
37. H. Gerischer, A. Mauerer, W. Mindt, Surf. 753–757.
Sci. 1966, 4, 431–439. 58. P. Allongue, V. Kieling, H. Gerischer, Elec-
38. H. Gerischer, Physical Chemistry: Vol. IXA/ trochim. Acta 1995, 40, 1353–1360.
Electrochemistry (Eds.: H. Eyring, D. Hend- 59. D. J. Blackwood, A. M. Borazio, R. Greef
erson, W. Jost) Academic Press, New York, et al., Electrochem. Acta 1992, 37, 889.
1970, pp. 463–542. 60. L. M. Peter, D. J. Blackwood, S. Pons, Phys.
39. H. Gerischer, J. Electroanal. Chem. 1977, 82, Rev. Lett. 1989, 62, 308–311.
133–143. 61. A. Venkateswara Rao, F. Ozanam, J.-N.
40. A. J. Bard, M. S. Wrighton, J. Electrochem. Chazalviel, J. Electrochem. Soc. 1991, 138,
Soc. 1977, 124, 1706–1710. 153–159.
41. R. Tenne, in Semiconductor Micromachin- 62. V. Lehmann, U. Gosele, Appl. Phys. Lett.
ing, Vol. 1: Fundamental Electrochem- 1991, 58, 856.
istry and Physics (Eds.: S. A. Campbell, 63. S. M. Sze, Physics of Semiconductor Devices,
H. J. Lewerenz), J. Wiley and Sons Ltd, New 2nd ed., Wiley and Sons, New York, 1981,
York, 1998, pp. 139–175. pp. 804.
42. A. Heller, B. Miller, F. A. Thiel, Appl. Phys. 64. R. Memming, Comprehensive Treatise of
Lett. 1981, 38, 282. Electrochemistry (Eds.: B. E. Conway, J.
43. A. J. Bard, R. L. Faulkner, Electrochemical O’M. Bockris, E. Yeager et al.,) Plenum,
Methods, Fundamendals and Applications, New York, 1983, Vol. 7, p. 529.
Wiley and Sons, New York, 1980. 65. H. Gerischer, Electrode Process in Solid State
44. M. J. Eddowes, J. Electroanal. Chem. 1990, Ionics (Eds.: H. Kleitz, J. Dupuy), D. Reidel:
280, 297–311. Dordrecht, The Netherlands, 1976, p. 299.
45. M. J. Eddowes, J. Electrochem. Soc. 1990, 66. X. G. Zhang, J. Electrochem. Soc. 1991, 138,
137, 3514–. 3750–3756.
67. M. J. J. Theunissen, J. Electrochem. Soc. 92. E. Galun, A. Lagoubi, R. Tenne, C. Lévy-

1972, 119, 351. Clément, J. Lumin, 1993, 57, 125–129.
68. H. Föll, Appl. Phys. A. 1991, 53, 8–19. 93. P. Williams, C. Lévy-Clément, J. E. Peou
69. V. Lehmann, H. Föll, J. Electrochem. Soc. et al., Thin Solid Films 1997, 298, 66–75.
1988, 135, 2831–2835. 94. G. Bomchil, A. Halimaoui, R. Herino, Mi-
70. C. Lévy-Clément, A. Lagoubi, M. Tomkie- croelectron. Eng. 1998, 8, 293–310.
wicz, J. Electrochem. Soc. 1994, 141, 95. T. Yonehara, K. Sakaguchi, N. Sato, Appl.
958–967. Phys. Lett. 1994, 64, 2108–2110.
71. V. Lehmann, H. Föll, J. Electrochem. Soc. 96. S. Strehlke, Q. N. Le, D. Sarti et al., Thin
1990, 137, 653–659. Solid Films 1997, 297, 291–295.
72. V. Lehmann, J. Electrochem. Soc. 1993, 140, 97. S. Strehlke, S. Bastide, C. Lévy-Clément,
2836–2483. Sol. Energy Mater. Sol. Cells 1999, 58,
73. V. Lehmann, W. Hönlein, H. Reisinger 399–409.
et al., Thin Solid Films 1996, 276, 138–142. 98. S. Strehlke, S. Bastide, J. Guillet et al.,
74. S. Ottow, V. Lehmann, H. Föll, J. Elec- Mater. Sci. Eng. 2000, B69–70, 81–86.
trochem. Soc. 1996, 390, 143, 385–390. 99. C. Lévy-Clément, A. Lagoubi, D. Ballutaud
75. U. Grüning, V. Lehmann, C. M. Engel- et al., Appl. Surf. Sci. 1993, 65/66, 408–414.
hardt, Appl. Phys. Lett. 1995, 66, 3254–3256. 100. E. Galun C. Reuben, S. Matlis et al., J. Phys.
76. E. Propst, P. A. Kohl, J. Electrochem. Soc. Chem. 1995, 99, 4132–4140.
1993, 140, L78–L80. 101. R. B. Wehrspohn, J.-N. Chazalviel, F. Oza-
77. E. Propst, P. A. Kohl, J. Electrochem. Soc. nam, J. Electrochem. Soc. 1999, 146,
1994, 141, 1006–1013. 3309–3314.
78. M. M. Rieger, P. A. Kohl, J. Electrochem. 102. V. Lehmann, S. Rönnebeck, J. Electrochem.
Soc. 1995, 142, 1490.
Soc. 1999, 146, 2968–2975.
79. E. Ponomarev, C. Lévy-Clément, Electro-
103. E. Ponomarev, C. Lévy-Clément, J. Porous
chem. Sol. State Lett. 1998, 1, 42–45.
Mater. 2000, 7, 51–56.
80. S. Lust, C. Lévy-Clément, Phys. Status. So-
104. W. M. Shen, M. Tomkiewicz, C. Lévy-
lidi. (A) 2000, 182, 17–21.
Clément, J. Appl. Phys. 1994, 76, 3636–3639.
81. P. C. Searson, X. G. Zhang, J. Electrochem.
105. A. Albu-Yaron, S. Bastide, D. Bouchet et al.,
Soc. 1990, 137, 2539.
J. Phys. I France 1994, 4, 1181–1197.
82. P. C. Searson, X. G. Zhang, Electrochim.
Acta 1991, 36, 499. 106. M. H. Al Rafai, S. Ottow, M. Christop-
83. F. Gaspard, A. Bsiesyl, M. Ligeon et al., J. hersen et al., J. Electrochem. Soc. 2000, 147,
Electrochem. Soc. 1989, 136, 3043–3046. 627–635.
84. I. Ronga, A. Bsiesy, F.Gaspard et al., J. 107. U. Grüning, V. Lehmann, Thin Solid Films
Electrochem. Soc. 1991, 138, 1403–1407. 1996, 276, 151–154.
85. M. I. J. Beale, Appl. Phys. Lett. 1985, 46, 108. U. Grüning, V. Lehmann, S. Ottow et al.,
86–88. Appl. Phys. Lett. 1996, 68, 747–749.
86. V. Lehmann, R. Stengl, A. Luigart, Mater. 109. A. Birner, S. Ottow, A. Schneider et al.,
Sci. Eng. 2000, B69–70, 11–22. Phys. Status. Solidi. (A) 1998, 165, 111–117.
87. E. Ponomarev, C. Lévy-Clément, J. Porous 110. F. Müller, A. Birner, U. Gösele et al., J.
Mater. 2000, 7, 51–56. Porous Mater. 2000, 7, 201–204.
88. P. Steiner, F. Kozlowski, W. Lang, Appl. 111. V. Lehmann, W. Hönlein, H. Reisinger
Phys. Lett. 1993, 62, 2700. et al., Thin Solid Films 1996, 276, 138–143.
89. M. I. J. Beale, J. D. Benjamin, M. J. Uren 112. S. Strehkle, Thesis, Paris, 1999.
et al., J. Cryst. Growth 1985, 73, 622–636. 113. M. Christophersen, J. Cartensen, H. Föll,
90. J. P. Gonchond, A. Halimaoui, K. Ogura, Phys. Status. Solidi (a) 2000, 182, 45–50.
Microscopy of Semiconducting Materials, 114. M. Christophersen, J. Cartensen, A. Feuer-
(Eds.: A. G. Cullis, N. J. Long), IOP, Bristol, hake, M. Föll, Mater. Sci. Eng. 2000,
1991, p. 235. B69–70, 194–198.
91. A. Albu-Yaron, S. Bastide, J. L. Maurice 115. P. Allongue, V. Costa-Kieling, H. Geris-
et al., J. Lumin, ‘‘Light Emission From cher, J. Electrochem. Soc. 1993, 140, 1009.
Silicon’’, (Eds.: J. C. Vial, L. T. Canham, 116. F. Ronkel, J. W. Schultze, J. Porous Mater.
W. Lang) 1993, 57, 67–71. 2000, 7, 11–16.
117. P. Allongue, M. C. Bernard, J. E. Peou et al., 141. A. Valence, Phys. Rev. B 1995, 52,
Thin Solid Films 1997, 297, 1–4. 8323–8336.
118. J. E. Peou, C. Henry de Villeneuve, 142. A. Valence, Phys. Rev. B 1997, 55,
P. Allongue et al., Pits and Pores: Formation, 9706–9715.
Properties and Significance for Advanced Lu- 143. R. B. Wehrspohn, F. Ozanam, J.-N. Chazal-
minescent Materials, Corrosion Luminescence viel, J. Electrochem. Soc. 1999, 146,
and Display Materials, (Eds.: D. J. Lockwood 3309–3314.
et al.), The Electrochemical Society Pro- 144. R. Tenne, V. Marcu, Y. Prior, Appl. Phys.
ceedings Series, Pennington, NJ, 1997, Lett. 1985, A37, 205–209.
Vol. 97-7, 83–91. 145. R. Tenne, H. Mariette, C. Lévy-Clément
119. R. Tenne, G. Hodes, Appl. Phys. Lett. 1980, et al., Phys. Rev. B 1987, 15, 1204–1207.
37, 428–430. 146. R. Tenne, C. Lévy-Clément, H. Mariette
120. R. Tenne, G. Hodes, Surf. Sci. 1983, 135, et al., J. Cryst. Growth 1987, 86, 826–833.
453. 147. H. Homyonfer, R. Tenne, W. Giriat et al.,
121. R. Tenne, H. Flaisher, R. Triboulet, Phys. Phys. Rev. B 1993, 47, 1244.
Rev. B 1984, 29, 5799. 148. R. Tenne, Appl. Phys. Lett. 1983, 43,
122. V. M. Dubin, Surf. Sci. 1992, 274,82. 201–203.
123. C. Lévy-Clément, A. Lagoubi, M. Neumann- 149. R. Tenne, V. Marcu, N. Yellin, Appl. Phys.
Spallart et al., J. Electrochem. Soc. 1991, 138, Lett. 1984, 45, 1219–1221.
L69–70. 150. C. Lévy-Clément, J. Rioux, S. Litch et al., J.
124. S. M. Prokes, W. E. Carlos, V. M. Bermu- Appl. Phys. 1985, 58, 4703–4708.
dez, Appl. Phys. Lett. 1992, 61, 1447. 151. C. Lévy-Clément, R. Triboulet, R. Tenne,
125. H. J. von Bardeleben, C. Ortega, A. Gros- Sol. Energy Mater. 1988, 17, 201–206.
mann et al., J. Lumin. 1993, 57, 301. 152. P. A. Kohl, F. W. Ostermayer Jr., Ann. Rev.
126. G. Oskam, P. M. Hoffman, J. C. Schmidt Mater. Sci. 1989, 19, 379–399.
et al., J. Phys. Chem. 1996, 100, 1801–1806. 153. F. Decker, D. A. Soltz, L. Cestato, Elec-
127. P. M. Hoffman, I. E. Vermeir, P. Searson, trochim. Acta 1993, 38, 95–99.
J. Electrochem. Soc. 2000, 147, 2999–3002. 154. N. G. Ferreira, D. A. Soltz, F. Decker et al.,
128. S. M. Sze, Physics of Semiconductor Devices, J. Electrochem. Soc. 1995, 142, 1348.
2nd ed., Wiley and Sons, New York, 1981, 155. R. Tenne, B. Theys, J. Rioux et al., J. Appl.
p. 122. Phys. 1985, 57, 141–145.
129. V. A. Burrows, Y. J. Chabal, G. S. Higasi 156. R. Tenne, A. Wold, Appl. Phys. Lett. 1985,
et al., Appl. Phys. Lett. 1988, 53, 998. 47, 707–709.
130. Y. J. Chabal, G. S. Higasi, K. Raghavachari 157. D. Mahalu, A. Wold, R. Tenne et al., J. Phys.
et al., J. Vac. Sci. Technol., A 1988, 7, 2104. Chem. 1990, 94, 8012.
131. J. J. Boland, Surf. Sci. 1992, 261, 17. 158. C. Lévy-Clément, R. Tenne, Photoelectro-
132. P. Kleimann, J. Linnros, S. Petersson, Mat. chemistry and Photovoltaics of Layered
Sci. Eng. 2000, B69–70, 29–33. Semiconductors, Chapter IV, (Ed.: A. Aru-
133. T. A. Witten, L. M. Sander, Phys. Rev. Lett. chamy), Kluwer Academic Publishers, Nor-
1981, 47, 1400–1403. well, MA, 1991, pp. 155–194.
134. W. W. Mullins, R. F. Sekerta, J. Appl. Phys. 159. D. Mahalu, L. Margulis, A. Wold, R. Tenne
1964, 35, 444. Rapid Comm. Phys. Rev. B 1992, 45, 1943.
135. R. L. Smith, S. D. Collins, Phys. Rev. A 1989, 160. F. Ozanam, R. B. Wehrspohn, J. N. Chazal-
39, 5409. viel, I. Solomon, Phys. Stat. Sol. (a) 1998,
136. R. W. Bower, S. D. Collins, Phys. Rev. A. 165, 15.
1991, 43, 3165. 161. K. H. Beckmann, Surf. Sci. 1966, 5,
137. H. Yan, X. Hu, J. Appl. Phys. 1993, 73, 187–196.
4324–4331. 162. M. Sendova-vassileva, N. Tzenov, D. Dim-
138. G. C. John, V. A. Singh, Phys. Rev. B. 1995, ova-malinskova et al., Thin Solid Films 1994,
52, 11125–11131. 255, 282–285.
139. Y. Kang, J. Jorné, J. Electrochem. Soc. 1993, 163. S. Miyazawa, K. Shiba, K. Hirose et al., Thin
140, 2258. Solid Films 1995, 255, 99.
140. Y. Kang, J. Jorné, Appl. Phys. Lett. 1993, 62, 164. S. Bayliss, Q. Zhang, P. Harris, Appl. Surf.
2224–2226. Sci. 1996, 102, 390–394.
165. H. C. Choi, J. M. Buriak, Chem. Commun. 191. J. N. Wang, Z. M. Chen, P. W. Woo et al.,
2000, 1669–1670. Appl. Phys. Lett. 1999, 74, 923–925.
166. A. Ksendzov T. George, W. D. Pike et al., 192. J. E. Spanier, I. P. Herman, J. Porous Mater.
Appl. Phys. Lett. 1993, 63, 200–202. 2000, 7, 139–142.
167. S. Gardelis, P. Dawson, J. S. Rimmer et al., 193. J. E. Spanier, I. P. Herman, Phys. Rev. B
Appl. Phys. Lett. 1991, 59, 2118–2120. 2000, 61, 10437–10450.
168. H. Shi, Y. Zeng, Y. Wang et al., Appl. Phys. 194. J. E. Spanier, G. T. Dunne, L. B. Rowland
A 1993, 57, 573–575. et al., Appl. Phys. Lett. 2000, 76, 3879–3881.
169. M. Schoisswohl, J. L. Cantin, M. Chamarro 195. S. Zangooie, H. Arwin, Phys. Status Solidi.
et al., Phys. Rev. B 1995, 52, 11898–11903. (A) 2000, 18, 213–219.
170. K. Kolic, E. Borne, A. Sibai et al., Thin Solid 196. T. Monguchi, K. Ono, Y. Baba et al., J.
Films 1995, 255, 279–281. Electrochem. Soc. 2000, 147, 741–743.
171. M. Schoisswohl, J. L. Cantin, S. Lebib et al., 197. J. P. Krumme, M. E. Straumanis, Trans.
J. Appl. Phys. 1996, 79, 9301–9304. Met. Soc. AIME 1967, 239, 395.
172. D. Buttard, J. L. Cantin, M. Schoisswohl 198. M. M. Faktor, D. G. Fiddyment, M. R. Tay-
et al., Thin Solid Films 1997, 297, 233–236. lor, J. Electrochem. Soc. 1975, 122,
173. B. Ünal, P. Phillips, E. H. C. Parker et al., 1566–1567.
Thin Solid Films 1997, 305, 274–279. 199. D. H. Holt, J. Mater. Sci. 1988, 23, 1131.
174. S. Lebib, J. L. Fave, J. Cernogora et al., J. 200. P. Schmuki, J. Fraser, C. M. Vitrus et al., J.
Lumin. 1999, 80, 153–157. Electrochem. Soc. 1996, 143, 3316–3322.
175. B. Ünal, M. Parkinson, S. C. Bayliss et al., 201. P. Schmuki, D. J. Lockwood, B. F. Mason
J. Porous Mater. 2000, 7, 143–146. et al., Appl. Phys. Lett. 1996, 69, 1620–1622.
176. S. Shor, B. Z. Weiss, A. D. Kurtz et al., 202. M. Hao, C. Shao, T. Soga et al., J. Cryst.
Appl. Phys. Lett. 1993, 62, 2836–2838. Growth 1997, 179, 6661.
177. T. Matsumoto, T. Tamaki, T. Futagi et al., 203. D. N. Goryachev, O. M. Sreseli, Semicon-
Appl. Phys. Lett. 1994, 64, 226–228. ductor 1997, 31, 1192–.
178. J. S. Shor, L. Bernis, A. D. Kurtz et al., J. 204. G. Oskam, A. Natarajan, F. M. Ross et al.,
Appl. Phys. 1994, 76, 4045–4049. Appl. Surf. Sci. 1997, 119, 160–168.
179. A. O. Konstantinov, A. Henry, E. Janzën, 205. F. M. Ross, G. Oskam, P. C. Pearson et al.,
Appl. Phys. Lett. 1994, 65, 2699–2701. Phil. Mag. A 1997, 75, 525–539.
180. H. Mimura, T. Matsumoto, Y. Kanemitsu, 206. I. M. Tiginyaru, G. Irmer, J. Monecke et al.,
Appl. Phys. Lett. 1994, 65, 3350–3352. Semicond. Sci. Technol. 1997, 12, 491–493.
181. V. Petrova-Koch, O. Sreseli, G. Polisski 207. P. Schmuki, L. E. Erickson, D. J. Lockwood
et al., Thin Solid Films 1995, 255, 107–110. et al., Appl. Phys. Lett. 1998, 72, 1039–1041.
182. A. O. Konstantinov, A. Henry, C. I. Harris 208. P. Schmuki, L. E. Erickson, D. J. Lockwood
et al., Appl. Phys. Lett. 1995, 66, 2250–2252. et al., J. Electrochem. Soc. 1999, 146,
183. H. Mimura, T. Matsumoto, Y. Kanemitsu, 735–740.
Solid-State Electron. 1996, 40, 501–504. 209. D. J. Lockwood, P. Schmuki, H. J. Labbé
184. M. F. MacMillan, J. Appl. Phys. 1996, 80, et al., Physica E 1999, 4, 102–110.
2412–2419. 210. Y. Morishita, S. Kawai, J. Sunagawa, Jpn. J.
185. V. P. Parkhutik, F. Namavar, E. Andrade, Appl. Phys. 1999, 38, L1156–L1158.
Thin Solid Films 1997, 297, 229–232. 211. C. M. Finnie, X. Li, P. W. Bohn, J. Appl.
186. O. Jessensky, F. Müller, V. Gösele, Thin Phys. 1999, 86, 4997–5003.
Solid Films 1997, 297, 224–228. 212. X. Li, P. W. Bohn, J. Electrochem. Soc. 2000,
187. H. W. Shim, K. C. Kim, Y. H. Seo et al., 147, 1740–1746.
Appl. Phys. Lett. 1997, 70, 1757–1759. 213. B. D. Chase, D. B. Holt, J. Electrochem. Soc.
188. V. P. Parkhutik, J. Appl. Phys. 1998, 83, 1972, 119, 314–317.
4647–4651. 214. B. H. Erné, D. Vanmaekelbergh, J. J. Kelly,
189. A. M. Danishevskii, V. B. Shuman, A. A. Adv. Mater. 1995, 7, 739–742.
Sitnikova et al., Semicond. Sci. Technol. 1998, 215. D. Vanmaekelbergh et al., Electrochim. Acta
13, 1111–1116. 1995, 10, 689.
190. W. Shin, T. Hikosaka, H. S. Ahn et al., J. 216. A. Anedda, A. Serpi, V. A. Karavanski et al.,
Electrochem. Soc. 1998, 145, 2456–2460. Appl. Phys. Lett. 1995, 67, 3316–3318.
217. A. Meijerink, A. A. Bol, J. J. Kelly, Appl. 240. B. H. Erné, A. Million, J. Vigneron et al.,
Phys. Lett. 1996, 69, 2807–2803. Electrochem. Solid Stat. Lett. 1999, 2,
218. B. H. Erné, D. Vanmaekelbergh, J. J. Kelly, 619–621.
J. Electrochem. Soc. 1996, 143, 305–314. 241. F. Zenia, R. Triboulet, C. Lévy-Clément
219. F. Iranzo Marin, M. A. Hamstra, D. Vanm- et al., Appl. Phys. Lett. 1999, 75, 531–533.
aekelbergh, J. Electrochem. Soc. 1996, 143, 242. F. Zenia, C. Lévy-Clément, R. Triboulet
1137–1142. et al., Electrochem. and Solid Stat. Lett. 2000,
220. I. M. Tiginyanu, C. Schwab, J.-J. Grob et al., 3, 73–76.
Appl. Phys. Lett. 1997, 71, 3829–3831. 243. F. Zenia, C. Lévy-Clément, R. Triboulet
221. I. M. Tiginyanu, G. Irmer, J. Monecke et al., et al., Thin Solid Films 2000, 361–362,
Phys. Rev. B 1997, 55, 6739–6742. 49–52.
222. K. Kuriyama, K. Ushiyama, K. Ohbora et al., 244. D. Lincot, J. Vedel, J. Cryst. Growth 1985, 72,
Phys. Rev. B 1998, 58, 1103–1105. 426.
223. A. Sarua, I. M. Tiginyanu, V. V. Ursaki 245. D. Lincot, J. Vedel, J. Electroanal. Chem.
et al., Solid. Stat. Comm. 1999, 112, 1987, 220, 179.
581–585. 246. L. A. Harris, R. H. Wilson, J. Electrochem.
224. F. J. P. Schhurmans, D. Vanmaekelbergh, Soc. 1976, 123, 1010–1015.
J. van de Lagemaat et al., Science 1999, 284, 247. Y. Nakato, H. Akanuma, J. I. Shimizu et al.,
141–143. J. Electroanal. Chem. 1995, 396, 35–39.
225. F. J. P. Schhurmans, M. Megens, D. Vanm- 248. L. A. Harris, D. R. Cross, M. E. Gerstner, J.
aekelbergh et al., Phys. Rev. Lett. 1999, 83, Electrochem. Soc. 1977, 124, 939–844.
2183–2186. 249. T. Sugiura, T. Yoshida, H. Minoura, Elec-
226. W. H. Lubberhuizen, D. Vanmaekelberg, trochem. Solid State Lett. 1998, 1, 175–177.
250. J. van de Lagemaat, M. Plakman, D. Vanm-
E. van Faassen, J. Porous Mater. 2000, 7,
aeckelbergh et al., Appl. Phys. Lett. 1996, 69,
147–152.
2246–2248.
227. I. M. Tiginyanu, I. V. Kravetsky, J. Monecke
251. (Ed.: P. M. Fauchet) IEEE J. Select. Topics
et al., Appl. Phys. Lett. 2000, 77, 2415–2417.
Quantum Electron. 1998, 4, 1020.
228. T. Takizawa, S. Arai, M. Nakahara, Jpn. J.
252. B. Lendt et al. Analyt. Chem. 1997, 69, 2877.
Appl. Phys. 1994, 33, L643–L645.
253. G. Lerondel et al. J. Imaging Sci. Technol.
229. E. Kikuno, M. Amiotti, T. Takizawa et al.,
1997, 41, 468.
Jpn. J. Appl. Phys. 1995, 34, 177–178. 254. H. Shinoda et al. Nature 1999, 400, 853.
230. T. Takizawa, M. Nakahara, E. Kikuno et al., 255. J. H. Song, M. J. Sailor, Comments Inorg.
J. Electron. Mater. 1996, 25, 657–660. Chem. 1999, 21, 69.
231. A. Hamamatsu, C. Kaneshiro, H. Fujikura 256. M. Wainwright et al. Lett. Appl. Microbiol.
et al., J. Electroanal. Chem. 1999, 473, 1999, 29, 224.
223–229. 257. V. G. Zubko et al. Astrophys. J. 1998, 501.
232. H. Fujikura, A. Liu, A. Hamamatsu et al., 258. T. E. Bell et al. J. Micromech. Microeng. 1996,
Jpn. J. Appl. Phys. 2000, 39, 4616–4620. 6, 361.
233. P. Schmuki, L. Santinacci, T. D. Djenizian 259. V. P. Bondarenko et al., IEEE Trans Nucl.
et al., Phys. Status Solidi (A) 2000, 182, Sci. 1997, 44, 1719.
51–61. 260. C. Nam et al. IEEE Microwave Guided Wave
234. A. Sarua, G. Gärtner, G. Irmer et al., Phys. Lett. 1997, 7, 236.
Status Solidi (A) 2000, 182, 207–211. 261. V. Parkhutik, E. Yu. Budnikiv, S. F. Tima-
235. K. Chernoutsan, V. Dneprovskii, O. Shali- shev, Mater. Sci. Eng. B 2000, 67/70, 53–58.
gina et al., Phys. Status. Solidi. (A) 2000, 262. L. T. Canham, Adv. Mater. 1995, 7, 1033.
182, 347–352. 263. N. Koshida, H. Koyama, Appl. Phys. Lett.
236. M. Mynbaeva, D. Tsvetkov, Inst. Phys. Conf. 1992, 60, 347.
Ser. 1997, 155, 365. 264. S. Zangooie, R. Bjorklund, H. Arwin, Sens.
237. J. van de Lagemaat, Thesis, Utrecht, 1998. Actuators B 1997, 43, 168.
238. M. Mynbaeva, A. Titkov, A. Kryganovskii 265. T. Laurell, J. Drott, L. Rosengre et al., Sens.
et al., Appl. Phys. Lett. 2000, 76, 1113–1115. Actuators B 1996, 31, 161.
239. B. H. Erné, C. Mathieu, J. Vigneron et al., J. 266. S. Setzu, G. Lérondel, R. Romestain, J.
Electrochem. Soc. 2000, 147, 3759–3767. Appl. Phys. 1998, 84, 3129.
267. Z. Zangooie, R. Janson, H. Arvin, J. Vac. 271. L. Pavesi, P. Dubost, Semicond. Sci. Technol.
Sci. Technol. A 1998, 16, 2901. 1997, 12, 570.
268. S. Fronhoff, M. G. Berger, Adv. Mater. 1994, 272. A. Loni, L. T. Canham, M. G. Ber-
12, 963. ger et al., Thin Solid Films 1996,
269. M. G. Berger, M. Thönissen, R. Arhens- 276, 143.
Fischer et al., Thin Solid Films 1995, 255, 273. V. P. Parkhutik, L. T. Canham, Phys. Status
313. Solidi (A) 2000, 182, 591–598.
270. C. Mazzoleni, L. Pavesi, Appl. Phys. Lett. 274. V. P. Parkhutik, L. T. Canhamll, Phy. Status
1995, 67, 2893. Solidi. (A) 2000, 182, 591–598.
287
4.1 applications. One of the most important as-

The Photoelectrochemistry of pects in using solar energy is its conversion
Semiconductor/Electrolyte Solar Cells from solar radiation into electrical energy.
This is achieved by using semiconductor-
Maheshwar Sharon
based photovoltaic (PV) cells. Today’s PV
Indian Institute of Technology, Bombay, India
market is 151 MW per year correspond-
4.1.1 ing to a value of about 0.7 − 1 billion
Introduction US$. This is a remarkable market but
still far away from being a noticeable
Statistical assessment suggests that contribution to the world energy con-
12 000 kg of coal per capita per year is be- sumption. The major reason for the low
ing used by developed countries, whereas penetration of PV today is the high cost.
150 kg is being used by developing coun- Broadly speaking, there are three types of
tries. It is estimated that in the last century, PV cells (1) n : p junction cell, (2) metal-
semiconductor cell also known as Metal-
man has consumed the energy equivalent
Schottky junction, and (3) semiconductor-
to 4 × 1021 J and in the next century would
electrolyte junction. This section discusses
need energy equivalent to 100 × 1021 J.
the last type of cell.
The rate of use of fossil fuels (e.g. oil,
In 1839, Becquerel [1] first discovered
gas, coal, etc.) has linearly increased with
the PV phenomena in electrochemical
respect to utilization time. However, it can-
systems. Brattain and Garret [2, 3] were pi-
not follow the same trend indefinitely. To oneers explaining aspects of the properties
avoid an undesirable situation arising due of semiconductor–electrolyte interfaces.
to the shortage of fossil fuels, scientists all Fujishima and Honda [4] reported the
over the world, are trying to take advantage first indication of a practical application
of renewable energy sources such as solar, of a photoelectrochemical (PEC) system
wind, ocean, and so on. in 1972. This paper sparked off a wave
The sun is the cleanest and most abun- of investigations all over the world. It
dant energy source. It is estimated that out would be appropriate, however, to suggest
of the total solar energy (3.8 × 1020 MW), that the interest in photoelectrochemistry
earth receives about 1.7 × 106 MW. There- of semiconductor blossomed only after
fore, in recent years, attention has been the pioneering work of Gerischer [5] and
paid to use solar energy for terrestrial Myamlin and Pleskov [6]. These studies
288 4 Solar Energy Conversion without Dye Sensitization
led to the discovery of wet PV solar cell. protected by insulation, so that only
This is popularly known as a PEC cell. It one of its surfaces is exposed to the
was believed that a PEC cell might be more redox electrolyte. In a regenerative PEC
economically viable as compared with the solar cell, the metal counter-electrode is
solid-state PV cell. Simplicity in fabricat- expected to perform an electrochemical
ing the cell was the main reason for such reaction that is the opposite of the
an optimistic view. Moreover, as discussed process occurring at the semiconductor
in a later chapter, it has been possible electrode. The counter-electrode should
to make a rechargeable battery with in be electrochemically stable. It would be
situ storage capability by using a PEC desirable that work function of counter-
cell [7–9]. The wet type PEC cell suffers electrode is compatible with that of the
from the instability of semiconductor in Fermi level of the semiconductor. The
aqueous media. It has been realized that matching of work function may not be
in spite of its simplicity and economical logical with metallic electrodes such as
viability, the wet-type PEC cell cannot eas- platinum, but may be important with
ily replace a silicon PV solar cell, unless semiconducting electrode such as SnO2 .
we discover photoelectrochemically stable On illumination of the semiconductor-
semiconductor materials possessing band electrolyte junction with a light having
gap approximately 1.4 eV. Till such time, energy greater than the band gap of
future of (unsensitized) wet-type PEC so- the semiconductor, photogenerated elec-
lar cell appears gloomy. However, research tron/holes are separated in the space
utilizing a standard PEC cell configuration charge region. The photogenerated minor-
is not complex, facilitating testing of the ity carriers (holes for n-type or electron
photoactivity of new semiconductor ma- for p-type semiconductor) arrive at the in-
terials. Current PEC cell research trends terface of the semiconductor–electrolyte.
have also been directed toward sensitized Photogenerated majority carriers (i.e. elec-
semiconductors and applications in the trons for n-type and holes for p-type
field of decontamination of water from semiconductor) accumulate at the back-
pathogenic bacteria [10], detoxification of side of the semiconductor (i.e. the side
water from toxic organic/inorganic ma- that is not illuminated and not in con-
terials, photography [11] or in any other tact with the redox electrolyte). With the
field where instability of the semiconduc- help of a connecting wire, photogener-
tor does not pose a real problem. ated majority carriers are transported to
The present section is devoted to various counter-electrode via a load (Fig. 1). These
aspects that are essential for the develop- carriers at the counter-electrode electro-
ment of a PEC cell. chemically react with the redox electrolyte
(i.e. reduction of redox electrolyte occurs
4.1.2 with n-type or oxidation of redox electrolyte
Description of a PEC Cell with p-type semiconductor).
Conversely, the photogenerated minor-
A PEC cell consists of a photoactive semi- ity carriers generated at the interface of
conductor electrode (either n- or p-type) the semiconductor perform the opposite
and a metal counter-electrode. Both these reaction, which occurred at the metal
electrodes are immersed in a suitable counter-electrode. Thus, photogenerated
redox electrolyte. The semiconductor is carriers are consumed. Because of these
4.1 The Photoelectrochemistry of Semiconductor/Electrolyte Solar Cells 289
Load
A
(Semiconductor) Counter electrode

Working electrode
Light hν
Quartz window Redox electrolyte
Fig. 1 A schematic diagram of a PEC cell.
two reactions, there is no net change in electrolyte to maintain the neutrality of

the electrolyte, and electrical power is pro- the solution. At the counter electrode e−
surf
duced in the load. reduces the redox electrolyte, that is,
Details of the electrochemical reac-
tions with respect to a PEC cell utiliz- e−
surf + Fe
3+
−−−→ Fe2+ (5)
ing a p-type semiconductor dipped in
FeSO4 /Fe2 (SO4 )3 electrolyte are given as Corresponding chemical reaction is
At the p-semiconductor electrode Fe2 (SO4 )3 + 2e−

surf
−−−→ 2FeSO4 , (6)
hν + p-semiconductor −−−→ e− /h+
that is, at the counter-electrode, one
−
× (photogenerated carriers) (1) additional SO4 − ion is formed that needs
to be transported to the semiconductor
Separation of charge carrier (e− /h+ )
or to the bulk solution to maintain its
caused by junction formation
neutrality.
(e− /h+ )depletion region −−−→ e− + The reaction in this regenerative PEC
surf + hbulk
(2) is the conversion of photon energy into
h+ is transported to the counter- electrical energy without destroying chem-
bulk
electrode via a load and oxidizes redox ical composition of the redox electrolyte or
electrolyte, that is, counter-electrode or semiconductor elec-
trode, that is, hν (photon energy) →
h+
bulk + Fe
2+
−−−→ Fe3+ (3) electrical energy.
Since in such a PEC cell, both electrodes
The corresponding chemical reaction is are immersed in the same electrolyte
2FeSO4 + 2h+ and specific reactions occur only at the
bulk −−−→ Fe2 (SO4 )3 , (4)
semiconductor and the metal, there is
that is, near this electrode, one additional no need for a separator. Hence, charge
−
SO4 − ion is required to be transported balance due to oxidation and reduction
from the counter-electrode or the bulk processes is maintained.
4.1.3 on illumination. This type of cell can be

Types of PEC Cells further classified as photoelectrolytic cell
("G > 0, Fig. 3) and photocatalytic cell
Figure 2 shows various possible PEC [12] ("G < 0, Fig. 3), depending on relative
cells. In a PEC cell (or electrochemical location of the potentials of the two redox
PV cell), optical energy is converted into couples (e.g. O/R and O$ /R$ ). In the latter
electrical energy, with zero change in the type of cells, anodic and cathodic com-
free energy of redox electrolyte. In this partments need to be separated to prevent
cell, the electrochemical reaction occur- mixing of the two redox couples.
ring at the counter-electrode is opposite
to the photoassisted reaction occurring at 4.1.4
the semiconductor electrode (Fig. 3). Thus, Aim of this Section
the light energy is converted into electri-
cal energy, with no change in the solution In this section, we shall be concerned
composition or electrode material. This with developing a regenerative PEC solar
is the case described earlier. In a photocell, where "G = 0. From previous dis-
electrosynthetic cell, optical energy is con- cussions, we understand how a PEC cell is
verted into chemical energy with nonzero fabricated. However, if we are to develop
free energy change in the electrolyte. In new materials for making an economically
this case, a net chemical change occurs viable PEC cell, it is essential that we
Electrochemical photovoltaic cells

(Optical energy converted into electrical energy)
∆G = 0
Photoelectrolytic cells
(Optical energy stored as chemical
energy in endorgic reactions
Photoelectrochemical e.g. H2O H2 + 1/2 O2)
cells
∆G ≠ 0 ∆G > 0
Photoelectrosynthetic cells
(Optical energy used to affect
chemical reactions)
∆G < 0
Photocatalytic cells
(Optical energy provides activation energy
for exorgic reactions
e.g. N2 + 3H2 2NH3)
Fig. 2 Classification of PEC cells [8].

e e
− −
e e
e Electrochemical
O O photovoltaic cell
R R
∆G = 0
e
+
(a) n -SC soln M p -SC + soln M
e e
− −
e e
O′ Photoelectrolytic
R′ O′ cell
R′ ∆G > 0
O O
R R
e
+ +
(b) n -SC soln M p -SC soln M
e e
− − Photocatalytic cell
O
R ∆G < 0
e O e
O′ R
R′ O′
R′
+ +
(c) n -SC soln M p -SC soln M
Fig. 3 A schematic representation of different types of PEC cells.
understand the principle, which controls technique to procure information about

the conversion of light energy into elec- the band position of the semiconductor.
trical energy. This knowledge will help us Finally, a simple experimental technique
to design a desirable semiconductor. It to maneuver the photocharacteristics of
is also necessary to realize that there are semiconductor-redox electrolyte is desir-
some constraints while employing a semi- able. Unless, we have this information,
conductor and a redox electrolyte to make a it would be an extremely difficult task
PEC cell. Therefore, knowledge regarding to launch production of new materials
potential energy of redox electrolyte and its for developing an economically viable and
relation with the band positions of semi- photoelectrochemically stable PEC cell. In
conductor is beneficial. To accomplish the forthcoming sections, efforts are made
this we also need to perceive a simple to provide this information as briefly as
possible because discussions on each of a specific Fermi level. For example,

these items are dealt extensively in other reduction of Fe3+ and Ce4+ requires 0.77
sections of this book. and 1.44 volts versus hydrogen electrode,
respectively. Alternatively, we can say that
4.1.5
the energy of electrons present in the
Semiconductor-electrolyte Junction
reduced product of these materials must
When a semiconductor comes in contact be equivalent to this potential. Hence,
with another material of Fermi level, dif- these potentials can be referred to as the
ferent from the semiconductor, a junction work function (or Fermi level) of the redox
is formed. This can be formed between electrolyte.
n-type and p-type semiconductors, or a Let us imagine a situation where an
metal and a semiconductor, or a semicon- n-type semiconductor is brought in con-
ductor and a redox electrolyte. tact with a redox electrolyte (Fig. 4a). Let
The semiconductor-electrolyte junction us also assume that the electrochemical
should not be confused with a metal- potential, that is, its redox potential, is al-
electrolyte junction. In the metal-electro- most equal to the intrinsic Fermi level of
lyte junction, the potential drop occurs the semiconductor. This situation forces
entirely on the solution side and practically the electrons (majority carriers) to flow
nil on the metal side, whereas for semicon- from the semiconductor to the electrolyte.
ductor–electrolyte interface, the potential This migration continues until the two
drop occurs on the semiconductor side as Fermi levels achieve an equilibrium posi-
well as the solution side. tion (Fig. 4b). At this condition, no further
For clarification of the type of junctions migration of electron occurs and a dynamic
formed at the semiconductor-electrolyte, equilibrium is established. In this situa-
let us take an example of n-type semi- tion, instead of electrochemically reacting
conductor. In addition to possessing free with the redox electrolyte, these major-
electrons (referred to as the majority car- ity carriers accumulate at the interface
rier), n-type semiconductor also possesses of semiconductor–electrolyte to maintain
holes (referred to as the minority carrier). neutrality of the material.
The concentration of holes is temperature- When electrons move toward the in-
dependent and is equivalent to the intrinsic terface, lattice sites from where electrons
concentration of the carrier (which is are originated become positively charged
related to the concentration of Frankel de- (Fig. 4b). Creation of positively charged
fects). It can be shown mathematically that carriers occurs randomly in the semicon-
the Fermi level of minority carrier lies at ductor (shown by a curved vertical line in
almost half the band gap position. On the Fig. 4b), but are situated not very far from
other hand, the concentration of majority the interface. Any mathematical treatment
carriers as well as the Fermi level depends for such random distribution or even
on doping concentration. Thus, the Fermi approximated exponential distribution of
level of the majority carrier can lie any- charged ions becomes a complicated sys-
where between the conduction band edge tem to deal with.
and the intrinsic Fermi level that is situated Hence, a simplified mathematical model
at ∼ 12 Eg . is visualized. It is assumed that all
Similarly, the redox electrolyte can positively charged species instead of being
also be viewed as a material possessing scattered are situated in one hypothetical
Vacuum level before forming junction

−Ve EA
+e +e +e +e +e +e
EF
+e +e +e +e +e +e E Fredox
+e +e +e +e +e +e
(a) Semiconductor electrolyte

level after forming junction
m
u
cu
Va
O
V +
−Ve EA + +
e +
e+ + +
e e+ +
+e +e +e + + E Fredox
e e +e + + + +
+e +e +e + +
+ ++
e e +e
+e +e + +
(b)
+ wn
e +e + E Fredox
+e +e +e + +
+
e +e +
+e +e +e + e
+
+
e +e + +
+e +e +
(c)
wn
+ +
e + +
+e +e +e + + + E redox
+
Lp +
+e +e +e +e
+
e e
+e + + +e
x
(d) −Lp −Wn O
Fig. 4 Schematic diagrams showing the formation of band bending
at a semiconductor–electrolyte interface. (a) A semiconductor
possessing its charge homogeneously distributed. +/e refers to a
situation in which lattice sites are neutral. A horizontal line of vacuum
scale showing no variation in energy of electrons of semiconductor.
(b) The contact of semiconductor with redox electrolyte disturbs the
neutrality, and nonhomogeneous charge separation occurs. Vacuum
scale takes the same shape as that of semiconductor to maintain its
electron affinity. (c) A hypothetical model showing charges retained in
two parallels planes separated by a distance w. (d) Creation of
diffusion region as a result of accumulation of minority carriers at
plane situated at distance x = −w.
plane at a distance ‘‘w’’ from the interface all charges present within the distance of
(Fig. 4c). We can get a better idea of x = 0 and x = −w are regular and fully
this hypothetical plane by considering ionized. Moreover, within the space charge
various positive charges q1 , q2 , q3 , . . . qn region, there are no free carriers. In other
situated at x1 , x2 , x3 , . . . xn distances from words, entire lattice sites falling within
the interface of the semiconductor. The the space charge region are fully ionized.
total electrical field created between the It is also assumed that electrostatic field
negative charges situated at the interface becomes zero beyond x = −wn .
(i.e. at x = 0) and positive charges situated Should the changes in distribution of
at various distances would be sum of electrons (majority carriers) in n-type be
the product of charges divided by their abrupt at the interface as compared to
respective distances. the bulk or should it change linearly or
That is, total electrical field created exponentially? The most reasonable as-
q1 e − q2 e− qn e − sumption would be that the change in con-
(F) = + + ··· + centration of majority carrier between the
x1 x2 xn
bulk and interface is exponential. This sit-
Q+ Q−
total total uation can be illustrated by an exponential
≈ (7)
w bending in the conduction band (Fig. 4).
where Q+ total is total positive charge of
Similarly, the valence band must also bend
minority carriers, equivalent to the total in a similar fashion to maintain the differ-
charge of majority carriers accumulated ence between the conduction band and va-
at the interface (Q− total ) and ‘‘w’’ is the
lence band, which is equal to the band gap
distance of the hypothetical plane from of the semiconductor. Since difference be-
the interface (x = 0), which contains Q+ total
tween the Fermi level and conduction band
total charge, such that (Q+ −
total Qtotal )/w =
is related to the magnitude of dopant’s con-
F. It is worth noting that ‘‘w’’ is not centration, this difference must be main-
equal to the sum of all distances (i.e. tained even after the formation of band
w ' = x1 + x2 + x3 + · · · + xn ). bending. However, it is difficult to present
This hypothetical visualization of charge it pictorially. Hence, we compromise by
distribution near the interface (i.e. its pres- depicting it as a horizontal broken line.
ence in one plane) reveals the possibility of The position of the conduction band
formation of charges in two parallel planes. on the vacuum scale also reflects the
One plane appears at the interface (which magnitude of electron affinity of the
is populated with the majority carriers at semiconductor. This value is not altered
x = 0) and another at a depth w (which because the semiconductor has come in
is populated with charges equivalent to contact with a redox electrolyte. Assuming
total charges developed at x = −w). The the vacuum level to follow a similar
depth between these two planes, (Fig. 4c), bend as the conduction band solves this
is called the space charge width designated problem (Fig. 4). Considering all these
as wn (for n-type) and wp (for p-type). It factors, the entire energy diagram of the
should be understood that concentration semiconductor after making its contact
of minority carriers at x = −wn would with redox electrolyte may be expressed as
be much greater than the intrinsic con- shown in Fig. 5. The variation in charges
centration of hole at x ≈ −α. In this and potential within the space charge
mathematical model, it is assumed that region are shown in Fig. 6.
Vacuum scale
−Ve
Dark Illuminated
E E
−
Ec
E N Light
w nE F
Ec nv
Eredox ∆E N Eredox
EF N
pE F +
Ev
+
Ev Redox Redox
(a) n -type semicond electrolyte n -type semicond electrolyte
E
Ec E Vacuum scale
−
Ec −
EF N
w
Eredox nE F Eredox
Ev
∆E N
EF pE F
Light
Ev
+
Redox Redox
(b) p -type semicond electrolyte p -type semicond electrolyte
Fig. 5 Nature of band bending in the dark and under illumination with
(a) an n-type semiconductor-electrolyte and (b) a p-type
semiconductor–electrolyte interface. Vacuum scale changes its shape to
match the shape of band bending. Variation in Fermi level for minority
carriers is shown by broken line (for illuminated semiconductor).
Space charge
+ layer
Charge
−
(a) Gouy layer
Fig. 6 Schematic diagrams showing

charge and potential distribution in an
n-semiconductor–electrolyte interface. Potential
(a) Variation in charge distribution
within the space charge region and in x
Gouy layer. (b) Variation in potential −wn O
within the space charge region, and
electrolyte. In Helmholtz layer it
changes linearly, whereas in Gouy n -semiconductor Helmholtz Electrolyte
layer it decreases exponentially. (b) layer
4.1.6 in magnitude, an equilibrium condition

What Kind of Force Exists in this Region? is established. This condition limits the
concentration of majority carrier at the
As per this model, minority carrier con- interface.
centration at x = −wn would be larger
than the intrinsic carrier concentration. 4.1.7
Moreover, it is also reasonable to assume Magnitude of Potential Developed at the
that intrinsic carriers are homogeneously Interface
distributed in the bulk region. Therefore,
the total magnitude of charge present at What would be the magnitude of potential
x = −wn must also decrease exponen- developed between the charges accumu-
tially until its concentration becomes equal lated at the interface (x = 0) and at plane
to the intrinsic equilibrium concentration x = −wn (or for p-type semiconductor at
(Fig. 4d). Let us assume that at x = −Lp x = −wp )? The driving force for these car-
we establish such a position, then the riers depends on the magnitude of the po-
depth between x = −wn and x = −Lp is tential difference between the Fermi level
called the diffusion region. In the diffusion of the semiconductor and the redox poten-
region, thus we have an exponential con- tial of the electrolyte (i.e. EF − EF(redox) ).
centration gradient established (Fig. 4d). This potential is known as the contact po-
What are possible reasons that prevent tential (θ). Can we fabricate a PEC cell,
majority carrier to increase to an infinite which gives a contact potential equal to the
value at x = 0? There must be an opposing band gap value of the semiconductor? In
force operating to establish an equilibrium other words, can we form a PEC cell with
condition. What is that force? When a a semiconductor (whose Ec ≈ EF ) and re-
semiconductor comes in contact with dox electrolyte (whose Eredox ≈ Ev ), such
redox electrolyte, the majority carrier tends that the contact potential (θ) = Eg ? The
to move toward the interface. When approximate Fermi level of the semicon-
these carriers move toward the interface, ductor (i.e. the intrinsic semiconductor) is
they create similar number of oppositely approximately equal to half the band gap
charged carriers (which can be termed as of the semiconductor (i.e. 12 Eg ). Therefore,
minority carriers). These minority carriers redox electrolyte cannot lower the Fermi
oppose the flow of majority carriers toward level of n-type semiconductor beyond 12 Eg .
the interface. Since no mobile carriers can This condition puts a restriction to the
be stationed in the space charge region, maximum achievable contact potential (θ),
the concentration of minority carriers has and is equal to 12 Eg value. This also sug-
to be established in the diffusion region gests that for a given semiconductor, the
only. Thus, we see that the force that most suitable electrolyte would be the one
allows establishment of an equilibrium that has a redox potential that is almost
condition (or the force that opposes equal to the intrinsic Fermi level of the
the flow of majority carrier toward the semiconductor.
interface) is the force that allows the flow of In conclusion, we see that when a
freshly generated minority carriers toward semiconductor comes in contact with a
x = −α. The flow of minority carrier redox electrolyte, two types of regions are
is also referred to as drift flow. When formed: a space charge region of width
these two confronting forces become equal ‘‘w’’ and diffusion region of width ‘‘L,’’
in addition to its bulk region. In the bulk kind to prevent them from recombining.
region, there is no accumulation of charge. Thus, the net result of these illuminations
is that photogenerated minority carriers
4.1.8 accumulate at the interface and photogen-
Effect of Junction Illumination erated majority carriers accumulate at the
bulk of the semiconductor.
We shall now examine the effect of illu-
While photogenerated holes at the
mination of these three regions separately.
interface perform oxidation of the redox
When photon of energy greater than band
electrolyte, photogenerated electrons move
gap (known as band gap light) falls in the
via the load toward the counter-electrode to
space charge region, electrons from the
perform reduction of the redox electrolyte
valence band get excited to the conduction
(Fig. 1). Thus, electrical energy is created
band, leaving behind a positive charge (i.e.
from light energy.
hole) in the valence band. These carriers
experience an electrostatic potential in the 4.1.9
space charge region. The magnitude of the Efficiency of Conversion of Light Energy
electrical field is very high in this region. into Electrical Energy
For example, with a contact potential of
1 V and space charge width of 1000 Å, an After constructing a PEC cell, it is neces-
electrostatic field of around 105 V/cm is sary to find the efficiency of conversion
formed. This field is strong enough to pre- of light energy into electrical energy. For
vent the recombination of photogenerated this purpose, a PEC cell is connected
electron/holes pairs in the space charge with a resistance (as a load), variable
region. As a result, photogenerated holes D.C. power source and an ammeter in
migrate toward the interface and elec- series (Fig. 7a). A current-voltage charac-
trons migrate toward the diffusion region teristic is measured by applying different
(Fig. 5). When illumination is conducted potentials to the PEC cell and measur-
in the diffusion region, electron/hole pairs ing the corresponding current. This is
also may be formed. In this region, there called dark current (Fig. 7b). The PEC cell
is only equilibrium drift current (estab- is then illuminated with a constant source
lished due to flow of minority carrier of light. The current-voltage characteristic
toward the bulk of semiconductor) and is again measured as before. This cur-
no electrostatic field. Therefore, photogen- rent is the photocurrent (Fig. 7b). It is
erated electrons/holes of diffusion region observed that the dark current is constant
experience only the drift current as a driv- in the reverse bias condition. This cur-
ing force to prevent their recombination. rent is known as the saturation current
To maintain the equilibrium condition, (I o ). The magnitude of the saturation cur-
rent (Ipho ) increases when the PEC cell is
electrons are directed to move toward the
bulk and holes toward the interface. Drift illuminated. The difference between the
current being not very strong, the major- two saturation currents is the photocur-
ity photogenerated carriers in this region rent (Iph ) that is produced from the PEC
undergo recombination. Finally, if bulk cell. However, the efficiency of power con-
region is illuminated, photogenerated elec- version by the PEC cell is calculated from
trons/holes would undergo 100% recom- the current-voltage characteristics, plotted
bination, as this region has no force of any in the fourth quadrant. It is observed that
R Fig. 7 Schematic diagram for

(a) measuring current-voltage
characteristics of a PEC connected in
PEC series with a load (R) and a variable dc
Cell power source and an ammeter (A).
(b) current-voltage characteristics of a
PEC cell under dark and illuminated
A
condition (c) current-voltage
dc Power characteristics (Fig. 7b) of a PEC cell
(a) source
replotted assuming its photocurrent to
be a positive quantity.
on
rk
inati
Da
I
Illum
−Ve V max
Io V
V oc
Imax
I Ph P max
I oPh
I sc
−I
(b)
I sc
FF = 1
Vmax P max
Photocurrent
I max
Voc
(c) Photovoltage
the maximum photopotential is obtained from the PEC cell, a square or rectangle
when photocurrent is zero. This poten- with maximum area is drawn (Fig. 7b,c).
tial is known as the open-circuit potential The intercept of the square at the current-
(Voc ). Similarly, a maximum photocurrent voltage curve is the magnitude of maxi-
is observed when photopotential is zero. mum power (Pmax ) that can be drawn from
This photocurrent is known as short-circuit the PEC cell. Similarly, the intercept on the
current (Isc ). One would get zero power potential axis is the magnitude of the max-
by working at these values. To find the imum photopotential (Vmax ) and intercept
maximum power, which can be drawn on current axis is the value of maximum
photocurrent (Imax ) that can be drawn space charge width (w). While the mate-
from the PEC cell. These characteristics rial should have large space charge width
are obtained when the PEC cell is con- (≈1/α) to ensure entire absorption of band
nected in series with a load of resistance gap light in the space charge region, the
(R). To determine the maximum power diffusion region should also be large. This
load that can be connected to the PEC cell, is because the magnitude of photocurrent
the current-voltage curve is drawn for var- flowing through the semiconductor de-
ious resistances, and the corresponding pends on the width of the diffusion region.
Pmax is determined. The resistance that It will be seen in the forthcoming section
gives the maximum Pmax is taken as the that concentration of dopant as well as
maximum load resistance (Rmax ) that can be the redox electrolyte forming the junction
connected in series with a PEC to drive the can control the width of the space charge
maximum power. The power efficiency of region. Moreover, the FF should also ap-
the PEC cell is calculated from the ratio of proach to unity. Both the space charge
Pmax to the power input (Pin ) to the cell. width and the FF depend mainly on qual-
The nature of the junction formed between ity of semiconductor-electrolyte junction.
the semiconductor and the electrolyte can However, width of the diffusion region is
also be obtained from the current-voltage the specific property of the material. Fur-
characteristic (Fig. 7b,c). This is obtained thermore, the contribution of redox elec-
from a factor known as fill factor (FF), trolyte in giving a better efficiency of a PEC
which is given by: cell is significant and is later discussed.
Imax Vmax 4.1.11
FF = (8)
Isc Voc Energy Levels in Redox Couples
FF decides how well the curve approxi- 4.1.11.1 How Do We Compare Redox
mates a rectangle. In an ideal condition Potential with Fermi Level of a
FF = 1. Generally, current-voltage curve Semiconductor?
instead of plotting in fourth quadrant is At the gas-solid interface, the vacuum
plotted in the first quadrant (Fig. 7c). Note, level is taken as the reference energy
that the photocurrent, although shown as a level; the energy of an electron in vacuum
positive current in figure, depends whether is taken as zero (i.e. Evaccum = 0). The
an n-type (i.e. negative current) or p-type energy of a bound electron in the solid
(i.e. positive current) semiconductor is uti- is referred as a negative value in this
lized. reference scale. In electrochemistry, the
4.1.10
standard zero level of the energy is the
Factors Considered in Selecting a potential of hydrogen ions (at unit activity),
Semiconductor which is in equilibrium with hydrogen
gas at one atmosphere pressure. The
From the previous discussions, it is clear electrode potential of any other redox
that material for a PEC cell must be able to electrolyte is referenced to this potential.
absorb band gap light in the space charge With this hydrogen scale, the value of
region. In other words, the reciprocal of redox potential can be either positive or
the absorption coefficient (i.e. 1/α) of band negative. Conventionally, positive redox
gap light preferably should be equal to the potentials are shown below zero, whereas
(ev)
Vacuum 0.0
−0.2
−0.2
−4.5 0.0 (NHE)
+0.2
(SCE)
+0.4
Ec −5.1 +0.6
Fe+2/Fe+3
+1.0
−5.5 +1.2
Ev
(a) n -semiconductor Electrolyte
E (ev)
Density of state
+1.0
W
re
d (E
)
+0.5
E ox (Unoccupied level)
E redox 0.0 2λ
E red (occupied level)

−0.5
)
(E
x
Wo
−1.0
(b) Semiconductor Electrolyte

Fig. 8 Comparison of energy levels of redox electrolyte with that of semiconductor.
(a) Band diagram of a semiconductor (in vacuum scale) and redox potential of electrolyte
(in normal hydrogen scale (NHE)). Redox potential is shown to possess one value for its
oxidized/reduced species. (b) Energy levels of redox electrolyte showing its distribution
and variation in energy for oxidized and reduced levels.
negative redox potentials are shown above semiconductor and back, it would come
the zero (Fig. 8a). back to the same energy level of the
When dealing with semiconductor-elec- electrolyte. Since the charge density on
trolyte junction, we have to consider both ions is different, the extent of hydration
of these energy scales. Therefore, there is differs. Therefore, it has been proposed
a need to find out the relationship between that the energy states of reduced and
them, so that one can draw useful conclu- oxidized species should not be discrete but
sion from the band diagram of the semi- are broadened into thermal distribution
conductor–electrolyte interface (Fig. 8a). of states caused by thermal fluctuation,
Considering the free energy changes interaction with solvents, and interaction
associated with the reduction of silver, with directly bonded ligands. The position
one can derive the relationship between of the localized energy states in a polar
Evacuum and the NHE. The energy involved solvent depends greatly on whether the
with various reactions can be presented by state is occupied or not because of
considering the following four steps [8]: the strong interaction with the solvent.
Ag+ − Fluctuation in the structure of hydrated
(g) + e = Ag(g) −7.64 eV
ions thus extends the energy of the
Ag(g) = Ag(s) −2.60 eV
electronic quantum states in the redox
Ag+ = Ag+ +4.96 eV
(aq) (g) couple around two most probable values.
Ag(s) + H+ (aq) = Ag+ 1
(aq) + 2 H2 +0.80 eV The energy change caused by the variation
H+ + e− = 1
H2 −4.48 eV in solvation shell can be represented in
2
(9) terms of harmonic oscillators. On the basis
In this analysis, ‘‘g’’ indicates the gas of the concepts proposed by Levich [13]
phase, ‘‘s’’ the solid, and ‘‘aq’’ the aqueous and the distribution function proposed by
solution phase. Thus, the relation between Gerischer [14–16] a model for the energy
the two scales, namely, Evacuum and NHE levels in solution is discussed in this
can be represented as: section. In this model, the ground state
vibration level of only the first electronic
o
EF(redox) (eV) = Eredox(NHE) (volt) state is considered relevant to electron
−4.48 (eV) (10) transfer processes at the electrode. The
o
Eredox(NHE) and EF(redox) are the potentials energy change caused by vibrations in
of a redox couple on the electrochemical the solvent shell is represented in terms
NHE scale and vacuum scale, respectively. of harmonic oscillations. The potential
energy is considered to vary around
the ground state to an equal degree.
4.1.11.2 Do the Energy Levels for
Reduction and Oxidation Process Possess
The resulting distribution functions for
the Same Value? oxidized and reduced species are given as:
The energy level of the redox electrolyte
o )2 /4πkT λ}]
[exp{−(E − Eox
shown in Fig. 8(a) is an oversimplified Wox (E) =
model because it indicates that the energy (4πkT λ)1/2
levels for the reduced and oxidized species (11)
of the redox couple, Fe2+/3+ , are the same. o )2 /4πkT λ}]
[exp{−(E − Ered
This also means that if electrons were Wred (E) =
(4πkT λ)1/2
to be transferred from electrolyte to the (12)
where Eox o and E o are the most probable must match with energy level of the redox
red
energies for the unoccupied and occupied electrolyte. For example, an electron trans-
quantum states in solution, respectively. ferred from the valence band to redox
The parameter λ represents the activation electrolyte would be facilitated if the level
energy for the process of transferring the of the former matches the oxidized level
solvation shell structure from equilibrium of the later. Alternatively, if electrons are
condition of one species to the most to be transferred to the conduction band
probable structure of the other. This is from the redox electrolyte, the conduction
also called reorientation energy. The density band must match the reduced level of the
of states for reduced and oxidized species redox electrolyte. Thus, knowledge of such
at a particular energy are given as distribution functions of the redox elec-
trolyte helps in understanding the I–V
Dred (E) = Cred × Wred (E) (13) characteristics of a semiconductor, espe-
Dox (E) = Cox × Wox (E) (14) cially when the λ value of the electrolyte
approaches the material’s band gap. Under
where Cred and Cox are the concentrations such a condition, the rectifying charac-
of the reduced and oxidized species, ter of semiconductor-electrolyte junction
respectively. The energy distribution is does not control the nature of anodic and
a Gaussian bell shaped curve (Fig. 8b). cathodic current, that is, an n-type semi-
In the harmonic oscillator approximation, conductor would give a cathodic current
the reorientation energy of reduced and in addition to the anodic current, and the
oxidized species are equal and are given by reverse would be true for a p-type semi-
o o o conductor. With this type of condition, one
Eredox = Eox − λ = Ered +λ ···
either has to select an electrolyte whose re-
(15)
dox potential (i.e. its Fermi level) is such
or, that only one of the two levels of the redox
o or E o as the requirement
electrolyte (Eox
o o
Eox − Ered = 2λ ··· (16) red
may be) matches the concerned band edge
Bockris and Khan [17], however, have level of the semiconductor. Alternatively,
suggested that energy is not symmetrically bias would have to be applied to the elec-
distributed around the central point (i.e. trode to stop the unwanted current.
the ground state) but in a Maxwell type
distribution. 4.1.11.3 Effect of Charge Distribution
The most interesting concept derived in the Electrolyte
from these views is that like a semicon- What is the role of the constituents of
ductor, the oxidized and reduced species the redox electrolyte present at the in-
of redox electrolyte are linked with a con- terface of the semiconductor-electrolyte
duction band (unoccupied electronic state) junction? Choice and optimization of the
and a valence band (occupied electronic electrolyte has a substantial effect on PEC
state). The energy necessary to transfer an solar cells [18–19]. Before a semiconduc-
electron from reduced to oxidized state is tor is immersed in an electrolyte, anions
2λ, analogous to the band gap of a semi- and cations freely and randomly move in
conductor Eg . the solution. As a result of this movement,
For an effective charge transfer to occur, no specific spatial accumulation of ions
the energy level of the semiconductor occurs in the solution. This situation alters
as soon as we insert a semiconductor and the bulk region, (where there is no

into the electrolyte. Because of the dif- accumulation of charge).
ference between the redox potential of the
electrolyte and the Fermi level of the semi- 4.1.12
conductor, accumulations of the majority Capacitance of the Space Charge Layer
carriers occur at the interface of the semi-
What is the role of the various re-
conductor. These majority carriers attract
gions formed on the property of the
(hydrated) cations toward the interface.
semiconductor-electrolyte junction? Can
Since they may be hydrated, instead of
we take some advantages from them to
directly coming in contact with the semi-
procure some useful parameters of the
conductor, they may approach the semi-
semiconductor-electrolyte junction? We
conductor spaced by water molecule. This
shall examine this briefly.
situation creates a double layer at the inter-
It has been seen earlier that knowledge
face: a negative charge at the interface of an
of band edge positions of the semicon-
n-type semiconductor and a positive charge
ductor and distribution of energy in the
(i.e. cation) separated by a water molecule,
redox system are helpful in selecting a
in the solution. Interface of the semicon-
suitable redox electrolyte to get an efficient
ductor that contains the majority carrier
charge transfer across the semiconduc-
can be assumed as one plane possessing
tor–electrolyte interface. The band edge
these charges. At a water molecule distance
from the interface, we have another plane positions of a semiconductor can be deter-
carrying the positive charge of cations. This mined if we know the value of the flat band
plane is termed as Helmholtz plane. This potential (i.e. Fermi level position) and the
double layer (known as Helmholtz double band gap of the semiconductor. The flat
layer) thus develops an electrostatic poten- band position is equivalent to the magni-
tial. Because of formation of the double tude of the biasing potential applied to the
layer, as compared to the bulk solution, a semiconductor so as to make the space
large concentration of cations thus gets charge region (i.e. depletion region) width
accumulated at the interface. In a di- (wn for n-type semiconductor) zero. This
lute solution, however, it is observed that potential is equivalent to Fermi level of the
the concentration of cations (i.e. after the semiconductor. One of the easiest meth-
double layer) decreases exponentially to a ods to determine the flat band potential
point where there is no accumulation of is by studying a differential capacitance of
charge and the solution behaves almost the semiconductor–electrolyte interface.
like the bulk solution. The region be- We have seen that a semiconductor-
tween the Helmhotlz plane and the plane electrolyte junction possesses three re-
where there is no accumulation of charge gions in the semiconductor and three in
is called the Guoy-Chapman diffusion re- the electrolyte. These layers act as parallel-
gion (Fig. 6a). The variation in potential in plate capacitors in series, so that the
Helmholtz region is expected to be linear, resultant reciprocal capacitance is the sum
but it is exponential in Guoy-Chapman of the reciprocal of the capacitance for all
region (Fig. 6b). Thus, as with a semicon- the capacitors.
ductor, we can visualize the formation of
1/C(total) = 1/CSC + 1/Cdiff
three charge regions in the solution: the
Helmholtz region, Gouy-Chapman region, + 1/CH + 1/CG (17)
Capacitance due to the bulk region in junction. Since, semiconductor–metal in-

an electrolyte and a semiconductor are terface behaves similar to semiconductor-
overlooked because there is no charge ac- electrolyte junction, Gärtner’s model has
cumulation in these two regions. It has been successfully applied to PEC cell, as
also been experimentally observed that well [21–22]. The photocurrent (Jscl ) due
magnitude of the capacitance of the diffu- to carriers generated in the space charge
sion layer in semiconductor (Cdiff ), Guoy- region is given as
Chapman layer (CG ), and the Helmhotlz
layer (CH ) are very high compared to that Jscl = qI o (1 − e−αw ), ··· ··· (18)
of the space charge layer (CSC ). Under
this condition, the reciprocal of capaci- where q = charge of the electron; I o =
tance of the space charge layer becomes incident photon flux (number of pho-
equivalent to reciprocal of total capac- tons s−1 cm−2 ); α = absorption coeffi-
itance of the semiconductor-electrolyte cient for photon of energy hν (cm−1 );
junction (i.e. 1/C(total) = 1/CSC ). Hence, w = space charge width (cm). This is
if we can measure the total capacitance designated as wn or wp for n-type and
formed with the semiconductor-electrolyte p-type semiconductor, respectively; Jscl =
junction, understanding of the capacitance photocurrent density (A cm−2 ).
of space charge layer can be obtained. Similarly, an equation for photocurrent
Measurement of the potential drop originated from the diffusion region is
across the space charge layer yields in- given as:
formation about the capacitance formed ! "
Lmin
with the semiconductor-electrolyte junc- Jdiff = −qI o e−αw
1 + Lmin
tion; this capacitance and its measurement # $
are described in a separate chapter. Dmin
− qImin · · · (19)
Lmin
4.1.13
Semiconductor-electrolyte Junction Under where Lmin = width of diffusion layer
Illumination (cm). In the n-type semiconductor, it
will be due to holes (designated as
In Sect. 4.4, we examined the effect of the Lp ) where as in p-type due to electron
illumination of semiconductor-electrolyte (designated as Ln ). Imin = equilibrium
junctions, using a light source of photon concentration of minority carriers (cm−3 ),
energy greater than the band gap of semi- which is designated as po and no in n-type
conductor. We shall now try to develop and p-type semiconductors, respectively.
a model to get a quantitative estimation Dmin = Diffusion coefficient for minority
of the photocurrent from the magnitude carrier (cm2 s−1 ). For holes in n-type
of the concentration of photogenerated it will be designated as Dp and for
carriers. electrons in p-type semiconductor it will
be designated as Dn . Jdiff = photocurrent
..
4.1.13.1 G artner’s Model due to photogenerated carriers in the
Wolfgang Gärtner [20] derived a relation diffusion region.
for photocurrent generated in space Total photocurrent passing through the
charge region and diffusion region af- junction would be sum of these two
ter illuminating a semiconductor-metal components.
Jtotal = {qI o (1 − e−αw )} useful to examine Eq. (22) by keeping one

% ! " of these variables constant.
o αLmin
+ −qI
1 + αLmin
# $& 4.1.13.1.2 When α is Constant during Pho-
Dmin
× e−αw − qImin (20) tocurrent Measurement This condition is
Lmin achieved when the junction is illuminated
For a large band gap semiconductor, Imin , with a monochromatic light and the pho-
Dmin , and Lmin would be very small; tocurrent is measured as a function of ap-
hence, their contribution can be neglected. plied potential (V ), using a potential close
Thus, the Eq. (20) simplifies to to the flat band potential. Substituting
the value w = [(2εεo /eNA )(V − Vfb )]1/2 to
! "
1 − e−αw Eq. (22) we have
Jtotal = −qI o · · · (21)
(1 + αLmin ) # $
2εo ε 1/2
− ln(1 − φ) = α (V − Vfb )1/2
It is important to remember that the qNA
polarity of photocurrent is negative, as one + ln(1 + αLmin ) (23)
would get experimentally (Fig. 7b).
Equation (23) suggests that if − ln(1 − φ)
.. is plotted against (V )1/2 , a linear graph is
4.1.13.1.1 Application of G artner Model
obtained. Intercept of the linear plot on
Equation (21) can be slightly modified to
the potential axis, when − ln(1 − φ) = 0,
include a term ‘‘quantum yield (φ)’’ to get
would give the value of the flat band
information about some useful parameters
potential (Vfb ) because none of the terms of
of semiconductors, for example, band gap,
α(2εo ε/qNA )1/2 (V − Vfb )1/2 except (V −
diffusion length, space charge width, and
Vfb )1/2 can be zero. A word of caution is
so on. Quantum yield is defined as the
necessary in this type of calculation. This
ratio of photocurrent (Jphoto ) to the total
equation is valid only under the condition
flux of light (I o ) used to illuminate the
of band bending. Therefore, it is always
semiconductor: that is, quantum yield
better to extrapolate the linear plot from
(φ) = Jphoto /I o .
the region where (V − Vfb ) > 0, that is,
Equation (21) can thus be rewritten as from the region where forward bias is
e−αw much greater than the Vfb values.
(1 − φ) = Slope of the linear region where (V −
1 + αLmin
Vfb ) > 0 gives the value of α(2εo ε/qNA )1/2
Taking the logarithm of this equation, for the wavelength of light used during il-
we have lumination. If NA is known from Hall
measurements or from capacitance mea-
− ln(1 − φ) = αw + ln(1 + αLmin ) (22) surement, α can be obtained.
If the experiment is repeated at different
In Eq. (22), there are two variables, α (ab- wavelengths of light, one can get the value
sorption coefficient of the semiconductor), of α for different wavelengths also. This
which depends on the wavelength of light, information is extremely useful because
and w, (the space charge width), which 1/α is almost equal to the penetration
is related to the dopant’s concentration depth of the corresponding wavelength of
and magnitude of the bias. Therefore, it is light. It is desirable to know the magnitude
of (1/α) for band gap light, so that of Aw(hν − Eg ) can be made zero except
efforts could be made to select a suitable (hν − Eg ). This would mean that the value
redox electrolyte such that (1/α) ∼ w. of (hν) (i.e. the intercept on (hν)-axis ‘‘s’’)
In addition, we can also calculate Ln should be equal to Eg . If we get the linear
from magnitude of the intercept (i.e. plot with n = 1, the band gap is a direct type
ln(1 + αLn )) of the linear plot on y-axis. and if n = 4, it is an indirect one. Some
From Ln the lifetime of the carrier (τ ) can be semiconductors may give linear plot for
calculated by using the following equation both values of n, indicating simultaneous
# $ presence of both types of band gap.
µn kT τ 1/2 Thus, application of the Gärtner model
Ln = (24)
q to data obtained from photocurrent ver-
sus potential measurement yields several
µn the mobility of the carrier, can be
important parameters of a semiconductor-
obtained from Hall measurement.
electrolyte junctions (e.g. Eg , Vfb , Lmin , τ ,
From these measurements, we can
and w).
thus evaluate the flat band potential
(which is equal to EF for heavily doped
4.1.14
semiconductor), absorption coefficient (α)
Effect of Counter-Electrode
for band gap light, diffusion length (L) and
lifetime of the carrier (τ ).
Besides the considerations on the param-
eters of the semiconductor and the redox
4.1.13.1.3 When ‘w’ is Constant During electrolyte, performance of the counter-
Photocurrent Measurement Using α = electrode also matters for designing an
{A(hν − Eg )n/2 }/ hν and keeping w con- efficient PEC cell. The counter-electrode
stant, Eq. (22) becomes in a PEC cell is a charge carrier collecting
' ( material, and catalyzes the electron trans-
A(hν − Eg )n/2 fer, but is not chemically involved in the
− ln(1 − φ) =
hν electrochemical reaction.
× w + ln(1 + αLmin ) (25) The counter-electrode material should
be electrically conducting, economical,
When (αLmin ) ( 1, Eq. (25) can be rear- and chemically stable in an electrolyte.
ranged to get It should minimize polarization losses
during exchange of electrons at the
−[{ln(1 − φ)}hν]2/n = Aw(hν − Eg ) interface. In other words, polarizability of
(26) the counter-electrode should be small in a
In order to utilize Eq. (26), a semiconduc- given redox electrolyte, indicating that the
tor-electrolyte junction is illuminated by electrode reaction occurring at the counter-
different wavelengths of light, keeping the electrode is reversible and fast. This
applied potential to a PEC cell constant maximizes the potential obtained from the
(i.e. under a constant forward potential). semiconductor–electrolyte interface.
A plot of [{ln(1 − φ)}hν]2/n versus hν For many redox reactions, noble met-
would thus result in a linear graph. Let als such as Pt, Au have been preferred
us assume that the value of the intercept as counter-electrode by electrochemists.
on the x-axis at [{ln(1 − φ)}hν]2/n = 0 is However, those are expensive for ter-
‘s’. At this condition, none of the terms restrial application, and materials such
as carbon (graphite, glassy carbon), and α-PbO/Fe (CN)6 4− (0.1M), Fe(CN)6 3−

degenerated transparent [23] semiconduc- (0.01M) pH 9.2/(counter-electrode). They
tors (SnO2 or ITO) have attracted the observed that while platinum counter-
attention of scientists for their use as electrode gives the best performance
counter-electrodes. For aqueous polysul- (Fig. 9a), surface modification of ITO
fide redox reactions, Hodes and cowork- counter-electrode with platinum (Fig. 9b)
ers [24] have shown that the use cobalt gives almost similar performance to that
sulfide as an electrocatalyst gives much obtained with platinum. However, un-
better performance than high surface area treated highly conducting ITO electrode
platinum or carbon electrodes. gives considerably low power charac-
When a counter-electrode other than teristics (Fig. 9c). Transparent counter-
inert platinum electrode is used, it is electrode, in addition, avoids the problem
necessary to confirm whether the elec- of light absorption by the redox elec-
trochemical reaction occurring at the trolyte [25]. With a transparent counter-
counter-electrode is as reversible as a electrode, it is possible to place the
platinum counter-electrode [25]. Sharon semiconductor very close to the counter-
and coworkers [26] has studied the ef- electrode, which also serves as a window
ficiency of three types of counter- for the PEC cell. This backwall config-
electrodes (ITO, Pt and Platinum-coated uration is especially helpful when very
ITO) in a PEC cell of configuration: concentrated electrolytes are used [27].
0.006
(a)
0.005
(b)
0.004 Pmax
(c)
0.003
[Amp]
0.002
I
0.001
0.000
−0.001
−0.002
−0.2 0.0 0.2 0.4 0.6 0.8 1.0
V
[Volts]
Fig. 9 Current-voltage characteristic of a PEC cell (α-PbO/Fe (CN)6 4− (0.1M), Fe(CN)6 3−
(0.01M) pH 9.2/(counter-electrode). Three types of counter-electrodes were used:
(a) conducting ITO electrode, (b) surface modified ITO electrode using platinum and
(c) platinum electrode.
4.1.14.1 Effect of Surface States charge, it is possible to neutralize their

It is important to appreciate that even if charges by controlling pH of the solution.
a material possess desirable characteris- Sharon and coworkers [28] have devel-
tics, as discussed in the previous sections, oped a simple technique to find the pH
it may not exhibit the desired efficiency at which the effect of surface states is
of light energy conversion into electrical minimum. Their model also gives infor-
energy. For example, in spite of semi- mation regarding the magnitude of other
conductor and redox electrolyte possessing parameters such as flat band potential and
desired Fermi levels experimentally deter- band bending potential. The experiment is
mined contact potential might be much based on the fact that concentration of a
less than the anticipated value. Presence of majority carrier at the surface of the semi-
surface state is one of the factors respon- conductor (in dark condition) is controlled
sible to show these deficiencies. Hence, by the extent of band bending formed in
efforts are required to minimize the ad- the space charge region. Similarly, free-
verse effects of surface states. dom of majority carrier to move freely
Surface states mainly originate from over the surface also depends on the na-
the lack of balance in the valence states ture and magnitude of the surface states
of atoms/ions present at the surface as present over the surface. Therefore, the
magnitude of surface states present on the
compared to atoms/ions present beneath
surface influences the surface conductiv-
the surface of thin film or single crystal.
ity. The activation energy for the surface
Consequently, atoms/ions present at the
conductivity should thus also be related to
surface possess either more or less number
the nature and magnitude of the surface
of electrons as compared with those
states.
present in the bulk material. This situation
For surface conductivity measurement,
creates surface states of various types. All
two ends of a thin film of semiconductor
of them behave like a donor or an acceptor
are connected with platinum wires. The
sites on the surface. Thus, irrespective of entire surface except the top surface of the
the type of surface states, they tend to trap film is covered with an insulating paint.
photo-generated carriers at the surface of This film is then immersed in a solution
the semiconductor. These trapped carriers of known pH whose temperature is con-
are released at a slow rate. Moreover, these trolled very accurately. Surface resistance
surface states can also cause pinning of is measured versus temperature (normally
the Fermi level. Pinning of the Fermi level in the range of −10 ◦ C to +25 ◦ C). This
takes away the freedom of Fermi level experiment is repeated at different pH
to adjust its position while attaining the values. Log of the surface conductivity is
equilibrium condition. Thus, pinning of plotted versus reciprocal of the absolute
the Fermi level decreases the magnitude temperature to get the activation energy.
of anticipated contact potential. Hence, for This calculation is carried out for each
an efficient PEC cell, the semiconductor pH. Finally, a plot of activation energy ver-
should be free from surface states. It is sus pH is plotted, which shows either a
impossible to fully eliminate surface states. minimum with n-type semiconductor or
However, this effect can be minimized. a maximum with p-type semiconductor.
Since these surface states behave either These pH values correspond to point of
as a site possessing positive or negative zero charge (pzc) of the material.
−8.4
−8.0 30°C 630
In conductivity × 10−2
21°C
−7.6 12°C 610 0°C
4°C 5°
−7.2 2°C 590 10°
In σ
−9°C 15°
−6.8 570 20°
−6.4 550 25°
−6.0 530
−5.6 510
5.4 5.8 6.2 6.6 7.0 7.4 7.6 8.2 8.6 9.0 9.4 0 1 2 3 4 5 6 7
(a) pH (b) pH
Barrier height (Activation energy)
Barrier height (Activation energy)

0.6 0.16
0.5 0.14
0.12
0.4
0.10
[eV]
[eV]
0.3
0.08
0.2
0.06
0.1
0.04
0.0
0.02
0 1 2 3 4 5 6 7 8 9 5 6 7 8 9 10
(c) pH (d) pH
Fig. 10 Surface conductivity measurements of thin films of α-Fe2 O3 and
p-SnSe21 . Log of surface conductivity versus pH at different temperatures for
(a) n-α-Fe2 O3 (b) p-SnSe. Activation energy of surface conductivity versus
pH for (c) n-α-Fe2 O3 and (d) p-SnSe.
In Fig. 10, typical results of thin films n- because the surface of the semiconductor
α-Fe2 O3 (a,c) and p-SnSe (b,d) are shown. would have no excess charge for pinning
n-α-Fe2 O3 gives a minimum at pH 4.1, the Fermi level. Moreover, photogenerated
which is equivalent to its pzc. A reverse carriers will also not be trapped by the sur-
behavior is obtained with p-SnS showing face states as their charges are neutralized
a maximum at pH 7.5, which is equivalent by pH of the solution.
to its pzc. This is the only nondestructive
method available to get information about 4.1.15
pH that corresponds to pzc for large area Photodecomposition of Semiconductor
thin-film semiconductor material.
A PEC cell operating at the pH equiv- In the dark many semiconductors are sta-
alent to the pzc will experience the least ble in contact with a given electrolyte, but
hindrance caused by the surface states, and start decomposing on illumination of the
may be able to show its anticipated PEC junction. This and energy conversion effi-
behavior. PEC operating at pzc would be ciency are the most serious problems faced
able to show its predicted contact potential in PEC devices. If the photo-generated
minority carriers are not rapidly used in Unfortunately, most of the low band gap
oxidation or reduction by the electrolyte materials studied so far have been found to
immediately, they have a tendency to re- undergo such corrosions. Therefore, there
act with the semiconductor material. The is a need to develop newer materials for
semiconductor electrode therefore starts developing a photoelectrochemically stable
corroding. PEC etching is being dealt semiconductor.
in a greater depth in a separate chap-
ter. However, the easiest way to establish 4.1.15.1 Materials for PEC Cell
PEC stability of the semiconductor is by Because of the limitation of space, it is
measuring the photocurrent of a PEC cell difficult to discuss details of materials
over a longer period under continuous il- studied for various types of PEC cell. Some
lumination. If semiconductor is stable, the of the materials and the redox electrolytes
photocurrent will show a constant value used for studying PEC cells are tabulated
over a period of illumination. in Tables 1 and 2. Nevertheless, Sharon
Tab. 1 Some of the chalcogenide materials studied for developing PEC cell in various redox
electrolytes
Material Electrolyte system Reference
n-CdS NaOH,S,Na2 S; NaOH,Na2 Te; 31–33, 35, 39

NaOH,Na2 (SCH2 COO)S; KCl,
KFe(CN)6 3−/4− ; NaI, I2 in CH3 CN
n-CdSe NaOH, Na2 S, S; NaOH, Na2 Se; NaOH, 31–33, 36–38, 40–42
Na2 Te; NaOH, Na2 S,S, Se;
NaOH,Na2 (SCH2 COO)S, Se;
KOH,Na2 S,Sn ; Fe(CN)6 3−/4− pH > 8,
Cesium polysulphides, S,CuSO4
n-CdTe/p-CdTe NaOH-Na2 Se, NaOH-Na2 Te, K2 Se-Se-KOH 34, 35, 40, 43
n-CdInS2 , NaOH-Na2 S-S 44–45,100
CuInSe2 , CuInS2
n-Re6 Se8 Cl2 I3 − -I− -H2 SO4 46
n-CdIn2 Se4 KOH-S-Na2 S 47
n-CuIn5 S8 Sulfide, Polysulfide, 48
n-WS2 Halide solution, 49
p-WS2 CH3 CN/(n-Bu4 N)ClO4 , 50
n-MoS2 Et2 N-halide ion in CH3 CN, NaI-I2 (aq), 51, 55, 56
Fe2+ /Fe3+ ,
n-MoSe2 I/I2 , Fe2+ /Fe3+ , (Et4 N)X/2 in CH3 CN; 52–53, 58
X = halogen, HBr/Br2 pH = 1
n-WSe2 Fe(CN)3−/4− , NaI/I− −
3 -Na2 SO4 ; 52, 80
H2 SO4 ,pH = 0
p-WSe2 Fe2+/3+ -H2 SO4 59
n-SnS Ce4+ /Ce3+ , 0.5 M; H2 SO4 70
n-Bi2 S3 NaOH-Na2 S-S 60–61
Tab. 2 Some of the phosphide/oxide and other materials studied for developing PEC cell in
various redox electrolytes
Compound Electrolyte Reference
n-InP NaOH/Fe(CN)6 3−/4− ; NaOH-Na2 Te, 62

VCl2 -VCl5 -HCl
p-InP NaOH/Fe(CN)6 3−/4− ; NaOH-Na2 Te, 63
VCl2 -VCl5 -HCl
p-GaP VCl2 -VCl5-Na2 SO4 ; pH = 2.5 64
ZnP2 , Zn3 P2 NaOH/Fe(CN)6 3−/4− 65
Fe2 O3 NaOH/Fe(CN)6 3−/4− 66
YFeo3 Fe(CN)6 3−/4− , pH = 7 & 9 67
Pb3 O4 IO3 − /I− 68
BaTiO3 Ce3+ /Ce4+ , Fe2+ /Fe3+ ; pH = 2 69
PbO(2−x) Fe(CN)6 3−/4− , pH = 9 71
n-GaAs Fe2+/3+ ; I− /I3 − ; S2− /S2−
n ; Fe(CN)6
3−/4− 43, 74–76, 79–80
p-GaAs I− /I3 − pH = 0; KCl-VCl2 /VCl; Eu3+ 57, 78, 81–83
p-Si HCl-VCl2 /VCl3 84–87
n-Si FeCp2 ; n-Bu4 N-ClO4 in ethanol; 54, 87, 89–93
FeCp2 = ferrocene; I− /I3 − -FeCl2 /FeCl3
GaAs0.72 P0.28 Nonaqueous electrolyte 78, 81
Tab. 3 Chronological improvements in solar to electrical conversion

efficiencies of some regenerative PEC solar cells
Year Solid/dissolved % Efficiency Reference

salts
1998 AlGaAs/Si/V2+/3+ ∼19 96

1991 Nano TiO2 (dye ∼7 97
sensitized)/I3 −
1990 n-CdSe/KFe(CN)6 3− ∼16 98
1987 n-GaAs/Se2− (Os3+ ) ∼15 99
1985 n-Cd(Se,Te)/S2− ∼13 100
1982 n-CuInSe2 /I− ∼9 101
1982 n-WSe2 /I− ∼10 102
1981 n-MoSe2 /I− ∼9 103
1981 p-InP/V3+ ∼11 104
1978 n-GaAs/Se2 -(Ru2+ ) ∼12 105
and coworkers [29, 30] have extensively Table 3. Highest reported conversion ef-
reviewed these materials. ficiencies include more than 9% for dye
Examples of chronological improve- sensitized thin film of n-TiO2 (nanos-
ments in high solar to electrical conver- tructure) in polyiodide electrolyte [97] and
sion efficiency PEC’s are summarized in more than 16% for single-crystal n-CdSe
in cyanide modified K3 Fe(CN)6 3− elec- efforts to extend the lifetimes of the requi-
trolyte [98]. The synthesis of single-crystal site dyes [97].
semiconductor material is not cost effec- But unfortunately none of these mate-
tive, being both time intensive and limited rials show a long-term PEC stability to
to formation of smaller surface area of warrant making of a commercially viable
materials. A variety of techniques have PEC cell. Interestingly, it has been hinted
been developed to provide polycrystalline that phosphides of Co, Fe, Mo, Ni, V,
or amorphous thin-film semiconductors. and W (band gap in vicinity of 1 eV) may
A particular advantage of electrodeposition be photoelectrochemically stable. Among
(plating) techniques is a one-step chemi- the oxide materials, PbO2 (Eg = 1.6 eV) is
cal synthesis of the semiconductor. PV, worth exploring because it is highly stable
as well as PEC solar cells requires the in acidic medium. Unfortunately, due to
use of thin-film semiconductors to be cost high oxygen vapor pressure, PbO2 exits
competitive. Each of these techniques con- as a degenerated semiconductor (i.e. it be-
sumes relatively small amounts of material haves like a metal). If suitable doping could
and can be applied to develop large surface open the band gap, then PbO2 can become
area electrodes. Modes of film preparation one of the best materials for making com-
include slurry techniques, chemical vapor mercially viable PEC cell. Semiconducting
deposition, wackercast, rf sputtering, vac- carbon is another class of material that
uum coevaporation, sol gel, chemical bath needs to be extensively studied for this pur-
deposition, hot pressing, and electroplat- pose. Earlier, diamond was the only known
ing. These techniques have been used semiconductor of carbon. But in the recent
to prepare thin films of n-CdS, n-CdSe, past, it has been possible to synthesize
n-CdTe, n-GaAs, p-InP, n-Si and p-Si, n- semiconducting carbon with a band gap
TiO2 and used in PEC solar cells [97, 105]. in the vicinity of 1.0 to 2.8 eV. Sharon and
Advantageously, thin film PEC’s have coworkers [107, 108] has recently reported
exhibited photocurrent lifetimes signifi- a PEC cell from a semiconducting carbon
cantly greater than single-crystal PEC’s. made from the pyrolysis of camphor. In
This phenomenon has been attributed addition, efforts should be made to reduce
to the high microscopic surface area, the band gap of stable oxides such as TiO2
and resulting lower effective microscopic into a low band gap material by manipu-
photocurrent density, passing through lating to obtain their mixed oxides. Mixed
the thin-film devices. The longest life- oxides of Mo & W also need similar at-
time PEC device to date, slurry de- tention. Sharon and coworkers [109, 110]
posited thin-film n-Cd (Se, Te) electrodes have tried to develop a theoretical model to
immersed in modified polysulfide so- predict the variation in band gap by varying
lutions, has demonstrated outdoor op- the composition of components in mixed
eration for approximately one year. In oxides. In addition to this, there is a need to
this system, photoelectrode and electrolyte develop a theoretical model, to predict PEC
stability was achieved by optimization stability of materials. Such models will go
of the solution alkalinity, using alter- a long way to minimize the number of ex-
native cations, and modifying the ratio periments needed in developing materials
of dissolved sulfur to sulfide in solu- for making a commercially viable PEC cell.
tion [106]. The recent advance in dye- It is necessary to emphasize that in spite
sensitized PEC devices requires future of the simplicity in fabrication, the future
of economically viable PEC depends en- beam is scanned over the entire surface of
tirely on the development of suitable low the film, and photocurrent is measured at
band gap, photoelectrochemically stable each illuminated point. A 3D plot of pho-
materials. tocurrent versus (x, y) distances is made. If
the composition of materials formed over
4.1.16 the large surface area electrode contains
Laser Scanning Technique for a Large Area photoactive oxide, a uniform distribution
Electrode of photocurrent would be observed. Fail-
ing to get uniform photocurrent would
There are materials, which can form va-
be an indication of nonuniformity of the
rieties of nonstoichiometric compounds.
surface with the desired material. Though
For example, PbO2 can coexist as PbOx
this technique is not specific about the
where x can have values from 0 to 2. composition, it is a very useful tool to give
Among these oxides, most photoactive ox- a guideline for improving the preparation
ide is PbO0.8 . If PbO0.8 oxide is to be technique especially for a large area elec-
prepared by anodic oxidation of lead, it trode. The resolution of this experiment,
is possible that surface of the lead film however, depends on the scanning size of
might contain other forms of oxides as the laser beam. For example, the 3D laser
well. Thus, a PEC cell prepared by anodiza- scanning map of an anodized lead sheet
tion of lead may give a low photoresponse, (2.5 cm2 ) shows that the photocurrent
not because PbO0.8 is an inappropriate ma- is not as uniform (Fig. 11a) as obtained
terial but because anodized surface may be with film anodized by condition shown in
contaminated with other forms of oxides Fig. 11(b).
as well. There seems to have no nonde-
structive technique available to confirm
4.1.17
the uniformity of surface with photoac- Summary
tive species, especially of a large area
electrode. Sharon and coworkers [111] has In this section, efforts are made to explain
recently developed a laser scanning sys- the basic principles of a regenerative PEC
tem to overcome this problem. A laser solar cell. Greater emphasis is made on
y
y
x
x
[a.u.]
[a.u.]
I ph
I ph
30
7.
30
8.4 8.4
0
7.
87
0 5.6
87
4.
5.6 0
4.
0
43
2.8 2.8
43
2.
0 0
2.
0.0
00
0.0 0
00
0.
(a) 0 (b)
0.
Fig. 11 3D photocurrent map obtained with scanning of He/Ne laser (3 mW) over the surface of lead
sheet anodized in the potential range (a) −0.32 to +0.08 V and (b) −0.22 to +0.08 V versus SCE [26].
criteria for selection of a semiconductor References

and a redox electrolyte to make a PEC cell
operating at its maximum performance. 1. A. E. Becquerel, C.R. Acad. Sci. 1839, 9, 561.
2. W. H. Brattain, C. G. B. Garret, Semicon-
Effect of light on illuminating a junction
ductor surface phenomena in Semiconducting
formed with semiconductor–electrolyte materials Proc. Conf. Univ. Reading, Eng-
interface is discussed with a view to charac- land, 1951, 37.
terize various parameters influencing the 3. W. H. Brattain, C. G. B. Garrett, Phys. Rev.
efficiency of a PEC cell. The role of counter- 1955, 99, 376.
4. A. Fujishima, K. Honda, Nature 1972, 238,
electrode and minimizing of the surface
37.
states to achieve a desirable performance 5. H. Gerischer in Semiconductor electrochem-
of a PEC cell are also discussed. Whenever, istry: in Physical Chemistry: An advanced
a large surface area electrode of semicon- treatise (Eds.: H. Eyring, D. Henderson,
ductor of Ax By type is prepared especially W. Jost), 9A, Academic Press, New York,
1970, p. 463.
by an electrochemical method, uniform
6. V. A. Myamlin, Yu. V. Pleskov, Electrochem-
deposition of material over the entire sur- istry of Semiconductors, Plenum Press, New
face with similar composition is always York, 1967.
a problem to establish. A laser scanning 7. M. Sharon, P. Veluchamy, C. Natrajan,
technique has been discussed for this pur- C. D. Kumar, Electrochim. Acta 1991, 36(7),
1107.
pose. Although PEC cells with efficiency
8. S. Licht, G. Hodes, R. Tenne et al., Nature
as high as 16% have been developed, be- 1987, 326, 863.
cause of its instability problem, it has not 9. S. Licht, B. Wang, T. Soga et al., Appl. Phys.
been possible to commercialize. A section Lett. 1999, 74, 4055.
is devoted to discuss the material aspect 10. B. Pal, M. Sharon, D. Kamat, Chem. Envi-
ron. Res. 1996, 5(1–4), 51.
of the PEC and a future direction of re-
11. H. Yoneyama, N. Takahashi, K. Susumu
search needed to fabricate a commercially et al., J. Chem. Soc. Chem. Commun. 1992,
viable PEC. A chronological improvement 9, 716.
made in PEC are also shown in a tabu- 12. A. J. Nozik in Photoelectrochemical devices for
lar form. solar energy conversion, NATO advance study
institute series, Series B; Physics, Plenum
press, New York, 1981, p. 280, Vol .69.
Acknowledgment 13. V. G. Levich in Physical Chemistry – An
Advanced Treatise, (Eds.: H. Eyring, D. Hen-
I am extremely grateful to all my stu- derson, W. Jost), Academic Press, New
York, 1970, Vol. 9B, Chapter 12.
dents who have contributed a lot via 14. H. Gerischer, Surf. Sci. 1969, 13, 265.
their Ph.D. thesis and research work, 15. H. Gerischer, in Special Topics in Electro-
which enabled me to perceive the com- chemistry, (Ed.: P. G. Rock), Elsevier, Ams-
plexity of PEC cell. Their interactions terdam, 1977.
enriched me to evolve a simpler way 16. H. Gerischer, Surf. Sci. 1969, 18, 97.
17. J.Ó.M Bockris, S. U. M. Khan, J. Electro-
to comprehend the intricacy of a PEC chem. Soc. 1985, 132, 2648.
cell. This has immensely helped me in 18. S. Licht, Nature 1987, 330, 148.
penning this section. I am also grateful 19. S. Licht, Sol. Energy Mater. Sol. Cells 1995,
to Dr. Michael Neumann-Spallart, CNRS, 38, 353.
Meudon, France, for going through the 20. W. W. Gärtner, Phys. Rev. 1959, 116, 84.
21. I. Mukhopadhyay, M. Sharon, Electrochim.
manuscript and for his valuable sugges- Acta. 1997, 42(1), 67–72.
tions, which helped me in formulating 22. I. Mukhopadhyay, M. Sharon, Sol. Energy
this chapter. Mater. Sol. Cells 1997, 45, 141.
23. J. F. Gibbons, G. W. Cogan, C. M. Gronet 49. J. A. Baglio, G. S. Calabrese, E. Kamieniecki

et al., Appl. Phys. Lett. 1984, 44, 539. et al., J. Electrochem. Soc. 1982, 129, 1461.
24. G. Hodes, J. Manassen, D. Cahen, J. Electro- 50. J. A. Baglio, G. S. Calabrese, D. J. Harrison
chem. Soc. 1980, 127, 544. et al., J. Am. Chem. Soc. 1983, 105, 2246.
25. R. Tenne, M. Peisach, C. A. Rabe et al., 51. H. Tributsch, Ber. Bunsen-Ges Phys. Chem.
J. Electroanal. Chem. 1989, 269, 389. 1977, 81, 361.
26. M. Sharon, S. Ghosh, J. Solid State Elec- 52. H. Tributsch, H. Gerischer, C. Clemen
trochem. 1999, 4, 52. et al., Ber. Bunsen-Ges Phys. Chem. 1979,
27. S. Licht, F. Forouzan, J. Electrochem. Soc. 83, 655.
1995, 142, 1539. 53. L. F. Schneemeyer, M. S. Wrighton, J. Am.
28. M. Sharon, B. M. Prasad, K. Basavaswaran, Chem. Soc. 1980, 102, 6964.
Indian J. Chem. 1989, 28A, 935. 54. T. Skotheim, L. G. Peterson, O. Inganäs
29. M. Sharon, K. Basavaswaran, N. P. Sathe, et al., J. Electrochem. Soc. 1982, 129, 1737.
J. Sci. Ind. Res. 1985, 44, 593. 55. G. Djemal, N. Muller, U. Lachish et al., Sol.
30. M. Sharon, G. Tamizhmani, J. Mater. Sci. Energy Mater. 1981, 5, 403.
1986, 21, 2193. 56. R. Audas, J. C. Irwin, J. Appl. Phys. 1981, 52,
31. R. N. Noufi, P. A. Kohl, A. J. Bard, J. Electro- 6954.
chem. Soc. 1978, 125, 375. 57. A. J. Bard, F.-R.-F. Fan, J. Am. Chem. Soc.
32. M. S. Wrighton, A. B. Ellis, S. W. Kaiser, 1980, 102, 3677.
J. Am. Chem. Soc. 1976, 98, 1635. 58. H. Tributsch, Ber. Bunsen-Ges Phys. Chem.
33. S. Chandra, S. K. Pandey, R. C. Agrawal, 1978, 82, 169.
Sol. Cell 1980, 1, 36. 59. H. Gerischer, W. Kautek, Ber. Bunsen-Ges
34. W. J. Danaher, L. E. Lyons, Nature 1978, Phys. Chem. 1980, 84, 645.
60. B. Miller, A. Heller, Nature 1976, 262, 680.
271, 139.
61. P. K. Mahapatra, C. B. Roy, Sol. Cells
35. J. M. Bolts, A. B. Ellis, K. D. Legg et al.,
1982–1983, 7, 225.
J. Am. Chem. Soc. 1977, 99, 4826.
62. M. S. Wrighton, A. B. Ellis, J. M. Bolts,
36. R. Noufi, D. Tench, L. F. Warren, J. Electro-
chem. Soc. 1980, 127, 2709.
63. Y. Ramprakash, S. Basu, D. N. Bose,
37. A. Aruchamy, A. Venkatarathnam, G. V.
J. Indian Chem. Soc. 1981, LVIII, 153.
Subba Rao et al., Electrochem. Acta. 1982,
64. M. Tomkiewicz, J. M. Woodall, Science
27, 701. 1977, 196, 990.
38. H. Gerischer, J. Electroanal. Chem. 1975, 58, 65. H. von Känel, L. Gantert, R. Hauger et al.,
263. Hydrogen Energy Prog. 1984, 5, 969.
39. M. Tsuiki, H. Minoura, T. Nakamura et al., 66. M. Sharon, B. M. Prasad, Sol. Energy Mater.
J. Appl. Electrochem. 1978, 8, 523. 1983, 8, 457.
40. A. B. Ellis, S. W. Kaiser, M. S. Wrighton 67. M. Sharon, B. M. Prasad, Electrochim. Acta
et al., J. Am. Chem. Soc. 1977, 99, 2839. 1985, 30(3), 331.
41. R. Tenne, G. Hodes, Appl. Phys. Lett. 1980, 68. M. Sharon, S. Kumar, S. R. Jawalekar, Bull.
37, 428. Mater. Sci. 1986, 8(3), 415.
42. K. Rajeshwar, J. Appl. Electrochem. 1985, 15, 69. M. Sharon, A. Sinha, Sol. Energy Mater.
1. 1984, 9, 391.
43. A. Heller, K. C. Chang, B. Miller, J.Am. 70. M. Sharon, K. Basavaswaran, Sol. Cells
Chem. Soc. 1978, 100, 684. 1988, 25, 97.
44. L. Fornarini, M. Lazzari, L. P. Bicelli et al., 71. P. Veluchamy, M. Sharon, M. Shimizu
J. Power Source 1981, 6, 371. et al., J. Electroanal. Chem. 1994, 371, 205.
45. Y. Mirovsky, D. Cahen, Appl. Phys. Lett. 72. A. Heller, B. Miller, S. S. Chu et al., J. Am.
1982, 40, 727. Chem. Soc. 1979, 101, 7633.
46. N. LeNagard, A. Perrin, M. Sergent et al., 73. A. J. Bard, F.-R.-F. Fan, B. Reichman, J. Am.
Mater. Res. Bull. 1985, 20, 835. Chem. Soc. 1979, 101, 7633.
47. R. Tenne, Y. Mirovsky, Y. Greenstein et al., 74. K. Rajeshwar, M. Kaneko, A. Yamado,
J. Electrochem. Soc. 1982, 129, 1506. J. Electrochem. Soc. 1983, 130, 38.
48. B. Scrosati, L. Fornarini, G. Razzini et al., 75. K. Rajeshwar, P. Singh, J. Du Bow, Trans.
J. Electrochem. Soc. 1985, 132, 593. ASME J. Sol. Energy Eng. 1982, 104, 146.
76. A. Barrasse, H. Cachat, G. Horowitz et al., 94. A. J. Nozik, Appl. Phys. Lett. 1976, 29, 150.
Rev. Phys. Appl. 1982, 17, 801. 95. A. Heller, M. Robbins, K. J. Bachmann
77. D. S. Ginley, R. M. Bie Feld, B. A. Parkinson et al., J. Electrochem. Soc., 1978, 125, 831.
et al., J. Electrochem. Soc. 1982, 129, 145. 96. S. Licht, O. Khaselev, T. Soga et al., Elec-
78. W. S. Hobson, A. B. Ellis, Appl. Phys. Lett. trochem. Soc. State Lett. 1998, 1, 20.
1982, 41, 891. 97. B. O’Regan, M. Grätzel, Nature 1991, 353,
79. S. Gourgaud, D. Elliott, J. Electrochem. Soc. 737.
1977, 124, 102. 98. S. Licht, D. Peramunage, Nature 1990, 345,
80. A. Heller, B. Miller, B. A. Parkinson, 330.
J. Electrochem. Soc. 1979, 126, 954. 99. S. Licht, D. Peramunage, Nature 1991, 354,
81. M. G. Chris, S. L. Nathan, Nature 1982, 300, 440.
733. 100. B. J. Tufts, I. L. Abrahams, P. G. Santangelo
82. R. Memming, J. Electrochem. Soc. 1978, 125, et al., Nature 1987, 326, 861.
117. 101. S. Licht, R. Tenne, G. Dagan et al., Appl.
83. A. J. Bard, R. E. Malpas, K. Itaya, J. Am. Phys. Lett. 1985, 46, 608.
Chem. Soc. 1979, 101, 2535. 102. S. Menezes, H. J. Lewerenz, K. J. Bach-
84. H. J. Lawerenz, M. Lubke, Appl. Phys. Lett. mann, Nature 1983, 305, 615.
1981, 39, 798. 103. G. Kline, K. Kam Keung, R. Ziegler et al.,
85. W. M. Ayers, J. Electrochem. Soc. 1982, 129, Sol. Energy. Mater. 1982, 6, 337.
1644. 104. G. Kline, K. Kam, D. Canfield et al., Sol.
86. A. J. Bard, F.-R.-F. Fan, G. A. Hope, Energy. Mater. 1981, 4, 301.
J. Electrochem. Soc. 1982, 129, 1647. 105. A. Heller, B. Miller, Appl. Phys. Lett. 1981,
87. T. S. Jayadevaiah, Appl. Phys. Lett. 1974, 25, 38, 282.
399. 106. B. A. Parkinson, A. Heller, B. Miller, Appl.
88. A. J. Bard, F.-R.-F. Fan, H. S. White et al., Phys. Lett. 1978, 33, 521.
J. Am. Chem. Soc. 1980, 102, 5142. 107. S. Licht, J. Phys. Chem. 1986, 90, 1096.
89. M. S. Wrighton, A. B. Bocarsly, E. G. Wal- 108. K. Mukhopadhyay, I. Mukhopadhyay,
ton et al., J. Electroanal. Chem. 1979, 100, M. Sharon et al., Carbon, 1997, 35, 863.
283. 109. M. Sharon, I. Mukhopadhyay, K. Mukho-
90. M. S. Wrighton, K. D. Legg, A. B. Ellis et al., padhyay, Sol. Energy Mater. Sol. Cells 1997,
Proc. Natln. Acad. Sci. USA 1977, 74, 4116. 45(1), 35.
91. M. S. Wrighton, R. G. Austin, A. B. Bocarsly 110. K. Murali Krishna, M. Sharon, M. K.
et al., J. Am. Chem. Soc. 1978, 100, 1602. Mishra, J. Phys. Solids 1996, 57(5), 615.
92. M. S. Wrighton, J. M. Bolts, A. B. Bocarsly 111. K. Murali Krishna, M. Sharon, M. K. Mishra
et al., J. Am. Chem. Soc. 1979, 101, 1378. et al., Electrochem. Acta 1996, 41(13), 1999.
93. R. M. Candea, M. Kastner, R. Goodman 112. M. Sharon, I. Mukhopadhyay, S. Ghosh,
et al., J. Appl. Phys. 1976, 47, 2724. J. Solid State Electrochem. 1999, 3, 141.
4.2 Photoelectrochemical Solar Energy Storage Cells 317
4.2 higher conversion efficiencies [1, 2a,b].

Photoelectrochemical Solar Energy Photoelectrochemical systems may facil-
Storage Cells itate not only solar to electrical energy
conversion but have also led to investi-
Stuart Licht
gations in photoelectrochemical synthesis,
Technion – Israel Institute of Technology,
photoelectrochemical production of fuels,
Haifa, Israel
and photoelectrochemical detoxification of
4.2.1 pollutants as discussed in other chapters
Introduction in this volume.
Although society’s electrical needs are 4.2.1.1 Regenerative Photoelectrochemical

largely continuous, clouds and darkness Conversion
dictate that photovoltaic solar cells have In illuminated semiconductor systems,
an intermittent output. A photoelectro- the absorption of photons generates ex-
chemical solar cell (PEC) can generate cited electronic states. These excited states
not only electrical but also electrochem- have lifetimes of limited duration. Without
ical energy and provide the basis for a a mechanism of charge separation their
system with an energy-storage component intrinsic energy would be lost through
(PECS). Sufficiently energetic insolation relaxation (recombination). Several dis-
incident on semiconductors can drive tinct mechanisms of charge separation
electrochemical oxidation/reduction and have been considered in designing ef-
generate chemical, electrical, or electro- ficient photoelectrochemical systems. At
chemical energy. Aspects include efficient illuminated semiconductor–liquid inter-
dye-sensitized or direct solar to elec- faces, an electric field (the space charge
trical energy conversion, photoetching, layer) occurs concurrent with charge–ion
photoelectrochemical water-splitting, en- redistribution at the interface. On photo-
vironmental cleanup, and solar energy generation of electron-hole pairs, this elec-
storage cells. This chapter focuses on tric field impedes recombinative processes
photoenergy storage concepts based on by oppositely accelerating and separating
photoelectrochemical processes, but in- these charges, resulting in minority car-
cludes a necessary comparison to other rier injection into the electrolytic redox
methods proposed for the conversion and couple.
storage of solar energy. The PEC uses This concept of carrier generation is il-
light to carry out a chemical reaction, lustrated in Fig. 1(a) (for an n-type PEC)
converting light to chemical energy. This and has been the theoretical basis for sev-
fundamental difference of the photovolatic eral efficient semiconductor-redox couple
solar cell’s (PV) solid–solid interface, and PEC cells. Illumination of the electrode
the PEC’s solid–liquid interface has sev- surface with light, whose photon energy
eral ramifications in cell function and is greater than the band gap, promotes
application. Energetic constraints imposed electrons into the conduction band leav-
by single band gap semiconductors have ing holes in the valance band. In the case
limited the demonstrated values of photo- of a photoanode, band-bending in the de-
electrochemical solar to electrical energy pletion region drives any electron that is
conversion efficiency to 16%, and multiple promoted into the conduction band into
band gap cells can lead to significantly the interior of the semiconductor and holes
Load Load
E E
Light Light
EC − D∗
− EC − −
EF EF
Energy Energy
− gain gain
Red Ox+ Red Ox+ −
D
EV + EV
+
n -type Redox Counter n -type Dye Redox Counter

semiconductor electrolyte electrode semiconductor electrolyte electrode
(a) (b)
Fig. 1 Carrier generation under illumination arising at (a) the semiconductor–liquid interface and
(b) the semiconductor–dye sensitizer–liquid interface.
in the valance band toward the electrolyte, charge is then injected into a wide band
where they participate in an oxidation re- gap semiconductor. This alternative car-
action. Electrons through the bulk drive an rier generation mode can also lead to
external load before they reach the counter effective charge separation as illustrated
electrode or storage electrode, where they in Fig. 1(b). The first high solar to elec-
participate in a reduction process. Under tric conversion efficiency example of such
illumination and open circuit, a negative a device was demonstrated in 1991 [3]
potential is created in a photoanode, and as through the use of a novel high surface area
a result the fermi level for the photoanode (nanostructured thin film) n-TiO2 , coated
shifts in the negative direction, thus reduc- with a well-matched trimeric ruthenium
ing the band-bending. Under illumination complex dye immersed in an aqueous
with increasing intensity, the semiconduc- polyiodide electrolyte. The unusually high
tor fermi level shifts continually toward surface area of the transparent semi-
negative potentials until the band-bending conductor coupled to the well-matched
effectively reduces to zero, which corre- spectral characteristics of the dye leads
sponds to the flat band condition. At this to a device that harvests a high proportion
point, a photoanode exhibits its maximum of insolation.
photovoltage, which is equal to the barrier
height. 4.2.1.2 Photoelectrochemical Storage
Excitation can also occur in molecules PECs can generate not only electrical
directly adsorbed and acting as a medi- but also electrochemical energy. Figure 2
ator at the semiconductor interface. In presents one configuration of a PEC com-
this dye sensitization mode, the func- bining in situ electrochemical storage
tion of light absorption is separated from and solar-conversion capabilities; provid-
charge carrier transport. Photoexcitation ing continuous output insensitive to daily
occurs at the dye and photogenerated variations in illumination. A high solar to
hn
S
P P= 0
S
POLYSULFIDE
OXIDATION
n -Cd(Se,Te) n -Cd(Se,Te)
POLYSULFIDE
TIN SULFIDE
REDUCTION
REDUCTION
POLYSULFIDE
REDUCTION
OXIDATION
MEMBRANE
MEMBRANE
Sn/SnS
TIN
Sn/SnS
CoS
1.0 m Cs2S4
0.8 m CsOS
0.8 m CsHS
1.8 m CsOH
1.0 m Cs2S4
0.8 m CsOS
1.8 m CsHS
0.8 m CsHS
1.8 m CsOH
1.8 m CsHS
L CoS LOAD
LOAD L
(a) (b)
Fig. 2 Schematic of a photoelectrochemical solar cell combining both solar conversion and storage
capabilities. (a) Under illumination; (b) in the dark.
electric conversion efficiency cell config- upon reversal of the storage process.
uration of this type was demonstrated in To date, the predominant nonbiologic
1987 and used a Cd(Se,Te)/Sx conversion utilization of solar energy is to heat a
half cell and a Sn/SnS storage system, working fluid that is maintained in an
resulting in a solar cell with a contin- insulated enclosure, storing a portion of
uous output [4]. Under illumination, as the incident solar radiation for future use.
seen in Fig. 2(a), the photocurrent drives Limitations of this approach include the
an external load. Simultaneously, a por- low energy available per unit mass of
tion of the photocurrent is used in the the storage medium, and low efficiencies
direct electrochemical reduction of metal of thermal to mechanical and thermal to
cations in the device storage half-cell. In electrical energy conversion. A variety of
darkness or below a certain level of light, passive and dynamic optical concentrators
the storage compartment spontaneously have been studied to compensate for these
delivers power by metal oxidation, as seen limitations.
in Fig. 2(b). The high temperatures generated by
concentrated solar power have been uti-
4.2.2 lized to drive highly endothermic re-
Comparative Solar-Storage Processes actions. The reverse reaction releases
the chemically stored energy as ther-
4.2.2.1 Thermal Conversion and Storage mal energy. Various systems have been
Solar insolation can be used to directly investigated such as in 1985 [5]:
activate a variety of thermal processes; the
enthalpy is stored physically or chemically SO3 −−−→ SO2 + 12 O2 !H = 98.94
and then either directly utilized or released (1)
4.2.2.2 Photochemical Storage yield, rapid back reaction, and difficulties

In photochemical processes, photon ab- separating the product species. An exam-
sorption creates a molecular excited ple of storage by such photodissociation
state or stimulates an interband elec- processes is exemplified by
tronic transition in a semiconductor
that induces a molecular change. Com- 2NOCl −−−→ 2NO + Cl2 (6)
prehensive reviews, including those by
Gratzel [6], Kalyanasundaram [7], and Har- In photoisomerization, an absorbed pho-
riman [8], have discussed various as- ton activates molecular rearrangement
pects of photochemical energy conver- and conversion of organic molecules into
sion. A photoactivated molecular excited strained isomers. The products are stored.
state can drive either (1) photodissociation Despite attractive features including a high
(2) photoisomerisation or (3) photoredox heat storage capacity and good thermal
reactions. Processes based on semicon- stability, most systems tested have poor
ductors may involve photovoltaic or photo- efficiencies. These systems necessitate
electrochemical systems. transformation to strained conformations
A substantial photochemical effort re- at high energies; energies consistent with
search has centered on the photoredox wavelengths below 450 nm. This excludes
storage of solar energy. Molecular photore- much of the energy inherent in the Air
dox processes use electron transfer from Mass 1 (AM1) solar spectra. The stored
photoinduced excited states: thermal energy is released on catalytic-
induced reversion to the starting compo-
A + hν −−−→ A∗ (2) nents. An example has been presented by
Kutal for the photocatalytic transformation
The electron transfer may be either of norbornadiene to quadricyclane [9].
direct Photochemical redox reactions can gen-
A∗ −−−→ A+ + e− (3) erate fuel formation, including H2 , CH3 ,
and CH3 OH. Because of its availability, the
or a variety of indirect processes as
splitting of water to produce H2 has been
exemplified by either
the focus of particular attention. H2 O is
A∗ −−−→ B; transparent to near UV or visible radiation,
and therefore sensitization is required to
followed by drive the water-splitting process. In early
B −−−→ B+ + e− (4) attempts on photoredox-splitting of water
by Heidt and McMillian, the process was
or A∗ + B −−−→ AB; sensitized by using solution redox species
followed by such as Ce3+/4+ [10].
AB −−−→ AB+ + e− (5) Ce3+ (aq) + H+ −−−→ Ce4+ (aq) + 12 H2
(7)
Solar-activated photodissociation pro-
cesses generally involve cleavage of a 2Ce4+ (aq) + H2 O −−−→ 2Ce3+ (aq)
simple molecule into several energetic + 2H+ + 12 O2 (8)
products. Limitations of this approach
include the limited absorption of solar These processes have displayed poor
energy by the molecule, low quantum quantum yields. As in photoisomerization
nν
Decomposition ∗Ru(bpy) 2+
3
∗TEOA
Ru(bpy)3 2+
Mv++
Ru(bpy)33+
Pt H2O
TEOA
Mv+•
TEOA N·(CH3·CH2·OH)3
+ H2
Mv++ CH3 — N N — CH3
+
Scheme 1
processes, these reactions are also gener- principle, this should lead to a higher level
ally driven only by high-energy radiation of photon absorption and more effective
(short wavelength) and cannot efficiently charge separation. Both effects can sub-
convert incident AM1 solar radiation. stantially increase solar photochemical-
H2 or O2 generation from water is conversion efficiency, but these systems
a multielectron process. Optimization of have not yet displayed high efficiencies of
photoredox-splitting of water necessitates fuel generation or long-term stabilities.
the presence of a catalyst to mediate Photoredox processes at semiconductor
this complex multielectron transfer. In electrodes generating fuels or products
one such process, a sacrificial reagent other than hydrogen, including methanol
triethanolamine (TEOA) is consumed ir- and ammonia, have been attempted with
reversibly in the process, as denoted in low overall yields. The photoelectrolysis of
Sch. 1 [9]: HI into H2 and I3 − at p-InP electrodes has
Direct multielectron processes are rare been described [11]. These H2 and I3 − pho-
and instead incorporate one or more rad- togenerated products are prime candidates
ical intermediate steps. These reactive for a fuel cell. Analogous advanced sys-
intermediates are susceptible to unfavor- tems, in which the photoelectrochemically
able side reactions, resulting in substantial generated fuels have been successfully re-
losses in the energy-conversion process. combined to generate electrical energy, are
Kinetically favored back reactions further discussed later in this section.
reduce the overall conversion efficiency. A system exemplifying photoelectro-
The engineering of these complex molec- chemical synthesis to generate hydro-
ular organizations provides a substantial gen is water photoelectrolysis. An early
scientific challenge and have generally re- demonstration of water photoelectroly-
sulted in systems with low conversion sis used TiO2 (band gap 3.0 eV) and
efficiencies. was capable of photoelectrolysis at ∼0.1%
solar to chemical energy–conversion ef-
4.2.2.3 Semiconductor Photoredox ficiency [12]. The semiconductor SrTiO3
Storage was demonstrated to successfully split wa-
Semiconductor surfaces have been used ter in a direct photon-driven process by
as sensitizers to drive photochemical con- Bolts and Wrighton (1976), albeit at low
version and storage of solar energy. In solar energy–conversion efficiencies [13].
The high SrTiO3 band gap, Eg , of 3.2 eV 4.2.3

creates sufficient energetic charge to drive Modes of Photoelectrochemical Storage
the photoredox process. This excludes
the longer wavelength photons and corre- Conversion of a regenerative PEC to a pho-
sponds to only a small fraction of incident toelectrochemical storage solar cell (PECS)
solar radiation. To improve the solar re- can incorporate several increasingly so-
sponse, Eg has to be lowered; in a single phisticated solar energy conversion and
band gap system, an optimum efficiency storage configurations.
can be expected around 1.4 eV.
In photoelectrochemical water-splitting 4.2.3.1 Two-Electrode Configurations
systems, corrosion of the semiconduc- A variety of two-electrode configurations
tor photoelectrodes can pose a significant have been investigated as PECS systems.
problem. Most surface-stabilizing redox Important variations of these photoelec-
reactions compete with oxygen and hy- trochemical conversion and storage con-
drogen generation and must be excluded figurations are summarized in Table 1. In
from these systems. To enhance the solar each case, and as summarized in Fig. 3
response of high band gap materials, tech- for the simplest configurations, exposure
niques such as dye sensitization and impu- to light drives separate redox couples and
rity sensitization have been attempted, al- a current through the external load. There
though with little improvement [14]. Semi- is a net chemical change in the system,
conductor surfaces have been modified to with an overall increase in free energy.
protect low band gap materials against In the absence of illumination, the gener-
ated chemical change drives a spontaneous
photocorrosion [15, 16]. A self-driven pho-
discharge reaction. The electrochemical
toelectrochemical cell consisting of Pt-
discharge induces a reverse current. In
coated p-InP and Mn-oxide-coated n-GaAs
each case in Table 1, exposure to light
has been demonstrated to operate at 8.2%
drives separate redox couples and current
maximum efficiency to generate H2 and
through the external load.
O2 under simulated sunlight [17], and
Consistent with Fig. 1, in a regenerative
more recently a two band gap cell in a tan-
PEC, illumination drives work through
dem arrangement has been used to split an external load without inducing a net
water at 12% efficiency [18]. A multijunc- change in the chemical composition of
tion GaAs, Si cell has been recently used the system. This compares with the two-
to drive water-splitting at over 18% solar to electrode PECS configurations shown in
electrical-conversion efficiency [19]. Fig. 3(a) and (b). Unlike a regenerative
Colloids and suspensions of semicon- system, there is a net chemical change
ductors have been used for the photoredox- in the system, with an overall increase
splitting of water. The principle advantage in free energy. In the absence of illu-
of a fine suspension is the large active sur- mination, the generated chemical change
face area available. Reaction rates of H2 drives a spontaneous discharge reaction.
and O2 generation have been enhanced by The electrochemical discharge induces
loading the particles with small deposit of a reverse current. Utilizing two quasi-
precious metals, and although significant reversible chemical processes, changes
progress was made in this direction, a prac- taking place in the system during illumina-
tical system is yet to be demonstrated [11]. tion can be reversed in the dark. Similar to
Tab. 1 Important two-electrode photoelectrochemical conversion and storage

configurations
Scheme Electrode 1 Electrolyte(s) Electrode 2
I SPE | Redox A Redox B | CE

II SPE | Redox A-membrane-Redox B | CE
III SPE | Redox A | Redox BCE -CE
IV SPE-Redox ASPE | Redox B | CE
V SPE | Redox A-membrane-Redox B | SPE
Note: Components of these systems include a semiconductor photoelectrode (SPE) and

a counter electrode (CE). At the electrode–electrolyte interface, redox couples ‘‘A’’ or ‘‘B’’
are either in solution (| Redox |), counter electrode–confined (| Redox BCE -CE) or
confined to the semiconductor photoelectrode (SPE-Redox ASPE |).
L e L
e−
R R′
O R′
hν hν
x−
x−
O R O′
O′
n-type Storage n-type Membrane Counter

semiconductor electrode semiconductor electrode
Fig. 3 Schematic diagram of a two-electrode storage cell. On the left the storage electrode
contains an insoluble redox couple, and on the right a soluble redox couple, with storage
represented as
hν
R + O% −−
←−−→
−− O + R
%
a secondary battery, the system discharges photoelectrode is kinetically unsuited to

producing an electric flow in the opposite perform as a counter electrode. In the
direction and the system gradually returns absence of illumination, the photoelec-
to the same original chemical state. trode P, in this case a photoanode, now
Each of the cells shown in Fig. 3 assumes the role of a counter electrode
has some disadvantages. For both bound by supporting a reduction process. For the
(Fig. 3left) and soluble (Fig. 3right) redox photoanode to perform efficiently during
couples, the redox species may chemi- illumination (charging), this very same re-
cally react with and impair the active duction process should be inhibited to
materials of the photoelectrode. Further- minimize photooxidation back reaction
more, during the discharge process, the losses. Hence, the same photoelectrode
cannot efficiently fulfill the duel role of configuration as shown in Fig. 4. In Fig. 4,
being kinetically sluggish to reduction dur- the switches E and F are generally alter-
ing illumination and yet being kinetically nated during charge and discharge. During
facile to the same reduction during dark the charging, only switch E may be closed,
discharge. The configuration represented facilitating the storage process, and dur-
in Fig. 3 has another disadvantage, the ing discharge, E is kept open while F
disparity between the small surface area is closed. In this case, chemical changes
needed to minimize photocurrent dark that took place during the storage phase
current losses and the large surface area are reversed, and a current flow is main-
necessary to minimize storage polarization tained from the storage electrode to a third
losses to maximize storage capacity [20]. (counter) electrode that is kept in the first
compartment. To minimize polarization
4.2.3.2 Three-Electrode Configurations losses during the discharge, this third elec-
Several of the two-electrode configura- trode should be kinetically fast to the redox
tion disadvantages can be overcome by couple used in the first compartment.
considering a three-electrode storage cell Still an improved situation would be to
E
• •
L
• •
F
R′
R R
hν
EL
O O′
O
P A M S
First compartment Second compartment

Fig. 4 Schematic diagram for a storage system with a third electrode (counter
electrode) in the photoelectrode compartment. P = Photoelectrode,
A = Counter electrode, M = Membrane, S = Storage electrode,
EL = Electrolyte, E, F = Electrical switch, and L = Load.
have both switches closed all the time. In Improvements relating to its stability and
this case, electric current flows from the conversion efficiency are of paramount
photoelectrode to both counter and stor- importance.
age electrodes. The system is energetically
tuned such that when insolation is avail- 4.2.4.1 Improvements of the
able, a significant fraction of the converted Photoresponse of a Photoelectrode
energy flows to the storage electrode. In To improve the solar response of a
the dark or diminished insolation, the photoelectrode, a proper match between
storage electrode begins to discharge, driv- the solar spectrum and the band gap of
ing continued current through the load. the semiconductor should be maintained.
In this system, a proper balance should When a single band gap semiconductor is
be maintained between the potential of used, a band gap in the vicinity of 1.4 eV
the solar energy–conversion process and is most desirable from the standpoint of
the electrochemical potential of the stor- optimum solar-conversion efficiency. An
age process. There may be residual electric important criterion is that the minority
flow through the photoelectrode during carrier that is driven toward the semi-
dark cell discharge, as the photoelectrode conductor–electrolyte interface should not
is sluggish, but not entirely passive, to a participate in a photocorrosion reaction
reduction process. This can be corrected that is detrimental to the long-term stabil-
by inserting a diode between the photo- ity of the photoelectrode. Photocorrosion
electrode and the outer circuit. can be viewed in terms of either kinetic
or thermodynamic considerations and the
4.2.4 real cause may be a mixture of both. From
Optimization of Photoelectrochemical thermodynamic perspective, a photoanode
Storage is susceptible to corrosion if the fermi level
for holes is at a positive potential with re-
The power obtained is the product of volt- spective to the semiconductor corrosion
age and current, and consideration of the potential [21]. The corrosion can be pre-
photocurrent is as important as the photo- vented or at least inhibited by choosing a
voltage. If the band-bending is sufficiently redox couple that has its Eredox more nega-
large, then the minority carrier redox re- tive than that for the corrosion process [22,
action, which is essential to maintain 23]. The kinetic approach has been to allow
the photocurrent, can compete effectively another desired redox process to occur at a
with the recombination of photogenerated much faster rate than the photocorrosion
electron hole pairs. This recombination reaction [13]. Other attempts to minimize
represents a loss of absorbed photo en- the photocorrosion has been to coat the
ergy. Therefore, an objective is to maintain photoelectrode surface with layers such as
a high band-bending and at the same time Se [24] and protective conductive polymer
a significant photovoltage. A photoanode films [25], and to search for alternate low
creates a negative photovoltage under il- band gap semiconductors [26]. Extensive
lumination, which results in reducing the reviews on the performance and stability
band-bending. In principal, one way to ac- of cadmium chalcogenides include those
complish high band-bending is to choose a by Cahen and coworkers, 1980 [27] and
very positive redox couple in the electrolyte. Hodes, 1983 [28]. Etching of photoelec-
The converse is true for a photoanode. trode surface has been recognized and
widely used as an important treatment a favorable influence on the PEC sys-

to achieve high-conversion efficiency [29]. tem by improved charge-transfer kinetics
This effect is mostly attributed to removal leading to improved stability of the pho-
of surface states that may act as trap- toelectrode [36]. Additives incorporated in
ping centers for photogenerated carriers. redox electrolytes are known to enhance
A related procedure called photoetching, the performance of PEC systems. Addition
initially developed for CdS and then ap- of small concentrations of Se in polysulfide
plied to a wide variety of semiconductors, electrolyte is known to improve the sta-
improves the photoelectrode performance bility of CdSe (single crystal)/polysulfide
and preferentially removes the surface de- system [22]. In this case, Se improves the
fects acting as recombination centers [30]. PEC performance by reducing S/Se ex-
In addition to the variety of etching pro- change and by increasing the dissolution
cedures, several other surface treatments of the photooxidized product S, which is
have been used to improve photoelectrode the rate-determining step in the oxidation
performance. Examples include a Ga3+ ion of sulfide at the anode. Addition of Cu2+
dip on CdSe [31], ZnCl2 dip on thin film into the I− /I3 − electrolyte is known to
CdSe [32], Ru on GaAs [33], Ru on InP [34], improve the stability of CuInX2 photoelec-
and Cu on CdSe [34]. Reasons explaining trode considerably [37], and in the same
the effectiveness of these dips range from a electrolyte, tungsten and molybdenum
decrease of dark current to electrocatalysis dichalcogenide photoelectrochemistry can
by surface-deposited metal atoms. Solu- be substantially improved by addition of
tion phase chemistry of the electrolyte is an Ag+ , or other metal cations, and shift of
important parameter that has been shown the I− /I3 − Eredox [38].
to dramatically influence photoeffects. The In the case of CdSe/polysulfide system,
equilibrium position of the redox cou- solution activity, conductivity, efficiency
ple will affect equilibrium band-bending. of the photoanode (fill factor), charge-
A photoanodic system with a solution transfer kinetics at the interface, and the
containing a more positive potential re- stability of the photoelectrode are known
dox couple causes a greater band-bending, to exhibit improvements in the trend
which in turn leads to a higher photovolt- Li > Na > K > Cs > for alkali polysulfide
age and efficient carrier separation under electrolyte. This trend is explained in terms
normal experimental conditions. of the secondary cation effect on electro-
chemical anion oxidation in concentrated
4.2.4.2 Effect of the Electrolyte aqueous polysulfide electrolytes [39]. In
Semiconductor photoeffects in a complex the case of Cd(Se,Te)/polysulfide system,
redox electrolyte are largely affected by the the efficiency of light energy conversion
solution properties such as solution re- is improved by using a polysulfide elec-
dox level, interfacial kinetics (adsorption), trolyte without added hydroxide because
conductivity, viscosity, overall ionic activ- of the combined effect of increasing the
ity, solution stability, and transparency solution transparency, relative increase of
within a crucial wavelength region. Re- S4 2− , and decrease in S3 2− in solution.
dox electrolytes are known to inhibit For the same photoelectrode–electrolyte
unfavorable phenomena such as surface system, an optimum photoeffect was ob-
recombination and trapping [35]. In addi- served for a solution containing a sul-
tion, the solution redox couple may induce fur–sulfide ratio of 1.5 : 2.1 with l : 2 molal
potassium sulfide concentrations because In PEC systems, a compromise is main-

of the combined effect of optimized solu- tained to simultaneously optimize the
tion viscosity, transparency, activity, and photoelectrode efficiency, stability, and
shift in solution redox level [40]. Stabil- electrolytic properties of the electrolyte.
ity of the polysulfide redox electrolyte, Practical PEC systems often require large
which is another parameter that deter- working and counter electrodes and their
mines the long-term performance of a geometric configuration within the PEC
PEC cell, has been shown to increase with system will effect mass transport and
sulfur and alkali metal sulfide concentra- effective cell current. In some cases, ad-
tion and to decrease with either increasing vantageous use has been made of selective
−OH− concentration or at high ratio of sluggish counter electrode kinetics toward
added sulfur to alkali metal sulfide [39, certain cathodic processes. For example,
41]. The combined polysulfide electrolyte carbon is a poor cathode for H2 evolution
optimization can substantially enhance compared to Pt, and the direct hydrogena-
cadmium chalcogenide photoelectrochem- tion of anthraquinone at a PEC cathode
ical conversion. has been avoided by using a carbon an-
Chemical composition of the electrolyte ode [44]. In this case, such hydrogenation
is a particularly important parameter represents an undesirable side reaction.
in PEC systems based on complex
electrolytes, such as polysulfide or ferro/ 4.2.4.4 Combined Optimization of Storage
ferricyanide. In the latter redox couple, and Photoconversion
replacement of a single hexacyano ligand An efficient photoelectrochemical conver-
strongly changes the photoelectrochemical sion and storage system requires not only
response of illuminated n-CdSe [42], and an efficient functional performance of the
addition of the KCN to the electrolyte can separate cell components but also a system
increase n-CdSe and n-CdTe photovoltage compatibility. In the combined photoelec-
by 200 mV [43]. trochemical storage system, simultaneous
parameters to be optimized include
4.2.4.3 Effect of the Counter Electrode
In a photoanodic system, even at mod- 1. minimization of light losses reaching
erate current densities, the occurrence the photoelectrode,
of sluggish counter electrode kinetics for 2. high photoelectrode–conversion effi-
the cathodic process will cause significant ciency of solar energy,
polarization losses and diminish the pho- 3. close potential match between the pho-
tovoltage. Minimization of these kinetic topotential and the required storage-
limitations necessitates a counter electrode charting potential,
with good catalytic properties. For exam- 4. high current and potential efficiency
ple, as shown by [34], CoS on stainless of the redox storage process,
steel or brass electrodes exhibits elec- 5. high energy capacity of the redox
trocatalytic properties toward polysulfide storage,
reduction and overpotentials as low as 6. reversibility (large number of charge–
1 mV cm2 mA−1 has been realized. Com- discharge cycles of the redox storage),
position of a particular redox electrolyte 7. stability of the photoelectrode,
may have a bearing on the extent of counter 8. stability of the electrolyte,
electrode polarization [32, 39]. 9. stability of the counter electrode,
10. economy and cost effectiveness, and a counter electrode. As an example of chal-
11. reduced toxicity and utilization of lenges that may arise in the combined
environmentally benign materials. photoconversion and storage system, con-
sider an n−CdSe/polysulfide/tin sulfide
A photoelectrochemical solar cell implic- version of Fig. 4 and consisting of Cell 1.
itly contains an electrolytic medium. In Cell: 1. CdSe | HS− , OH− , Sx 2− |
the majority of laboratory PEC configu- Membrane | HS− , OH− | SnS | Sn
rations, incident light travels through the With illumination, the cell exhibits si-
electrolytic medium before illuminating multaneous photoelectrode, counter elec-
the photoelectrode. The resultant light ab- trode and storage reactions, and equilibria
sorbance by the electrolyte is a significant including
loss, which is avoided by use of a back cell
configuration. For example, the substan- Photoanode:
tial absorptivity of dissolved polysenide
HS− + OH− −−−→ S + 2e− + H2 O
species has been avoided in a n-GaAs pho-
(9)
toelectrochemistry through the use of the
Photocompartment equilibria:
back wall cell configuration presented in
Fig. 5 [45]. S + S2−
x −−−→ S( x + 1)
2−
(10)
The photoelectrochemical system shown
Counter electrode:
in Fig. 4 is a combination of a photoelec-
trode, electrolyte, membrane, storage, and S + 2e− −−−→ HS− + OH− (11)
Photoelectrode Counter
contact electrode
contact
Sapphire Cu wire
window
Mounted
hν thin-film
PEC
hν Ag paste
Au grid
hν
1.9 µm Photoelectrode
thick contact
n -GaAs
Polyselenide
(aq)
Window
Epoxy
Fig. 5 A back wall n-GaAs/aqueous polyselenide photoelectrochemical cell.

Storage electrode: variation in electrolyte composition may

− have a profound influence on the stability
SnS + H2 O + 2e −−−→
of the photoelectrode and electrolyte and
Sn + HS− + OH− (12) on cell potential. Hence, to minimize these
effects, either excess polysulfide must be
Unlike the case of the analogous re- included in the photocompartment or a
generative PEC system, in the preceding limit must be set to the maximum depth
equations, sulfur formed at the photoan- of cell charge and discharge.
ode (and dissolved as polysulfide species, Another important consideration is the
S(x+1) 2− for x = 1 to 4) is not balanced by energy compatibility between the photo-
the reduction that is taking place at the conversion and the storage processes. This
counter electrode because of the simulta- compatibility is referred to as the volt-
neous reduction process taking place at age optimization. Figure 6 presents the
the storage electrode. As a result, sulfur combined IV characteristics for an ideal
is accumulated in the photoelectrode com- photoelectrode and current–voltage curves
partment and is removed only in the sub- for two alternative redox processes; process
sequent discharge process. This dynamic A and process B. Vph is the maximum
Redox process B Redox process A
A: P: IVphoto
B: ISC
Current
Vmax
EB VPh EA
Potential
Fig. 6 Current–voltage curves for power, Pmax , and Isc and Vph are the short-circuit
electrochemical storage processes, A or B. current and open-circuit photopotentials,
Process A may be charged by the photodriven respectively. EA and EB refer to the redox
current–voltage curve P, whereas process B may potentials for redox processes, A and B,
not. In the photodriven IV curve P, Vmax is the respectively.
voltage corresponding to the point of maximum
photovoltage that can be generated. Isc is Ideally, the membrane used to separate
the short-circuit photocurrent correspond- the two-cell compartments, as indicated
ing to maximum band-bending. In Fig. 6, in Fig. 4, must be permeable only to ions
consider the electrochemical process rep- that will transport charge, but that will not
resented by curve B. This process is located chemically react or otherwise impair any
outside the region of potentials gener- electrode. The permeability of membranes
ated by the photoelectrode; it does not is generally less than ideal. Different mem-
represent a potential storage system to branes permit other ions and water to per-
be driven by a single photoelectrode. In meate to a varying degree [46, 47]. Gross
such a case, a serial combination of more mixing of active materials across the mem-
than one photoelectrode would be neces- brane causes them to combine chemically
sary. For a redox process to be a potential and in the process lose energy. Favorable
candidate for a redox storage system, the qualities that a membrane should exhibit
storage and photodriven current–voltage are low permeability toward chemically
curves should intersect. Whereas Vph and reactive ions, low resistivity, mechanical
Isc correspond to zero power, the point integrity, and cost effectiveness.
Pmax shown in Fig. 6 corresponds to the
maximum power point. Solar energy con- 4.2.5
version is accomplished at its maximum High Efficiency Solar Cells with Storage
efficiency only during operation in the po-
4.2.5.1 Multiple Band Gap Cells with
tential vicinity of Pmax . Storage
By adjusting the electrical load L, shown A limited fraction of incident solar photons
in Fig. 4, the system can be constrained have sufficient (greater than band gap) en-
to operate near its maximum power effi- ergy to initiate charge excitation within a
ciency. In this case, if the counter electrode semiconductor. Because of the low frac-
is not polarized, the potential difference tion of short wavelength solar light, wide
between photoelectrode and the counter band gap solar cells generate a high photo-
electrode will be close to Vmax . If one voltage but have low photocurrent. Smaller
chooses a facile redox process for the stor- band gap cells can use a larger fraction of
age electrode, as indicated by the sharply the incident photons, but generate lower
rising IV curve for process A in Fig. 6, photovoltage. Multiple band gap devices
with Eredox in the vicinity of Vmax , then the can overcome these limitations. In stacked
potential during charge and discharge of multijunction systems, the topmost cell
the storage process will remain near Vmax . absorbs (and converts) energetic photons,
As a result, the potential will be a highly but it is transparent to lower energy pho-
invariant current variation through the tons. Subsequent layer(s) absorb the lower
load L, regardless of insolation intensity. energy photons. Conversion efficiencies
This situation represents an ideal match can be enhanced, and calculations predict
between solar energy conversion and stor- that a 1.64-eV and 0.96-eV two–band gap
age processes within a PECS. Nonideality system has an ideal efficiency of 38% and
occurs with poor voltage-matching or ki- 50%, light of 1 and 1000 suns intensity,
netic limitations and polarization losses respectively. The ideal efficiency increases
associated with the counter, storage, or to a limit of 72% for a 36–band gap solar
photoelectrodes. cell [48].
Recently, high solar conversion and stor- storage cell described by the Fig. 7 en-
age efficiencies have been attained with ergy diagram. The single cell contains
a system that combines efficient mul- both multiple band gap and electrochem-
tiple band gap semiconductors, with a ical storage, which unlike conventional
simultaneous high capacity electrochem- photovoltaics, provides a nearly constant
ical storage [49, 50]. The energy diagram energetic output in illuminated or dark
for one of several multiple band gap cells conditions. The cell combines bipolar Al-
is presented in Fig. 7, and several other GaAs (Eg = 1.6 eV) and Si (Eg = 1.0 eV)
configurations are also feasible [1, 2a, b]. and AB5 metal hydride/NiOOH storage.
In the figure, storage occurs at a potential Appropriate lattice-matching between Al-
of Eredox = EA+/A − EB/B+ . On illumina- GaAs and Si is critical to minimize dark
tion, two photons generate each electron, current, provide ohmic contact without ab-
a fraction of which drives a load, whereas sorption loss, and maximize cell efficiency.
the remainder (1/xe− ) charges the storage The NiOOH/MH metal hydride storage
redox couple. Without light – the poten- process is near ideal for the AlGaAs/Si be-
tial falls below Eredox – the storage couple cause of the excellent match of the storage
spontaneously discharges. This dark dis- and photocharging potentials. The electro-
charge is directed through the load rather chemical storage processes utilizes MH ox-
than through the multijunction semicon- idation and nickel oxyhydroxide reduction:
ductor’s high dark resistance.
MH + OH− −−−→ M + H2 O + e− ;
Cell: 2. In Fig. 8, an operational form
of the solar conversion is presented and a EM/MH = −0.8 V vs SHE (13)
1/xe–
(1–1/x)e–
V = Vw + Vo Rload
Electrocatalyst anode
e– 1/xA → 1/xA+
Electrocatalyst cathode
ECs
ηA
– EFermi(ns)
e EAredox
Vs
FCw
hν E
Gs
EFermi(nw=ps) 1/xB → 1/xB+
+
E Bredox ηB
Vw EGw > hν > EGs ⇒ h E Vs
hν EGw
Electro-
Semiconductor Electrolyte
catalyst
EFermi(pw)
Storage bipolar MPEC
+
hν > EGW ⇒ h E VW
Ohmic
p n p n
junction
Wide gap Small gap
Fig. 7 Energy diagram for a bipolar band gap indirect ohmic storage multiple band
gap photoelectrochemical solar cell (MBPEC).
1/xe–
(1−1/x)e-
Storage
Rload
current
current
current
Photo
Load
e–
350 µm
300 nm
(4 × 1019cm–3) 800 nm
Au-Zn/Au
1.0 µm
1.0 µm
1.7 nm
50 nm
20 nm
10 nm
P+ -GaAs
NiOOH + H2O + e–
MH + OH–
Separator with KOH electrolyte
19cm–3)
(8 × 1015cm–3)
(1 × 1018cm–3)
(0.3-0.15)Ga(0.7-0.85)As (1 × 10 cm )
(2 × 1017cm–3)
(1 × 1018cm–3)
–3
Metal hydride anode

AR coating [ZnS(50nm)/MgF2(70nm)]
x
18
Nickel cathode
(1 × 10
2hν
Au-Sb/Au
⇒
MHx–1 + H2O + e–
Ni(OH)2 + OH–
GaAs(buffer layer)
Illumination
N+-Al(0.15)Ga(0.85)As
N-Al(0.15)Ga(0.85)As
P+-Al0.8Ga0.2As
N-GaAs
N+-Si
P+-Al
+-Si
N-Si
P
Fig. 8 The bipolar AlGaAs/Si/MH/NiOOH MBPEC solar cell.
NiOOH + H2 O + e− −−−→ over approximately an eight-month period.

Figure 10 presents representative results
Ni(OH)2 + OH− ;
for two-day periods occurring 0, 40, 140,
ENiOOH/Ni(OH)2 = 0.4 V vs SHE (14) and 240 days into the experiment. As sum-
marized in the lower curves of the figure,
As reported [49] and as shown in
the load potential is again nearly constant,
Fig. 9, the cell generates a light vari-
despite a 100% variation in illumination
ation insensitive potential of 1.2–1.3 V
(AM0/dark) conditions. Over a 24-hour pe-
at total (including storage losses) so-
riod, the load potential increases by ∼2%
lar–electrical energy conversion efficiency
as the cell charges with illumination, fol-
of over 18%.
A long-term indoor cycling experiment lowed by a similar decrease in potential as
was conducted to probe the stability of stored energy is spontaneously released in
the AlGaAs/Si metal hydride storage solar the dark. The cycles exhibited in Fig. 10
cell [50]. Unlike the variable insolation of are representative, and as observed exhibit
Fig. 9, in each 24-hour cycle, a constant little variation on the order of weeks, and
simulated AM0 (135.3 mW cm−2 ) illumi- exhibit a variation of ∼1% over a period of
nation was applied for 12 hours, followed months. In this figure, photopower is de-
by 12 hours of darkness, and the cell po- termined as the product of the measured
tential, and storage (charge and discharge) cell potential and measured photocurrent.
currents monitored as a function of time Power over load is determined as the
1.5
Load potential
1.0 Generated VCell insensitive to illumination variation
[mA]
0.5
AlGa/Si MH multiple band gap storage solar cell
4
I Photo
Area = 0.22 cm2

[mA]
0
Illumination Dark Illumination Dark
Discharge Discharge
0
I Storage,
[mA]
Charge Charge
−2
insitu AB5/NiOOH
Metal hydride storage
0 6 12 18 24 30 36 42 48
Time,
[h]
Fig. 9Two days measured conversion and storage characteristics of the
AlGaAs/Si/MH/NiOOH MBPEC solar cell.
product of the measured cell potential and cell, and the remainder consists of the
measured load current. power over load during illumination, as
Under constant 12-hour (AM0) illumi- illustrated in Fig. 10. In the dark, inclusive
nation, the long-term indoor cycling cell of storage losses, the stored energy is spon-
generated a nearly constant photocurrent taneously released and this power over
density of 21.2 (constant to within 1% or load during both 12-hour illumination and
±0.2 mA cm−2 ), and as seen in the top 12-hour dark periods is also summarized.
curves of Fig. 10, a photopower that var- The cell is a single physical–chemical de-
ied by ±3%. The cell’s storage component vice generating load current without any
exhibits the expected increase in charging external switching.
potential with cumulative charging, which
moves the system to a higher photopoten- 4.2.5.2 PECS Driving an External Fuel Cell
tial. The observed increase in photopower In the early 1980s, Texas Instruments,
during 12 hours of illumination is be- Inc. developed an innovative program
cause of this increase in photopotential based on a hybrid photovoltaic storage that
with cumulative charging. A majority of used imbedded multilayer photoanode and
the photogenerated power drives the redox photocathode silicon spheres and was
7 7
6 6
5 5
Potential over load Illumination cycle:

Load potential
4 Photopower 12 hour 135.5 mW cm−2 & 12 hour dark 4

Power over load Area: 0.22 cm2
Power
[mW]
[V]
3 3
2 AlGaAs/Si MH bipolar gap storage solar C 2

2/3 year summary
1 1
0 0
0 2 40 42 140 142 240 242
Time, days
Fig. 10 Eight-months photopotential and power characteristics of the AlGaAs/Si/MH/NiOOH
solar cell under fully charged AM0 conditions. Each day, the cell is illuminated for 12 hours and
is in the dark for 12 hours.
designed to provide a close match between Photo cathode:

their maximum power point voltage and a Contact metal | Ohmic contact | p-Si |
solution phase bromine oxidation process n-Si | Surface metal | Solution
in acidic solution. The program was
discontinued, but the system has several A description of the Texas Instrument cell
attractive features. action is provided in Fig. 11.
Cell: 3 [51]. In the bromine-imbedded Si
sphere system, energy stored as bromine 4.2.6
is recovered in an external hydrogen Other Examples of Photoelectrochemical
bromine fuel cell. The conversion and Storage Cells
storage reaction and cell configuration are
summarized by Photoelectrochemical storage cells of dif-
ferent configurations have been suggested,
2HBr + H2 O + hν −−−→ designed, and tested for their performance,
H2 + Br2 + H2 O (15) under sunlight or artificial illumination.
Although none of these configurations
Photo anode: has attained the high solar to electri-
Contact metal | Ohmic contact | n-Si | cal conversion and storage efficiency of
p-Si | Surface metal | Solution the system in the earlier sections, they
Glass panel
Separator cover
Br2
Hydrogen HBr
bromide anode
H2 cathode
Br h 1/2 Br2
e− 1/2 H2
H Metal
e h
p n np
e
h Glass p n h− p n
e
Conductive plane
Storage
Hydrogen Fuel cell and heat
storage converter exchanger
Electrical energy Thermal energy

Fig. 11 Schematic diagram of the Texas Instruments Solar Energy
System. Illumination occurs through the electrolyte to produce hydrogen
gas that can be stored as a metal hydride and bromine that can be stored
as aqueous tribromide. A hydrogen-bromine fuel cell is used to convert
chemical to electrical energy and regenerate the hydrogen bromide
electrolyte in a closed loop cyclic system. Thermal energy can be extracted
through a heat exchanger.
are of significant scientific interest and The back reaction between Leucoth-
form a solid basis for further develop- ionene (TH4 2+ ) and Fe3+ is slow. Leucoth-
ment toward future systems. The following ionene is oxidized at the SnO2 electrode.
three sections, covering PECS with either
a solution, solid, or intercalation storage TH4 2+ −−−→ TH+ + 2e− + 3H+ (17)
redox processes, and provides a brief sum- Ferric cation is reduced at a Pt electrode
mary of many of these investigations, in a second compartment.
with a particular emphasis on their per- As a result, the concentration ratio of
formance. Fe3+ /Fe2+ is increased in the first com-
partment and decreases in the second
4.2.6.1 PECS Cells with Solution Phase compartment, which is equivalent to a dif-
Storage ference in chemical potential. The system
Cell: 4 [52]. This is an example of the use returns to its original uncharged state by
of photoexcitable absorbers to promote a discharging in the dark. Only 60 mV of po-
redox process using the following reaction tential difference is equivalent to a decade
sequence: change of the Fe3+ /Fe2+ concentration ra-
tio. Hence, this concentration cell does
not generate a significant potential and
TH+ +2Fe2+ +3H+ −−−→ TH4 2+ +2Fe3+
the power density is low. The cell has the
(TH+ = Thionene acetate) (16) configuration
SnO2 | 0.0001 M Thionene Acetate, difference between Pt electrodes before

0.01 M FeSO4 , 0.001 M Fe2 (SO4 )3 charging = 720 mV, and after charging =
H2 SO4 , pH = 1.7–2 | Ion-exchange 840 mV.
membrane | 0.01 M FeSO4 0.001 M Cell: 6 [54]. In this cell, a wide gap
Fe2 (SO4 )3 | Pt semiconductor (Eg = 3.3), BaTiO3 is used,
capable of absorbing only near UV ra-
Studied cell characteristics: Polycrys-
diation and comprising less than 5% of
talline photoelectrode, Pt counter elec-
available solar energy. This limits its prac-
trode, Eredox of Fe3+ /Fe2+ = 0.77 V vs
tical use for solar energy conversion. The
SHE, standard hydrogen electrode illu-
following storage couple is used:
mination 40–50 mW cm−2 by a tungsten
lamp, initial current during discharge = Light
9.1 µA, and initial voltage = 10.9 mV. Ce3+ + Fe3+ −
←−−
−−
→ 4+
− Ce + Fe
2+
(19)
Cell: 5 [53]. The photoelectrode used in Dark
this investigation was a powder-pressed The storage system operates well away
sintered pellet. The high band gap of from the maximum power point of the
the semiconductor used (n-Pb3 O4 ; Eg = semiconductor device, and therefore stor-
2.1 eV) and electron hole recombinations age and discharge efficiency is poor. The
at grain boundaries have contributed to an cell uses a salt bridge between the two
observed low-conversion efficiency, which compartments and is of the form
drives the overall cell reaction:
Single crystal | n-BaTiO3 | 0.1 M Ce2
Dark
− −
I + 6OH + 6Fe 3+ −−−
← −−
→− (SO4 )3 , 0.005 M Ce(SO4 )2 | Salt bridge |
Light 0.1 M Fe2 (SO4 )3 , 0.005 M FeSO4 | Pt
IO3 − + 3H2 O + 6Fe2+ (18) Studied cell characteristics: Pt counter
electrodes, three-electrode configuration
A saturated salt bridge is used between as in Fig. 4, Eredox of Ce4+ /Ce3+ =
the cell compartments to minimize mem- 1.45 V vs SHE, Eredox of Fe3+ /Fe2+ =
brane IR loss, but it allows the active redox 0.77 V vs SHE, illumination sunlight,
species to mix and chemically combine conversion efficiency = 0.01%, FF = 0.26,
across the junction. The cell has the form photopotential = 730 mV, Vmax = 0.33 V,
charge efficiency 15%, potential across two
n-Pb3 O4 | 0.1 M Fe3+ , Saturated
Pt terminals of the charged cell = 0.60,
Fe2+ | Salt bridge | 0.1 M IO3 − ,
and short-circuit current = 0.12 mA.
0.1 M I− | Pt
Cell: 7 [55]. In this study, attempts have
Studied cell characteristics: Polycrys- been made to improve the behavior of a
talline photoelectrode, Pt counter elec- MoS2 electrode in HBr electrolyte. Equili-
trodes, three-electrode configuration as in bration between the electrode and HBr has
Fig. 4., Eredox of Fe3+ /Fe2+ = 0.77 V vs been improved by subjecting the electrode
SHE, Eredox of IO3 − /I− = −0 − 26 V vs to a dark anodic potential. A ratio of 10
SHE, Illumination 60 mW cm−2 quartz between the areas of counter and work-
halogen lamp, conversion efficiency = ing electrode is another favorable feature
0.09%, FF = 0.38, photopotential = in this study to minimize the polariza-
172 mV, Vmax = 172 mV, charge effi- tion resistance at the counter electrode.
ciency of the battery = 74%, potential Nafion-315 membrane contributes only a
moderate resistance of 20 ohm cm2 . The anthraquinone redox couple have been
storage reaction and cell configuration are demonstrated in this study. Any H2 evolu-
tion would carry out direct hydrogenation
Light of AQ and associated side reactions, and
2Br− + 3I2 ←
−−−
−−
→− Br2 + 2I3
−
(20)
Dark
therefore a carbon electrode has been se-
lected because of the H2 over potential
n-MoSe2 | 0.1 M HBr, 0.01 M Br2 |
on this electrode. The cell underwent sev-
Nafion 315 | 1 M KI, 0.18 M I2 | Pt eral deep charge and discharge cycles with
Studied cell characteristics: Single reproducible performance. The storage re-
crystal photoelectrode, Pt counter action and cell form are
electrodes, cell configuration as in Fig. 4, Light
Eredox of Br/Br− = 1.087 vs SHE, Eredox 2I− + AQ + 2H+ −
←−−
−−
→− AQH2 + I2 (22)
of I3 − /I− = 0.534 V vs SHE. Illumination Dark
200 mW cm−2 Xe lamp, conversion n-WSe2 | 1 M KI, 0.1 M Na2 SO4 , 0.5 M
efficiency = 6.2%, potential across two Pt H2 SO4 | Saturated KCl bridge | 5 × 10−2
terminals of the charged cell = 0.49, and M AQ, 0.5 M H2 SO4 | C
short-circuit current = 0.5 mA.
Cell: 8 [55]. The next cell also uses a Studied cell characteristics: Single-crys-
Nafion membrane, but makes use of n- tal photoelectrode, C counter electrode
CdSe to drive a polysulfide–polyselenide during charging, Pt during discharg-
storage couple. Low-output power density ing. The cell configuration is similar
is the biggest drawback in this cell. The to that in Fig. 4. Eredox of I3− /I− =
storage reaction and cell configuration are 0.534 V vs SHE, Eredox of AQ/AQH2 , il-
lumination 150 mW cm−2 He–Ne laser
Dark (632.8 nm), conversion efficiency = 9%,
S0 + Se2− ←
−−−
−−
→ 2−
− S + Se
0
(21)
Light discharge across a 10 ohm load produces
a current of 1 mA cm−2 , and open-circuit
n-CdSe | 1 M in Na2 S, S, NaOH | Nafion voltage 200 mV.
315 | 1 M in M Na2 Se, Se, NaOH | Pt Cell: 10 [44]. This study uses a p-WSe2
Studied cell characteristics: Poly- photocathode rather than n-WSe2 . During
crystalline photoelectrode, Pt counter the cell discharge, oxidation of AQH2 at
electrodes, cell configuration as in the surface of p-WSe2 indicates that the
Fig. 4, Eredox of polysulfide electrolyte = electrode has the duel role of being a
−700 mV vs SCE, saturated calomel elec- cathode during the charging and being the
trode, Eredox of Se2 2− /Se2− = −800 mV vs anode during the discharge. As discussed
SCE, illumination 100 mW cm−2 Xenon earlier, this limits the activity and low-
lamp, conversion efficiency 4%, FF = current densities were observed. The
0.45, photovoltage = −400 mV, charged storage reaction and cell configuration are
cell has an open-circuit voltage of 60 mV, Light
and initial current across a 100 ohm across AQ + 2H+ + 2I −
←−−
−−
→− AQH2 + I2 (23)
Pt electrode = 0.5 mA. Dark
Cell: 9 [44]. This study uses organic re- p-WSe2 | 5 × 10−2 M AQ | Saturated |
dox species for energy-storage purposes. 1 M KI, 0.5 M H2 SO4 | Pt
Stability of the n-WSe2 photoanode in Single crystal | 0.5 M H2 SO4 | Salt
iodine electrolyte and the stability of bridge | 0.5 M Na2 SO4 |
Cell: 11 [55]. The theoretical band gap of Studied cell characteristics: Polycrys-
WSe2 provides a near ideal single band gap talline photoelectrode, Pt counter elec-
match for the solar spectrum. But the fol- trodes, cell configuration is similar
lowing cell has some disadvantages. These to Fig. 4, Eredox of O2 , H+ /H2 O
include the low solubility of the storage re- couple = 1.23 V vs NHE at pH = 1,
dox couple employed, MV2+ and MV+• and Eredox of Ag/Ag+ = 0.80 V vs NHE,
the possibility of undesirable side reactions normal hydrogen electrode illumination
of the radical ion MV+• . Using dual (n-type 500 W Hg lamp, conversion efficiency =
and p-type) photoelectrodes expands the 1%, photopotential = 0.28 V vs NHE,
potential regime one can access for the open-circuit voltage of the charged
redox-storage couple. The storage reaction cell = 0.28 V, and short-circuit current =
and cell configuration are 0.3 mA cm−2 .
Light Cell: 13 [57]. Of the four TiO2 Cells
2I− + 2MV2+ −
←−−
−−
→ +•
− 2MV + I2 (24) 10–13, the following cell exhibited the
Dark highest short-circuit discharge current and
−
n-WSe2 | I | MV2+ | p-WSe2 voltage. However, during the charging
process, a stationary concentration of
4.2.6.2 PECS Cells Including a Solid
Ce4+ was observed in the photoanode
Phase–Storage Couple compartment. This suggest the existence
The earlier experimental investigations, of competing process that consumes the
Cells 2–9, use only solution phase redox oxidized species Ce4+ . The later is known
couples. However, as indicated in the to participate in photochemical reactions
following examples, a solid phase–storage under illumination [10]. Considering the
couple may also be employed, which low concentration of the reduced form of
in principle tends to increase the cell’s active materials used with the photoanode,
storage capacity. there is a possibility that the water
Cell: 12 [57]. Having at least one oxidation becomes the dominant process
component in insoluble form may add during charging. In this study, it was
compactness into the cell configuration, observed that with a passage of a charge
although low conductivity of the insoluble of 10 coulomb during charging, Ce4+
active component may cause significant present in the photoanode compartment
polarization losses associated with the accounted for only 22% of the charge. In
storage electrode, as exemplified by the this second TiO2 photoelectrode cell, the
low conductivity of silver (chloride) in one storage reaction and cell configuration are
of the next cells. The next four cells use Light
a TiO2 polycrystalline photoelectrode. In Ce3+ + Ag+ ←
−−−
−−
→ 4+
− Ce + Ag (26)
the first cell, the storage reaction and cell Dark
configuration are TiO2 | 1 M HNO3 , 0.05 M Ce2 (SO4 )3 ,
Light 0.1 M Ce(SO4 )2 | Anion Specific Mem-
2H2 O + 4Ag+ −
←−−
−−
→ +
− 4H + 4Ag + O2 brane | 1 M AgNO3 , 1 M KNO3 | Ag
Dark
(25) Studied cell characteristics: Polycrys-
TiO2 | 1 M HNO3 , 1 M KNO3 | Anion talline photoelectrode, Pt counter elec-
Specific Membrane | 1 M AgNO3 , 1 M trode, cell configuration illumination etc.
KNO3 | Ag are similar to the earlier cell, Eredox
of Ce4+ /Ce3+ vs SHE, charge effi- competing reaction of H2 evolution. Cell

ciency of the cell without stirring = voltage of the charged cell is higher than
18%, open-circuit voltage of the charged the photovoltage available, which indicates
cell = 0.76, and initial short-circuit cur- the possible influence of another redox
rent 1.3 mA cm−2 . couple Ni(OH)2 /NiOH− occurring at a
Cell: 14 [57]. In this third TiO2 photo- higher redox potential. Only about 55% of
electrode cell, the storage reaction and cell the charge stored can be recovered during
configuration are discharge. The possibility of self-discharge
reactions because of imperfect permeabil-
Light
ity of the membrane has been cited as a
2Fe2+ + Cu2+ −
←−−
−−
→ 3+
− 2Fe + Cu (27)
Dark possible cause, and is further complicated
by the complex ferro/ferricyanide equilib-
TiO2 | 1 M KNO3 , 0.01 M FeSO4 | Anion
ria that is known to occur (Licht, 1995).
Specific Membrane | 0.025 M CuSO4 , 1 M
In this cell, the storage reaction and cell
KNO3 | Cu
configuration are
Studied cell characteristics: Poly-
crystalline photoelectrode, Eredox of Light
Cu2+ /Cu = 0.34 V vs NHE, Eredox of 2Fe(CN)6 4− + Ni2+ ←
−−−
−−
→ 3−
− Fe(CN)6 + Ni
Dark
Fe3+ /Fe2+ = 0.77 vs NHE, open-circuit (29)
voltage of the charged cell = 0.3, and short-
circuit current 1.5 mA cm−2 . n-GaP | 0.2 M K2 SO4 , pH = 6.7, 0.05 M
Cell: 15 [57]. The wide band gap of K3 Fe(CN)6 , 0.05 M K4 Fe(CN)6 | Anion
Specific Membrane | 0.05 M K2 SO4 , 0.2 M
TiO2 is not an appropriate match to
NiSO4 , 0.06 M NiCl2 | Pt
the solar spectrum. In this fourth TiO2
photoelectrode cell, the storage reaction
Studied cell characteristics: Single crys-
and cell configuration are
tal photoelectrode, Pt counter electrodes,
Light cell configuration is similar to Fig. 4.
−
Fe2+ + AgCl ←
−−−
−−
→ 3+
− Fe + Ag + Cl Eredox of Fe(CN)6 3− /Fe(CN)6 4+ is 0.36 V
Dark vs NHE, Eredox of Ni2+ /Ni = −0.25 V
(28)
vs NHE, illumination 500 W Hg lamp,
conversion efficiency 13% for 450–540-
TiO2 | 0.2 M KCl, 0.01 M FeCl2 |Anion
nm region, photovoltage = 0.63 V, open-
Specific Membrane | 0.2 M KCl | AgCl | Pt
circuit voltage of the charged cell = 0.75 V,
Studied cell characteristics: Polycrys- short-circuit current = 4.3 mA cm−2 , and
talline photoelectrode, cell configuration charge efficiency = 55%.
and illumination are the same as in the Cell: 17 [55]. The conversion efficiency
previous cell, Eredox Fe3+ /Fe2+ = 0.77 V data in the following cell reflect the
vs NHE, Eredox of AgCl/Ag+ = 0.22 V vs poor quality of the GaAs material that
NHE, open-circuit voltage of the charged was used, although in other studies,
cell = 0.39 V, and short-circuit current = there has been higher efficiency GaAs
0.4 mA cm−2 . PEC (without storage). In this study,
Cell: 16 [58]. In this next cell, Ni is significant polarization was observed and
deposited during charge at 80% charge performance data of the storage cell was
efficiency. Losses may be because of the not reported. The storage reaction and cell
configuration are discharge through Pt and Cd electrodes

with a 100-ohm load, and a current of
Dark
Cd + Se2 2− + 2OH− ←
−−−
−−
→ 8.3 mA cm−2 flowed at cell voltage close to
−
Light 175 mV.
Cell: 19 [59]. In this study, the possi-
Cd(OH)2 + 2Se2− (30) bility of using organic semiconductors to
drive storage processes is demonstrated.
n-GaAs | 0.1 M Na2 Se, 0.1 M Se, 1 M The process is in principle similar to
NaOH | Nafion | 2 M NaOH | Cd a concentration cell. During photocharg-
Studied cell characteristics: Single crys- ing, Prussian Blue (PB, Fe4 [FeII (CN)6 ]3 ) is
tal photoelectrode, Pt counter electrode, reduced at the photocathode and PB is ox-
cell configuration is similar to Fig. 4, Eredox idized at the anode. In the dark, the redox
of Se2− /Se2 2− = −800 mV vs SCE, Eredox process involving PB is reversed producing
of Cd/Cd(OH)2 = −1050 mV vs SCE. an electron flow. Process ability, stability,
Illumination 100 mW cm−2 Xe lamp, and lack of photocorrosion make these
conversion efficiency = 4%, FF = 0.53, low band gap organic materials very attrac-
photopotential = −500 mV, and short- tive for photoelectrochemical applications.
circuit discharge current of the storage However, they are defect-based systems,
cell in the dark using Pt electrodes = and the very low conversion efficiencies
14.6 mA cm−2 . and self-discharge appear to outweigh
Cell: 18 [55]. The next two cells use a these benefits. The storage reaction and
polycrystalline n-CdSe photoanode. The cell configuration are
following cell exhibited steady current-
bilayer electrode: FeII II
4 [Fe (CN)6 ]3
4−
+ 4h+
time and voltage-time curves during the
photoelectrochemical charging and dark Light
−−−
−−
→ III II
discharging. The flat discharge curve pre- ← − Fe4 [Fe (CN)6 ]3 (32)
Dark
vailed until the capacity of the sulfide
−
electrolyte is exhausted. The storage re- counter electrode: FeIII II
4 [Fe (CN)6 ]3 + 4e
action and cell configuration are Light
−−−
−−
→ II II 4−
← − Fe4 [Fe (CN)6 ]3 (33)
Dark Dark
−−−
−−
→ − 2−
Cd + S + 2OH ← − Cd(OH)2 + S
Light ITO | P3MT | PB | 0.2 M KCl, 0.1 M
(31) HCl | PB | ITO
Studied cell characteristics: Illumination
n-CdSe | 0.1 M in NaOH, Na2 S, 1 M in S,
500 W Xenon lamp, the ITO/P3MT elec-
Na2 Se, Se | Nafion | 2 M NaOH | Cd
trode has open-circuit voltage = 0.44 V,
Studied cell characteristics: Polycrys- short-circuit photocurrent 0.09 µA cm−2 ,
talline photoelectrode, Pt counter elec- and charge efficiency of the storage cell =
trodes, cell configuration is similar to 40%.
Fig. 4. Eredox of Sx 2 /S2− = −700 mV vs Cell: 20 [60]. Metal ions introduced into
SCE, Eredox of Cd(OH)2 /Cd = −1050 mV a solid β-alumina lattice behave like ions in
vs SCE, illumination 100 mV cm−2 Xe solution. This study illustrates a compact
light, conversion efficiency = 4%, FF = solid-state storage cell that can be charged
0.45, photovoltage = −400 mV during using solar energy. During charging, Fe
and Ti change their oxidation state and the n − CdSe | 1 M in NaOH, Na2 S,
charge balance is maintained by the mi- S Nafion-315 | 0.1 M ZnO, 1 M NaOH | C
gration of Na+ ions from one phase to the
other. In the actual cell design, an n-type Studied cell characteristics: Polycrys-
semiconductor is connected to the alu- talline photoelectrode, Ni counter elec-
mina phase containing Ti and p-material trode, basic cell configuration is based
is connected to the phase containing Fe. on Fig. 4, Eredox of Sx 2 /S2− = 0.500 V vs
Limitations are the comparatively slow SHE, Eredox of Zn/Zn(OH)4 2− = −1.25 V
diffusion of ions in the solid electrolyte vs SHE, artificial illumination, conversion
and resistance to ionic movement at var- efficiency = 3%, photovoltage = −0.50 V,
ious phase boundaries, and lower the during discharge through 75-ohm load
energy output during discharge. In this between C and Ni discharge current =
device, back wall illumination demands 10 mA, and voltage = 0.6 V.
the use of very thin semiconductor layers Cell: 22 [61]. This cell takes advantage
to minimize absorption losses and has the of photocorrosion to drive a storage cell.
general form Under illumination, n-CdSe is decom-
posed and p-CdSe is electroplated, and
n-semiconductor | Na2 O.11(AlFeO3 ) | the reverse occurs during cell discharge.
Na2 O.11(Al2 O3 ) | Na2 O.11(AlTiO3 ) | However, photoactivity depends on an op-
p-semiconductor timized semiconductor surface, and in an
environment where the surface is changed
Cell: 21 [47]. In this detailed study, selec- constantly, the surface optimization is lost.
tion of a Nafion-315 membrane was done This and the poor kinetics of the p-type
on the basis of (1) stability in high alkaline photoreduction result in a continual dete-
sulfide solutions, (2) low IR drop, and (3) rioration of the photoactivity and cause low
low permeability to sulfide. Maintaining an photoefficiency and low-discharge power
area ratio of 1 : 8 between photo and storage density. The storage reaction and cell con-
electrodes has minimized polarization at figuration are
the storage electrode. The storage system
was driven by three semiconductor PEC Light
devices connected in series. Charging was CdSe + 2h+ ←
−−−
−−
→ 0
− Se + Cd
2+
(35)
Dark
done up to 90% of the capacity followed
by complete discharge. Overall observed and the other electrode in photoelectro-
charge efficiency was 83%. Although the plated by Cd
system was not fully optimized with re-
Light
spect to photoelectrode, electrolyte, and CdTe + 2e− ←
−−−
−−
→ 0 2−
− Cd + Te (36)
storage, voltage efficiency of 75% was ob- Dark
tained during discharge. Discharge curves n-CdSe | 0.1 M CdSO4 | p-CdTe
were flat until the stored active material
was fully consumed. The storage reaction Cell: 23 [46]. This is a detailed study of a
and cell configuration are thin film cell with moderately high outdoor
solar efficiency, high storage efficiency,
Light
− and an output that is highly invariant de-
S2− + Zn(OH)4 2− ←
−−−
−−
→ 0
− S + Zn + 4OH
Dark spite changing illumination. This study
(34) provides extensive details of the choice
of photoelectrode, membrane, and elec- has the design as shown in beginning of

trochemistry of the tin–tin sulfide redox the chapter (Fig. 2) and uses conversion
storage. Cd(Se,Te) electrodes, compared and storage reactions described in the
to CdSe, improve the band gap match and earlier cell and a configuration
increase solar-conversion efficiency. Two
photoelectrodes in series were used to pro- n-Cd(Se,Te) | 0.8 M Cs2 S, 1 M Cs2 S4 |
vide a voltage match to the storage redox Redcad Membrane | 1.8 M Cs2 S | SnS | Sn
couple in a cell of the form of Fig. 12. Studied cell characteristics: The PEC had
The conversion and storage reactions a power conversion efficiency of 12.7%
and cell configuration are presented as under 96.5 mW cm−2 insolation and volt-
Light age at maximum power point was of
−−−
−−
→ 2−
SnS + 2e ← − S + Sn Storage (37) −1.1 V vs SHE, sufficient to drive the
Dark SnS/Sn storage system. Under direct illu-
Light mination, the 0.08 cm2 single crystal pho-
S2− −
←−−
−−
→− S + 2e Photo electrode (38) toelectrode generated more than 1.5 mA
Dark
through the 3 cm3 SnS electrode driving
n-Cd(Se,Te) | 2 M in NaOH, Na2 S, S | SnS reduction while supporting 0.33 mA
Redcad Membrane | 2 M in NaOH, Na2 S | through a 1500 load simultaneously at a
SnS | Sn photogenerated 0.495 V. In the dark spon-
Studied cell characteristics: Bipolar se- taneous oxidation drive, the load with
ries polycrystalline photoelectrode, CoS storage efficiency over 95%. The total con-
counter electrodes, cell configuration is version efficiency, including conversion
as shown in Fig. 4 without the need and storage losses, was 11.8%.
of switches E or F . Eredox of S/S2− =
−0.48 V vs NHE, Eredox of SnS/Sn,S2− = 4.2.6.3 PECS Cells Incorporating
−0.94 V vs NHE. Illumination sunlight, Intercalation
500 mWhr cm−2 per day, conversion ef- In photointercalation, illumination drives
ficiency 6–7%, photovoltage = −600 mV, insertion storage into layer type com-
and storage efficiency >90%. After two pounds [62]. The photointercalation pro-
weeks of continuous operation the over- cess can be characterized as
all solar to electrical efficiency (including
conversion and storage losses) is 2–7%. TX2 + e− (h) + p+ (h) + Msol + −−−→
Cell: 24 [4]. The earlier cell is improved TMIN X2 + p+ (39)
by a series of solution phase optimizations
(cesium electrolyte with low hydroxide where TX2 is generally a nonintercalated
and optimized polysulfide), to provide transition metal dichalcogenide. For this
a higher photopotential and improved process to occur without the assistance of
stability and also the use of a single an external power source, a counter elec-
crystal, rather than thin film, Cd(Se,Te) trode is driven at an electrode potential
to also improve photopotential and cell negative to that of the layer type interca-
efficiency, as described earlier in Fig. 12. lating electrode. The process is generally
Because of the higher photopotential, restricted to p-type materials. The develop-
only a single photoelectrode is required ment of this concept has been slow because
to match the storage potential and high of dearth of materials that are stable semi-
overall efficiencies are observed. The cell conductors and at the same time behave as
4.2 Photoelectrochemical Solar Energy Storage“hbox Cells 343
Bipolar Cd Se Tel [polysulfide]lCoS PEC

with in-situ tini[sulfide]ll[polysulfide]lCos storage
Cd Se Te photo hν hν Cd Se Te photo
CoS counter
A2 CoS counter
A
Membrane
B
Tin storage Load
Fig. 12 A bipolar thin film photoelectrochemical solar cell with in situ storage.
Compartments A and A2 contain alkali polysulfide solution and compartment B
contains alkali sulfide solution.
intercalating compounds that are able to because to reach the junction, light has to
exchange guest ions and molecules with an travel several layers. The cell functions in
electrolyte in a reversible manner, and yet the same manner as the earlier cell, and
that is not disruptive to photon absorption. the configuration is given by
Cell: 25 [63]. In this cell, Eredox of copper
thiophosphate is variable depending on Conductive Glass | Cu | Cu+ Conducting
the degree of intercalation. A limitation of solid electrolyte | p-Cu2 Te | n-CdTe | Mo
this system is poor-discharge kinetics and
low-energy density of the discharge. The
4.2.7
cell configuration is given by Summary
Cu3 PS4 | 0.02 M CuCl | CH3 CN | Cu2 S
Conversion and storage of solar energy
Studied cell characteristics: Eredox of is of growing importance as fossil fuel
Cu+ /Cu0 = −0.344 vs NHE, illumination energy sources are depleted and stricter
117 mW cm−2 Xe lamp, photopotential = environmental legislation is implemented.
l00 mV, charging current < 50 µA cm−2 , Although society’s electrical needs are
and discharge current < 10 µA cm−2 . largely continuous, clouds and darkness
Cell: 26 [64]. This cell illustrates another dictate that photovoltaic solar cells have
all solid state design for a thin storage an intermittent output. Photoelectrochem-
cell. p-Cux S changes its electrode potential ical systems have the potential to not
with changes in its composition. During only convert but also store incident so-
charging, Cu is oxidized at n-CdS surface lar energy. Design component and system
while it is reduced at the Cu electrode. considerations and a number of photoelec-
Between the two electrodes Cu+ ion trochemical solar cells with storage have
transport process takes place in the solid been reviewed in this chapter.
state electrolyte. The cell configuration is
given by
Acknowledgment
Cu | n-CdS | p-Cu2 S | RuCl4 I5 Cl3.5 | Cu
S. Licht is grateful to Dharmasena Per-
Cell: 27 [64]. As with the earlier cell, amunage and for support by the BMBF
this final cell requires a very thin design Israel–German Cooperation.
References 26. B. Miller, S. Licht, M. E. Orazem et al., Pho-

toelectrochemical Systems, Crit. Rev. Surf.
1. S. Licht, O. Khaselev, T. Soga et al., Elec- Chem. 1994, 3, 29.
trochem. Solid State Lett. 1998, 1, 20. 27. D. Cahen, G. Hodes, J. Manassen et al., ACS.
2. (a) S. Licht, O. Khaselev, T. Soga et al., J. Symp. Ser. No. 1980, 146, 369.
Phys. Chem. 102, 2536; (b) ibid, 2546. 28. G. Hodes, in Energy Resources Through Pho-
tochemistry and Catalysis, (Ed.: M. Gratzel),
3. B. O’Regan, M. Grätzel, Nature 1991, 353,
Academic Press, New York, 1983, p. 242.
737.
29. A. Heller, K. C. Chang, B. Miller, J. Elec-
4. S. Licht, G. Hodes, R. Tenne et al., Nature
trochem. Soc. 1977, 124, 697.
1987, 326, 863.
30. R. Tenne, G. Hodes, Appl. Phys. Lett. 1980,
5. G. De Maria, L. D’Alesso, E. Coffari et al., 37, 428.
C. A. Tiberio, Solar Energy 1985, 35, 409. 31. M. Tomkiewicz, I. Ling, W. S. Parsons, J.
6. M. Grätzel, Ber. Bunsenges. Phys. Chem., Electrochem. Soc. 1982, 129, 2016.
1980, 84, 981. 32. J. Reichman, M. A. Russak, J. Appl. Phys.
7. K. Kalyanasundaram, Coord. Chem. Rev. 1982, 52, 708.
1982, 46, 159. 33. B. A. Parkinson, A. Heller, B. Miller, J. Elec-
8. A. Harriman, Photochemistry 1986–1987, 19, trochem. Soc. 1979, 126, 954.
509. 34. G. Hodes, J. Manassen, D. Cahen, J. Elec-
9. K. Kutal, Chem. Educ. 1984, 60, 882. trochem. Soc. 1980, 127, 544.
10. L. J. Heidt, A. F. McMillan, Science 1953, 35. A. J. McEroy, M. Etman, R. Memming, J.
117, 712. Electroanal. Chem. 1985, 190, 2212.
11. R. Memming, Top. Curr. Chem. 1988, 143, 36. J. Reichman, M. A. Russak, J. Electrochem.
79. Soc. 1984, 131, 796.
12. A. Fujishima, K. Honda, Nature 1972, 238, 37. D. Cohen, Y. W. Chen, Appl. Phys. Lett. 1984,
37. 45, 746.
13. J. M. Bolts, A. B. Bocarsly, M. C. Palazzotto 38. S. Licht, N. Myung, R. Tenne et al., J. Elec-
et al., J. Am. Chem. Soc. 1979, 101, 1378. trochem. Soc. 1995, 142, 840; S. Licht,
14. H. P. Maruska, K. A. Ghosh, Solar Energy N. Myung, ibid, L129; ibid, 8412.
1978, 20, 443. 39. S. Licht et al., J. Electrochem. Soc. 1985, 132,
15. G. Hodes, L. Thompson, J. Dubow et al., J. 1076; ibid, 1986, 133, 52; ibid, 272; ibid, 277;
Am. Chem. Soc. 1983, 105, 324. ibid, 269.
16. R. C. Kaintala, B. Zelenay, J. OM. Bockris, J. 40. S. Licht, Solar Energy Mat. Solar Cells 1995,
Electrochem. Soc. 1986, 133, 248. 38, 3012.
17. R. C. Kaintala, J. OM. Bockris, Int. J. Hydro- 41. S. Licht, Nature 1987, 330, 148.
gen Energy 1988, 13, 3712. 42. S. Licht, D. Peramunage, Solar Energy 1994,
52, 197.
18. O. Khaselev, J. Turner, Science 1998, 280,
43. S. Licht, D. Peramunage, Nature 1990, 345,
4212.
330.
19. S. Licht, B. Wang, T. Soga et al., H. Tribut-
44. B. Keita, L. Nadjao, J. Electroanal. Chem.
sch, J. Phys. Chem. B. 2000, 104, 8920.
1984, 163, 171.
19. S. Licht, B. Wang, T. Soga et al., H. Tribut- 45. S. Licht, F. Forouzan, J. Electrochem. Soc.
sch, J. Phys. Chem. B. 2000, 104, 8920. 1995, 142, 1539; ibid, 1546.
20. J. Manassen, G. Hodes, D. Cahen, CHEM- 46. S. Licht, J. Manassen, J. Electrochem. Soc.
TECH 1981, 112. 1987, 134, 1064.
21. A. J. Bard, M. S. Wrighton, J. Electrochem. 47. P. Bratin, M. Tomkiewicz, J. Electrochem. Soc.
Soc. 1977, 124, 1706. 1982, 129, 2649.
22. A. Heller, G. P. Schawartz, R. G. Vadimisky 48. C. H. Henry, J. Appl. Phys. 1980, 51, 4494.
et al., J. Electrochem. Soc. 1978, 25, 1156. 49. S. Licht, B. Wang, T. Soga et al., Appl. Phys.
23. G. Hodes, J. Manassen, D. Cahen, Nature Lett. 1999, 74, 4055.
1976, 261, 403. 50. B. Wang, S. Licht, T. Soga et al., Solar Energy
24. K. W. Frese, Appl. Phys. Lett. 1982, 40, 2712. Mat. Solar Cells. 2000, 64, 311.
25. F.-R. F. Fan, B. Wheeler, A. J. Bard et al., J. 51. J. White, F-R. Fan, A. J. Bard, J. Electrochem.
Electrochem. Soc. 1981, 128, 2042. Soc. 1985, 132, 544.
52. G. W. Murphy, Solar Energy 1978, 21, 403. 58. Y. Yonezawa, M. Okai, M. Ishino et al., Bull.
53. M. Sharon, S. Kumar, N. P. Sathe et al., Solar Chem. Soc. Jpn 1983, 56, 2873.
Cells 1984, 12, 353. 59. M. Kaneko, K. Takagashi, E. Tsuchida, J.
54. M. Sharon, A. Singha, Int. J. Hydrogen Energy Electroanal. Chem. 1987, 227, 2512.
1982, 7, 557. 60. A. Sammells, A. Ang, Patent 1980, US
55. P. G. Ang, P. A. F. Sammells, Faraday Dis- 4,235,9512.
cussions of the Chemical Society, Gen- 61. H. J. Gerritsen, W. Ruppel, P. Wurfel, J. Elec-
eral Discussions 1980, No. 70, Photoelectro- trochem. Soc. 1984, 131, 2037.
chemistry, St. Catherine’s College, Oxford, 62. H. Tributsh, Appl. Phys. 1986, 23, 61.
Sept 8–10. 63. G. Betz, S. Fiechter, H. Tributsch, J. Appl.
56. F. -R. F. Fan, H. S. White, B. L. Wheeler Phys. 1987, 62(11), 4597.
et al., J. Am. Chem. Soc. 1980, 102, 5142. 64. T. Tonomura, K. Teratoshi, 1986, JP
57. H. Hada, K. Takaoka, M. Saikawa, Y. Yone- Patent 62,249,366; ibid., JP Patent 62,249,
zawa, Bull. Chem. Soc. Jpn. 1981, 54, 1640. 3612.
4.3 each environmentally benign. Insolation

Solar Photoelectrochemical Generation of (solar radiation) on semiconductors can
Hydrogen Fuel generate significant electrical, electro-
chemical, or chemical energy [1–3]. Effi-
Maheshwar Sharon cient solar-driven water splitting requires
Indian Institute of Technology, Bombay, India a critical balance of the energetics of
the solar conversion and the solution
Stuart Licht
phase redox processes. The UV and visible
Israel Institute of Technology, Technion, Israel
energy-rich portion of the solar spectrum
4.3.1 is transmitted through H2 O (Fig. 1 [4] and
Introduction Fig. 2 [5]). Therefore, sensitization, such as
via semiconductors, is required to derive
Photoelectrolysis is a vast field and it is dif- the water-splitting process. In a solar pho-
toelectrolysis system, the redox-active in-
ficult to cover all its aspects in this chapter.
terfaces can be in indirect or direct contact
Therefore, this chapter confines its dis-
with the photosensitizer and can comprise
cussions to photoelectrolysis of water to
either an ohmic or a Schottky junction.
obtain hydrogen gas. Hydrogen (H2 ) is an
Independent of this interface composi-
important renewable source of energy, as
tion, the various parameters in models
water is the main precursor for hydrogen.
predicting solar water-splitting conversion
Water (H2 O), on electrolysis produces hy-
efficiency may be combined into general
drogen and oxygen (O2 ) in the ratio of parameters: (1) related to losses in optical
2 : 1. Moreover, when hydrogen is utilized energy conversion, ηphoto or (2) related to
as a fuel and oxidized to release its heat losses in redox conversion of H2 and O2 ,
chemically (burnt) or electrochemically (in ηelectrolysis . A combination of these param-
a hydrogen or air fuel cell), it produces eters yields an overall solar to electrolysis
water by consuming similar ratio of hydro- efficiency (excluding storage and utiliza-
gen and oxygen. Unlike other sources of tion losses) as
energy such as coal, gas, and oil, hydrogen
is a very clean source of energy. Hence, ηphotoelectrolysis = ηphoto × ηelectrolysis (1)
the hydrogen ⇔ water cycle seems to pro-
Early photoelectrolytic attempts used so-
vide the most appropriate path toward a
lution redox species such as Ce3+/4+ and
renewable source of energy.
displayed poor quantum yield [6]. Further
studies utilized semiconductors, such as
4.3.2 TiO2 (band gap, Eg = 3.0 eV) [7] or SrTiO3
Theoretical Consideration for Water (Eg = 3.2 eV) [8]. The wide Eg excludes
Electrolysis longer wavelength insolation, leading to
poor efficiency. H2 and O2 evolutions have
Solar energy–driven water splitting com- been enhanced using large surface area
bines several attractive features for energy or catalyst addition, but energy conversion
utilization. Both the energy source (sun) efficiency remains low [9]. Early photoelec-
and the reactive media (H2 O) are readily trolysis systems also combined p-type and
available and are renewable, and the re- n-type photoelectrodes [10, 11]. A two or
sultant fuel (generated hydrogen) and the more band gap configuration can provide
emission with fuel consumption (H2 O) are efficient matching of the solar spectra.
4.3 Solar Photoelectrochemical Generation of Hydrogen Fuel 347
Fig. 1 Absorption spectrum of

A
water (after K. W. Atanabe and
500
ε cm−1 Atm−1
M. Zelikoff, J. Opt. Soc. Am.
1953, 43, 753).
250
0
100 150 200
nm
Fig. 2 Solar spectral irradiance

(w m−2 nm−1 ) at air-mass ratio 2.0 a/m = 0
(a/m) of 0, 1, and 2 (after 1.6 a/m = 1
W m−2 nm−1
Nikola Getoff, Int. J. Hydrogen a/m = 2

Energy 1990, 15(6), 407). 1.2
0.8
0.4
0
500 1000 1500 2000
nm
p-InP and n-GaAs were demonstrated at or a Schottky junction electrochemical or

8.2% efficiency to generate H2 and O2 [12]. a solid-state junction can inhibit charge
A GaInP2 /GaAs photoelectrolysis cell was recombination, driving charge at photopo-
demonstrated at 12.4% efficiency [13] and tential, Vphoto . In these junctions, Vphoto is
a AlGaAs/Si photoelectrolysis cell was constrained by the saturation current, jo ,
demonstrated at 18.3% efficiency [14]. through to yield a maximum power pho-
A limited fraction of incident solar photopotential, Vpmax , less than 60% of the
tons has sufficient energy greater than band gap in efficient devices [15].
Eg to initiate charge excitation within a
Vphoto (jphoto ) = Vw + Vs =
semiconductor. A semiconductor configu-
! " # $ # $
ration can drive water electrolysis under nkT jphoto jphoto
ln 1 + × 1+
the conditions in which the generated e jo−w jo−s
photovoltage, Vphoto , is greater than the (2)
electrolysis potential, VH2 O . A complex
energetic challenge exists to choose a Vpmax < 0.6(EGw + Egs ) (3)
semiconductor system with band gap at a
To initiate electrolysis Vphoto must be
combined maximum power point voltage
greater than VH2 O . The thermodynamic
tuned to VH2 O , in a system also providing o
potential, EH 2O
, for the water-splitting
high ηphoto (high fill factor, photopotential,
reaction is given by
and photocurrent). Incident light of suf-
ficiently energetic photons are absorbed; H2 O −−−→ H2 + 12 O2 ;
less energetic photons are transmitted. o o o
EH 2O
= EO − EH ;
Absorbed photons, hν > EGw , can stim- 2 2
o ◦
ulate e− / h+ pair excitation. A pn-type EH 2O
(25 C) = 1.229 V (4)
The subsequent challenge is to opti- a thermodynamic photoefficiency describ-

mize sustained water electrolysis, without ing the generation of electrochemical free
considerable additional energy losses. Ef- energy for electrolysis:
fective water electrolysis must occur at a
#GH2 O
potential, VH2 O , near the photocell point (7)
of maximum power. VH2 O is greater than #Hph
EHo (= −#GH2 O /nF ) as a result of over-
2O In reality, how small can #Hph be made,
potential losses, ξ , in driving an electroly- as compared with #GH2 O ?
sis current density, j , through both the O2 For photosynthesis, it has been esti-
and the H2 electrodes: mated that of the 1.8-eV excitation energy
of chlorophyll, at least 0.6–0.8 eV are
VH2 O (j ) = EO2 (j ) − EH2 (j ) lost before the energy can be stored in
o o stable chemical products. Photosynthe-
= [EO 2
+ ξO2 (j )] − [EH 2
+ ξ H2 (j )] (5)
sis, however, requires many subsequent
Planar platinum and Ptblack are ex- electron transfer steps, all of which con-
amples of effective H2 electrocatalysts. tribute to efficiency losses. The molecules
Minimization of ξO2 is a greater chal- involved are also complicated and quite
lenge. In the absence of competing redox unstable compounds maintained through
couples, the faradic efficiency of H2 and self-organization. Photoelectrolysis sys-
O2 evolution approaches 100%, and the tems with many fewer components, and
ηelectrolysis is determined by the current with fewer degrees of freedom in terms of
limited VH2 O (j ): chemical reactivity, can better approach
the ideal energy conversion efficiency.
o
EH 2O
Contrarily, when fuel is consumed to
ηelectrolysis = ; reduce power (resulting in a fuel cell
VH2 O (j )
with efficiency #G/#H ) the optimal ef-
◦ 1.229 V
ηelectrolysis (25 C) = (6) ficiency for electrolysis must include en-
VH2 O (j ) tropy considerations, and this strongly de-
pends on the temperature enthalpy change
Thermodynamic heat considerations
(#HH2 O (T )):
can be applied to photoconversion. Some
amount of heat, Q, is always be lost due #HH2 O
to charge carrier relaxation and vibrations. ηelectrolysis−opt = (8)
#GH2 O
This heat difference involves an entropy
turnover Q/T = #S. Photoenergy conver- Practical electrolysis has to include
sion efficiencies can be optimized by min- losses so that the efficiency becomes
imizing #Hph . For band gap excitation, #HH2 O
#Hph corresponds to Eg . If the absorbed ηelectrolysis−opt = (9)
#GH2 O + losses
photon energy is much higher than Eg ,
then #Hph is correspondingly higher. If A thermoneutral potential, Etn (Etn =
Eg is much larger than the required elec- #HH2 O /zF ), is defined in which no heat
trolysis energy, #GH2 O , then the predicted turnover is observed and Etn = 1.48 V for
efficiency is low. The efficiency is higher water electrolysis [16]. If, because of effec-
if the photovoltaic system is matched to tive catalysis, the total electrolysis cell volt-
the chemical one. Consequently, there is age is close to 1.48 V, then ηelectrolysis ≈ 1.
If we return to Eq. (1) and substitute photoelectrochemical (PEC) solar cell con-
Eqs. (7) and (9), we obtain figuration similar to that described in
the earlier chapter. In brief, a PEC cell
ηphotoelectrolysis = ηphoto × ηelectrolysis
! " consists of two electrodes separated by
#GH2 O 1 + losses a suitable redox electrolyte. Both elec-
= (10)
#Hph #GH2 O trodes could either be n-type and p-type
Two conclusions arise from Eq. (10). semiconductors or one electrode could
First, it is seen that efficiency can be be a semiconductor (either n-type or p-
maximized if electrochemical ‘‘losses’’ can type) and the other electrode could be
be made small as compared with the Gibbs a noncorrosive metal. If both electrodes
free energy change for water electrolysis, are made of semiconductor, then the an-
#GH2 O . Second, it is seen that #HH2 O , the odic electrode should be made from n-type
enthalpy of water electrolysis (with little semiconductor and the cathodic electrode
temperature dependence), is involved in should be made from p-type semicon-
determining the overall efficiency as well ductor. This last configuration adds the
as #Hph , the enthalpy of photogenerated additional requirement of photocurrent
charge carriers. If these two can be matching through the two photoelec-
properly matched, a maximum overall trodes. Regarding n-type semiconductor
efficiency may be accomplished. (Fig. 3), during its illumination photogen-
erated carriers are generated, which assist
4.3.3 in the oxidation of electrolyte at the inter-
Photoelectrochemical Cell for face of the semiconductor and electrolyte.
Photoelectrolysis A reverse phenomenon takes place if a
p-type semiconductor is used, that is,
Conversion of solar energy into electrical photogenerated electrons are generated at
energy can be achieved by using a the semiconductor–electrolyte interface to
RL
0.3 cm
C A
Fig. 3 A schematic representation of a
photoelectrochemical cell. A, B, and C
are counter electrode, electrolyte, and hν
semiconductor electrode, respectively.
RL is used to vary the potential (V) to B
measure photocurrent (A) flowing
across two electrodes.
initiate reduction of the electrolyte. Thus, apply an external bias (Eb ) to facilitate the
in a PEC cell (Fig. 3), if water is used in- photoelectrolysis of water. The energy level
stead of a redox electrolyte, oxidation of of semiconductor and other photoelectro-
water can take place at the illuminated chemical reactions are shown in Fig. 4(b).
n-type semiconductor (C) and reduction The semiconductor anode was illuminated
can take place at counter electrode (A). with UV radiation. On illumination of
However, for photodecomposition of wa- TiO2 electrode, the photogenerated holes
ter rather than a single redox couple, two oxidize water to produce oxygen and the
different redox couples are used at the two photogenerated electrons are transferred
immersed electrodes, with a provision to to counter electrode to perform the reduc-
collect hydrogen and oxygen separately. tion of water at the platinum electrode.
4.3.4 4.3.4.1 Energetics of Photodecomposition

Photoelectrolysis of Water of Water
To accomplish photodecomposition of
Fujishima and Honda [7, 17] were the first water, so that the PEC cell supplies
to show the possibility of decomposing wa- the entire potential, energetic of the cell
ter through a PEC cell (Fig. 4a). They used should meet the requirements as shown
a photoelectrochemical cell with an an- in Fig. 5. It is assumed that Fermi energy
ode made up of a semiconductor electrode of n-type semiconductor (or its flat band
of n-TiO2 connected to a platinum black potential) is equivalent to water reduction
counter electrode through an external load. potential. Alternatively, instead of using
Because photopotential developed by the metal as a counter electrode, one could use
TiO2 electrode was insufficient, they had to a p-type semiconductor (Fig. 6). Because
E-bias
E-bias
e−
O2 H2 e−
e−
H2
e− H e− EF
e−
EC
H2O Eb
hν
EF
Pt EG
e− OH + H+
EV P+
NaOH Na2SO4 O2
Frit n -TiO2 Electrolyte
n -TiO2 Cathode Metal Helmholtz Pt
Anode contact layer
(a) (b)
Fig. 4 (a) Fujishima-Honda cell with n-TiO2 photoanode and Pt-cathode. (b) Schematic
energy level diagram of the cell. EV -valence band, EC -conduction band, EF -Fermi level,
EG -energy gap (for n-TiO2 ; EG = 3.0 eV), Eb -bias voltage, p+ -hole (after Nikola Getoff,
Int. J. Hydrogen Energy 1990, 15(6), 407).
−
Conduct band
− −
EF,sc 2− − −
H2 2H+ Fermi-level
nEF
Stored
hν ∆V energy
1 O + 2H+
H2O
pEF 2 2
2+
Valence band +
+
n- semiconductor Electrolyte Metal

Fig. 5 Energy scheme of a cell with one n-type semiconductor electrode for
photoelectrolysis of water. #V is stored energy for electrolysis. p EF is Fermi level of
photogenerated holes known as quasi-Fermi level, n EF is Fermi level of electron.
(after Heinz Gerischer, Pure Appl. Chem. 1980, 52, 2649).
Load
h+
e− EC
e 2H2O + 2e− 2OH− + H2
hν
e
EC ∼1.7 V EF
EF Eν
+
hν
+ 2h + 2OH− H2O + 1/2 O2

Eν
Fig. 6 Energy scheme of a cell utilizing n-type semiconductor and
p-type semiconductor. The difference between the energy levels of
valence band edge of n-type and conduction band edge of p-type
should be approximately equal to potential needed to electrolyze
water (VH2 O ).
the objective is to electrolyze water with calculate theoretical achievable efficiency

a freely available energy, light source for various band gaps. This relationship
for the illumination has to be solar gives a parabolic curve with a maximum
energy. Considering factors responsible theoretical efficiency of about 30% for
to produce photocurrent or photopotential a band gap of 1.4 eV (Fig. 7). But, a
generated by semiconductors of various semiconductor with a band gap of 1.4 eV
band gaps utilizing solar energy of various would give a maximum photopotential of
wavelengths and intensities, one can ∼
=700 mV (assuming its flat band potential
35
Black-body limit (AMO)

30
25 AM1.5 GaAs
Cu2S a-Si : H
Si a-Si : H : F
Efficiency
[%]
20 AMO
15
T = 300 K
Ge CdS
10
5
0.5 1.0 1.5 2.0 2.5
Semiconductor band gap
[eV]
Fig. 7 Theoretically calculated conversion efficiency of solar cell materials versus band
gap for single junction cells (after Adolf Goetzberger, Christopher Hebling, Sol. Energy
Mater. Sol. cells 2000, 62, 1).
matches with the reduction potential of with photopotential larger than VH2 O .
water), which is not sufficient to electrolyze Sharon and Rao [19] developed a photo-
water. On the contrary, for obtaining the electrochemical cell with a semiconductor
necessary electrolysis photopotential using separating two types of electrolytes: one
a single semiconductor, a semiconductor electrolyte forming an ohmic contact and
of much larger band gap is needed (Fig. 5). the other forming a Schottky-type contact.
With such material, biasing is needed, as It is postulated that like the formation of
is case with n-TiO2 (Fig. 4). Moreover, as a Schottky-type barrier, if the magnitude
a result of high band gap, n-TiO2 absorbs of the Fermi level of the semiconductor
only about 2% of the solar spectra (Fig. 2). and the redox electrolyte is same, then the
Therefore, a semiconductor of large band contact between them should be ohmic.
gap would correspondingly give very low This provided an opportunity to visualize a
photocurrent (hence, lowphotoconversion PEC cell in which the semiconductor acts
efficiency), making the photoelectrolysis like a separator between two types of re-
processes uneconomical. dox electrolytes. The electrolyte is selected
such that the front side of the semicon-
4.3.4.2 Multiple-type PEC Cells ductor (i.e. the side to be illuminated)
The calculations from the earlier section forms a Schottky junction. The backside
suggest that there is a need to introduce of the semiconductor is kept in contact
some improvements in the PEC cell to at- with another redox electrolyte, which gives
tain a sufficient solar energy conversion an ohmic contact. The front side of the
semiconductor acts as a normal PEC cell electrons migrate in opposite directions to

and its backside acts as a counter electrode. arrive at the metal electrode (lead). Because
This type of cell has been classified as a this metal gives an ohmic contact, photo-
Sharon-Schottky-type cell [20]. To simplify generated electrons are easily transferred
this type of cell, this concept was extended to the redox electrolyte to perform reduc-
to make a multiple-type PEC cell. It is tion of the electrolyte. Thus, while the front
assumed that a metal forms an ohmic side (i.e. the illuminated side) performs
contact with its own oxide. In this cell, oxidation of the redox electrolyte, the back-
three electrodes made of n-type semicon- side performs reduction of the electrolyte
ductor (n-Pb3 O4 ) deposited over a metal (and also acts like a counter electrode).
(lead) are arranged in an array (Fig. 8a). When more than one such electrodes are
The front side of each electrode contains used (Fig. 8a), a vectorial migration of
n-Pb3 O4 , which is dipped in a suitable holes takes place from left to right side
redox electrolyte to form a Schottky-type while vectorial migration of electrons takes
barrier. On illumination of n-Pb3 O4 , elec- place from right to left side. These carriers
trode photogenerated holes perform oxida- (electrons and holes) are transferred to the
tion of redox electrolyte. Photogenerated load by the two outer electrodes, that is,
Load
Semi- hν Counter hν hν
conductor electrode h+
electrode
e e e B
A
e− + OX− Red
carbon electrode
Z
h+ h+ h+
PEC cells
e
Red e + OX− Pb3O4
Pb3O4 h+ − Red OX−
Pb + X h+
Lead
Y
Fig. 8 Schematic representation of multiple between the metal and the semiconductor and
PEC cells connected by the help of redox the direction of flow of photogenerated electron
electrolyte. Inset X shows the energy level of and hole. Electron is extracted to the load via
semiconductor Pb3 O4 electrode deposited over metal A and the hole is extracted via inert metal
metal lead. Photoelectrochemical reactions B. Inset Z describes reduction process occurring
occurring at two ends of the semiconductor is at inert carbon electrode-B.
shown. Inset Y shows the ohmic contact
Fig. 9 A photograph of a multiple PEC

cells (7 in number) connected in series
by the help of redox electrolyte as
described in Fig. 7. The output of this
cell can be given to a normal electrolytic
cell for electrolysis of water (after
M. Sharon et al., Electrochim. Acta 1991,
36(7), 1107–1126).
by the backside of the last left side elec- radiation (assuming each semiconductor
trode (A) and metallic electrode on the to be of the same area). Employing this
extreme right side of the cell (B). The total cell as a source of electrical power, it also
photopotential of such system is equal to can be used for electrolyzing water (Fig. 9).
the photopotential of one PEC cell times
the number of semiconductor electrodes 4.3.4.3 Bipolar Cell
used [20]. This arrangement gives an op- The bipolar cell of Sharon-Schottky cell has
portunity to get the required electrolysis been further developed. Bard and cowork-
photopotential, even with the semiconduc- ers [21, 22] developed a bipolar semicon-
tor of band gap ≈1.4 eV. In addition, the ductor photoelectrode array (Fig. 10) and
requirement for matching the conduction studied its application to light-driven water
band edge (or flat band potential) with the splitting and electrical power generation.
electrode potential for hydrogen evolution They used five n-TiO2 electrodes in series.
and the valence band edge with elec- In the subsequent developments [23], they
trode potential of oxygen evolution (Figs. 5 devised a similar bipolar cell (Fig. 11) with
and 6) becomes redundant. Photocurrent, CoS/CdSe. In this cell, also a salt bridge
however, would depend on the area of the is used to complete the electrical circuit.
individual semiconductor exposed to solar CoS is used to make ohmic contact with
Bipolar
n -SC electrode
1/4 O2 O O O 1/2 H2
1/2 H2O R R R H+
(1) (2) (3) (4)

DC(H2)
Salt bridge DC Salt bridge
Fig. 10 A bipolar cell to show the vectorial transfer of electrons in one direction.
On the extreme left hand side oxygen is evolved and on the extreme right hand
side hydrogen gas is released. All other cells behave similar to a PEC cell. Salt
bridge is used to complete electrical contact. n-SC − n-TiO2 , dark electrocatalyst
reaction (O → R)j , light-initiated electrocatalyst (R → O)j , and dark
electrocatalyst for H2 evolution DC(H2 ) (after A. J. Bard et al., J. Phys. Chem. 1986,
90, 4606 and A. J. Bard et al., J.Electrochem. Soc. 1988, 135, 567).
Visible light flux
Oxygen Hydrogen
collector collector
CdSe Pyrex tube

1 2 3 5 6
4 I
A B C D E F G H J
Pt Pt
CoS
E
− CB
CdSe/Electrolyte −
interface *
n EF EF E°
hν H2 / H2O
CdSe Pt Pt / Electrolyte
E ° 2− s2
s / 2 *
pEF
interface
+ VB
+
Direction of electron flow
Fig. 11 (a) Schematic representation of water
photoelectrolysis cell. A, J: Pt; B, D, F, H: CoS; C, E, G, I: CdSe.
Solutions 1,6: KOH (1M); 2–5 Na2 S (1M), S (1M), KOH (1M).
For H2 and O2 generation solutions 1 and 6 are connected
with KOH bridge. (b) Expansion shows energetics of bipolar
panel (after A. J. Bard et al., J. Phys. Chem. 1987, 91, 6).
n-CdSe as well as with the electrolyte. Four for sensitization of semiconductor elec-
CoS/CdSe electrodes are used. Polysulfide trode. Hence, this will not be discussed
is used as an electrolyte in these PEC cells. here any further.
In this arrangement, the PEC cell is formed
among (B, C), (D, E), (F, G), and (H, I). 4.3.5
Sides C, E, G, and I form Schottky-type Recent Developments
barriers while sides B, D, F, and H act
like a counter electrode. The first and the Principal solar water-splitting models had
last electrodes are made of platinum. The predicted similar dual band gap photo-
energy level diagram to show the flow of electrolysis efficiencies of only 16% and
electron is shown in Fig. 11. 10–18% [3, 28], respectively, whereas re-
Several oxides (e.g. TiO2 [7], SrTiO3 [8], cently dual band gap systems were cal-
n-SiC [24], p-GaP [25], Fe2 O3 [26], etc.) and culated to be capable of attaining over
chalcogenides such as CdSe [23] have been 30% solar photoelectrolysis conversion ef-
used as anodes for water photodecompo- ficiency [14]. The physics of the earlier
sition. n-TiO2 , even today seems to be one models were superb, but their analy-
of the most important materials, because sis was influenced by dated technol-
it has been possible to extend its spec- ogy, and underestimated the experimental
tral response into visible portion of the ηphoto attained by contemporary devices
solar spectrum through sensitization with or underestimated the high experimen-
organic dyes. A separate chapter is devoted tal values of ηelectrolysis which can be
attained. For example, Ref. 3 estimates low developed in bipolar cells, amounts to
values of ηphoto (less than 20% conversion) connecting large number of PEC cells
because of the assumed cumulative rel- in series. The advantage of bipolar cells
ative secondary losses that include 10% is that it can operate with low band
reflection loss, 10% quantum-yield loss, gap semiconductor, yet provide desired
and 20% absorption loss. Experimentally, photopotential and high solar to electri-
a cell containing illuminated AlGaAs/Si cal efficiency. Although photopotential of
RuO2 /Ptblack was demonstrated to evolve such device is equal to the photopotential
H2 and O2 at record solar–driven water of one-photoelectrode times the number
electrolysis efficiency. Under illumination, of PEC cells connected in series, photocur-
bipolar configured Al0.15 Ga0.85 As (Eg = rent of system depends on the intensity
1.6 eV) and Si (Eg = 1.1 eV) semiconduc- of solar radiation falling on an individual
tors generate open circuit and maximum photoelectrode. But it is important to re-
power photopotentials of 1.30 and 1.57 V, alize that none of semiconductors so far
respectively, well suited to the water elec- developed can be used for prolonged pho-
trolysis thermodynamic potential: toelectrolysis of water, because of their
inherent instability towards photocorro-
H2 O −−−→ H2 + 12 O2 ;
sion. Success of making bipolar cells as a
o commercial viable system entirely depends
EH 2O
= EO2 − EH2 ;
o ◦ on the development of photoelectrochem-
EH 2O
(25 C) = 1.229 V (11)
ically stable low band gap semiconductor.
o
The EH /photopotential-matched semi-
2O
conductors are combined with effective Acknowledgment
water electrolysis O2 or H2 electrocat-
alysts, RuO2 , or Ptblack . The resultant S. Licht is grateful to Helmut Tributsch for
solar photoelectrolysis cell drives sustained his review and suggestions on theoretical
water-splitting at 18.3% conversion effi- sections of this chapter and thankful for
ciencies [14]. These recent developments the support by the BMBF Israel-German
in hydrogen generation at high solar en- Cooperation and the Berman-Shein Solar
ergy conversion efficiency are detailed in Fund. Maheshwar Sharon is thankful to
this volume in the chapter titled ‘‘Opti- his students and in special to G. Ranga
mizing Photoelectrochemical Solar Energy Rao who contributed toward development
Conversion: Multiple Band Gap and Solu- and modification of PEC cell.
tion Phase Phenomena.’’
References
4.3.6
Conclusion
1. M. A. Green, K. Emery, K. Bucher et al.,
Propgr. Photovolt. 1999, 11, 31.
In this section, efforts are made to discuss 2. S. Licht, B. Wang, T. Soga et al., Appl. Phys.
the thermodynamics of photoelectrolysis Lett. 1999, 74, 4055.
of water using a PEC cell. To facilitate 3. J. R. Bolton, S. J. Strickler, J. S. Connolly,
Nature 1985, 316, 495.
the generation of required potential for
4. K. W. Atanabe, M. Zelikoff, J. Opt. Soc. Am.
the photoelectrolysis of water, discussions 1953, 43, 753.
are made on some modified PEC cells 5. Nikola Getoff, Int. J. Hydrogen Energy 1990,
popularly known as bipolar cell. Potential 15(6), 407.
6. L. J. Heidt, A. F. McMillan, Science 1953, 17. A. Fujishima, K. Honda, Bull. Chem. Soc.
117, 75. Jpn. 1971, 44, 1148.
7. A. Fujishima, K. Honda, Nature 1972, 238, 18. Heinz Gerischer, Pure Appl. Chem. 1980, 52,
37. 2649.
8. J. M. Bolts, M. S. Wrighton, J. Phys. Chem. 19. Maheshwar Sharon, G. Ranga Rao, Indian J.
1976, 80, 2641. Chem. 1986, 25A, 170–172.
9. R. Memming, Top. Curr. Chem. 1988, 143, 20. M. Sharon et al., Electrochim. Acta 1991,
79. 36(7), 1107–1126.
10. A. Nozik, Appl. Phys. Lett. 1976, 29, 150. 21. A. J. Bard et al., J. Phys. Chem. 1986, 90,
11. J. White, F. -R. Fan, A. J. Bard, J. Electrochem. 4606.
Soc. 1985, 132, 544. 22. A. J. Bard et al., J. Electrochem. Soc. 1988, 135,
12. R. C. Kaintala, J. Ö. M. Bockris, J. Int. Hydro- 567.
gen Energy 1988, 13, 375. 23. Tooru Inoue, Toshihiro Yamase, Chem. Soc.
13. O. Khaselev, K. Turner, Science 1998, 280, Japan, Chem. Lett. 1985, 869.
425. 24. H. Honeyama, H. Sakamoto, H. Tamura,
14. S. Licht, B. Wang, S. Mukerji et al., J. Phys. Electrochim. Acta 1979, 277, 637.
Chem. B 2000, 104, 8920. 25. Lynn C. Schumacher, Suzanne Mamiche-
15. R. Memming in Photochemical Conversion Afara, Michael F. Weber et al., J. Electrochem.
and Storage of Solar Energy, (Eds.: E. Pelizzetti, Soc. 1985, 132(12), 2945.
M. Schiavello), Kluwer Academic Publishers, 26. M. F. Weber, M. J. Digman, Int. J. Hydrogen
Netherlands, 1991, pp. 193–212. Energy 1986, 11, 225.
16. F. Gutman, O. J. Murphy in Modern Aspects 27. Ibid, J. Electrochem. Soc. 1984, 131, 1258.
of Electrochemistry (Eds.: White, Bockris, 28. Adolf Goetzberger, Christopher Hebling,
Conway) 1983, P5. Sol. Energy Mater. Sol. cells 2000, 62, 1.
4.4 reduction and generate chemical, electri-

Optimizing Photoelectrochemical Solar cal, or electrochemical energy. Energetic
Energy Conversion: Multiple Bandgap and constraints imposed by single band gap
Solution Phase Phenomena semiconductors had limited values of pho-
Stuart Licht toelectrochemical solar to electrical energy
Technion – Israel Institute of Technology, conversion efficiency to date to 12–16% [9,
Haifa, Israel 10]. Multiple band gap devices can pro-
vide efficient matching of the solar spec-
4.4.1 tra [11–15]. A two or more band gap con-
Introduction figuration will lead, per unit surface area,
to more efficient solar energy conversion,
This chapter focuses on two concerted ef-
and in the solid-state, multiple band gap
forts to achieve stable, high solar energy
solar cells have achieved more than 30%
conversion efficiency using a variety of
conversion efficiency of solar energy [14].
photoelectrochemical systems. The first
The fundamental benefits of multiple band
effort explores the use of multiple band
gap photoelectrochemistry have been rec-
gap semiconductor systems. Limiting con-
straints of multiple band gap photoelec- ognized [16]. In this section an overview of
trochemical energy conversion as well as the energetics of distinct multiple band gap
practical configurations for efficient solar photoelectrochemical solar cell (MPEC)
to electrical energy conversion have been configurations is introduced. The MPEC
probed [1–6]. Such systems are capable of configurations can lead to higher conver-
better matching and utilization of incident sion efficiency than previously observed
solar radiation (insolation). Efficient solar for single band gap solar cells.
cells, solar storage cells, and solar hydro- A limited fraction of incident solar pho-
gen generation systems are discussed and tons have sufficient (greater than band gap)
demonstrated. energy to initiate charge excitation within a
The principles of photoelectrochemi- semiconductor. Because of the low fraction
cal phenomena systems can differ sub- of short wavelength solar light, wide band
stantially from conventional solid-state gap solar cells generate a high photovoltage
physics. Photoelectrochemical systems but have low photocurrent. Smaller band
can be characterized not only by semicon- gap cells can use a larger fraction of the
ductor but also often by electrolytic limi- incident photons but generate lower pho-
tations. The second half of the chapter fo- tovoltage. As shown in Fig. 1, schematic,
cuses on the substantial improvements to multiple band gap devices can overcome
photoelectrochemical energy conversion these limitations. In stacked multijunc-
attained by understanding and optimizing tion systems, the topmost cell absorbs
of such solution phase phenomena [7, 8]. (and converts) energetic photons but is
transparent to lower energy photons. Sub-
4.4.2
sequent layer(s) absorb the lower energy
Multiple Band Gap Photoelectrochemistry
photons. Conversion efficiencies can be
4.4.2.1 Theory of Multiple Band Gap Solar enhanced, and calculations predict that a
Cell Configurations 1.64-eV and 0.96-eV two-band gap system
Radiation incident on semiconductors has an ideal efficiency of 38% and 50% for 1
can drive electrochemical oxidation or and 1000 suns concentration, respectively.
4.4 Optimizing Photoelectrochemical Solar Energy Conversion: 359
Fig. 1 In stacked multijunction solar

cells, the top cell converts higher energy
photons and transmits the remainder
onto layers, each of smaller band gap
cell than the layer above, for more
effective utilization of the solar
spectrum.
The ideal efficiency increases to a limit of

72% for a 36-band gap solar cell [11].
Present solid-state (photovoltaic) mul- Eg1 > Eg2 > Eg3
tijunction photovoltaics exist in one of
two modes, either splitting (in which the
solar spectrum is optically separated be-
fore incidence on the cells) or stacked Cell 1 (Eg1)
devices. The latter mode has been devel-
oped as either monolithic (integrated) or
mechanically (discrete cells layered with Cell 2 (Eg2)
transparent adhesive) stacked cells. Most
monolithic stacked multijunction photo-
voltaics under development use III–V Cell 3 (Eg3)
semiconductors in at least one of the
layers and take advantage of the varia-
tions in band gap and close lattice match
Depending on the relative rates of charge
achieved with other related III–V al-
transfer, jo,Sch may be constrained by
loys. For example, GaInP has been used
either solid-state or electrochemical lim-
as a wide gap top cell, or GaSb and
itations, and is respectively termed the
GaInAsP as small gap lower cells, for
saturation current or the equilibrium ex-
III–V top-layer multijunction cells; other
change current [17]. A single representa-
cells use silicon or CIS cells as a lower
layer [12–15]. tion of either a pn or Schottky junction
The photopotential of a single pn is schematized in the upper center of
junction is: Fig. 2, by a junction generating a volt-
! " # $ age V .
nkT 1 + jph,pn A variety of distinct MPEC configura-
Vpn = ln (1)
e jo,pn tions are possible, each with advantages
and disadvantages [1]. The simplest MPEC
where Vpn is constrained by the photocur-
configurations contain two adjacent band
rent density through the junction, jph,pn .
gaps. Adjacent band gap layers can be
jo,pn is the saturation current (of a re-
aligned in the cell in either a bipolar or
versely polarized diode) and is described
a less conventional inverted manner. In
by the Shockley equation [17].
either the bipolar or inverted cell configura-
The photopotential of a single liquid
tion, the PEC solid–electrode interface can
Schottky junction is given by:
consist of either an ohmic or a Schottky in-
! " # $
nkT 1 + jph,Sch terface. The ohmic interface can consist of
VSch = ln (2)
e jo,Sch either direct (semiconductor–electrolyte)
360
Configurations of Multiple Band Gap Photoelectrochemical Solar Cells
Single band gap p-Schottky
or pn junction
∆V
representation Common node
Bipolar gap Inverted gap
∆V small solar cell Ohmic
solar cell
contacts
∆V wide ∆V wide ∆V small
Semiconductor(sc) /electroyte junctions
Bipolar pnpn 2
Inverted pnnp 2
band gap structure: Schottky Ohmic
band gap structure :
V = ∆V wide + ∆V small sc Electrolyte Electrolyte
Eredox V = ∆V wide, ∆V small
Bipolar gap EFermi
Eredox Inverted gap
ohmic contact ∆V ohmic contact
photoelectro- ∆V photoelectro-
chemical chemical
solar cell pn Ohmic solar cell
(PEC) junction junction (PEC)
4 Solar Energy Conversion without Dye Sensitization
Bipolar gap Bipolar gap Bipolar gap Inverted gap Inverted gap Inverted gap
indirect direct Schottky contact indirect direct Schottky contact
(semiconductor- (semiconductor- photoelectro- (semiconductor- (semiconductor- photoelectro-
metal-solution) solution) chemical metal-solution) solution) chemical
ohmic contact ohmic contact solar cell ohmic contact ohmic contact solar cell
(BGIO PEC) (BGDO PEC) (BGS PEC) (IGIO PEC) (IGDO PEC) (IGS PEC)
Regen- Storage Regen- Storage Regen- Storage Regen- Storage Regen- Storage Regen- Storage
erative BGIO erative BGDO erative BGS erative IGIO erative IGDO erative IGS
BGIO PEC BGDO PEC BGS PEC IGIO PEC IGDO PEC IGS PEC
PEC PEC PEC PEC PEC PEC
Fig. 2 Relation of the twelve representative MPEC configurations comprising regenerative ohmic cells: (1) bipolar gap direct ohmic regenerative,
(2) bipolar gap indirect ohmic regenerative, (3) inverted gap direct ohmic regenerative, (4) inverted gap indirect ohmic regenerative; storage ohmic cells:
(5) bipolar gap direct ohmic storage, (6) bipolar gap indirect ohmic storage, (7) inverted gap direct ohmic storage, (8) inverted gap indirect ohmic
storage; regenerative Schottky cells: (9) bipolar gap Schottky regenerative, (10) inverted gap Schottky regenerative; and storage Schottky cells:
(11) bipolar gap Schottky storage, and the (12) inverted gap Schottky storage configuration.
or indirect (semiconductor–metal and/or Wide band gap layer charge separation

electrocatalyst–electrolyte) interfaces. Fi- occurs across a pn junction space charge
nally, each of these modes can be con- field gradient, while charge separation in
figured as either regenerative or storage the small band gap is maintained with
solar cells. As shown in Fig. 2, these a field formed by the Schottky semicon-
modes provide twelve alternate configu- ductor–electrolyte interface. In the bipolar
rations available for MPECs. Schottky MPEC configuration, generated
Compared with the inverted case, the charge flows through all layers of the cell,
bipolar arrangement provides a concep- providing the additional constraint:
tually simpler PEC but generates a large
open-circuit photopotential, Voc . As shown jph,pn = jph,Sch (5)
for a regenerative MPEC in either scheme
on the left side of Fig. 2, the generated In an alternate bipolar regenerative con-
bipolar photovoltage, Vphoto , is the sum of figuration, the bipolar (or multiple) band
the potentials of the individual band gap gap configurations may contain consecu-
layers minus cathodic and anodic polariza- tive space charge field gradients generated
tion losses in driving a regenerative redox only via solid-state phenomena. This is
couple: presented in Fig. 3(b) for the case of two
consecutive bipolar pn junctions. The low-
Vphoto = Vw + Vs − (ηcathode + ηanode ) est semiconductor layer (the small band
(3) gap n-type layer in the figure) may remain
The energy diagram of a bipolar in direct contact with the electrolyte, but
band gap photocathodic electrochemical the contact is ohmic and is not the source
Schottky configuration is presented in of the small band gap space charge field.
Fig. 3(a). The scheme comprises a two- A similar indirect ohmic contact bipolar
photon/one-electron photoelectrochemi- regenerative configuration may also be de-
cal process (2hν → e− ), which may be rived from this figure. In this case, the
generalized for an n band gap configura- lowest semiconductor layer is restricted to
tion, to an n photon process (nhν → e− ). electronic and not ionic contact with the
Light shown incident from the left side of electrolyte through use of an intermediate
the configuration first enters the wide band (bridging) ohmic electrocatalytic surface
gap layer(s) in which more energetic pho- layer. This can facilitate charge transfer
tons are absorbed; less energetic photons to the solution phase redox couple, and
are transmitted through this upper layer prevent any chemical attack of the semi-
and are absorbed by the small band gap conductor. In the bipolar cases (including
layer. The resultant combined potential of Schottky, direct or indirect ohmic config-
the photodriven charge sustains reduction
urations), the photopower generated by a
at the photocathode interface, and drives
bipolar regenerative MPEC is given by the
extractable work through the external load,
product [17]:
Rload . The wide ‘‘w’’ and small ‘‘s’’ band
gap layers are denoted with respective va- Pbipolar regenerative = jph [Vw + Vs
lence and conduction bands, EV and EC ,
and band gap: − (ηcathode + ηanode )] (6)
EGw = ECw − EVw ; EGs = ECs − EVs Adjacent band gap layers in a multiple
(4) band gap configuration can also be aligned
V = Vw + Vs − η cathode + η anode
e−
R load
ECs e−
e−
A+ → A → A+
A
ηcathode
Semiconductor
ηanode
EFermi (ns) EGs
e− Eredox
hν Vs
ECw
EFermi (nw = ps) EVs
Vw h+
EGw > hν > EGs⇒
h ν EGw
EFermi (pw)
+
h ν > EGw ⇒ h EVw
Ohmic p-Schott.
p n Electrolyte
junction small gap
(a) Wide gap
V = Vw + Vs − η cathode + η anode
e−
R load
e−
A+ → A A → A+
e− E
electrocatalyst cathode
Cs
Semiconductor or
ηcathode ηanode
e− EFermi(ns)
Eredox
Ecw Vs
hν
EGs
EFermi (nw=ps)
Vw EGw > h ν > EGs ⇒ h+ EVs

hν
EGw
Ohmic
Electrolyte
EFermi (pw) junction
h ν > EGw ⇒ h+ EVw

Ohmic
p n p n
junction
(b) Wide gap Small gap
Fig. 3 Energy diagrams for multiple band gap photoelectrochemistry. Elements of bipolar or
inverted band gap, Schottky, ohmic, regenerative, and storage configurations are illustrated.
(a): Bipolar band gap Schottky regenerative MPEC. (b): Bipolar band gap (direct or indirect)
ohmic regenerative MPEC.(c): Inverted band gap indirect ohmic regenerative MPEC.(d): Inverted
gap (direct or indirect) ohmic storage MPEC.
Vw − ηc+ ηaw
Rload-w
Rload-s
Semiconductor/metal
ECs Vs− η c+ ηas
e− e− e− e−
ECw ηas ηc ηc ηaw
EFermi (nw=ns) E
Vs hν Gs Eredox Eredox
Vw A → A+
h ν EGw EVs EFermi (ps) A+ ← A
EGw > h ν >EGs ⇒ h+

EFermi (pw) Ohmic A+ → A
junction Electrolyte
+ EVw
h ν > EGw ⇒ h
Ohmic
p n n p
Wide gap junction Small gap
(c)
e− e−
Es = V s Ew = Vw
ECs
e− e− + +
electrocatalyst anode
electrocatalyst anode
B →B C →C
ECw
Semiconductor or
Semiconductor or
ηcB ηcC
EFermi (nw=ns) hν EGs EredoxB EredoxC
Vs ηaA
Vw
hνE EredoxA D → D+
Gw EVs EFermi (ps) ηaD
EGw > h ν > EGs ⇒ h+ A→ A+
EFermi (pw) EredoxD
Ohmic
hν > EGw ⇒ h+ EVw junction Electrolyte Electrolyte
p n n p
(d) Wide gap Ohmic Small gap

junction
Fig. 3 (Continued)
in a less conventional inverted manner. As regenerative ohmic configuration, com-

shown in Fig. 3(c), a dual gap inverted pared to the pn/pn in Fig. 3(b). As seen for
cell generates two smaller photopoten- the inverted pn/np case, a quasi fermi level
tials, which can be separately applied to is shared by the conduction bands of the
the same PEC, minimizing electrolysis wide band gap n-type and the small band
losses. This figure presents an inverted gap n-type layers. Light incident from the
pn/np (wide band gap or small band gap) figure left side generates charge through
a common intermediate node, comprising Secondary redox storage must be ac-

a two photon/two electron photoelectro- complished at sufficiently low potentials
chemical process (2hν → 2e− ): to prevent losses due to simultaneous
(undesired) solvent electrolysis. However,
2h+ e− + hν(hν > EGw ) + hν(hν > EGs )
bipolar band gap photoelectrochemistry
−−−→ 2h+ + 2ew −∗ + 2es −∗ (7) imposes large photopotentials. These are
avoided through the inverted band gap
Photoinduced holes, generated in accord
configuration, as exemplified in Fig. 3(d),
with Eq. (7), drive solution phase oxida-
in which the photopotentials generated in
tion at separate electrocatalytic anodes. A
the respective small and wide band gap
challenge to bipolar MPEC use is that all
portions of the tandem cell, Vw and Vs
generated charge must flow through sub-
drive two separate electrochemical storage
sequent cell layers, as seen in the diagrams
processes:
in the left side of Fig. 3. This imposes a
current matching constraint on each of D + C+ −−−→ D+ + C;
the individual junctions of any monolithic
bipolar device. This constraint is removed EC+ /C − ED+ /D < Vw and
with inverted band gaps, in which pho-
A + B+ −−−→ A+ + B;
tocurrent is independent for each band
gap, flowing through the common node EB+ /B − EA+ /A < Vs (9)
illustrated in the figure. Unlike the bipo-
lar cases, for the inverted regenerative 4.4.2.2 Bipolar Band Gap PECs
(Schottky, direct or indirect ohmic) con- A bipolar gap direct ohmic photoelectro-
figurations, the photopower generated is a chemical system comprises either a bipolar
separate sum of the different layers, where band gap pnpn/electrolyte ohmic photo-
for the two band gap case [17]: electrochemical cell, with reduction oc-
Pinverted regenerative = jph−w Vw + jph−s curring at the photoelectrode–electrolyte
interface and regenerative oxidation oc-
× Vs − (ηcathode + ηanode ) (8) curring at the electrolyte–counter elec-
Electrochemical energy storage config- trode (anode) interface or alternately: a
urations provide an energy reservoir that npnp/electrolyte bipolar band gap with
may compensate for the intermittent na- oxidation occurring at the semiconduc-
ture of terrestrial insolation. When differ- tor–electrolyte interface and regenera-
ent anodic and cathodic redox processes tive reduction occurring at the elec-
are driven, electrochemical energy stor- trolyte–counter electrode interface. In the
age can be accomplished. Storage can bipolar gap direct ohmic photoelectro-
consist of both in situ or photoelectroly- chemical system, direct refers to the direct
sis cells. Until recently, photoelectrolysis contact between semiconductor and solu-
cells, in which the electrolyte solvent is tion, and ohmic indicates this interface is
oxidized and/or reduced to provide a chem- an ohmic rather than a Schottky junction.
ical fuel, had been inefficient [6]. In situ This facilitates study of several characteris-
secondary redox couples, added to the tics of bipolar multiple band gap systems,
electrolyte, have been simpler to optimize, without the added complication of si-
and efficient single band gap in situ PEC multaneous parameterization of a direct
storage cells have been demonstrated [18]. Schottky barrier at the electrolyte interface.
The examples presented here utilize interference will occur for the top AlGaAs
combine multijunction solid-state layers layers, and bottom Si layers through this
consisting of a bipolar AlGaAs (EGw = or substantially shorter electrolyte path-
1.6 eV) wide band gap, overlaid on a lengths. The solid-state component in-
Si (EGs = 1.0 eV) small band gap, and cludes a graded band emitter, varying from
used in an electrolytic cell [2]. Light ab- Al(0.3 – 0.15) Ga(0.7 – 0.85) As with overlayers
sorption by the electrolyte can interfere of p + -Alx Gax−1 As on n-Alx Gax−1 As. The
with the cell and should be avoided. growth sequence and graded band emitter
Figure 4 overlays the optical characteris- layer improve collection efficiency [2]. The
tics of the solid and solution phase of Si bottom cell consists of a p + -Si, n-Si,
the AlGaAs/Si solid-state and V3+/2+ elec- and n+ -Si multijunction. The band edges
trolyte optimized components within a observed in the figure at approximately
bipolar gap photoelectrochemical cell. So- 800 nm and 1100 nm are consistent with
lution transmission is measured through a the respective AlGaAs and Si band gaps.
pathlength that is typical (1 mm) of many For efficient electron/hole pair charge
experimental front wall photoelectrochem- generation, incident photons need to be
ical cells. As is evident, light-transmission localized within the multiple band gap
1.0 0
0.8 20
Q(AlGaAs layers) Q(Si layers)

Quantum efficiency
0.6 40
Transmittance
[%]
0.4 60
0.2 T(0.35 M V 2+/3 +) 80

1 mm pathlength
0.0 100
400 500 600 700 800 900 1000 1100 1200
Wavelength
[nm]
Fig. 4 Overlay of the optical characteristics of measured through a pathlength of 1 mm. As
the solid and solution phase of the AlGaAs/Si described in the text, the Si bottom cell consists
solid-state and V3+/2+ electrolyte constituents of a p+ -Si, n-Si and n+ -Si multijunction. The
within a bipolar gap photoelectrochemical cell. Al(0.3 – 0.15) Ga(0.7 – 0.85) As top cell utilizes a
Transmission of the V3+/2+ electrolyte is graded band emitter.
semiconductor small and wide band gap GaAs layer. Internally, a bridging GaAs
regions, rather than lost through compet- buffer layer provides an ohmic contact be-
itive electrolyte light absorption. As seen tween the wide band gap AlGaAs junctions
in Fig. 4, the vanadium electrolyte can sig- and the lower Si layers. An intermediate
nificantly block light, over a wide range contact layer indicated as ‘‘Au’’ is used
of visible and near infrared wavelengths, only for probing separated characteristics
from entering the wide and small band gap of the wide and small band gap junc-
layers of the multiple band gap photoelec- tions, and is not utilized in the complete
trochemical cell. This deleterious effect is cell. Photo generated charge at the indi-
prevented by use of the back wall mul- cated silicon electrolyte interface induces
tiple band gap photoelectrochemical cell solution phase vanadium reduction, and
presented in Fig. 5. Light does not pass a carbon counter electrode provides an
through the solution. As shown, illumina- effective (low polarization) electrocatalytic
tion enters directly through antireflection surface for the reverse process in a regen-
films of 50 nm ZnS situated on 70 nm erative cell, in accord with:
MgF2 . An evaporated Au-Zn/Au grid pro-
V3+ (+hν) −−−→ V2+ + h+ ;
vides electrical contact to the wide gap
AlGaAs layers through a bridging p + = V2+ −−−→ V3+ + e− (10)
e−
Au
Au-Zn/Au
p+ - GaAs
300 nm
1.0 µm
1.0 µm
1.7 nm
350 µm
800 nm
50 nm
20 nm
10 nm
AR coating [ZnS(50 nm)/MgF2(70 nm)]
V 3+ + e− → V 2+
V 2+ → V 3+ + e−
(1 × 1018 cm−3)
p+ - Al(0.3 − 0.15)Ga(0.7− 0.85)As (1 × 1018 cm−3)
(1 × 1018 cm−3)
(1 × 1019 cm−3)
(8 × 1015 cm−3)
(4 × 1019 cm−3)
(2 × 1017 cm−3)
Illumination ⇒
Carbon
GaAs (Buffer layer)
n - Al(0.15)Ga(0.85)As
n+Al(0.15)Ga(0.85)As
p+ - Al0.8Ga0.2As
E°,V 2+/3+
n - GaAs
Electrolyte
p+ - Si
n - Si
− 0.3 V vs H2
n - Si
+
Fig. 5Schematic description of the components in the bipolar gap direct

ohmic AlGaAs/Si-V3+/2+ PECs.
Heller, Miller and coworkers have previ- concentration of HF to the electrolyte can
ously shown that the p-Si surface is capable remove this passivating layer and permit
of sustaining minority carrier injection sustained photocurrent. This is illustrated
into the solution and stabilizes reduction in the inset of Fig. 6, in which the addi-
of the V3+/2+ redox couple [19]. However, tion of 0.02 and 0.2 M HF improves the
at n-Si this does not appear to be the case photocurrent stability of this bipolar direct
for reduction of the V3+/2+ redox couple by ohmic MPEC. Also as shown, no further
majority carrier injection into the solution. improvement in the photocurrent stability
Thermodynamically silicon can be sponta- is observed when the HF concentration is
neously oxidized in aqueous solution. This further increased from 0.2 to 0.5 M HF. As
process may occur despite the photoin- presented in the main portion of Fig. 6, the
duced generation of reductive charge at 0.2 M HF modified electrolyte stabilizes
the semiconductor–electrolyte interface. the photocurrent at the n-Si interface for
This is observed as the time-dependent the measured period of several hours [2].
decrease of photocurrent under constant As will be subsequently shown, photocur-
illumination, during photoinduced reduc- rent stability is further improved through
tion of V2+ (by majority carrier injection at use of a electrode catalyst bridging the
n-Si) indicated in the inset Fig. 6. This silicon–electrolyte interface.
is consistent with the onset of surface As can be seen in Eq. (6), maximiza-
passivation forming a layer that is pas- tion of the photopower necessitates
sive to charge transfer and diminishes minimization of the anodic and ca-
the photocurrent. The addition of a low thodic polarization losses, ηanode and
14
Bipolar gap ohmic photoelectrochemical cell Electrolyte :

AlGaAs/Si - V2+/3+ Electrolyte solar cell 0.35 M V(II)+V(III), 4 M HCl, 0.2 M HF
12 Illuminated area : 0.22 cm2
Photocurrent density
Illumination : simulated AM 1.5

Cathode : 0.2 cm2 n+ - Si 14
Anode : 0.2 cm2C
[mA cm−2]
12
10 No HF
Photocurrent
[mA cm ]
0.02 MHF
−2
10 0.2 MHF
0.5 MHF
8
Photocurrent stability, n-Si photoanode
in direct contact with solution
8
0 10 20 30 40 50 60
Time
[min]
0 1 2 3 4 5
Time
[h]
Fig. 6 Photocurrent stability in several V3+/2+ aqueous electrolytes of the bipolar band gap
direct ohmic AlGaAs/Si-V3+/2+ photoelectrochemical cell (measured indoors using a tungsten
halogen lamp to simulate outdoor AM 1.5 insolation).
ηcathode , during charge transfer through illumination, the AlGaAs/Si/V2+/3+ elec-

the photoelectrode and counter electrode trolyte photoelectrochemical solar cell ex-
interfaces. In the current domain investi- hibits an open-circuit potential, Voc =
gated, polarization losses are highly linear 1.4V, a short-circuit photocurrent, Jsc =
for both anodic and cathodic processes at 12.7 mA cm−2 , a fill factor, FF = 0.81,
2.5 to 3.5 mV cm2 mA−1 , and can create determined from the fraction of the max-
small but significant losses on the order of imum power, Pmax , compared to the
10–100 millivolts in the MPEC. product of the open-circuit potential and
Figure 7 presents the outdoor char- short-circuit current.
acteristics of the bipolar direct ohmic The multiple band gap solar to electrical
AlGaAs/Si/V2+/3+ photoelectrochemical conversion efficiency of 19.2% compares
cell under solar illumination. The system favorably to the maximum 15 to 16% solar
comprises the individual components il- to electrical energy conversion efficiency
lustrated in Fig. 5 and uses a HF previously reported for single band gap
containing aqueous vanadium electrolyte PECs [9, 10]. Small photoelectrochemi-
to improve photocurrent stability (0.35 M cal efficiency losses can be attributed to
V(II)+V(III), 4 M HCl, 0.2 M HF). The polarization losses accumulating at the so-
photoelectrochemical characteristics of lution interfaces. Under illumination, a
the cell were determined under 75- photocurrent density of 13 mA cm−2 seen
mW cm−2 insolation. As shown under in Fig. 7 is consistent with polarization
15 15
Bipolar gap direct ohmic photoelectrochemistry
AlGaAs/Si - V2+/3+ electroyte solar cell
FF = 0.81
12 Electrolyte: Aq.0.35 M V(II)+V(III), 4 M HCl, 0.2M HF 12
Electrolyte cathode: 0.2 cm2 n+ - Si
Electrolyte anode: 0.2 cm2 C

Insolation: 75 mWcm−2
Photopower density
9 Illuminated area: 0.22 cm2 9

[mA cm−2]
Solar electrical conversion

[mW cm−2]
Efficiency: 19.1%
15 15
6 6
Photopowder density
Current Indirect ohmic AlGaAs/Si V2+/3+ cell

Power Insolation: 80 mW cm−2
Illuminated area : 0.22 cm2
Anode + Cathode : 0.2 cm2 C
3 3
FF = 0.79
Solar conversion efficiency : 19.2%

Electrolyte : Aq.0.35 M V(II) + V(III), 4 M HCl
0 0
Photovoltage
0 0
0.0 1.2
[V]
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
Photovoltage
[V]
Fig. 7 Measured outdoor photocurrent/voltage characteristics of the
bipolar gap direct ohmic AlGaAs/Si-V3+/2+ photoelectrochemical solar cell.
Inset: Measured outdoor photocurrent/voltage characteristics of the bipolar
gap indirect ohmic AlGaAs/Si-V3+/2+ photoelectrochemical solar cell.
losses of approximately 0.04 V, accumu- the electrolyte–counter electrode inter-

lating both at the anode and the cathode. face. In these systems, indirect refers
At the observed maximum power point to the catalyst interface that bridges the
photovoltage, in excess of 1.2 V in Fig. 7, semiconductor and solution, and ohmic
this combined 0.08 V loss is consistent indicates this interface is an ohmic
with an effective loss in efficiency of 0.3 to rather than a Schottky junction. The pho-
0.4% compared to an analogous solid-state tocathodic bipolar direct ohmic MPEC
photovoltaic cell configurations. photoelectrochemistry comprises a two-
A common disadvantage of photoelec- photon/one-electron photoelectrochemi-
trochemical systems is photoinduced cor- cal process (2hν → e− ) injecting charge
rosion of the semiconductor that originates into an electrocatalyst and then into solu-
at the semiconductor–solution interface. tion. Energy conversion can occur either
The corroded surface inhibits charge trans- through Fig. 8’s photocathodic driven re-
fer that diminishes photocurrent. A stable dox process:
solid–solution interface, which both facil-
itates charge transfer and impedes semi- hν pn(wide gap)|pn (small gap)|
conductor photocorrosion, is provided by electrocatalyst|redox couple|
an electrocatalyst placed between the semi- electrocatalyst anode (11)
conductor and the electrolyte. The multiple
band gap photoelectrochemical cell can or a photoanodic driven redox couple:
utilize this electrocatalyst interface, as well
as a bipolar series arrangement of wide hν np(wide gap)|np (small gap)|
and small band gap semiconductors to en- electrocatalyst|redox couple|
hance energy conversion. The photoelec- electrocatalyst cathode (12)
trochemical characterization presented in
the inset of Fig. 7 summarizes a modified Two highly stable aqueous phase redox
GaAs/Si-V3+/2+ MPEC. In this indirect couples, iodide and polysulfide, are used
ohmic photoelectrochemistry, electrolyte in these studies to further this bipolar
induced photocorrosion of the silicon is indirect multiple band gap photoelectro-
entirely inhibited by utilization of an elec- chemistry. Platinum and cobalt sulfide,
trocatalyst (a second carbon electrode) respectively, provide an effective (low over-
bridging charge transfer between the semi- potential) electrocatalyst for a wide range
conductor and the electrolyte. of iodide and polysulfide electrolytes [2].
Bipolar gap indirect ohmic photoelectro- These electrolytes have also been utilized
chemistry comprises either: a bipolar band in bipolar band gap solar cells. Table 1
gap pnpn/electrolyte ohmic photoelectro- summarizes the results of a variety of
chemical cell, with reduction occurring at bipolar two band gap solar cells [2, 3].
the semiconductor–electrocatalyst–elec- In this table, comparison of the solid-
trolyte interface and regenerative oxida- state and direct photoelectrochemical cells
tion occurring at the electrolyte–counter shows that energy conversion efficiency
electrode (anode) interface or alter- of the photoelectrochemical cell approach
nately: an npnp/electrolyte cell, with that of the solid-state device. In the bipolar
oxidation occurring at the semiconduc- direct MPEC cell, the majority of the pho-
tor–electrocatalyst–electrolyte interface topower loss in these photoelectrochem-
and regenerative reduction occurring at ical cells was attributed to polarization
20 Inverted gap direct ohmic photoelectrochemistry 12

GaAs/Si-lodide electrolyte solar cell
Electrolyte: Aqueous 10.4 M Hl, 0.01M I2, 0.8 M HF
Electrolyte electrodes:Top cell : 0.2 cm2 Pt FF = 0.73
Bottom cell anode: 0.2 cm2 p+- Si
15 9
Bottom cell cathode: 0.2 cm2 Pt
Photopower density
Insolation: 80 mW cm−2 Top cell
Illuminated area: 0.22 cm2 characteristics
[mA cm−2]
[mW cm−2]
Solar/Electrical conversion
10 Efficiency: 19.0% 6
8 3.2
Photopower density
Current FF = 0.74
5 Power 4 1.6
3
Bottom cell
characteristics
0 0 Photovoltage 0.0 0
0.0 0.6
0.0 0.2 0.4 0.6 0.8
Photovoltage
[V]
Fig. 8 Measured outdoor characteristics of the portion of the cell. Main figure: top layer cell
inverted band gap direct ohmic GaAs/Si/I3 − , photocurrent/voltage characteristics. Inset:
3/2I− /Pt MPEC. The top cell consists of the bottom layer cell photocurrent/voltage
GaAs/Pti/I3 − , 3/2I− /Pt portion of the cell. The characteristics.
lower cell consists of the Si/I3 − , 3/2I− /Pt
Tab. 1 Comparison of bipolar solid-state or bipolar regenerative ohmic PECs under solar
illumination
Cell Voc Isc Pmax FF Insolation Conversion

configuration [V] [mA cm−2 ] [mW cm−2 ] [mW cm−2 ] [Efficiency %]
Bipolar solid state

AlGaAs/Si 1.409 15.6 17.7 0.81 90.2 19.6
Bipolar direct V2+/3+
AlGaAs/Si/V2+ /V3+ /C 1.393 12.7 14.4 0.81 75.0 19.1
Bipolar indirect V2+/3+
AlGaAs/Si/C/V2+ /V3+ /C 1.383 14.0 15.3 0.79 80.1 19.2
Bipolar indirect sulfide
AlGaAs/Si/CoS/S2 2− /S4 2− /CoS 1.456 15.8 18.5 0.80 94.1 19.7
Bipolar indirect iodide
AlGaAs/Si/Pt/I3 − /I− /Pt 1.409 15.9 18.1 0.81 94.1 19.2
Note: Cells utilize multijunction wide band gap AlGaAs layers over smaller band gap Si layers and one
of the indicated pairs of electrolyte/electrocatalyst electrodes indicated as Direct V2+/3+ : 0.35 M
V(II)+V(III), 4 M HCl, 0.2 M HF at carbon; Indirect V2+/3+ : 0.35 M V(II)+V(III), 4 M HCl at carbon;
Indirect Sulfide 1 M K2 S2 , 1 M KOH at CoS; Indirect Iodide: 10.4 M HI, 0.01, 4 M I2 at Pt.
PEC, photoelectrochemical solar cells.
losses arising at the solid–solution inter- I3 − /3/2I− portion of the cell under solar
faces. In certain indirect MPEC cases, as illumination.
seen in Table 1, the photoelectrochemical Photoelectrochemical characteristics we-
energy conversion efficiency is comparable re determined under 80 mW cm−2 inso-
to the solid-state cell process. In this case, lation and the generated photocurrent is
combined electrolytic polarization losses highly stable. As shown under illumina-
are less than or equal to the potential drop tion, the Si bottom cell exhibits an open-
from resistance losses in the back contact circuit potential, Voc = 0.51V, a short-
of the solid-state device. circuit photocurrent, Jsc = 7.4 mA cm−2
a fill factor, FF = 0.73, and a maximum
power, Pmax = 2.8 mW cm−2 . Simultane-
4.4.2.3 Inverted Band Gap PECs
ous to the photopower generated by the
Inverted photoelectrochemistry permits
bottom portion of the MPEC, the top
efficient energy, and maintained smaller
(GaAs) portion also generates photopower.
photopotentials than in comparable bipo-
The main portion of Fig. 8 presents the
lar MPECs. To probe inverted semicon-
outdoor characteristics of the top cell
ductor direct ohmic regenerative electro-
(GaAs/Pt/I3 − , 3/2I− /Pt) portion during
chemistry, an electrolyte such as iodide
solar illumination of the complete in-
was driven by a GaAs (EGw = 1.4 eV) wide
verted direct ohmic GaAs/Si/I3 − , 3/2I− /Pt
band gap, and an Si (EGs = 1.0 eV) small
MPEC. In this portion of the cell, GaAs
band gap were utilized [3]. In the inverted
is isolated from the iodide electrolyte via
configuration, bottom and top cells are a Pt electrocatalytic anode. The Pt elec-
utilized simultaneously. Simultaneous to trode is stable in this electrolyte, and
the photopower generated by the bottom hence the top cell portion of the pho-
portion (Si) of the MPEC, the top (GaAs) tocurrent appears to be fully stable. Un-
portion also generates photopower. A plat- der the same 80 mW cm−2 insolation,
inum counter anode, provides an effective photopower is generated simultaneous
(stable, low polarization) electrocatalytic to the photopower presented in Fig. 8.
surface for the reverse process in both As shown under illumination, the GaAs
the top (wide band gap) and bottom (small cell exhibits an open-circuit potential,
band gap) driven regenerative cells. At the Voc = 0.81 V, a short-circuit photocurrent,
n-Si interface, photocurrent stability can Jsc = 21.0 mA cm−2 , a fill factor, and a
be improved with the addition of HF to maximum power, Pmax = 12.4 mW cm−2
an acidic iodide electrolyte, and the elec- at FF = 0.74. The total power generated is
trolyte used was 10.4 M HI, 0.01 M I2 , the sum of the simultaneous extractable
0.8 M HF. Previous studies have shown power generated by each component.
that other electrolytes and surface modi- Table 2 compares inverted solid-state
fications will also enhance photocurrent photovoltaics with direct and indirect re-
stability through the (single band gap) generative ohmic PECs, each containing
silicon–electrolyte interface [19–21]. The a GaAs wide band gap and an inverted
inset of Fig. 8 presents the inverted di- aligned Si small band gap, and a variety of
rect ohmic outdoor characteristics of the redox couples. As in the case of the bipolar
(Si/I3 − /3/2I− ) bottom cell portion, and systems summarized in Table 1, the solid-
the main portion of the figure presents state device exhibits a marginally enhanced
the characteristics of the GaAs driven energy conversion efficiency compared to
Tab. 2 Comparison of inverted solid-state or inverted regenerative ohmic PECs under solar
illumination
Cell Voc Isc Pmax FF Insolation Conversion

configuration [V] [mA cm−2 ] [mW cm−2 ] [mW cm−2 ] [Efficiency %]
Inverted solid-state
bottom: Si 0.552 22.8 3.6 0.74
top: GaAs 0.855 8.7 14.2 0.74
GaAs/Si 17.7 90.2 19.6
Inverted direct iodide
Bottom: Si/I3 − /I− /Pt 0.510 7.4 2.8 0.74
Top: GaAs/Pt/I3 − /I− /Pt 0.813 21.0 12.4 0.73
GaAs/Si/I3 − /I− /Pt 15.2 80.1 19.0
Inverted indirect iodide
Bottom: Si/Pt/I3 − /I− /Pt 0.541 9.2 3.7 0.74
Top: GaAs/Pt/I3 − /I− /Pt 0.866 24.5 14.9 0.70
GaAs/Si/Pt/I3 − /I− /Pt 18.5 94.1 19.7
Inverted indirect sulfide
Bottom: Si/CoS/S2 2− /S4 2− /CoS 0.556 8.9 3.8 0.76
Top: GaAs/CoS/S2 2− /S4 2− /CoS 0.855 24.0 14.9 0.73
GaAs/Si/CoS/S2 2− /S4 2− /CoS 18.7 94.1 19.8
Inverted indirect V2+/3+
Bottom: Si/C/V2+ /V3+ /C 0.523 7.7 3.0 0.73
Top: GaAs/C/V2+ /V3+ /C 0.823 20.8 12.3 0.72
GaAs/Si/C/V2+ /V3+ /C 15.3 80.1 19.2
Note: Cells utilize multijunction wide band gap GaAs layers over smaller band gap Si layers and one
of the indicated pairs of electrolyte or electrocatalyst electrodes, indicated as Direct Iodide: 10.4 M
HI, 0.01, 4 M I2 , 0.8 M HF at Pt; Indirect Iodide: 10.4 M HI, 0.01, 4 M I2 at Pt; Indirect Sulfide 1 M
K2 S2 , 1 M KOH at CoS; Indirect V2+/3+ : 0.35 M V(II)+V(III), 4 M HCl at carbon.
the analogous direct regenerative MPEC. the V2+/3+ indirect regenerative MPEC in
This difference may be attributed to dimin- this table is consistent with the higher
ished charge transfer through the semi- polarization losses of 3 mV cm2 mA−1
conductor–electrolyte interface. However that we have measured for V2+/3+ at
these differences are small and are only carbon, compared to those of less than
slightly larger than those caused by the 2 mV cm2 mA−1 previously measured at
uncertainty of ±1 mW cm−2 in the mea- Pt and CoS, respectively, for the iodide
sured insolation level. Each of the sulfide and sulfide MPECs [2, 3].
or iodide indirect regenerative MPECs
probed in this study exhibit conversion 4.4.2.4 Bipolar Band Gap Solar Storage
efficiency comparable to, or better than, Cells
the solid-state device in Table 2. This indi- One limitation to solar cells is that, while
cates that electrolytic polarization losses most of our electrical needs are contin-
in these cells are comparable to resis- uous, clouds and darkness dictate that
tance losses of the semiconductor back solar energy is intermittent in nature.
contact in the solid-state MPEC. The PECs can generate not only electrical but
marginally lower conversion efficiency of also electrochemical energy, and provide
a single device to solve the problem of band gap and electrochemical storage,
the intermittent nature of solar energy. which unlike conventional photovoltaics,
In 1987 we presented a highly efficient provides a nearly constant energetic output
single band gap photoelectrochemical cell in illuminated or dark conditions. The cell
combining in situ electrochemical storage combines bipolar AlGaAs (Eg = 1.6 eV)
and solar conversion capabilities, pro- and Si (Eg = 1.0 eV) and AB5 metal hy-
viding continuous output insensitive to dride/NiOOH storage. The NiOOH/MH
daily variations in illumination. The 1987 metal hydride storage process is near ideal
configuration was demonstrated with a for the AlGaAs/Si because of the excellent
n-Cd(Se,Te)/polysulfide electrolyte conver- match of the storage and photocharg-
sion half cell and a Sn/SnS storage system, ing potentials. The electrochemical storage
resulting in a single cell operating contin- processes utilizes MH oxidation and nickel
uously at an overall efficiency of 11% [18]. oxyhydroxide reduction [22]:
Under illumination, photocurrent droves
an external load. Simultaneously, a portion MH + OH− −−−→ M + H2 O + e− ;
of the photocurrent was used in the direct EM/MH = −0.8 V versus SHE (13)
electrochemical reduction to a metal (Sn)
in the device storage half cell. In darkness NiOOH + H2 O + e− −−−→
or below a certain level of light, the stor- Ni(OH)2 + OH− ;
age compartment delivers power by metal
oxidation. ENiOOH/Ni(OH)2 = 0.4 V vs SHE (14)
Recently, high solar conversion and stor- As shown in Fig. 10, the cell gener-
age efficiencies have been attained with ates a light variation insensitive potential
a system that combines efficient mul- of 1.2–1.3 V at total (including storage
tiple band gap semiconductors, with a losses) solar/electrical energy conversion
simultaneous high-capacity electrochem- efficiency of 18%. Over an eight-month pe-
ical storage [4, 5]. The energy diagram for riod of daily cycles, under constant 12-hour
one of several multiple band gap cells is (AM0) illumination, the long-term indoor
presented in Fig. 3(d), and as described in cycling cell generated a nearly constant
Fig. 2, several other configurations are also photocurrent density of 21.2 (constant to
feasible. In the figure, storage occurs at a within one percent or ±0.2 mA cm−2 ), and
potential of Eredox = EA+/A − EB/B+ . On a photopower that varied by ±3% [5]. The
illumination, two photons generate each cell is a single physical/chemical device
electron, a fraction of which drives a load, that generates load current without any
while the remainder (1/xe− ) charges the external switching.
storage redox couple. Without light, the
potential falls below Eredox and the stor- 4.4.2.5 Bipolar Band Gap Solar Hydrolysis
age couple spontaneously discharges. This (hydrogen generation) Cells
dark discharge is directed through the Solar energy–driven water splitting com-
load, rather than through the multijunc- bines several attractive features for energy
tion semiconductor’s high dark resistance. utilization. Both the energy source (sun)
In Fig. 9 is presented an operational and the reactive media (water) are readily
form of the solar conversion and storage available and are renewable, and the re-
cell described by the Fig. 2 energy diagram. sultant fuel (generated hydrogen) and the
The single cell contains both multiple emission with fuel consumption (water)
374
Fig. 9
2hν
Illumination ⇓
In dark
Au-Zn/Au Au-Zn/Au Photo
+ current
AR coating [ZnS(50 nm)/MgF2(70 nm)] p+-GaAs AR coating [ZnS(50 nm)/MgF 2(70 nm)] p -GaAs
18 −3 18 −3
significant back current

p+-Al0.8Ga0.2As (1 × 10 cm ) 50 nm p+-Al0.8Ga0.2As (1 × 10 cm ) 50 nm
p+-Al(0.3 – 0.15)Ga(0.7– 0.85)As (1 × 1018 cm−3) 300 nm p+-Al(0.3 – 0.15)Ga(0.7–0.85)As (1 × 1018 cm−3 ) 300 nm
17 −3 17 −3
Reverse resistance prevents

n -Al(0.15)Ga(0.85)As (2 × 10 cm ) 1.0 µm n-Al(0.15)Ga(0.85)As (2 × 10 cm ) 1.0 µm
Rload
Rload
+ 18 −3 + 18
n -Al(0.15)Ga(0.85)As (1 × 10 cm ) 1.7 nm n Al(0.15)Ga(0.85)As (1 × 10 cm−3 ) 1.7 nm
n -GaAS 20 nm n-GaAS 20 nm
←1/xe−
GaAs(Buffer layer) 10 nm GaAs(Buffer layer) 10 nm

e−→
p+-Si (1 × 1019 cm−3) 1.0 µm p+-Si (1 × 1019 cm−3 ) 1.0 µm

n -Si (8 × 1015 cm−3) 350 µm n-Si (8 × 1015 cm−3 ) 350 µm
←e
19 −3 n+-Si
n+-Si
−
(4 × 10 cm ) 800 nm (4 × 1019 cm−3 ) 800 nm
solar spectrum is transmitted through

and visible energy rich portion of the
are each environmentally benign. The UV
H2 O. Therefore sensitization, such as

Au-Sb/Au Au-Sb/Au Load
4 Solar Energy Conversion without Dye Sensitization
←1 − 1/xe−
e− →
Metal hydride anode Metal hydride anode current
MHx + OH− → MHx−1 + H2O + e− MHx−1 + H2O + e− → MHx + OH−
Separator with KOH electrolyte Separator with KOH electrolyte

− −
Day-time energy conversion of the
The bipolar AlGaAs/Si/MH/NiOOH MBPEC solar cell.

Storage/bipolar band gap solar cell.
NiOOH + H2O + e → Ni(OH)2 + OH
Storage/bipolar band gap solar cell.

Night-time energy conversion of the
Ni(OH)2 + OH → NiOOH + H2O + e
− −
Nickel cathode Nickel cathode

Storage
current
the water-splitting process. In a solar

photoelectrolysis system, the redox active
via semiconductors, is required to drive
interfaces can be in indirect or direct

28 28
Bipolar gap in situ storage solar cell
AlGaAs/Si-MH/Ni cell
24 24
IV curve at full charge
Solar to electrical conversion efficiency: 19.6%
20 Illumination: Simulated AMO (135.5 mW cm2), Illuminated area 0.22 cm2 20
1.5
Load potential
Photopower density
1.0 Generated cell potential is insensitive
16 [V]
to variation in illumination
16
[mA cm−2]
[mA cm−2]
0.5
Illumination Dark Illumination Dark
12 IPhoto 12
4
IDischarge IDischarge
8 2 8
[mA]
0
I
IDark IDark
4 -2 ICharge 4
0 6 12 18 24 30 36 42 48
0 Time 0
[hours]
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

Photovoltage
[V]
Fig. 10 Photoconversion current and power solar cell under simulated AM0 insolation. In the
characteristics of the AlGaAs/Si/MH/NiOOH first 12 hours of each of the two 24-hour cycles,
solar cell under fully charged AM0 conditions. the AM0 illumination is varied from 0 to AM0
Inset: Two days conversion and storage back to 0, using a graded diffuse filter; the
characteristics of the AlGaAs/Si/MH/NiOOH second 12 hours are without illumination.
contact with the photosensitizer, and com- The subsequent challenge is to optimize
prise either an ohmic or Schottky junction. sustained water electrolysis, without con-
Independent of this interface composition, siderable additional energy losses. Effec-
the various parameters in models predict- tive water electrolysis must occur at a
ing solar water–splitting conversion effi- potential, VH2 O , near the photocell point
ciency, may be combined into two general of maximum power. VH2 O is greater than
parameters: (1) related to losses in optical EHo (=−#GH2 O /nF) as a result of over-
2O
energy conversion, ηphoto , or (2) related to potential losses, ζ , in driving an electroly-
losses in redox conversion of water to H2 sis current density, j , through both the O2
and O2 , ηelectrolysis . Combined, these yield and the H2 electrodes:
an overall solar electrolysis efficiency (ex-
VH2 O (j ) = EO2 (j ) − EH2 (j )
cluding storage and utilization losses) as: % o & o
= EO + ζO2 (j ) − (EH + ζH2 (j ))
ηphotoelectrolysis = ηphoto × ηelectrolysis 2 2
(17)
(15)
A cell containing illuminated AlGaAs/Si
The thermodynamic potential, E o H 2O , for
RuO2 /Ptblack is demonstrated to evolve H2
the water splitting reaction is given by:
and O2 at record solar-driven water elec-
H2 O −−−→ H2 + 1/2O2 ; trolysis efficiency [6]. Under illumination,
o o ◦ bipolar configured Al0.15 Ga0.85 As (Eg =
EH 2O
= EO − EH2 ;
2 1.6 eV) and Si (Eg = 1.1 eV) semiconduc-
o ◦
EH 2O
(25 C) = 1.229 V (16) tors generate open-circuit and maximum
power photopotentials of 1.30 and 1.57 V, cell are available [6]. Under illumination,
well suited to the water electrolysis ther- this bipolar cell generates an open-circuit
modynamic potential of EH o
2O
(25 ◦ C) = potential of 1.57 V, which is considerably
o
1.229 V. The EH2 O /photopotential-mat- larger the thermodynamic potential for
ched semiconductors are combined with the water-splitting reaction. Also included
effective water electrolysis O2 or H2 elec- are the photopotential dependence of so-
trocatalysts, RuO2 or Ptblack . The resultant lar to electrical conversion efficiency. A
solar photoelectrolysis cell drives sustained portion of the high ηphoto domain lies at
water splitting at 18.3% conversion effi- a potential above EH o , and in principle
2O
ciencies. Alternate dual band gap systems is sufficiently energetic to drive efficient
are calculated to be capable of attaining photoelectrolysis.
more than 30% solar photoelectrolysis con- Planar platinum and Ptblack are effective
version efficiency. H2 electrocatalysts. At low current densi-
Figure 11 presents the measured cur- ties (0.5 mA cm−2 ), the observed Pt over-
rent/voltage characteristics for the Al- potential is low (ζH2 = −17 mV), and even
GaAs/Si photocell at AM0 illumination. smaller (ζH2 = −13 mV) with the Ptblack
Further details and characterization of the electrocatalyst, and smaller yet (ζH2 =
Jphoto E oH2O,the
24 minimum
water electrolysis
potential
AlGaAs/Si cell characteristics
20
Area: 0.22 sq.cm
AM0 (135 mW/sq.cm) illumination
Open-circuit potential: 1.57 V
16 Short-circuit current: 5.2 mA
Fill factor: 0.772
[mA sq.cm]
Jphoto
12 ηphoto
Photoelectrolysis conversion
efficiency upper limit
4 = η photo × E °H2O/Vphoto
0
0 200 400 600 800 1000 1200 1400 1600
Vphoto
[millivolt]
Fig. 11 AlGaAs/Si I–V characteristics at AM0 light intensity (135 mW cm−2 ).
ηphoto = 100% × (Jphoto ∗ Vphoto )/Pillumination . Using the indicated, measured ηphoto
and Vphoto , the upper limits of photoelectrolysis efficiency are calculated as
ηphoto × 1.229 V/Vphoto .
−7 mV) with a low level of convection portion of Fig. 12 presents the water elec-
(stirring) to improve mass transport and trolysis potential, VH2 O , determined under
prevent observed gas buildup on the elec- stirred or quiescent conditions, and mea-
trode surface. Minimization of ζO2 is a sured in 1 M HClO4 using equal areas of
greater challenge. In this case Pt is a the optimized RuO2 and Ptblack electrodes.
poor electrocatalyst. ζO2 may be decreased These measurements of the electrolysis
∼500 mV, utilizing an RuO2 electrode. current, as a function of potential, enable
The kinetics of this effective oxygen elec- us to predict the AlGaAs/Si photocurrent
trocatalyst [23] have been attributed to at which photoelectrolysis will occur.
catalysis by the intervening RuO4 /RuO2 Our photoelectrolysis cell consists of
redox couple [24, 25]. In the absence of illuminated AlGaAs/Si RuO2 /Ptblack elec-
competing redox couples, the faradaic effi- trolysis. With these active electrocatalysts,
ciency of H2 and O2 evolution approaches high values of ηelectrolysis are insured by
100%, and ηelectrolysis is determined by the using large surface areas of the electrolysis
current limited VH2 O (j ): electrodes, compared to the illuminated
area. This is accomplished without in-
o
ηelectrolysis = EH /VH2 O (j ); creasing the illuminated electrode area,
2O
◦
as schematically represented in the lower
ηelectrolysis (25 C) = 1.229V/VH2 O (j ) portion of Fig. 13, by utilizing a large
(18) vertical depth of electrolysis electrodes
The Fig. 12 inset contains the Eq. (18)–de- compared to the cross section of illu-
termined ηelectrolysis . The limiting maxi- mination. Specifically, the 10 cm2 Ptblack
mum ηphotoelectrolysis can be readily de- and RuO2 electrodes utilized are large
termined from the solar to electrical compared to the 0.22 cm2 illuminated
conversion efficiency. Expanding Eq. (15) area. In the Fig. 13(a) is the H2 and O2
with Eq. (18), ηphotoelectrolysis is diminished electrolysis current generated by this cell
from ηphoto by the potential of the stored as a function of time. The average pho-
energy compared to the potential at which tocurrent of 4.42 mA (generating a current
the water electrolysis occurred: density of 20.1 mA cm−2 at the illumi-
nated electrode area, and 0.44 mA cm−2 at
ηphotoelectrolysis (T) the electrolysis electrodes) corresponds to a
ηphoto (T) × EH2 O (T) photopotential of 1.36 V in Fig. 11 compa-
= ; rable to the equivalent electrolysis potential
EO2 (T) − EH2 (T)
at 0.5 mA cm−2 in Fig. 12. The overall effi-
ηphotoelectrolysis (25 ◦ C) = 1.229V × ηphoto ciency is determined by the 1.229 V energy
VH2 O stored as H2 and O2 and the incident pho-
(19) topower (135 mW cm−2 ):
Figure 12 includes these limiting
ηphotoelectrolysis values because the Al- ηphotoelectrolysis = 100% × 20.1 mA cm−2
GaAs/Si system is determined from mea- × 1.229 V/135 mW cm−2 = 18.3% (20)
sured values of ηphoto at various values of
Vphoto . AlGaAs/Si solar photoelectrolysis 4.4.2.6 Higher Solar Production Rates of
at conversion efficiencies exceeding 20% Hydrogen Fuel are Attainable
(Fig. 11) are in principle possible at poten- In Eq. (15), the significance of the elec-
tials approaching the EH2 O limit. The main trolysis compared to photo components
100 20
90
80
ηelectrolysis
16
[%]
70
60
12
[mA sq.cm]
50
Jelectrolysis
1200 1600 2000 2400
Velectrolysis
[mv] 8
4
Electrolysis in 1 M HClO4 Stirred
Anode: Pt black on Pt mesh
Quiescent
Cathode: 2.3 µm RuO2
0
1200 1300 1400 1500 1600
Velectrolysis
[millivolt]
Fig. 12 Measured variation of the VH2 O with current density in 1 M HClO4 , using
equal area 2.3 µm RuO2 and Ptblack . Inset: Calculated ηphoto , in a 100%
faradaic-efficient process, as a function of VH2 O .
of the conversion efficiency is evident on in series, effectively increasing the relative

analysis of a report of a 12.4% photo- photoelectrolysis efficiency by 50%.
electrolysis cell. The cell contained a Pt Previous principal solar water splitting
coated GaInP2 hydrogen electrode in bipo- models predict similar dual band gap pho-
lar contact with a GaAs junction driving toelectrolysis efficiencies of only 16%, and
an oxygen electrode [26]. At their specified 10–18%, respectively [27, 28]. Each are
11 sun–illuminated electrolysis, current lower than our observed water splitting
density of 120 mA cm−2 , ζH2 will exceed efficiency discussed below. The physics of
300 mV at Pt. Oxygen losses will be even these models were superb, but their analy-
larger, and the cell will have a total overpo- sis was influenced by dated technology and
tential loss >700 mV. Two such cells will underestimated the experimental ηphoto
have losses >EH o ; unnecessary losses if attained by contemporary devices or un-
2O
the cell was illuminated at 1 sun or em- derestimated the high experimental values
ployed larger surface area or more effective of ηelectrolysis , which can be attained. For
electrocatalysts. Under these conditions, example, Boltonand coworkers, estimates
two GaInP2 /GaAs cells placed in series low values of ηphoto (less than 20% conver-
should drive three water electrolysis cells sion) due to assumed cumulative relative
AIGaAs/Si RuO2/Ptblack Photoelectrolysis

Illumination power: 135 mW cm−2
5 Illuminated & electrolysis area: 0.22 cm2 & 10 cm2
Photoelectrolysis current
Solar water-splitting system

4
hν
Cross Illumination ⇓
hν section AR coating [ZnS(50 nm)/MgF2(70 nm)]
expanded P+-AI0.8Ga0.2As 50 nm
P+-AI(0.3−0.15)Ga(0.7−0.85)As 300 nm
P+-GaAs Au-Zn/Au
[mA]
3 n -AI(0.15)Ga(0.85)As 1.0 µm
n+-AI(0.15)Ga(0.85)As 1.7 nm
n -GaAs 20 nm
GaAs (Buffer Layer) 10 nm
RuO2
Ptblack
2
+
p -Si 1.0 µm
n -Si 350 µm
n+-Si 800 nm
Ru02
Au − Sb/Au
Ptblack
1
1M HCIO4
0
2 4 6 8 10 12
Time
[hours]
Fig. 13 Schematic representation (inset) and measured characteristics of the illuminated
AlGaAs/Si RuO2 /Ptblack photoelectrolysis cell. Further details of the layered AlGaAs/Si structure
are given in Ref. 6.
secondary losses that include: 10% re- such cells in series will efficiently drive
flection loss, 10% quantum-yield loss, three 1.3 to 1.4 V water electrolysis cells
20% absorption loss. As summarized in in series, and as discussed here at a wa-
Table 3 demonstrated ηphoto are substan- ter electrolysis at 1.36 V (Fig. 12 inset)
tially higher than 20%. Many ηphoto that yielded electrolysis efficiencies of more
have been carefully reviewed or monitored than 90%. It is reasonable that with larger
are in excess of 30% [29]. surface area, or more effective electrocatal-
Each of the cells in the Table 3 exhibits ysis, these efficiencies will approach 95%.
an open-circuit photopotential signifi- Using this range of ηelectrolysis = 90–95%:
cantly greater than the minimum potential
ηphotoelectrolysis (predicted maximum)
necessary to split water. The majority of
= ηphoto × 90–95% (21)
these cells can generate a photopotential
in excess of 2 volts. An unnecessary limit Table 3 includes predicted maximum
of one multiple band gap photoexcitation ηphotoelectrolysis using observed ηphoto of var-
per electrolysis would under-utilize Vphoto , ious dual band gap sensitizers. It is seen
diminishing ηelectrolysis . For example, a that solar water splitting efficiencies may
GaInP/GaAs cell has a maximum power be viable at up to double the amount of that
photopotential of 2.0 to 2.1 V and an previously predicted. Efficient, three or
open-circuit potential of 2.3 V [12]. Two more multiple band gap photoelectrolysis
Tab. 3 Predicted and measured photoelectrolysis efficiencies. Calculated ηphotoelectrolysis are from
Eq. (21)
Photovoltaic Light ηphoto ηphotoelectrolysis

Level Measured Predicted Maximum Measured
GaInP/GaAs 1 sun 30.3% (Ref. 30) 27–29% (Ref. 6)

GaInP/GaAs 180 sun 30.2% (Ref. 30) 27–29% (Ref. 6)
GaAs/Si 350 sun 29.6% (Ref. 30) 27–28% (Ref. 6)
GaAs/GaSb 100 sun 32.6% (Ref. 30) 29–31% (Ref. 6)
InP/GaInAs 50 sun 31.8% (Ref. 30) 29–30% (Ref. 6)
GaAs/GaInAsP 40 sun 30.2% (Ref. 30) 27–29% (Ref. 6)
AlGaAs/Si 1 sun 21.2% (Ref. 15) 19–20% (Ref. 6) 18.3% (Ref. 6)
GaInP2 /GaAs 11 sun 12.4% (Ref. 27)
InP/GaAs 1 sun 8.2% (Ref. 30)
will be expected to be capable of attaining phenomena. This systematically probes

even higher efficiencies. the approach in studies on polysulfide,
ferrocyanide, polyselenide, and polyiodide
4.4.3 PECs.
Solution Phase Phenomena
4.4.3.2 n-Cd Chalcogenide/Aqueous
4.4.3.1 Solution Phase Chemistry Polysulfide Photoelectrochemistry
Optimization In 1976, the first regenerative PECs with
In PECs, the manifold conditions in which substantial and sustained solar to electrical
solution phase chemistry constrains the conversion efficiency were demonstrated.
kinetics and thermodynamics of pho- These PECs are based on n-type cad-
toelectrochemical charge transfer have mium chalcogenide (S, Se or Te) electrodes
been probed. Modification of the solu- immersed in aqueous polychalcogenide
tion phase chemistry, can substantially electrolytes. The cells were introduced
impact on photoelectrochemical proper- by Hodes, Cahen, and Manassen [31],
ties by (1) enhancing facile charge trans- Wrighton and coworkers [32], and Heller
fer, (2) suppressing competing reactions and Miller [33] and were capable of con-
and suppressing both (3) electrode and verting up to 7% of insolation to electrical
(4) electrolyte decomposition products, as energy. Most investigations of these sys-
well as (5) substantially affecting the open- tems focused on solid-state and interfacial
circuit photovoltage. These limitations aspects of these PECs and photodriven ox-
have been studied in a variety of redox idation of polysulfide at the photoelectrode
couples and electrolytes. A chemical mech- was represented:
anism was presented [7, 8] for photoelec-
− n-cd(Se,Te)
trochemical solar to electrical energy con- S− + 2hn −−−−−−→ S + 2e− (22)
version which emphasized understanding
of the distribution of species in photoelec- However as represented in Fig. 14(c), the
trochemical electrolytes. The approach sys- number and type of species in polysulfide
tematically modifies observed photoelec- photoelectrolytes is considerably more
trochemical kinetic and thermodynamic complex than represented by Eq. (22). The
Photoelectrochemical solar cell solution phase studies

hν hν hν
Photoelectrode Photoelectrode Photoelectrode

In solution In solution
H2S Fe(CN)64− In solution
HS− HxFe(CN)6(x − 4) H2Se
S= KyFe(CN)6(y − 4) HSe−
S2= HxKyFe(CN)6(y + x − 4) Se=
Se2=
Load
Load
Load
S2− HxFe(CN)6(x − 3)
S3= Se3=
KyFe(CN)6(y − 3) Se4=
S4= HxKyFe(CN)6(y + x − 3)
S5= OH−
CN−
H+
OH− OH−
H+ H+
Counter electrode Counter electrode Counter electrode
(a) Aqueous polysulfide PEC (b) Aqueous ferrocyanide PEC (c) Aqueous polyselenide PEC
Fig. 14Chemical species coexisting in polysulfide (a), ferrocyanide (b) and polyselenide (c)
aqueous electrolytes used in regenerative PECs.
addition to water of a simple soluble preferred cation, and a sulfur to sulfide

sulfide salt and sulfur to water gives ratio of approximately 1.5 to 1 resulted
rise to a wide distribution of species in in a near doubling of the conversion effi-
solution. The equilibria constraining this ciency [47]. Figure 15 presents n-Cd(Se,Te)
distribution has been investigated by Licht photocurrents measured at various applied
and coworkers [34–39], Gigenbach [40], potentials in a traditional and modified
and Teder [41]. electrolyte. As seen in the figure, the cu-
It has been shown that modification mulative effect of polysulfide electrolyte
of the distribution of species in aqueous modifications on photoelectrochemical so-
polysulfide solution can be correlated to lar to electrical energy conversion by n-
variations in transparency, conductivity, Cd(Se,Te)/aqueous polysulfide PECs can
activity, and cadmium chalcogenide photo- be considerable.
electrochemistry. Specifically, the separate A greater percentage of photogenerated
effects of hydroxide and pH modifica- holes utilized in constructive oxidation of
tion [42], sulfur [43] and sulfide [44], and polysulfide results in enhanced photocur-
cation [45, 46] were investigated in terms rents. This has the additional benefit of
of speciation and photoelectrochemical fewer oxidizing holes available for attack
phenomena. Previously cadmium chalco- on the semiconductor (photocorrosion).
genide polysulfide PECs had generally As seen in Fig. 16, this results in en-
employed electrolytes composed 1 molar hanced photocurrent stability of the PEC.
in sodium sulfide, sulfur and sodium hy- This study showed that with solution opti-
droxide and resulted in solar conversion mization, not only the photocurrent but
efficiencies of approximately 7% [31–33]. also the polysulfide electrolyte exhibits
However, added hydroxide is to be enhanced lifetime, both approaching one-
minimized in these cells, cesium is the year operation outdoors [48].
n-CdSe0.65 Te0.35 in aqueous polysulfide Fig. 15 Potentiostatic photocurrent

voltage characteristics for an
illuminated n-CdSe0.65 Te0.35 single
crystal immersed in either of two types
12.7% efficiency of aqueous polysulfide electrolyte: Top
curve in 1.8 M Cs2 S and 3 M sulfur:;
Photocurrent
1.6 bottom curve is in 1 M NaOH, 1 M
7.7% efficiency Na2 S, 1 M sulfur. The photocurrent
[mA]
voltage curves were obtained outdoors,
and solar to electrical conversion
0.8 efficiencies are indicated.
~AM1 illumination
−800 −600 −400 −200
Voltage
[mV]
0.5
0.4
Photovoltage over load
0.3
[Volts]
0.2
0.1
Jan July Jan July

Time
Fig. 16 Long-term outdoor stability curves of thin film
CdSe0.75 Te0.35 in electrolytes of either 1.8 M S (solid line) or 2.0 M
KOH, 1.4 M Na2 S, and 2.6 M S (shaded region). The
sodium/potassium electrolyte cell results are average results over
separate cells, with initial conversion efficiencies from 3.5 to 4.5%.
The cesium electrolyte cell had 4.6% conversion efficiency potential
measured under a load chosen for initial maximum power.
4.4.3.3 n-Cd Chalcogenide/Aqueous The role of the various redox active

Ferrocyanide Photoelectrochemistry constituents in constraining n-CdSe/
Reichman and Russak [49] and Freeze [50] Fe(CN)6 3−/4− photo-oxidative energy con-
had reported high (12 to 14%)% conversion& version had not been previously investi-
efficiencies for n-CdSe/ Fe(CN)6 3−/4− gated, and ferricyanide and ferrocyanide
solar cells in electrolytes containing a 1 : 1 salts exhibit a complex aqueous equilib-
ratio of aqueous Fe(CN)6 4− to Fe(CN)6 3− ria. However, in a manner analogous to
salts in a highly alkaline environment. the polysulfide electrolytes, and as illus-
However, Rubin and coworkers [51] had trated in Fig. 14(b), these salts exhibit a
shown that these n-CdSe/Fe(CN)6 3−/4− complex aqueous equilibria, resulting in
solar cells can have a limited stability and a variety of Fe(II) and Fe(III) species. As
photocurrents decrease within the order with the polysulfide electrolytes, we have
of hours. Their attempts to enhance the probed the equilibria constraining the dis-
photoconversion stability of these systems tribution of species in these solutions. This
have included cationic surface modifica- speciation can be controlled and substan-
tion of the n-CdSe photoelectrode and tially effects n-CdSe photo-oxidative energy
isolation of the wavelength dependence conversion. The photoelectrochemical ef-
of the surface instability [51, 52]. In these fect and solution stability effects of pH,
studies, photo-oxidative processes in n- cation modification and ratio of ferro-
CdSe/Fe(CN)6 3−/4− PECs were typically cyanide to ferricyanide, have been studied
represented: by Licht and Peramunage [9, 53–55], and
n-CdSe also the systematic variation of the primary
Fe(CN)6 4− + hn −−−−→ Fe(CN)6 3− + e− photo-oxidized species. As seen in Fig. 17,
(23) substitution of a single ligand in the
10 Photoelectrochemistry of n -CdSe in
[Fe(CN)5 -L]3−/2− aq
L
a
Photocurrent
a) −CN−
[mA cm−2]
5 b b) −NH3
c) −NO2−
c
d d) −Fe(CN)53−
e e) −NO+
0
−0.8 −0.4 0.0

Potential
[volts vs Pt]
Fig. 17 Potentiostatic photocurrent voltage curves for an
illuminated n-CdSe single crystal in several aqueous modified
ferro/ferricyanide electrolytes in which 1 of the hexacyano
(Fe(CN)6 ) has been replaced by the indicated ligands.
hexacyanoferrate species has a substantial 4.4.3.4 n-GaAs/Aqueous Polyselenide

effect on photovoltage and photocurrent, Photoelectrochemistry
Licht and Peramunage [56]. n-GaAs/aqueous polyselenide PECs have
As seen in Fig. 18, addition of potassium shown stable efficient solar to electrical
cyanide to an alkaline ferrocyanide elec- conversion. Photodriven oxidation of poly-
trolyte substantially enhances both photo- selenide can be written as
current stability (figure inset) and pho- − n-GaAs
tovoltage (figure outset), Licht and Pera- Se− + 2hn −−−−→ Se + 2e− (24)
munage [9]. It was proposed by Bocarlsy or:
and coworkers [57] that in addition to
− n-GaAs −
Fe(CN)6 4− , CN− was also photooxidized 2Se− + 2hn −−−−→ Se2 − + 2e− (25)
(to CNO− ) by the photoelectrode. Our sub-
sequent investigations indicated that al- Parkinson, Heller, and Miller [58] and
though cyanide may be chemically oxidized Lewis and coworkers [59] have shown that
to cyanate, no cyanide was photoelectro- metal ion (Ru3+ , Os3+ ) treatment of the
chemically oxidized by the semiconductor n-GaAs surface leads to high solar to elec-
Licht and Peramunage [53–56]. trical conversion efficiencies in these cells
40 mW cm−2 illuminated single crystal n-CdSe in

20
0.5 mK4Fe(CN)6, 25 mm K3Fe(CN)6, 0.5 m KOH
without (dashed) or with (solid line) 0.1 m KCN
15
Current Density
Iphoto (normalized)
[mA cm−2]
1 θ
10
d
b
5 0
0 1 Time 3
[days]
No Illumination
0
−1.2 −1.0 −0.8 −0.6 −0.4 −0.2 0.0

Voltage
[volts vs Pt]
Fig. 18 Photocurrent-voltage curves of illuminated single crystal
n-CdSe immersed in alkaline potassium ferrocyanide electrolytes
with and without added cyanide. Inset: Photocurrent stability of in
several electrolytes. ‘‘e’’ is the only electrolyte with cyanide.
Electrolytes ‘‘d’’ and ‘‘e’’ contain low ferricyanide. Electrolyte ‘‘b’’
contains high ferricyanide. Specifically, e: 0.25 m K4 Fe(CN)6 ,
0.01 m K3 Fe(CN)6 , 0.5 m KOH 0.5 m KOH, 0.1 m KCN; d: 0.25 m
K4 Fe(CN)6 , 0.01 m K3 Fe(CN)6 , 0.5 m KOH; b: 0.25 m K4 Fe(CN)6 ,
0.25 m K3 Fe(CN)6 , 0.5 m KOH. In an electrolyte comparable to
‘‘b’’ but with pH = 3.8, photocurrent diminished within seconds.
(with up to 15% conversion efficiencies − − −

Se3 − + Se− ←−→ 2Se2 − (26)
reported). Lewis and coworkers [60] have − − −
− − −
further investigated 0.1 M to 1.0 M con- 2Se4 + Se ←−→ 3Se3 (27)
centration K2 Se electrolytes. The first and with equilibrium constants for the equilib-
second acid dissociation constant of hy- rium given by K23 and K34 , respectively.
drogen selenide of pK1 = 3.9 and pK2 = These equilibria were probed by study of
13.0 have been well characterized by sev- rest potential variation with solution com-
eral investigators including Myers and position and by isolation of the near UV ab-
coworkers [61]. However, as represented sorption peaks of the various polyselenide
in the right hand portion of Fig. 14, the species, and yield equilibrium constants
fundamental equilibria constraining the of pK23 = −0.65 and pK34 = −4.2. When
distribution of polyselenide species was combined with K1 and K2 these provide a
not characterized. Measurements of these description of polyselenide speciation [62,
equilibria are a first step in determining 63].
the regenerative reactive species in pol- Using the understanding of polyselenide
yselenide medium and in probing how speciation in solution, the rationale elec-
they effect photoelectrochemical energy trolyte modification of aqueous polyse-
conversion. lenide photoelectrochemical solar cell gas
The equilibria constraining the forma- been investigated. n-GaAs photocurrent,
tion of diselenide, triselenide, and tetrase- photovoltage, and photopower are af-
lenide may be written: fected by the distribution of hydroselenide,
19
J, Stable
18
J, Photocurrent density
J, Fluctuating
[mA cm−2]
(Corrosion)
J, Diminishing
17
16
15
0 1 2 3 4
KOH
[M]
Fig. 19 Short-circuit photocurrent density, Jsc , for a 75 mW cm−2
tungsten-halogen illuminated single crystal n-GaAs immersed 1 M K2 Se,
0.01 M Se and in varying KOH concentration.
selenide, and polyselenide in solution. cell configuration discussed in Licht and

Polyselenide electrolytes containing 1 to Forouzan [62, 63].
2 m hydroxide and 1 m selenide (as K2 Se)
4.4.3.5
enhance n-GaAs photocurrent density sta-
Aqueous Polyiodide Photoelectrochemistry
bility (Reported as ‘‘J’’ in Fig. 19), whereas
photocurrent is unstable at either 0.5 m The aqueous polyiodide (I− /I3 − ) is the
KOH or 3.0 m KOH. The presence of poly- only redox couple shown to be compa-
selenide species in low concentrations (as tible with efficient oxidative photoelec-
0.01 m dissolved selenium) are important trochemistry at n-type transition metal
for desorption of selenium from the n- dichalcogenide including WSe2 , MoS2 ,
GaAs. Furthermore, electrolytes enriched WS2 , and MoSe2 . Following the pioneering
in dissolved selenium, (as up to 0.2 m work of Tributsch and coworkers [64], sev-
dissolved selenium) enhance n-GaAs pho- eral groups including Parkinson, Heller,
tovoltage. As seen in Fig. 20, the pho- and Miller [65] and Tenne and Wold [66]
topower (the product of the photocurrent have reported n-WSe2 or n-MoSe2 /aque-
and the applied potential) as well as the ous polyiodide solar to electrical en-
photopower improves with an increase ergy conversion of more than 10%
in selenium. However, these electrolytes and/or photocurrent stability in excess
diminish electrolyte transmittance neces- of 105 coulombs cm−2 . The photodriven
sitating use of an alternate back wall oxidation of iodine at tungsten diselenide
3.0
0.01 M Se
2.5 0.02 M Se
0.20 M Se
2.0
[mW cm−2]
1.5
Power
1.0
0.5
0.0
0.0 0.2 0.4 0.6

Photovoltage
[volts]
Fig. 20Photopower variation of a 92 mW cm−2 tungsten-halogen illuminated backwall
n-GaAS/polyselenide PEC as a function of increasing selenium concentration.
has been described in one of two manners: To a lesser extent the solution phase
n-WSe2 species IO3 − , HIO, H2 O I+ , I5 − , and
◦
I− + hn −−−−→ I + e− (28) I6 2− can also occur. Figure 21 presents the
relative variation of polyiodide speciation
or: in solutions containing 0.01 molal iodine
n-WSe2 and 1 to 12 molal iodide and which may
3I− + 2hn −−−−→ I3 − + 2e− (29) be compatible with n-WSe2 regenerative
photoelectrochemistry. As emphasized in
Significant species in aqueous polyiodide the figure inset, in these electrolytes the
solution can include: I− , I3 − , I4 2− , I− , concentration of added iodide dominates
OH− , and H+ . Constrained by pH and the iodine in these solutions. The distribution
equilibria [67, 68]: of species in these solutions is calculated in
accordance with Eqs. (30) and (31). As can
I− + I2 ←−→ I3 − K3 = 723 (30)
be seen in the figure inset, little (less than
I3 − + I− ←−→ I4 2− K4 = 0.20 (31) 0.3%) of the added iodine exists as solution
100%
Aqueous polyiodide speciation

0.01 M iodine in the
indicated concentration of sodium iodide
80%
Total iodine = 0.01 m = [I2] + [I3− ] + [I 4= ]
Total iodide = [Nal] = [I −] + [I3−] + 2[I4=]
60%
Percent of total
[I3−] / Total iodine [I 4=] / Total iodine
40%
0.3%
[I2] / Total iodine
% of total
Total iodine / Total iodide

20%
0.0%
0 4 8 12
NaI
[m]
0%
0 2 4 6 8 10 12
NaI
[molal]
−
Fig. 21 The fractions of I3 − and I4 − (compared to total iodine) in aqueous polyiodide
solutions as a function of added iodide concentration. Inset: The fractions of solution
phase iodine (compared to total iodine) and total iodine (compared to total iodide).
phase iodine, and the maximum [I2 ] is These improvements are sustained in the
(1 × 10−4 m. As seen in the figure outset, silver-bearing electrolytes. On returning
the bulk (greater than 99.7%) of the added the electrode to the polyiodide electrolyte
iodine resides as the polyiodide species without silver, the PEC gradually (on the
I3 − and I4 2− . In iodide concentrations less order of hours) returns to the silver free
than 5 m, I3 − is the predominant polyio- PEC behavior. This improvement appears
dide, whereas at higher iodide concentra- to provide a mechanism for long-lasting
tions, I4 2− predominates. Longer chain suppression of n-WSe2 exposed edges and
species, I5 − and I6 2− , if they exist, have recombination sites as further discussed
upper limit concentrations of 1 × 10−6 m by Licht and Myung [67, 68].
and 1 × 10−8 m, respectively. Hence, in
these aqueous polyiodide electrolytes, I2 , 4.4.4
I5 − and I6 2− are not at substantial concen- Concluding Remarks
trations in these photoelectrolytes, and a
primary oxidizable ion is I− and a primary In the 1970s through 1990s health con-
reducible ion is either I3 − or I4 2− . cerns (air pollution) and political concerns
Choice of cation in aqueous polyiodide (localized shortages) were cited in the
solution can effect n-tungsten dichalco- need for development of alternative en-
genide photoelectrochemistry. The alkali ergy sources. Today, in addition to these
cations do not substantially interact with concerns, growing awareness of carbon
solution phase iodide. However, other
dioxide emissions as a green house gas,
metal cations including Ag+ , or Zn2+ or
as well as the economic realization that
Cd2+ , will create a series of complexes
oil and coal are better allocated as raw
with dissolved iodide including: AgI2 − ,
materials for pharmaceuticals and plas-
AgI3 2− , ZnI3 − , ZnI4 2− , CdI4 2 and so
tics, than for electrical energy generation,
on. These complexes can shift rest po-
provide impetus to the technological de-
tentials or either enhance or diminish
velopment of renewable energy sources.
charge transfer from the semiconductor
Solar energy remains the principal sym-
surface. Because of the limitations on sil-
ver iodide solubility (Ksp (AgI) = 10−16 ) bol of a clean, abundant renewable energy
high (molal) level concentrations of silver source. Society’s energy needs are continu-
will dissolve only in solutions that facili- ous. Unlike purely solid-state, photovoltaic
tate complex formation as in concentrated processes, the advantage of combined
NaI or KI solutions. In particular, we have solar/electrochemical processes is that en-
shown that the presence of high concentra- ergy is not only converted but also may be
tions of dissolved silver is advantageous to stored for future use when sunlight is not
n-tungsten diselenide photoelectrochem- so frequently available.
istry. As seen in Fig. 22, silver dissolved as Limiting constraints of multiple band
AgNO3 enhances the voltage at maximum gap photoelectrochemical energy conver-
point. The extent of the improvement sion, as well as practical configurations for
varies with the initial condition of the efficient solar to electrical energy conver-
individual n-WSe2 crystal, and general im- sion have been probed [1–6]. Such systems
provements are a 15 to 45 mV increase are capable of better matching and utiliz-
in the voltage of maximum power, Vmax , ing of incident solar radiation (insolation).
and a 5 to 20% relative increase in power. Efficient solar cells, solar storage cells,
Illuminated n-Wse2
in 6 m iodide, 0.01 m iodine
1.0
0.8
Relative power
0.6
0.4
0.2
0.0
in Nal in Kl
Without silver Without silver
With 0.7 m silver nitrate With 1.0 m silver nitrate
−0.6 −0.4 −0.2 0.0 −0.6 −0.4 −0.2 0.0

Potential
[V vs Pt]
Fig. 22 The effect of dissolved silver on the relative photopower-voltage
characteristics for illuminated single crystals of n-WSe2 . As indicated on the figure,
four polyiodide electrolytes are used consisting of either: (6 m KI, 0.01 m I2 )
or (6 m KI, 0.01 m I2 , 1.0 m AgNO3 ) or (6 m NaI, 0.01 m I2 ) or (6 m NaI, 0.01 m
I2 , 0.7 m AgNO3 ) as indicated. Relative power is determined by comparison to the
maximum photopower measured in the silver bearing electrolyte.
and solar hydrogen generation systems are in polysulfide, ferrocyanide, polyselenide,

discussed and demonstrated. and polyiodide electrolyte regenerative
Along with conventional parameters PECs.
constraining photovoltaic devices, modifi- The advantage of multiple band gap
cation of the electrolyte (solution phase) semiconductor processes is in the abil-
chemistry is a key to understanding ity to improve the energy match to
the mechanism of energy conversion the solar spectra, and hence improved
and device characteristics of PECs. The utilization/conversion of the incident solar
fundamental importance of modification radiation (insolation). Limiting constraints
of the electrolyte in terms of the of multiple band gap semiconductor/elec-
distribution of species in photoelectrolytes trolyte energy conversion, as well as prac-
and the pragmatic importance in terms tical configurations for solar to electrical
of enhanced solar to electrical conver- energy conversion have been probed. Such
sion efficiency is reiterated by studies combined systems are capable of highly
efficient utilization of solar energy. Effi- 2. S. Licht, O. Khaselev, P. A. Ramakrishnan

cient solar cells, solar storage cells, and et al., J. Phys. Chem. B 1998, 102, 2536.
3. S. Licht, O. Khaselev, P. A. Ramakrishnan
solar hydrogen generation systems are
et al., J. Phys. Chem. B 1998, 102, 2546.
discussed and demonstrated. Alternate 4. S. Licht, B. Wang, T. Soga et al., Appl. Phys.
processes discussed include both an in- Lett. 1999, 74, 4055.
verted (1 photon per e− ) or bipolar (n ≥ 2 5. B. Wang, S. Licht, T. Soga et al., Sol. Energy
photons per e− ) arrangement of succes- Mater. Sol. Cells 2000, 64, 311.
6. S. Licht, H. Tributsch, B. Wang et al., J.
sive band gaps, which can contain either a
Phys. Chem. B 2000, 104, 8920.
Schottky or an ohmic photoelectrochemi- 7. S. Licht, Sol. Energy Mater. Sol. Cells 1995,
cal solution interface, and can drive either 38, 305.
regenerative single, or multiple different, 8. S. Licht, Nature 1987, 330, 148.
redox reactions. The latter case permits so- 9. S. Licht, D. Peramunage, Nature 1990, 345,
330.
lar energy storage or solar water splitting
10. B. Miller, S. Licht, M. E. Orazem et al., Pho-
evolving H2 , and provides an energy reser- toelectrochemical systems, Crit. Rev. Surface
voir which may compensate for the inter- Chem. 1994, 3, 29.
mittent nature of solar energy. Generated 11. C. H. Henry, J. Appl. Phys. 1980, 51, 4494.
H2 is attractive as a clean, renewable fuel. 12. D. J. Friedman, S. R. Kurtz, K. Bertness
et al., Prog. Photovolt. 1995, 3, 47.
Experimental configurations examined
13. J. P. Benner, J. M. Olson, T. J. Coutts, in
include GaAs/Si or AlGaAs/Si photo- Advances in Solar Energy, Chap. 4 (Eds.:
driven redox couples, each with a solar K. W. Boer), American Solar Energy Soci-
to electrical conversion efficiency of 19 to ety, Inc., 1992, 125, Vol. 7.
20%. A related solar cell is configured with 14. M. A. Green, K. Emery, K. Bucher et al.,
Progr. Photovolt. 1996, 4, 321.
electrochemical storage, and which pro-
15. T. Soga, T. Kato, M. Yang et al., J. Appl.
vides a nearly constant energetic output Phys. 1995, 78, 4196.
in illuminated or dark conditions. Simi- 16. J. White, F-R. Fan, A. J. Bard, J. Electrochem.
larly, multiple band gap semiconductors Soc. 1985, 132, 544.
can also be utilized to generate hydro- 17. R. Memming, in Photochemical Conver-
sion and Storage of Solar Energy, (Eds.:
gen fuel by solar-driven water splitting.
E. Pelizzetti, M. Schiavello), Kluwer Acad.
A cell containing illuminated AlGaAs/Si Pub., Netherlands, 1991, pp. 193.
RuO2 /Ptblack evolves H2 and O2 at record 18. S. Licht, G. Hodes, R. Tenne et al., Nature
(18.3%) efficiency. Contemporary models 1987, 330, 148.
underestimate the attainable efficiency of 19. A. Heller, H. J. Lewerenz, B. Miller, J. Am.
solar energy conversion to water splitting, Chem. Soc. 1981, 103, 200.
20. C. Lévy Clément, A. Lagoubi, M. Neumann-
and future multiple band gap systems are Spallart et al., J. Electrochem. Soc. 1991, 138,
calculated as capable of attaining more L69.
than 30% solar efficiency. 21. H. Kobayashi, T. Kubota, N. Toshikawa
et al., J. Electroanal. Chem. 1995, 398, 165.
Acknowledgment 22. S. R. Ovshinsky, M. A. Fetcenko, J. Ross,
Science 1993, 260, 176.
23. G. Lodi, E. Sivieri, A. Debattisti et al., J.
S. Licht is grateful for support by the BMBF Appl. Electrochem. 1978, 8, 135.
Israel-German Cooperation. 24. N. Alonso-Vante, H. Colell, U. Stimming
et al., J. Phys. Chem. 1993, 97, 7381.
References 25. P. Salvador, N. Alonso-Vante, H. Tributsch,
1. S. Licht, O. Khaselev, T. Soga et al., Elec- 26. O. Khaselev, J. Turner, Science 1998, 280,
trochem. Solid State Lett. 1998, 1, 20. 425.
27. J. R. Bolton, S. J. Strickler, J. S. Connolly, 48. S. Licht, J. Phys. Chem. 1986, 90, 1096.
Nature 1985, 316, 495. 49. J. Reichman, M. Russak, J. Electrochem. Soc.
28. M. F. Weber, M. J. Digman, Int. J. Hydrogen 1984, 131, 796.
Energy 1986, 11, 225. 50. K. W. Freeze Jr., Appl. Phys. Lett. 1982, 40,
28a. M. F. Weber, J. Electrochem. Soc. 1984, 131, 275.
1258. 51. H. D. Rubin, B. D. Humphrey, A. B. Bo-
29. M. A. Green, K. Emery, K. Bucher et al., carsley, Nature 1984, 308, 339.
Prog. Photovolt. 1999, 11, 31. 52. H. D. Rubin, D. J. Arent, A. B. Boc-
30. R. C. Kainthla, B. Zelenay, JO’M. Bockris, J. arsley, J. Electrochem. Soc. 1985, 132,
Electrochem. Soc 1986, 133, 248. 523.
31. G. Hodes, J. Manassen, D. Cahen, Nature 53. S. Licht, D. Peramunage, Nature 1991, 354,
1976, 261, 403. 440.
32. A. B. Ellis, S. W. Kaiser, M. S. Wrighton, J. 54. S. Licht, D. Peramunage, J. Electrochem. Soc.
Am. Chem. Soc. 1976, 98, 1635. 1992, 139, L23.
33. B. Miller, A. Heller, Nature 1976, 262, 680. 55. S. Licht, D. Peramunage, J. Electrochem. Soc.
34. S. Licht, G. Hodes, J. Manassen, Inorg. 1992, 139, 1791.
Chem. 1986, 25, 2486. 56. S. Licht, D. Peramunage, Sol. Energy 1994,
35. S. Licht, J. Manassen, J. Electrochem. Soc. 52, 197.
1987, 134, 918. 57. G. Seshardi, J. K. M. Chun, A. B. Bocarsly,
36. S. Licht, J. Electrochem. Soc. 1988, 135, 2971. Nature 1991, 352, 508.
37. F. Forouzan, K. Longo, D. Peramunage 58. B. A. Parkinson, A. Heller, B. Miller, J. Elec-
et al., Anal. Chem. 1990, 62, 1356. trochem. Soc. 1979, 126, 954.
38. K. Longo, F. Farouzan, D. Peramunage 59. B. J. Tufts, I. L. Abrahams, P. G. Santengelo
et al., J. Electroanal. Chem. Interfacial Elec- et al., Nature 1987, 326, 861.
trochem. 1991, 318, 111. 60. L. G. Casasgrand, B. J. Tufts, N. S. Lewis, J.
39. D. Peramunage, F. Forouzan, S. Licht, Anal. Phys. Chem. 1991, 95, 1373.
Chem. 1994, 66, 378. 61. D. E. Levy, R. J. Myers, J. Phys. Chem. 1990,
40. W. Giggenbach, Inorg. Chem. 1971, 10, 94, 7842.
1333. 62. S. Licht, F. Forouzan, J. Electrochem. Soc.
41. A. Teder, Sr. Papperstodn 1969, 72, 245. 1995, 142, 1539.
42. S. Licht, Joost Manassen, J. Electrochem. 63. S. Licht, F. Forouzan, J. Electrochem. Soc.
Soc. 1985, 132, 1077. 1995, 142, 1546.
43. S. Licht, G. Hodes, Joost Manassen, J. 64. W. Kautek, H. Gerischer, H. Tributsch, J.
Electrochem. Soc. 1986, 133, 272. Electrochem. Soc. 1981, 128, 2471.
44. S. Licht, J. Manassen, J. Electrochem. Soc. 65. B. A. Parkinson, A. Heller, B. Miller, Appl.
1986, 133, 277. Phys. Lett. 1978, 33, 521.
45. S. Licht, R. Tenne, H. Flaisher et al., J. 66. R. Tenne, A. Wold, Appl. Phys. Lett. 1985,
Electrochem. Soc. 1986, 133, 52. 47, 707.
46. S. Licht, R. Tenne, H. Flaisher et al., J. 67. S. Licht, N. Myung, J. Electrochem. Soc.
Electrochem. Soc. 1984, 131, 5920. 1985, 142, 845.
47. S. Licht, R. Tenne, G. Dagan et al., Appl. 68. S. Licht, N. Myung, J. Electrochem. Soc.
Phys. Lett. 1985, 46, 608. 1995, 142, L129.
397
5.1 5.1.2
Dye-Sensitized Regenerative Solar Cells Device Concepts
..
Augustin J. McEvoy and Michael Gratzel It should be remembered that the first
Ecole Polytechnique Fédérale de Lausanne, observation of the photovoltaic effect
Lausanne, Switzerland by A. -E. Becquerel [1] in 1839 was at
a solid–liquid interface, with a semi-
5.1.1 conductor photoelectrode contacting an
Overview electrolyte, and was technically a pho-
toelectrochemical cell. Nevertheless, the
At present, dye-sensitized photosystems modern photovoltaic industry is estab-
provide the only technically and econom- lished exclusively on solid-state devices,
ically credible alternative to solid-state with semiconductors of appropriate band
photovoltaic devices. The concept recon- gap as optical absorbers and junctions to
ciles the electrochemical stability of wide solids with different conduction mecha-
band gap photoelectrodes with the ef- nisms as the sites of charge separation and
ficiency of photovoltaic devices having photovoltage generation. In commerce are
a more extended absorption spectrum found homojunction devices, in which the
in the visible range. Optical absorption semiconductors are chemically identical,
and charge separation take place on dis- differing only in impurity content and
therefore in free carrier polarity; hetero-
tinct sites within these photovoltaic cells.
junctions, in which the semiconductors
Hence, oppositely charged species are re-
of opposite carrier polarity are chemically
stricted to separate phases, so conventional
different, for example, cadmium sulfide or
recombination losses are inhibited. In copper indium-gallium selenide contacts;
consequence, device photoconversion ef- and finally Schottky junctions in which
ficiency is better maintained at low light the contacting phase is a metal. However,
levels than with conventional semicon- all these are solid-state devices in which
ductor solid-state junction devices. Vari- conduction throughout is by an electronic
ous configurations of the dye-sensitized mechanism. None is a photoelectrochem-
concept are under investigation, includ- ical device that has an electrolyte as the
ing electrochemical and heterojunction contacting phase or that uses an ionic
variants. conduction mechanism. Awareness of
398 5 Dye-Sensitized Photoelectrochemistry
photoelectrochemistry in the photovoltaic The realization in practice of this option

world therefore remains marginal, al- is simply presented. Analogous to the
though this awareness is rapidly grow- band gap of the semiconductor is the
ing. Intensive research over the past two HOMO-LUMO (highest occupied molec-
decades has led to an inescapable con- ular orbital ⇒ valence band/lowest un-
clusion that those semiconductors, whose occupied molecular orbital ⇒ conduction
band gaps are sufficiently wide to give a band) gap of a molecule. When the lat-
long-term stability under illumination, in ter lies above the conduction band edge
contact with an electrolyte and a techni- of a semiconductor substrate, an electron
cally useful photovoltage are insensitive to entering the LUMO on photoexcitation of
visible light, requiring higher-energy ul- the dye by a photon within its absorption
traviolet photons to excite charge carrier spectrum energy range can be injected into
pairs. A narrower band gap, compatible the semiconductor. The positively charged
with the photoconversion of visible light, oxidized dye molecule can then be neu-
is indicative of weaker chemical bonding tralized and its ground state restored by
of the semiconductor and hence liability to reaction and charge exchange with the
photocorrosion and incompatibility with electrolyte, which in turn receives an elec-
a stable extended lifetime as an energy- tron from a metallized counterelectrode.
conversion device. The resolution of this With a closed external circuit and under
dilemma lies in the separation of light ab- illumination, the device then constitutes
sorption and charge-separation functions a photovoltaic energy-conversion system,
by sub–band gap sensitization of the semi- which is regenerative and stable, and is
conductor with an electroactive dye. functionally comparable to its solid-state
Conducting
glass
TiO2 Dye Electrolyte Cathode
E /V vs NHE
Injection
S*
−0.5
Maximum
0 voltage
hν
0.5 Red ox
Mediator
Interception
Diffusion
1.0 S°/S+
Fig. 1 Schematic of the structure and function of the dye-sensitized

electrochemical photovoltaic cell. The nanoporous sensitized
semiconductor photoanode receives electrons from the photoexcited
dye, which is thereby oxidized and which in turn oxidizes the mediator, a
redox species dissolved in the electrolyte. The mediator is regenerated
by reduction at the metallic cathode, while the electrons circulate in the
external circuit.
5.1 Dye-Sensitized Regenerative Solar Cells 399
junction analog. A typical wide band gap was not realistic of the visual impres-
semiconductor for photoelectrochemistry sion of the same scene. So many early
with stability against photocorrosion is photographs therefore give an impres-
titanium dioxide. Titanium dioxide has a sion of an unnatural ‘‘cleanliness’’ of
band gap of 3.1 eV and is therefore in- the scene imaged that considerable ef-
sensitive to visible light. The selected dye forts were expended to find emulsions
must have a HOMO-LUMO gap corre- capable of more realistic response. The
sponding to a photon energy in the visible problem is now recognized to be due to
or infrared parts of the spectrum, with an the semiconductor nature of the halide
optimum near the solar spectra maximum grains responsible for optical absorption
of 1.4 eV. This mechanism is illustrated in photography, which are activated es-
schematically in Fig. 1. However, before sentially by the photoelectric effect. They
presenting the relevant characteristics of have band gaps typically in the range 2.7
a semiconductor electrode that is apt for to 3.2 eV, and the photoresponse is ef-
chemisorption of a dye and therefore pho- fectively negligible for wavelengths longer
tovoltaic application, a background of the than 460 nm. It was however noted that the
history and development of the sensitiza- use of gelatine as a support medium for
tion of semiconductors is necessary. the alkali halide grains in the photographic
emulsion could significantly modify the
5.1.3 film spectral response. Only in the last
Photography and Photoelectrochemistry century was it finally demonstrated that
an organo-sulfur compound present in
Although artisanal ‘‘virtual reality’’ by calf-skin gelatine was responsible for this
painting is as old as human culture, extension of sensitivity [3]. The mecha-
synthetic ‘‘virtual reality’’ by technical nism now recognized is the induction
reproduction of images is relatively recent, of a nanostructure of silver sulphide on
with the first photography by Daguerre the halide surface of each grain, giving
in 1837 and the introduction of the sil- the first application of the sensitization
ver halide process by Fox-Talbot in 1839. of a semiconductor heterojunction! Even
Thereafter, the art advanced so rapidly more significant in the history of sensiti-
that the nineteenth century events – the zation is the work of Vogel, Professor of
wars, the explorations, the early factory age photochemistry, spectroscopy and photog-
and the introduction of steam power – are raphy at the then Königliche Technische
vividly documented by photographs. Pho- Hochschule of Berlin, who in the 1870s in-
tography remained an art for a century, vestigated the effect of dyes on emulsions
becoming a science only with the theo- and found that the halide spectral response
retical analysis of the process by Gurney extended through the visible and even into
and Mott [2] in 1938. Nevertheless, an the infrared region [4]. Not only did this
empirical skill in the formulation of pho- facilitate his spectroscopic research and
tographic emulsions had evolved in the contribute to astronomy but also made
interval, even in the absence of a funda- possible the commercial ‘‘panchromatic’’
mental understanding. Early emulsions, broad-spectrum black-and-white film with
referred to as ‘‘orthochromic,’’ had a dis- realistic gray scale reproduction, and later
tinct insensitivity to midspectral and red with spectrally selective dyes – the modern
light, so that their gray scale reproduction era of color photography.
Meanwhile, the photoelectric effect was This parallel evolution of sensitization in

a focus of scientific interest, with a photography and in photoelectrochemistry
history synchronized with that of pho- still seems to surprise each generation of
tography. The concept of dye enhance- photochemists. That the same dyes were
ment of the photoeffect was carried effective for both processes was recognized
over from photography to the sensiti- among others by Namba and Hishiki [7] at
zation of an electrode already in 1887 the 1964 International Conference on Pho-
by Moser [5] (Fig. 2) using the dye ery- tosensitization in Solids, which can now be
throsin on silver halide electrodes and seen as a seminal event in the history of
confirmed by Rigollot in 1893 [6]. In the dyes in photochemistry. It was also recog-
retrospectively quaint report, written sub- nized then that the dye should be adsorbed
stantially before the Einstein theory of on the semiconductor surface as a closely
the photoelectric effect, Moser records packed monolayer for maximum sensitiza-
his observations on dye-induced enhance- tion effectiveness [8]. The conference also
ment as an increased photopotential (V ) provided the venue for the identification
rather than the more fundamental cur- of the sensitization mechanism through
rent (A), despite the title – ‘‘Strengthened charge transfer from a dye excited state;
Photoelectric Current Through Optical until then, the matter was in dispute,
Sensitization.’’ with an energy-coupling process being
Fig. 2A copy of the first paper on sensitization published by

James Moser in 1887.
Fig. 3 The dilemma of 1968 – energy Vacuum Level 0ev

resonance or electron injection?
I I
proposed, as illustrated in Fig. 3 taken
from the conference proceedings [9]. The Conduction band e
understanding became definitive with the
hν
work of Gerischer and Tributsch on sen- D
Br −
sitization of zinc oxide [10]. However, the Br+ Dye
fact remained that sensitization efficiency hν
remained low, as research systematically
investigated dyes in solution or those Dye
adsorbed on single-crystal substrates. In Valence band
Energy Electron
BrAg
retrospect, it appears that there was an transfer transfer
excessive influence of solid-state engineer-
ing practice in that the disordered surfaces modifications of certain side groups.
were expected to promote surface recombi- Chlorophyll is in turn similar in struc-
nation and therefore loss of excited charge ture to hemoglobin, the oxygen-carrying
carriers, whereas single-crystal surfaces, iron-based pigment found in blood. Given
being better defined physically and chem- that the development of the dye-sensitized
ically, make the effects of sensitization cell arose out of an interest in artificial pho-
more evident. The results of this presump- tosynthesis, the adoption of porphyrin-like
tion were slow to emerge. organometallic dyes as sensitizers was log-
ical. However, although nature confines
5.1.4 itself to magnesium and iron for its princi-
Sensitization of Powders and Rough pal pigments, the synthetic chemist can
Surfaces access the whole range of metallic el-
ements. The use of ruthenium pyridyl
complexes now has more than twenty years
With the ongoing interest in sensitization
of development history. Clark and Sutin
during the 1970s, and the idea that
had already used a trisbipyridyl ruthenium
photocatalysts could possibly carry out
complex in 1977 to sensitize titanium diox-
electrolytic reactions such as the evolution
ide to sub–band gap illumination [12], but
of hydrogen from water, a phase of experi-
in solution only. Charge transfer could
mental development evolved with the sen- only occur after diffusion of the ion to the
sitization of semiconductor powders [11]. semiconductor, so the efficiency of the sen-
With the ecological awareness of time, sitization was very low. By 1980, the idea of
photochemical production particularly of chemisorption of the dye through an acid
energy-related chemicals was seen as a carboxylate group bonding to the metal ox-
form of ‘‘artificial photosynthesis.’’ The ide surface had also been established [11]
prototype energy-converting dye provided so that the sensitizer was immobilized
by nature is indeed chlorophyll, a molecule and it formed the required monomolecular
consisting of a central magnesium atom film on the semiconductor substrate which
surrounded by a nitrogen-containing por- facilitated charge transfer by electron in-
phyrin ring. Variations are due to minor jection. The carboxylated trisbipyridyl dye
(‘‘RuL3 ’’) therefore became the prototype of nanoporous ceramic films is the basis
sensitizer in photochemistry. The subse- for the understanding of semiconductors.
quent development of sensitizers through The nanoporous structure permits the
to the panchromatic ‘‘black dye’’ that con- specific surface concentration of the sen-
stitutes the current state of the art is dealt sitizing dye to be sufficiently high for the
with in detail in a separate chapter. total absorption of the incident light that is
As a substrate for the chemisorption necessary for efficient solar energy conver-
of sensitizing dyes, titania, TiO2 , par- sion, since the area of the monomolecular
ticularly in the anatase form, gradually distribution of adsorbate is 2 to 3 orders of
became dominant [13]. It has many advan- magnitude higher than the geometric area
tages for sensitized photochemistry and of the substrate. Moreover, in the elec-
photoelectrochemistry, being a low-cost, trochemical system, this high degree of
widely available, nontoxic and biocompat- roughness does not promote charge carrier
ible material, and as such is even used loss by recombination, since the electron
in health care products and domestic
and the positive charge find themselves
applications such as paint pigmentation.
within picoseconds on opposite sides of
The objective had also evolved to concen-
the liquid–solid interface. In addition, the
trate on photovoltaic devices rather than
photoexcited carrier in the n-type TiO2 is
on photosynthesis. Progress thereafter was
an electron, a majority carrier, and there-
incremental until the announcement in
fore not subject to depletion by carriers of
1991 [14] of the sensitized electrochemical
the opposite polarity. Solid-state junction
solar cell, with the remarkable conversion
efficiency at that time of 7.1% under so- cells, on the other hand, are minority car-
lar illumination, a synergy of structure, rier devices. The carrier loss mechanisms
substrate roughness, dye photochemistry, are comparatively slow [16], and although
counterelectrode kinetics and electrolyte conventionally referred to as recombina-
redox chemistry. That evolution has con- tion by analogy with the solid-state process,
tinued progressively since then, with certi- the loss of a photoexcited electron from the
fied efficiency now more than 10% [15]. semiconductor should rather be regarded
There has been a continuous develop- as a recapture of the electron by an oxidized
ment of the electrolyte solvent, aqueous dye species, or a redox capture, when the
systems being eliminated in favor of or- electron reacts directly with the iodine in
ganics and molten salts. However, finding the electrolyte. Either can occur on a mil-
substitutes for the original iodine-iodide lisecond timescale, but much slower than
redox system has proved to be difficult electron injection or transport to the rear
although recent work in our laboratory contact layer.
shows that alternative redox couples are The original substrate structure used
finally at hand. for our early photosensitization experi-
ments was a fractal derived by hydrolysis
5.1.5 of an organo-titanium compound but it
Substrate Development and Fabrication has since been replaced with a nanostruc-
ture deposited from colloidal suspension.
If molecular design and engineering has This evidently provides a much more re-
underpinned the evolution of efficient, producible and controlled porous high
stable sensitizer dyes, the materials science surface area nanotexture. Further, since
it is compatible with screen-printing tech- the electron loss by exchange with the re-
nology, it anticipates future production dox system in solution. In the analogous
requirements. nonliquid electrolyte case, it is not im-
While commercially available titania mediately evident that an interpenetrating
powders, produced by a pyrolysis route network of two conducting solids can be
from a chloride precursor, have been suc- so easily established that an immobilized
cessfully employed, the present optimized molecule at their interface can exchange
material is the result of a procedure de- charge carriers with both. However, ini-
scribed by Brooks and coworkers [17]. A tial results [18,19] are promising. In both
specific advantage of the hydrothermal cases, the charge transport materials are
technique is the ease of control of the par- deposited by spin coating from the liquid
ticle size and hence of the nanostructure phase in order to achieve the necessary
and porosity of the resultant semiconduc- intimate contact. In the latter case, a mix-
tor substrate. The relevant preparation flow ture of polymers was used, which phase-
diagram is given and the product is illus- separates spontaneously on removal of a
trated by the accompanying micrograph solvent, whereas the Lausanne laboratory
(Fig. 4; Table 1). Figure 5 presents data on introduces a solution of the conducting
the control of substrate porosity by the compound into a previously sensitized
powder-preparation parameters. nanostructure. The charge transfer ma-
terial currently used is a spirobifluorene,
5.1.6 proprietary to Hoechst [20] as shown in
Dye Sensitization in Heterojunctions Fig. 6. As a matter of technical precision, if
these materials function in the cell as a hole
Because the sensitizing dye itself does
conductor, the device is a sensitized nanos-
not provide a conducting functionality
tructured heterojunction in which the
but is distributed at an interface in the
charge transport is entirely of electronic
form of immobilized molecular species,
nature (Fig. 7). However, if the molecules
it is evident that for charge transfer,
accept positive charge to become cations,
each molecule must be in intimate con-
for which there is initial spectroscopic evi-
tact with both conducting phases. It is
dence [18], the charge transfer mechanism
clear that this applies to the porous
within this organic phase can be consid-
wide band gap semiconductor substrate
ered a redox equilibration, and the device
into which the photoexcited chemisorbed
molecules inject electrons. It is also ev-
ident that in the photoelectrochemical
form of the sensitized cell, the liquid
electrolyte penetrates into the porosity,
thereby permitting the intimate contact
with the charged dye molecule that is
necessary for charge neutralization after
Fig. 4 SEM image of the surface of a

mesoporous film prepared from the 100 nm
hydrothermal TiO2 colloid.
Tab. 1 Flow diagram for the preparation of TiO2 colloids and mesoporous films [17]
Precursor preparation Modify Ti-isopropoxide with acetic acid
Hydrolysis Rapidly add precursor to water
Peptization Acidify with HNO3, reflux
Hydrothermal growth Autoclave: 12 hours at 230 °C
Concentrate colloid Rotovap: 45 °C, 30 mbar
Solvent exchange, Flocculation, centrifuging

Ethyl cellulose addition
Homogenize paste 3-roll mill, 15 minutes
Screen-print films
Dry and anneal films Anneal: 450 °C, 20 minutes
9000
8000 190 °C
210 °C
7000 230 °C
[(cm3/g)/nm] [10−5]
Differential volume
250 °C
6000 270 °C
5000
4000
3000
2000
1000
0
0 10 20 30 40 50 60 70 80
Pore diameter
[nm]
Fig. 5 Control of semiconductor substrate porosity by temperature of
hydrothermal processing.
OCH3
N
OCH3
OCH3
OCH3
N
N OCH3
OCH3 N
OCH3
OCH3
Fig. 6 Structure of spirobifluorene conducting material. The
bisfluorene structures are perpendicular, conjoined through a
carbon site common to both.
is photoelectrochemical. This distinction cells, in contrast, fabrication procedures

may have practical implications. require development and costing based on
estimates of the requirements of chemical
5.1.7 processes rather than the silicon met-
Commercial Prospects allurgy with elevated temperatures and
vacuum technology required for conven-
The status of the dye-sensitized device as tional cells. Equally, it is well known that
the only verified alternative to solid-state the substitution of an established technol-
junction devices has already been dis- ogy by an upcoming alternative requires
cussed. It must be recognized that the that the new concept has definite ad-
solid-state devices, particularly the sili- vantages and no clear disadvantages. It
con p-n junction cells benefit from forty is therefore noted with some satisfaction
years of industrial and development ex- that several companies in Europe, Japan,
perience, the technology transfer from and Australia have taken up the challenge
the silicon-based electronics industry, and and are currently engaged under license
even the widespread availability of high-
quality silicon at low cost resulting from
the expansion of that industry. The proce-
dures for high-yield fabrication of silicon
devices, both crystalline and amorphous, cb Inj.(D + /D*)
are well understood, with costing well
established on the basis of decades of solid
industrial experience. For dye-sensitized hν
Reg. Hopping
Fig. 7 Charge transport by hopping in (D+ /D)

the dye-sensitized solid-state
TiO2 Dye Au
photovoltaic cell.
in the venture of industrialization and 2. R. W. Gurney, N. F. Mott, Proc. Roy. Soc. A

commercialization of dye-sensitized pho- 1938, 164, 151.
3. T. H. James, in The Theory of the Photographic
tovoltaic cells. Process, 4th. Ed., MacMillan, New York,
5.1.8 1977.
4. W. West, Proc. Vogel Centenary Symp., Pho-
Conclusion
togr. Sci. Eng. 1974, 18, 35.
5. J. Moser, Monatsh. Chem. 1887, 8, 373.
The development of reproducible and 6. H. Rigollot, C. R. Acad. Sci. 1893, 116,
stable photovoltaic devices adapted for 873.
manufacturing processes has proceeded in 7. S. Namba, Y. Hishiki, J. Phys. Chem. 1965,
evolutionary steps, with each component 69, 774.
optimized and verified for compatibility 8. R. C. Nelson, J. Phys. Chem. 1965, 69,
714.
with system requirements. The dye-sensi- 9. J. Bourdon, J. Phys. Chem. 1965, 69, 705.
tized nanocrystalline solar cell, either the 10. H. Gerischer, H. Tributsch, Ber. Bunsen-Ges.
electrochemical device or the closely re- Phys. Chem. 1968, 72, 437.
lated sensitized heterojunction [18], pro- 11. M. P. Dare-Edwards, J. B. Goodenough, A.
vides a credible alternative to the conven- Hamnet et al., Faraday Discuss. Chem. Soc.
1980, 70, 285.
tional semiconductor junction solid-state 12. W. D. K. Clark, N. Sutin, J. Am. Chem. Soc.
cell. Time and the market will tell if it can 1977, 99, 4676.
compete successfully. 13. J. DeSilvestro, M. Grätzel, L. Kavan et al.,
J. Am. Chem. Soc. 1985, 107, 2988.
Acknowledgments 14. B. O’Regan, M. Grätzel, Nature 1991, 335,
737.
15. M. K. Nazeeruddin, P. Pechy, Th. Renouard
The team at EPFL greatly appreciates the et al., J. Am. Chem Soc. 2001, 123, 1613.
vote of confidence represented by the 16. S. A. Hague, Y. Tachibana, R. Willis, J. E.
licenses taken up by the industry for this Moser, M. Grätzel, D. R. Klug, J. R. Durrant,
type of solar cell. In the present work, we do J. Phys. Chem. B 2000, 104, 538.
acknowledge the initiative and innovative 17. K. G. Brooks, S. D. Burnside, V. Shklover
et al., Ceramic Transactions 109: Processing
spirit of our coworkers, past and present, and Characterisation of Electrochemical Mate-
in the areas of dye synthesis, catalysis and rials and Devices, (Eds.: P. N. Kumta), Ameri-
electrochemistry, and semiconductor thin can Ceramic Society, 2000, p. 115.
film fabrication. 18. U. Bach, D. Lupo, P. Comte et al., Nature
1998, 395, 583.
References 19. J. J. M. Halls, C. A. Walsh, N. C. Greenham
et al., Nature 1995, 376, 498.
1. A. -E. Becquerel, C. R. Acad. Sci. 1839, 9, 20. D. Lupo, J. Salbeck, Intern. patent PCT/
561. EP96/03944.
5.2 Dyes for Semiconductor Sensitization 407
5.2 dye molecules, which absorb light, free-

Dyes for Semiconductor Sensitization ing electrons to carry electric current.
.. The low production costs, stability, and
Md. Khaja Nazeeruddin and Michael Gratzel the efficiency of these cells has gener-
Swiss Federal Institute of Technology, Lau-
ated renewed interest in many groups
sanne, Switzerland
around the world [6–15]. The added ad-
5.2.1 vantages of these cells are availability
Introduction and nontoxicity of the main component
that is TiO2 , which is even used in
5.2.1.1 General Background paints, cosmetics, and health care pro-
The extensive use of conventional energy ducts like hip joints and other orthopedic
supplies is causing many serious problems implants.
to the global environment. Moreover, Efficient sensitization of large band gap
sooner or later the fossil fuels are going semiconductors to the visible and the
to be exhausted and if nuclear power is near infrared (IR) solar spectrum can be
not to be depended on solely, alternative achieved by engineering dyes at a molecu-
methods for harnessing solar power, lar level, which is the topic of this chapter.
which is clean, nonhazardous, enormous, A dye-derivatized mesoporous titanium
film is one of the key components in dye-
and ever lasting have to be developed.
sensitized solar cells. The electrochemical
There are numerous ways to convert the
and photophysical properties of the ground
solar radiation directly into electrical power
and the excited state of the dye play an im-
or chemical fuel [1]. However, the capital
portant role in the charge-transfer (CT)
cost of such devices is not attractive for
dynamics at the semiconductor interface
large-scale applications.
(Chapter 5.3 in Volume 6).
Dye-sensitized solar cell technology is an
The photophysical and photochemical
interesting, inexpensive, and promising al-
properties of group VIII metal complexes
ternative to those of traditional solid-state using polypyridyl ligands have been thor-
photovoltaics. The process of extending oughly investigated during the last three
the sensitivity of transparent materials decades [16–20]. The main thrust behind
such as titanium dioxide (TiO2 ) to the these studies is to understand the energy
visible spectra is called spectral sensitiza- and electron-transfer processes in the ex-
tion. The topic of sensitization is very cited state and to apply this knowledge
old [2–5] and without going into the his- to potential practical applications such as
tory of the sensitization, the progress made dye-sensitized solar cells and light-driven
during last 10 years on dye-sensitized information processing [21].
solar cells is reviewed. The intention Particularly, ruthenium (Ru) (II) com-
here is to restrict the discussion ex- plexes have been used extensively as
clusively to inorganic dyes (transition CT sensitizers on nanocrystalline TiO2
metal complexes) based largely on our films [22]. The choice of ruthenium metal
own work. is of special interest for a number of rea-
The photoelectrode of a dye-sensitized sons: (1) because of its octahedral geome-
solar cell consists of 8–20 µm film of try, one can introduce specific ligands in a
mesoporous nanocrystalline TiO2 parti- controlled manner; (2) the photophysical,
cles containing a monolayer of anchored photochemical, and the electrochemical
properties of these complexes can be tuned However, the photophysical and the
in a predictable way; (3) it possesses sta- electrochemical properties of its coordi-
ble and accessible oxidation states from I nation complexes are difficult to tune in
to IV; (4) it forms very inert bonds with an expected fashion [25]. The other no-
imine nitrogen centers [23, 24]. table disadvantage of this metal is weaker
Iron (Fe) that is in the first row ligand field splitting compared to ruthe-
transition metal of the periodic table, nium and osmium (Os). On the other
and the same group as ruthenium, hand, osmium being in the third row tran-
is an inexpensive and abundant metal. sition metal of the periodic table has a
Load
e− e−
S+/S∗
e− e−
−1.0
cb
e−
e− e− (e−)
−0.6
e−
∆v −0.2
0
(R / R−)
TiO2 0.2
e−
Electrolyte
0.6
S+/S
1.0
TCO TiO2 Counter electrode V vs SCE

COOH
HOOC
N
N NCS
= Ru
Anchored dye N NCS
N
HOOC
HOOC
Fig. 1 Operating principles and energy level diagram of dye-sensitized
solar cell. S/S+ = Sensitizer in the ground and oxidized state;
S∗ /S+ = Sensitizer in the excited state R/R− = Redox mediator.
stronger ligand field splitting compared conduction band and the redox couple
to ruthenium. Moreover, the spin-orbit because of differences in recombination
coupling in osmium complexes leads to rates [28].
enhanced spectral response in the red re-
gion [26]. Nevertheless, the low abundance IPCE =
of this metal restricts its use for large-scale photocurrent density
(1.25 × 103 ) × −2
applications. (mA cm )
photon flux
wavelength (nm) × −2
(W m )
5.2.1.2 Operating Principles of the
Dye-Sensitized Solar Cell (1)
The details of the operating principles of
the dye-sensitized solar cell are given in 5.2.2
Molecular Sensitizers
Fig. 1. The photo-excitation of the metal-
to-ligand charge transfer (MLCT) of the
5.2.2.1 Formation of Complexes
adsorbed dye leads to injection of elec-
One can view the formation of a complex
trons into the conduction band of the
between metal ion and ligands as an
oxide. The original state of the dye is sub-
electrostatic attraction between positively
sequently restored by electron donation
charged metal and negatively charged ions
from an electrolyte, containing the re-
or the negative ends of the dipoles of
dox system (iodide/triiodide). The injected
neutral ligands. A metal ion or atom can
electron flows through the semiconduc-
act as a discrete center about which a set of
tor network to arrive at the back contact
ligands is arranged in a definite way. The
and then through the external load to the
general concept of a ligand (unidentate) is
counter electrode. At the counter electrode,
donating of an electron pair to the central
reduction of triiodide in turn regenerates
metal atom. The number of ligands per
iodide, which completes the circuit.
metal center is generally either four or six,
with others being rare. The approach of the
5.2.1.3 Incident Photon to Current negative, or the neutral, ligands toward a
Efficiency and Open-Circuit Photovoltage charged metal repels electrons residing in
The incident monochromatic photon-to- d-orbitals and raises their energies both
current conversion efficiency (IPCE), de- with respect to those residing on the
fined as the number of electrons generated ligands and the metal.
by light in the external circuit divided In an octahedral ligand field, the five
by the number of incident photons as a degenerate d-orbitals split into degenerate
function of excitation wavelength is ex- t2g (dxy , dxz , dyz ), and eg (dx2 −y2 , dz2 ) sets
pressed in the Eq. (1) [27]. The open-circuit of orbitals. The splitting value increases
photovoltage is determined by the en- by about 40% as one moves from the first
ergy difference between the fermi level row (3d) to the second (4d) and third (5d)
of the solid under illumination and the row transition metal ions [29]. A schematic
Nernst potential of the redox couple in the representation for splitting of d-orbital
electrolyte (Fig. 1). However, the experi- and the position of ligand orbitals in
mentally observed open-circuit potential an octahedral complex containing three
using various transition metal complexes bidentate 2,2$ -bipyridyl ligands is shown
is smaller than the difference between the in Fig. 2.
π*
π* π*
Energy
eg
∆Oh ∆Oh ∆Oh
d t2g
Metal orbitals Fe2+ Ru2+ Os2+

Fig. 2 Fe2+ ,
A d-orbital splitting pattern for Ru2+ and Os2+
in an octahedral
field. The relative eg levels with respect to t2g and ligand π ∗ -orbitals are shown
arbitrarily.
5.2.2.2 Photophysical Properties is called intersystem crossing (ISC). The

The photophysics and photochemistry of possible deexcitation processes are radia-
polypyridyl complexes of ruthenium can tive and nonradiative. The radiative decay
be understood with the aid of the en- of a singlet and triplet excited states
ergy diagram shown in Fig. 3. In these are termed fluorescence and phosphores-
complexes there are three possible excited cence, respectively.
states: (1) metal centered (MC), which The intense colors in 2,2$ bipyridyl com-
are due to promotion of an electron plexes of iron, ruthenium, and osmium
from t2g to eg orbitals; (2) ligand cen- are due to excitation of an electron from
tered (LC) that are π –π ∗ transitions; and metal t2g orbitals to the empty π ∗ –orbitals
(3) CT excited states, which are either of the conjugated 2,2$ bipyridyl. The photo-
MLCT or ligand-to-metal (LMCT). An elec- excitation of this MLCT excited state
tronic transition from metal t2g orbitals can lead to emission. However, not all
to empty ligand orbitals without spin- complexes are luminescent because of
change allowed, is called a singlet-singlet the different competing deactivation path-
optical transition. The allowed transitions ways. This aspect is beyond the scope of
are identified by large extinction coeffi- this chapter; the interested reader can re-
cients. The transitions with spin change fer to a number of publications on this
are termed singlet-triplet optical transi- subject [16–20]. The other potential deac-
tions, which are forbidden and are usually tivation pathways for the excited dye are
associated with a small extinction coeffi- donation of an electron (called oxidative
cient. However, the excited singlet state quenching, Eq. 2) or the capture of an
may also undergo a spin flip, resulting electron (reductive quenching, Eq. 3) or
in an excited triplet state. This process transfer of its energy to other molecules or
MC
eg
1
p LC MLCT
π*
ISC
s
3MLCT
d Em
t2g
LMCT
σ
Metal orbitals Molecular orbitals Ligand orbitals
Fig. 3 Schematic presentation of a molecular orbital diagram for an
octahedral d6 metal complex involving 2,2$ -bipyridyl type ligands in which
various possible transitions are indicated.
solvent (Eq. 4). process.
E(S∗ /S− ) = E(S/S− ) + E0 – 0 (5)

D∗ + Q −−−→ D+ + Q− (2)
E(S+ /S∗ ) = E(S+ /S) − E0 – 0 (6)
D∗ + Q −−−→ D− + Q+ (3)
D∗ + Q −−−→ D + Q∗ (4) 5.2.2.4 Requirements of the Sensitizers
The ideal sensitizer for a single junction
5.2.2.3 Ground and Excited State Redox photovoltaic cell should absorb all the
Potentials solar photons below a wavelength of about
The ground state oxidation and reduction 920 nm [31]. In addition, it should fulfill
potentials of a complex can be obtained several demanding requirements: (1) It
by cyclic voltammetric studies. An ap- should possess suitable excited state redox
proximate value of the excited state redox properties; (2) It must firmly be grafted to
potential can be extracted from the po- the semiconductor oxide surface and inject
tentials of the ground state couples and electrons into the conduction band with
the zero-zero excitation energy (E0 – 0 ) ac- a quantum yield of unity; (3) Its ground
cording to Eqs. (5) and (6). The zero-zero state redox potential should be sufficiently
energy can be obtained from 77 K emis- high that it can be regenerated rapidly
sion spectrum of the sensitizer [30]. The via electron donation from the electrolyte
excited state redox potential of a sensitizer or a hole conductor, (4) It should have a
plays an important role in the sensitization long-term photo and thermal stability to
sustain at least 108 redox turnovers under of ruthenium polypyridyl complexes can
illumination corresponding to about 20 be tuned in two ways. First, by introducing
years of exposure to natural sunlight at Air a ligand with a low-lying π ∗ -molecular or-
Mass (AM) 1.5. bital and second, by the destabilization of
On the basis of extensive screening the metal t2g orbitals with a strong donor
of hundreds of ruthenium complexes, ligand. The limiting conditions for an effi-
it was discovered that the sensitizer ex- cient sensitizer makes these two extreme
cited state oxidation potential should be at situations incompatible. Because, in the
least – 0.9 V versus Saturated calomel ref- former case, the excited state is not suffi-
erence electrode (SCE), to inject electrons ciently energetic to enable electron transfer
efficiently onto TiO2 conduction band. The into the conduction band of the TiO2 semi-
ground state oxidation potential should be conductor, and in the latter case, the easily
about 0.5 V versus SCE, to be regener- oxidizable complex cannot be reduced back
ated rapidly via electron donation from the by a suitable electron relay like I− .
electrolyte (iodide/triiodide redox system) To illustrate the tuning aspects of
or a hole conductor. A significant decrease the MLCT transitions in ruthenium
in electron injection efficiencies will oc- polypyridyl complexes, the well known
cur if the excited and ground state redox [RuL3 ] (L = 4,4$ dicarboxylic acid-2,2$ -bi-
potentials are lower than these values [32]. pyridine) type of complex can be consid-
ered. This complex shows strong visible
5.2.2.5 Absorption Spectral Properties of band at 466 nm, because of CT tran-
Metal Complexes sition from metal t2g highest occupied
molecular orbitals (HOMO) to π ∗ -lowest
5.2.2.5.1 Tuning of MLCT Transitions To unoccupied molecular orbitals (LUMO) of
absorb the light below 920 nm, the sensi- the ligand (Fig. 3). The Ru(II)/(III) oxida-
tizer’s redox level needs to be tuned. The tion potential is at 1.3 V, and the ligand
MLCT transitions and the redox properties based reduction potential is at −1.5 V
Ligand π *t
un
π*orbitals ing
∆E
Energy
∆E ∆E
ing
tun
t 2g
Metal
t2g orbitals
[Ru(dcbpy)3] [Ru(dcbiq)3] [Ru(dcbpy)2(phpy)]
Fig. 4 Tuning of HOMO (t2g ) and LUMO (π ∗ ) orbital energy in various

ruthenium polypyridyl complexes.
versus SCE [33]. From spectro chemical are tuned from 466 nm to 560 nm by
and electrochemical studies of polypyridyl introducing different donor ligands. The
complexes of ruthenium, it has been con- oxidation potential in these complexes
cluded that the oxidation and reduction were tuned from 1.3 V to 0.3 V ver-
potentials are the best indicators of the en- sus SCE, going from [RuL3 ] to [RuL2 L$ ]
ergy levels of the HOMO and LUMO [34]. L$ = phenylpyridine). Thus, the energy of
The energy between the metal t2g orbitals the HOMO is varied to an extent of 1 V
and π ∗ -orbitals can be reduced either by in this series of complexes [35]. Meyer and
raising the energy of the t2g or by de- coworkers have used the same concept to
creasing the energy of the π ∗ -orbitals synthesize black absorbers by judiciously
with donor acceptor ligands, respectively selecting push and pull ligands [36].
(Fig. 4). In the following sections, ways It is interesting to note the magnitude
to tune HOMO and LUMO energy lev- of the spectral shift for the lowest energy
els by introducing various ligands will be CT transition is ≈0.5 eV, and the shift in
discussed. the oxidation potential ≈1 eV. This clearly
shows that not all the HOMO tuning is
5.2.2.5.2 Spectral Tuning in ‘‘Push-Pull’’ translated into the spectral shift of the
Type of Complexes The lowest energy complex. The apparent 0.5 eV, difference
MLCT transition of ruthenium polypyridyl is involved in raising the energy of π ∗ -
complexes of the type [RuL3 ], (L = orbitals of the pulling ligands caused by
4,4$ dicarboxylic acid-2,2$ -bipyridine), can the pushing ligands [35].
be lowered so that it absorbs more in the
red region of the visible spectrum by sub- 5.2.2.5.3 Spectral Tuning in Complexes
stituting one of the bidentate ligand by a Containing Hybrid Donor Ligands The ab-
donor bidentate ligand (pushing ligand). sorption and luminescence spectral prop-
The remaining two electron-withdrawing erties of complexes of the type [Ru(dmbip)
bidentate ligands act as pull ligands. A list (dcbpy)(X)] and [Ru(dmbip)(dcbiq)(X)],
of complexes is included in Table 1, where where dmbip = 2,6-bis(1-methylbenzimid-
the absorption maxima of the complexes azol-2$ -yl)pyridine, dcbpy = 4,4$ -(COOH)2 -
Tab. 1 Absorption, emission and redox properties of ruthenium complexes containing

‘‘push-pull’’ ligands
Complex Abs. max. Emission max. V vs SCE Reference

(nm) (nm)
[Ru(dcbpy)3 ]4− 466 640 1.3 33

[Ru(dcbpy)2 (dmg)]2− 470 665 1.2 35
[Ru(dcbpy)2 (Me2 bpy)]2− 474 655 1.2 35
[Ru(dcbpy)2 (2COOpy)]3− 490 712 0.80 35
[Ru(dcbpy)2 (bpe)]2− 500 716 0.78 35
[Ru(dcbpy)2 (dea-bpy)]2− 506 732 0.60 35
[Ru(dcbpy)2 (dh-phen)]2− 516 760 – 35
[Ru(dcbpy)2 (NCS)2 ]4− 518 750 0.60 22
[Ru(dcbpy)2 (phpy)2 ]3− 560 865 0.3 35
Tab. 2 Absorption, emission, and redox properties of ruthenium complexes containing ‘‘hybrid
donor’’ ligands

(nm) (nm)
[Ru(dmbip)(dcbpyH)(Cl)] 516 – 0.67 32

[Ru(dmbip)(dcbpy)(H2 O)] 508 780 0.71 32
[Ru(dmbip)(dcbpy)(OH)] 528 800 – 32
[Ru(dmbip)(dcbpy)(NCS)] 502 770 0.80 32
[Ru(dmbip)(dcbpy)(CN)] 486 750 0.92 32
[Ru(dmbip)(dcbpy)(N(CN)2 )] 492 760 0.60 38
[Ru(dmbip)(dcbpy)(Cl3 -pcyd)] 502 – 0.74 38
[Ru(dmbip)(dcbpy)(NCN)] 528 – – 39
[Ru(dmbip)(5,5dcbpy)(NCN)] 542 – – 38
[Ru(dmbip)(dcbiqH2 )(NCS)] 580 900 1.15 32
[Ru(dmbip)(dcbiqH2 )(Cl)] Cl 608 >950 0.83 32
2,2$ -bipyridine, dcbiq = 4,4$ -(COOH)2 -2,2$ - the π-acceptor strength of the ancillary
biquinoline and X = Cl− , NCS− , CN− , ligand, CN− , NCS− , H2 O, NCN− Cl− .
N(CN)2 − , NCN− , and H2 O are summa- The red shift of the absorption maxima
rized in Table 2. The 2,6-bis(1-methyl- in complexes containing dcbiq compared
benzimidazol-2$ -yl)pyridine ligand is a hy- to dcbpy acceptor ligand is because
brid ligand that contains donor groups of the low π ∗ -orbitals of biquinoline.
(benzimidazol-2$ -yl) and an acceptor group The resonance Raman spectra of these
(pyridine). The complexes containing complexes for excitation at 568 nm show
these ligands show intense ultraviolet bands predominantly associated with the
(UV) bands at 362 and 346 due to dcbpy and dcbiq ligands indicating that the
intraligand π –π ∗ transition of 2,6-bis lowest excited state is a metal to dcbpy or
(1-methylbenzimidazol-2$ -yl)pyridine. The dcbiq ligand CT state [32, 37, 38].
bands at 308 and 244 nm are because of
the π –π ∗ transition of dcbpy ligand. On 5.2.2.5.4 Influence of Nonchromophoric
acidification, the 308 nm band shifts to Ligands on MLCT Transitions This section
314 nm as a result of the protonation of compares spectral response of the three
the carboxylate groups, whereas the 362 complexes of the type [Ru(P-terpy)(Me2
and 342 nm bands remain unchanged in bpy)(NCS)] 1, (P-terpy = 4-phosphonato-2,
acidic and basic pH. 2$ : 6$ ,2$$ -terpyridine, Me2 bpy] = 4,4$ -di-
The MLCT bands of these complexes methyl-2,2$ -bipyridine), [Ru(H2 dcbpy)2
are broad and red shifted by approximately (NCS)2 ] 2, (H2 dcbpy = 4,4$ -dicarboxy-2,
140 nm, compared to [Ru(dcbpy)3 ]. The 2$ -bipyridine) and (TBA)2 [RuL(NCS)3 ] 3,
lowest-energy MLCT transitions within (L$ = 4,4$ ,4$$ -tricarboxy-2,2$ : 6$ ,2$$ -terpy-
this series were shifted from 486 to ridine) where the number of nonchro-
608 nm, and the HOMO level varied over mophoric ligands such as NCS− are
an extent of 0.45 V versus SCE. The varied from one to three. The absorption
energy of the MLCT transition in these and luminescence spectral properties of
complexes decreases with the decrease in complex 1–3 are summarized in Table 3.
Tab. 3 Absorption, emission, and redox properties of ruthenium complexes containing NCS
ligands

(nm) (nm)
[Ru(pterpy)(Me2 bpy)(NCS)] 500 780 0.86 39

[Ru(tcterpy)(Me2 bpy)(NCS)] 510 800 0.94 39
[Ru(H2 dcbpy)2 (NCS)2 ] 535 830 0.85 22
[Ru(dcbpy)(Me2 bpy)2 (NCS)2 ] 527 760 0.78 39
[Ru(terpy)(NCS)3 ] 580 820 0.54 40
[Ru(pterpy)(NCS)3 ] 580 820 0.52 43
[Ru(tcterpy)(NCS)3 ] 620 950 0.60 40
The absorption maxima are listed for carboxylic acid groups at the 4,4$$ -position
the intense lowest-energy MLCT bands in of 2,2-bipyridine ligand the MLCT maxima
ethanol, although the spectra possess ad- red shifted to 535 nm. However, substitu-
ditional absorption features at lower and tion at the 5,5$ -position of 2,2$ -bipyridine
higher energies. shifted the maxima further into the red,
The interesting feature of these com- 580 nm. On the contrary, the MLCT max-
plexes is a broad MLCT absorption bands ima blue shifted (500 nm) by substituting
in the visible region. The most intense at the 6,6$ positions of 2,2$ -bipyridine.
MLCT transition maxima of complexes 1, The enhanced red response of com-
2, and 3 are at 500, 535, and 620 nm, plexes containing the 5,5$ -dicarboxylic
respectively. The 120 nm red shift of 3 acid-2,2$ -bipyridine is because of a de-
compared to 1 is the result of an increase in the energy of the π ∗ -orbitals,
crease in the energy of the metal t2g which makes them attractive as sensitizers
orbitals caused by introducing nonchro- for nanocrystalline TiO2 films. Bignozzi
mophoric ligands and a slight decrease and coworkers found that the IPCE of com-
in the LUMO of the 4,4$ ,4$$ -tricarboxy- plexes having the 5,5$ -dicarboxylic acid-
2,2$ : 6$ ,2$$ -terpyridine. The electrochem- 2,2$ -bipyridine ligands were lower than
ical data are consistent with the above the analogous complexes that contain 4,4$ -
assignments [22, 39, 40]. dicarboxylic acid-2,2$ -bipyridine [43]. They
rationalized the low efficiency of the sen-
5.2.2.5.5 Influence of the Position of Car- sitizers containing 5,5$ -dicarboxylic acid-
boxyl Groups on MLCT Transitions Ruthe- 2,2$ -bipyridine ligands in terms of low
nium polypyridyl compounds, of the type excited state redox potentials.
[RuL2 (X)2 ], (where L = 4,4$ -dicarboxylic
acid-2,2$ -bipyridine, 5,5$ -dicarboxylic acid- 5.2.2.5.6 MLCT Transitions in Geometric
2,2$ -bipyridine, and 6,6$ -dicarboxylic acid- Isomers Isomerization is another excit-
2,2$ -bipyridine; X = Cl− , CN− and NCS− ) ing approach toward tuning the spectral
have been reported [41–43]. The lowest properties of metal complexes (Fig. 5)
MLCT absorption maxima of [RuL2 (NCS)2 ] [44–46]. The electronic properties of cis-
complex (L = 2,2$ -bipyridine) is seen at and trans-isomers of [RuL2 (X)2 ], (L = 4,4$ -
510 nm in ethanol. By substituting two dicarboxylic acid-2,2$ -bipyridine; X = Cl− ,
0.8 COOH
HOOC COOH
HOOC
N
0.6 NCS N N
N
Ru SCN Ru NCS
N
NCS
Absorbance
N N
HOOC N
0.4
HOOC COOH
HOOC
0.2
0.0
300 400 500 600 700 800
Energy
[nm]
Fig. 5 UV-visible absorption spectra of cis-, trans-
bis(2,2$ -bipyridine-4,4$ -dicarboxylic acid) ruthenium (II) (NCS)2 complexes.
Tab. 4 Electronic Spectral Data of the cis- and trans-bis(4,4$ -dicarboxylic

acid-2,2$ -bipyridine)Ruthenium (X2 ) Complexes (X = Cl− , H2 O, NCS− ) in DMF
Complex Abs. max. Emission max. Reference

(nm) (nm)
cis-[Ru(H2 dcbpy)2 (Cl)2 ] 590 847 47

cis-[Ru(H2 dcbpy)2 (H2 O)2 ]a 500 750 44
cis-[Ru(H2 dcbpy)2 (NCS)2 ] 535 830 47
trans-[Ru(H2 dcbpy)2 (Cl)2 ] 690 >900b 47
trans-[Ru(H2 dcbpy)2 (H2 O)2 ] 530 800b 47
trans-[Ru(H2 dcbpy)2 (NCS)2 ] 581 870 47
cis-[Ru(bpy)2 (H2 O)2 ] 480 660 45
trans-[Ru(bpy)2 (H2 O)2 ] 495 700 46
DMF, dimethylformamide.
a in water at pH 4.5.
b Weak and broad emission maximum.
H2 O and NCS− ) are collected in Table 4. and 440 nm. On the other hand, the
The UV-visible absorption spectrum of the cis-dichloro complex in DMF solution
trans-dichloro complex in DMF solution shows only two distinct broadbands in
shows at least three MLCT absorption the visible region at 590 and 434 nm,
bands in the visible region at 690, 592, which was assigned to a MLCT transitions.
The lowest energy MLCT band in the the L3 ligands. In these complexes, L3 be-
trans-dichloro complex is significantly red ing the dithiocarbamate ligand, shifts the
shifted compared to the cis-dichloro com- Ru(II) t2g level upward without signifi-
plex. This red shift is due to stabilization cantly affecting the π ∗ -level of the dcbpy
of the LUMO of the dcbpy ligand in the ligand thereby increasing the MLCT ab-
trans-species relative to the cis-species [47]. sorption in the red portion of the visible
The red shift (108 nm) of the lowest region.
energy MLCT absorption in the spectrum
of the trans-dichloro complex compared to 5.2.3
the spectrum of the trans-dithiocyanato is Hydrophobic Sensitizers
due to the strong σ donor property of the
Cl− compared to the NCS− ligand. The The other important aspect in dye-
chloride ligands cause destabilization of sensitized solar cells is water-induced
the metal t2g orbitals, and raising them desorption of the sensitizer from the
in energy relatively closer to the ligand surface. Extensive efforts have been made
π ∗ -orbitals result in a lower energy MLCT in our laboratory to overcome this problem
transition. According to Lever’s electro- by introducing hydrophobic properties in
chemical parameterization, the chloride
the ligand. The heteroleptic complexes
ligand has a parameter value of −0.24 and
containing hydrophobic ligands of
the thiocyanate ligand has a value of −0.06.
the type [Ru(dcbpy)(mhdbpy)(NCS)2 ] 1,
The electrochemical and the absorption
[Ru(dcbpy)(dtdbpy)(NCS)2 ] 2 [Ru(dcbpy)
data are in good agreement with reported
(mddbpy)(NCS)2 ] 3 (dcbpy = 4,4$ -dicar-
parameterization scale for ligands [48].
boxy-2,2$ -bipyridine, mhdbpy = 4-methyl-
4$ -hexadecyl-2,2$ -bipyridine and dtdbpy =
5.2.2.5.7 Spectral Tuning in Heteroleptic
4,4$ -ditridecyl-2,2$ -bipyridine, mddbpy =
Sensitizers
4-methyl-4$ -didodecyl-2,2$ bipyridine) have
The synthesis and characterization of
been synthesized (Fig. 6). The photo-
heteroleptic complexes of ruthenium
[Ru(L1 )(L2 )(L3 )] starting from a poly- current action spectra of these complexes
meric complex Ru(CO)2 Cl2 have been show broad features covering a large part
reported [49–51]. The synthetic method- of visible spectrum and displays a maxima
ology of such heteroleptic complexes in- around 550 nm, where the incident
volves several steps, resulting in very low monochromatic IPCE exceeds 80%.
yields. Recently, a novel synthetic route The performance of these hydrophobic
has been developed to introduce differ- complexes as CT photosensitizers in
ent ligands on a ruthenium (II) precur- nanocrystalline TiO2 -based solar cell
sor, namely, the RuCl2 (DMSO)4 complex, shows excellent stability toward water-
which is a versatile and well-characterized induced desorption [55].
starting material [52]. A range of mixed The rate of electron transport in dye-
ligand sensitizers of the type [Ru(L1 )(L2 ) sensitized solar cell is a major element of
(L3 )] have been prepared and their spectro- the overall efficiency of the cells. The elec-
scopic, luminescent, and redox properties trons injected into the conduction band
are reported [53, 54]. The absorption spec- from optically excited dye, can traverse the
tral data of these complexes show a red TiO2 network and can be collected at the
shift with increasing donor strength of transparent conducting glass or can react
Complex 1 Complex 2 Complex 3

Fig. 6 Structure of complexes [Ru(dcbpy)(dtdbpy)(NCS)2 ] 1,
[Ru(dcbpy)(mhdbpy)(NCS)2 ] 2, and [Ru(dcbpy)(mddbpy)(NCS)2 ] 3.
with a redox mediator. The reaction of in- current conversion yields remaining un-
jected electrons into the conduction band der 1%, which is insufficient for solar cell
with a redox mediator gives undesirable applications [57–60]. One of the reasons
dark currents, significantly reducing the for such low efficiencies is aggregation of
charge-collection efficiency, and thereby the dye on the TiO2 surface. This asso-
decreasing the total efficiency of the cell. ciation often leads to undesirable photo-
This problem is drastically suppressed by physical properties such as self-quenching
using hydrophobic sensitizers of the type and excited state annihilation. However,
shown in Fig. 6. The most likely explana- the advantage of this class of complexes
tion for the dark-current suppression is is the near IR response which is very
that the long chains of the sensitizer inter- strong, having extinction coefficients of
acts laterally to form an aliphatic network, close to 50 000 M−1 cm−1 at 650 nm com-
thereby preventing triiodide from reaching pared to the polypyridyl complexes, which
the TiO2 surface. have small extinction coefficients at this
wavelength.
5.2.4
Near IR Sensitizers
5.2.4.2 Ruthenium Phthalocyanines
5.2.4.1 Phthalocyanines The bis(3,4-dicarboxypyridine)Ru(II)1,4,8,
Phthalocyanines possess intense absorp- 11,15,18,22,25-octamethyl-phthalocyanine
tion bands in the near IR region and are (JM3306) shows a visible absorption band
known for their excellent stability render- at 650 nm (ε 49 000 M−1 cm−1 ) and a
ing them attractive for photovoltaic ap- phosphorescence band located at 895 nm.
plications [56]. They have been repeatedly The triplet state lifetime is 474 ns under
tested in the past as sensitizers of wide anaerobic conditions. The emission is
band gap oxide semiconductors and ob- entirely quenched when JM3306 is ad-
tained poor incident photon to electric sorbed onto a nanocrystalline TiO2 film.
The very efficient quenching of the the π –π ∗ -excitation which is responsible

emission of JM3306 was found to be due to for the 650 nm absorption band. This
the electron injection from the excited sin- phenomenon shows that the electronic
glet/triplet state of the phthalocyanine into coupling of the excited state of the dye
the conduction band of the TiO2 [61]. The to the Ti (3d) conduction band manifold
photocurrent action spectrum is shown in is strong enough through this mode of
Fig. 7, where the incident IPCE is plotted attachment to render charge injection very
as a function of wavelength. The feature efficient. These results establish a new
is extending well into the near IR region pathway for grafting dyes to oxide surfaces
displaying a maximum at around 660 nm through axially attached pyridine ligands.
where the IPCE exceeds 60%. These are
by far the highest conversion efficien- 5.2.4.3 Phthalocyanines Containing 3d
cies obtained with the phthalocyanine type Metals
sensitizers. Several zinc (II) and aluminum (III)
It is fascinating to note that this class of phthalocyanines substituted by car-
dyes inject efficiently into the conduction boxylic acid and sulfonic acid groups
band of TiO2 , despite the fact that the were anchored to nanocrystalline TiO2
pyridyl orbitals do not participate in films and tested for their photovoltaic
100
80
60
IPCE
[%]
40
20
0
400 500 600 700 800
Wavelength
[nm]
Fig. 7 Photocurrent action spectra of nanocrystalline TiO2
films sensitized by bis(3,4-dicarboxypyridine)Ru(II)
1,4,8,11,15,18,22,25-octamethyl-phthalocyanin. The incident
IPCE is plotted as a function of wavelength.
behavior using LiI/LiI3 /propylene car- from the lower energy MLCT band of
bonate electrolyte [62]. Interestingly, the complexes containing 5,5$ -dicarboxylic
zinc (II) 2,9,16,23-tetracarboxyphthalocya- acid-2,2$ -bipyridine ligand because of lack
nine exhibited 45% monochromatic cur- of driving force for electron injection.
rent conversion efficiency at 700 nm. It
is shown that electron injection to TiO2 5.2.5.2 Osmium Complexes
occurs from the excited singlet state Although there are a large number of
of the phthalocyanine derivatives. The osmium polypyridyl complexes, very few
inherent problem of phthalocyanine ag- have been used as sensitizers in dye-
gregation is reduced considerably in 3d sensitized solar cells. The osmium com-
metal phthalocyanines by introducing 4- plexes have several advantages compared
tert-butylpyridine and 3α,7α-dihydroxy-5β- to their ruthenium analogues: (1) osmium
cholic acid (cheno) into the dye solution. has a stronger ligand field splitting com-
The added 4-tert-butylpyridine in all likeli- pared to ruthenium, (2) the spin-orbit
hood coordinates to the metal in the axial coupling leads to excellent response in the
position and thereby prevents aggregation red region because of enhanced singlet-
of the phthalocyanines. triplet CT transitions [26]. Heimer and
This type of sensitizers opens up new av- coworkers have synthesized [OsL2 (CN)2 ]
enues for improving the near IR response complex (L = 4, 4$ -dicarboxylic acid-2,2$ -
of nanocrystalline injection solar cell. In bipyridine) and found that the complex
addition, important applications can be is extremely stable under irradiation in a
foreseen for the development of photo- homogeneous aqueous solution compared
voltaic windows transmitting part of the to the analogous ruthenium complex [64].
visible light. Such devices would remain The greater photo-stability for osmium
transparent to the eye, while absorbing is consistent with a stronger crystal field
enough solar energy photons in the near splitting of metal d-orbital that inhibits ef-
IR to render efficiencies acceptable for ficient population of d-d states. However,
practical applications. a factor of 10 lower injection efficiency for
the osmium complex when anchored on
5.2.5 TiO2 electrode was reported compared to
Mononuclear and Polynuclear Metal the ruthenium analog and rationalized in
Complexes of Group VIII terms of slow regeneration rates for the
oxidized dye [65].
5.2.5.1 Iron Complexes Lewis and coworkers have synthesized
The photoconversion efficiency for iron and characterized two osmium com-
complexes [FeL3 ] and [FeL2 (X)2 ], (L = plexes [OsL2 (CN)2 ] and [OsL3 ] (L = 4,4$ -
4,4$ -dicarboxylic acid-2,2$ -bipyridine and dicarboxylic acid-2,2$ -bipyridine). In their
5,5$ -dicarboxylic acid-2,2$ -bipyridine) has studies they found nearly 80% incident
been reported [63]. The [FeL2 (X)2 ] type monochromatic IPCE [66].
complexes show two broad MLCT bands
in the visible region. In both cases, 5.2.5.3 Polynuclear Complexes
the photoconversion efficiency of the Polynuclear systems in which more than
high-energy band was reported in the one complex is connected by a bridging
range of 40%. Nonetheless, there is ligand have been studied in several labo-
virtually no photocurrent contribution ratories [67–72]. Here studies have been
restricted exclusively to the cyano-bridged with the three metal centers, the one with
(CN-bridged) trinuclear ruthenium com- lowest energy is based on the central
plexes of the type shown in Fig. 8. Scandola unit carrying the ligand 4,4$ -dicarboxy-
and coworkers have synthesized and char- 2,2$ -bipyridine (dcbpy). When coated on
acterized several of these complexes with a TiO2 -electrode, the complex anchors
respect to their photophysical and elec- through the carboxyl unit of this central
trochemical properties [73]. In such com- unit. Photons absorbed by the higher en-
plexes, one can invoke possible antenna ergy peripheral units efficiently transfer
effects in the light harvesting process by in- excitation energy to this central unit, ac-
troducing various fragments in which the counting for its excellent performance.
photophysical properties are tuned to have To understand the principles of opera-
directionality in the excited state. The pres- tion of polynuclear complexes and possibly
ence of numerous metal centers should improve their performance, the ‘‘tunabil-
allow for different tailored fragments con- ity’’ of photophysical and redox properties
taining various ligands [72, 74]. has been examined in a number of such
Potential application of these complexes complexes. These complexes are com-
as photosensitizers for light energy har- posed of polypyridyl units of Ru and
vesting relies heavily on demanding syn- Os as chromophores and cyanide as the
thetic schemes. The CN-bridged trinu- bridging ligand [76]. By appropriate syn-
clear complex 1 (Fig. 8) has been found thetic procedures, it is possible to control
to be an excellent photosensitizer for the mode of coordination of the bridg-
spectral sensitization of the wide band ing cyanide to a given chromophoric unit
gap semiconductor TiO2 [75]. Among the as a nitrile (C-bonded) or isonitrile (N-
possible MLCT excited states associated bonded) [74–77]. By varying the nature of
R
COOH
R
R
N
HOOC N N
N Ru
N C C
N N
Ru N
N N
N R
N R
HOOC C
C N
Ru
HOOC N N
N
R
R
R
1 : R = H ; 2 : R = COOH ; 3 : R = CH3 ; 4 : R = C6H5
5 = [(CN)(bby)2 Os-CN-Ru(dcbpy)2-NC-Os(bpy)2(CN)]
Fig. 8 Structure of CN-bridge
trinuclear ruthenium complex . 6 = [(H2O)(bby)2 Ru-CN-Ru(dcbpy)2-NC-Ru(bpy)2(H 2O)]
the metal center, the mode of coordination the complex and the substrate, is an
of the bridging ligand, and the spectator important target in dye-sensitized solar
ligands, it is possible to vary the location cells. Several ruthenium complexes con-
and the direction of excited state electron taining carboxylic acid and phosphonated
transfer process. polypyridyl ligands are described [78–83].
The carboxylic and phosphonate function-
5.2.6 ality serves as an anchoring group to
Surface Chelation of Polypyridyl Complexes immobilize the complex on the nanocrys-
onto the TiO2 Surface talline TiO2 films. The immobilized sen-
sitizer absorbs a photon to produce an
5.2.6.1 Anchoring Groups excited state, which efficiently transfers
The functional groups serve as grafting its electron onto the TiO2 conduction
agents for the oxide surface of the TiO2 band. To achieve high quantum yields
films. The grafting of polypyridyl com- of the excited state electron-transfer pro-
plexes onto the oxide surface, which allows cess the dye ideally needs to be in in-
for electronic communication between timate contact with the semiconductor
HOOC COOH
HOOC COOH
N N N N
4,4′-dicarboxyl-2,2′-bipyridine 5,5′-dicarboxyl-2,2′-bipyridine
HOOC COOH
COOH
HOOC COOH
N
N N
N N
4,4′-(paradicarboxyl)diphenyl-2,2′-bipyridine 4,4′,4′′-tricarboxylic-2,2′;6,2′′-terpyridine
P(O)(OH)2
P(O)(OH)2 P(O)(OH)2
N
N N
4,phosphonate-2,2′;6,2′′-terpyridine 4,4′-diphosphonate-2,2′;bipyridine
HOOC COOH
HOOC COOH
N N
N N
4,4′-dicarboxyl-2,2′-biquinoline 4,7-dicarboxyl-1,10′-phenanthroline
Fig. 9 Structure of the ligands containing anchoring groups.

surface. The ruthenium complexes that 5.2.6.2 Acid-Base Equilibria of the

have carboxylic acid and phosphonic acid Anchoring Groups
groups show electron-transfer processes in The grafting properties of complexes con-
the range of 80 to 90%. The near quantita- taining the anchoring group onto the TiO2
tive electron injection efficiency indicate a surface and the factors responsible for
close overlap of the ligand π ∗ -orbitals and differences in the attachment of various
the titanium 3d orbitals [84]. anchoring groups are essential for the
Figure 9 shows the polypyridine ligands design of novel sensitizers [89]. The dif-
with different anchoring groups that have ference in binding properties of the com-
been employed for surface derivation plexes containing carboxyl and phospho-
of ruthenium complexes. Goodenough nate anchoring groups may stem from the
and coworkers have derivatized a mixed differences in the pKa values of the ligands.
ligand complex [Ru(bpy)2 (dcbpy)] on TiO2 Knowledge of the pKa values of ionizable
and SnO2 films in the presence of groups may aid in indicating the possible
dicyclohexylcarbodimide [84]. nature of the attachment. The acid base
The interaction between the adsorbed properties of transition metal complexes
sensitizer and the semiconductor sur- containing ligands with protonatable func-
face has been addressed using reso- tional groups can provide important in-
nance Raman and Fourier transform formation about the nature of binding
infrared (FT-IR) spectroscopy. The car- properties on the oxide surface.
boxylic acid functional group could ad-
sorb on the surface in a unidentate, 5.2.6.3 Acid-Base Equilibria of the
bidentate, or bridging fashion. Yanagida 4,4-Dicarboxy-2,2!-Bipyridine and its
and coworkers concluded that the sensi- Complexes
tizer cis-dithiocyanato bis(2,2$ -bipyridine- The ground state pKa values of
4,4$ -dicarboxylate) ruthenium (II) complex [Ru(dcbpy)2 (NCS)2 ] (N3) were determined
(N3) binds to the surface using ester-like from the relationship between the change
and chelating linkages [85]. Woolfrey and in the optical density or the peak maximum
coworkers have reported that the N3 com- with the pH for a given wavelength [90].
plex anchor on the surface of TiO2 as When acid is added to an alkaline solution
a bidentate or bridging mode using two of N3, changes in the electronic spectrum
carboxylate groups per dye [86]. However, occur as shown in Fig. 10. The plot of λ-
Fillinger and Parkinson studied the ad- max change versus pH for N3 shows the
sorption behaviour of the N3 sensitizer expected sigmoidal shape, with the pH at
and found that the initial binding involves the inflection point giving two ground state
one carboxylate, with subsequent binding pKa values at 3 and 1.5 ± 0.1.
of two or more carboxylate groups on the In a related complex of the type
surface [87]. [Ru(bpy)2 (dcbpy)], two inflection points at
It is noteworthy that, Bignozzi and pH 2.7 and <0.5 were found suggesting
coworkers achieved sensitization of TiO2 that the dissociation of the carboxylic acid
with a cyanobridged dimer of Ru-Re groups of 4,4-dicarboxy-2,2$ -bipyridine lig-
polypyridyl complex where the coupling and is a sequential process. The pKa1 of the
of the MLCT exited state to the conduction free 4,4-dicarboxy-2,2$ -bipyridine ligand is
band manifold was through space and did 4.2 and the second one is below 2. The
not involve the anchoring group [88]. difference between the free ligand and
1.2
pH 9.0
1.0 pH 6.2
pH 4.4
pH 3.9
0.8 pH 3.0
pH 2.4
Absorbance
pH 1.0
[OD]
0.6
0.4
0.2
0.0
400 450 500 550 600 650 700
Wavelength
[nm]
Fig. 10 Absorption spectral changes of cis-dithiocyanatobis(2,2$ -bipyridine-4,4$ -dicarboxylic acid)
ruthenium (II) complex as a function of pH.
the complex pKa1 can be considered as the 4-phosphonato-2,2$ : 6$ ,2$$ -terpyridine

a measure of donor strength of the lig- is sterically restricted and less basic
and. The difference in the pKa values of because of the doubly protonated form.
[Ru(dcbpy)2 (NCS)2 ] and [Ru(bpy)2 (dcbpy)] The dissociation constants of 4-
is rationalized in terms of the donor phosphonato-2,2$ : 6$ ,2$$ -terpyridine lig-
or acceptor properties of nonprotonatable and were measured by pH-nuclear mag-
ligands [91]. netic resonance (NMR) and potentiometric
titration methods. The 4-phosphonato-
5.2.6.4 Acid-Base Equilibria of the 2,2$ : 6$ ,2$$ -terpyridine ligand shows three
Phosphonato Group measurable pKa s at 6.0, 4.65, and 3.55.
The 4-phosphonato-2,2$ : 6$ ,2$$ -terpyridine The fourth pKa was estimated to be
is protonated in four steps in the pH re- below 2 [92]. Spectrophotometric titration
gion from 1.5 to 7 as shown in Sch. 1. of the ruthenium complex of the type
Out of four protons, two are involved in [Ru(P-terpy)(Me2 bpy)(NCS)] (P-terpy = 4-
protonation of the pyridyl nitrogens of ter- phosphonato-2,2$ : 6$ ,2$$ -terpyridine and
pyridyl, and the remaining two protons Me2 bpy = 4,4$ -dimethyl-2,2$ -bipyridine)
are on the phosphonated group. Since exhibits two ground state pKa values at
the adjacent nitrogens of 2,2$ -bipyridyl 6 and 4, which can be assigned to pKa1 and
ligand accept only one proton down to pKa2 , respectively.
pH 1.6, it can be safely assumed that The emission maximum of [Ru(P-
4-phosphonato-2,2$ : 6$ ,2$$ -terpyridine ac- terpy)(Me2 bpy)(NCS)] shows a blue shift
cepts protons in nonadjacent positions as with decreasing pH, and apparent
shown in Sch. 1. The central nitrogen of pH dependent excited state lifetimes.
O 2+ O +
HO OH HO O−
P P
pKa4 < 2
H H H H
N N N N
6 2 N N
5 3
4
pKa3 = 3.55
O
HO O−
P
H
N N
N
pKa2 = 4.65
O 2− O −
−
O O− HO O−
P P
pKa1 = 5.99
N N N N
N N
Scheme 1 Acid-base equilibria of 4-phosphonato-2,2$ : 6,2$$ -terpyridine.
The complex [Ru(P-terpy)(Me2 bpy)(NCS)] state mostly 4-phosphonato-2,2$ : 6,2$$ -

shows two excited state pK∗a s at 6.5 terpyridine peaks are enhanced compared
(pK∗a1 ) and 4.5 (pK∗a2 ). Comparison of to the Me2 bpy [39].
the ground and excited state pKa s Schmehl and coworkers has measured
of complex [Ru(P-terpy)(Me2 bpy)(NCS)] ground and excited state pKa of sev-
demonstrate that the excited state eral ruthenium polypyridyl complexes
has a metal to 4-phosphonato-2,2$ : containing phosphonate groups. They
6$ ,2$$ -terpyridine ligand CT character. reported two ground state pKa values
The resonance Raman data of [Ru(P- approximately 2 and 6.3 for [Ru(bpy)2
terpy)(Me2 bpy)(NCS)] show that the (bpppH2 )]. The excited pK∗a1 of [Ru(bpy)2
energy levels of the two ligands (bpppH2 )] is more basic than the ground
are very close and in the excited state pKa1 suggesting that the MLCT
transition is localized on the bppH2 Higher temperature processes, including

ligand [93]. deamination of the counterion as well as
decarboxylation and finally decomposition
5.2.7 of the complex, occur between 200 to
Stability and Performance of the Dyes 400 ◦ C with different characteristics ob-
served in air and in nitrogen (Fig. 11).
5.2.7.1 Stability of the Inorganic Dyes The decarboxylation reaction is an endo-
The thermal stability of ruthenium (II) thermic process in a nitrogen atmo-
complexes of the type [Ru(H2 dcbpy)2 sphere and overlaps with decomposition
(NCS)2 ] 1, (Bu4 N)2 [Ru(Hdcbpy)2 (NCS)2 ] of the complexes. In air, on the other
2, (Bu4 N)4 [Ru(dcbpy)2 (NCS)2 ] 3, and (Im)4 hand, it is an exothermic process dis-
[Ru(dcbpy)2 (NCS)2 ] 4, where dcbpy = 2, 2$ - tinctively separated from decomposition.
bipyridyl-4,4$ dicarboxylate, Bu4 N = tetra- Higher thermal stability is observed for
butylammonium, and Im = dimethylethy- 1 and 2 when anchored onto nanocrys-
limidazolium, has been studied using ther- talline TiO2 films. The activation energy
moanalytical techniques, IR, UV-visible, of decarboxylation is estimated for 1 in
and 1 H NMR spectroscopic methods [94]. the free form (≈115 kJ mol−1 ) and on
These complexes show remarkable stabil- TiO2 (≈126 kJ mol−1 ). In addition, the
ity in both nitrogen and air atmospheres deamination temperature increases by
at high temperatures, ranging from 180 ◦ C about 35 ◦ C when Bu4 N are replaced by
to 250 ◦ C for 1. The only process that is Im cations. A surprising observation is
observed at lower temperatures is the dehy- the two-step deamination in dyes with
dration that occurs between 40 to 110 ◦ C. four organic cations, which is due to the
100 TG 0
98
−5
Deriv. mass
Exo(heat flow)
DTG
Mass
[%]
[%]
96
−10
94
0 −15
DSC
92
100 200 300 400
Temperature
[°C]
Fig. 11 The combined thermogravimetric (TG), derivative
thermogravimetric (DTG), and differential scanning calorimetry (DSC)
curves of complex (Bu4 N)4 [Ru(dcbpy)2 (NCS)2 ] adsorbed on TiO2 in
dynamic air atmosphere. Loss of water, exothermic decomposition of
amine (between 200–330 ◦C) and decarboxylation starting at 360, ending
at 390 ◦ C.
difference in the extent of interaction be- anionic complex, although consequently,

tween these cations and the carboxylate the short-circuit photocurrent is lower.
groups. Thus, there should be an optimum de-
gree of protonation of the sensitizer (for
5.2.7.2 Effect of Protons Carried by the example (Bu4 N)2 [Ru(Hdcbpy)2 (NCS)2 ] 2,
Sensitizer on the Performance which contains two protons) to obtain
To obtain high conversion efficiencies, optimum short-circuit photocurrent and
optimization of the short-circuit photo- open-circuit potential, which determines
current (isc ) and open-circuit potential the power conversion efficiency of the cell.
(Voc ) of the solar cell is essential. The The performance of the three sensi-
conduction band of the TiO2 is known tizers 1, 2, and 3 that contain differ-
to have a Nernstian dependence on ent degrees of protonation were studied
pH [80]. The fully protonated sensitizer
on nanocrystalline TiO2 electrodes [90].
[Ru(H2 dcbpy)2 (NCS)2 ] 1, on adsorption
The photocurrent action spectra obtained
transfers most of its protons to the TiO2
with a monolayer of these complexes
surface charging it positively. The elec-
coated on TiO2 films are shown in
tric field associated with the surface dipole
Fig. 12. The redox electrolyte consisted
generated in this fashion enhances the
of a solution of 0.6 mol dimethylpropyl
adsorption of the anionic ruthenium com-
plex and assists electron injection from imidazolium iodide, and 0.05 mol I2 , in
the excited state of the sensitizer in the methoxyacetonitrile.
titanium conduction band, favoring high The incident monochromatic IPCE is
photocurrents. However, the open-circuit plotted as a function of excitation wave-
potential is lower as a result of the positive length. The IPCE value in the plateau
shift of the conduction band edge induced region is 80% for complex 1, whereas
by the surface protonation. for complex 3 it is only about 66%. In
On the other hand, the sensitizer the red region, the difference is even
(Bu4 N)4 [Ru(dcbpy)2 (NCS)2 ] 3, that carries more pronounced. Thus, at 700 nm the
no protons shows high open-circuit poten- IPCE value is twice as high for the
tial compared to the complex 1, because of fully protonated complex 1 as compared
the relative negative shift of the conduction to the deprotonated complex 3. Con-
band edge induced by the adsorption of the sequently, the short-circuit photocurrent
80
1
2
60
3
IPCE
40
[%]
Fig. 12 Photocurrent action spectra of

nanocrystalline TiO2 films sensitized by 20
the complexes [Ru(H2 dcbpy)2 (NCS)2 ] 1,
(Bu4 N)2 [Ru(Hdcbpy)2 (NCS)2 ] 2, and 0
(Bu4 N)4 [Ru(dcbpy)2 (NCS)2 ] 3. The
400 500 600 700 800
incident photon to current conversion
efficiency is plotted as a function of Wavelength
wavelength. [nm]
is 17.5–18.5 mA cm−2 for complex 1, yridine) under similar conditions, show an

whereas it is only about 12–13 mA cm−2 IPCE value between 70–80% in the plateau
for complex 3. However, there is a trade- region. Although the IPCE values are com-
off inasmuch as the photovoltage is about parable with that of (Bu4 N)3 [RuL(NCS)3 ],
200 mV higher for the latter as compared the total integrated current decreased
to the former sensitizer. Nevertheless, this significantly because of increasing blue
is insufficient to compensate for the cur- shift of the spectral response of the later
rent loss. Hence, the photovoltaic perfor- complexes. This shows that the ideal sen-
mance of complex 2 carrying two protons sitizer should absorb most of the visible
is superior to that of compounds 1 and light in order to achieve high currents.
3 that contains four and no protons, re-
spectively. The doubly protonated form of 5.2.8
the complex is therefore preferred over the Synthesis and Characterization
other two sensitizers for sensitization of
5.2.8.1 Synthetic Strategies for Ruthenium
nanocrystalline TiO2 films.
Complexes
Several groups have synthesized various
5.2.7.3 Comparison of IPCE Obtained with ruthenium complexes using different
Various Sensitizers
starting materials. Kelly, Meyer, Vos, and
Figure 8 in chapter 5.3, shows the pho-
Decon’s groups have used a polymeric
tocurrent action spectrum of a cell contain-
starting complex [Ru(CO)2 Cl2 ] for step-
ing various sensitizers. The broad feature
wise synthesis of complexes containing
covers the entire visible spectrum and
different ligands [49, 51, 95]. Recently,
extending into the near IR region up
the synthesis of heteroleptic complexes
to 920 nm, for (Bu4 N)3 [RuL(NCS)3 ] (L$ =
starting from the [Ru(Cl)2 (dmso)4 ] com-
4,4$ ,4$$ -tricarboxy-2,2$ : 6$ ,2$$ -terpyridine)
plex [54] have been reported. For homolep-
complex. The incident IPCE value in
tic complexes Dwyer and coworkers used
the plateau region being about 80% for
K4 [Ru2 Cl10 O] in weakly acidic aqueous
(Bu4 N)3 [RuL(NCS)3 ] complex. Taking the
solution containing the appropriate lig-
light losses in the conducting glass into
account the efficiency of electric current and [96]. The same procedure has been
generation is practically 95% over a broad employed with RuCl3 . 3H2 O as source ma-
wavelength range extending from 400 terial. Wilkinson’s group used ruthenium
to 700 nm. The overlap integral of the blue solutions in the synthesis of variety
absorption spectrum with the standard of complexes [97]. Other groups have used
global AM 1.5 solar emission spectrum RuCl3 . 3H2 O as starting material for the
for (Bu4 N)3 [RuL(NCS)3 ] complex yields a synthesis of complexes of the type [RuL3 ],
photocurrent density of 20.5 mA cm2 . The [RuL2 L$$ ] and [RuL2 (X)2 ] in DMF [98]. The
open circuit potential is 720 mV and the fill advantage of the solvent DMF is that it
factor is 0.7. These results were confirmed provides high refluxing temperature and
at the National Renewable Energy Labora- at the same time acts as a reducing agent
tory (NREL), Golden, Colorado, USA (see for the ruthenium (III).
Chapter 5.3, Fig. 10).
The complexes [RuL3 ], [RuL2 (RuL$$2 5.2.8.2 Purification
(CN)2 )2 ], and [RuL2 (NCS)2 ] (L = 4,4$ - Purity is an indispensable requirement
dicarboxy-2,2$ -bipyridine; L$$ = 2,2$ -bip- of any sensitizer for better efficiency of
the dye-sensitized solar cell. Although well engineering aspect of the sensitizers for
worked-out procedures exist for the effi- enhanced spectral response into the visible
cient purification of metal complexes, it and near IR region is clearly illustrated.
was found that the isolation of complexes The development of efficient dye-
at its isoelectric point and followed by col- sensitized solar cells would expand our
umn purification using Sephadex LH-20 ability to better harvest the sun’s energy
gel resulted in analytically pure samples. and decrease our dependency on fossil
fuels and nuclear power.
5.2.8.3 Characterization
The most commonly used methods for Acknowledgment
characterization of ruthenium sensitizers
We acknowledge financial support of this
are NMR, IR, Raman, UV-vis, emission,
work by the Swiss Federal Office for En-
cyclic voltammetry, high-performance liq-
ergy (OFEN) and the Institute for Applied
uid chromatography (HPLC) and X-ray
Photovoltaics (INAP), Gelsenkirchen, Ger-
crystallography.
many. We thank Drs. R. Humphry-Baker,
5.2.9 S. M. Zakeeruddinn, F. P. Rotzinger and
Conclusion K. Kalyanasundaram for their valuable
discussions. Authors are indebted to
By appropriate synthetic methodology, it Dr. P. Infelta for his assistance in handling
has been possible to prepare a wide range various applications during the prepara-
of complexes and to examine the concept of tion of this chapter.
tunability of their photophysical and redox
properties. Different synthetic methods to References
introduce diverse ligands, which have been
judiciously selected to reconcile the tasks 1. R. D. McConnell, Future Generation Photo-
voltaic Technologies, AIP Conference Proceed-
of the sensitizer, to afford vectorial elec- ings 404, Denver, 1997.
tron injection into the semiconductor and 2. H. W. Vogel, Ber. Dtsch. Chem. Ges. 1873, 6,
efficient solar light harvesting, were dis- 1302.
cussed. Analysis of the acid-base properties 3. R. Gleria, R. Memming, Z. Phys. Chem. 1976,
of transition metal complexes contain- 98, 303.
4. W. D. K. Clark, N. Sutin, J. Am. Chem. Soc.
ing ligands with protonatable functional 1977, 99, 4676.
groups provide important information 5. A. Hamnett, M. P. Dare-Edwards, R. D.
about the nature of binding properties Wright et al., J. Phys. Chem. 1979, 83, 3280.
on the oxide surface. Comparison of the 6. M. Grätzel, Prog. Photovoltaics 2000, 8, 171.
ground state and excited state pKa s provide 7. S. Y. Huang, G. Schlichthörl, A. J. Nozik
et al., J. Phys. Chem. B 1997, 101, 2576.
useful data for the identification of the lig-
8. J. van de Lagemaat, N.-G. Park, A. J. Frank,
and involved in the lowest energy MLCT J. Phys. Chem. 2000, 104, 2044.
transition, which is essential in the design 9. R. Argazzi, C. A. Bignozzi, G. M. Hassel-
of photosensitizers for nanocrystalline so- mann et al., Inorg. Chem. 1998, 37, 4533.
lar cells. The effects exerted by the proton 10. K. Murakoshi, G. Kano, Y. Wada et al., J.
Electroanal. Chem. 1995, 396, 27.
content of the sensitizer on both the short-
11. A. Zaban, S. Ferrere, J. Sprague et al., J. Phys.
circuit photocurrent and open-circuit pho- Chem. B 1997, 101, 55.
tovoltage of dye-sensitized nanocrystalline 12. H. Sughihara, L. P. Sing, K. Sayama et al.,
solar cells were also shown. The molecular Chem. Lett. 1998, 1005.
13. C. Nasr, S. Hotchandani, P. V. Kamat, J. 39. S. M. Zakeeruddin, M. K. Nazeeruddin,

Phys. Chem. B 1998, 102, 4944. P. Pechy et al., Inorg. Chem. 1997, 36, 5937.
14. A. Solbrand, A. Henningsson, S. Södergren 40. M. K. Nazeeruddin, P. Pechy, M. Grätzel,
et al., J. Phys. Chem. B 1999, 103, 1078. Chem. Commun. 1997, 1705.
15. K. Tennakone, G. R. R. A. Kumara, I. R. M. 41. M. K. Nazeeruddin, R. Humphry-Baker,
Kottegoda et al., Chem. Commun. 1999, 15. P. Pechy et al., 10th International Conference
16. G. A. Crosby, Acc. Chem. Res. 1975, 8, 231. on Photochemical Conversion and Storage of
17. T. J. Meyer, Acc. Chem. Res. 1978, 11, 94. Solar Energy, (Interlaken, Switzerland), 1994,
18. A. Juris, V. Balzani, F. Barigelletti et al., Co- pp. 201.
ord. Chem. Rev. 1988, 84, 85. 42. P. -H. Xie, Y. J. Hou, B. W. Zhang et al., J.
19. K. Kalyanasundaram, Coord. Chem. Rev. Chem. Soc., Dalton Trans. 1999, 4217.
1982, 46, 159. 43. R. Argazzi, C. A. Bignozzi, T. A. Heimer
20. R. Krause in Structure and Bonding, Vol. 67, et al., Inorg. Chem. 1994, 33, 5741.
Springer-Verlag, Berlin, Heidelberg, 1987, 44. M. K. Nazeeruddin, S. M. Zakeeruddin,
pp. 1–52. R. Humphry-Baker et al., Inorg. Chim. Acta
21. K. Kalyanasundaram, M. Grätzel, Coord. 1999, 296, 250.
Chem. Rev. 1998, 77, 347. 45. M. A. Masood, B. P. Sullivan, D. J. Hodges,
22. M. K. Nazeeruddin, A. Kay, I. Rodicio et al., Inorg. Chem. 1994, 33, 5360.
J. Am. Chem. Soc. 1993, 115, 6382. 46. B. Durham, S. R. Wilson, D. J. Hodges et al.,
23. J. J. Rack, H. B. Gray, Inorg. Chem. 1999, 38, J. Am. Chem. Soc. 1980, 102, 600.
2. 47. M. K. Nazeeruddin, S. M. Zakeeruddin,
24. H. Taube, Angew. Chem., Int. Ed. Engl. 1984, R. Humphry-Baker et al., Coord. Chem. Rev.
23, 329. 2000, 208, 213.
25. P. S. Braterman, J. I. Song, R. D. Peacock, 48. A. B. P. Lever, Inorg. Chem. 1990, 29, 1271.
Inorg. Chem. 1992, 31, 555. 49. J. M. Clear, J. M. Kelly, C. M. O’Connell
26. E. M. Kober, T. J. Meyer, Inorg. Chem. 1983, et al., J. Chem. Soc., Chem. Commun. 1980,
22, 1614. 750.
27. A. Hagfeldt, M. Grätzel, Chem. Rev. 1995, 95, 50. D. J. Cole-Hamilton, J. Chem. Soc., Chem.
49. Commun. 1980, 1213.
28. D. Cahen, G. Hodes, M. Grätzel et al., J. 51. D. S. C. Black, G. B. Deacon, N. C. Thomas,
Phys. Chem. 2000, 104, 2053. Inorg. Chim. Acta 1982, 65, L75.
29. D. M. Roundhill in Photochemistry and Pho- 52. I. P. Evans, A. Spencer, G. Wilkinson, J.
tophysics of Metal Complexes, Plenum Press, Chem. Soc., Dalton Trans. 1973, 204.
New York, 1994. 53. K. A. Maxwell, M. Sykora, J. M. DeSimone
30. S. Ernst, W. Kaim, Inorg. Chem. 1989, 28, et al., Inorg. Chem. 2000, 39, 71.
1520. 54. S. M. Zakeeruddin, M. K. Nazeeruddin,
31. M. X. Tan, P. E. Laibinis, S. T. Nguyen et al., R. Humphry-Baker et al., Inorg. Chem. 1998,
Prog. Inorg. Chem. 1994, 44, 21. 37, 5251.
32. M. K. Nazeeruddin, E. Muller, R. Humphry- 55. S. M. Zakeeruddin, M. K. Nazeeruddin,
Baker et al., J. Chem. Soc., Dalton Trans. 1997, P. Péchy et al., Chem. Mater., communicated.
4571. 56. D. Wöhrle, D. Meissner, Adv. Mater. 1991, 3,
33. M. K. Nazeeruddin, K. Kalyanasundaram, 129.
M. Grätzel, Inorg. Synth. 1997, 32, 181. 57. C. D. Jaeger, F. F. Fan, A. J. Bard, J. Am.
34. V. Balzani, A. Juris, M. Venturi et al., Chem. Chem. Soc. 1980, 102, 2592.
Rev. 1996, 96, 759–833. 58. A. Giraudeau, F. Ren, F. Fan et al., J. Am.
35. K. Kalyanasundaram, M. K. Nazeeruddin, Chem. Soc. 1980, 102, 5137.
Chem. Phys. Lett. 1992, 193, 292. 59. A. Giraudeau, F. F. Fan, A. J. Bard, J. Am.
36. P. A. Anderson, G. F. Strouse, J. A. Treadway Chem. Soc. 1980, 102, 5137.
et al., Inorg. Chem. 1994, 33, 3863. 60. H. Yanagi, S. Chen, P. A. Lee et al., J. Phys.
37. S. Ruile, O. Kohle, P. Péchy et al., Inorganica Chem. 1996, 100, 5447.
Chimica Acta 1997, 261, 129. 61. M. K. Nazeeruddin, R. Humphry-Baker,
38. O. Kohle, S. Ruile, M. Grätzel, Inorg. Chem. B. A. Murrer et al., J. Chem. Soc., Chem. Com-
1996, 35, 4779. mun. 1998, 719.
62. M. K. Nazeeruddin, R. Humphry-Baker, 80. S. G. Yan, J. T. Hupp, J. Phys. Chem. 1996,

M. Graetzel et al., J. Porphyrins and Phthalo- 100, 6867.
cyanines 1999, 3, 230. 81. G. Will, G. Boschloo, S. Nagaraja Rao et al.,
63. S. Ferrere, Chem. Mater. 2000, 12, 1083. J. Phys. Chem. B 1999, 103, 8067.
64. T. A. Heimer, C. A. Bignozzi, G. J. Meyer, J. 82. B. Lemon, J. T. Hupp, J. Phys. Chem. B 1999,
Phys. Chem. 1993, 97, 11987. 103, 3797.
65. M. Alebbi, C. A. Bignozzi, T. D. Heimer 83. B. Jing, H. Zhang, M. Zhang et al., J. Mater.
et al., J. Phys. Chem. 1998, 102, 7577. Chem. 1998, 8, 2055.
66. G. Sauvé, M. E. Cass, S. J. Doig et al., J. Phys. 84. S. Anderson, E. C. Constable, M. P. Dare-
Chem. 2000, 104, 3488. Edwards et al., Nature 1979, 280, 571.
67. C. A. Bignozzi, M. T. Indelli, F. Scandola, J. 85. K. Murakoshi, G. Kano, Y. Wada et al., J.
Am. Chem. Soc. 1989, 111, 5192. Electroanal. Chem. 1995, 396, 27.
68. Y. Lei, T. Buranda, J. F. Endicott, J. Am. 86. K. S. Finnie, J. R. Bartlett, J. L. Woolfrey,
Chem. Soc. 1990, 112, 8820. Langmuir 1998, 14, 2744.
69. C. A. Bignozzi, R. Argazzi, C. Chiorboli et al., 87. A. Fillinger, B. A. Parkinson, J. Electrochem.
Coord. Chem. Rev. 1991, 111, 261. Soc. 1999, 146, 4559.
70. G. Denti, S. Serroni, S. Campagna et al., in 88. R. Argazzi, C. A. Bignozzi, T. A. Heimer
Perspectives in Coordination Chemistry, (Eds.: et al., Inorg. Chem. 1997, 36, 2.
A. F. Williams, C. Floriani, A. E. Merbach), 89. P. Péchy, F. P. Rotzinger, M. K. Nazeeruddin
Helvetica Chimica Acta Verlag and VCH, et al., J. Chem. Soc. Chem. Commun. 1995, 65.
Basel, Switzerland and Weinheim, Germany, 90. M. K. Nazeeruddin, S. M. Zakeeruddin,
1992, p. 153. R. Humphry-Baker et al., Inorg. Chem. 1999,
71. K. Kalyanasundaram, M. K. Nazeeruddin, 38, 6298.
Inorg. Chim. Acta 1994, 226, 213. 91. K. Kalyanasundaram, M. K. Nazeeruddin,
72. K. Matsui, M. K. Nazeeruddin, R. Humphry- Inorg. Chem. 1990, 29, 1888.
Baker et al., J. Phys. Chem. 1992, 96, 10 587. 92. M. K. Nazeeruddin, S. M. Zakeeruddin,
73. F. Scandola, M. T. Indelli, C. Chiorboli et al., R. Humphry-Baker et al., Inorg. Chem. 2000,
Top. Curr. Chem. 1990, 158, 73. 39, 4542.
74. M. K. Nazeeruddin, M. Grätzel, K. Kalyana- 93. M. Montalti, S. Wadhwa, W. Y. Kim et al.,
sundaram et al., J. Chem. Soc., Dalton Trans. Inorg. Chem. 2000, 39, 76.
1993, 323. 94. M. Amirnasr, M. K. Nazeeruddin, M. Grätzel,
75. M. K. Nazeeruddin, P. Liska, J. Moser et al., Thermochim. Acta 2000, 348, 105.
Helv. Chim. Acta 1990, 73, 1788. 95. P. A. Anderson, G. F. Strouse, J. A. Treadway
76. M. K. Nazeeruddin, R. Humphry-Baker, M. et al., Inorg. Chem. 1994, 33, 3863.
Grätzel, J. Phys. Chem., communicated. 96. F. P. Dwyer, J. E. Humpoletz, R. S. Nyholm,
77. K. Kalyanasundaram, M. Grätzel, M. K. J. Proc. Roy. Soc. N. S. Wales. 1946, 80, 212.
Nazeeruddin, Inorg. Chem. 1992, 31, 5243. 97. D. Rose, G. Wilkinson, J. Chem. Soc. A 1970,
78. C. A. Bignozzi, R. Aragazzi, C. J. Kleverlaan, 1791.
Chem. Soc. Rev. 2000, 29, 87. 98. R. Krause in Structure and Bonding, Vol. 67,
79. S. A. Trammell, J. A. Moss, J. C. Yang et al., Springer-Verlag, Berin, Germany, 1987,
Inorg. Chem. 1999, 38, 3665. pp. 1–52.
5.3 Dye-sensitized systems are not suited

Charge Transport in Dye-sensitized to interpretation by conventional device
Systems physics methods on account of two fea-
tures: firstly, the mesoscopic phase sep-
Jenny Nelson
aration of electron and hole conductors,
Imperial College of Science, Technology and
which makes the porous material unable
Medicine, London, United Kingdom
to sustain large electric fields; and sec-
5.3.1 ondly, the separation, through the use
Introduction of sensitizers, of optical absorption from
charge transport in either material. Efforts
5.3.1.1 Dye-sensitized Systems for to understand the photovoltaic action of
Photovoltaics the DSSC are leading to a reassessment
In this chapter, ‘‘Dye-sensitized System’’ of basic principles and the possibilities of
is used to refer to the electrochemical novel photovoltaic designs.
junction, or heterojunction between a
porous, wide band gap semiconducting 5.3.1.2 Dye-sensitized Systems for Other
film, which has been sensitized for visible Applications
light by an adsorbed dye species and Other applications for dye-sensitized sys-
a hole conducting medium. By far the tems are emerging. A particular example
most important application is the dye- is the dye-sensitized photoelectrochromic
sensitized solar cell (DSSC), which has window, where the absorption of light in
been responsible for much of the research a dye-sensitized electrode leads to charg-
aimed at understanding and optimizing ing of a counter electrode, resulting in
these systems. [1–3]. a change in its optical density [5, 6].
Charge transport in DSSCs is partic- Novel, indirectly related applications are
ularly intriguing because of the long emerging in biotechnology, where the sen-
lifetimes and long diffusion lengths for sitization of semiconducting films with
charge-separated species in these systems. proteins, leads to an electronic signal in re-
In other photovoltaic devices, based on sponse to the biochemical action of the pro-
organic materials, diffusion lengths for tein [7]. Charge transport in mesoporous
charge-separated species are typically tens systems is relevant to non-dye-sensitized
or hundreds of nm, severely limiting the applications, such as lithium intercalation
quantum efficiency. In the DSSC, photo- batteries [8] and electrochromics [9].
carrier diffusion lengths are of the order Another aspect is the use of sensitized
of microns, in spite of the proximity of systems to study the properties of either
positive and negative charge carriers. The the sensitizer or the film. For example,
improvement is in part due to the func- the use of transparent porous TiO2 films
tion of the sensitizer, which separates the to study the optoelectronic properties of
process of photoinduced charge separa- PbS quantum dots (QD) [10] and the use
tion from the main recombination process. of sensitizers to study electron dynamics
These features give the DSSC its remark- in the nanocrystalline film [11].
ably high light to electric current efficiency This chapter is set out as follows: in
first reported in 1988 [4], and has enabled Sect. 5.3.2 we summarize the structure
the development of 10% efficient solar and function of a dye-sensitized junction;
cells. Sect. 5.3.3 reviews the material properties
5.3 Charge Transport in Dye-sensitized Systems 433
of the electron and hole conducting com- cell at operating point, light forward biases
ponents; Sect. 5.3.4 reviews some basic the junction to stimulate a reverse current,
theoretical principles and attempts to de- which opposes the photocurrent. In a dye-
velop a ‘‘device physics’’ for dye-sensitized sensitized device, unusually, this reverse
devices; Sect. 5.3.5 reviews the experimen- current contains both light and dark com-
tal technique for studying charge transport ponents. The key charge transfer reactions
and discusses the main results and the are illustrated in Fig. 1.
models that have been proposed; finally
some points of consensus and outstanding Key Processes
questions are summarized in Sect. 5.3.6. Light absorption:
5.3.2 hν + D −−−→ D∗ (1)

Structure and Function of Dye-sensitized
Junctions: An Overview Light-induced forward electron trans-
fers:
5.3.2.1 General Structure of
Dye-sensitized Electrochemical Junction D∗ −−−→ e(sc) + D+ (2)
Dye-sensitized devices are based on an +
HCRed + D −−−→ HCOx (3)
electrochemical junction or heterojunc-
tion, containing three media: electron transport of e(sc) (4)
conductor, sensitizer, and hole conduc- transport of HCRed and HCOx (5)
tor. Light absorbed in the sensitizer causes
the injection of an electron into the elec- e(sc) −−−→ e(n) (6)
tron conductor and a positive charge into e(p) + HCOx −−−→ HCRed (7)
the hole conductor. Electron and hole con-
ductor are separately connected through Dark reverse electron transfers:
selective contacts to the terminals of the e(sc) + HCOx −−−→ HCRed (8)
device. Charges successfully reaching the
external circuit as a result of illumination e(n) −−−→ e(sc) (6∗ )
constitute the photocurrent. In addition to HCRed −−−→ HCOx + e(p) (7∗ )
light-induced charge transfers, we need
Light induced reverse electron transfers:
to consider bias-induced, dark charge-
transfer reactions. In the case of a solar e(sc) + D+ −−−→ D (9)
hν 1
8 6 2 3 9
8* 4 7 9*
1
hν
2
Fig. 1 Key electron transfer 5
reactions in a dye-sensitized Electron conductor Hole conductor
device. n p
where D, D∗ , and D+ refer to the dye, the the continual photoionization and regen-
excited state of the dye, and the dye cation; eration of the sensitizer under steady state
e(sc), e(n), and e(p) refer to the electron in illumination. In energetic terms, the Gibbs
the semiconductor (electron conductor), in free energy of the excited sensitizer, D∗ ,
the n-contact and in the p-contact material; plus the vacant electron acceptor state in
and HCRed and HCOx refer to the reduced the semiconductor, should exceed that of
and oxidized state of the hole conductor. the dye cation plus the occupied acceptor
Following charge transfer, electrons and state, D+ + e(TiO2 ). This is usually rep-
holes are transported through the respec- resented on a simplified potential energy
tive mediums by electron conduction (6) diagram as a sensitizer excited state of
and hole conduction (7). We have assumed potential energy higher than the conduc-
that hole transport is more appropriately tion band of the semiconductor. Similarly,
described as a redox process and electron the redox potential µRedOx of the redox
transport as a simple electron conduction. couple in the hole conductor should be
In discussing the net transport of elec- sufficient to reduce the dye cation. This
trons, we need to include processes (4), is usually represented as a dye ground
(5), and (6), whereas net hole transport state of potential energy which is lower
involves (3), (4), and (7). than the redox potential of the electrolyte.
In addition, there is the energy transfer These relationships are illustrated in Fig. 2
process because of relaxation of the excited together with the energetically preferred
state of the dye directions for charge transfer. Such po-
tential energy diagrams do not normally
D∗ −−−→ D + hν or heat (1∗ ) distinguish the entropic contribution to
the driving force, which arises from the
which does not result in a forward electron much greater density of states per unit sur-
transfer and may be counted with the face area in the semiconductor than in the
reverse electron transfers as a loss process. dye. This effect is estimated to contribute
We define forward as the direction of up to 0.1 eV to the driving force [12]. Con-
charge transfer under photovoltaic oper- tact materials should allow facile transfer
ating conditions, and will use ‘‘forward’’ of electrons from electron conductor and
bias to mean the direction of potential dif- positive charge from electrolyte.
ference in those conditions (p terminal
positive with respect to n). Note that this 5.3.2.3 Physical Structure and Materials
condition corresponds to negative applied In the most commonly studied configura-
potentials in electrochemical terms. tion of the DSSC, the electron conductor
is a wide band gap, nanocrystalline, metal
5.3.2.2 Energetic Structure oxide film. Colloidal titanium dioxide is
In a functioning dye-sensitized device, the most often used although other wide band
net current output is the sum of all the gap metal oxides are candidates. The sen-
forward electron transfer processes less all sitizer is a transition metal-based (usually
of the reverse processes. For a net forward ruthenium), organic dye, with excited state
current under (steady state) illumination, free energy sufficient to reduce the semi-
processes (2) and (3) must be kinetically conductor, and containing ligands, such
more favorable than (4), and (3) must be as carboxylates or phosphonates, which
preferred to (5). These conditions enable facilitate bonding to the semiconductor
D*/D +
E cb
Ef Photovoltage
Energy
µ RedOx
D/D +
E vb
Semiconductor Sensitiser Hole conductor
− +
− +
Distance
Fig. 2 Potential energy diagram of DSSC. For net forward electron transfer the
oxidation potential of the dye-excited state, D∗ /D+ should be higher than the
semiconductor conduction band, and the redox potential of the hole conductor
should be higher than the dye-ground state potential D/D+ . Absorption of light by
the dye causes charge separation across the interface, resulting in a splitting of the
electrochemical potential and hence a photovoltage.
surface. The hole conductor is tradition- 5–20 nm in size are sintered together
ally a polar liquid electrolyte, such as into a random porous network, leaving
propylene carbonate or acetonitrile, con- a network of voids of similar volume
taining the iodide–triiodide redox couple, fraction and dimensions. In the DSSC,
although latterly solid-state ionic and hole the dye is adsorbed on to the entire
conductors have been substituted. In many film surface, whereas the voids are filled
cases, the electrolyte contains a lithium with the hole conductor. The enormous
salt, which facilitates electron accumula- microscopic surface area, which is typically
tion in the TiO2 and apparently improves 1000 times the geometrical surface area for
charge separation [13]. In the TiO2 –liquid a 10-µm-thick film, leads to a much higher
electrolyte system, ohmic contact is made optical depth than for a planar interface,
to the oxide through a fluorine doped SnO2 whereas the pores allow redox ions to
conducting glass substrate, and to the elec- travel between all points on the surface
trolyte through a platinum coated counter and the counter electrode [2]. Figure 4
electrode, which catalyzes the regeneration illustrates the typical device structure of
of the redox couple. Material properties the DSSC.
and requirements are described here and
further details are given in Refs. 2, 14. 5.3.2.4 Charge Carriers and Charge
A unique feature of DSSCs is that Transfer Processes
the electron conductor (the oxide film) is The charge transfer processes are sum-
both porous and electronically connected. marized here. In the dark, at equilibrium,
The morphology of a typical TiO2 film there is no net charge transfer. When a
is illustrated in Fig. 3. Crystallites of forward bias (negative potential) is applied
Fig. 3 Scanning electron

micrograph of film surface
(from Ref. 15).
300 nm
Dye layer Electrolyte Platinum layer

Glass
+ + + + + +
+V
0 Seal
− − − − − −
Glass
TiO2 film SnO2 layer

3I − I 3− + 2e
e−
Fig. 4 Schematic illustration of arrangement of dye-sensitized PV cell.

Electron conduction to the conducting glass contact proceeds through the
TiO2 , whereas hole conduction by the liquid electrolyte proceeds through the
pores of the TiO2 film.
in the dark, electrons flow from substrate Probable intermediate reaction schemes
into the film and oxidized species from are detailed in Ref. 16 and the solution
counter electrode toward the interface. phase redox couple has been probed in
Electrons and oxidized species combine depth [17]. In the steady state, ion popu-
at the interface through process (4), which lations are maintained by the oxidation of
in the iodide–triiodide redox couple can I3 − at the counter electrode, which deliv-
be summarized as ers the electron to the external circuit. This
1 3 recombination process is responsible for
eTiO2 + I3 − −−−→ I − . (10) the dark current from the junction.
2 2
Under illumination, an absorbed photon The primary species involved in charge

results in the injection of an electron transport are thus electrons in the ox-
into the oxide film (2). The electron ide film and the reduced and oxidized
is transported through the film to the forms of the hole conductor. The ex-
substrate (6). The reduced form of the tremely high surface area means that
hole conductor (I − ) restores the electron interfacial electron transfer reactions and
to the dye cation (3), summarized by the electrostatic interactions between species
reaction in the two conducting media are particu-
larly important. In addition, the following
3 − 1 secondary species may influence charge
I + D+ −−−→ D + I3 − . (11)
2 2 transport:
The reaction may be detailed by the • Cations in the hole conductor. These in-
sequence: fluence the electrostatics of the system
by reorganizing in response to motion
D+ + I − −−−→ D + I • of electrons in the oxide film (which
I • + I − −−−→ I2 •− may improve electron mobility), and by
screening redox species from electrons.
2I2 •− −−−→ I3 − + I − • Intercalated or surface adsorbed species in
the electron conductor. Protons and Li
The (I3 − ) is transported to the counter ions, if present in the hole conductor,
electrode (7) and is there reduced to re- may be capable of intercalating into the
generate I − through the donation of oxide film, affecting both the potential
electrons from the external circuit. This energy for electron stabilization and the
charge transfer constitutes the photocur- ease of electronic conduction. Surface
rent. Some electrons are lost through absorbed cations influence electronic
recombination with the oxidized form of states near the surface. Adsorption
the hole conductor (4) or the dye cation (5) of OH− groups influences electronic
before being collected. Both recombina- states and proton transport. Water
tion processes amount to the light-induced molecules, which are expected to be
recycling of charge around the inter- present around the oxide surface even
face. Both are accelerated by the light in dry solvent, may be dissociatively
driven forward biasing of the junction, adsorbed leaving adsorbed OH species
leading to kinetic competition for the and mobile protons, which influence
dye cations ((3) versus (5)) and, effectively, electronic states and conductivity.
for the photoinjected electrons ((4) and (5) • Charge hopping between dye cations. To a
versus transport, (6)). Note that, for the io- first approximation, dyes are normally
dide–triiodide redox couple, the reaction assumed to be immobilized on the
schemes for (3) and (4) are not necessar- surface and function only to inject
ily first order and cannot be described by charge to other media. However, there
a rate constant. The rates of these pro- is limited evidence for hole migration
cesses depend on the concentration of dye between dyes, as discussed later.
cations (i.e. illumination) and electrons • Holes. In the case of band gap illu-
(i.e. applied bias), leading to illumination mination, holes will be generated in
and bias dependent reaction kinetics. the electron conductor. These may act
as recombination centers for electrons the porous material. This leaves two re-
(a loss process), or may be scavenged gions where the potential can be dropped:
by the hole conductor (a photovoltaic at the interface between the TiO2 film
effect). However, the wide band gap and the SnO2 : F (fluorine doped SnO2 )
means that holes are seldom generated, rear conducting contact; and across the
and except in the case of ultraviolet (UV) interface layer between film and elec-
irradiation, can be ignored. trolyte, which includes outer atomic layers
of film and Helmholtz layer. It is most
5.3.2.5 Potential Distribution likely that the potential drop is shared be-
To sustain the forward electron transfer tween the two interfaces [19, 20]. Bisquert
described earlier, the hole-conducting ma- has shown by solving Poisson’s equation
terial should have a lower effective work for a columnar model of the film that
function than the electron conductor. As in the largest electric fields within the film
any electronic junction, the electrochem- are established at the interface with the
ical potential or Fermi level in the two SnO2 substrate [19]. Large fields cannot
materials aligns at equilibrium, leaving a extend beyond a few nanoparticle radii
potential difference equal to the difference from the rear substrate (Fig. 5). The same
in work functions to be accommodated in conclusion was reached earlier with more
the two materials. In a conventional elec- qualitative arguments in Refs. 21, 22. The
trochemical junction, the charge density proportion of the potential difference that
in the layer of electrolyte close to the inter- can be taken up by the film and rear sub-
face (the Helmholtz layer) is much greater strate is determined by morphology and
than the fixed charge density in the semi- materials properties; the remainder must
conductor, and the potential difference is be taken up by the Helmholtz layer [23,
accommodated mainly in the semiconduc- 12], leading to large electric fields at the in-
tor by band bending. In a lightly doped terface, extending only a few nm into the
semiconductor, this band bending extends electrolyte [24], and a film which is largely
over a depletion region some microns in depleted of electrons in equilibrium. The
width, beyond which the semiconductor is size of the potential drop at the Helmholtz
neutral. layer is believed to be influenced by dye ad-
In the porous geometry of the DSSC, sorption, which polarizes the surface layer
the situation is different. The high charge by introducing protons, and by the redox
density of the electrolyte means that the potential of the electrolyte. Under applied
electrolyte tends to maintain the outer sur- reverse bias, the potential drop increases
face of the metal oxide film at a constant and an increasing proportion is taken up
potential, and individual nanoparticles in the Helmholtz layer. When the junction
making up the film are too small to sustain is forward biased either electrically or by
the junction potential difference. Calcu- illumination, the concentrations of both
lations using the Albery–Bartlett model electrons in the film and oxidized species
indicate that only some 50 mV of poten- (‘‘holes’’) in the electrolyte increase. The
tial can be dropped between the surface applied bias will be taken up, in general, by
and center of a 10-nm radius particle of both the TiO2 –TCO (transparent conduct-
(lightly n-type) TiO2 [18–20]. Therefore, if ing oxide) and the TiO2 –hole conductor
the film surface is an equipotential, large interface layers, reducing the potential dif-
electric fields cannot be sustained within ference and modifying the electric field
at those interfaces. However, according to charge on the TiO2 electrode will attract
Cahen and coworkers, neither potential cations toward the film surface, where they
drop is sufficient on its own to account act to screen and stabilize electrons in the
for the open circuit voltage of the cell, film. Very small cations such as Li+ , which
and this is instead controlled by the dif- are able to intercalate into the interstices
ference between the redox potential of the of the crystal structure, may be drawn
electrolyte and the conduction band of the in. These processes clearly modify the
TiO2 [12]. This is an important distinction electrostatic potential distribution. In the
from conventional photovoltaic devices, steady state, they appear as changes in the
where the difference in work functions effective transport parameters; in transient
controls Voc . It is supported by studies by studies, currents due to ion movement
Pichot and Gregg of the photovoltage of may be resolved [26].
cells deposited on different work function
substrates, which indicate that the photo- 5.3.2.6 Relevance to Device Performance
voltage is not influenced by the potential Key requirements for good performance in
drop at the TiO2 –TCO interface [25]. any photovoltaic device are high light ab-
A key consequence of the potential sorption, efficient charge separation, and
distribution in the cell is that once efficient charge transport. In the DSSC,
separated, the electrons and holes in their light absorption is ensured by the large
respective media do not experience large optical depth of the sensitized porous film.
electrostatic forces because of the junction, Efficient charge separation in DSSC is
and are not driven primarily by built-in provided by the much faster rate of in-
electric fields. jection (2) relative to excited state decay,
An additional effect of applied bias is the and is discussed elsewhere in this Vol-
reorganization of solvent ions to minimize ume. Efficient charge transport is therefore
the energy of the system. A more negative the key. Having identified the species and
Red
+
Ox Ox 10
+ 15
+ + 8
Red 6
φ 10
+ Red 5 4 z/a
+ Ox
0 2
−0.51
−
0
0
0.5
1
r/a
(a) (b)
Fig. 5 (a) Schematic illustration of film surface, electrostatic potential distribution for a
showing lines of constant potential. columnar electrode with constant surface
Reorganization of cations in the electrolyte potential, showing how significant potential
allows the screening of electrons in the film, drops can be maintained only within a few tens
reducing their electrostatic effect. (b) Calculated of nm from the rear substrate (from Ref. 20).
processes involved in charge transport in substrates coated with a fluorine doped

dye-sensitized systems, we may now iden- SnO2 conducting layer (SnO2 : F), which
tify a number of relevant issues for device is preferred to Indium-Tin Oxide (ITO)
performance: for the robustness of its conductive prop-
erties under heating. Alternative routes
• Recombination. Because of the porous to porous TiO2 films, more amenable to
structure, the interfacial electron trans- large-scale production, are being studied.
fer processes discussed earlier compete The particle synthesis and film prepara-
with transport at all times. This limits tion is discussed elsewhere [14, 29, 27], as
both the cell voltage and the cell current are the electronic and structural properties
at positive biases. of the material [30–34]. Here we will fo-
• Resistive losses. Current flow leads to gra- cus only the material and morphological
dients in the quasi Fermi levels. In the features relevant to charge transport.
light this causes a drop in quasi Fermi
level separation between interface and 1. TiO2 is a polar material. Valence band
terminals. Increased resistance through states are associated with the valence
charge trapping increases this loss. orbitals of the O2− ion, whereas
• Electrostatics. Mobile electrolyte ions re- conduction states are associated with
organize to effect charge screening and the 3d orbitals of the Ti4+ ions. Band
compensate charging of nanoparticles. gap excitation corresponds to a shift in
This may improve conduction but at electron density from O to Ti sites.
possible cost to the cell voltage. 2. Colloidal particles are crystalline and
• Leakage. This includes shunting of tend to adopt the anatase polymorph,
charges across the edges of the sam- rather than the more widespread
ple and charge transfer directly from and widely studied rutile (this has
working electrode to electrolyte. been confirmed by Raman and XRD
studies [27, 28]). The preference for
5.3.3 anatase at nanometer sizes has been
Material Properties attributed to surface tension argu-
ments [35]. Anatase is less dense than
5.3.3.1 Electron Conductor rutile, with a different arrangement
The porous semiconducting film used in of TiO6 octahedra in the unit cell,
the DSSC should be transparent to visi- which leaves open channels parallel
ble light, n-type, stable, and have a large to the c-axis, and may assist interca-
surface area. Many materials satisfy these lation [30, 34, 36]. Anatase has a band
requirements in principle, but in prac- gap of 3.2 eV in the bulk, slightly larger
tice TiO2 is by far the best studied and than rutile, which makes it superior
most widely used. Usually the porous films for dye-sensitized photovoltaic appli-
comprise colloidal particles of anatase cations (less loss through absorption
TiO2 some 10–20 nm in diameter assem- of blue light).
bled into a random porous network with 3. Random orientation of different crys-
a porosity of typically 50–60% [27, 28]. tallites leads to grain boundaries at
Particle and pore size can be varied by mod- the interfaces, which may reduce
ifications to synthesis and film deposition. electronic mobility through enhanced
Films are usually deposited on glass scattering, small cross sectional area
and, possibly, through space charge rutile, and exploited for applications
induced potential barriers. However in gas sensing [39].
quantum confinement effects are not 5. The enormous surface area leads to a
expected because the typical crystal- high-volume density of surface-related
lite radius greatly exceeds the exciton defects. Apart from oxygen vacancy
radius (∼2 nm [18]). Space charge ef- induced Ti3+ states, these include de-
fects are also expected to be small on fects due to surface adsorption of ions
account of the low doping density of and intercalated species, particularly
the crystallites. in the presence of an electrolyte. Sur-
4. The material is nonstoichiometric and face adsorbed cations may lead to
has a natural tendency to acquire intraband gap states, similar in na-
oxygen vacancies through loss of O2 , ture to the oxygen vacancy [40], and
in an oxygen-poor environment: intercalated lithium leads to a sim-
ilar intraband gap density of states,
TiO2 −−−→ TiO2−x + xVO + 12 xO2 as observed by photoemission spec-
(12) troscopy [41]. Such defects can trap
where VO represents an oxygen va- electrons. The profusion of defects
cancy and x the fraction of oxygen leads to the formation of a surface
vacancies in the film. The rate of O2 ef- layer of different electronic structure
fusion can be controlled by the partial to the bulk.
pressure of oxygen and temperature. 6. Under UV illumination, TiO2 is highly
Considered as a point defect, the oxy- hydrophilic and attracts a surface layer
gen vacancy contains two electrons in of hydrogen bonded H2 O molecules.
its ground state but is readily ionized to In certain conditions [42, 43], the
deliver one or both electrons to empty water dissociates to leave surface
3d orbitals on adjacent Ti sites, chang- adsorbed OH groups – which may
ing the Ti oxidation state from Ti4+ to create electron traps – and itinerant
Ti3+ and increasing the population of protons. Protons intercalate easily and
electrons available for conduction [30]. can strongly influence conductivity
The electron in the Ti3+ defect state by assisting the formation of Ti3+
is localized but easily accessible to states [19, 44].
the conduction band. Calculations [37] 7. Ti3+ defect states have a characteris-
and experiment [38] place its energy tic red-infrared optical absorbance [45,
around 2–3 eV above the valence band 46, 19]. This can be used as a measure
edge in the case of rutile. Function- of the density of electrons accumu-
ally, these states are important because lated in the film [47–50]. Although
electrons trapped in these defect lev- there is some debate about whether
els are readily promoted to conduction the states giving rise to the absorbance
band states, where they contribute to are delocalized conduction band states
electrical conduction, whereas unoc- or intraband trap states, the evidence
cupied states act as traps for conduc- favors the latter [19, 51].
tion electrons. Oxygen vacancies thus 8. The nanocrystalline films are insu-
effectively dope the material n-type. lating in normal conditions, with a
The control of conductivity through dark resistivity of 108 –1010 " cm in
stoichiometry is well established for air. This is substantially larger than
the dark resistivity of single crystal or determining an electronic mobility.

polycrystalline anatase [32, 52] or ru- Dittrich and coworkers find a mo-
tile [53]. Conductivity can be increased bility of around 10−5 cm2 V−1 s−1
by reducing the films by heating in at room temperature in an oxy-
vacuum (introducing oxygen vacan- genic atmosphere, using time of
cies) [32, 54] or by doping with lithium flight techniques [59]. Konenkamp
or other electron donors. There also and Hoyer [60] use junction re-
appears to be some sensitivity to am- covery transients to derive a mo-
bient water content, indicating a pos- bility lifetime product of around
sible contribution from proton dop- 10−10 cm2 V−1 , which leads to in-
ing or protonic conductivity [44, 55]. jection dependent mobilities of
Definitive results on the conductiv- 10−7 –10−5 cm2 V−1 s−1 . These val-
ity of ‘‘dry’’ films is rare, given the ues are substantially smaller than
high resistivity and sensitivity to am- the electronic mobility in crystalline
bient conditions. However, it appears anatase of around 1 cm2 V−1 s−1 [52].
that conduction is thermally activated
with an activation energy between It is apparent that trap states influence
0.8 eV [56] and 1.6 eV [57]. Loss of conduction in several ways: through dis-
thermal activation in strongly reduced persive photocurrent and dark current
polycrystalline anatase has been at- transients; strongly space charge limited
tributed to a metal-insulator like tran- transport [59]; and sub–band gap photo-
sition, which is specific to anatase [32], conductive and photovoltaic effects.
although the equivalent observation in
nanocrystalline films is yet to be made. 5.3.3.1.1 ZnO and Other Oxides Alterna-
Because the material is polar, charge tive electron conductors to TiO2 have been
transport within crystallites is likely to considered. For efficient function, the con-
be by small polaron hopping. In rutile, duction band edge should lie between the
an activation energy of around 60 meV conduction band of the substrate and the
is observed and attributed to the hop- excited state potential energy of the dye. A
ping barrier [53]; similar observations wide range of oxides are potential materials
for nanocrystalline films have not been but only a few (ZnO, SnO2 , and SrTiO3 )
reported. have been studied for dye-sensitized ap-
9. Under band gap illumination the plications. Some oxides, such as SrTiO3
films become photoconducting, with offer the possibility of higher open-circuit
DC conductivity increasing by sev- voltages through closer alignment of sen-
eral orders of magnitude [2, 57]. Sub- sitizer excited state levels with conduction
band gap illumination causes a small band edge [61]. Others such as ZrO2 , are
photocurrent [55] and photovoltaic ef- ruled out by a low electron affinity, which
fect [58] in dry films and a small pho- places the conduction band edge above the
toconductivity in electrolytically con- excited state potential of most dyes. This is
tacted films [19], suggesting excitation illustrated in Fig. 6.
of carriers out of trap states. By far the best studied of these al-
10. Transient photocurrent and dark ternative materials is ZnO [18, 63–65].
current measurements are strongly Nanocrystalline ZnO can be prepared by
dispersive, leading to difficulties in sol-gel routes and sintered into porous
D*/D+
E c.b.
3.2 eV
D/D+
3.2 eV
3.2 eV
3.4 eV ZrO2
SrTiO3
ZnO TiO2
E v.b.
SnO2
Fig. 6 Schematic energy band diagram of suitable and unsuitable metal

oxides for dye sensitization. Electron injection from dye-excited state to
conduction band is energetically favorable for SnO2 , ZnO, and SrTiO3 , as
well as TiO2 , but forbidden for ZrO2 [2, 62].
films on TCO substrates (usually SnO2 : F edge (Burstein shift) can be detected
or ITO) by similar procedures. It is sim- as an indicator of electron accumula-
ilar in band gap (3.2 eV) and band edge tion [18, 63].
position to TiO2 [2] with similar (13 nm di- 3. The exciton radius is, at around 5 nm,
ameter) or smaller (5 nm) crystallite sizes larger than in TiO2 (on account of
than for typical TiO2 films. As with TiO2 , a smaller effective mass), and may
it may be self doped n-type through oxygen lead to quantum confinement effects
deficiency. in nanocrystalline films [18, 65, 67].
Charge transport in nanocrystalline ZnO These affect the ease of transport
is less well studied than in TiO2 , but we between particles, leading to barriers
may make a few relevant comments: between large and small particles for
polydisperse particle size distributions.
1. Nanocrystalline ZnO is highly nonsto- 4. There is some evidence for a barrier at
ichiometric with a defect density of up the ZnO/ITO contact [67].
to 1018 cm−3 . This leads to a higher 5. ZnO is not capable of intercalating Li
electronic mobility than TiO2 , of over ions.
10−3 cm2 V−1 s−1 [65].
2. The oxygen vacancy defects appear to 5.3.3.1.2 p-type Metal Oxides Apart from
function as deep traps for holes and re- the sensitization of n-type semiconduc-
combination centers for electrons and tors with electron injecting dyes, there
shallow traps for electrons [66] (the is, in principle, the possibility of sensi-
limited dispersion of time constants tization of a p-type semiconductor with
in transient current experiments has a hole-injecting dye. The requirement for
been attributed to the absence of deep hole injection is that the potential of the
electron traps). No optical absorbance ground state of the dye lies below the
is associated with occupied defects but valence band of the semiconductor. There
instead a bleaching of the absorption has been some interest in nanoporous
Cu2 O, which can be sensitized with viscosity on current transients, whereas

methylviologen [68], and NiO. Solbrand and coworkers find, from sim-
ilar experiments, that an effective diffu-
5.3.3.2 Hole Conductors sion coefficient derived from photocurrent
transients increases with electrolyte con-
5.3.3.2.1 Liquid Electrolytes The usual ductivity [74]. It has been pointed out by
hole-conducting medium is a polar sol- Kopidakis and coworkers [75] that ambipo-
vent, such as propanol carbonate, ethylene lar diffusion is expected to apply in dye-
glycol, or acetonitrile, containing the io- sensitized systems, and although electron
dide–triiodide redox couple. In such liquid diffusion appears to dominate transport,
electrolytes the main issue for charge the ambipolar model allows for some in-
transport is the mobility of the iodide fluence of electrolyte diffusion coefficient
and triiodide ions through the pores of on transport.
the structure. The relevant comparison is Redox anions in the electrolyte must
between the ionic conductivity of the elec- be charge balanced by cations. Typically,
trolyte and the electronic conductivity of these are metal ions, such as lithium or
the TiO2 . Bulk diffusion coefficients for I − sodium, or large organic cations such as
and I3 − are typically 10−6 cm2 V−1 s−1 , al- tetrabutylammonium. Lithium is an im-
though the value for I3 − in particular is portant factor, as it appears to be required
strongly dependent on electrolyte [69–71]. for very efficient charge injection and for
Diffusion coefficients in solid-state ion stabilization of electrons in the TiO2 film.
conducting electrolytes can be one to two For aqueous electrolytes, protons are also
orders of magnitude smaller [70]. The ef- present. Although these cations do not take
fect of the porous medium on triiodide part in charge-transfer reactions and are
diffusion has been studied, and is found electrochemically inert, they influence the
to reduce the diffusion coefficient by an electrostatics of the system, and are widely
order of magnitude [72], which is still believed to neutralize the electric fields
insufficient to make triiodide diffusion created by charge injection by reorganiza-
the limiting factor in current transport. tion at the surface. In the case of lithium
Concentration of redox species in the elec- and protons, intercalated cations are asso-
trolyte is typically 0.1 to 0.5 M, which mean ciated with the formation and stabilization
a much higher charge density than in the of Ti3+ species in TiO2 films.
semiconducting film at equilibrium. These
values lead to ionic conductivities substan- 5.3.3.2.2 Solid-state Hole Conductors Re-
tially larger than the electronic conductivity placement of the liquid electrolyte with
in the porous films and so ionic diffu- a solid-state analog is considered essen-
sion is normally disregarded in analysis of tial for widespread industrial application.
charge transport. However, at high injec- However, at the time of writing, solid-state
tion levels where the electronic mobility hole conductors in dye-sensitized devices
improves, ionic conductivity may become have not been widely studied. In a solid
rate limiting. Evidence for the effect of state, hole-transporting material charge
electrolyte on charge transport is incon- can be carried by a redox couple trans-
clusive: it is concluded in Ref. 73 that ported through an inert polymer, or by the
ionic diffusion is not rate limiting, from hopping of charged excitations through the
studies showing no effect of electrolyte molecules of the hole-conducting material.
Mobile cations may or may not be present, complex. Under photoexcitation the charge
depending on the material used. Key density shifts from around the ruthenium
requirements are clearly the rapid regen- ion to around the carboxylated bipyridal
eration of the dye cation, and slow recom- groups (a metal to ligand charge transfer
bination with electrons in the porous film. reaction (MLCT)). This encourages rapid
In addition, there are a number of physical injection into acceptor states in the metal
(permeability into the pores of the film, sta- oxide. It is controversial whether MLCT is
bility, and optical transparency) and chem- essential for rapid charge injection. Sub-
ical (stability, redox potential consistency ps charge injection has also been observed
with TiO2 film and dye) requirements. for porphyrin dyes in which MLCT does
Several approaches have been tried, not occur [82]. Although the basic require-
including polymer gel electrolytes con- ments for electronic structure of the dye
taining I − /I2 [76], inert ionic conduct- are not clear, it has been pointed out [2] that
ing polymers containing I − /I2 [77], hole- electronic coupling between the π ∗ orbital
conducting polymers [78] and amorphous of the dye complex and the conduction
molecular hole conductors such as the band of the semiconductor (a manifold of
chiral organic, OMeTAD [79]. Molecular 3d orbitals in the case of TiO2 ) may enable
hole conductors are preferred to ionic electron injection from dye to semiconduc-
conductors because of the elimination of tor with high yield.
the volatile component, but with penal- Another much less well studied aspect
ties for the charge transport. A peak of charge transport in dyes is the lateral
short-circuit incident photon to current transfer of excitations and charge between
efficiency (IPCE) for OMeTAD of 33% in- dye molecules anchored on the metal
dicates poor charge transport. It is not clear oxide surface. Some evidence for this
whether this results from reduced hole mo- ‘‘hole hopping’’ is provided by studies
bility relative to the ionic mobility of the of quantum efficiency as a function of
liquid electrolyte, or from increased inter- dye coverage, indicating that collection
facial recombination because of reduced probability is affected by the density
charge screening. In general, both are of dye molecules at low coverage [83].
likely to matter. Although hole-transport Hole hopping in triarylamine dyes on
properties are unimportant in the liquid nanocrystalline films has been inferred
cell relative to electron transport, the same in Ref. 84 from studies of the redox
will not be true of solid-state devices. activity of these dye molecules, which
A limited number of inorganic solid- are energetically incapable of transferring
state hole conductors have been consid- charge to the metal oxide films.
ered. These need to be p-type semicon- A limited amount of work has been re-
ductor materials, which can be grown ported on inorganic sensitizers such as
within the pores of the TiO2 film. The II-VI semiconductor QDs in dye-sensitized
most successful materials reported to date systems [10, 85, 86]. QD sensitizers offer
are copper (I) compounds such as copper the advantages of a tunable absorption
iodide and copper thiocyanate [80, 81]. edge and, in principle, tunable excited
state energy level to optimize the matching
5.3.3.3 Dyes of energy levels between sensitizer and
The most commonly used dye is a electron conductor. However, charge sep-
carboxylated ruthenium bipyridal dye aration is evidently poorer than for organic
dyes, possibly due to poorer coupling recombination rate per unit volume. G and
between QD excited state levels and the R may include capture and release of elec-
semiconductor conduction band, and the trons by traps. A similar equation can be
injection yield is correspondingly less. constructed for each type of charge carrier
The remainder of this chapter is con- present in the system. If the form of Jn ,
cerned primarily with electron transport G, and R are known, the continuity equa-
in the porous semiconducting film. This tions can be solved together with Poisson’s
is on the grounds that the high mobili- equation and appropriate boundary condi-
ties and concentrations of charge carriers tions to yield the carrier density functions
in liquid electrolytes make ionic trans- for particular physical situations. Carrier
port a secondary consideration, although populations are related to the physically
it may affect the ease of electronic trans- relevant measurables, current and voltage,
port through effective parameters. Charge through the definition of the current terms
transport within solid-state electrolytes, and the relationship between carrier den-
on the other hand, cannot be neglected sity and electrochemical potential (Fermi
in considering solid-state systems, but, level) at contacts.
at the time of writing, the field is too For a solar cell, the most relevant quan-
young to make a fair review of the mech- tity is the current density at the terminals
anisms and observations. Similarly, hole of the device, J , delivered under steady
transport within the dye layer cannot be state illumination when a given bias V
dismissed, but lies beyond the scope of the is applied between the cell terminals, the
present work. J (V ) characteristic. In the steady state, the
left hand side of Eq. (13) is zero, and equa-
5.3.4 tions are solved to determine J (V ) for a
Theoretical Background given incident spectrum and applied bias.
We can make some general comments
In this section we introduce the equations about G and Jn . For a slab of material
of device physics as normally applied to of uniform absorption oriented normal to
solar cells and outline the ways in which the direction of incidence of the light,
they must be modified for dye sensitized photogeneration at depth x below the
systems. surface is given by
!
5.3.4.1 Basic Equations of Device Physics G(x) = g(E, x) dE (14)
In any electronic device, charge carriers
must obey a set of continuity equations, where
which balance the generation and recom-
bination of carriers [87]. For instance, the g(E, x) = α(E)b0 (E)[1 − r(E)]e−α(E)x
electron density n(r, t) as a function of (15)
position r and time t must obey, where α(E) is the absorption coefficient of
the material at photon energy E, b0 (E) is
∂n 1
= ∇ · Jn + G − R (13) the incident spectral photon flux, and r(E)
∂t q
the surface reflectivity at energy E.
where Jn (r, t) is the electron current den- For a semiconductor, the electron cur-
sity, q the electronic charge, G is the elec- rent can usually be defined in terms of the
tron generation rate, and R is the electron electron quasi Fermi level, EFn , and the
electron density, n 1. The system has one-dimensional sym-

metry,
Jn = µn n∇EFn (16) 2. Most of the photogeneration occurs in
the doped, neutral regions adjacent
where µn is the electron mobility. (This
to the field bearing p–n junction.
result is derived in the Boltzmann re-
In the neutral regions E = 0 and
laxation time approximation, where relax-
carriers move mainly by diffusion. For
ation within a band is much faster than
electrons in p-type material,
transitions between bands [88, 89]). The
factor ∇EFn is sometimes called the driv- dn
ing force of the carrier population. n is Jn = qDn (20)
dx
related to EFn in the usual way through
3. Most of the recombination occurs in
Fermi Dirac statistics:
! the neutral regions and is dominated
1 by the minority carriers, so that the
n = g(E) (E−E )/kT dE (17)
e Fn +1 recombination rate R varies linearly
where g(E) is the density of electron with the minority carrier density. For
acceptor states at energy E, T is tem- electrons in p-type material
perature, and k is Boltzmann’s constant. n
R= (21)
For a nondegenerate semiconductor with τn
no acceptor states in the band gap, the
where τn is the electron lifetime. (In
Boltzmann approximation is valid and the
the Boltzmann relaxation approxima-
integral is taken up from the conduction
band edge energy, EC . Then tion, τn is a constant).
n = NC e(EFn −EC )/kT (18) Given these simplifications, the follow-

ing linear differential equation can be con-
where NC is the effective density of states structed and solved for the (minority) elec-
in the conduction band. In this limit, Jn tron density in the p region at steady state
can be broken down into contributions
from diffusion and drift, d2 n n
Dn − + g(E, x) = 0 (22)
dx 2 τn
Jn = qDn ∇n + qµn En (19)
with g(E, x) given by Eq. (15). Holes in the
where Dn is the electron diffusion coeffi- n-region obey an analogous equation. For
cient and E the electrostatic field. constant Dn and τn , Eq. (22) can be solved
exactly at a given photon energy E. The
5.3.4.2 Charge Transport in a boundary conditions are (under current
Conventional Photovoltaic Device flow) that n = n0 eqV /kT at the edge of the
First we consider the solution of the prob- depleted p–n junction region, where n0
lem for a conventional p–n homojunction is the electron density in the dark and V
solar cell. Although a very different system, the applied bias, or (in open circuit) that
dn
this is instructive, given that some of the dx = 0 at that plane. A second boundary
dn
terms and concepts have been applied to condition relates dx at the surface to
dye-sensitized devices. surface recombination velocity. In the
In a p–n junction solar cell a number of steady state Eq. (22) yields hyperbolic
simplifications can be made: profiles for both electron and current
density, a dark current density that is the system is in general more complicated.
proportional to (eqV /kT − 1), resembling The following factors are important:
that of an ideal diode. More refined
models of p–n junction device physics 1. More carrier types A continuity equa-
allow for recombination in the junction tion should be constructed for each of
region and lead to nonideal J –V behavior. the species involved in charge transfer:
Evidently, the current voltage response electrons, reduced and oxidized ions in
can be described with knowledge of the the hole conductor. In addition, cations
transport parameters Dn and τn for either in the hole conductor are mobile and
carrier, and the device structure. A useful will affect electrostatics, and therefore
derived parameter is the minority carrier they should be included, although these
diffusion length, Ln , do not contribute to charge transfer.
2. Three-dimensional geometry Because of
"
Ln = Dn τn (23) the structure of the junction, the
geometry is three dimensional on the
which determines the thickness of the microscopic scale. The electrochemical
neutral region from which photoexcited junction gives rise to large local electric
minority carriers can be collected effec- fields within the hole conductor at
tively. all points of the interface. These are
The transport parameters, although well typically a few nm in extent in the case
known for common photovoltaic materi- of liquid electrolytes [23, 24].
als, are determined by transient measure- 3. Boltzmann approximation invalid Be-
ments on doped semiconductor samples cause the nanocrystalline film contains
of known composition (for example, Dn a high density of intraband gap states,
from time of flight measurements of the Boltzmann approximation for n
mobility, τn from time resolved photolu- (Eq. 18) may not be valid. Instead, n will
minescence). be determined by some distribution of
Notice that although both electrons and intraband gap states, with distribution
holes are present, current in the neutral re- g(E), and Eq. (17) should apply.
gions is dominated by the minority carrier 4. Nature of photogeneration Photogener-
because light and bias induced carrier gra- ation is a two-stage process that requires
dients are larger. Majority carrier transport absorption of a photon by a sensitizer
is described by similar diffusion equations molecule and injection of an electron
but because the recombination probability from dye to semiconductor.
is so low, the majority carrier density is 5. Nature of current The nature of the
effectively constant across the neutral re- driving force for the electron current is
gion. Gradients in the carrier density may not definitely known. For example, the
arise from resistive losses in low mobility electron mobility is not expected to be
materials. a constant because the relaxation time
approximation, which is used to derive
5.3.4.3 Device Physics for Dye-sensitized Eq. (18), is not necessarily valid.
Systems 6. Nature of recombination The defini-
Now we consider the case of a dye- tion of the electron recombination
sensitized device. On account of the rate is not known. In a dye-sensitized
complex geometry and multiple media, cell, unusually, two routes for electron
recombination exist: with hole conduc- detrapping is given by

tor and with dye cation, as detailed
earlier. Neither of them are necessar- βe−(E−EFn )/kT (26)
ily expected to be linear with electron
The energetic distribution and activ-
density.
ity of such trap states is not known,
7. Trapping Electron capture into and re-
and will be a function of film compo-
lease from intraband gap ‘‘trap’’ states
sition and morphology and chemical
is expected to occur. In this case, it
environment.
is useful to distinguish electrons that
8. Coulombic effects Because of the small
contribute to photocurrent from those capacitance of the nanocrystals forming
that cannot. If we use nc to represent the porous film, the Coulombic energy
the density of ‘‘free’’ electrons (those required to charge a nanoparticle on ad-
with energy above the conduction band, dition of an electron may be significant.
or mobility edge), then a net transfer A change in electrical potential energy
of carriers to trap states should be in- by application of bias, V , will not be
cluded in the continuity equation for nc : translated solely into a shift in electron
∂nc 1 Fermi level, but will be divided between
= ∇ · Jn + G − R a shift in Fermi level )EFn and a shift
∂t q
! EC in electrostatic potential energy )Ee ,
df
− nc g(E) dE (24)
EV dt qV = )EFn + )Ee (27)
where g(E) is the density of states For a spherical nanoparticle of radius r,

at energy E in the intraband gap
q 2 n1
region, EC/V the conduction/valence )Ee = (28)
band edge energy, and f (E) is the 8πεε0 r
(nonequilibrium) electron distribution where n1 is the number of electrons
function. The final term can be resolved already contained in the particle, and ε
into trapping and detrapping terms, the effective dielectric constant, which
! EC will in general depend on the chemical
nc β(E)g(E)(1 − f (E)) dE environment. This issue is relevant to
EV both transport (Coulombic forces op-
! EC posing further accumulation of charge
− nc β(E)e−(E−EFn )/kT on a nanoparticle) and cell photovolt-
EV
age (not all of the electric potential
× g(E)f (E) dE (25) of the electrons contributes to external
voltage).
where β(E) is the rate per unit volume
of electron capture by a vacant trap of
energy E, given by 5.3.4.4 Simplifications for Dye-sensitized
Systems
β = σν (25) Although complex in general, the fol-
lowing simplifications can be made for
where σ is the electron capture cross dye-sensitized systems. Many of these are
section and ν the electron thermal justified by experimental studies discussed
velocity. The volume rate of electron in Sect. 5.3.5.
1. Generation = absorption In the case of electrolyte is concentrated enough and

bipyridal Ru dyes, electron injection the ions mobile enough within the
from excited dye molecules occurs in porous structure that electric fields are
subpicosecond time scales with unit not expected to exist within the elec-
efficiency [82, 90, 91]. Therefore light trolyte over more than a few nm [12,
absorption by the dye can be considered 24, 92, 93]. The small size of the
equivalent to electron photogeneration. nanocrystals adds to the restriction that
2. One dimensional geometry Because the the porous film cannot sustain elec-
scale of phase separation in the junc- tric fields. Together these mean that no
tion is small compared to the device macroscopic electric fields are expected
size, an effective medium approach to exist within the film away from the in-
is often applied, and one-dimensional terface with the SnO2 substrate. (Local
symmetry assumed. This means ne- electric fields associated with individual
glecting the effect of local electric fields charge carriers and the polarization of
at the interface, and differences in be- their environment are still expected, but
havior between electrons or ions that are generally neglected.)
are near or further from the interface or 5. Collection at SnO2 interface The higher
grain boundaries. The effect of elec- electron affinity of doped SnO2 than
trolyte on electron transport can be TiO2 implies that a space charge re-
included through an effective diffusion gion encouraging electron flow into the
coefficient. substrate should exist at this interface.
3. Neglect of hole transport Because the This has been confirmed experimen-
density and mobility of charged species tally in both dry and wet films by
in the electrolyte is high, transport of the observation of directed photocharge
ionic species is presumed to be facile transients [94] and direction dependent
and only electron transport needs to photocurrent efficiencies [73, 95]. The
be considered to explain transient phe- interface can be treated similarly to
nomena. This is analogous to the focus the depleted region of a p–n junction
on minority carrier transport in the (although it requires electron accumu-
neutral regions of a p–n junction so- lation, rather than depletion, in the
lar cell. Although electrons are majority TiO2 ) in that all electrons reaching the
carriers within the oxide film, in an interface are collected, and the excess
effective medium that also contains photogenerated carrier density at that
electrolyte they are minority species. plane is zero.
The solar cell contains an electron den-
sity of around 0.1 per nanoparticle in the These simplifications reduce the prob-
dark, and a ‘‘hole’’ density of some tens lem to that of electrons diffusing in a
(30–50 mM of I3 − ) [14, 20]. Transport fieldless environment bordered by an ab-
in this system should more properly sorbing boundary, that is,
be regarded as ambipolar diffusion, but # $
∂n ∂ ∂n
the relative ease of hole relative to elec- = Dn (n(x)) + G − R (29)
∂t ∂x ∂x
tron transport means that the electron
transport properties dominate [75]. In the one-dimensional approximation,
4. Neglect of electric fields At least in the the boundary conditions are that n(0) =
case of liquid electrolyte junctions, the nV (0) at the absorbing boundary x = 0,
0.6
0.5
Electron concentration
0.4
[arbitrary units]
Light 0.3
0.2 High OD, V > 0

High OD, V = 0
0.1 Low OD, V = 0
Low OD, V > 0
0
0 2 4 6 8
Depth
[µm]
Fig. 7 Illustration of electron density satisfying the general diffusion equation for
different optical densities and applied biases. It is assumed that light is incident from
the negative × direction. In the diffusion picture the electron density is always depleted
near to the interface with the TCO at x = 0, providing a statistical driving force for
electron collection.
where nV (0) is the electron concentration the most relevant to photovoltaic device
in the dark when a bias V is applied performance, but time and frequency
dn
across the terminals, and dx = 0 at the resolved measurements, usually on model
extreme semiconductor–electrolyte inter- systems, are the most revealing about the
face. Equation (29) is more complex than nature of charge transport.
the case of electrons in the p-type region
of a p–n junction solar cell (Eq. 22) be- 5.3.5.1 DC Measurements
cause of the generally nonlinear behavior In the steady state the left-hand side
of the recombination and diffusion, and of Eq. (13) is equal to 0. Measurables
the failure of the Boltzmann approxima- are the steady state photocurrent, pho-
tion. The diffusion coefficient is expressed tovoltage, light and dark current voltage
as a function of n(x) to reflect this. Some characteristics, and the quantum efficiency
approaches to this problem are reviewed spectrum (or IPCE). If the physical origin
later. Elucidation of the mechanisms of of Jn is known, the dependence of such
recombination and diffusion to provide DC measurements on variations in in-
appropriate functional forms for Dn and tensity, wavelength, and bias, deliver the
R is a major experimental challenge, and parameters controlling Jn (for example,
is discussed in the following section. the values of the diffusion length and dif-
fusion constant in the case of diffusion
5.3.5 limited transport).
Experimental Techniques and Results
5.3.5.1.1 Incident Photon to Current
We divide these into three groups: DC, Efficiency IPCE, also called external
time domain, and frequency domain. quantum efficiency, is the ratio of electrons
DC measurements under illumination are collected at the terminals to incident
photons under monochromatic illumina- of device thickness, sensitizer, and sub-

tion at short circuit. A high IPCE is a strate and (2) is near unit efficient, (3) is
prerequisite for a useful photovoltaic de- the key quantity probed. We will refer to
vice. At short circuit, recombination can this probability as ηcoll , the quantum ef-
usually be neglected and the continuity ficiency for charge collection. IPCE’s of
equation becomes, in the one-dimensional around 80% at the absorption maximum
geometry, of the dye have been reported for sev-
eral Ru based sensitizer dyes on ∼10 µm
1 dJn (E, x) thick nanocrystalline TiO2 films [14, 96,
+ g(E, x) = 0 (30)
q dx 97]. These are illustrated in Fig. 8.
Such high IPCE’s correspond to almost
for light of a given photon energy, E. IPCE 100% efficient collection because around
is defined as 20% of the light is lost by absorption in the
1 Jn (E, 0) SnO2 layer, by reflection and scattering.
η(E) = (31) In terms of a simple diffusion model
q b0 (E)
of electron transport, this indicates that
where Jn (E, 0) is the current passing the diffusion length of the electron is
through the SnO2 interface at x = 0. Be- comparable to the optical depth of the
cause the recombination rate, R, does not cell at that wavelength. The very high
enter, this is a direct probe of the current IPCE’s first reported for dye-sensitized
mechanism. It is convenient to think of nanocrystalline films were a surprising
the IPCE as the product of the probabili- and promising feature.
ties of (1) photon absorption, (2) electron More information about the diffusion
injection, and (3) electron transport to the mechanism can be extracted by studying
terminals. Because (1) is a known function the IPCE as a function of direction of
Incident photon to current efficiency
Rul′(NCS)3
60
Rul2(NCS)2
Rul2[Ru(bpy)2(CN)2]2
40 TiO2
L = 2,2′ − dcbpy
20 L′ = tc − terpy
RuL3
0
400 600 800 1000
Wavelength
[nm]
Fig. 8 IPCE’s of some different sensitizer dyes, including the most
widely used Ru(bpy)2 (NCS)2 , and the recently developed ‘‘black’’ dye,
..
developed by Prof. Gratzel’s group at EPFL [97].
illumination. In normal operation the de- the operating bias of the solar cell, IPCE
vice is illuminated from the glass substrate at peak wavelength falls to about 45%
side, so that light is absorbed preferen- of its saturation value [93]. This indicates
tially close to the SnO2 : TiO2 interface. If competition between recombination and
the diffusion length is comparable with transport, although it is not trivial to
the absorption depth and collection at the disentangle the contributions from Jn
interface is easy, then high quantum effi- and R.
ciencies are expected for illumination from The IPCE for sensitized ZnO films
this direction. For illumination from the is smaller [64], although it is not clear
rear side (electrolyte side) of a film which whether this is because of poorer electron
is optically thick, most electrons have to transport or poorer electron collection at
cross the film to be collected, and study of the interface with substrate [67].
the IPCE as a function of film thickness
provides information on the electron diffu- 5.3.5.1.2 Dark-current Voltage In the
sion length. Sodergren and coworkers [21, dark, G = 0 and the steady state continuity
98] have applied the diffusion solutions to
Eq. (13) becomes
fit the IPCE of a non-dye-sensitized device
for illumination from electrolyte side and 1 dJn
−R =0 (32)
substrate side, and derive an electron dif- q dx
fusion length of 0.8 µm. Their results also
where R depends (usually strongly) on
confirm, importantly, that electrons gen-
the applied bias V . The dark current
erated close to the SnO2 : TiO2 interface
Jdark (V ) is the current passing though
are collected most efficiently, indicating
the presence of an electric field driving x = 0 when a forward bias V is applied
collection at that interface. Measurements between the terminals. Note that Jdark acts
by Lindstrom and coworkers indicate that in the opposite sense to the short circuit
the electron diffusion length is longer in photocurrent. It is normally this process
dye-sensitized devices [95], as would be ex- which limits the performance of a solar
pected from the absence of photogenerated cell. Analysis of Jdark (V ) can in principle
holes. teach us something about R.
The IPCE at the optimum wavelength In a conventional solar cell, R results
can be interpreted as an indicator of the from the recombination of injected minor-
relative efficiency of electron transport ity carriers. The dark current is rectifying
in variations of the dye-sensitized TiO2 and can be approximated by the form
device. For instance, the addition of
electron acceptors such as O2 and I2 to Jdark = J0 (eqV /mkT − 1) (33)
the electrolyte lead to reduced IPCE at all
wavelengths [95], which can be interpreted where J0 is a constant and the ideality
in terms of a shorter electron lifetime or factor m normally lies in the range 1 ≤ m ≤
shorter diffusion length. 2. m and J0 are related to the mechanism
Under forward bias (electrolyte at a more for charge transport and recombination.
negative potential relative to the electrode The same parameters have been used
potential) the electron density in the to characterize dark currents of dye-
film increases and recombination becomes sensitized cells, although their meaning
significant. At a bias that corresponds to is not clear.
In a dye-sensitized cell in the dark, R re- this number is related to the kinetics of
sults mainly from the scavenging of electri- the reduction reaction, but because inter-
cally injected electrons by oxidized species pretation in terms of transport parameters
(usually I3 − ions) in the hole conductor. depends on the type of junction, for in-
Dark currents typically reach current den- stance, p–n, p–i–n, Schottky barrier, it is
sities of several mA cm−2 at the operating not useful for DSSCs. Interpretations of
bias of the solar cell, and exhibit recti- dark-current behavior in terms of various
fying behavior before being taken over barrier [99] or kinetic [92] models have
by series resistance at current densities been made but are speculative.
of ∼10 mA cm−2 (Fig. 9a). Unlike solid- The magnitude of the dark current
state junctions, the dark current does not can be increased by various treatments
saturate at negative biases because of the that are expected to enhance electron
nonnegligible probability of electron trans- recombination. These include the addition
fer to the film at high reverse biases where of electron scavengers such as oxygen [100]
the film is electron poor. The high series re- to the electrolyte, and pretreatment to
sistances are due to the series resistance of increase the density of defects (oxygen
the conducting glass of some 10–15 ohm- vacancies) in the film [54]. The latter
square. Ideality factors of around 1 [99] effect indicates that oxygen vacancies
and 2 [92] have been reported. In principle act as recombination centers. Any such
107 dark
22 °C
106 40 °C
55 °C
105 70 °C
Current
[mA cm 2]
104
103
102
101
−0.2 −0.1 0.0 0.1 0.2 0.3 0.4
Potential
(a) [V]
0.75
0.50 Fig. 9 (a) Dark current voltage

slope = 0.085 V
characteristics of dye-sensitized cells at
Uphoto
[mV]
different temperatures, from Ref. 99. At

0.25
high forward bias the current deviates
from the exponential form on account of
series resistance. Notice the bias
dependence of the current at reverse
0.00 bias. (b) Light intensity dependence of
1010 1011 1012 1013 1014 1015 1016
the open circuit voltage. At high
Photon flux intensities Voc varies as expected from
(b) [cm−2 s−1] Eq. (36) (from Ref. 16).
treatment that enhances dark current will short circuit, and that recombination is
reduce the efficiency of charge collection. negligible in this bias range.
At larger forward bias, the net photocur-
5.3.5.1.3 Light-current voltage Although rent drops because of increased electron re-
the most relevant measurement for solar combination. In a conventional solar cell,
energy conversion, steady state J (V ) is the process responsible for the recombina-
not the most useful for analyzing charge tion current is normally the same in light
transport. The J (V ) curve for a typical and dark, and it is common to make the ap-
dye-sensitized cell in standard solar il- proximation that J (V ) = Jsc − Jdark (V ).
lumination is presented in Fig. 10. The Although this may be good for semi-
short-circuit photocurrent, Jsc = J (0), cor- conductor p–n junction devices, it is not
responds to the IPCE integrated over the reliable for dye-sensitized photovoltaic de-
spectrum, vices. Ionization of the sensitizer opens
!
Jsc = η(E)b0 (E) dE (34) up a new path for recombination in light,
which is not active in dark (Fig. 11). There
In a good device, J (V ) remains constant is some evidence [69] that the reverse cur-
into small forward bias, indicating that rent is larger under illumination than in
the photocurrent has indeed saturated at the dark.
X-25 IV System
4.0
Sample PL0710/2
3.5 Oct 30, 1998
tape aperture, black dye
3.0
Area = 0.1863 cm2
2.5 Irradiance = 1000 Wm−2
2.0
Current
VOC = 0.7210 V
[mA]
ISC = 3.824 mA
1.5
Vmax = 0.5465 V
Imax = 3.552 mA
1.0 JSC = 20.53 mA cm−2
Pmax = 1.941 mW
0.5 Fill factor = 70.41%
Efficiency = 10.4%
0.0
−0.5
−0.2 0.0 0.2 0.4 0.6 0.8
Voltage
[V]
Fig. 10 Current voltage response of dye-sensitized cell measured at NREL [101]. This cell
has a short-circuit current density of over 20 mA cm−2 and an open circuit voltage of over
0.7 V. The relatively low slope of I–V near Voc indicates a significant series resistance, largely
due to the sheet resistance of the conducting glass substrate.
Dark Light
Fig. 11 Schematic illustration of charge recombination pathways in a

dye-sensitized device in dark and light conditions. In the dark, electrons in
the nanocrystalline semiconductor recombine only with the hole
conductor. Under illumination, some of the dye molecules are ionized and
electrons may recombine either with the hole conductor or with the dye
cation. The second recombination route is often neglected, but may be
significant at forward bias.
Study of the dark current is thus useful R, as does the dark J (V ), but without
for understanding the reaction between the problem of series resistance. For a
electrons and redox couple, but not for dye-sensitized cell it is a more relevant
understanding the limits to photovoltage comparison as it is probing recombination
of the DSSC. Nevertheless, simple models under illumination, and not in the dark.
of J (V ) behavior derived from solid-state Variation of Voc with light intensity has
device physics can been applied to DSSC’s been reported by various authors including
to derive values of effective transport pa- Huang and coworkers [69] and Fisher
rameters. For example, Sodergren and and coworkers [16] for TiO2 and by de
coworkers [21] and Ferber and cowork- Jongh and coworkers [67] for ZnO. For
ers [102] have used a model, based on both materials the relationship tends to
diffusion-controlled current, to fit the J (V ) the form
behavior of DSSCs. mkT
The open circuit photovoltage, Voc , Voc = ln X + Voc (1) (36)
q
corresponds to the point, at which the
recombination current and photocurrent at high light intensities, where X is the
are equal, light concentration factor, Voc (1) is the
open-circuit voltage at unit concentration,
1 dJn and m the apparent ideality factor. Values
− Rn (Voc ) + G(x) = 0 (35)
q dx of m are in the range 1.3 to 1.5, but,
as commented earlier, these values are
There is no net current at Voc . If local not meaningful without a model for the
currents are also small – for instance, if the recombination current.
generation rate is constant so that there are Voc is influenced by surface treatments
no internal diffusion currents – then the and chemical environment. Huang and
first term in Eq. (35) can be neglected and coworkers show that Voc decreases with
the variation of Voc with light intensity increasing concentration of triiodide, as ex-
tells us something about the form of pected for a recombination current which
is dominated by the reaction of electrons TiO2 ) [12, 103] and with evidence for
with I3 − . Several authors have reported directed photocurrents in dry films [94].
the variation of Voc by surface treatments A comparison of the SPV spectra of an
of the TiO2 . Treatment with pyridine anatase single crystal and nanocrystalline
derivatives [69] and tetrabutylammonium film indicate the presence of a tail of states
phosphate [2] is observed to increase Voc . in the band gap below the conduction band
The effect is attributed to the masking of edge [58].
electrons at the TiO2 surface from elec-
tron scavengers in the electrolyte by inert 5.3.5.2 Time Domain
adsorbed molecules. Treatment of TiO2
Traditionally, time resolved measure-
surface by benzoic acid derivatives is ob-
ments are used to study the mechanisms
served to alter the Voc of solid-state devices
of charge transport and recombination in
in a way that is consistent with modifi-
semiconductor materials.
cation of the surface dipole field by the
adsorbed acids [103].
5.3.5.2.1 Photocurrent Transients Time-
resolved photocurrent at zero or reverse
5.3.5.1.4 Photoconductivity and Photovol-
bias is, like IPCE, a probe of the forces
taic Properties Relevant to the light re-
driving conduction rather than recombi-
sponse of dye-sensitized systems are the
nation. If recombination can be neglected,
photoconductivity and photovoltaic prop-
the transient electron current resulting
erties of nanocrystalline films. Earlier
from an instantaneous or steady state light
it was noted that the conductivity of
stimulus G(x, t), is defined by
nanocrystalline TiO2 is very small in the
dark, around 10−10 –10−8 "−1 cm−1 , ris- ∂n 1 ∂Jn
ing by several orders of magnitude in UV = + G(x, t) (37)
∂t q ∂x
light. Sensitized films are photoconduc-
tive in visible light, even in the absence in one dimension, where the form of J is
of a hole conductor [10]. The relationship not known.
between light intensity and photoconduc- Several configurations have been studied
tivity is superlinear, indicating increased exhibiting some common features. Steady
electron mobility through trap filling [104]. state and square-wave illumination of
It can be shown by simulation that this be- dye-sensitized electrochemical junctions
havior result is consistent with dispersive produces a slow photocurrent, which rises
transport (discussed below) in the pres- to a saturation level [73, 106]. Photocurrent
ence of traps [105]. rise times are extremely long and intensity
Steady state surface photovoltage (SPV) dependent (ms to s, depending on inten-
spectroscopy is useful for determining sity). Both features are attributed to the
the nature of the junction between the trapping and release of photogenerated
nanocrystalline film and substrate. SPV electrons. Schwarzburg and Willig [107]
studies of TiO2 films on SnO2 confirm show with a simple rate equation treatment
that an electric field exists at the interface, that the filling of a single trap level leads
driving electrons into the substrate. This to intensity-dependent photocurrent kinet-
is consistent with the values of the ics, which are qualitatively similar to those
work functions of the two materials in observed. In their case of a single trap level
vacuum (4.85 eV for SnO2 : F, 5.15 eV for Eq. (17) can be resolved into the sum of
conduction (nc ) and trapped (nt ) electrons, filling: at higher injection levels, diffusion
and two continuity equations can be con- should be easier because fewer deep traps
structed for nc and nt with well defined rate are available. Observations [73] of faster
constants for trapping and detrapping. Cao rise times under bias illumination and ki-
and coworkers [106] present transients at netics which depend on the direction of
different light intensities and solve Eq. (37) illumination (effectively, on the intensity
for a light step (G(x, t) = G%0 for t > 0) and& reaching the SnO2 interface) are consis-
dn
diffusive electron current Jn = qDn dx . tent with intensity-dependent diffusion.
Although diffusion leads to a photocur- Transient photocurrents in response to
rent which saturates at a level proportional a laser pulse have been studied in elec-
to the light intensity, it predicts kinetics, trolyte supported and dry films [57, 60,
which are intensity independent. This can 74, 108]. Such transient experiments are
be seen by normalizing n with respect to most informative when the optical depth
G0 in Eq. (37): the solution for n(x, t)/G0 of the film is high, and the light pulse
is independent of G0 . In the data, how- effectively creates a sheet of photogener-
ever, a clear intensity dependence is seen ated carriers, which travels through the
(Fig. 12). The authors propose that the ef- film to the absorbing boundary. This is
fective diffusion coefficient is n-dependent achieved by exciting in the UV where
and demonstrate that the numerical so- absorption of the TiO2 is high. Solbrand
lution of the diffusion equation when and coworkers [74, 108] present laser pulse
D(n) = D0 n, leads to a rise time which induced transient photocurrents for a TiO2
decreases with increasing light intensity. electrochemical junction as a function of
This form of D(n) is somewhat arbitrary, film thickness, electrolyte composition and
but is qualitatively consistent with trap direction of illumination (Fig. 13). The
0.1
0.09
0.08
0.07
I (t) / G 0 / A W1
0.06
0.05
0.04
G 0 = 0.05 mW cm−2
0.03
G 0 = 4 mW cm−2
0.02
0.01
0
0 0.1 0.2 0.3 0.4
Time
[s]
Fig. 12 Transient photocurrent in response to step illumination of different
intensities normalized with respect to the light intensity. At higher intensities,
the photocurrent rise time is faster, indicating an electron density dependent
diffusion coefficient. Replotted from Ref. 106.
Fig. 13 In a transient photocurrent

experiment, charges created near the t1
Back-contact
TiO2 –electrolyte interface travel
n(x,t)
through the film to deliver a current at
the collecting interface, from Ref. 74.
t2
The thickness dependence of the t3
transient currents is consistent with
diffusion, but contains some features of
(a) 0 x W
dispersive transport.
7.5 µm
Current 11.3 µm
[a.u.] 14.9 µm
17.4 µm
0 20 40 60 80 100
Time
(b) [ms]
observed film thickness dependence is Solbrand and coworkers [74] is that the
consistent with a diffusion picture, where apparent diffusion coefficient is dependent
the carrier profile generated near the semi- on the conductivity and composition of the
conductor–electrolyte interface at x = W electrolyte. This is consistent with an acti-
spreads out over time to deliver a time- vation energy for electron hopping which
dependent flux of carriers through the depends on the energy required to polar-
absorbing boundary at x = 0. It is evident ize its environment, a phenomenon which
that the transit times are long (tens of ms is familiar from conduction in doped dis-
to s) and the current signal becomes more ordered materials [110]. It is also a clear
dispersive with film thickness. They fit the signal that the electrochemical interface is
slow portion of their transients (an initial very important to transport, and consistent
fast decay is attributed to capacitive effects) with the idea that electrons move along or
to the solution of the diffusion equation for close to the surface.
a delta function stimulus, and derive an ef- Laser pulse photocurrent transients
fective diffusion coefficient, D, of the order in solid-state Schottky barrier junctions
of 10−6 cm2 s−1 . The facts that (1) the ex- between dry TiO2 and a platinum con-
perimental tail in the photocurrent is more tact have been studied by Konenkamp and
persistent than predicted by simulation coworkers [57]. They observe a power law
and (2) the kinetic shape of the nor- dependence of the collected photocurrent
malized current, J (0, t)/G0 , is intensity on time,
dependent, indicate that simple diffusion J (0, t) ∝ t −β (38)
is not a complete description. These fea-
tures are, however, consistent with a model which is characteristic of ‘‘dispersive’’
of ‘‘anomalous diffusion’’, a random transport. Such algebraic photocurrent
walk through energetically distributed trap transients are familiar from the physics
states, as discussed by Nelson [109]. A of disordered semiconductors and have
second and important observation by been shown to result from the anomalous
diffusion of carriers within a medium the recovered charge is measured as a func-

containing a wide energetic dispersion tion of forward current level. Konenkamp
of trap states. A continuous time ran- and coworkers [55, 60] have applied the
dom walk model due to Scher and technique to Schottky barrier nanocrys-
Montroll [111] has been adapted by Nel- talline TiO2 junctions, and extract an elec-
son [109] for the case of nanocrystalline tron mobility-lifetime product of approxi-
electrodes, and shown to be qualitatively mately 10−10 cm2 V−1 , by fitting their data
consistent with observed photocurrent with a simple model. They find that the
decays. mobility-lifetime product is apparently in-
There is some evidence in the work dependent of injection level, whereas the
cited earlier [57] for a dependence of recombination time, which is in the ms– µs
the photocurrent kinetics on chemical range, falls with increasing injection. This
environment, suggesting that the different behavior is typical of disordered materi-
environments or surface treatments (air, als such as amorphous semiconductors,
vacuum, and pH of electrolyte) influence where charge transport is dominated by
the distribution of states acting as electron the filling of localized states. The observa-
traps. Related evidence is provided by tion is explained in terms of trap filling by
Hagfedlt and coworkers [112], who use arguing that the mobility µ varies like
time-resolved absorption at 900 nm to nc
study the formation and disappearance µ = µc (39)
n
of electrons (in absorbing Ti3+ states)
following electron-hole pair generation where nc is the free (conduction band) elec-
by a UV laser pulse. They observe that tron density, n the total including trapped
in air, the signal initially decays rapidly electrons, and µc represents the mobility
(tens of ns), whereas in ethanol, which of conduction band electrons in the bulk
crystal. The ratio of nc to n is calculated
is an efficient hole scavenger, the signal
using an exponential density of trap states,
persists for µs at least, implying that once
which appears to be consistent with obser-
holes have been removed the injected
vations of sub band gap photoconductivity.
electron density decays very slowly. This
The authors show that the junction recov-
suggests that it may be the presence
ery current transients are both intensity
of a hole-scavenging electrolyte, and not
dependent and tend toward a power law
dye sensitization, which is the necessary
form, as shown in Fig. 14. They argue
condition for majority carrier injection in
that this behavior is consistent with the
nanocrystalline films.
thermal stimulation of electrons out of an
exponential distribution of trap states.
5.3.5.2.2 Potential Step Transients This A variant of the potential step technique,
refers to the current transient in response chronoamperometry, has been applied to
to a potential step. In solid-state devices electrochemical junctions in the dark to
a technique known as junction recov- estimate the charge stored in the film.
ery is applied to barrier junctions as a The transient current resulting from an
probe of the charge density in localized applied potential step is measured and in-
states, particularly in amorphous materi- tegrated to yield a stored charge. Hoyer
als [104]. A large negative potential step is and Weller used this technique to estimate
applied to a forward biased junction and the charge stored in nanocrystalline ZnO
Fig. 14 Transient currents from a −1

junction recovery experiment for a
TiO2 – Schottky barrier device [60]. The 2 mA α = 0.4
Log10 (Recovery current)

currents are dependent on injection −2
level, and tend to a power law at long
times, as expected for dispersive 0.4 mA
transport. −3
[A]
0.1 mA
−4
−5 20 µA
−6
−8 −7 −6 −5 −4 −3
Log10 (Time)
[s]
films [63] and others have used it to esti- dye cation [90]. These and subsequent [82,
mate charge stored in TiO2 [113]. Results 91] studies showed that injection occurs in
indicate an accumulated electron density sub picosecond time scales in a wide range
of several electrons per ∼15 nm diameter of chemical environments, supporting the
nanoparticle, and are consistent with – but assertion made earlier that absorption is
do not prove – an exponential density of equivalent to photogeneration in a DSSC.
surface states. A variant of the junction Although transient absorption is not a
recovery technique, where recovered cur- direct probe of transport, it can be used
rent from an illuminated electrochemical to probe the kinetics of recombination
junction at open circuit is measured as a processes, which compete with transport.
function of light intensity, has been ap- Haque and coworkers [115] studied the re-
plied to nanocrystalline TiO2 by Duffy and combination between electrons in the TiO2
coworkers [114]. These authors also infer electrode and photoionized dye molecules
an exponential density of states from their by monitoring the concentration of the
data and find that the recovered charge dye cation following a laser pulse. By
varies algebraically with time, which they using a redox inactive environment, all
attribute to a reaction of electrons with I3 − , other mechanisms to reduce the dye cation
which is second order in n. were suppressed. They found that this
recombination reaction is slow (∼ms) at
5.3.5.2.3 Transient Absorption Time-re- zero applied bias, but it is very strongly
solved absorption can be used to mon- dependent on the bias applied to the
itor the appearance or disappearance of junction, indicating that electrically in-
the reduced or oxidized states of molec- jected electrons are capable of reducing
ular species. Ultra fast time-resolved the dye cation. Importantly, they find that
absorption was first used by Tachibana this recombination reaction becomes suf-
and coworkers to measure the rate of ficiently fast (ns-µs time scale) to compete
injection from dye-excited state into semi- with the reduction of dye cation by iodide
conductor, probing the absorption of the at voltages near open circuit, and so should
not be neglected in estimating the limits to energies are distributed according to an

performance of the DSSC. Simultaneous exponential density of states function with
determination of the electron density in the characteristic temperature T0
film using the known optical absorbance 1 −(EC −E)/kT0
of electrons trapped at Ti3+ states showed g(E) = e (42)
kT0
that the half life for decay, t50% , varies like
then, in the limit where n exceeds [D+ ], the
t50% ∝ n−1/α (40) cation population evolves with time like a
stretched exponential [116]
where n is the total initial electron density
[D+ ] ∝ e−n(t/τ )
α
including optically and electrically injected (43)
electrons, and α a fraction in the range
where τ and α are constants and
from 0.25 to 0.5, which depends on
chemical environment [11]. T
α= . (44)
Neither the electron density dependence T0
nor the shape (which is approximately It can be shown from Eq. (43) [117] that
stretched exponential) of the kinetics in these conditions the half life for cation
can be explained with second order decay, t50% , varies with n like
reaction kinetics, where it is assumed
that the reaction is controlled only by t50% ∝ n−1/α (45)
the concentrations of electrons and dye as observed. Nelson and coworkers [11]
cations, nor are they consistent with simple have used Monte Carlo simulations of the
electron transfer theory. An explanation CTRW to model observed recombination
was proposed by Nelson based on the kinetics in different chemical environ-
continuous time random walk [109]. In ments with different values of α, (Fig. 15).
the CTRW, electrons perform a random If this is a valid description, it appears
walk on a lattice, which contains trap sites that recombination kinetics are strongly
distributed in energy, according to some influenced by the presence and energetic
distribution function, g(E). In contrast to distribution of trap states, and therefore
normal diffusion, where the mean time may be controlled by controlling the film
taken for each step is a constant, in the treatment and chemical environment. In
CTRW the time taken for each electron to a separate study of the kinetics of the re-
move is determined by the time for thermal combination reaction between electrons
escape from the site currently occupied, in TiO2 electrodes and rhenium sensitizer
τhop ∝ βe(EC −E)/kT (41) dyes, Hasselmann and Meyer [118] arrived
at the conclusion that charge recombina-
where EC is the conduction band edge tion is limited by the diffusional encoun-
energy, E is the energy of the trap ters of the electron with the dye cation,
site occupied, and β is the capture and not by the kinetics of the interfacial
cross section defined in Eq. (26) earlier. electron transfer step.
The lattice is populated with a number A less-used form of transient absorp-
of mobile electrons, n, and a number tion is the monitoring of the electron
of static dye cations, [D+ ], and it is density within the metal oxide electrode by
supposed that a dye cation is reduced measuring changes in the optical density
whenever it meets an electron. If the trap in the red-IR, where Ti3+ states are
0.9
0.8
0.7
−200 mV
∆OD(t) / OD0
0.6
0.5 −300 mV
0.4
−400 mV
0.3
−500 mV
0.2
0.1
0
10 000 100000 1000000 10000000 100000000
Time
(a) [ns]
1000
Electrons per cation
n = At−0.46
100
10
1
10 1000 100000 10000000
t50%
(b) [ns]
Fig. 15 (a) Kinetic decay curves for the rereduction of the dye cation by
electrons in the TiO2 film. The strong bias dependence indicates that
electrically injected electrons are capable of reducing the dye. The smooth
lines are fits using a continuous time random walk model of electron
transport and assuming diffusion limited recombination. (b) Power law
relationship between the electron density and the half life, t50% , for
reduction of the dye cation. In the random walk model, the exponent α is
related to the characteristic temperature of the density of trap states.
known to absorb. This has been used Olson and coworkers to observe the ki-
by Hagfeldt and coworkers to observe netics of electron accumulation following
electron recombination kinetics following the application of a potential step in
injection by a UV laser pulse [112] and by the dark [113]. Another novel probe of
transient electron density is the use of their application to nanocrystalline sys-

time-resolved microwave conductivity to tems are reviewed in Ref. 120.
sense changes in electron density follow- In general, a driving force, which may be
ing application of light or bias. Salafsky optical or electrical is applied and the phase
has applied this technique to deduce an and amplitude of the response, which may
electron transit time of 100s of ms to sec- be a current, voltage, or optical density,
onds for a dye-sensitized device at short is measured as a function of modulation
circuit [119]. frequency. For compactness, the results
are often presented in the complex plane,
5.3.5.3 Frequency Domain as the locus of the complex response as
Here we refer to a class of techniques frequency is varied, although phase and
where the response of a system to small, amplitude data plotted against frequency
periodic modulations of the driving force may be easier to interpret. For a simple
is studied as a function of the frequency system with a single time constant, τ , the
of the disturbance. These give the same response is a semicircle in the complex
information as time-domain studies but plane, and it is easy to show that the
without the need for large amplitude frequency at the point where the imaginary
disturbances. This is particularly useful component of the response is a minimum,
for dye-sensitized – and for a wide range ωmin , is equal to 1/τ . Distorted semicircles
of complex systems – because the optical imply multiphasic responses, which would
and electronic response is nonlinear in be modeled with several time constants in
n, and therefore diffusion coefficient and the time domain.
recombination times are not constants.
The use of small modulations permit lin- 5.3.5.3.1 Intensity Modulated Photocur-
earization of the transport problem, and rent Spectroscopy IMPS was used on un-
yields effective diffusion and recombina- sensitized nanocrystalline TiO2 electrodes
tion parameters, which can be related to by de Jongh and Vanmaekelbergh [121].
the injection level through the underlying In this configuration the driving force is a
steady state driving force. modulated light source and the response
We will distinguish various modes. The is the modulated short-circuit photocur-
techniques most widely used in DSSCs are: rent. These authors showed that ωmin
intensity modulated photocurrent spec- is dependent on light intensity, X, vary-
troscopy (IMPS) which is, like IPCE, con- ing approximately like ωmin ∝ X 0.5 (as
cerned with electron transport under short shown in Fig. 16) and were able to at-
circuit conditions; Intensity modulated tribute this to the trapping and detrapping
photovoltage spectroscopy IMVS, which of electrons in a distribution of trap states
probes the competition between transport which is wide compared to kT . They de-
and recombination at open circuit; and duced a density of trap states of around
electrical impedance spectroscopy (EIS), 1017 cm−3 and from this value concluded
which probes the bias-dependent electrical that the traps are located on the large
response, analogous to dark-current tran- TiO2 –electrolyte interface and not at the
sients. Other variants, such as frequency- TiO2 –TCO contact [122]. The trapping hy-
resolved transmittance, the frequency do- pothesis is supported by studies on porous
main analogs of transient absorption, have GaP electrodes, where ωmin is intensity in-
been developed [50]. The techniques and dependent and the trap density is known
80
0.000 60
Im [∆i (ω)/e Φ(ω )]
ω min/rad s−1
0.1 Hz
1 KHz
1 Hz 40
−0.005 100 Hz
8 Hz
20
−0.010 (c) (c)
0
0.00 0.01 0.02 0 1 2 3
(a) Re [∆i (ω )/e Φ(ω)] (b) (Φ/1016 cm−2s−1)0.5
Fig. 16 (a) Complex plane plot of characteristic frequency, which varies
intensity-modulated photocurrent spectrum for a approximately like the square root of the light
nanocrystalline TiO2 film. A characteristic time is intensity. In a trap-free system this frequency
assigned from the frequency at which the would be independent of light intensity (from
imaginary part of the response function is Ref. 121).
greatest. (b) Intensity dependence of the
to be low; in this case multiple scattering is increases and so does Dn . Later work by
believed to dominate transport. The theory the same authors [50, 16] showed that the
of trap limited IMPS response is devel- diffusion coefficient derived this way varies
oped in Ref. 123, where a distinction is with light intensity approximately like
made between multiple trapping and dis-
Dn ∝ X 0.6 (47)
persive transport in the frequency domain,
and in Ref. 124. Fisher and coworkers combine this ob-
Dloczik and coworkers [92] later used servation with the observed (exponential)
IMPS to obtain an estimate of the dependence of photovoltage on light inten-
electron diffusion coefficient in sensitized sity to derive a relationship between Deff
nanocrystalline TiO2 electrodes. This is and electron Fermi level,
done by transforming the time dependent
Dn ∝ e(1−β)EFn /kT (48)
diffusion equation (Eq. 22) into frequency
space and fitting the result for the which is consistent with an exponential
photocurrent modulation to the measured density of trap states of the form,
values to extract values for an effective
g(E) ∝ eβE/kT (49)
diffusion coefficient, Dn and electron
lifetime, τ . This analysis does not explicitly as Eq. (42), with β = α. They deduce a
account for the trapping and detrapping value of 0.57 for the exponent β from their
of electrons. However, it is pointed out data, which corresponds to a characteristic
that for the case of a single trap level, temperature T0 of 526 K [16].
the effective diffusion coefficient is related
to the true diffusion coefficient, Dc , for 5.3.5.3.2 IMVS In IMVS, the driving
conduction band electrons through force is again a modulated light source and
nc the measured response is the modulation
Dn = Dc (46)
n in open circuit photovoltage. The tech-
in analogy with Eq. (39). As injection level nique was first applied to dye-sensitized de-
increases, the proportion of free electrons vices by Schlichthorl and coworkers [125].
The time constant, 1/ωmin , delivered by The origin of the power law dependence
IMVS, corresponds to an effective lifetime of recombination time on light intensity,
for recombination, τrec , whereas the time consistent with Eq. (52) is not clear. Such
constant delivered by IMPS results from a behavior has been observed elsewhere
combination of recombination and charge but attributed alternatively to a reaction
collection, between electrons and triiodide, which
# $ is second order in n [16], and to charge
1 1 −1 trapping [11, 128].
τIMPS = + (50)
τrec τcoll
although at short circuit the time constant 5.3.5.3.3 Electrical Impedance Spectro-
for collection, τcoll , often dominates. It is scopy (EIS) EIS measures the current
clearly important to be able to relate these response to a modulation in applied
experimentally accessible time constants bias, and can be used in the light
to the functionally relevant quantity, the or dark. Traditionally, EIS results are
quantum efficiency ηcoll for charge collec- interpreted in terms of an equivalent
tion, but that is not trivial in a nonlinear circuit and information extracted about
system. Schlichthorl and coworkers [126] relevant capacitive, resistive, and faradaic
processes. The technique has been applied
showed that although the linear approxi-
to nanocrystalline TiO2 junctions in the
mation,
dark by various groups [19, 22] and a
# $
1 + τcoll −1 film capacitance extracted and interpreted
ηcoll = (51) using a simple equivalent circuit. Results
τrec
confirm that a potential drop does exist
is not adequate, a model based on elec- within the film in the dark. In reverse
tron diffusion with nonlinear relationships bias, where the film is depleted, the film
between electron density and recombina- capacitance appears to be bias independent
tion rate and dominated by the potential drop at
R ∝ nm1 (52) the TiO2 –TCO interface. In forward bias
the film becomes populated and a high
where m1 ≈ 2.2, and between total and and a bias dependent capacitance develops
free electron density at the large TiO2 –electrolyte interface,
suggesting an increased charge storage
ncb ∝ nm2 (53) capacity as surface or conduction band
states become filled.
where m2 ≈ 2.7, is able to explain observed Van de Lagemaat and coworkers [93]
IMVS and IMPS behavior. The two apply EIS to illuminated nanocrystalline
power law relationships are deduced from TiO2 electrodes and again observe a
observations of τrec versus light intensity capacitance which increases, approxi-
using IMVS, and Jsc versus electron mately exponentially, with applied forward
density using IMPS. The strong power bias. They attribute this to the charging of
law dependence of ncb on n implied surface states and show that the behavior
by the comparison is consistent with is consistent with an exponential density
an exponential density of states, again of states. They further conclude that the
of the form (40), but with α = 0.37 or potential drop across the TiO2 –TCO in-
T0 = 810 K [127]. terface extends over a narrow region, and
drops under illumination, which would be photopotential is developed mainly across

expected as the Fermi level of the TiO2 the semiconductor–electrolyte interface.
is raised. EIS under illumination provides In electrolyte-supported systems, trans-
information about the time constant for port is rate limited by electron transport.
recombination, as does IMVS, but is not This has some characteristics of diffusion,
limited to open circuit. This means that but is not well described by a simple diffu-
with an appropriate model for the extrac- sion equation. Both the apparent diffusion
tion of ηcoll from τrec and τcoll , EIS and coefficient and the apparent recombina-
IMPS can provide an alternative measure tion time are electron density dependent.
of the bias dependent quantum efficiency The n-dependence of D is consistent with
of a dye-sensitized device. Those authors the concept of increasing ease of transport
find ηcoll = 0.45 at open circuit. as deep traps are filled.
Frequency-resolved experiments in na- There is widespread evidence for elec-
nocrystalline electrodes often reveal a high tron traps in the nanocrystalline films,
degree of frequency dispersion in the largely Ti3+ states because of oxygen va-
underlying processes. In simple planar cancies and states because of adsorbed and
interface models this may be dealt with intercalated species.
using a constant phase element (CPE) Nonlinear diffusion has been app-
in the equivalent circuit. The CPE con-
roached with a number of models:
tains a degree of frequency dispersion but
(1) n-dependent diffusion coefficient [106],
is hard to relate to physical processes.
which can be related to the fraction of
Bisquert [129] has taken the analysis of
free electrons in a medium containing
electrical impedance spectra one stage fur-
traps, specifically, Dn ∝ nc /n [16, 55, 75,
ther by invoking a transmission line model
127]; (2) rate-equation approach involving
of the interface. The model allows for
a distribution of trap states [123]; (3) Monte
two resistive channels representing charge
Carlo simulations of anomalous diffusion
flow in the two media, and distributed
in the presence of trap states [109]. Numer-
charge transfer at different points on the
interface by means of a resistor-CPE el- ous authors have invoked an exponential
ement. They obtain excellent agreement distribution of trap states with a charac-
with measured EIS in light and dark, and teristic temperature of between 500 K and
suggest that two characteristic relaxation 810 K. There appears to be some consen-
frequencies can be resolved. sus on this point.
Nonlinear recombination such as τrec ∝
5.3.6 n−β has been observed by several
Summary groups [11, 16, 60, 93, 115, 125]. This
has been attributed to different origins:
In summary, we can make the following recombination reactions which are second
observations about the nature of charge order in n, and diffusion limited
transport in dye-sensitized systems. recombination, where the electron motion
There is virtually no electric field within is limited by transport through traps.
the porous film. There is an electric field The dispersion of time constants for
driving charge collection at the interface conduction and recombination lead to dis-
with TCO, and an electric field at the persive features in both time-resolved and
Helmholtz layer. Under illumination, a frequency-resolved measurements. These
are not easily addressable with equivalent is considered only in one dimension.
circuit models for planar interfaces. Some consider transport within a single
The fact that recombination is slow nanoparticle, or other highly simplified
and transport relatively efficient at short geometry. There is clearly a need to move
circuit, is qualitatively attributed to the ef- on to models incorporating the interpene-
fects of screening of electrons by cations trating features of the morphology.
in the electrolyte; and to the absence of Looking to the future, the following is-
mobile photogenerated holes. Recombina- sues remain to be addressed quantitatively:
tion with the oxidized species in the hole the relationship between material proper-
conductor is normally assumed to be the ties and the apparent density of states; the
main loss process for electrons, but re- effect of hole conductor composition and
combination with dye cations cannot be mobility on electron transport; Coulom-
ruled out. bic effects of nanoparticle charging and
Continuity equation approaches involv- electron-hole interactions. An improved
ing nonlinear forms for the electron diffu- description of charge transport is needed to
sion and recombination terms are capable replace the simple diffusion model. Tech-
of reproducing some aspects of observed niques developed to deal with disordered
behavior [123, 126]. The challenge is to conductors in the past may prove use-
predictively relate the forms for recombi- ful here. Multiple-trapping models such
nation and diffusion to known material as the continuous time random walk can
properties of the system. explain certain dispersive features of cur-
Chemical environment strongly influ- rent and recombination kinetics. Another
ences transport. Adsorption and interca- tool that offers the potential to describe
lation influence the potential distribution some features of the behavior of these sys-
at the interface, affecting the tendency for tems is percolation theory. This has already
electrons to accumulate in the film and the been used to describe hole conduction in a
photovoltage. Composition of electrolyte polymeric hole conductor [130]; to describe
influences rate of electron scavenging. charge hopping between quantum sized
Film history influences the density of traps ZnO nanoparticles [65]; and to describe
and recombination centers. hole migration between dyes [83, 84].
In nanoparticulate semiconductors other
than TiO2 , quantum confinement may be- Acknowledgment
come important.
In solid-state systems charge transport The author is grateful to James Durrant,
within the electrolyte is expected to be rate Sven Sodergren, Thierry Lutz, Carol Ol-
limiting. Even in liquid electrolyte systems son, Mili Biswas, and Ellen Moons for
there is some evidence for an influence helpful discussions, and acknowledges the
of electrolyte on electron transport. This financial support of the Engineering and
is expected from the screening and reor- Physical Sciences Research Council and
ganization energy. In solid-state systems the Greenpeace Environmental Trust.
the absence of macroscopic fields cannot
be assumed and drift may be expected to References
become significant. 1. M. Grätzel, Nanocrystalline electronic junc-
Most models applied to date use an effec- tions, Stud. Surf. Sci. Catal. 1997 103,
tive medium in which electron transport 353–375.
2. A. Hagfeldt, M. Gratzel, Light-induced re- (X = Cl-, Br-, I-, Cn-, and Scn-) on nanocrys-
dox reactions in nanocrystalline systems, talline Tio2 electrodes, J. Am. Chem. Soc.
Chem. Rev. 1995, 95(1), 49–68. 1993, 115(14), 6382–6390.
3. B. O’Regan, M. Gratzel, A low-cost, high- 15. E. Topoglidis, T. Lutz, R. L. Willis et al., Pro-
efficiency solar-cell based on dye-sensitized tein adsorption on nanoporous TiO2 films:
colloidal TiO2 films, Nature 1991, 353(6346), a novel approach to studying photoinduced
737–740. protein/electrode interactions, Faraday Dis-
4. N. Vlachopoulos, P. Liska, J. Augustyn- cuss. 2000, 116, 35–46.
ski et al., Very efficient visible-light energy 16. A. C. Fisher, L. M. Peter, E. A. Ponomarev
harvesting and conversion by spectral sensi- et al., Intensity dependence of the back
tization of high surface-area polycrystalline reaction and transport of electrons in dye-
titanium-dioxide films, J. Am. Chem. Soc. sensitized nanocrystalline TiO2 solar cells,
1988, 110(4), 1216–1220. J. Phys. Chem. B 2000, 104(5), 949–958.
5. C. Bechinger, S. Ferrer, A. Zaban et al., 17. N. Myung, S. Licht, J. Electrochem. Soc.
Photoelectrochromic windows and displays, 1995, 142, L129–L132.
Nature 1996, 383(6601), 608–610. 18. D. Fitzmaurice, Using spectroscopy to
6. J. Sotomayor, G. Will, D. Fitzmaurice, Pho- probe the band energetics of transpar-
toelectrochromic heterosupramolecular as- ent nanocrystalline semiconductor-films,
semblies, J. Mater. Chem. 2000, 10(3), Sol. Energy Mater. Sol. Cells 1994, 32(3),
685–692. 289–305.
7. E. Topoglidis, A. E. G. Cass, G. Gilardi 19. F. Cao, G. Oskam, P. C. Searson et al., Elec-
et al., Protein adsorption on nanocrystalline trical and optical-properties of porous
TiO2 films: an immobilization strategy for nanocrystalline Tio2 films, J. Phys. Chem.
bioanalytical devices, Anal. Chem. 1998,
1995, 99(31), 11974–11980.
70(23), 5111–5113.
20. J. Bisquert, G. GarciaBelmonte, F. Fabregat-
8. S. Y. Huang, L. Kavan, I. Exnar et al., Rock-
Santiago, Modelling the electric potential
ing chair lithium battery based on nanocrys-
distribution in the dark in nanoporous
talline Tio2 (anatase), J. Electrochem. Soc.
semiconductor electrodes, J. Solid State Elec-
1995, 142(9), L142–L144.
trochem. 1999, 3(6), 337–347.
9. C. Bechinger, B. A. Gregg, Development of
21. S. Sodergren, A. Hagfeldt, J. Olsson et al.,
a new self-powered electrochromic device
for light modulation without external power Theoretical-models for the action spectrum
supply, Sol. Energy Mater. Sol. Cells 1998, and the current-voltage characteristics of
54(1–4), 405–410. microporous semiconductor-films in pho-
10. P. Hoyer, R. Konenkamp, Photoconduction toelectrochemical cells, J. Phys. Chem. 1994,
in porous Tio2 sensitized by Pbs quantum 98(21), 5552–5556.
dots, App. Phys. Lett. 1995, 66(3), 349–351. 22. A. Zaban, A. Meier, B. A. Gregg, Elec-
11. J. Nelson, S. A. Haque, D. R. Klug et al., tric potential distribution and short-range
Observation of trap limited recombina- screening in nanoporous TiO2 electrodes,
tion in dye-sensitized nanocrystalline metal J. Phys. Chem. B 1997, 101(40), 7985–7990.
oxide electrodes, Phys. Rev. B 2001, 63, 23. D. Matthews, P. Infelta, M. Gratzel, Calcu-
20 5321. lation of the photocurrent-potential char-
12. D. Cahen, G. Hodes, M. Gratzel et al., Na- acteristic for regenerative, sensitized semi-
ture of photovoltaic action in dye-sensitized conductor electrodes, Sol. Energy Mater. Sol.
solar cells, J. Phys. Chem. B 2000, 104(9), Cells 1996, 44(2), 119–155.
2053–2059. 24. B. A. Gregg, A. Zaban, S. Ferrere, Dye sen-
13. C. A. Kelly, F. Farzad, D. W. Thompson sitized solar cells: energetic considerations
et al., Cation-controlled interfacial charge and applications, Zeitschrift Fur Physikalis-
injection in sensitized nanocrystalline TiO2 , che Chemie-International, J. Res. Phys.
Langmuir 1999, 15(20), 7047–7054. Chem. Chem. Phys. 1999, 212(Pt1), 11–22.
14. M. K. Nazeeruddin, A. Kay, I. Rodicio et al., 25. F. Pichot, B. A. Gregg, The photovoltage-
Conversion of light to electricity by cis- determining mechanism in dye-sensitized
X2bis(2,2* -Bipyridyl-4,4* -Dicarboxylate) solar cells, J. Phys. Chem. B 2000, 104(1),
Ruthenium(Ii) charge-transfer sensitizers 6–10.
26. H. Lindstrom, S. Sodergren, A. Solbrand 39. U. Kirner, K. D. Schierbaum, W. Gopel

et al., Li+ ion insertion in TiO2 (anatase) et al., Low and high-temperature TiO2
chronoamperometry on CVD films and oxygen sensors, Sens. Actuators B-Chem.
nanoporous films, J. Phys. Chem. B 1997, 1990, 1(1–6), 103–107.
101(39), 7710–7716. 40. J. Muscat, N. M. Harrison, G. Thornton,
27. C. J. Barbe, F. Arendse, P. Comte et al., First-principles study of potassium adsorp-
Nanocrystalline titanium oxide electrodes tion on TiO2 surfaces, Phys. Rev. B: Conden.
for photovoltaic applications, J. Am. Ceram. Matter 1999, 59(23), 15457–15463.
Soc. 1997, 80(12), 3157–3171. 41. S. Sodergren, H. Siegbahn, H. Rensmo
28. V. Shklover, M. K. Nazeeruddin, S. M. et al., Lithium intercalation in nanoporous
Zakeeruddin et al., Structure of nanocrys- anatase TiO2 studied with XPS, J. Phys.
talline TiO2 powders and precursor to Chem. B 1997, 101(16), 3087–3090.
their highly efficient photosensitizer, Chem. 42. A. Vittadini, A. Selloni, F. P. Rotzinger et al.,
Mater. 1997, 9(2), 430–439. Structure and energetics of water adsorbed
29. A. Hagfeldt, B. Didriksson, T. Palmqvist at TiO2 anatase (101) and (001) surfaces,
et al., Verification of high efficiencies for Phys. Rev. Lett. 1998, 81(14), 2954–2957.
the Gratzel-cell – a 7-percent efficient solar- 43. P. J. D. Lindan, N. M. Harrison, M. J.
cell based on dye-sensitized colloidal TiO2 Gillan, Mixed dissociative and molecular ad-
films, Sol. Energy Mater. Sol. Cells 1994, sorption of water on the rutile (110) surface,
31(4), 481–488. Phys. Rev. Lett. 1998, 80(4), 762–765.
30. P. A. Cox, Transition Metal Oxides, in Inter- 44. H. Haerudin, S. Bertel, R. Kramer, Sur-
national Series of Monographs on Chemistry, face stoichiometry of ‘‘titanium subox-
Oxford, Oxford University Press, 1992. ide’’ – Part I–Volumetric and FTIR study,
J. Chem. Soc., Faraday Trans. 1998, 94(10),
31. H. Tang, H. Berger, P. E. Schmid et al.,
1481–1487.
Optical-properties of anatase (TiO2 ), Solid
45. R. F. Howe, M. Grätzel, Electron-para-
State Commun. 1994, 92(3), 267–271.
magnetic-res observation of trapped elec-
32. H. Tang, K. Prasad, R. Sanjines et al., Elec-
trons in colloidal TiO2 , J. of Phys. Chem.
trical and optical-properties of TiO2 anatase
1985, 89(21), 4495–4499.
thin-films, J. Appl. Phys. 1994, 75(4),
46. U. Kolle, J. Moser, M. Grätzel, Dynam-
2042–2047.
ics of interfacial charge-transfer reactions
33. H. Tang, F. Levy, H. Berger et al., Urbach in semiconductor dispersions – reduction
tail of anatase TiO2 , Phys. Rev. B: Conden. of cobaltoceniumdicarboxylate in colloidal
Matter 1995, 52(11), 7771–7774. TiO2 , Inorg. Chem. 1985, 24(14), 2253–2258.
34. R. Sanjines, H. Tang, H. Berger et al., 47. B. O’Regan, M. Grätzel, D. Fitzmaurice,
Electronic-structure of anatase TiO2 oxide, Optical electrochemistry steady state spec-
J. Appl. Phys. 1994, 75(6), 2945–2951. troscopy of conduction-band electrons in
35. H. Z. Zhang, J. F. Banfield, Thermodyna- a metal-oxide semiconductor electrode,
mic analysis of phase stability of nanocrys- Chem. Phys. Lett. 1991, 183(1–2), 89–93.
talline titania, J. Mater. Chem. 1998, 8(9), 48. G. Rothenberger, D. Fitzmaurice, M. Grat-
2073–2076. zel, Optical electrochemistry spectroscopy
36. L. Kavan, M. Grätzel, S. E. Gilbert et al., of conduction-band electrons in transparent
Electrochemical and photoelectrochemical metal-oxide semiconductor-films – optical
investigation of single-crystal anatase, J. determination of the flat-band potential
Amer. Chem. Soc. 1996, 118(28), 6716–6723. of colloidal titanium-dioxide films, J. Phys.
37. W. C. Mackrodt, E. A. Simson, N. M. Har- Chem. 1992, 96(14), 5983–5986.
rison, An ab initio Hartree-Fock study of 49. G. Redmond, D. Fitzmaurice, Spectrosco-
the electron-excess gap states in oxygen- pic determination of flat-band potentials for
deficient rutile TiO2 , Surf. Sci. 1997, polycrystalline Tio2 electrodes in nonaque-
384(1–3), 192–200. ous solvents, J. Phys. Chem. 1993, 97(7),
38. R. G. Egdell, S. Eriksen, W. R. Flavell, Oxy- 1426–1430.
gen deficient SnO2 (110) and TiO2 (110) – a 50. G. Franco, J. Gehring, L. M. Peter et al.,
comparative-study by photoemission, Solid Frequency-resolved optical detection of
State Commun. 1986, 60(10), 835–838. photoinjected electrons in dye-sensitized
nanocrystalline photovoltaic cells, J. Phys. semiconductors, Chem. Mater. 1997, 9(11),

Chem. B 1999, 103(4), 692–698. 2341–2353.
51. G. Boschloo, D. Fitzmaurice, Electron accu- 63. P. Hoyer, H. Weller, Potential-dependent
mulation in nanostructured TiO2 (anatase) electron injection in nanoporous colloidal
electrodes, J. Phys. Chem. B 1999, 103(37), ZnO films, J. Phys. Chem. 1995, 99(38),
7860–7868. 14096–14100.
52. L. Forro, O. Chauvet, D. Emin et al., High- 64. H. Rensmo, K. Keis, H. Lindstrom et al.,
mobility N-type charge-carriers in large High light-to-energy conversion efficiencies
single-crystals of anatase (TiO2 ), J. Appl. for solar cells based on nanostructured ZnO
Phys. 1994, 75(1), 633–635. electrodes, J. Phys. Chem. B 1997, 101(14),
53. B. Poumellec, J. F. Marucco, F. Lagnel, 2598–2601.
Electron-transport in TiO2 -X at intermedi- 65. E. A. Meulenkamp, Electron transport in
ate temperatures 300-K less- than T less- nanoparticulate ZnO films, J. Phys. Chem. B
than 1500-K, Phys. Status Solidi A-Appl. Res. 1999, 103(37), 7831–7838.
1985, 89(1), 375–382. 66. A. vanDijken, E. A. Meulenkamp, D. Van-
54. J. Weidmann, T. Dittrich, E. Konstantinova maekelbergh et al., The kinetics of the
et al., Influence of oxygen and water related radiative and nonradiative processes in
surface defects on the dye sensitized TiO2 nanocrystalline ZnO particles upon pho-
solar cell, Sol. Energy Mater. Sol. Cells 1999, toexcitation, J. Phys. Chem. B 2000, 104(8),
56(2), 153–165. 1715–1723.
55. R. Konenkamp, Carrier transport in nano- 67. P. E. de Jongh, E. A. Meulenkamp, D. Van-
porous TiO2 films, Phys. Rev. B 2000, 61(16), maekelbergh et al., Charge carrier dy-
11057–11064. namics in illuminated particulate ZnO
electrodes, J. Phys. Chem. B 2000, 104,
56. T. Dittrich, J. Weidmann, V. Y. Timoshenko
7686–7693.
et al., Thermal activation of the electronic
68. P. E. de Jongh, D. Vanmaekelbergh, J. J.
transport in porous titanium dioxides,
Kelly, Photoelectrochemistry of electrode-
Mater. Sci. Eng., B-Solid State Materials
posited Cu2 O, J. Electrochem. Soc. 2000,
For Advanced Technology 2000, 69(SISI),
147(2), 486–489.
489–493.
69. S. Y. Huang, G. Schlichthorl, A. J. Nozik
57. R. Konenkamp, R. Henninger, P. Hoyer,
et al., Charge recombination in dye-sensiti-
Photocarrier transport in colloidal TiO2 zed nanocrystalline TiO2 solar cells, J. Phys.
films, J. Phys. Chem. 1993, 97(28), Chem. B 1997, 101(14), 2576–2582.
7328–7330. 70. A. Hauch, R. Kern, J. Ferber et al., Charac-
58. E. Moons, M. Grätzel, Photovoltage in Nano- terisation of the electrolyte-solid interfaces
crystalline TiO2 Films 1998, Unpublished. of dye-sensitized solar cells by means of
59. T. Dittrich, E. A. Lebedev, J. Weidmann, impedance spectroscopy, 2nd World Confer-
Electron drift mobility in porous TiO2 ence on Photovoltaic Solar Energy Conversion,
(anatase), Phys. Status Solidi A-Appl. Res. Vienna, European Communities, 1998.
1998, 165(2), R5–R6. 71. N. Papageorgiou, M. Gratzel, P. P. Infelta,
60. R. Konenkamp, R. Henninger, Recombina- On the relevance of mass transport in thin
tion in nanophase TiO2 films, Appl. Phys. layer nanocrystalline photoelectrochemical
A-Materials Science & Processing 1994, 58(1), solar cells, Sol. Energy Mater. Sol. Cells 1996,
87–90. 44(4), 405–438.
61. S. Burnside, J. E. Moser, K. Brooks et al., 72. Z. Kebede, S. E. Lindquist, The obstructed
Nanocrystalline mesoporous strontium ti- diffusion of the I-3(−) ion in mesoscopic
tanate as photoelectrode material for pho- TiO2 membranes, Sol. Energy Mater. Sol.
tosensitized solar devices: increasing pho- Cells 1998, 51(3–4), 291–303.
tovoltage through flatband potential engi- 73. P. M. Sommeling, H. C. Rieffe, J. M. Kroon
neering, J. Phys. Chem. B 1999, 103(43), et al., Spectral response and response time
9328–9332. of nanocrystalline dye-sensitized TiO2 solar
62. J. M. Stipkala, F. N. Castellano, T. A. cells, 14th European Photovoltaic Solar En-
Heimer et al., Light-induced charge sep- ergy Conference, Barcelona, H.S. Stephens
aration at sensitized sol-gel processed and Associates, 1997.
74. A. Solbrand, H. Lindstrom, H. Rensmo 85. R. Vogel, P. Hoyer, H. Weller, Quantum-

et al., Electron transport in the nanos- sized Pbs, Cds, Ag2s, Sb2s3, and Bi2s3 par-
tructured TiO2 -electrolyte system studied ticles as sensitizers for various nanoporous
with time-resolved photocurrents, J. Phys. wide-bandgap semiconductors, J. Phys.
Chem. B 1997, 101(14), 2514–2518. Chem. 1994, 98(12), 3183–3188.
75. N. Kopidakis, E. A. Schiff, N.- G. Park et al., 86. A. Zaban, O .I. Micic, B. A. Gregg et al.,
Ambipolar diffusion of carriers in elec- Photosensitization of nanoporous TiO2
trolyte filled, nanoporous TiO2 , J. Phys. electrodes with InP quantum dots, Lang-
Chem. B 2000, 104(16), 3930–3936. muir 1998, 14(12), 3153–3156.
76. F. Cao, G. Oskam, P. C. Searson, A solid- 87. M. Shur, in Physics of Semiconductor De-
state, dye-sensitized photoelectrochemi- vices, Prentice-Hall, Englewood Cliffs, 1990,
pp. 86–102.
cal cell, J. Phys. Chem. 1995, 99(47),
88. M. Lundstrom, in Fundamentals of Carrier
17071–17073.
Transport, Addison Wesley, Wokingham,
77. S. A. Haque, S. Sodergren, A. Holmes et al.,
1990.
Solid state dye-sensitized photovoltaic cells
89. S. J. Fonash, in Solar Cell Device Physics,
employing a polymer electrolyte, Interna- Academic Press, New York, 1981.
tional conference on photochemical conversion 90. Y. Tachibana, J. E. Moser, M. Gratzel et al.,
and storage of solar energy, Colorado, 2000. Subpicosecond interfacial charge separa-
78. T. J. Savenije, J. M. Warman, A. Goossens, tion in dye-sensitized nanocrystalline tita-
Visible light sensitization of titanium diox- nium dioxide films, J. Phys. Chem. 1996,
ide using a phenylene vinylene polymer, 100(51), 20056–20062.
Chem. Phys. Lett. 1998, 287(1–2), 148–153. 91. T. Hannappel, B. Burfeindt, W. Storck et al.,
79. U. Bach, D. Lupo, P. Comte et al., Solid- Measurement of ultrafast photoinduced
state dye-sensitized mesoporous TiO2 solar electron transfer from chemically anchored
cells with high photon-to-electron conver- Ru-dye molecules into empty electronic
sion efficiencies, Nature 1998, 395(6702), states in a colloidal anatase TiO2 film, J.
583–585. Phys. Chem. B 1997, 101(35), 6799–6802.
80. B. O’Regan, D. T. Schwartz, Efficient dye- 92. L. Dloczik, O. Ileperuma, I. Lauermann
sensitized charge separation in a wide-band- et al., Dynamic response of dye-sensitized
gap p–n heterojunction, J. Appl. Phys. 1996, nanocrystalline solar cells: characterization
80(8), 4749–4754. by intensity-modulated photocurrent spec-
81. K. Tennakone, G. Kumara, A. R. Kumaras- troscopy, J. Phys. Chem. B 1997, 101(49),
inghe et al., A Dye-sensitized nano-porous 10281–10289.
solid-state photovoltaic cell, Semicond. Sci. 93. J. van de Lagemaat, N. G. Park, A. J. Frank,
Technol. 1995, 10(12), 1689–1693. Influence of electrical potential distribution,
82. Y. Tachibana, S. A. Haque, I. P. Mercer charge transport, and recombination on the
et al., Electron injection and recombination photopotential and photocurrent conver-
sion efficiency of dye-sensitized nanocrys-
in dye sensitized nanocrystalline titanium
talline TiO2 solar cells: a study by elec-
dioxide films: a comparison of ruthenium
trical impedance and optical modulation
bipyridyl and porphyrin sensitizer dyes, J.
techniques, J. Phys. Chem. B 2000, 104(9),
Phys. Chem. B 2000, 104(6), 1198–1205.
2044–2052.
83. A. Fillinger, B. A. Parkinson, The adsorp- 94. B. Levy, W. Liu, S. E. Gilbert, Directed pho-
tion behavior of a ruthenium-based sensitiz- tocurrents in nanostructured TiO2 /SnO2
ing dye to nanocrystalline TiO2 – coverage heterojunction diodes, J. Phys. Chem. B
effects on the external and internal sensiti- 1997, 101(10), 1810–1816.
zation quantum yields, J. Electrochem. Soc. 95. H. Lindstrom, H. Rensmo, S. Soder-
1999, 146(12), 4559–4564. gren et al., Electron transport properties in
84. P. Bonhote, E. Gogniat, S. Tingry et al., Ef- dye-sensitized nanoporous-nanocrystalline
ficient lateral electron transport inside a TiO2 films, J. Phys. Chem. 1996, 100(8),
monolayer of aromatic amines anchored on 3084–3088.
nanocrystalline metal oxide films, J. Phys. 96. M. K. Nazeeruddin, P. Pechy, M. Grätzel,
Chem. B 1998, 102(9), 1498–1507. Efficient panchromatic sensitization of
nanocrystalline TiO2 films by a black dye 109. J. Nelson, Continuous-time random-walk

based on a trithiocyanato-ruthenium com- model of electron transport in nanocrys-
plex, Chem. Commun. 1997, 18, 1705–1706. talline TiO2 electrodes, Phys. Rev. B: Con-
97. M. Grätzel, Photoelectrochemical solar en- dens. Matter 1999, 59(23), 15374–15380.
ergy conversion by dye sensitization, Future 110. N. F. Mott, Conduction in glasses contain-
Generation Photovoltaic Technologies: First ing transition metal ions, J. Non-cryst. Solids
NREL Conference, Denver: AIP, 1997. 1968, 1, 1–17.
98. A. Hagfeldt, U. Bjorksten, S. E. Lindquist, 111. H. Scher, E. W. Montroll, Anomalous tran-
Photoelectrochemical studies of colloidal sit time dispersion in amorphous solids,
TiO2 -films – the charge separation process Phys. Rev. B 1975, 12, 2455.
studied by means of action spectra in the 112. A. Hagfeldt, S. E. Lindquist, M. Gratzel,
UV region, Sol. Energy Mater. Sol. Cells 1992, Charge-carrier separation and charge-tran-
27(4), 293–304. sport in nanocrystalline junctions, Sol. En-
99. T. Dittrich, P. Beer, F. Koch et al., Barrier ergy Mater. Sol. Cells 1994, 32(3), 245–257.
lowering in dye-sensitized porous-TiO2 113. C. L. Olson, R. L. Willis, T. Lutz et al. Cur-
solar cells, Appl. Phys. Lett. 1998, 73(13), rent Transients, Optical Absorbance and
1901–1903. Charge Recombination in Dye-sensitized
100. A. Stanley, B. Verity, D. Matthews, Min- Nanocrystalline Titanium Dioxide Elec-
imizing the dark current at the dye- trodes, 13th International conference on pho-
sensitized TiO2 electrode, Sol. Energy Mater. tochemical conversion and storage of solar
Sol. Cells 1998, 52(1–2), 141–154. energy (IPS-2000), Snowmass, Colorado,
101. A. Hagfedlt, M. Grätzel, Acc. Chem. Res. 2000, Unpublished.
2000, 33, 269–277. 114. N. W. Duffy, L. M. Peter, R. M. G. Ra-
japakse et al., Electrochemistry Communica-
102. J. Ferber, R. Stangl, J. Luther, An electrical
tions 2, 2000, 658–662.
model of the dye-sensitized solar cell, Sol.
115. S. A. Haque, Y. Tachibana, D. R. Klug et al.,
Energy Mater. Sol. Cells 1998, 53(1–2),
Charge recombination kinetics in dye-
29–54.
sensitized nanocrystalline titanium dioxide
103. J. Kruger, U. Bach, M. Grätzel, Modifica-
films under externally applied bias, J. Phys.
tion of TiO2 heterojunctions with benzoic
Chem. B 1998, 102(10), 1745–1749.
acid derivatives in hybrid molecular solid-
116. A. Blumen, G. Zumofen, J. Klafter,
state devices, Adv. Mater. 2000, 12(6), 447 (6 Target annihilation by random walkers,
pages). Phys. Rev. B: Condens. Matter 1984, 30(9),
104. R. A. Street, in Hydrogenated Amorphous 5379–5382.
Silicon, Cambridge University Press, Cam- 117. A. Blumen, J. Klafter, G. Zumofen, Models
bridge, 1991. for reaction dynamics in glasses, in Optical
105. J. A. Anta, J. Nelson, N. Quirke, Charge Spectroscopy of glasses, (Ed.: I. Zschokke),
transport model for disordered materials: D. Reidel, Dordrecht, The Netherlands,
application to sensitized TiO2 , Phys. Rev. B 1986, pp. 199–266.
2000(Submitted). 118. G. M. Hasselmann, G.J. Meyer, Diffusion-
106. F. Cao, G. Oskam, G. J. Meyer et al., Elec- limited interfacial electron transfer with
tron transport in porous nanocrystalline large apparent driving forces, J. Phys.
TiO2 photoelectrochemical cells, J. Phys. Chem. B 1999, 103(36), 7671–7675.
Chem. 1996, 100(42), 17021–17027. 119. J. S. Salafsky, W. H. Lubberhuizen, E. van-
107. K. Schwarzburg, F. Willig, Influence of Faassen et al., Charge dynamics following
trap filling on photocurrent transients in dye photoinjection into a TiO2 nanocrys-
polycrystalline TiO2 , Appl. Phys. Lett. 1991, talline network, J. Phys. Chem. B 1998,
58(22), 2520–2522. 102(5), 766–769.
108. A. Solbrand, A. Henningsson, S. Sodergren 120. L. Peter, D. Vanmaekelbergh, Time and
et al., Charge transport properties in dye- frequency resolved studies of photoelec-
sensitized nanostructured TiO2 thin film trochemical kinetics, in Advances in Elec-
electrodes studied by photoinduced current trochemical Science and Engineering, (Eds.:
transients, J. Phys. Chem. B 1999, 103(7), R. Alkire, D. Kolb), Wiley WCH, Weinheim,
1078–1083. 1999, pp. 77–163.
121. P. E. de Jongh, D. Vanmaekelbergh, Trap- ciency of dye-sensitized nanocrystalline

limited electronic transport in assemblies TiO2 solar cells, J. Phys. Chem. B 1999,
of nanometer-size TiO2 particles, Phys. Rev. 103(5), 782–791.
Lett. 1996, 77(16), 3427–3430. 127. J. van de Lagemaat, A. J. Frank, Effect
122. P. E. de Jongh, D. Vanmaekelbergh, Inves- of the surface-state distribution on elec-
tigation of the electronic transport prop- tron transport in dye-sensitized TiO2 so-
erties of nanocrystalline particulate TiO2 lar cells: Nonlinear electron-transport ki-
electrodes by intensity-modulated photocur- netics, J. Phys. Chem. B 2000, 104(18),
rent spectroscopy, J. Phys. Chem. B 1997, 4292–4294.
101(14), 2716–2722. 128. N. W. Duffy, L. M. Peter, K. G. U. Wija-
123. D. Vanmaekelbergh, P. E. de Jongh, Elec- yantha, Characterisation of electron trans-
tron transport in disordered semiconduc- port and back reaction in dye-sensitized
tors studied by a small harmonic modula- nanocrystalline solar cells by small am-
tion of the steady state, Phys. Rev. B 2000, plitude laser pulse excitation, Electrochem.
61(7), 4699–4704. Commun. 2000, 2(4), 262–266.
124. J. J. Kelly, D. Vanmaekelbergh, Charge car- 129. J. Bisquert, G. GarciaBelmonte, F. Fabregat-
rier dynamics in nanoporous photoelec- Santiago et al., Doubling exponent models
trodes, Electrochim. Acta 1998, 43(19–20), for the analysis of porous film electrodes by
2773–2780. impedance. Relaxation of TiO2 nanoporous
125. G. Schlichthorl, S. Y. Huang, J. Sprague in aqueous solution, J. Phys. Chem. B 2000,
et al., Band edge movement and recombi- 104(10), 2287–2298.
nation kinetics in dye-sensitized nanocrys- 130. J. S. Salafsky, Exciton dissociation, charge
talline TiO2 solar cells: a study by intensity transport, and recombination in ultra-
modulated photovoltage spectroscopy, J. thin, conjugated polymer-TiO2 nanocrystal
Phys. Chem. B 1997, 101(41), 8141–8155. intermixed composites, Phys. Rev. B:
126. G. Schlichthorl, N. G. Park, A. J. Frank, Condens. Matter 1999, 59(16), 10885–
Evaluation of the charge-collection effi- 10894.
5.4 Solid State Dye-sensitized Solar Cells – An Alternative Route Towards Low-Cost Photovoltaic Devices 475
5.4 performance of these cells, the corrosive

Solid State Dye-sensitized Solar Cells – An and volatile nature of the electrolyte rep-
Alternative Route Towards Low-Cost resents an important drawback of this
Photovoltaic Devices technology. Practical advantages might be
gained by the replacement of the liquid
Udo Bach
NTera Ltd., Dublin, Ireland electrolyte with a solid charge transport
material (CTM). Organic materials and in-
5.4.1 organic p-type semiconductors have been
Introduction tested in this regard. The major focus of
this chapter is to describe these concepts
To date, the field of applied photovoltaics is and to compare them to other recent ideas
dominated exclusively by conventional infor combining new materials and concepts
organic semiconductor technologies such to solid-state photovoltaic junctions.
as silicon or thin-film solar cells. Since
the invention of silicon solar cells in the 5.4.2
1940s, significant improvements in device Solid-state Dye-sensitized Systems
manufacture and the introduction of novel
thin-film techniques have led to a gradual When replacing the liquid electrolyte in
increase in energy conversion efficiencies dye-sensitized solar cells with a solid p-
and made their production more energy- type conductor, a range of requirements
and cost-effective. However, classical in- have to be fulfilled to preserve the high
organic p-n junction high-efficiency solar external quantum efficiencies common to
cells most probably will always demand the original photoelectrochemical devices.
the use of extremely pure starting materi- First, the p-type material needs to have
als and somewhat sophisticated deposition
electronic levels into which holes can be
procedures.
injected from the oxidized or excited state
A very promising alternative to classi-
of the dye. The redox levels of the dye and
cal, inorganic p-n junction solar cells is
the p-type material therefore have to be
the concept of nanoporous, dye-sensitized
adapted carefully. An intimate contact be-
photoelectrochemical solar cells, which
tween the sensitized metal oxide and the p-
was first introduced by Grätzel and cowork-
ers [1]. In these cells, dye molecules, type material is vital to assure fast injection
adsorbed onto a nanostructured TiO2 and regeneration processes (Fig. 1). This
electrode sensitize photoinduced charge implies either the growth or deposition
separation across the interface. These cells of one semiconductor inside a preformed,
exhibit external quantum efficiencies of sensitized porous film of its counterpart or
up to 10% and can be made from less the in situ formation of the sensitized com-
expensive starting materials and by sim- posite. Direct formation of the sensitized
ple printing techniques [2]. Therefore, they junction would be appreciable; however,
represent a promising alternative to clas- charge collection within the two indepen-
sical inorganic p-n junction solar cells. dent semiconductor networks, in which
Both industry and academia have re- at least one semiconductor is formed
search efforts toward development and from nanometer-sized inorganic semi-
commercialization of dye-sensitized solar conductor particles, demands intimate
cells. However, despite the very promising contact between the particles. Reduced
φ (V ) Fig. 1 Scheme for the electron

transfer processes (injection,
−1 regeneration, and recapture)
cb inj. (D+/D*) that occur in solid-state
dye-sensitized heterojunctions.
0
rec.
hν
reg.
1
(D+/D)
TiO2 Dye HTM Au
interparticle contact was specified as a rea- lithium salts were added to the phase of
son for the poor photovoltaic performance organic p-type conductors [6, 7]. It was
of early sensitized TiO2 /polymer com- postulated that the lithium salt might
posites [3–5]. However, Greenham and screen space charge layers, which are
coworkers realized a CdSe/MEH-PPV expected to build up in the illuminated het-
(polymethoxyethyl hexyloxy-p-phenylene- erojunction. The elimination of the space
vinylene) solar cell made via simple spin charge control of the photocurrent could
coating techniques, with a maximum in- thereby explain the observed increase in
cident photon-to-electron conversion effi- photovoltaic performance.
ciency (IPCE) of 12%, suggesting efficient The contacting electrodes also have to
electron percolation in the CdSe net- be carefully chosen. The cathode mate-
work [4]. Decapping of the nanoparticles rial needs to form an ohmic contact to
before the formation of the composite was the p-type material so as to avoid injec-
essential for the functioning of the de- tion barriers that would impede efficient
vice, a fact that has to be considered in charge collection. The anode needs to
future attempts to form composites with form a blocking contact to the p-type
dye-derivatized nanometer-sized particles. material so as to avoid internal shortcir-
Once a positive charge is injected suc- cuits. Generally, transparent conducting
cessfully into the p-type material, it has to metal oxide layers such as FTO (fluo-
be transported efficiently to the contacting rine doped tin(IV)oxide) or ITO (indium
electrode before interfacial recombination doped tin oxide) are used as the backcon-
can occur. To allow for such efficient tact onto which the nanocrystalline film
charge collection, both the p- and n-type is applied. Iodine is unique, because it
material need sufficient conductivities so shows a very high overpotential at these
as not to impose high internal resis- materials. However, most solid p-type ma-
tance and thereby limit charge collection. terials will form ohmic contacts with these
Screening of charges by the displace- transparent metal oxides causing internal
ment of ions plays an important role short circuits at the points of contact [8].
in the charge collection process in pho- This applies both to organic and inorganic
toelectrochemical solar cells. Solid-state materials. Consequently, whenever trans-
dye-sensitized heterojunctions also exhibit parent conducting metal oxide layers are
increased photovoltaic performance, when used in solid heterojunctions, care has to
ionic species introduced in the form of be taken to avoid direct contact between
the p-type component of the sensitized metal oxide electrode and the CTM inside
heterojunction and the back contact of the the sensitized heterojunction.
photoanode. The application of organic CTMs in
photovoltaic devices is a new research
5.4.2.1 Organic CTMs Applied to field, which still is in its infancy. Figure 2
Nanoporous Dye-sensitized Metal Oxide overviews the organic charge transfer ma-
Electrodes terials that were scrutinized with regard
Organic electronic materials nowadays are to their application in solid-state dye-
successfully replacing inorganic semicon- sensitized solar cells. A much wider range
ductor components in a growing num- of materials is mentioned in the patent
ber of electronic devices such as xero- literature [13–15].
graphic ink drums or light emitting diodes The charge transfer events taking place
(LEDs) [9, 10]. Further applications will in dye-sensitized heterojunctions are de-
soon follow including transistors [11], or picted in Fig. 1. Visible light is absorbed
entire integrated circuits fabricated on by the sensitizing dye. Electron injection
plastic [13]. Very recently, these materials from the excited state of the dye into the
conduction band of TiO2 (1) is followed
also have been scrutinized with regard to
by subsequent hole transfer from the pho-
their applicability in photovoltaic devices.
tooxidized dye to the organic CTM, regen-
In the following section, the application
erating the dye’s original ground state (2).
of organic materials in conjunction with
dye-sensitized mesoporous metal oxide Dye∗ −−−→ Dye+ + e− (cb TiO2 ) (1)
electrodes is reviewed, whereas alterna- + +
Dye + CTM −−−→ Dye + CTM (2)
tive conceptual approaches toward organic
solar cells that do not directly involve dye- The latter process competes with recap-
sensitization are briefly discussed in the ture of the injected electron by the oxidized
final section of this chapter. dye (3):
Organic synthetic chemistry is a pow-
erful tool used to develop materials with Dye+ + e− (cb TiO2 ) −−−→ Dye (3)
very specific electronic and optical proper-
The double injection process described
ties. A large variety of electronically active
by Eqs. (1) and (2) generates mobile major-
materials with excellent, reproducible ma-
ity carriers in both the TiO2 and the hole
terial properties became available over the conductor phases, which are subsequently
last few years. These organic materials are collected at the contact electrodes. A sec-
still lagging their inorganic counterparts ond mechanism can be postulated that
regarding their semiconducting proper- involves photoinduced hole transfer from
ties such as conductivity and mobility. the electronically excited dye to a neighbor-
Organic materials, however, offer an al- ing charge transport molecule, followed
most unlimited potential to vary their by regeneration of the photoreduced state
specific properties. Furthermore, most of by electron injection into the TiO2 . Both
them can be prepared easily in their amor- mechanisms lead to the same initial charge
phous state. This is important because separated state, which results in an elec-
crystallization would most probably impair tron in the conduction band of the TiO2
the formation of a good contact between and a positive charge, localized on a hole
the mesoporous surface of the sensitized conductor molecule.
OMe
H
N
N N
n
Polypyrrol
MeO
p-OMeTPD
OCH3
N OCH3
OCH3
OCH3
N
N
OCH3
OCH3 N
OCH3
OCH3
spiro-OMeTAD
Fig. 2 Different CTMs used in solid-state dye-sensitized solar
cells.
It is important to note, that hole con- because of recombination. Because they

duction to the counter-electrode does not are generated throughout the semicon-
involve mass-transfer. When using low- ductor and away from the junction, ex-
molecular-weight CTMs, the underlying pensive high-purity materials are required
mechanism for charge transport generally to maintain the minority carrier diffusion
is assumed to be a hopping mechanism as length at a level at which current losses
described by Bässler and coworkers [16], caused by recombination are avoided.
whereas charge transport in conjugated The first to describe a solid-state device
polymers is generally explained via a po- based on an organic CTM (p-OMeTPD,
laron mechanism [17]. Fig. 2) were Hagen and coworkers [18].
Similar to photoelectrochemical solar The organic hole conductor was applied to
cells, the concept of solid-state sensitized a Ru(dcbpy)2 (SCN)2 -sensitized nanocrys-
heterojunctions has the great advantage talline TiO2 electrode via thermal evapo-
over conventional inorganic solar cells that ration. A thin layer of gold, applied via
only majority carriers are involved in the vacuum deposition techniques, formed the
photoelectric conversion process. Photo- back contact. IPCEs of these devices were
electric conversion in conventional inor- still low (IPCE = 0.2%). Space charge ef-
ganic p-n junction solar cells involves mi- fects were found to play a dominating
nority carriers whose lifetime is restricted role in this type of devices. The authors
attribute the low mobility of electrons in higher than previously reported, but still
the final device responsible for the poor did not exceed 0.1%.
performance. Incomplete pore filling and Significant improvement in efficiency
low conductivities might have been fur- were achieved recently in a novel solar
ther reasons for the low efficiency, because cell device applying the amorphous
thermal evaporation is not likely to trans- organic CTM 2,2$ ,7,7$ -tetrakis(N ,N -di-p-
port organic material deep into the porous methoxyphenylamine)9,9$ -spirobifluorene
structure and because the charge carrier [7] (spiro-OMeTAD; Fig. 2). Photoinduced
density inside the hole conductor matrix charge carrier generation was shown to
was low and merely influenced by acciden- be very efficient. A solar cell based on
tal doping as a result of the presence of spiro-OMeTAD converts photons to elec-
oxygen. tric current with a strikingly high yield
The first solid-state dye-sensitized het- of 50% [20]. The new hole conductor
erojunction of TiO2 and a semiconduct- contained a spiro-center, which was in-
ing polymer was reported by Murakoshi troduced to improve the glass-forming
and coworkers [19, 7]. They formed a properties and prevent crystallization of
solid-state heterojunction by photoelec- the material. Its glass-transition tempera-
trochemical polymerization of pyrrole ture of Tg = 120 ◦ C, measured by differen-
into a nanoporous TiO2 film, sensitized tial scanning calorimetry, is much higher
with Ru(dcbpy)2 (SCN)2 . Direct contact of than that of comparable hole conduc-
polypyrrole and the FTO back contact was tors. The methoxy groups were introduced
suppressed by electrodeposition of a thin to match the oxidation potential of the
TiO2 film before the adsorption of the dye CTM to that of the applied sensitizer
onto the nanocrystalline TiO2 electrode. Ru(dcbpy)2 (SCN)2 .
The doping density of the polymer was Figure 3 shows a cross-section scheme
controlled electrochemically. Solar energy of the sensitized TiO2 /spiro-OMeTAD
conversion efficiencies were somewhat heterojunction mentioned earlier. The
e− Thin gold electrode

Dye-sensitized nanocrystalline
heterojunction
− Dense TiO2 layer
+
Transparent, conducting layer
(F-doped SnO2)
Glass support
h+ Organic hole conductor

e− Dye molecule
Light Semiconductor particle
Fig. 3 Cross-section scheme for a solid-state dye-sensitized solar

cell as described in by Bach and coworkers [6].
working electrode consisted of conducting photovoltaic performances were reported

glass (F-doped SnO2 ) onto which a com- recently for comparable systems [25].
pact TiO2 layer was deposited by spray Picosecond transient absorption laser
pyrolysis [21]. This avoids direct contact spectroscopy was used to scrutinize the
between the hole conductor and the SnO2 , dynamics of the photoinduced charge sep-
which would short-circuit the cell. A 4.2- aration process [26]. The dye regeneration
µm thick mesoporous film of TiO2 was process [2] could be time-resolved, and was
deposited by screen printing onto the shown to proceed primarily on the picosec-
compact layer [22] and derivatized with ond timescale with multiphasic kinetics.
Ru(dcbpy)2 (SCN)2 . The hole conductor Therefore it is at least one order of mag-
was introduced into the mesopores by nitude faster than the dye regeneration
spin coating a concentrated solution of process involving iodine or iodide elec-
OMeTAD onto the TiO2 film and sub- trolytes [2, 27]. The high hole injection rate
sequent evaporation of the solvent. A insures that the quantum efficiency for the
semitransparent gold back contact was overall charge separation process across
evaporated on top of the hole conductor the junction is close to unity. This may be
under vacuum. a key factor behind the high external quan-
The conductivity of the device was con- tum efficiency for charge carrier collection
trolled by partial chemical oxidation of the obtained for spiro-OMeTAD compared to
hole conductor, before deposition onto the other solid-state hole conducting materials
sensitized TiO2 substrate to increase the studied to date. Transient absorption spec-
number of free charge carriers [23]. Addi- troscopy also gave indications that in the
tion of the lithium salt Li[(CF3 SO2 )2 N] to specific junction, electron injection into
the spin coating solution of the hole con- TiO2 proceeds before hole injection into
ductor resulted in a strong performance spiro-OMeTAD.
increase in the final device. The under- Analysis of the front and backside il-
lying mechanism remained unidentified lumination photoresponse spectra proved
although charge screening due to partial to be a powerful tool to probe interfa-
ionic mobility inside the hole conductor cial charge recombination during charge
matrix and/or the effect of the present collection [20]. Figure 5 shows a typical ex-
lithium ions on the flat band potential ample of a set of photoresponse spectra
of TiO2 [24] were postulated as possible obtained on front and backside illumina-
mechanisms. tion, compared to the absorptivity of the
Figure 4 shows the photocurrent ac- sensitized TiO2 layer. At the absorption
tion spectra of three typical cells under maximum of the dye, the IPCEfront (illu-
short circuit conditions [20]. Three dif- mination through the conducting glass) ex-
ferent dyes were used to sensitize the hibits a maximum, whereas the IPCEback
heterojunction. In all three cases the action spectrum (illumination through the semi-
spectrum matches closely the absorptivity transparent gold layer) exhibits a mini-
spectrum of the sensitizing dye. Solid-state mum. Even when corrected for the trans-
heterojunctions sensitized with the mero- mission losses of the conducting glass
cyanine dyes (structure see Fig. 4) used and gold electrode, respectively, the IPCE
in this study showed higher peak IPCE value for backside illumination mode is
values when compared to Ru(II)L2 (SCN)2 - one order of magnitude smaller than the
sensitized junctions. Strongly improved corresponding value for IPCEfront . This
50
40
30
IPCE
[%]
20
Mc2
10
Mc3
Ru(II)L 2(SCN) 2
0
400 500 600 700 800
Wavelength
[nm]
O OH
O
N
S
S S
N
O OH
O
N
S
S S
N S
Mc3 O N
OH
O
Fig. 4 Photocurrent action spectra (IPCE) obtained for solid
TiO2 /spiro-OMeTAD heterojunctions, sensitized with three different
sensitizers. (For device structure, see Fig. 3.) The dyes were absorbed from
above solution for 48 hours at room temperature. All devices are based on
4-µm thick nanocrystalline TiO2 electrodes and spiro-OMeTAD as hole
conducting layer (0.070 mol% spiro-OMeTAD++ (PF6 ) in respect to
spiro-OMeTAD; the spin coating solution contained 15 mM
Li[(CF3 SO2 )2 N]). The counter electrode was a 10-nm thick gold layer.
100 Fig. 5 Set of photoresponse spectra

obtained under front and backside
80 illumination conditions. The dotted
curves show the IPCE spectra as
60
measured, whereas the full curves
IPCE
corresponds to the corrected IPCE

[%]
spectra, which take adsorption and

40
reflection losses through the
corresponding window materials (FTO
20 glass or semitransparent gold layer,
respectively). The top curve represents
0 the absorptivity of the device, which
400 500 600 700
corresponds to a calculated IPCE value
Wavelength assuming no reflection or
[nm] recombination losses.
front-IPCE back-IPCEcorr
front-IPCEcorr (100-T) [%]
back-IPCE
finding strongly suggests that recombi- sites are represented by Ti (IV) ions on the
nation of photogenerated electrons while surface, which, because of their Lewis acid-
migrating toward the collecting electrode ity, are prone to interact with the pyridine
is a major loss mechanism. Applying a derivative. Similar results were obtained by
simple recombination model, a free elec- Huang and coworkers [28], who attributed
tron diffusion length of about 1 µm could the observed increase in Voc to either a
be determined for the observed junction. charging of surface states or a shift of the
Charge recombination occurring during conduction band edge.
charge collection therefore could be iden- Different techniques have been pro-
tified as major loss mechanism of the cell. posed to fill sensitized nanoporous struc-
A strong increase in energy conversion tures with organic material. Evapora-
efficiency could be observed when ter- tion [19], electrodeposition [7] and spin
tiarybutylpyridine was introduced into the coating [6] techniques were described in
matrix of the organic hole conductor [20]. this respect. Little is known regarding inti-
Although the typical open circuit volt- macy of the contact between the sensitized
age (Voc ) in absence of tertiarybutylpyri- metal oxide surface and the organic CTM
dine reaches approximately 350 mV, it when simple spincoating procedures are
can easily exceed 700 mV in the pres- used. Therefore complete pore filling still
ence of tertiarybutylpyridine. This leads remains a critical problem. Although spin
to an increase in white light conversion coating or dip coating techniques are very
efficiency from 0.74% to 1.75% (AM1.5, simple and inexpensive, electrodeposition
10 mW cm−2 ). Very similar effects were might still be a more controlled technique.
reported for classical sensitized I − /I3 − In situ polymerization of monomeric pre-
electrolyte/TiO2 junctions [2]. The authors cursors appears to be the most reliable
rationalized the effect in terms of 4-tert- way to fill nanoporous electrodes with
butylpyridine adsorption at the TiO2 sur- large-molecular-weight polymers. Very re-
face blocking surface states that are active cently it was shown that low-molecular-
as intermediates in the interfacial charge weight organic semiconductors also can
recombination. They presumed that these be electrodeposited into nanoporous TiO2
electrodes [29], which thereby represents a combinations of wide band gap p-type and
very powerful technique to deposit poorly n-type semiconductors are energetically
soluble materials sush as perylenes or ph- favorable. Organic and organo-metallic
thalocyanines into porous structures. dyes easily can be adjusted with regard
Many questions regarding the function to their redox potentials and absorption
of these sensitized inorganic-organic hy- characteristics to match the band positions
brid materials still remain unsolved. Most of these materials. In principle, the dye also
importantly, the factors controlling the might be chosen to function as a flexible
electronic properties of the junction have buffer between the two nanocrystalline
to be identified and understood. First steps phases, reducing strain effects, which arise
toward a better control of the interface have as a result of the lattice mismatch of the
been made. The dipole field, present at the two semiconductor materials. A potential
interface because of the self-assembled advantage of inorganic semiconductors
monolayer of organic molecules proved over their novel organic counterparts
to strongly affect the junction’s electronic is their generally high charge mobility.
properties [30]. Further systematic studies However, the very limited choice of
will hopefully help to widen the under- potentially interesting materials is a clear
standing of such devices, leading the way drawback, compared to the wide choice of
to increased device performances. organic CTMs.
Two mechanisms can contribute to the
5.4.2.2 Inorganic Nanoporous formation of the initial charge separated
Dye-sensitized Heterojunctions state. The electronic excitation of the dye
In dye-sensitized inorganic heterojunc- either could be followed by electron injec-
tion solar cells, a monolayer of dye is tion into the n-type semiconductor or by
sandwiched between two wide band gap hole injection into the p-type semiconduc-
semiconductors; one of them exhibits a p- tor. Both mechanisms have been shown
type and the other an n-type conduction to occur and both are likely to contribute
mechanism. Inorganic p-type semicon- to the charge separation mechanism [32].
ductors were successfully applied in an However, no studies were undertaken to
attempt to replace the liquid electrolyte differentiate which of the two possible
in photoelectrochemical solar cells, and mechanisms mainly is responsible for the
in fact the first solid-state dye-sensitized initial formation of the charge separated
photovoltaic device described in the litera- state.
ture was based on a wide band gap p-type Although dye sensitization of wide band
semiconductor material [31]. gap n-type semiconductors has been com-
Such junctions mainly are of inorganic prehensively studied, little is known re-
character, however, the dye present at the garding the sensitization of p-type materi-
junction’s interface still plays the central als. Sensitized electron injection into an n-
role in the initial formation of the charge type semiconductor and hole injection into
separated state. The sensitizer takes on a p-type material are analogous processes.
the role of both the light absorber and The misbalance might be explained simply
acts as the link between the two inorganic in terms of the wider availability of sta-
materials. The choice of inorganic wide ble wide band gap n-type materials, which
band gap semiconductors applicable in allow for the preparation of nanoporous
such junctions is small, and only few films with excellent electronic properties.
Only a few materials were studied con- a columnar morphology, oriented per-
cerning their applicability to dye-sensitized pendicular to the substrate surface could
hole injection processes. Among those be realized, using an electrodeposition
are different copper(I) compounds (e.g. technique. These highly anisotropic dye-
Cu(I)SCN, Cu(I)I, Cu2 (I)O [33–35]) and sensitized films were shown to form a
nickel(II) oxide [36]. Photovoltaic perfor- complete inner surface electrical contact
mances of such devices are orders of when p-type CuSCN was electrodeposited
magnitudes poorer than those of classical into the pores (Fig. 6). These junctions ex-
dye-sensitized photoelectrochemical solar hibited external quantum efficiencies of
cells based on n-type materials. Substan- up to 50%, and incident energy conver-
tial advantages could arise if an efficient sion efficiencies of 1.5%. Dye-sensitized
photo–hole injection process would be heterojunctions made from colloidal ZnO
available. The formation of solid-state tan- particles on the other hand, showed maxi-
dem solar cells would become feasible, mum external quantum efficiencies of only
and a quantum step in device efficiency of 20%. It could thereby be shown that lattice
dye-sensitized solar cells could be at reach. mismatch is not an intrinsic problem of
However, because of the poor performance dye-sensitized inorganic heterojunctions
of all known photocathodes, a combination but can be overcome by choosing ap-
propriate device assembly techniques and
of available photoanodes and photocath-
materials. The sensitizing dye might also
odes to a tandem device always results in a
exhibit a certain buffer function between
device that is photovoltaically less efficient
the two crystal structures, especially if the
than the photoanode on its own. The con-
dye is substituted with two sets of at-
cept for electrolyte-based tandem cells ex-
tachment groups, each of them strongly
ists. However, it contains strong potential
interacting with one of the semiconductor
to improve the photovoltaic performance
components.
in both electrolytic and in solid-state, dye-
Solid-state dye-sensitized TiO2 /CuI het-
sensitized solar cells.
erojunction were recently realized by Ten-
Analogous to their photoelectrochemi- nakone and coworkers [38]. The authors
cal counterparts, efficient dye-sensitized reported on surprisingly high solar energy
solid-state solar cells require a nanoporous conversion efficiencies, showing that in
junction structure. The first solar cell the future it might be feasible to form solid-
of this type was reported in 1995 [33] state dye-sensitized solar cells with similar
in form of a TiO2 /Cu(I)I heterojunction, efficiencies as their photoelectrochemical
sensitized with a cyanidine dye. When counterparts.
forming an inorganic interpenetrating net-
work on the nanometer scale, the crystal 5.4.3
lattice mismatch of the two inorganic Alternative Organic Solid-state Approaches
semiconductors is likely to impair the for-
mation of a good contact between the 5.4.3.1 Organic Solar Cells
two nanocrystalline structures. Recently, Thin films of organic pigments are known
O’Regan and coworkers reported on a sen- to exhibit semiconducting properties from
sitized nanocomposite of ZnO and CuSCN their use in xerography; Morel, Ghosh,
forming a bicontinuous network hetero- Feng and coworkers [39, 40] reported on
junction [37]. Nanoporous ZnO films with a solar cell in 1978 based on the Schottky
Fig. 6 Scheme for a sensitized −
SnO2
nanocomposite of ZnO and CuSCN, as
described by O’Regan and
coworkers [37]. (a) Layered structure of Dense ZnO
the dye-sensitized heterojunction.
Glass support
(b) Fractured edge of a complete DSH, CuSCN
showing the SnO2 layer, the porous
ZnO +
ZnO/dye/CuSCN layer, and some of the dye + +
overlying CuSCN. (c) Fractured edge of CuSCN h+
a porous ZnO layer, without CuSCN, e−
showing the columnar structure
Aluminium
perpendicular to the substrate.
Graphite
(d) Schematic of the ZnO/dye/CuSCN
Light
interface with dye molecules shown to
scale, but at low packing density.
(a)
1 µm
junction of aluminum and a merocyanine

dye, which showed an overall sunlight
conversion efficiency of 0.7% – about 35
times higher than any efficiency that
had been reported on organic films (b)
before. Following that, interest in organic 100 nm
solar cells strongly increased, particularly
driven by the research effort undertaken
by companies such as Exxon, Xerox,
Kodak, and Shell in the 1980s. However,
white light conversion efficiencies never
exceeded 1%, so that the general interest in
this technology decreased after a number (c)
of years. Several comprehensive reviews 20 nm
describing organic solar cells in great in-
depth were published since [41–44]. A e− ZnO
short summary follows, to allow for the
comparison with alternative technologies.
In thin-film organic solar cells, light is CuSCN h+
absorbed by a dye layer, typically of a thick-
(d) • = Dye
ness of several tens of nanometers. Light
absorption leads to the formation of an
exciton (an electron/hole pair, which is
strongly bound). These excitons can give the formation of a charge carrier depletion
rise to free charge carriers by dissociation layer. According to the origin of this built-
under the influence of a strong electric in field, one can distinguish three different
field. In general, such fields develop close types of organic solar cells. Most organic
to the interface of two materials exhibit- photodiodes are based on the rectifying
ing different work functions because of junction of an organic semiconductor and
a metal, forming a Schottky type cell such control of the charge carrier density via
as the cell reported by Morel and cowork- chemical doping with electron acceptors
ers. A rectifying contact can also be formed such as iodine for p-type materials [48]
by the formation of a p-n junction between or with donors such as H2 for n-type
the light-absorbing, semiconducting dye materials [49] can improve drastically the
and a second semiconductor of opposite conductivity and device performance.
conduction type. This second semicon- An interesting approach partially over-
ductor either can be an inorganic wide coming the problem of low conductivity
band gap semiconductor or a second light- and limited exciton diffusion widths is
absorbing, semiconducting dye, giving rise to combine two organic p-n junctions in
to an organic/inorganic or organic/organic series, leading to a tandem cell with a dou-
p-n heterojunction. bled photovoltage, compared to the simple
The main limitation of all three device p-n structure device [50].
types is the limited diffusion length of ex-
citons in organic materials, which usually 5.4.3.2 Interpenetrating Networks
does not exceed several tens of nanome- As described in the previous section, the re-
ters. Therefore the photovoltaically active combination of an exciton before it reaches
region in a thin-film solar cell is limited the built-in field region is the main loss
to a very thin layer close to the junc- mechanism in organic thin-film solar cells.
tion. This usually is too thin to assure This problem can be overcome by the
good light harvesting over a wide spectral formation of an interpenetrating p-n het-
range. Very similar to conventional inor- erojunction in which the average width
ganic p-n junction solar cells, the exciton of the p-type and n-type domains are
lifetime strongly decreases in the presence smaller than the exciton diffusion length,
of bulk impurities, demanding highly pu- thereby guaranteeing that every exciton
rified dyes. Heterojunction formation by can reach the interface before it decays.
two different organic dyes can improve the However, the role of the electric field in
light harvesting, especially if their absorp- these devices is questionable because the
tion spectra are complementary to each theoretical depletion layer width generally
other [45, 46]. However, layer thicknesses is much wider than the dimension of phase
have to be optimized to prevent opti- segregation. It might be more exact to as-
cal filter effects caused by photovoltaically sume the totality of the interpenetrating
inactive areas away from the junction. At- network to be depleted with exciton split-
tempts were made to widen the thickness ting occurring via an interfacial electron
of the photovoltaically active area by in- transfer mechanism. Similar to thin-film
corporation of a coevaporated layer of the solar cells, the rectifying junction can be
two organic, semiconducting device com- formed by two organic semiconductors, or
ponents, resulting in a significant increase by a blend of an organic and an inorganic
in device efficiency [47]. semiconductor.
Another limitation of organic thin-film Semiconducting polymers are superior
solar cells is the generally poor conductivity to their low-molecular-weight analogs,
of the organic layer. Organic devices with regard to their applicability in in-
commonly show strong sublinearity of the terpenetrating junctions. This is due to the
photocurrent at illumination intensities fact that charge transport in semiconduct-
close to AM1.5 (100 mW cm−2 ). However, ing polymers mainly proceeds along the
polymer chain, so that percolation starts al- and the polymer take over the function of
ready at low concentration thresholds (for the light absorber. Among those, nanocrys-
a review on polymer solar cells see Ref. 51). talline heterojunctions of TiO2 either with
Poly(p-phenylenevinylene) derivatives PPV [59] or polythiophene [60] derivatives
(PPVs) are the most prominent examples have received particular interest. Several
of polymers used in organic solar cells. different techniques were applied to form
Depending on their substituents they ei- the heterojunctions. Commonly the first
ther can exhibit p-type or n-type behavior. step was the formation of a nanoporous
The first junction of this type was re- thin film of TiO2 via spray pyrolysis or
ported in 1995, using a phase-segregated via deposition of a colloidal paste followed
thin film of MEH-PPV as p-type material by a sintering step while the polymer was
and CN-PPV (cyano-p-phenylenevinylene) applied via a spincoating technique or (in
as n-type material [52]. The phase segrega- the case of polythiophene) via electrode-
tion was on the scale of 10–100 nm and position. IPCE values reported on these
IPCE values as high as 6% were recorded. systems are still very low (below 3%).
Recently, significantly improved device However, just recently, highly efficient
performances were reported based on a TiO2 /polythiophene solar cells exhibiting
polythiophene/PPV-derivative heterojunc- peak IPCE values of up to 10% were
tion, assembled via a lamination technique accomplished [61]. Photovoltaic cells of
with IPCE values reaching 29% [53]. The comparable efficiency were obtained when
cross-section scheme and the molecular polycrystalline films of low band gap semi-
structures can be seen in Fig. 7. conductors such as CdSe, CdS, or CuInSe2
A different approach to form or- are contacted with a thin film of polythio-
ganic interpenetrating heterojunctions is phene via electrodeposition [62]. Overall
to dope strongly a p-type conducting white light conversion efficiencies of 1.3%
polymer with a low-molecular-weight ac- at 56 mW cm−2 were reported for a nanos-
ceptor until charge percolation occurs tructured CdSe/poly(3-methylethiophene)
inside the acceptor phase. The blends junction [63].
attracting most attention are those of One particularly interesting aspect of
PPV- or thiophene derivatives doped with nanoparticle/polymer solar cells is that
fullerenes as acceptor, reaching 23% IPCE charge ‘‘hopping’’ from one semiconduc-
for polythiophene/C60 junctions [54] and tor particle to another takes place even in
29% for PPV/C60 junctions [55]. Perylene blends prepared from spin coating solu-
nanocrystals were also used as doping actions containing both the polymer and the
ceptor [56]. Different theoretical models colloidal nanoparticles. Usually, those de-
were developed to describe the functioning vices are not exposed to successive heat
principals of these devices [57, 58]. treatments. This is surprising because
nanocrystalline semiconductor films used
5.4.3.3 Nonsensitized Inorganic/Organic in photoelectrochemical solar cells gen-
Junctions erally are believed to need such a heat
Recently, increased interest was raised in treatment to induce interparticle contact
the photovoltaic properties of composites and ‘‘necking.’’ Less expensive, flexible
of semiconducting polymers and nano- dye-sensitized solar cells based on plas-
sized semiconductor particles, in which tic substrates could be realized if this heat
either the polymer or the semiconductor treatment could be made redundant.
O O
CN
n
NC
O O
S n
POPT MEH-CN-PPV
(a)
AI or Ca
Glass substrate
Acceptor layer
Donor layer
ITO or
PEDOT on Au
Glass substrate
(b)
Fig. 7 Chemical structures and device structure for an organic
..
polymer photodiode as described by Granstrom and
coworkers [53]. (a) Chemical structures of the polymers used in the
devices. POPT is a regioregular phenyl-octyl substituted
polythiophene. MEH/CN-PPV is a cyano-derivative of
poly(p-phenylene vinylene), with a large electron affinity as a result
of the electron-withdrawing cyano groups. (b) Device structure.
Both polymers (see (a) for chemical structures) were dissolved in
chloroform or toluene (5 mg ml−1 ). For the top half of the device,
aluminum or calcium contacts were thermally evaporated on glass
substrates and the acceptor material, MEH/CN-PPV (and a small
amount of POPT, usually 5%), was spin-coated on top of the metal
electrode. The other half of the device comprises the donor
material, POPT (and a small amount of MEH/CN-PPV, usually
5%), either spin-coated on ITO substrates or glass-coated with
poly(ethylene dioxythiophene) doped with polystyrene sulfonic
acid. To ensure a low contact resistance, a thin layer of gold
(∼10 nm) was thermally evaporated on the glass slide before
spin-coating the PEDOT film from a water solution. The thickness
of the PEDOT layer was 100 nm. To get the desired structure of the
POPT, this half of the device was heated to 200 ◦ C under vacuum
before the device was laminated together by applying a light
pressure while one half was still at elevated temperature. The total
thickness of the organic semiconductive layer was 70–80 nm, and
the active area 2.5 mm2 .
Various examples of composite nanopar- components. Those organic materials have

ticle or polymer films made via spin found their way into fields, which until re-
coating can be found in the literature [3, cently were purely dominated by inorganic
4, 64]. Most examples actually show that device physics. Organic LEDs, for ex-
charge percolation inside the nanopar- ample, are already outperforming their
ticle network is reduced strongly, com- inorganic counterparts in certain types
pared to nanocrystalline, sintered films. of applications. Very often the pioneer-
Arango and coworkers [3] reported a de- ing work that was done to realize those
crease in photoconductivity of 2 orders novel applications finally gave access to
of magnitude when comparing a sin- materials that also showed very interest-
tered polymer-coated TiO2 film, to a ing photovoltaic properties. Examples are
composite made via spincoating of a PPV derivatives developed for their use
TiO2 -colloid/polymer solution. The rein organic LEDs [51] and pentacene single
duced interparticle contact also leads to crystals studied by Lucent Technologies for
low photovoltaic performances. An excep- use in organic transistors [11]. It was only
tion to these observations is the CdSe/PPV at a later stage that interesting photovoltaic
composite solar cell reported by Green- properties of such materials were scruti-
ham and coworkers [38]. Peak IPCE values nized [52–55, 65]. The organic compounds
of 12% were obtained from spin-coated used in these early studies were poorly op-
composite thin films with a weight ra- timized with regard to their photovoltaic
tio of CdSe (φ 5 nm): MEH-PPV = 9 : 1 properties; nonetheless, those initial pro-
(film thickness = 300–500 nm). Removal totypes showed impressive performances.
of surface-adsorbed species from the col- Triggered by these initial promising re-
loid before the cell assembly was shown to sults, new efforts currently are under way
be crucial for interparticle electron transfer to develop tailor-made organic materi-
to occur. als with improved optical and electronic
properties for their use in solar cells.
5.4.4 These second-generation organic photo-
Conclusion and Outlook voltaic materials almost necessarily will
lead to significant improvements in device
A large number of different approaches to- performances. Along with the develop-
ward solid-state dye-sensitized or organic ment of novel materials, there is the need
solar cells were recently described in the to gain a better understanding of the dif-
literature, and some of these are briefly ferent concepts described in this chapter.
described in this chapter. The efficien- Most of these are only poorly understood,
cies of these novel prototype devices are when compared to well-established sys-
constantly rising, some of them showing tems such as photoelectrochemical solar
energy conversion efficiencies of up to 2%. cells. Nevertheless, very exciting advance-
Such concepts not only are of scientific but ments were realized over the last years
also of potentially economic interest, indi- and the field of organic photovoltaics is
cating that the realization of less expensive, steadily progressing. Organic solar cells
efficient plastic solar cells might soon be at have the potential for very low produc-
hand. These recent developments strongly tion costs at reasonable high efficiencies.
benefit from the continued success in These properties make them interesting
the synthesis of novel organic electronic for their application in mobile electronic
consumables, in which high efficiencies 19. K. Murakoshi, R. Kogure, S. Yanagida, Chem.

and long-term stability in extreme environ- Lett. 1997, 5, 471–472.
20. U. Bach, J. Krüger, M. Grätzel, SPIE proceed-
ments are not an ultimate requisite. There-
ings San Diego, San Diego, Calif., 2000, in
fore, the predescribed concepts emerge as print.
a very interesting and viable option for 21. L. Kavan, M. Grätzel, Electrochim. Acta 1995,
future low-cost plastic solar cells. 40, 643–652.
22. C. J. Barbé et al., J. Am. Ceram. Soc. 1997, 80,
3157–3171.
Acknowledgment 23. M. Abkowitz, D. M. Pai, Philos Mag. B 1986,
53, 193–216.
Greg Smestad is gratefully acknowledged 24. B. Enright, G. Redmond, D. Fitzmaurice,
for proofreading this chapter. J. Phys. Chem. 1994, 97, 1426–1430.
25. J. Krüger, R. Plass, L. Cavey et al., Appl. Phys.
Lett. 2001, 79, 2085–2087.
References 26. U. Bach, Y. Tachibana, J.-E. Moser et al.,
J. Am. Chem. Soc. 1998, 121, 7445–7446.
1. B. O’Regan, M. Grätzel, Nature 1991, 353, 27. S. A. Haque, Y. Tachibana, D. R. Klug et al.,
737–739. J. Phys. Chem. B 1998, 102, 1745–1749.
2. M. K. Nazeeruddin et al., J. Am. Chem. Soc. 28. S. Y. Huang, G. Schlichthörl, A. J. Notzik
1993, 115, 6382–6390. et al., J. Phys. Chem. B 1997, 101, 2576–2582.
3. A. C. Arango, S. A. Carter, P. Brook, J. Appl. 29. A. Zaban, Y. Diamant, J. Phys. Chem. B 2000,
Phys. Chem. 1999, 74, 1698–1700. 104, 10043–10046.
4. N. C. Greenham, X. Peng, A. P. Alivisatos, 30. J. Krüger, U. Bach, M. Grätzel, Adv. Mater.
Phys. Rev. B 1996, 54, 17628. 2000, 12, 447–451.
5. M. Kocher, T. K. Däubler, E. Harth et al., 31. K. Tennakone, K. P. Hewaparakkrama,
Appl. Phys. Lett. 1998, 72, 650–652. M. Dewasurendra et al., Semicond. Sci.
6. U. Bach, D. Lupo, P. Compte et al., Nature Technol. 1988, 3, 382–387.
1998, 395, 583–585. 32. B. O’Regan, D. T. Schwartz, Semicond. Chem.
7. K. Murakoshi, R. Kogure, Y. Wada et al., Sol. Mater. 1998, 10, 1501–1509.
Energy Mater. Sol. cells 1998, 55, 113–125.
33. K. Tennakone, G. R. R. A. Kumara, A. R.
8. U. Bach, Solid-state dye-sensitized meso-
Kumarasinghe et al., Semicond. Sci. Technol.
porous TiO2 solar cells, 2000, Ph.D. thesis
1995, 10, 1689–1693.
N◦ 2187, EPFL Lausanne, Switzerland.
34. C. A. N. Fernando, Sol. Energy. Mater. 1993,
9. G. Hadziioannou, P. F. van Hutten in Semi-
30, 211–220.
conducting polymers, Wiley-VCH, Weinheim,
35. C. R. N. Fernando, Sol. Energy. Mater. 1992,
2000.
10. K. Müllen, G. Wegner in Electronic Materials: 28, 255–271.
The Oligomer Approach, Wiley-VCH, Wein- 36. J. He, H. Lindström, A. Hagfeldt et al.,
heim, 1998. J. Phys. Chem. B 1999, 103, 8940–8943.
11. C. D. Dimitrakopoulos, S. Purushothaman, 37. B. O’Regan, D. T. Schwartz, M. S. Za-
J. Kymissis et al., Science 1999, 283, 822–824. keerudin et al., Adv. Mater. 2000, 12,
12. B. Crone, A. Dodabalapur, Y.-Y. Lin et al., 1263–67.
Nature 2000, 403, 521–523. 38. K. Tennakone, G. R. A. A. Kumara, I. R. M.
13. D. Lupo, J. Salbeck 1999, US5885368. Kottegoda et al., J. Phys. D: Appl. Phys. 1998,
14. L. Hausslinger et al., 1997, US5683833. 31, 1492–1496.
15. K. Shiratsuchi, H. Takizawa 2000, US 39. A. K. Ghosh, T. Feng, J. Appl. Phys. 1978, 49,
6084176. 5982–5989.
16. P. M. Borsenberger, L. Pautmeier, H. 40. D. L. Morel, A. K. Ghosh, T. Feng et al., Appl.
Bässler, J. Chem. Phys. 1991, 94, 5447–5454. Phys. Lett. 1978, 32, 495–497.
17. M. N. Bussac, L. Zuppiroli, Phys. Rev. 1996, 41. G. A. Chamberlain, Sol. Cells 1983, 8, 47.
B54, 4674. 42. G. Horowitz, Adv. Mater. 1990, 2, 287.
18. J. Hagen, W. Schaffrath, P. Otschik et al., 43. D. Wöhrle, D. Meissner, Adv. Mater. 1991, 3,
Synth. Met. 1997, 89, 215–220. 129.
44. J. Simon, J. J. André, in Molecular Semicon- 56. J. J. Dittmer et al., Sol. Energy Mater. Sol. Cells
ductors, Springer-Verlag, Berlin, 1985. 2000, 61, 53–61.
45. C. W. Tang, Appl. Phys Lett. 1986, 48, 183. 57. A. Köhler, D. A. D. Santos, D. Beljonne et al.,
46. K. Kudo, T. Moriizumi, Jpn. J. Appl. Phys. Nature 1998, 392, 903–906.
1981, 20, L553–L556. 58. L. A. A. Pettersson, L. S. Roman, O. Inganäs,
47. M. Hiramoto, H. Fujiwara, M. Yokoyama, J. Appl. Phys. 1999, 86, 487–496.
J. Appl. Phys. 1992, 72, 3781–3787. 59. T. J. Savenije, J. M. Warman, A. Goossens,
48. K. Yamashita, Y. Harima, T. Matsubayashi, Chem. Phys. Lett. 1998, 287, 148–153.
J. Phys. Chem. 1989, 93, 5311–5315. 60. K. Kajihara, K. Tanaka, K. Hirao et al.,
49. M. Hiramoto, Y. Kishigami, M. Yokoyama, Jpn. J. Appl. Phys. 1997, 36, 5537–5542.
Chem. Lett. 1990, 119–122. 61. S. Spiekermann, S. Mayer, W. Blau et al.,
50. M. Hiramoto, M. Suezaki, M. Yokoyama, unpublished results.
Chem. Lett. 1990, 327–330. 62. S. A. Gamboa, H. Nguyen-Cong, P. Chartier
51. G. Hadziioannou, P. F. van Hutten, in Semi- et al., Sol. Energy Mater. Sol. Cells 1998, 55,
conducting Polymers, Wiley-VCH, Weinheim, 95–104.
2000. 63. P. Chartier, H. N. Cong, C. Sene, Sol. Energy
52. J. J. M. Halls, C. A. Walsh, N. C. Greenham Mater. Sol. Cells 1998, 52, 413–421.
et al., Nature 1995, 376, 498–500. 64. J. S. Salafsky, W. H. Lubbrthuizen, R. E. I.
53. M. Granström, K. Petrisch, A. C. Arias et al., Schropp, Chem. Phys. Lett. 1998, 290,
Nature 1998, 395, 257–260. 297–303.
54. L. S. Roman, W. Mammo, L. A. A. Pettersson 65. J. H. Schoen, C. Kloc, E. Bucher et al., Nature
et al., Adv. Mater. 1998, 10, 774–777. 2000, 403, 408–410.
55. G. Yu, J. Gao, J. C. Hummelen et al., Science
1995, 270, 1789.
497
6.1 1970s, and 1980s provided an excellent

Fundamentals of Photocatalysis basis for work on photocatalysis. Each of
the basic principles related to photoelectro-
Akira Fujishima and Donald A. Tryk chemistry is also involved in photocataly-
The University of Tokyo, Tokyo, Japan
sis. To a first approximation, the frame-
6.1.1 work of photocatalysis diverges from other
Introduction PEC effects in two aspects: (1) scale, that is,
micrometer or nanometer instead of mil-
Remarkable progress has been made in the limeter; and (2) in most cases, lack of an
area of photocatalysis in almost 25 years external electrical circuit. The latter situa-
since the earliest work appeared [1–3]. It tion is analogous to that of electrochemical
was recognized that TiO2 is a highly suit- corrosion, in which anodic and cathodic
able material for photocatalysis because electrochemical reactions short-circuit lo-
of its complete resistance to photocor- cally on a conductive metal surface. In the
rosion. This virtue of TiO2 had been case of photocatalysis, a PEC reaction, usu-
recognized only a few years earlier in the ally an oxidation reaction, is short-circuited
work on photoelectrochemical (PEC) water with a purely electrochemical reduction
splitting [4–7]. Other virtues include the reaction, for example, that of molecular
low cost and nontoxicity of TiO2 . These oxygen, as discussed later in detail in this
have allowed TiO2 to be used in an in- chapter.
creasing number of applications involving In a solar PEC cell, photogenerated
both indoor and outdoor air purification, electron-hole pairs are driven efficiently
water purification, self-cleaning surfaces, in opposite directions by an electric
and antifogging surfaces. This introduc- field existing at the interface between a
tory chapter focuses on the basic aspects semiconductor and an electrolyte solution.
of photocatalysis. The same processes can occur on a
single semiconductor particle, although
6.1.2 the electric field may be vanishingly
Background small [9, 10]. Thus, the carriers must
be separated by means of diffusion,
As outlined by Heller [8], the extensive with nanoscale sources and sinks. In a
work that was carried out in the field of system where semiconductor particles are
photoelectrochemistry during the 1960s, suspended in a liquid solution, excitation
498 6 Nonsolar Energy Applications
of the semiconductor can lead to redox acid solution or −1.2 V (SHE) in alka-
processes in the interfacial region around line solution. Among the oxide semicon-
each particle. These types of systems have ductors, TiO2 , SrTiO3 , CaTiO3 , KTaO3 ,
drawn the attention of a large number of Ta2 O5 , and ZrO2 satisfy this requirement,
investigators over the past twenty years. although in most cases the band gaps are
For example, in the area of photocatalytic relatively large (>3.0 eV), and thus the
water splitting, there has been a very active efficiency of solar energy conversion is
effort, as reviewed recently [11, 12]. very low.
In the early work on PEC solar cells Following early TiO2 -induced water
for water electrolysis, one of the first splitting studies, it was discovered that
types of electrode materials examined was TiO2 particularly excels in photocatalyti-
TiO2 , partly because it has a sufficiently cally breaking down organic compounds
positive valence band edge to oxidize water and other pollutants to innocuous prod-
to oxygen [4–7]. It is also an extremely ucts, neutralizing foul odors, and killing
stable material in the presence of aqueous bacteria. Frank and Bard in 1977 were
electrolyte solutions. The possibility of among the first to examine the possibili-
solar photoelectrolysis was demonstrated ties of using TiO2 to decompose pollutants;
for the first time with a system in which in their case, cyanide, in water [2, 3]. Since
then, there has been an increasing inter-
an immersed n-type TiO2 semiconductor
est in environmental applications. These
electrode, connected through an electrical
authors quite correctly pointed out the
load to an immersed platinum black
implications of their result in the field
counter electrode, was exposed to near-
of environmental purification. Their pre-
UV (ultraviolet) light. When the surface
diction has indeed been borne out, as
of the TiO2 electrode was irradiated with
evidenced by the extensive international
light consisting of wavelengths shorter
efforts in this area [8, 12–23].
than ∼415 nm, photocurrent flowed from
the platinum counter electrode to the
6.1.3
TiO2 electrode through the external circuit. TiO2 Characteristics
The direction of the current revealed that
the oxidation reaction (oxygen evolution) 6.1.3.1 General
occurs at the TiO2 electrode and the Titanium dioxide occurs in nature in
reduction reaction (hydrogen evolution) at three crystalline forms, rutile, anatase,
the Pt electrode. and brookite. The crystal structures are
When the conduction band energy, ECB , shown in Fig. 1 [24]. Some of the physical
is higher (i.e. closer to the vacuum level properties are given in Table 1. Much of
or more negative on the electrochemi- the fundamental work on TiO2 has been
cal scale) than the hydrogen evolution carried out with rutile single crystals be-
potential, photogenerated electrons can cause these are relatively straightforward
flow to the counter electrode and drive to produce. Anatase single crystals are
the reduction of protons, resulting in hy- more difficult to produce, and some work
drogen gas evolution without an applied has been carried out with natural crystals
potential. Hence, for this to occur, ECB but it is possible to grow such crystals
should be at least as negative as −0.4 V with the use of the vapor-phase trans-
[standard hydrogen electrode (SHE)] in port method [25]. Photocatalytic reactions
6.1 Fundamentals of Photocatalysis 499
Fig. 1 TiO2 crystal structures: c

(a) rutile; (b) anatase;
(c) brookite. Structure data were
taken from Wyckoff [24].
a b
(a)
a
b
(b)
a
(c) c
have been shown to exhibit larger rates on surface with gas or liquid environment.
anatase compared to rutile, and it will be The electrochemical flat band potentials,
very useful to carry out further comparative Efb , can be used to estimate the band edge
studies with single crystals. energies for titanium dioxide. Efb values
for rutile single crystals have been reported
6.1.3.2 Band Edge Energies by several authors [26]. In acid solution
The energies of the valence and conduction (pH 2.1), a value of −0.05 V versus SHE
bands of TiO2 are determined by the has been proposed by Morrison, based
chemical termination of the material, on the results of several workers [27]. In
together with the interactions of the alkaline solution, values of −0.6 [27] and
Tab. 1 Selected properties of TiO2 arises from the fact that the surface oxy-
gens of the oxide interact strongly with
Property Rutile Anatase protons and hydroxide ions [27]. To un-
derstand the processes underlying photo-
Crystal structure Tetragonal Tetragonal catalysis on TiO2 , it is useful to take the
Space group P42/mnm I41/amd
pH dependencies of the band edge ener-
(136) (141)
Density, g cm−3 4.250 3.894 gies into perspective. The relationships of
Band gap, eV 3.02 3.2 these energies to those corresponding to
Type Indirect Indirect various redox couples, for example, those
for various forms of reduced oxygen (see
Note: Data taken from Finklea [24] and next section) or for various organic ox-
Wyckoff [26] idations, then become apparent (Fig. 2).
Here we have adapted the commonly used
−0.76 V [28] have been given. Recently, Pourbaix representation in a modified
Kavan and coworkers have measured Efb form (i.e. with negative potentials upper-
values for both rutile and synthetically most), to be consistent with the prevailing
grown anatase single crystals in 1 M practice in energy band diagrams [30].
H2 SO4 and obtained values of +0.04 and Therefore, the ubiquitous lines for the
−0.16 V against SHE, respectively [25]. H+ /H2 and O2 /H2 O redox couples are
The authors point out that the latter inverted and slope upward with increasing
value is sufficiently negative so that no pH (more negative electrochemical poten-
external bias needs to be applied to tials). Likewise, the valence band (VB) and
photoelectrochemically evolve hydrogen. conduction band (CB) energies for TiO2
On the basis of the flat band potential, slope upward (higher electron energies)
the actual location of the conduction band with increasing pH.
edge energy, ECB , then depends on the The VB energies lie approximately
doping level, according to the following 1.8 eV below those necessary to evolve
equation: molecular oxygen from water over the en-
n tire pH range. Electrons in the CB for
ECB − EF = ξ = −kT ln (1)
NC rutile are very close in potential to that
where EF is the Fermi level energy, n necessary to evolve molecular hydrogen
is the occupation of conduction band from water but not at appreciable rates. In
states with electrons, and NC is the early TiO2 -based PEC solar cells, this prob-
density of conduction band states at the lem was circumvented through the use of
band edge [29]. For normal semiconductor a two-compartment cell. The TiO2 elec-
doping levels, ξ could be in the range of 0.1 trode was immersed in a more alkaline
to 0.2 V [27]. In subsequent discussions we half-cell and a Pt counter electrode was
will refer to the Efb values, which should immersed in a more acid electrolyte in the
then be considered to be lower (energy) second half-cell. The electrons generated
limits for the ECB values. at the photoelectrode diffuse at the back of
As already stated, the band positions de- the electrode and then toward the counter
pend on the environment external to the electrode. In this case, electrons photogen-
solid surface, for example, the hydrogen erated at the alkaline TiO2 electrode have
ion concentration or pH. This is a situation a relatively high electrochemical energy,
common to all oxide semiconductors and which is sufficient to evolve hydrogen in
Fig. 2 TiO2 band edge −1

energies. Data were taken from
Finklea [26] and from the data TiO2 CB H+/H2
given in Table 2. O2/O2−
0 O2/H2O2
O2/HO2−
[V vs. SHE]
Potential
+1 O2/H2O
O−/OH−
+2
OH/H2O
TiO2 VB
+3
0 2 4 6 8 10 12 14
pH
the more acidic solution. In the case of the topic of the actual energetics of various
photocatalysis, the same type of effect may oxidation reactions pertinent to photo-
also be operative: specifically, if the local catalysis is beginning to be explored in
pH is high, for example, at a reducing site detail [35]. For example, because of the
at a particular location on a TiO2 parti- presence of water, the actual reaction could
cle, additional reducing power is available be considered to be an electron transfer
for the reduction of molecular oxygen to followed by an extremely fast chemical
superoxide. reaction with water.
In contrast, the oxidizing power of
photogenerated holes decreases with in- R −−−→ • R+ + e− (2)
creasing pH but the standard potential for •
R+ + H2 O −−−→ • RH + OH− (3)
the one-electron oxidation of any given or-
ganic compound RH to its radical cation The effective redox potential of this re-
•
RH+ is not pH-dependent. For example, action would be shifted significantly in the
the standard potential for this reaction for negative direction because of consumption
n-octane in acetonitrile is +2.71 V against of the oxidized product. Thus, it is quite
Ag/Ag+ [34], which can be converted to possible that trapped holes with redox po-
+2.84 V against SHE. This value may not tentials of +1.5 V [36] to +1.75 [37], have
be sufficiently positive because solvent sufficient oxidizing power to induce the
reactions make elucidation of the reproduction of the neutral radical, which
versible potential problematic. At present, would involve Eqs. (2) and (3).
Tab. 2 Standard redox potentials for various electrochemical reactions involving molecular oxygen
and its reduced forms
Reaction pH E0 G pKa References

[V] [eV]
Acid dissociations
HO•2 ←−→ •
O 2 − + H+ 4.88 31
H2 O 2 ←−→ HO2 − + H+ 11.63 31
• −
•
OH ←−→ O + H+ 11.9 32
Redox reactions
O2 + e− ←−→ •
O2 − all −0.284 +0.284 33
O2 + H+ + e− ←−→ HO•2 0 +0.005 −0.005 c
O2 + H+ + e− ←−→ HO•2 7 −0.409 +0.409 c
O2 + 2H+ + 2e− ←−→ H2 O 2 0 +0.693 −1.386a c
O2 + 2H+ + 2e− ←−→ H2 O 2 7 +0.279 −0.558a c
O2 + H+ + 2e− ←−→ HO2 − 14 −0.065 +0.130a 33

O2 + H+ + 2e− ←−→ HO2 − 7 +0.142 −0.284a c
O2 + 4H+ + 4e− ←−→ 2H2 O 0 +1.229 −4.916a 33

O2 + 4H+ + 4e− ←−→ 2H2 O 7 +0.815 −3.260a c
O2 + 4H+ + 4e− ←−→ 2H2 O 14 +0.401 −1.603a c

•
O2 − + 2H+ + e− ←−→ H2 O 2 7 +0.842 −0.558a c
HO•2 + e− ←−→ HO2 − all +0.694 c
HO•2 + H+ + e− ←−→ H2 O 2 7 +0.967 −0.558a c
H2 O2 + H+ + e− ←−→ •
OH + H2 O 7 +0.433 −0.991a c
H2 O2 + 2H+ + 2e− ←−→ 2H2 O 7 +1.351 −3.260a c

•
OH + H+ + e− ←−→ H2 O 0 +2.683 −2.233b 32
•
OH + H+ + e− ←−→ H2 O 7 +2.269 −0.991b 32
•
OH + e− ←−→ OH− 14 +1.857 +0.252b 32
• −
O + 2H+ + e− ←−→ H2 O 7 +2.559 −0.701b c
• −
O + 2H+ + e− ←−→ H2 O 14 +1.733 +0.128b c
a The free energy is for the product.

b The free energy is for • OH or • O− .
c Calculated.
Note: Data were taken from Hoare [31], Buxton and coworkers [32], and Tarasevich and
coworkers [33].
Solely considering Eq. (2), it would be reactions are often higher at higher
expected that the reaction of a hole in pH [18]. For organic oxidation reactions
the TiO2 valence band with the organic in aqueous solution, not involving radi-
compound would become less favorable cal species, standard redox potentials are
at high pH. However, considering Eq. (3), known (Table 3) and many exhibit pH
a normal pH dependence, that is, about dependencies of about 60 mV pH−1 , al-
60 mV/decade at 25 ◦ C should exist. Over- though the behavior can be somewhat
all, it might be expected that there should more complex, with at least one change
not be significant pH effects in terms in slope [38] because of involvement of
of changes in excess oxidizing power of acid-base equilibria.
valence band holes or trapped holes. In The majority of the VB and CB values
fact, the rates of photocatalytic oxidation (the latter being approximated by Efb as
Tab. 3 Selected free energies and standard potentials for CO2 reduction reactions
Reaction pH !G0 E0 , V vs. SHE References

[kJ mol−1 ]
CO2 (g) + 2H+ + 2e− ←−→ CO(g) + H2 O 0 +20.06 −0.104 39

CO2 (g) + 2H+ + 2e− ←−→ HCOOH(aq) 0 +38.32 −0.199 38, 40
CO2 (g) + 4H+ + 4e− ←−→ HCOH(aq) + H2 O 0 +27.49 −0.071 38, 40
CO2 (g) + 6H+ + 6e− ←−→ CH3 OH(aq) + H2 O 0 −17.24 +0.030 38, 40
CO2 (g) + 8H+ + 8e− ←−→ CH4 (g) + 2H2 O 0 −130.62 +0.169 39
HCOOH(aq) + 4H+ + 4e− ←−→ CH3 OH(aq) + H2 O 0 −55.96 +0.145 40
HCHO(aq) + 2H+ + 2e− ←−→ CH3 OH(aq) 0 −44.77 +0.232 40
HCOOH(aq) + 2H+ + 2e− ←−→ HCHO(aq) + H2 O 0 −10.81 +0.056 40
•
CH3 OH + (CH3 CN) + e− ←−→ CH3 OH(CH3 CN) NA > −268.23a > +2.78b 41
•
CH3 CH2 OH + (CH3 CN) + e− ←−→ CH3 CH2 OH(CH3 CN) NA > −256.65a > +2.66b 41
2-propanol(CH3 CN) NA 41
•+
2-propanol (CH3 CN) + e− ←−→ > −246.04a > +2.55b
•
n-pentane + (CH3 CN) + e− ←−→ n-pentane(CH3 CN) NA > −281.74a > +2.92c 34
n-hexane(CH3 CN) NA 34
•+
n-hexane (CH3 CN) + e− ←−→ > −277.88a > +2.88c
•
n-heptane + (CH3 CN) + e− ←−→ n-heptane(CH3 CN) NA > −274.02a > +2.84c 34
n-octane(CH3 CN) NA 34
•+
n-octane (CH3 CN) + e− ←−→ > −274.02a > +2.84c
a Indicatesthat the actual "G0 values are more negative than those given.
b converted from values versus ferrocene/ferrocinium [135].
c converted from values versus Ag/Ag+ (0.1 M) [135].
6.1 Fundamentals of Photocatalysis
503
already discussed) shown in Fig. 2 are generally forms of reduced oxygen, in-
for rutile single crystals. It has already cluding superoxide • O2 − , the hydroperoxyl
been pointed out that the crystalline form radical • O2 H, hydrogen peroxide H2 O2 ,
more commonly used in photocatalysis, the peroxide anion HO2 − , the hydroxyl
anatase, has a higher CB energy (more neg- radical • OH, and the oxyl radical anion
• −
ative electrochemical potential). A point O . The relevant reactions involving these
is included for single crystal anatase, ap- species are shown in Table 2. The pH de-
proximately 200 mV above the line drawn pendences of these reactions are shown
through the points for rutile, and a point in the same figure with the TiO2 band
is also shown for a nanocrystalline anatase edge energies (Fig. 2). It is interesting
thin film: its CB energy is elevated approx- that the reducing power of the conduc-
imately 100 mV with respect to that for tion band electrons is barely sufficient to
rutile. reduce O2 to • O2 − at pH values below the
Lying approximately 0.6 to 0.8 eV below pKa for • O2 H (4.88). With increasing pH,
the CB energy, there are surface states however, there is an increasing excess of
on TiO2 ; three experimental points are in- reducing power for this reaction.
cluded in Fig. 2 at specific pH values but The pH dependence of the oxidizing
additional experimental evidence exists for power of the • OH radical parallels that for
such surface states under conditions with valence band holes, except that there is a
no pH control, for example, in the gas discontinuity at pH 11.9, which is the pK
phase, and therefore such data cannot be for the equilibrium
included in the figure. A reasonable char- •
OH ⇐'⇒ H+ + O− (5)
acterization for this type of surface state is
that it involves a trapped electron with an At pH 0, the potential for the one-
accompanying proton, viz.: electron reduction of • OH is +2.683 V [32]
(Table 2). As already mentioned, the redox
H potential for trapped holes has been
estimated to be significantly less positive,
O O +1.50 to +1.75 V against SHE.
Ti (IV) Ti (IV) Ti (IV) Ti (III) The standard potentials for the various
(4) redox processes, involving molecular oxy-
This type of species is important for pho- gen, can be conveniently visualized in
tocatalysis because it provides a surface a Frost diagram, given here for pH 7
sink for photogenerated electrons. Addi- (Fig. 3) [33, 43, 44]. In this type of dia-
tionally, this type of species is particularly gram, the relative state of reduction of the
important with respect to the wetting prop- oxygen is plotted along the horizontal axis,
erties of TiO2 , which is of importance to with O2 itself being taken as zero and H2 O
self-cleaning surfaces. being taken as four. The vertical axis is
the free energy of formation in units of
6.1.4 electron volts. The slopes of the lines be-
Photocatalytic Reactions tween species of differing oxidation states
are the negative of the potential for the cor-
6.1.4.1 Oxygen Electrochemistry responding redox process. The free energy
The active chemical species that are change for the process is the slope multi-
involved in photocatalytic reactions are plied by the number of electrons involved
Fig. 3 Frost diagram for O2 HO2 O2−

reduction at pH 7. Data were HO2−
taken from Table 2. 0.0 H2O2
O2 O−
Free energy
−1.0 OH
[eV]
−2.0
H2O
−3.0
0 1 2 3 4
State of reduction
electrons
in the oxidation state change. At any partic- follows:

ular oxidation state, there can be more than
one state of protonation depending on the hν −−−→ h+ + e− (7)
difference of pH 7 and the pKa value for h+ + RH −−−→ RH+ (8)
a particular equilibrium. The free energy
+ • +
change is obtained from the relationship RH −−−→ R + H (9)
h+ + H2 O −−−→ • OH + H+ (10)
"G 2.303RT
= (pH − pKa ) (6) RH + • OH −−−→ R• + H2 O (11)
F F
As shown here, the initial attack on
where "G/F gives the free energy change the organic compound can be carried out
directly in eV. This diagram shows that either by the hole (h+ ) or the hydroxyl
each of the species with oxidation states radical (• OH). The question of which
−1, −2 and −3 are unstable with respect species is more important has been a
to disproportionation. It shows why the subject of much debate. One reason for
hydroxyl radical is such a powerful oxidant: the uncertainty is that it is often difficult to
it is simply because it gains so much distinguish between products of reactions
stability by being reduced to water. It also with holes and products of reactions with
shows that superoxide is not a very strong hydroxyl radicals. Similar to other aspects
one-electron oxidant and that peroxide is of photocatalysis, the answer must be
a moderately strong two-electron oxidant. given in very specific terms. One must
The redox potentials are given in Table 2. specify the precise conditions under which
a particular result is obtained. For example,
6.1.4.2 one must consider the surface coverage of
Basic Reactions, Kinetics, and Active the organic compound when discussing
Species the active species. If the coverage is
high, direct attack by holes appears to be
The basic photocatalytic reactions have possible. For relatively low coverages, it
been reviewed by several authors. A appears that • OH is the active species.
generic set of reaction steps for a given Lawless and coworkers argued that
organic compound RH can be written as an adsorbed • OH radical is identical
to a trapped hole [36]. By reacting col- and hydroxylated products of organic com-
loidal TiO2 (13 nm) in aqueous suspen- pounds. The desorption of free • OH is
sion with radiolytically generated • OH, considered to be unlikely because of the ex-
they observed a broad absorption band tremely large rate constant for the reverse
(ca. 350 nm). The intensity of this signal reaction.
exhibited first-order kinetics, with a rate A later study by Goldstein and cowork-
that was proportional to the • OH concen- ers involved phenol as the reactant [45].
tration as a result of the self-reaction of The photocatalytic reaction products were
•
OH to produce hydrogen peroxide: compared with those of radiolytically gen-
erated • OH on one hand and SO4 − on the
•
•
OH + • OH −−−→ H2 O2 (12)
other. The latter is a strong one-electron
with a time constant of about 50 ms. oxidant and would mimic the effect of
In homogeneous solution, this reaction a hole (see section on mechanisms). At
is second-order (Table 4) but becomes low phenol concentrations, the oxidation
first-order on the two-dimensional sur- products of the photocatalytic reaction are
face, as discussed later. The trapped hole similar to those for • OH, whereas at high
can be represented as TiIV − O− . On phenol concentration, they are similar to
those for reaction with SO4 − , indicating
•
the basis of the reaction with SCN− ,
the E 0 for this species was estimated to involvement of holes. This study shows
be about 1.5 V (pH 3), which is approx- that the trapped hole acts as a • OH trans-
imately 1.3 eV above the valence band. fer agent. At high phenol concentrations,
These workers conclude that the exis- phenol adsorbs on the TiO2 surface and
tence of adsorbed • OH (surface-trapped competes for mobile holes, decreasing the
holes) can explain most of the observed rates of hole localization and electron-hole
results, including the production of H2 O2 recombination.
Tab. 4 Homogeneous reaction rate constants for selected reactions involving hydroxyl
radicals
Reactants Products Rate constant

[M−1 s−1 ]
HO•2 + • OH −−−→ H2 O + O2 1 × 1010

•
O2 − + • OH −−−→ OH− + O2 1 × 1010
•
OH + • OH −−−→ H2 O 2 4.2 × 109
CH3 CHO + • OH −−−→ H2 O + CH3 CO 3.6 × 109
CH3 COCH3 + • OH −−−→ Products 1.3 × 108
CH3 CH2 CH3 + • OH −−−→ Products 2.3 × 109
2-propanol + • OH −−−→ Products 1.6 × 109
1,4-C6 H4 (CO2 H− ) + • OH −−−→ Products 3.3 × 109
Coumarin + • OH −−−→ Products 2 × 109
C6 H5 OH + • OH −−−→ dihydroxycyclohexadienyl 6.6 × 109
DMPO + • OH −−−→ DMPO-OH 4.3 × 109
Note: Values were taken from A. B. Ross, NDRL/NIST Solution Kinetics Database, v. 2.0,
Standard Reference Data, NIST, Gaithersburg, MD 20899, USA (http://allen.rad.nd.edu),
also found in Ref. 32. DMPO stands for 5,5-dimethyl-1-pyrroline N-oxide.
Bahnemann and coworkers have dis- with neutral molecular oxygen O2 . The ef-
tinguished between shallow and deeply fect of raising the pH from 3 to 6 was
trapped holes [46]. They point out that the consistent with the second pathway be-
absorption features commonly used to in- cause both the • Q− and • O2 − would be
dicate trapped electrons (650 nm, 1.7 eV) stabilized at the higher pH.
and holes (430 nm, 2.7 eV), could not A different approach was taken by
be the actual ones involved with pho- Kesselman and coworkers, who examined
tocatalytic reactions because they would the oxidation of some organic anions
correspond to energy levels near the mid- (chloroacetates) at heavily Nb-doped TiO2
dle of the band gap, for example, 1.7 eV electrodes. Because the selectivity for
below the CB or 2.7 eV above the VB. reaction of these species with • OH in
The latter is assigned to an excitation homogeneous solution is known, as is
of a surface peroxide. The active oxidiz- the ordering of the electrochemical redox
ing species is thought to be a shallow potentials, it is possible to distinguish
trapped hole, which can equilibrate with between pathways. In addition, the extent
VB holes and would have a similar oxi- to which the oxidation current depends on
dizing power. These authors showed that the reactant concentration is a measure
strongly adsorbing species, such as cit- of the involvement of the direct-hole
rate and acetate, react with free holes, pathway. If there is no dependence on
whereas more weakly adsorbed species, concentration, this indicates that • OH
such as ethanol, could react with deeply is produced and then reacts with the
trapped holes (i.e. surface-bound • OH). organic species. These authors found a
They also showed that dichloroacetate and full range of behavior from 12% to 97%
thiocyanate adsorb strongly and react with direct-hole involvement for formate and
either freely or shallowly trapped holes but 4-chlorocatechol, respectively. The results
not deeply trapped holes. were found to vary significantly from one
Cermenati and coworkers used a very electrode to another (15% vs. 65% direct-
interesting approach to assess the rel- hole involvement for acetate).
ative importance of holes and hydroxyl From the work of Nosaka and coworkers
radicals [47]. For a model reactant, they with electron paramagnetic resonance
employed quinoline, which has two aro- (EPR) spectroscopy, the proportion of the
matic rings with differing electron density, direct hole and • OH-mediated pathways
the nitrogen-containing ring with less. can be determined for various types of
Because • OH is electrophilic, the attack TiO2 samples. These workers used the
should occur on the benzene ring. On photocatalytic oxidation of acetate as a
probe reaction and found mixtures of EPR-
the basis of a detailed analysis of prod-
active products. They have proposed the
ucts as a function of time, the major
following schemes:
mechanistic pathways involving • OH were
discounted. Instead, a mechanism involv-
CH3 COOH + H+ −−−→
ing oxidation with holes, together with
superoxide, was proposed (see section on •
CH3 + CO2 + H+ (13)
mechanisms). A third mechanism was •
CH3 COOH + OH −−−→
proposed in which initial oxidation by ei-
•
ther holes or • OH is followed by reaction CH3 + CO2 + H2 O (14)
H2 O + h+ −−−→• OH + H+ Ti4+ O− Ti4+ OH− + H2 O −−−→

•
(15)
CH3 COOH + • OH −−−→ Ti4+ O2 − Ti4+ OH2 + • OH (17)
•
CH2 COOH + H2 O (16)
takes place more easily than the reaction
The methyl radical • CH3 can be pro-
duced by reaction with either h+ or Ti4+ O2 − Ti4+ O− + H2 O −−−→
•
•
OH, whereas the carboxymethyl radical
•
CH2 COOH can be produced only by re- Ti4+ O2 − Ti4+ OH− + • OH (18)
action with • OH. Thus the ratio of • CH3 to
•
CH2 COOH radical yields the proportion The structural arrangements of these
of the direct h+ path against the • OH path. species are shown in Fig. 4.
It was found that the • OH path was favored From work in our laboratory, it was
for TiO2 samples with higher coverages of found that the yields of fluorescent prod-
OH groups. On such surfaces, the type of ucts from the reactions of • OH with
active surface species has been proposed coumarin and 1,4-phthalate could be mea-
by Howe and Grätzel [48], based on EPR sured as a function of time, with a linear
measurements, to be dependence being found for fluorescence
intensity versus illumination time [51]. It
Ti4+ O− Ti4+ OH−
•
(S1) was concluded that the • OH pathway
predominates for the low reactant concen-
On TiO2 surfaces that contain fewer trations used (10−4 to 10−3 M) in aqueous
OH groups, perhaps as a result of heat
solution.
treatment, the active species may instead
In later work, it was found that the
be [49, 50]:
quantum yield for the reaction of • OH
Ti4+ O2 − Ti4+ O−
•
(S2) with 1,4-phthalate was relatively small (7 ×
10−5 ) under the experimental conditions
Hirakawa and coworkers propose that used (4 × 10−4 M, 2 × 10−3 M NaOH; UV
the reaction intensity, 1 mW cm−2 ) [52]. In contrast,
A B C D E
•
O
OH OH OH −O
•
O O O •
O O +
O O O
Ti Ti Ti Ti Ti Ti Ti Ti Ti
O O O −
O O O O O
O O• O O O O O O O
Ti +Ti Ti Ti Ti Ti Ti Ti
O O O O O O O O− O
O O O O O O O •
O
Ti Ti Ti Ti Ti Ti Ti Ti Ti
O O O O O O O O
O O O O O O O O O
Ti Ti Ti Ti Ti Ti Ti Ti
Fig. 4 TiO2 surface species, from Nakaoka and coworkers [45].

the quantum yield for the photocatalytic diffusing species and could then be con-
oxidation of iodide I− to iodine I2 was verted to • OH on reaching the organic film,
found to be relatively high (5.7 × 10−2 ) as in the following reactions:
for the conditions used (10−2 M KI; UV
intensity, 1 mW cm−2 ). The quantum 2HO•2 −−−→ H2 O2 + O2 (19)
yields usually observed for photocatalytic hν
reactions at similar light intensities are H2 O2 −−−→ 2• OH (20)
more in line with the latter, and therefore
it was concluded that the hole is the active In experiments with chemically gener-
species in most cases. ated • OH (Fenton reagent), similar effects
As already mentioned, the desorption were observed in water contact angle for
of active oxygen species, particularly • OH, PS and ODS.
from the TiO2 surface, with subsequent In earlier work, it was found that
reaction in homogeneous aqueous solu- Escherichia coli cells were killed in aqueous
tion, has been downplayed as a significant suspensions in contact with illuminated
pathway for TiO2 photocatalysis, although TiO2 films, even when separated by a 50-
not by direct experiments. Nevertheless, µm-thick porous membrane [55]. On the
our group has found it possible to observe basis of the inhibitory effect of catalase,
the oxidative degradation of organic films, it was concluded that the active species
for example, methylene blue, that are sepa- is hydrogen peroxide but there appeared
rated from the illuminated TiO2 surface by to be a cooperative effect, possibly due to
air gaps of up to 125 µm [53]. In an effort to superoxide, which significantly enhanced
identify the active species, various types of the lethal effect.
organic reactant films were examined [54]. Now that many of the important photo-
Changes in mass were also monitored generated species have been identified, it
by use of the quartz-crystal-microbalance is useful to consider the kinetics of forma-
(QCM) technique. In addition to methy- tion and decay of some of these species.
lene blue, a polyaniline derivative, poly(2- The absorption of a photon and creation
methoxyaniline-5-sulfonic acid) (PMAS), of an electron-hole pair in a TiO2 parti-
polystyrene (PS), octadecyltriethoxysilane cle occurs in a time range that has, thus
(ODS)-coated glass, and polyvinyl alcohol far, been too short to measure, that is,
were examined. With the ODS-coated glass a few femtoseconds. These charge carri-
separated from the illuminated TiO2 film ers are initially considered to exist in the
by 50 µm, oxygen-containing functional conduction band and valence band, re-
groups were produced, as observed with spectively but are then quickly trapped.
X-ray photoelectron spectroscopy, creating Skinner and coworkers have determined
a hydrophilic surface. For PMAS, PS, and the trapping time for electrons to approx-
ODS, the QCM results showed decreases imately 180 fs based on measurements
in frequency, that is, increases in mass due on TiO2 sols in acetonitrile, which was
to oxygenation. The fact that aliphatic hy- used to avoid the absorbance of hydrated
drocarbons are oxidized suggests that the electrons [56]. The nanoparticle surface
active oxidizing agent may not be HO•2 , consists of many different types of local ge-
H2 O2 , or 1 O2 . Thus the most likely can- ometries and corresponding trap energies,
didate is • OH. However, it may also be leading to a broad range for the absorbance
possible that HO•2 or H2 O2 could be the spectrum.
Holes have been estimated to be trapped greatly increase the lifetime of the trapped
within 1–10 ps by Serpone and coworkers, electrons [58].
and these workers find that 90–100% of all It had already been observed in early
electron-hole (e-h) pairs have undergone work with flash photolysis, albeit on
recombination much longer timescales, that recombi-
nation could be greatly inhibited in the
e− + h+ −−−→ heat (21) presence of either hole scavengers or
electron scavengers [67]. Lifetimes of sec-
within 10 ns for colloidal TiO2 particles onds or even hours have been reported
from 2–27 nm in diameter [57]. However, for trapped electrons in the presence of
Colombo and coworkers estimate the half-
hole scavengers such as 2-propanol [68].
life of an e-h pair to be approximately
Lifetimes of milliseconds (in acid) or mi-
30 ps [58]. This is just one example of
croseconds (in base) have been reported for
the wide variety of kinetic results that
trapped holes in the presence of an electron
have been reported (Table 5). Second-order
scavenger such as methylviologen [67].
kinetics are observed particularly at short
The decay process is usually assumed
times and high light intensities, that is,
to involve recombination of electrons and
when there are multiple e-h pairs existing
holes in various states of trapping or de-
simultaneously in a single particle [57, 59].
trapping. Because at short time intervals
Results for time-resolved microwave
and relatively high light intensities, there
conductivity measurements, which reflect
may be approximately equal numbers
the number of charge carriers, indicate
that the decay kinetics span many or- of each type of carrier in a single
ders of magnitude in time, for example, particle, the recombination process can
from ns to s [60, 61, 63]. Striking differ- be expected to be second-order. When
ences were observed, comparing anatase there is on the order of only a single e-
and rutile, with the latter exhibiting h pair in a particle, the decay process is
much faster decay kinetics [60]. Major dif- known to become first-order. This has
ferences were observed among various been explained in terms of stochastic
types of colloidal TiO2 preparations and processes [59, 69–71], that is, because of
in comparison to Degussa P25 in terms the discrete nature of the species and
of recombination kinetics [63]. However, corresponding events, they must be treated
Colombo and coworkers, using diffuse statistically rather than continuously. The
reflectance flash photolysis experiments, simple fact that reactions are taking
have found the behavior of TiO2 colloidal place on a two-dimensional surface is
solutions versus powder form, rutile ver- also important [72]. However, it is well
sus anatase, and nanoparticle versus P25, known in the field of solid-state electronics
all to be similar [58, 65]. These workers that direct recombination in a doped
point out that the diffuse reflectivity re- semiconductor behaves as a first-order
sults are specific for trapped electrons, process for low injection rates, simply
rather than for carriers in general and because the magnitude of the squared
assign the long-lived species observed term in the rate expression becomes
in this and other studies to such elec- small [73]. This is the analog of pseudo-
trons [65]. They also show that the effect of first-order kinetics. In the case of indirect
a hole scavenger, such as thiocyanate, is to recombination (i.e. mediated by a band gap
Tab. 5 Kinetics of photogenerated charge carrier trapping and decay
Reaction TiO2 Particle Medium pH Gas e-h/ Techniquea Half-lifeb Comments Year Author Refe-
type size, nm particle rences
e− -trapping Sol – aqueous 2.7 Inert 67 LFP <30 ps 1985 Rothenberger 59

h+ -trapping Sol – aqueous 2.7 Inert 67 LFP ∼250 ns 1985 Rothenberger 59
e− -decay Sol – aqueous 2.7 Inert 67 LFP 200–500 ps 1985 Rothenberger 59
e− -decay Sol – aqueous 2.7 Inert 6 LFP 20–50 ns 1985 Rothenberger 59
e− -decay Sol – aqueous 2.7 Inert 0.85 LFP ∼100 ns 1985 Rothenberger 59
e− -decay Anatase dry – – – TRMC 1 µs to 10 s 1990 Schindler 60
powder
e− -decay Rutile dry – – – TRMC 1–5 µs 1990 Schindler 60
powder
e− -decay P25 29 nm dry – – – TRMC 1 µs–10 s 1990 Schindler 60
powder
•
OH decay Sol 13 nm aqueous Acid Inert – PR 20 ns •
OH self- 1991 Lawless 36
reaction
e− -decay P25 29 nm dry – – 0.8 PR-TRMC 100 ns 1991 Warman 61
e− -decay P25 29 nm 2-propanol – – 0.8 PR-TRMC 25 µs 1991 Warman 61
e, h-trapping P25 29 nm aqueous 2.5 O2 – LFP 0.25–1 ns 1993 Lepore 62
(124 nm)
e, h-decay P25 10 ns aqueous 2.5 O2 – LFP ∼10 ns 1993 Lepore 62
e− -decay P25 29 nm paste – – – TRMC 100 ns to trans-decalin 1994 Martin 63
10 ms
e− -decay S13 – paste – – – TRMC >60 ms trans-decalin 1994 Martin 63
e− -decay P25 29 nm dry – – – TRMC 230 ns to Various 1994 Martin 64
1.6 ms adsorbates
e− -decay Q-size 2–4 nm dry – – – TRMC 1.6 µs to Various 1994 Martin 64
11 ms adsorbates
(continued overleaf )
6.1 Fundamentals of Photocatalysis
511
512
Tab. 5 (continued)
Reaction TiO2 Particle Medium pH Gas e-h/ Techniquea Half-lifeb Comments Year Author Refe-
type size, nm particle rences
e− -decay Q-size – aqueous – – – TRDRS ∼5 ps 1995 Colombo 65

e− -decay Q-size – dry – – – TRDRS ∼20 ps 1995 Colombo 65
6 Nonsolar Energy Applications
e− -decay Rutile – aqueous – – – TRDRS ∼20 ps 1995 Colombo 65

e− -decay Anatase – aqueous – – – TRDRS ∼5 ps 1995 Colombo 65
e− -decay P25 – aqueous – – – TRDRS ∼10 ps 1995 Colombo 65
e− -trapping Q-size 2 nm aqueous 2.7 – – TRDRS ∼500 fs 1995 Colombo 66
e− -decay Q-size 2 nm aqueous 2.7 – 17.3 TRDRS ∼10 ps 1995 Colombo 66
e− -decay Q-size 2 nm aqueous 2.7 – 2.9 TRDRS ∼20 ps 1995 Colombo 66
e, h-trapping Sol 2.1–26.7 nm aqueous – Air 0.2–1504 LFP 1–10 ps 1995 Serpone 57
e− -decay Sol 2.1 nm aqueous – Air 0.2 LFP 483 ps 1995 Serpone 57
e− -decay Sol 13.3 nm aqueous – Air 157 LFP 120 ps 1995 Serpone 57
e− -decay Sol 26.7 nm aqueous – Air 1504 LFP 38 ps 1995 Serpone 57
e-trapping Sol 2 nm CH3 CN – – – LFP 180 fs 1995 Skinner 56
e− -decay Sol – aqueous 4.5 – – TRDRS ∼3 ps 1996 Colombo 58
e− -decay P25 29 nm dry – – – TRDRS ∼10 ps 1996 Colombo 58
e− -decay P25 29 nm wet – – – TRDRS ∼7 ps 1996 Colombo 58
e− -decay P25 29 nm aqueous – – – TRDRS ∼7 ps 1996 Colombo 58
e− -decay P25 29 nm aqueous – – – TRDRS >100 ps adsorbed 1996 Colombo 58
SCN−
e− -trapping Sol 2.4 nm aqueous 2.3–2.6 Air <1 LFP <20 ns 1996 Bahnemann 46
e− -decay Sol 2.4 nm aqueous 2.3–2.6 Air <1 LFP ∼200 ns 1996 Bahnemann 46
h+ -decay Sol 2.4 nm aqueous 2.3–2.6 Air <1 LFP ∼500 ns 1996 Bahnemann 46
a LFP = laser flash photolysis; TRMC = time-resolved microwave conductivity; PR = pulse radiolysis; time-resolved diffuse reflectance spectroscopy.
b Values in parentheses are times for which the intensities decrease by one-half.
recombination level), first-order kinetics exemplified, in the presence of appropriate

may also be observed. scavengers, there is no corresponding
Although e-h recombination is certainly pathway for trapped electrons to self-react.
important, particularly for large numbers
of e-h pairs per particle, an additional decay 6.1.4.3 Quantum Yield
channel that must not be overlooked is the
hole-hole process: 6.1.4.3.1 General There is an important
+ + + need to assess the rates of photocatalytic
h + h −−−→ (h )2 (22)
reactions in terms of various other pa-
or the related process with adsorbed • OH: rameters, such as the amount of catalyst,
the catalyst surface area, the number
•
OH + • OH −−−→ H2 O2 (12) of active catalytic sites, and the illumi-
nation intensity. This is based on the
These processes could also, in principle, desire to compare results for various
be either second-order or first-order. For substrates, catalysts, photocatalytic reac-
example, as already mentioned, Lawless tors, light sources, and other experimental
and coworkers generated • OH radicals ra- conditions [76–79]. Progress in the pho-
diolytically and allowed them to react with tocatalytic field is facilitated by a clear set
colloidal TiO2 particles. There were no ex- of standard definitions and methodologies,
cess electrons present in this case, and and several ambiguities need to be clarified
the decay of the absorbance at 350 nm or resolved in this area.
was first-order, assigned to Eq. (12). The For example, the quantum yield (QY or
existence of peroxide-like species on illu- $) is a fundamental parameter in hetero-
minated TiO2 surfaces has been reported geneous photocatalysis, whose definition
in cases in which molecular oxygen was should be carefully considered. A standard
absent [74, 75]. As already mentioned, definition has been given in terms of a
Bahnemann and coworkers have pointed particular reaction, with a defined reactant
out that the optical absorbance signal at or product, at a given wavelength λ [78]:
430 nm (2.7 eV) could not correspond to
a trapped hole because it was too deep moles of reactant or product
$λ =
within the band gap and have instead as- moles of photons (einsteins)
signed this absorbance to a peroxide-like absorbed
species [46]. In an EPR study by Howe and (24)
Grätzel of illuminated TiO2 particles in This definition raises a number of is-
vacuo at 4.2 K, Eq. (12) was proposed to sues. For example, commonly used light
occur on warming to 77 K, followed by sources are rarely monochromatic. How-
ever, it is also possible to use an averaged
H2 O2 + h+ −−−→ • O2 − + 2H+ (23) value for a given wavelength range or it is
also possible to use a close approximation
to explain the existence of • O2 − in the of a monochromatic source. A second is-
absence of O2 . We should also note that sue is that the measurement of the number
these decay channels, that is, hole-hole of photons absorbed by the photocatalyst
or hydroxyl-hydroxyl, help to explain the is a nontrivial task, as has been discussed
fact that trapped holes, in general, are for the cases of liquid suspensions [76, 79]
not as long-lived as trapped electrons. As and immobilized films in gas phase [77].
For photocatalytic reactions in aqueous 6.1.4.3.2 Factors Determining Quantum

suspensions, it has been proposed that a Yield Several groups have found that $
standard reactant (phenol) and standard values tend to approach a maximum at
photocatalyst (Degussa P25) be used to low light intensities and/or high reactant
calibrate the results from a particular labo- concentrations. Early work by Egerton and
ratory or a particular experimental arrange- King, with TiO2 suspensions in liquid 2-
ment in terms of quantum yield [78]. This propanol, found $ values between 0.5 and
proposal appears to have some merit, in 1.0 for light intensities of 2 × 1013 to 5 ×
that it is not difficult to implement, requir- 1015 quanta s−1 for 2-propanol oxidation
ing no specialized equipment. Specifically, to acetone [89]. Above this range, a square-
one would measure the effective quantum root dependence of rate on intensity
yield for any arbitrary reactant and photo- was observed. Kormann and coworkers
catalyst, normalized to that for phenol and measured a limiting value of 0.56 for the
P25 TiO2 . oxidation of chloroform in aqueous TiO2
For gas-phase photocatalytic reactions, suspension at a light intensity of 2.8 ×
we would like to propose that isopropanol 10−6 einstein L−1 min−1 [90]. Lepore and
be used as a standard reactant because it coworkers found an experimental value of
approximately 0.6 for gas-phase propanol
degrades relatively rapidly to acetone in
oxidation, with an extrapolated value of
a pathway that involves only the hydroxyl
1.0 for high concentration [62]. In EPR
radical, with a much slower decomposition
spin-trapping experiments, in transparent
of acetone [80, 81]:
aqueous TiO2 sols, Grela and coworkers
measured a limiting value of 0.54 for high
H3 C−CHOH−CH3 + • OH −−−→ spin-trap concentrations [71].
H3 C−• COH−CH3 + H2 O (25) In gas-phase experiments with TiO2
films on glass, we have found that
under specific conditions (i.e. high 2-
Before getting into the details of the propanol coverage and low light intensity
oxidation reaction for 2-propanol, we will (36 nW cm−2 to 45 uW cm−2 ), the quan-
first mention the use of TiO2 films on tum yield $ tends to approach a maximum
various solid substrates, which are now value that is intrinsic to a particular TiO2
being used in many gas-phase applica- film [81]. Even films made from the same
tions. Initially, to avoid the difficulties of starting materials can have small differ-
using TiO2 powder for water purification, ences in this maximum value. It was
which entails a separation step, various proposed that this value depends on such
researchers began to work on ways of im- factors as the levels of impurities in the
mobilizing TiO2 particles, for example, in bulk of the TiO2 particles and the degree
thin-film form. One of the first reports on of crystallinity [12], which would control
the preparation of TiO2 films was that of the rate of bulk recombination. Another
Matthews [82]. Such films have also been idea has been advanced that suggests that
developed by Anderson [77], by Heller [8], the limiting $ value at low intensity is due
and by our group [83–88]. Our group has to particle-size effects, with lower $ values
been developing ways to put photocatalytic exhibited by small particles [69].
TiO2 coatings on various types of support Furthermore, we found that the $
materials, for example, ceramic tiles. values are constant for a particular
ratio of absorbed photons to adsorbed of the following: either e-h recombination

molecules [81]: (Eq. 21), hole-hole annihilation (Eq. 22), or
the hydroxyl radical self-reaction (Eq. 12).
Jhν
Inorm = (26) Assuming that • OH is the active species,
&2−p
one can write an equation for the branch-
where Inorm is termed the normalized light ing of the light flux into these two types of
intensity, Jhν is the light flux in number of processes, based on the assumption that
moles per unit apparent area per unit time a steady state surface concentration of hy-
(mol cm−2 s−1 ), and &2−p is the number of droxyl radicals or trapped holes is reached.
moles of adsorbed 2-propanol per unit ap-
parent area per unit time (mol cm−2 s−1 ). Jhν = k25 &2−p &• OH + k12 &•2OH (27)
For any given initial value for &2−p , we k25 &2−p &• OH
propose that the shape of the curve (Fig. 5) $=
Jhν
is governed by the competition between re-
actions of either trapped holes or hydroxyl k25 &2−p &• OH
= (28)
radicals with isopropanol (Eq. 25) and one k25 &2−p &• OH + k12 &•2OH
0.5
Simulation
0.4 QY(1000 ppm)
QY(100 ppm)
QY(10 ppm)
0.3 QY(1 ppm)
Φ
0.2
0.1
0
10−7 10−6 10−5 10−4 10−3 10−2 10−1 100
I norm
(a) [s−1]
Simulation
QY(1000 ppm)
1 QY(100 ppm)
QY(10 ppm)
QY(1 ppm)
Φ
0.1
Fig. 5 Quantum yield as a 0.01

10−7 10−6 10−5 10−4 10−3 10−2 10−1 100
function of normalized light
intensity: 2-propanol, taken I norm
from Ohko and coworkers [81]. (b) [s−1]
This equation can be rearranged to the that the ratio of absorbed photons to ad-
following form: sorbed molecules is the same, 1 : 3000. We
find that $ is the same in both cases, 0.14,
2 or half the limiting value of 0.28. Hence,
$= ! (29)
4Jhν k12 each • OH is approximately as likely to react
1+ 1+ 2 2 with a 2-propanol molecule, as it is to react
k25 &2−p
with another • OH. Thus, the $ value is
determined only by the probabilities that
It can be seen that this function
the • OH radicals will either react with a
approaches unity at low light intensity and
2-propanol molecule or with each other.
zero at high intensity. This type of curve
However, this simple model requires
was used to fit the data in Fig. 5. Of course,
further modification because the $ values
it is also possible that such a relationship
tend to deviate from the theoretical
could also result from a scheme, based
curve at higher light intensities; this
on e-h recombination instead of the • OH
becomes apparent when log $ is plotted
self-reaction. It is highly intriguing that
(Fig. 5b), which indicates a mixing of first
this type of behavior could be exhibited
and second-order kinetics. Second-order
over such a wide range of initial reactant behavior is approached for the lowest 2-
surface concentrations (three orders of propanol concentration but deviates for
magnitude) and light intensities (seven higher concentrations. For the latter,
orders of magnitude). trapped holes would become increasingly
We have proposed a simple model to favored as the active species. Additionally,
rationalize the fact that the quantum yield the adsorbed molecules may block sites at
versus light intensity behavior tends to re- which holes can self-react, and therefore
main constant for a given ratio of adsorbed another decay channel might become
photons to adsorbed molecules. Basically, important, for example, trapping at a
this model involves the assumption that different type of site at which the h-h
the diffusion of • OH on the surface is fast reaction is favored or e-h recombination.
enough, that it is not a limiting factor, In any case, the results show clearly that 2-
and every • OH generated has an oppor- propanol decomposition can be used as
tunity to react with all of the adsorbed a type of standard for quantum yield,
2-propanol molecules in its vicinity. Let us if high reactant concentrations and low
imagine for a moment that the reactions light intensities are used to estimate the
are taking place on a perfectly flat, non- maximum, limiting value. However, it is
porous surface with a certain number of still necessary to estimate the amount of
adsorbed 2-propanol molecules, say 3000. absorbed light, for example, using the
For a low initial gas-phase 2-propanol con- methodology proposed by Anderson and
centration, this number of molecules will coworkers [77].
occupy a certain surface area Alow . For a The idea of • OH radicals self-reacting,
high initial concentration, the molecules has also been proposed for the decompo-
will occupy a smaller surface area Ahigh , sition of chloroform at illuminated TiO2
simply due to the adsorption equilibrium. aqueous suspension by Kormann and
For each surface area, we now assume coworkers [18, 90]. These authors have
that a certain number of photons is ab- given a limiting form of Eq. (29) at high
sorbed, for example one every second, so light intensity:
krds &chloro such as CO2 and H2 O. In some cases,

$= √ (30)
k12 Jhν it is possible to monitor specific inter-
krds mediate compounds, and this approach
kobs = √ = 2.0 × 10−3 (M s)−1/2 can be quite powerful in elucidating the
k12
mechanistic pathways, as discussed in the
(31) following sections.
where kobs is the observed rate constant,
krds is the rate constant for the reaction 6.1.4.4.2 Active Species The first set
between a surface-bound • OH radical and of reactions that follow those already
a chloroform molecule, corresponding to given, involve either molecular oxygen or
our k25 , and k12 is the rate constant for superoxide radical anion • O2 − :
the • OH self-reaction. From the value of
kobs , they conclude that the rds is slow R• + O2 −−−→ ROO• (32)
compared to Eq. (10). This conclusion
In many cases, this radical can then
is consistent with that of Ohko and react with superoxide, which has been ex-
coworkers for 2-propanol oxidation on perimentally demonstrated to be produced
TiO2 films [81]. on illuminated-TiO2 [91–93]:
It is an important challenge to resolve
this issue of whether e-h recombination or ROO• + • O2 − −−−→ ROOOO• (33)
another reaction, such as 12 or 25, is more
important in determining $ at higher light The latter species is termed a tetroxide
intensities. The answer could help to guide anion. It can become protonated to the
•
further efforts to improve $ values. ROOOOH form, a tetroxide radical, which
We continue to emphasize that low light can take on a cyclic structure, depending
intensities are not only interesting from a on the nature of the reactant. These
will be discussed further in the section
fundamental standpoint but are also highly
on hydrocarbons and related compounds.
important for indoor environmental appli-
We shall now proceed to describe the
cations, in which intense UV light sources
reactions that have been proposed for
cannot be used [12, 20]. For these appli-
various specific types of reactants.
cations, the most important point is that
the photocatalytic reaction can be quite ef-
6.1.4.4.3 Aldehydes For aldehydes, it has
ficient, even at UV light intensities that
been known for quite some time that
are compatible with safe human exposure radical initiators can induce chain-type ox-
levels. idation in the presence of oxygen [94–97].
6.1.4.4 Reaction Mechanisms •
OH + RH −−−→R• + H2 O
6.1.4.4.1 General The reaction steps al- (initiation) (11)

ready given are just the beginning of R• + O2 −−−→ROO•
a complex mix of reactions that can be
(propogation) (34)
proposed to explain the experimental pho-
tocatalysis results. These results often ROO• + RH −−−→ROOH + R• (35)
involve the complete mineralization of the • •
ROO + ROO −−−→nonradical products
starting organic compound, that is, com-
plete conversion to inorganic compounds (termination) (36)
where we have inserted • OH as the radical values exceeding unity, as high as 1.4, for
initiator. the photocatalytic dechlorination of carbon
We were able to observe clear evidence tetrachloride [100].
for the chain-type mechanism in experi-
ments, involving acetaldehyde decomposi- 6.1.4.4.4 Hydrocarbons and Related Com-
tion in the gas-phase [98], similar to those pounds Reaction pathways involving su-
already discussed for 2-propanol. With peroxide have been treated in detail
acetaldehyde, the $ values exceeded the by Schwitzgebel and coworkers [101], as
maximum value obtained for a similar film shown in Fig. 7. These reactions can oper-
for 2-propanol oxidation (0.28) (Fig. 6). As ate on alkanes and their oxidation products
already discussed, the latter value may be so that with the exception of aldehydes, as
considered to be an intrinsic maximum $ noted in the earlier section, any of the
value for this particular film. Therefore, if main types of CH and C−H−O com-
$ exceeds the intrinsic maximum value, pounds (i.e. alkanes, alcohols, ketones,
it indicates that radical chain reactions are and carboxylic acids) can be completely
important, that is, a single photon can decomposed in this fashion to CO2 and
cause more than one photodecomposition H2 O. These reactions involve cyclic tetrox-
reaction. ide intermediates, similar to those orig-
Other groups have also obtained rela- inally proposed by Russell [102]. It is
tively high $ values for chain-type re- quite interesting that TiO2 is so well
actions. As already mentioned, Lepore suited to the decomposition of organic
and coworkers found values, approach- compounds, with a combined attack of ox-
ing unity for propanol oxidation [62]. idatively and reductively produced radical
Raupp and Junio found values exceed- species.
ing unity, as large as 3.0, for the ox- In Fig. 7(b), a possible three-dimensional
idation of acetone and methyl-t-butyl structure for the tetroxide intermediate for
ether [99]. Stark and Rabani measured the photocatalytic oxidation of n-octane
1000
Quantum yield (total)
100
10
[%]
0.1
10−6 10−5 10−4 10−3 10−2 10−1 100 101 102
I normal
[s−1]
Fig. 6 Quantum yield as a function of normalized light intensity:
acetaldehyde, taken from Ohko and coworkers [98].
n-octane: RCH(OOOOH)R′ RCOR′ + O2 + H2O

3-octanol: RCOH(OOOOH)R′ RCOR′ + O2 + H2O2
3-octanone: RCH2COCH(OOOOH)R′ RCHO + R′CHO + CO2 + H2O
n-octanoic acid: RCH2COOH + h+ ( •OH) RCH2• + CO2 + H+(H2O)
RCH2• + O2 RCH2 OO•
RCH2 OO• + •OOH
RCH2OOOOH
RCH2OOOOH + RCH2COOH RCH2OH + RCHO + O2 + CO2
(a)
(b)
Fig. 7 (a) Photocatalytic reactions involving tetroxides, taken from Schwitzgebel and
coworkers [101]. (b) Proposed structure for the tetroxide intermediate in the oxidation of
octane, with the hydrogen bond shown as a solid rod. Carbon atoms are gray, hydrogen atoms
are a light stipple, oxygen atoms are a dark stipple, and oxygen lone pairs are white.
is shown. Very similar intermediates can facilitate the breakage of the chain, with
be drawn for the corresponding alcohol the formation of two aldehydes and a CO2
and ketone. For the alcohol (3-octanol), molecule.
the cyclic tetroxide structure can form The pathways delineated in the preced-
at C-2, similar to the structure shown ing paragraph are probably also important
for n-octane but the intermediate given in the decomposition of long-chain hy-
in the equation in Fig. 7(a) actually cor- drocarbons, carboxylic acids, fats, oils,
responds to the cyclic structure, form- and even long-chain polymers [103–105].
ing at C-3. In that case, the alcohol In the latter case, we have direct in-
−OH group becomes part of the ring. frared evidence that the hydrocarbon
For 3-octanone, the terminal oxygen of chains can be attacked at any point along
the tetroxide may hydrogen-bond across their length, with C−C bond scission,
to the hydrogens on C-4, which could presumably with subsequent degradation
0.15 Fig. 8 Photocatalytic

0h decomposition of octadecane:
IR spectra, taken from Minabe
and coworkers [105].
2h
0.10 4h
Absorbance
0.05
End of reaction
0.00
3000 2950 2900 2850 2800
Wavenumber
[cm−1]
3.0 2.0
Weight of octadecane
2.5
CO2 concentration
[10 µg cm−2 ]
1.5
[10 µmol cm−2]
2.0
1.5 1.0
1.0
0.5
0.5
0 0
2.0 100
Carbon concentration
Carbon proportion
80
[µmol C cm−2 ]
1.5
60
[%]
1.0
40
5.0
20
0 0
0 20 40 60 80 100
Fig. 9Photocatalytic decomposition of octadecane: weight loss
and CO2 generation, taken from Minabe and coworkers [105].
via the resulting aldehydes (Fig. 8). The 6.1.4.5 Kinetic and Mass Transport
aldehydes may be unstable enough to Considerations
account for the fact that they are not
detected by IR on the TiO2 film. Also, 6.1.4.5.1 Adsorption For moderate-in-
it is important to note that no gas-phase tensity illumination, over a wide range
products, other than CO2 , are detected in of experimental conditions, photocatalytic
the decomposition of octadecane, stearic decomposition follows first-order kinetics
acid, and glycerol trioleate, demonstrating in the adsorbed concentration of the
the safety of the photocatalytic approach organic compound and one-half-order in
to self-cleaning technology. For exam- light intensity due to recombination.
ple, a 36-µm-thick film of octadecane
was completely converted to CO2 after rate = koxidation &reactant I 1/2 (37)
approximately 50 hours of illumination at
0.8 mW cm−2 (Fig. 9). Where koxidation is a first-order rate
constant, &reactant is concentration per
unit real surface area, and I is the light
6.1.4.4.5 Aromatics As already briefly intensity. As pointed out by Emeline and
mentioned, the photocatalytic oxidation coworkers, however, this equation is a
of phenol was studied by Goldstein special case of a more general one [106].
and coworkers [45]. On the basis of the
analysis of the possible reactions (Fig. 10), n
rate = koxidation &reactant Im (38)
these authors were able to conclude
that either • OH or h+ can be the where the n and m values can vary with I
principal active species. Also, the use of and &reactant , respectively.
quinoline as a probe molecule was already Many organic compounds and noxious
mentioned. The full scheme developed by gases, such as H2 S, follow Langmuir ad-
Cermenati and coworkers is shown in sorption behavior so that Eq. (37) can
Fig. 11 [47]. This scheme shows that the be converted into the familiar Lang-
combined action of holes and superoxide muir–Hinshelwood form by substituting
is important. the adsorption equilibrium expression for
R
R R
•OH Ox
+ •
a d
H OH
OH
H+ H2O
c
R R
SO4• − + +
•
Side chain fragmentation
b e
Fig. 10 Photocatalytic decomposition of phenol: scheme, based on

Goldstein and coworkers [45].
&reactant : the gas-phase decomposition of formalde-

−1 −1 hyde and acetaldehyde, which have been
&reactant = &reactant,0 + (Keq creactant )−1
(39) implicated in the ‘‘sick-building’’ syn-
where &reactant,0 is the surface concentra- drome [107]. These two compounds were
tion at full coverage, Keq is the adsorption found to obey Langmuir adsorption and
equilibrium constant, and creactant is the Langmuir–Hinshelwood kinetics were fol-
gas-phase or liquid-phase concentration. lowed (Fig. 12). It was also found that
An example of this type of behavior formaldehyde adsorbs relatively strongly
is provided by our recent work with on the TiO2 surface, showing that TiO2
OH H HO H OH
OH° O2 − HO2°
•
N N N N
H OO°
OH OH HO OO° OH
OH° O2 − HO2°
•
N N N N
HO H HO H OH
+ OH° − OH°
− H2O − H2O
O° O O
O2 − OH°
N N N
(a) H OO° O
O transfer N
H
O
O
TiO2 CHO H2O CHO
Q,O2 Q+°,O2−°
(h+,e) CHO CHO
N N CH CHO NH2
O2 (2-formyl)phenyliminoethonal
O O°
CHO
O −CO2
O
N NH − CHO
(b) O
Fig. 11 Photocatalytic decomposition of quinoline: scheme, based on [47].

O O°
CHO
+
Q+° + O2
+ O2
N H2O NH2
− H+ O − HO2°
O
− CHO − CHO
N OH
O +O2
O2 − HCOO− CHO
Q + OH ° OH
N H NH − CHO
O
+ O2 − HO2°
N O
(c) H
Fig. 11 (Continued)
0.2
0.15
[µmol min−1]
20
r
0.1
[µmol−1 min]
15
1/r
10
0.05 5
0
0 0.1 0.2 0.3 0.4 0.5
1/Ceq
0 [µM−1]
0 5 10 15 20 25 30 35 40
Ceq
[µM]
Fig. 12 Photocatalytic decomposition of acetaldehyde:
Langmuir–Hinshelwood plot, based on Noguchi and
coworkers [107].
can be effective even at low formaldehyde Many groups have obtained results that
concentrations. The adsorption character- are consistent with Langmuir–Hinshel-
istics of TiO2 in this case are similar, if not wood kinetics. However, Emeline and co-
better than those of activated carbon. workers have emphasized the well-known
difficulties in distinguishing between out electrochemical measurements on a

various mechanisms purely based on the particulate suspension or slurry. If the
observed kinetic rate laws [106]. suspension is illuminated, then one has
a PEC slurry cell, a technique introduced
6.1.4.5.2 Mass Transport Under condi- by Bard and coworkers [109–111] and later
tions of high light intensity and low extended by other groups [112–114], in-
reactant concentration, it is possible to cluding our group [93]. For example, a
reach a point at which the mass trans- TiO2 suspension in aqueous electrolyte is
port of the organic compound is no longer stirred and current is measured at a po-
sufficient to maintain the adsorption equi- tentiostated platinum electrode. With illu-
librium. Such diffusion-controlled condi- minated TiO2 in a deaerated solution with
tions have been explored in further work no hole scavenger, the results show a net
with 2-propanol as a model reactant [108]. anodic current due to oxidation of trapped
These results show that it is possible electrons. As soon as the illumination is
to clearly distinguish the experimental stopped, the anodic current decays over a
conditions under which various types of period of up to 250 ms [112]. The negative
kinetics are operative (e.g. light-limited charge build-up on the illuminated parti-
conditions, a middle range of light inten- cle is due to the need for a pathway by
sity in which either Eqs. (37) or (38) are
which photogenerated holes can be con-
valid, and a high light intensity range, in
sumed and the product desorbed. In the
which mass-transport limitations become
work of Peterson and coworkers and that
important). These types of conditions can
of our group, this was assumed to be the
be seen in a plot of light intensity versus
hydroxyl radical itself (Fig. 14). However,
reactant concentration (Fig. 13). Such plots
based on our results with photocatalytic
must also take into account the Langmuir
effects at a distance from an illuminated
isotherm characteristics of the particular
TiO2 film, it appears to be more likely that
reactant of interest on the specific type of
TiO2 being considered. the hydroxyl radicals self-react (Eq. 12) and
hydrogen peroxide desorbs [53–55]. With
6.1.4.6 Special Techniques the addition of hole scavengers, such as
In addition to the techniques that have formate, acetate or methanol, the anodic
already been mentioned, with which to in- photocurrent is enhanced greatly and de-
vestigate photocatalytic reactions, includ- cays much more slowly when illumination
ing laser flash photolysis and EPR, there is stopped [112–114].
are several others that can yield very use- In related work, we have found that
ful information. These include the slurry illuminated suspensions of TiO2 particles
electrode technique, the use of biased in silicone oil exhibit a tendency to form
electrodes or photoelectrodes, the scan- bridges between biased electrodes that
ning electrochemical microscopic (SECM) varies, depending on the water content of
technique, and the Langmuir–Blodgett the TiO2 [115, 116]. For relatively dry types
(LB) film technique. These will be dis- of TiO2 , such as P25 (ca. 1%), the bridge-
cussed briefly. forming tendency is larger because the
photogenerated charges can be polarized
6.1.4.6.1 Slurry Electrodes The slurry at opposite ends of the particles. With TiO2
electrode technique enables one to carry that contains greater amounts of water,
106
1018
Mass-transport
105 controlled region 1017
104
1016
UV-light intensity
[quanta • cm−2 • s−1]

Absorbed photons
103 Curve A
[µW • cm−2]
1015
102 Target for outdoor
air purification 1014
101
Target for 1013
1 indoor air purification
1012
Curve B
10−1
1011
Light-limited
10−2 controlled region
1010
10−3 10−2 10−1 1 101 102 103

Reactant concentration
[ppmv]
Fig. 13 Master plot for determining operating light intensity as
a function of reactant concentration, based on Ohko and
coworkers [64].
Fig. 14 Reactions taking place between O2

an illuminated TiO2 particle in aqueous +
solution, based on Cai and
H2O
coworkers [93].
Catalase
e− H2O2
there is a tendency for the particles to
develop excess charge when they make e−
contact with the electrode, so that they SOD
e− e− O +
move under the influence of the electric e− 2
field and circulate electrophoretically back e− e−
− − O2
and forth between the electrodes. This is e− e e e−
e−
because the photogenerated carriers can be e− hν
effectively trapped via reactions involving TiO2
h+
hydroxide ions or water. Because of the
absence of bulk water surrounding each
OH−
particle, excess charge cannot be carried OH−
away, for example, by the production of
solution-phase peroxide.
6.1.4.6.2 Biased Electrodes It is possible W

to significantly enhance the oxidation + 0.6 V vs. SCE
of organic compounds, such as 4- magnitude of the current, is limited by

chlorophenol, at a nanoparticulate pho- the rate of the cathodic process, that is,
toelectrode by applying a positive bias O2 reduction (Fig. 15). In this case, the
potential [117, 118]. This result provides magnitudes of both currents at open cir-
convincing evidence that recombination is cuit were much lower than the maximum
important because it seems clear that the anodic current, which is due to photo-
effect of the electric field is to enhance the assisted O2 evolution and is limited only
separation of electrons and holes. by the rate of photogeneration of holes.
It was mentioned at the beginning of It was found that the deposition of ∼2 Å
this chapter that the photocatalytic process of Pt was effective in increasing the cur-
is closely related to the corrosion of metals, rent density for O2 reduction, such that
in that both anodic and cathodic processes the magnitudes of the anodic and cathodic
are occurring on the same surface. Indeed, reactions at open circuit were then lim-
for an electrically isolated material, the ited by the anodic reaction. These results
anodic and cathodic currents must exactly imply that photocatalytic reaction rates on
match to maintain charge neutrality. One illuminated TiO2 could be significantly ac-
can measure the separate anodic and celerated by the use of catalytic metals, as
cathodic current-potential curves for an discussed later. One point in which the
appropriate electrode to understand the current-matching experiments appear to
overall behavior. There is a potential at deviate from the situation with practical
which the anodic and cathodic currents photocatalytic phenomena is that the an-
are equal in magnitude and this potential odic process in the former case is found to
is termed the ‘‘mixed potential.’’ The be almost purely O2 evolution. This cannot
basic aspects of metal corrosion have been be true, in general, for TiO2 photocatalysis
treated by Vetter [119] and Sato [120]. because the quantum yields would then be
A similar approach has been developed very low.
for the photocatalytic process by Kessel- It is interesting to note that, in
man and coworkers [121]. These work- Vetter’s treatment of corrosion, there are
ers, using a single-crystal rutile rotating- three different special cases: (1) chemically
disk electrode, have shown that the and physically homogeneous surfaces;
mixed potential, and more importantly the (2) chemically inhomogeneous surfaces;
4
(a)
2
[µA cm−2]
0
J
−2
−4 (b)
−1000 −800 −600 −400 −200 0 Fig. 15 Current–potential

curves for an illuminated TiO2
Potential photoelectrode, taken from
[mV vs. SCE] Kesselman and coworkers [121].
and (3) chemically homogeneous, physi- the type of situation that could be expected
cally inhomogeneous surfaces. An exam- for a pure polycrystalline TiO2 sample.
ple of case 2 in a metal would be a Zn-Cu A direct comparison of the anodic and
alloy, in which Zn undergoes corrosion cathodic reactions on the various single
but Cu does not. A corresponding exam- crystal faces of either anatase or rutile does
ple in photocatalysis would be TiO2 , with not yet appear to have been studied.
deposited Pt, in which the anodic process
might occur predominantly on TiO2 and 6.1.4.6.3 Scanning Electrochemical Micro-
O2 reduction might occur predominantly scopy Another way in which it can be con-
on Pt. An example of case 3 in terms of cor- firmed that oxidative and reductive pho-
rosion would be any polycrystalline metal, tocatalytic reactions simultaneously occur
in which different crystallographic faces on TiO2 particles is to set up a simple
are exposed to the electrolyte, such that heterogeneous model electrode system.
the relative proportions of the anodic and This can be used to simulate model indi-
cathodic processes are different, although vidual particles and can be compatible with
both reactions occur on both types of sur- microelectrode detection of reaction prod-
faces. In terms of photocatalysis, this is ucts [122–126]. Such a system, shown in
Low viscosity
Epoxy resin Mercury epoxy resin
Cu wire
Pyrex glass tube (o.d. = 2 mm)
7 µm Carbon fiber
ca. 20 µm
(a)
Potentiostat
Pt SCE
(CE) (RE)
hν Micromanipulator
Carbon microelectrode (WE)
Pd
TiO2 Microscope
(b)
Fig. 16 Schematic diagram of a scanning electrochemical microelectrode
system, taken from Fujishima and coworkers [12].
Fig. 16, involves a film that is half-covered an increase in O2 concentration near the
with TiO2 and half-covered with metallic TiO2 surface and a decrease in the O2
regions, for example, Pd or ITO (In-Sn concentration near the TiO2 surface un-
oxide). A scanning microelectrode can be der illumination (Fig. 17) [124]. Oxygen is
positioned close to the surface, as close as produced via water oxidation on TiO2 and
50 µm. Its potential can be set at values at is then consumed at the metallic electrode
which, for example, either O2 or H2 O2 can (ITO or Pd) via reduction. Principally,
be monitored as a function of time. reduction to peroxide occurs although re-
The oxidation and reduction reactions duction to superoxide or to water may also
can be monitored separately to obtain in- contribute on an actual TiO2 microparticle.
formation regarding the mechanism of the There can also be varying contributions by
photocatalytic reaction. In photocatalytic these three reduction reactions, depending
reactions with metal-deposited TiO2 par- on whether there is a deposited catalyst.
ticles and films, the deposited metal acts As already discussed, some photocatalytic
as a reduction site, thereby increasing the reaction pathways can involve • O2 − ; this
efficiency of photogenerated charge sepa- is a desirable product, and metallic cata-
ration. Thus, a TiO2 -ITO composite film lysts may in fact be counterproductive, if
can act as a simple model for the metal- they promote the two- or four-electron O2
deposited photocatalyst. A carbon micro- reduction reactions.
electrode is employed to detect the reaction In the presence of ethanol, although
products. By positioning this electrode dissolved oxygen was consumed at both
close to either the TiO2 or the ITO surface TiO2 and Pd sites, the consumption of
(50–500 µm), the oxidation and reduction oxygen was larger at the TiO2 surface [122].
reactions, respectively, can be monitored A possible scheme that is consistent with
on these regions. This method was used to O2 consumption at TiO2 is:
detect hydrogen peroxide at the two types •
OH + CH3 CH2 OH −−−→
of regions, separately, by use of a ‘‘wired’’
horseradish peroxidase microsensor fabri- CH3 C• HOH + H2 O (40)
cated on the carbon microelectrode sur- •
CH3 C HOH + O2 −−−→
face [122]. Later, it was found that peroxide
could also be detected adequately by setting CH3 CH(OH)OO• (41)
the microelectrode potential at a very pos-
producing an organoperoxyl radical, which
itive potential (+1.0 V versus SCE) [124].
can either participate in a chain-type
Additionally, the carbon microelectrode is
process
able to detect O2 by setting the potential at
a negative potential (−1.0 V versus SCE). CH3 CH(OH)OO• + CH3 CH2 OH −−−→
In one of the early studies, it was found
CH3 CH(OH)OOH + CH3 C• HOH (42)
that H2 O2 is produced largely at the ITO
portion of the illuminated TiO2 -ITO com- CH3 CH(OH)OOH −−−→
posite surface due to O2 reduction [122].
CH3 CHO + H2 O2 (43)
In a subsequent study on the separate
detection of dissolved oxygen in aque- or can react further with • O2 − or HO•2
ous solution at reducing regions (in this to produce a tetroxide intermediate, as al-
case, Pd) and oxidizing regions (TiO2 ), ready discussed. The importance of radical
using a carbon microelectrode, we found intermediates was found to be great, as
UV irradiation UV irradiation
−500
−500
Current
−400
Current
[pA]
[pA]
−400
−300
−300
0 1 2 3 4 0 1 2 3
Time Time
(a) [min] [min]
−300 −300
Current
−200 −200
Current
[pA]
[pA]
−100 −100
0 0
0 1 2 3 4 5 0 1 2 3 4 5
Time Time
(b) [min] [min]
+100 +100
Current
Current
[pA]
[pA]
+50 +50
0 0
0 1 2 3 4 0 1 2 3
Time Time
(c) [min] [min]
Above TiO2 Above Pd
Fig. 17 Microelectrode current versus time curves, taken from Ikeda and
coworkers [124].
evidenced by the strong decrease in O2 is faster for the photodecomposition of

consumption in the presence of phenol, a acetaldehyde compared to ethanol because
radical scavenger. of the strong contribution of radical chain
It was also found that the consumption reactions for the former [125]. In the
of O2 over the illuminated TiO2 portion presence of superoxide dismutase, the
Without SOD
Ethanol With SOD
(a)
Acetaldehyde
0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0

(b) The amount of O2 consumption
[10−9 mol cm−2 ]
Fig. 18 The effect of SOD on the rate of oxygen consumption for
ethanol versus acetaldehyde [125].
O2 consumption decreased significantly results obeyed first-order kinetics, which

in the case of ethanol but negligibly in is reasonable because of the extremely
the case of acetaldehyde (Fig. 18). Thus limited amount of material.
it appears that the tetroxide pathway
may be important for ethanol. However, 6.1.4.7 New Photocatalysts
aldehydes can decompose via radical-
initiated chain reactions, consuming O2 6.1.4.7.1 General The results presented
but not • O2 − [95–97]. thus far certainly demonstrate the superior
characteristics of TiO2 as a photocatalyst.
6.1.4.6.4 LB Techniques Fundamental However, there are various characteristics
information concerning the photocatalytic of TiO2 that could, in principle, be im-
decomposition of stearic acid LB films proved. For example, various researchers
on both polycrystalline anatase films and have probed the idea of improving the
single crystal rutile can be obtained charge separation through catalysis of elec-
with the AFM technique [127, 128]. Such tron transfer to electron acceptors, such
studies have shown that, for less than as O2 , by means of deposited metal par-
full monolayer coverage, the photocatalytic ticles [22]. TiO2 has a drawback, in that
decomposition process takes place at the proportion of ambient light as ei-
random locations on the film. In the case ther sunlight or even indoor light that
of full monolayers, the film structure is can be utilized to induce photocatalytic
disturbed very quickly, with molecules reactions is still quite small. The value
aggregating into relatively tall ridges. It for full sunlight, which has a total power
appears that there are randomly located of approximately 100 mW cm−2 , is just
active sites, which can become passivated about 1% or 1 mW cm−2 . For indoor light,
by reaction products. New sites can also the values are much lower, on the or-
form. In contrast to results obtained with der of 1 µW cm−2 under a study lamp
thicker layers (100 nm), for which zero- or 0.1 µw cm−2 on the walls or floor of
order kinetics were obtained, the LB film a room [20]. In addition, for photocatalytic
air or water treatment, it is necessary to use In addition, metal ions can enhance
relatively intense UV light sources. It may oxidation rates by acting as electron
be possible to greatly increase the light- acceptors in themselves. Litter points out
harvesting ability of TiO2 through band that metal ions can be either beneficial
gap engineering. or detrimental [22]. Examples are Cu2+ ,
Fe3+ , and Ag+ , all of which can have
6.1.4.7.2 Metal-modified Photocatalysts beneficial effects in certain concentration
The area of metal-modified TiO2 ranges and detrimental effects at higher
photocatalysts is an extensive one but some concentrations. Metal ions can be present
aspects have been reviewed recently [22]. in water that is being photocatalytically
We will only cite a few examples here. treated and the effects must be taken into
We have already introduced two of account. Finally, Litter points out that the
the main themes: (1) improvement of addition of metal ions is generally too
electron transfer; and (2) improvement of expensive in large-scale applications.
light harvesting. One extremely important In terms of the modification of the light
application, the photocatalytic deposition absorption properties of TiO2 , there have
of metals, has also been treated in depth been several notable efforts. Anpo and
by Litter. coworkers have used the ion implantation
In the area of photocatalytic water split- technique extensively to extend the absorp-
ting, the use of deposited noble metals tion spectrum into the visible range [130].
has been studied for many years [11]. With the use of implanted Cr, the spectrum
The objective in this case is to enhance was shifted by as much as about 80 nm,
the transfer of photogenerated electrons with significant absorption at 450 nm. One
from the TiO2 conduction band to water challenge is to try to devise methods of
to produce hydrogen gas. This approach
preparing such photocatalysts with less
does indeed lead to improvements in H2
technology-intensive approaches.
production due to the well-known elec-
Zang and coworkers have prepared
trocatalytic activity of platinum for this
Pt(IV) chloride-modified amorphous TiO2
reaction. In the area of photocatalytic de-
that is effective at visible wavelengths [131].
composition of organic compounds, as
These authors have shown that the Pt is
already discussed, it is often desirable to
produce superoxide from O2 to assist in present in a complex ion form, still sur-
the formation of tetroxides, for example, rounded by chloride. This constitutes a
Ref. 101. The energetic requirement for highly effective use of the precious metal.
the production of superoxide is substan- Photocatalytic activity was found at wave-
tially less than that for hydrogen, so that in lengths as high as 546 nm. Hamerski and
principle, an electrocatalyst is not needed. coworkers have reported the improvement
In fact, as already mentioned, the use of a of photocatalytic activity for oil decomposi-
metallic catalyst may be counterproductive tion under sunlight with calcium-modified
if it leads to a multielectron product such as TiO2 [132].
H2 O2 or H2 O. In terms of simply increas- Murata and coworkers have thermally
ing the electron transfer rate for oxygen re- prepared rutile films on alloys of Ti with
duction a metallic catalyst, such as Pt or Pd, various 4d and 5d transition metals [133].
could be beneficial as shown by Gerischer, With the Ti-Ta and Ti-W alloys, visible
Heller, and coworkers [113, 129]. light activity was found. These results are
promising and should be followed up with occurring simultaneously on small TiO2

efforts to produce practical materials. particles. A fourth might be a thorough
understanding of the electronic effects of
6.1.4.7.3 Mixed Oxides In the area of the doping of TiO2 with other elements.
modification of TiO2 with other oxides, There are certainly a host of other topics
one of the rationales is to increase the worth intensive study.
surface acidity and thereby modify the Even so, the number of applications has
adsorption and catalytic properties. Fu advanced, proving that complete under-
and coworkers have reviewed the previous standing is never a necessity in developing
literature in this area and have prepared new technology. However, we hope that
TiO2 /SiO2 and TiO2 /ZrO2 materials via there will continue to be a healthy synergy
sol-gel techniques [134]. Improvements by between fundamental and applied work in
a factor of three were observed for the photocatalysis.
gas-phase oxidation of ethylene.
An important area of activity in terms References
of modification of TiO2 with other oxides
is that of adsorption enhancement. This 1. J. H. Carey, J. Lawrence, H. M. Tosine, Bull.
area has been pursued by the Bard Environ. Contam. Toxicol. 1976, 16, 697.
2. S. N. Frank, A. J. Bard, J. Am. Chem. Soc.
group [135, 136] and by the Yoneyama
1977, 99, 303–304.
group [137, 138]. The rationale is to use 3. S. N. Frank, A. J. Bard, J. Phys. Chem. 1977,
a metal oxide, with optimized adsorption 81, 1484–1488.
properties for a particular target pollutant, 4. A. Fujishima, K. Honda, S. Kikuchi, Kogyo
to capture the pollutant more effectively Kagaku Zasshi 1969, 72, 108–113.
than would be possible on TiO2 by itself. If 5. A. Fujishima, K. Honda, Nature 1972, 238,
37–38.
a microheterogeneous material is used, the
6. A. Fujishima, K. Kobayakawa, K. Honda,
pollutant can be captured and delivered to J. Electrochem. Soc. 1975, 122, 1487–1489.
active TiO2 sites over distances that are as 7. A. Fujishima, K. Kobayakawa, K. Honda,
short as possible. Both groups reported Bull. Chem. Soc. Jpn. 1975, 48, 1041–1042.
improvements in the performance of 8. A. Heller, Acc. Chem. Res. 1995, 28,
photocatalyst systems using this approach. 503–508.
9. B. O’Regan, J. Moser, M. Anderson et al.,
J. Phys. Chem. 1990, 94, 9820–8726.
6.1.5 10. L. Kavan, T. Stoto, M. Grätzel et al., J. Phys.
Conclusion Chem. 1993, 97, 9493–9498.
11. D. A. Tryk, A. Fujishima, K. Honda, Elec-
Although a great deal of progress has been trochim. Acta 2000, 45, 2363–2376.
12. A. Fujishima, T. N. Rao, D. A. Tryk, J. Pho-
made in the fundamental aspects of photo-
tochem. Photobiol. C, Photochem. Rev. 2000,
catalysis in the last 25 years, there are still 1, 1–21.
many areas that are incompletely under- 13. E. Pelizzetti, N. Serpone, Eds., Homoge-
stood. One example is the understanding neous and Heterogeneous Photocatalysis, D.
of the nature of the photocatalytically ac- Reidel Publishing Company, Dordrecht,
tive sites on TiO2 . Another is a detailed The Netherlands, 1986.
14. N. Serpone, E. Pelizzetti, Eds., Photocatal-
understanding of the factors that deter-
ysis – Fundamentals and Applications, John
mine quantum yield. A third might be Wiley & Sons, New York, 1989.
the understanding of the nanoscale details 15. E. Pelizzetti, M. Schiavello, Eds., Photo-
of the oxidation and reduction reactions, chemical Conversion and Storage of Solar
Energy, Kluwer Academic Publishers, Dor- 34. A. Taghizadeh, M. F. Lawrence, L. Miller

drecht, The Netherlands, 1991. et al., J. Photochem. Photobiol. A: Chem.
16. M. A. Fox, M. T. Dulay, Chem. Rev. 1993, 2000, 130, 145–156.
93, 341–357. 35. F. R. Keene, in Electrochemical and Electro-
17. D. F. Ollis, H. Al-Ekabi, Eds., Photocatalytic catalytic Reactions of Carbon Dioxide, (Eds.:
Purification and Treatment of Water and B. P. Sullivan, K. Krist, H. E. Guard), Else-
Air, Elsevier, Amsterdam, The Netherlands, vier, Amsterdam, The Netherlands, 1993,
1993. pp. 1–18.
18. M. R. Hoffmann, S. T. Martin, W. Choi 36. G. V. Buxton, C. L. Greenstock, W. P. Hel-
et al., Chem. Rev. 1995, 95, 69–96. man et al., J. Phys. Chem. Ref. Data 1988,
19. A. L. Linsebigler, G. Lu, J. J. T. Yates, Chem. 17, 513–535.
Rev. 1995, 95, 735–758. 37. A. A. Frost, J. Am. Chem. Soc. 1951, 73,
20. A. Fujishima, K. Hashimoto, T. Watanabe, 2680–2682.
in TiO2 Photocatalysis: Fundamentals and 38. M. R. Tarasevich, A. Sadkowski, E. Yeager,
Applications, BKC, Inc., Tokyo, 1999. in Comprehensive Treatise of Electro-
21. J. -M. Herrmann, Catal. Today 1999, 53, chemistry, Kinetics and Mechanisms of
115–129. Electrode Processes, (Eds.: B. E. Conway,
22. M. I. Litter, Appl. Catal. B: Environ. 1999, J. O. M. Bockris, E. Yeager et al.), Plenum
23, 89–114. Press, New York, 1983, pp. 301–398, Vol. 7.
23. D. F. Ollis, C. R. Acad. Sci. Paris, Serie IIC, 39. D. F. Shriver, P. W. Atkins, C. H. Langford,
Chim. 2000, 3, 405–411. in Inorganic Chemistry, Oxford University
24. R. W. G. Wyckoff, in Crystal Structures, In- Press, Oxford, UK, 1995, pp. 300–304.
terscience Publishers, New York, 1963, 40. S. Goldstein, G. Czapski, J. Rabani, J. Phys.
Chem. 1994, 98, 6586–6591.
p. 205.
41. D. W. Bahnemann, M. Hilgendorff,
25. L. Kavan, M. Grätzel, S. E. Gilbert et al.,
R. Memming, J. Phys. Chem. B 1997, 101,
J. Am. Chem. Soc. 1996, 118, 6716–6723.
4265–4275.
26. H. O. Finklea, in Semiconductor Electrodes,
42. L. Cermenati, P. Pichat, C. Guillard et al.,
(Eds.: H. O. Finklea), Elsevier, Amsterdam,
J. Phys. Chem. B 1997, 101, 2650–2658.
1988, pp. 43–145, Vol. 55.
43. R. F. Howe, M. Grätzel, J. Phys. Chem. 1987,
27. S. R. Morrison, in Electrochemistry at Semi-
91, 3906–3909.
conductor and Oxidized Metal Electrodes, 44. O. I. Micic, Y. Zhang, K. R. Cromack et al.,
Plenum Press, New York, 1980, p. 183. J. Phys. Chem. 1993, 97, 7277–7283.
28. A. J. Nozik, in Annual Review of Physical 45. Y. Nakaoka, Y. Nosaka, J. Photochem. Photo-
Chemistry, (Eds.: B. S. Rabinovitch, J. M. biol. A: Chem. 1997, 110, 299–305.
Schurr, H. L. Strauss), 1978, pp. 189–222. 46. K. Ishibashi, Y. Nosaka, K. Hashimoto et al.,
29. W. Jaegermann, in Modern Aspects of Electro- J. Phys. Chem. B 1998, 102, 2117–2120.
chemistry, (Eds.: R. E. White, B. E. Conway, 47. K. Ishibashi, A. Fujishima, T. Watanabe
J. O. M. Bockris), Plenum Press, New York, et al., J. Photochem. Photobiol A: Chem. 2000,
1996, pp. 1–185, Vol. 30. 134, 139–142.
30. M. R. Prairie, B. M. Stange, L. R. Evans, in 48. T. Tatsuma, S. Tachibana, T. Miwa et al.,
Photocatalytic Purification and Treatment of J. Phys. Chem. B 1999, 103, 8033–8035 .
Water and Air, (Eds.: D. F. Ollis, H. Al- 49. T. Tatsuma, S. Tachibana, A. Fujishima,
Ekabi), Elsevier Science, Amsterdam, The submitted.
Netherlands, 1993, pp. 353–363. 50. Y. Kikuchi, K. Sunada, T. Iyoda et al., J. Pho-
31. M. Herlem, F. Bobilliart, A. Thiebault, in tochem. Photobiol. A: Chem. 1997, 106,
Encyclopedia of Electrochemistry of the El- 51–56.
ements, Organic Section, (Eds.: A. J. Bard, 51. D. E. Skinner, D. P. Colombo, Jr., J. J.
H. Lund), Marcel Dekker, Inc., New York, Cavaleri et al., J. Phys. Chem. 1995, 99,
1978, pp. 5–16, Vol. XI. 7853–7856.
32. D. Lawless, N. Serpone, D. Meisel, J. Phys. 52. N. Serpone, D. Lawless, R. Khairutdinov
Chem. 1991, 95, 5166–5170. et al., J. Phys. Chem. 1995, 99, 16655–16661.
33. Y. Nakato, A. Tsumura, H. Tsubomura, 53. D. P. Colombo, Jr., R. M. Bowman, J. Phys.
J. Phys. Chem. 1983, 87, 2402. Chem. 1996, 100, 18445–18449.
54. G. Rothenberger, J. Moser, M. Grätzel et al., 80. S. I. Matsushita, T. Miwa, D. A. Tryk et al.,
J. Am. Chem. Soc. 1985, 107, 8054–8059. Langmuir 1998, 14, 6441–6447.
55. K. -M. Schindler, M. Kunst, J. Phys. Chem. 81. T. A. Egerton, C. J. King, J. Oil Col. Chem.
1990, 94, 8222–8226. Assoc. 1979, 62, 386–391.
56. J. M. Warman, M. P. Haas, P. Pichat et al., 82. C. Kormann, D. W. Bahnemann, M. R.
J. Phys. Chem. 1991, 95, 8858–8861. Hoffmann, Environ. Sci. Technol. 1991, 25,
57. S. T. Martin, H. Herrmann, W. Choi et al., 494–500.
J. Chem. Soc. Faraday Trans. 1994, 90, 83. G. P. Lepore, C. H. Langford, J. Vi-
3315–3322. chova et al., J. Photochem. Photobiol. A:
58. D. P. Colombo, Jr., R. M. Bowman, J. Phys. Chem. 1993, 75, 67–75.
Chem. 1995, 99, 11752–11756. 84. M. Anpo, N. Aikawa, Y. Kubokawa et al.,
59. D. Bahnemann, A. Henglein, J. Lilie et al., J. Phys. Chem. 1985, 89, 5689.
J. Phys. Chem. 1984, 88, 709–711. 85. R. F. Howe, M. Grätzel, J. Phys. Chem. 1985,
60. A. Henglein, Pure & Appl. Chem. 1984, 56, 89, 4495–4499.
1215–1224. 86. R. Cai, R. Baba, K. Hashimoto et al., J. Elec-
61. Y. Nosaka, J. Phys. Chem. 1990, 94, troanal. Chem. 1993, 360, 237–245.
3752–3755. 87. K. U. Ingold, Acc. Chem. Res. 1969, 2, 1–9.
62. M. O’Neil, J. Marohn, G. McLendon, J. Phys. 88. N. A. Clinton, R. A. Kenley, T. G. Traylor,
Chem. 1990, 94, 4356–4363. J. Am. Chem. Soc. 1975, 97, 3746–3751.
63. M. A. Grela, M. E. J. Coronel, A. J. Colussi, 89. N. A. Clinton, R. A. Kenley, T. G. Traylor,
J. Phys. Chem. B 1996, 100, 16940–16946. J. Am. Chem. Soc. 1975, 97, 3752–3757.
64. R. Kopelman, Science 1988, 241, 1620–1626. 90. N. A. Clinton, R. A. Kenley, T. G. Traylor,
65. B. G. Streetman, in Solid State Electronic J. Am. Chem. Soc. 1975, 97, 3757–3762.
Devices, Prentice-Hall International, En- 91. Y. Ohko, D. A. Tryk, K. Hashimoto et al.,
glewood Cliffs, New Jersey, USA, 1995, J. Phys. Chem. B 1998, 102, 2699–2704.
pp. 103–106. 92. G. B. Raupp, C. T. Junio, Appl. Surf. Sci.
66. D. Duonghong, M. Grätzel, J. Chem. Soc., 1993, 72, 321–327.
Chem. Commun., 1984, 1597–1599. 93. J. Stark, J. Rabani, J. Phys. Chem. B 1999,
67. Y. Nosaka, Y. Yamashita, H. Fukuyama, 103, 8524–8531.
J. Phys. Chem. B 1997, 101, 5822–5827. 94. J. Schwitzgebel, J. G. Ekerdt, H. Gerischer
68. V. Augugliaro, L. Palmisano, M. Schiavello, et al., J. Phys. Chem. 1995, 99, 5633–5638.
AIChE J. 1991, 37, 1096–1100. 95. G. A. Russell, J. Am. Chem. Soc. 1957, 79,
69. M. A. Aguado, M. A. Anderson, C. G. 3871–3877.
Hill, Jr., J. Mol. Catal. 1994, 89, 165–178. 96. S. Sitkiewitz, A. Heller, Nouv. J. Chem.
70. N. Serpone, J. Photochem. Photobiol. A: 1996, 20, 233–241.
Chem. 1997, 104, 1–12. 97. T. Minabe, P. Swanyama, Y. Kikuchi et al.,
71. L. Davydov, P. G. Smirniotis, S. E. Pratsinis, Electrochem. 1999, 67, 1132–1134.
Indust. Eng. Chem. Res. 1999, 38, 1376–1383. 98. T. Minabe, D. A. Tryk, P. Sawunyama et al.,
72. R. I. Bickley, J. Catal. 1973, 31, 398–407. J. Photochem. Photobiol. A: Chem. 2000, 137,
73. Y. Ohko, K. Hashimoto, A. Fujishima, 53–62.
J. Phys. Chem. A 1997, 101, 8057–8062. 99. A. V. Emiline, V. Ryabchuk, N. Serpone,
74. R. W. Matthews, J. Phys. Chem. 1987, 91, J. Photochem. Photobiol. A: Chem. 2000, 133,
3328–3333. 89–97.
75. I. Sopyan, S. Murasawa, K. Hashimoto 100. T. Noguchi, A. Fujishima, P. Sawunyama
et al., Chem. Lett. 1994, 723–726. et al., Environ. Sci. Technol. 1998, 32,
76. N. Negishi, T. Iyoda, K. Hashimoto et al., 3831–3833.
Chem. Lett. 1995, 841–842. 101. Y. Ohko, A. Fujishima, K. Hashimoto,
77. I. Sopyan, M. Watanabe, S. Murasawa et al., J. Phys. Chem. B 1998, 102, 1724–1729.
J. Photochem. Photobiol. A: Chem. 1996, 98, 102. W. W. Dunn, Y. Aikawa, A. J. Bard, J. Am.
79–86. Chem. Soc. 1981, 103, 3456–3459.
78. I. Sopyan, M. Watanabe, S. Marasawa et al., 103. W. W. Dunn, Y. Aikawa, A. J. Bard, J. Elec-
J. Electroanal. Chem. 1996, 415, 183–186. trochem. Soc. 1981, 128, 222–224.
79. S. Matsushita, T. Miwa, A. Fujishima, 104. M. D. Ward, A. J. Bard, J. Phys. Chem. 1982,
Chem. Lett. 1997, 925–926. 86, 3599–3605.
105. M. W. Peterson, J. A. Turner, A. F. Nozik, 123. M. Anpo, Y. Ichihashi, M. Takeuchi et al.,

J. Phys. Chem. 1991, 95, 221–225. Res. Chem. Intermed. 1998, 24, 143–149.
106. C. -M. Wang, A. Heller, H. Gerischer, J. Am. 124. L. Zang, C. Lange, I. Abraham et al., J. Phys.
Chem. Soc. 1992, 114, 5230–5234. Chem. B 1998, 102, 10765–10771.
107. W. -Y. Lin, N. R. de Tacconi, R. L. Smith 125. M. Hamerski, J. Grzechulska, A. W. Mora-
et al., J. Electrochem. Soc. 1997, 144, wski, Sol. Energy 1999, 66, 395–399.
497–502. 126. Y. Murata, S. Fukuta, S. Ishikawa et al.,
108. Y. Komoda, N. Sakai, T. N. Rao et al., Lang- Sol. Energy Mater. Sol. Cells 2000, 62,
muir 1998, 14, 1081–1091. 157–165.
109. Y. Komoda, T. N. Rao, D. A. Tryk et al., 127. X. Fu, l. A. Clark, Q. Yang et al., Environ.
J. Electroanal. Chem. 1998, 459, 155–165. Sci. Technol. 1996, 30, 647–653.
110. K. Vinodgopal, S. Hotchandani, P. V. Ka- 128. C. Anderson, A. J. Bard, J. Phys. Chem.
mat, J. Phys. Chem. 1993, 97, 9040–9044. 1995, 99, 9882–9885.
111. K. Vinodgopal, U. Stafford, K. A. Gray et al., 129. C. Anderson, A. J. Bard, J. Phys. Chem. B
J. Phys. Chem. 1994, 98, 6797–6803. 1997, 101, 2611–2616.
112. K. J. Vetter, in Electrochemical Kinetics: The- 130. N. Takeda, T. Torimoto, S. Sampath et al.,
oretical and Experimental Aspects, Academic J. Phys. Chem. 1995, 99, 9986–9991.
Press, New York, 1967, pp. 406–417. 131. N. Takeda, M. Ohtani, T. Torimotoet al., J.
113. N. Sato, in Electrochemistry at Metal and Phys. Chem. B 1997, 101, 2644–2649.
Semiconductor Electrodes, Elsevier, Amster- 132. P. W. Atkins, in Physical Chemistry, Oxford
dam, 1998, p. 400. University Press, Oxford, 1998.
114. J. M. Kesselman, G. A. Shreve, M. R. Hoff- 133. J. VanMuylder, M. Pourbaix, in Atlas of Elec-
mann et al., J. Phys. Chem. 1994, 98, trochemical Equilibria in Aqueous Solutions,
13385–13395. (Eds.: M. Pourbaix), Pergamon, Oxford,
115. H. Sakai, R. Baba, K. Hashimoto et al., 1966, pp. 449–457.
J. Phys. Chem. 1995, 99, 11896–11900. 134. V. D. Parker, G. Sundholm, U. Svanholm
116. K. Ikeda, H. Sakai, R. Baba et al., Chem. et al., in Encyclopedia of Electrochemistry of
Lett. 1995, 979–980. the Elements, Organic Section, (Eds.: A. J.
117. K. Ikeda, H. Sakai, R. Baba et al., J. Phys. Bard, H. Lund), Marcel Dekker, New York,
Chem. B 1997, 101, 2617–2620. 1978, pp. 181–225, Vol. XI.
118. K. Ikeda, K. Hashimoto, A. Fujishima, 135. J. N. Butler, in Advances in Electrochem-
J. Electroanal. Chem. 1997, 437, 241–244. istry and Electrochemical Engineering (Eds.:
119. H. Maeda, K. Ikeda, K. Hashimoto et al., P. Delahay), Interscience Publishers, New
J. Phys. Chem. B 1999, 103, 3213–3217. York, 1970, pp. 77–139, Vol. 7.
120. P. Sawunyama, L. Jiang, A. Fujishima et al., 136. J. P. Hoare, in Encyclopedia of Electrochem-
J. Phys. Chem. B 1997, 101, 11000–11003. istry of the Elements, (Eds.: A. J. Bard),
121. P. Sawunyama, A. Fujishima, K. Hashi- Vol. 2, Marcel Dekker, New York, 1974,
moto, Langmuir 1999, 15, 3551–3556. p. 220.
122. H. Gerischer, A. Heller, J. Phys. Chem. 137. Y. Ohko, K. Ikeda, T. N. Rao et al., Z. phys.
1991, 95, 5261–5267. Chem. 1998, 213, 33–42.
6.2 1700 [1–4]. Besides a rapid increase in the

Applications of TiO2 Photocatalysis number of publications, the number of
patents is also growing as a result of high
Tata N. Rao, Donald A. Tryk, and Akira
commercial interest in applications, many
Fujishima
of these being developed in Japan. The
University of Tokyo, Tokyo, Japan
topics covered are shown in Fig. 2, to-
6.2.1 gether with the number of publications
Introduction in each area [5]. In addition to the compre-
hensive listings of publications that have
The applications of titanium dioxide (TiO2 ) been assembled by Blake, there have been
photocatalysis date from the mid-1970s, a number of reviews of the photocatalysis
with the solar-assisted purification of area [6–32].
contaminated water, but now include a The Section ‘‘Fundamentals of Cataly-
staggering array of diverse ideas from sis’’ provides an overview of the fundamen-
the purification and sterilization of indoor tal aspects of TiO2 photocatalysis, and the
air to self-cleaning ceramics and glass, to present chapter is devoted to the presen-
antifogging mirrors, and to light-induced tation of some of the myriad applications
cancer treatment. These technologies are (Table 1). The intent of this chapter is to
becoming more attractive because of the
present an overview of the state of the art
environmentally benign nature of TiO2 ,
in the applications of TiO2 photocataly-
its low cost, nontoxicity, and ability to self-
sis for sterilization, disinfection, air and
regenerate combined with the possibility
water purification, and self-cleaning and
of using only solar light or ambient indoor
antifogging surfaces. To indicate the broad
light together with ambient oxygen (O).
Additional attractive features include its range of applications, a brief overview of
unique chemical properties, for example, selected applications is given, including
its resistance to photocorrosion, its ability discussion of their practicability and future
to simultaneously produce at least two prospects. Emphasis is given to photocat-
types of highly reactive oxygen radical alytic systems that either are already being
species, and its ability to modify its commercialized or are ready for commer-
wetting properties under illumination. cialization.
Part of the appeal stems from the recent
emphasis on indoor environments in 6.2.2
which photocatalysis is expected to play Photocatalytic Sterilization and Disinfection
a significant role. Another appealing
aspect is the ability of TiO2 to maintain In the medical field, the development of
sterile conditions on illuminated surfaces ways to control outbreaks of disease caused
without the use of antiseptics. by new strains of deadly bacteria and
The growth in the number of publica- viruses has been a topic of serious dis-
tions on photocatalysis has been accelerat- cussion. Sterilization and disinfection are
ing steadily over the past quarter century, seen to be increasingly important, even in
as seen in Fig. 1. The total number of environments other than hospitals, such
papers that have appeared up to Septem- as bathrooms and kitchens. The photo-
ber 2000 is slightly more than 5000 and catalytic action of TiO2 can be extremely
the number of patents is just more than useful in preventing disease, particularly
6.2 Applications of TiO2 Photocatalysis 537
Photocatalysis citations
1200
1000
Total citations
800
Patents
Number
600
400
200
0
1969 1973 1977 1981 1985 1989 1993 1997
Year
Fig. 1Numbers of papers and patents related to photocatalysis up to
September 2000 [4].
Tab. 1 Selected applications of photocatalysis
Property Category Application
Self-cleaning Materials for residential Exterior tiles, kitchen and bathroom components,
and office buildings interior furnishings, plastic surfaces, aluminum
siding, building stone and curtains, paper
window blinds
Indoor and outdoor lamps Translucent paper for indoor lamp covers, coatings
and related systems on fluorescent lamps and highway tunnel lamp
cover glass
Materials for roads Tunnel wall, soundproofed wall, traffic signs and
reflectors
Others Tent material, cloth for hospital garments and
uniforms and spray coating for cars
Air cleaning Indoor air cleaners Room air cleaner, photocatalyst-equipped air
conditioners, and interior air cleaner for factories
Outdoor air purifiers Concrete for highways, roadways and footpaths,
tunnel walls, soundproofed walls, and building
walls
Water purification Drinking water River water, groundwater, lakes, and water-storage
tanks
Others Fish feeding tanks, drainage water, and industrial
wastewater
Antitumor activity Cancer therapy Endoscopic-like instruments
Self-sterilizing Hospital Tiles to cover the floor and walls of operating
rooms, silicone rubber for medical catheters,
and hospital garments and uniforms
Others Public rest rooms, bathrooms, and rat-breeding
rooms
Generic information on photocatalysis

(up to January 1999)
Patents Active companies

(792)
Reviews and articles Miscellaneous topics

(329) H2O2 and other oxidants (361)
(107)
Photocatalysts Engineering issues
Reactor and system design System analysis

(211) (38)
Modified TiO2 Metallized TiO2 Hydrophobic/hydrophilic surface

(354) (179) (22)
Dye-sensitized TiO2 Metal-doped TiO2 or mixed oxides

(65) (122)
Immobilized photocatalysts Other semiconductors

(348) (189)
Fig. 2 Generic information on photocatalytic processes up to January 1999.
because it operates in a passive fashion, (UV) illumination [33]. Since then there
that is, without the need for electrical have been several reports and patents ap-
power or chemical reagents, only light and pearing on this topic [5, 34–44]. Blake
oxygen being required. For these reasons and coworkers have published a compre-
photocatalysis recently has attracted the at- hensive review of this, including a list
tention not only of the medical world but of patents on photocatalytic disinfection
also of the construction industry in Japan. and killing of cancer cells using TiO2 [5].
Most of the work reported on photocat- Many of the commercial products avail-
alytic disinfection has been carried out in able in the market at present are based on
aqueous suspensions, in view of applica- self-sterilizing photocatalytic surfaces and
tions in water sterilization. In 1985, for disinfection of indoor air.
the first time, Matsunaga and coworkers Bacteria find places in bathrooms,
reported the possibility of killing mi- kitchens, and hospital operating rooms
crobial cells in water using a TiO2 -Pt where they can multiply. Once they find
(platinum) photocatalyst under ultraviolet a suitable environment, they can increase
in number at an exponential rate. The key TiO2 -coated tile and, when the surface
point regarding photocatalysis is that it is irradiated with UV light, the metal
can be very effective in controlling bacte- cations are photocatalytically converted to
ria while they still are relatively small in the metallic form as ultrafine particles that
number, before they start multiplying out are firmly attached to the surface. The
of control. photocatalytic deposition of various metals
on TiO2 is also discussed in a later section.
6.2.2.1 Photocatalytic Antibacterial Tiles The present method of metal deposi-
In general, the effective use of photocat- tion was found to have an advantage over
alytic technology depends heavily on the the conventional method in which the
specific way in which the TiO2 powder metal powders are mixed with the glaze
is used. One of the simplest approaches, and fired on a tile [25]. With the conven-
often used in model studies on water treat- tional method, most of the metal particles
ment, is to mix the powder with the water are submerged in the glaze, only a few
and expose the suspension to UV light or being exposed on the surface. Thus, the
sunlight. Although the separation of the particles in the glaze are not effective for
catalyst powder from the water is not an the antibacterial effect. The new photo-
insurmountable problem, an alternate ap- catalytic approach allows the formation of
proach is to immobilize the powder on a a high-density layer of antibacterial metal
substrate [25]. particles on the TiO2 layer, which then ex-
In view of indoor applications, various hibits a strong antibacterial activity. The
kinds of substrates such as glass plates and photocatalytic antibacterial tiles prepared
tiles have been examined as substrates. this way are highly effective and durable.
It has been found that even very thin
films (less than a few micrometers in 6.2.2.2 Sterilizing Effect of TiO2
thickness) are photocatalytically active [45, As already indicated, one of the promising
46]. Moreover, (∼1 µm) TiO2 films on the applications of TiO2 is photocatalytic ster-
surfaces of ceramic tiles have been found ilization. Escherichia coli has been studied
to be effective for antibacterial functions. widely, but work on other organisms also
These films can be prepared by spraying has been reported. As recently reviewed,
a liquid suspension containing TiO2 on various organisms have been examined for
a glazed ceramic tile, followed by heating photocatalytic sterilization [5]. Although
at 800 ◦ C [25, 47, 48]. Films prepared this the feasibility of sterilization was realized
way are attached to the tiles so strongly from the beginning, the mechanism of
that they are difficult to scrape off. cell death was not understood completely.
In the case of indoor antibacterial Initially, the hydroxyl radical was thought
tiles, one has to keep in mind the to be the main species responsible for
possibility of accumulation of bacteria in cell death. However, peroxide currently is
areas either temporarily or permanently considered to be one of the most active
inaccessible to light. To overcome this agents. Evidence against the involvement
problem, silver or copper metal particles of hydroxyl radicals is that there is lit-
have been deposited photocatalytically on tle effect of mannitol, a hydroxyl radical
catalyst surfaces [25]. These metals are scavenger [39].
widely used as antimicrobial agents. Either A typical experiment involves plac-
of the metal solutions is sprayed on the ing 150 µL of an E. coli suspension
containing 3 × 104 cells on an illuminated damage caused by peroxidation of lipids in

TiO2 -coated glass plate (1 mW cm−2 UV the cell membrane [42]. After eliminating
light) [39]. Under these conditions, there the protection of the cell wall, the oxidative
were no surviving cells after one hour of damage takes place on the underlying cy-
illumination. By contrast, in the absence toplasmic membrane [44]. Photocatalytic
of a TiO2 film, it took four hours for the action progressively increases the cell per-
destruction of 50% of the cells. In another meability and subsequently allows the
experiment, the E. coli suspension was sep- free efflux of intracellular contents, even-
arated from the TiO2 surface by a porous tually leading to cell death. Free TiO2
PTFE membrane (pore size, 0.4 µm). The particles also are expected to gain access
into membrane-damaged cells, where they
function of this membrane is that it pre-
can attack the intracellular components,
vents the passage of E. coli cells through
accelerating cell death.
its pores. Because of its thickness (50 µm),
In view of practical applications, pho-
relatively reactive radical species such as
tocatalytic antibacterial tiles mentioned
•
OH generated at the irradiated TiO2 sur-
earlier also were tested by placing live
face are expected to be deactivated during bacteria on them. The bacteria on these
passage through it. Under these condi- tiles were killed completely in an hour un-
tions, the complete destruction of cells was der illumination equal to the brightness
observed in four hours. This result tends of the surface of a study desk (∼4 µW
to rule out the involvement of • OH, but cm−2 ) (Fig. 3). On the basis of these re-
not that of peroxide. However, the latter sults, such tiles were tested on the floor
by itself is not as effective, so the addi- and walls of a hospital operating room
tional presence of other oxygen species where sterile conditions are crucial. Af-
must be considered. Further experiments ter installing the tiles, the bacterial counts
are necessary to understand this mech- on the walls decreased to negligible lev-
anism in more detail. Regardless of the els in a period of one hour. Surprisingly,
oxidative species involved, the cell mem- the bacterial counts in the surrounding
brane damage was attributed to oxidative air also decreased significantly (Fig. 4).
Photocatalytic
Ordinary tile
antibacterial tile
Methicillin-
resistant Staph.
aureus (MRSA)
E.coli
Pseudomonas
aeruginosa
Fig. 3 Demonstration of the

1000 lux self-sterilizing action of
1000 lux
illumination photocatalytic tiles before and
illumination
(1 hour) after illumination [25].
(CFU min−1 cm−2 L−1)

Before installation
Number of airborne
30
bacteria
20 Installation of photocatalytic tile
10
0
Immediately 1 1.5 2 2.5 3
after Month Months Months Months Months
installation later later later later later
Immediately after installation of the photocatalytic tiles,

the bactericidal effect on airborne bacteria is apparent
and continues indefinitely.
Fig. 4 Sterilizing effect of illuminated photocatalytic tiles on airborne

bacteria as a function of time [25].
The applications of these tiles can be ex- TiO2 is capable of decomposing not only
tended further to public toilet facilities, dead cells but also the toxins released
shower rooms, and biological experimen- during cell death. Jacoby and coworkers
tal facilities such as breeding rooms for have demonstrated the complete oxidation
experimental rats. The tests conducted in of E. coli cells to carbon dioxide (CO2 ) [40].
these facilities have indicated that the an- In the case of E. coli, TiO2 decomposes
tibacterial tiles are effective not only in both the living cells and the endotoxin
killing the bacteria but also in significantly released from these cells at the time
reducing the odor caused by ammonia, of death [49]. This unique function of
which is common in such facilities. The de- TiO2 has attracted tremendous attention in
odorizing functions of TiO2 are discussed Japan. Endotoxin is a cell wall constituent
in a later section. of bacteria that consists of a sugar chain,
an O-antigen, and a complex lipid referred
6.2.2.3 Destruction of Microbial Toxins to as lipid A. This endotoxin is toxic and
A major advantage of photocatalytic an- can cause critical problems in medical
tibacterial tiles over conventional antibac- facilities and in factories manufacturing
terial tiles is that they not only kill bacte- pharmaceuticals and medical devices, as
rial cells but also completely decompose well as in food-processing factories. In the
them. Conventional self-sterilizing sur- summer of 1996 in western Japan, there
faces, such as those with silver coatings, are was a serious outbreak of food poisoning,
not useful for long-term use because dead in which nearly 1800 people, including
cells can eventually cover the surface, ren- many children, were hospitalized, and
dering it ineffective. The advantage of TiO2 twelve died. The poisoning was caused
is that it continues to work well even when by the O-157 endotoxin.
there are cells covering the surface and Figure 5 shows the concentration ch-
while the bacteria are actively propagating. ange of endotoxin and the survival ratio
120 120
Remaining endotoxin
(in dark)
100 100
Remaining endotoxin
80 80
Live bacteria
[%]
[%]
60 60
Remaining endotoxin
(under illumination)
40 40
20 20
Live bacteria
0 0
0 1 2 3 4 5
Irradiation time
[hours]
Fig. 5Sterilization of E. coli and endotoxin decomposition (endotoxin
amount: 3–8 EU cc−1 = 100%). Both sterilization and endotoxin
decomposition commence immediately [49].
of E. coli as a function of illumination time new cell growth and old cell death are
on a TiO2 -coated glass plate. It was found kept in balance. In cancer, this balance
that two hours were required to decom- is disrupted. This disruption can result
pose most of the toxin and four hours in the formation of a malignant tu-
were required to decompose it completely mor. The selective killing of cancer cells
under UV illumination (0.4 mW cm−2 ) on in vivo by photocatalysis is a challeng-
a TiO2 film [49]. This can be contrasted ing task.
with other techniques that have been de- The possibility of applying photocatalytic
veloped, for example, thermal treatment at techniques to the photodynamic therapy
250 ◦ C for 30 minutes or chemical treat- (PDT) of cancer led to early efforts in this
ment in ethanol containing NaOH. These area. It was found that the multiplication
results clearly indicate that the antibacte- of human malignant cells was suppressed
rial effects of TiO2 -coated materials involve remarkably both in vitro [50–54] and in
not only the nullification of the viability of vivo [55, 56] in the presence of photoirra-
the bacteria but also the destruction of the diated TiO2 . Initially, TiO2 was found to
bacterial cell. be effective in killing HeLe cells under UV
irradiation [50]. Later, various experimen-
6.2.2.4 Photocatalytic Cancer Therapy tal conditions were examined, including
Cancer treatment is carried out mainly by the effect of superoxide dismutase, which
means of radiation and chemical thera- enhances the effect, caused by the produc-
pies, which may generate a wide range tion of peroxide [51–53]. The possibility of
of side effects. Photocatalysis offers a selectively killing a cancer cell was demon-
new approach in cancer treatment. Can- strated using a polarized, illuminated TiO2
cer arises from a loss of normal growth microelectrode [54]. Photocatalytic killing
control. In normal tissues, the rates of of human U937 monocytic leukemia cells
also was reported by Chinese researchers to cause tumors to form. When the size
using colloidal TiO2 [57]. A complete of the tumors grew to about 0.5 cm, a
photocatalytic destruction of these cells solution containing fine particles of TiO2
was observed after 30 minutes of UV was injected. After two or three days, the
illumination. The destruction process in- skin was cut open to expose the tumor,
volved membrane blistering and DNA which then was irradiated. This treatment
fragmentation. All of these effects are char- clearly inhibited tumor growth (Fig. 6).
acteristics of apoptosis (programmed cell After 13 more days, the treatment was
death) that is known to occur when can- repeated with TiO2 photocatalyst and a
cer cells are treated with antitumor agents, further marked antineoplastic effect was
radiation, or high temperature. observed [55].
In collaboration with urologists at Yoko- Although this technique offers some
hama City University, animal experiments promise for cancer therapy, it may not be
also have been conducted. Cancer cells effective in treating large tumors. To make
were implanted under the skin of mice this technique more useful, a device (a
2
2
1
1
(a) (b)
Fig. 6 Animal test of photocatalytic cancer injected into Tumor 1. Tumor 2, which was not
therapy; photograph of nude mouse just after injected with TiO2 particles, was also opened
initial treatment (a) and four weeks after surgically. Both tumors were irradiated with a
treatment (b). TiO2 powder (0.4 mg) was mercury lamp for one hour [25].
Eye
TiO2 powder
UV light
Optical fiber
Fig. 7 Proposed instrument for

photocatalytic cancer
Cancer
therapy [25].
modified endoscope) has been developed such as trichloroethylene (TCE), tetra-

to allow the tumor to be exposed to light chloroethylene, 1,3-dichlorobenzene, and
while TiO2 powder is being added to the dichloromethane, which are widely used
tumor (Fig. 7). Because a photocatalytic as extracting solvents in industrial pro-
reaction occurs at the illuminated areas, it cesses, are released into the environment
is possible to make an attack on cancer cells in large quantities. Air stripping followed
alone. There are special types of dyes that by adsorption on activated carbon is a
have been developed that selectively bind common method employed to treat wa-
to cancer cells. They can be used along with ter and soil that is contaminated with
TiO2 powder to trace the location of cancer such volatile organics. Such a treatment
cells and treat with the new endoscope-like method, merely transferring toxic materi-
instrument. Thus it is possible to avoid als from one medium to another, is not a
damage to normal cells. long-term solution to the problem of haz-
Obviously, the excitation light should ardous waste disposal. TiO2 photocatalysis
not cause mutations in normal cells. The is advantageous in handling such prob-
results of animal experiments have shown lems because of its ability to completely
that the near UV light that is used in and efficiently mineralize a wide range
photocatalytic reactions, with wavelengths of pollutants. Purification of indoor air is
of 300–400 nm, is safe. another area where photocatalysis plays
a dominant role. Photocatalytic treatment
6.2.3
of various kinds of gaseous pollutants is
Photocatalytic Air Purification
summarized in this section.
Photocatalytic oxidation of various air
pollutants currently is a rapidly de- 6.2.3.1 Deodorization of Indoor Air
veloping field of heterogeneous pho- TiO2 photocatalysis is very successful for
tocatalysis. Volatile organic compounds the purification of indoor air in which
(VOCs), nitrogen oxides, and sulfur the air pollutants are relatively low in
oxides are the most important an- concentration. Experimental results have
thropogenic pollutants generated in ur- indicated that the quantum efficiency of
ban and industrial areas. For example, photocatalytic reactions is high at low
volatile chlorinated organic compounds light intensities (Fig. 8) [25]. Thus, the
100
80
Quantum efficiency
60
[%]
40
20
0
Fig. 8 Inverse relation between
100 101 102 103 104 105 quantum efficiency and UV light
UV intensity intensity for low-pollutant
[µW cm−2] concentrations [25].
most appropriate applications for the that TiO2 particles should be aggregated
photocatalytic approach are those that before they are added to the paper pulp to
involve low concentrations of pollutants maintain the paper strength. If the TiO2
but that also involve serious risks to health particles are dispersed in the paper, there
or comfort. A malodorous pollutant is a is photocatalytic attack on the fibers and
good example because the actual quantity loss of strength. Acetaldehyde degradation
of the chemical may be very small. The was seen with the TiO2 -containing papers
photocatalytic sterilization discussed in the even under very weak UV irradiation. The
previous section is another example. decomposition rate was found to increase
The Japanese government regulates the with increasing amounts of TiO2 con-
concentrations of major malodorous sub- tained in the paper [61]. Furthermore, the
stances as ammonia, hydrogen sulfide, papers showed high stability under illumi-
methyl mercaptan, and acetaldehyde in the nation for about 600 hours. These papers
environment. Methyl mercaptan, for ex- also were tested for the decomposition of
ample, is regulated to concentrations lower cigarette smoke residue. Smoke from one
than 0.002–0.01 ppm. The human sense of cigarette was adsorbed on a 5 × 5-cm sheet
smell, however, senses it at a concentration of TiO2 -containing paper. TiO2 -free paper
as low as 0.00012 ppm. A conventional flu- was also used for control experiments. The
orescent lamp can be used as a light source
smoke residue on the TiO2 paper disap-
for a photocatalytic reactor to control this
peared after only two hours under black
substance [25]. Regarding acetaldehyde,
light (1 mW cm−2 ) illumination and after
which is a highly malodorous compound,
one week under regular fluorescent light
a fluorescent lamp is also sufficient to
(10 µWcm−2 ).
decompose it at an air concentration of
On the basis of these findings, vari-
5 ppm, which is 10–100 times higher than
ous kinds of commercial photocatalytic air
that at which it becomes offensive. Several
cleaners currently are being produced in
studies have been conducted on acetalde-
hyde decomposition under various light Japan by Daikin Industries and are being
intensities and concentrations to under- marketed worldwide (Fig. 9). The Ishihara
stand the reaction mechanism [45, 58–60]. Techno Corporation has developed pho-
Another interesting application of TiO2 tocatalytic filter elements for air cleaners.
photocatalysis was demonstrated using They contain TiO2 and activated carbon
TiO2 -containing paper. In collaboration (C), the latter trapping pollutants for sub-
with the Gifu Prefectural Paper Re- sequent destruction by the irradiated TiO2 .
search Institute, TiO2 -containing papers These filters feature a honeycomb-type
possessing a high catalytic activity toward construction for minimum pressure drop.
the decomposition of acetaldehyde and The air cleaners are available in various
cigarette smoke residues were prepared. sizes, which have been developed to clean
The TiO2 -containing paper was prepared the air in a car or in an entire factory or
from a mixture of softwood kraft pulp and hospital.
TiO2 aqueous solution. The details of the
preparation were reported earlier [61]. The 6.2.3.2 Treatment of Industrial Gaseous
amount of added TiO2 was varied from 2 Effluents
to 10% based on the weight of the pulp. Because gas-phase photocatalysis is in gen-
The important point in this preparation is eral more than 10 times more effective
Pre-filter Plasma-ionizing unit Double-layer roll Photocatalyst 546
photocatalyst filter
Double-layer construction means
more power. Energy savings by using
an inverter motor.
6 Nonsolar Energy Applications
Dirty air Clean air
Newly developed fan and DC motor

Special inverter lamp
Double lumps mean more power.
Fig. 9 Schematic diagram of the components of an indoor room air-purification unit (Courtesy of Daikin Industries).
than that in liquid phase [22] the use of high levels of water vapor (>4000 ppm)
photocatalytic purification methods for in- were found to decrease the degradation
dustrial gas effluents has gained much efficiency for acetone [76]. The role of
attention. Almost all kinds of organic com- water vapor in gas-phase photocatalysis
pounds are photocatalytically degradable is not completely understood. Even for
on TiO2 . Gas-phase carbon tetrachloride the same compound, divergent results
(CCl4 ) appears to be an exception, in that can be obtained by different groups [77,
it is resistant to photocatalytic degrada- 78]. This may result from subtle dif-
tion [62, 63]. The presence of methanol ferences in the film microstructure. For
helps in slightly increasing the photo- example, one film may have relatively
catalytic reduction of CCl4 , the effect small pores and may become flooded
being higher in aqueous solutions [63, easily in the presence of high humidity,
64]. There have been several reports on but such a film may be more effective
the photocatalytic oxidation of individual in dry conditions because water that is
gaseous substrates such as halogenated produced during the photocatalytic reac-
organics [62, 65–67], alkanes [68], alco- tion may be trapped more effectively [78].
hols [62, 69], aldehydes [60, 70], aromat- This is an area that requires further
ics [62, 66, 71, 72], and heterocycles [73]. study.
Among these organic substrates, only a
The design of suitable reactors is very
few compounds such as TCE, methanol,
crucial for the achievement of maxi-
and ethanol exhibited quantum efficien-
mum conversion efficiencies. Ollis has
cies between 80 and 100% [22, 62]. In
made comparisons of various types of
contrast, the quantum efficiencies for the
photoreactors for air purification [9]. Al-
degradation of nonhalogenated organic
though monolithic photocatalytic reactors
substrates such as benzene, toluene, and
are well suited for air handling in build-
xylene were found to be below 15% [62].
ings and exhausts, fluidized-bed reactors
However, a drastic enhancement in the
efficiencies for the degradation of some are advantageous for large-scale opera-
of these compounds has been reported tions. The monolithic catalysts allow a
to be possible by mixing with chlorocar- drastic reduction in the pressure drop
bons such as TCE that cause radical chain produced by the passage of the gas
reactions [62]. through the catalyst. Monolithic structures
The presence of water vapor plays an increase the illumination area of the cata-
important role in the gas-phase photo- lyst in a three-dimensional way. Although
catalysis. For example, moderate levels fluidized-bed reactors offer good catalyst-
of water vapor promote the photocatalytic light and catalyst-gas contacts, bubble-
degradation of toluene and m-xylene [66, phase formation is a disadvantage. Re-
74], whereas they inhibit the degradation cently, Vorontsov and coworkers [76], have
of ethylene [75]. In some cases, such as demonstrated photocatalytic oxidation of
TCE [62], deactivation of the photocata- acetone using a vibrofluidized-bed reac-
lyst was observed in the absence of water tor. In this type of reactor, fluidization
vapor, which probably is due to the exis achieved by vibration. The advantage
haustion of surface hydroxyls or to the is that it can operate under low-gas flow
formation of an intermediate species that rates compared to conventional fluidized-
blocks the active catalyst sites. However, bed reactors.
6.2.3.3 Outdoor Air Purification (NOx gas may be trapped. The accumulation of
Removal) HNO3 on the catalyst surface retards the
Air pollution, especially nitrogen oxides photocatalytic action. Therefore it is nec-
(NOx ) contamination from the combus- essary to wash the catalyst after certain
tion of hydrocarbons, is a particularly time intervals. However, if the photocata-
serious problem in urban areas. Despite lyst is used outdoors, the products may be
serious efforts toward emission control, washed off by rainfall. A similar mecha-
the concentrations of NOx often ex- nism is expected for SO2 removal as well.
ceed the air-quality standard, especially The range of concentrations of air
in large cities. TiO2 photocatalysis ap- pollutants that can be removed efficiently
pears to be a promising technology for is from 0.01 ppm to 10 ppm. These
the removal of low concentrations of NOx concentrations range from those present
from ambient air [79]. Daikin Industries typically in the atmosphere to those
has demonstrated the efficient removal in highway tunnels. However, at the
of NOx from indoor air using a photo- end of an automobile’s exhaust pipe,
catalyst coated on activated carbon [25]. the concentration of pollutants is higher
According to their results, the indoor NOx than 100 ppm. Therefore, attaching TiO2
concentration decreases from 0.1 ppm to directly to the car exhaust system does
0.06 ppm (the air-pollution standard) in not seem to be a good idea. This is better
25 minutes. handled by a catalytic converter.
The mechanism of nitrogen oxide (NO) Further efforts are being made to im-
removal by a photocatalyst is somewhat prove the catalytic response of TiO2 for
complicated. It is assumed that the NO is NOx removal. For example, Nakamura
photocatalytically oxidized to NO2 , which and coworkers [80], have used hydrogen
is finally converted to nitric acid (HNO3 ) plasma–treated TiO2 to sensitize the
(Fig. 10) [25]. At the NO2 stage, part of catalytic response in the visible wave-
the gas may escape from the photocatalyst length region to make use of sunlight
surface but with an adsorbent such as acti- more efficiently. Anpo and coworkers
vated carbon mixed with the catalyst, this have used chromium (Cr) ion–implanted
O2
UV light NOx
(300–400 nm)
NO NO
e−
O2 −
NO2 NO2
O−
TiO2 OH Activated carbon
oxidation h+ HNO3 (adsorption)
H2O Rinse H2 O
Drainage
Neutralization with lime
Fig. 10 NOx removal by TiO2 photocatalyst (courtesy of Dr. Koji Takeuchi).

TiO2 to obtain enhanced response for adsorption period of one hour in the
decomposition in the visible region [81]. dark. After four hours of photocatalytic
Because photocatalytic reactions are not treatment under UV light, the chemical
thermochemical, it is not necessary to oxygen demand (COD) decreased by 95%,
supply heat. They proceed at temperatures the water turning clear and odorless.
available in the ambient environment. The advantage of photocatalysis in water
The required UV light intensity is not purification is the complete mineralization
more than about 0.1 mW cm−2 , which of organics caused by the photogeneration
corresponds to the intensity level of of • OH radicals originating from water
outdoor sunlight that can be found on via the OH groups of the TiO2 surface.
a cloudy winter day. Outdoor experiments In the case of gas-phase catalysis, water
have confirmed that photocatalytic effects vapor must be present to achieve com-
can be observed from sunrise to sunset. plete mineralization for most compounds.
Photocatalysts, however, are by no means However, a disadvantage with regard to the
a cure-all. Further efforts are necessary to photocatalytic purification of water is the
use fully their special characteristics. lower quantum efficiencies in comparison
to gas-phase photocatalysis. The principal
6.2.4 reason for this is the smaller availability of
Photocatalytic Water Purification oxygen in liquid water compared to air.
Several reports have appeared in the lit-
6.2.4.1 Wastewater Purification erature on the photocatalytic treatment
In a recent article, Herrmann has shown of industrial effluents such as textile
the feasibility of photocatalytic treatment effluents [82–84], rinse waters of pes-
of an actual highly loaded industrial ticide containers [85], wastewater from
wastewater (Fig. 11) [26]. In that exper- a phenolic resin factory [86], distillation
iment, the foul-smelling black-colored effluents from a pharmaceutical com-
waste was diluted 1000 times before treat- pany [86], waters that are produced dur-
ment. The treatment started after an ing recovery of natural gas and crude
1000
+UV
800
600
COD
400
200
Fig. 11 Kinetics of the decrease

of the COD (in mg kg−1 ) during 0
0 2 4 6
the photocatalytic treatment of
a real industrial wastewater Time
(diluted 1000 times) [26]. [h]
Solar or artificial UV light
Catalyst pane
Waste Treated
water Recycle if necessary water
Fig. 12 Flow sheet of the thin-film fixed-bed reactor [91].
oil [87], explosive-contaminated water [29, treatment [91]. Photocatalytic purification

88, 89], and oil-contaminated water [90]. of drinking water is also promising.
Various reactor configurations and their However, there are other questions related
applications for wastewater treatment have to equipment design, for example, how to
been summarized recently [29]. A compar- increase the contact area between the pho-
ative study made on different solar reactors tocatalyst and water and how to improve
indicated that the thin-film fixed-bed reac- the light-irradiation efficiency.
tor (Fig. 12) offers the advantages of a high
degree of degradation, lower-cost construc- 6.2.4.2 Drinking Water Purification
tion, and reduced operating costs [91]. Disinfection and decontamination of pub-
Although photocatalysis is attractive be- lic drinking water supplies is an extremely
cause of its ability to break down or- important environmental matter because
ganic compounds completely to CO2 and a large percentage of the world’s popula-
H2 O, that is, mineralization, industrial tion relies on these waters, especially in
applications of this technology for solar developing countries. As pointed out by
wastewater purification still involve some Matthews [92], one of the most promising
fundamental questions, such as low hy- applications of photocatalysis is for domes-
droxyl radical–production efficiency and tic photocatalytic water purification, which
slow kinetics. These two points are barriers would supply sufficient water for drink-
to marketing the technology. However, so- ing and cooking requirements. A 15-W
lar photocatalytic technology may become to 40-W photocatalytic water purifier that
practical for the purification of low-level could process 10–20 liters per day was
contaminants, similar to the case with in- estimated to be less expensive than an
door air purification. Freudenhammer and activated carbon filter. Application of tradi-
coworkers have indicated the suitability of tional methods to water treatment appears
photocatalysis for final-stage purification to be impractical for the low concentrations
of wastewater after treatment with low-cost of organic pollutants in drinking water.
conventional techniques such as biological The complete mineralization of many
Fig. 13 Drinking water Photocatalyst

treatment with TiO2
Water-circulation system
photocatalyst [25].
UV lamp
Water
types of organic compounds is possible coworkers demonstrated the rapid removal

by photocatalysis. Matthews has demon- of cyanobacterial microcystins in lake wa-
strated the photocatalytic degradation of ters [97]. However, humic substances in
several organic compounds that are typical drinking water could be removed only par-
water contaminants [93]. tially with the formation of fluorescent
Several research groups recently have intermediates [98]. Overall, the photocat-
demonstrated the feasibility of drinking alytic process seems to be very promising
water purification by photocatalysis. Pla- for drinking water purification, especially
tinized TiO2 was used to photocatalyze on the small scale required for domestic
the destruction of bromate ions in tap purposes (Fig. 13). Recently, the photocat-
water, at the ppb levels typically found alytic destruction of endocrine disruptors,
in ozonated water [94]. Chen and Jenq such as bisphenol A [99], and that of natu-
have examined the degradation of dis- ral and synthetic estrogens [100] has been
solved organic carbon in drinking water demonstrated in water.
sources [95]. Vidal has conducted pilot-
plant studies to examine the effectiveness 6.2.4.3 Remediation of Metal
of the solar photocatalytic process for Contamination
the destruction of low-level contaminants Precious metals are released into wa-
such as lindane (an organochlorine pes- ters by industrial processes such as hy-
ticide) in drinking water [96]. He has drometallurgy, plating, and photography.
conducted tests with the municipal wa- Silver contamination from photographic
ter supply of Madrid spiked with lindane processes, Cr from electroplating, and
and found that the lindane concentra- lead (Pb) from various biological and
tion was reduced from 200–500 µg L−1 industrial sources are typical examples.
to maximum permitted levels (0.1 µg L−1 ) Many of the metals are highly toxic, and
during a 30-min solar exposure. These accumulation of such metals in effluents
studies have indicated that inorganic sub- and industrial wastes is of great envi-
stances found in municipal water slightly ronmental concern. Photocatalysis again
reduced the oxidation rates expected from helps in either the transformation or depo-
related laboratory data. Shephard and sition of metals from aqueous solutions.
These metallic species can be recovered only metal ions with potentials more
from the photocatalyst by mechanical or positive than ∼0.4 V can be reduced on
chemical methods. TiO2 . Prairie and coworkers demonstrated
Dissolved metals in the form of ions the reduction of CrVI , AuIII , HgII , PdII ,
can affect the photocatalytic reaction rates PtIV , and AgI in deaerated conditions in
and efficiencies of organic degradation the presence of salicylic acid [106]. In-
reactions to a significant extent. Recently, significant metal deposition is observed for
Litter has published an extensive review platinum (Pt), silver (Ag), and gold (Au) in
on this subject [27]. Both increases [101, the absence of a hole scavenger [107]. With
102] and decreases [101, 103] in the pho- an alcohol as a hole scavenger, metal depo-
tocatalytic reaction rates were reported, sition may be an indirect reduction process
depending on the metal ion concentrations involving • ROH radicals produced from al-
in the aqueous solutions. At low metal cohol degradation by OH• radicals, rather
concentrations, trapping of photogener- than a direct capture of conduction band
ated electrons by the metal ions, with electrons. These are the two possibilities
resulting inhibition of electron-hole re- for the direct photocatalytic deposition of
combination, was proposed to be one of metals. A third, indirect pathway involves
the reasons for the enhanced reaction metal species such as Pb2+ , Mn2+ , and
rates [101, 102]. Photo-Fenton-type reac- Tl+ , with relatively negative redox poten-
tions may also take place in the presence tials, which may follow an oxidative route
of metal ions by reaction with photo- through hole attack because of the exis-
generated H2 O2 [104]. Photogeneration of tence of stable higher oxidation states [27].
H2 O2 caused by oxygen reduction has been Litter has summarized the pathways for
demonstrated [105]. the photocatalytic deposition of several
The detrimental effect on the oxidation metals [27].
rate of the presence of high metal
ion concentrations was attributed to the 6.2.5
oxidation of the reduced metal ions, Self-cleaning and Antifogging Surfaces
either by • OH radicals or directly by
photogenerated holes [104]. Other reasons In 1997, it was reported that TiO2 surfaces
include the filter effect as a result of can become highly hydrophilic (super-
absorption of UV light by the metal species hydrophilic) under UV illumination [108,
(e.g. FeII , CuII , and NiII ) and precipitation 109]. Although the mechanism for this
of dissolved metals as hydroxides on the effect still is not completely understood,
photocatalyst [103]. great progress has been made, and it
Photocatalytic removal of metals is an is clear that it involves reactions that
important application in its own right. are distinct from those involved in the
There are three pathways for metal re- photocatalytic decomposition of organic
moval, two direct and one indirect. Direct compounds [109–111]. In fact, the pos-
reduction of metal ions by conduction sibility that the conversion of the TiO2
band electrons is the simplest way. How- surface from hydrophobic to hydrophilic
ever, the redox potential of the metallic was due to the photocatalytic removal of
couple (Mn+ /M0 ) should be more positive a thin layer of organic contamination was
than the conduction band edge (Fig. 14). considered, but mounting evidence has
Experimental results have indicated that shown this not to be true. For example,
−1.0 Mn2+/Mn0
Zn2+/Zn0
~
~
e− Cd2+/Cd0
vb
−0.3
Ni2+/Ni0
Pb2+/Pb0
0.0
Cu2+/Cu0
Pt4+/Pt0
+0.5
hν
Ag+/Ag0
Hg2+/Hg0
+1.0 AuCl4−/Au0
Cr2O72−/Cr 3+
PbO2/Pb2+
+1.5
~
~
cb +2.9
h+
Fig. 14 Positions of the redox potentials of various metallic couples

related to the energy levels of the conduction and valence bands of
TiO2 [27].
the surface wettability, which increases The latter is thought to oxidize reduced
under relatively low intensity UV illumi- surface states (Ti3+ ) by • OH radicals or
nation over the span of about one hour hydrogen peroxide. The surface can then
(Fig. 15), can be converted back to the be reconverted to the hydrophilic state by
hydrophobic state by ultrasonic treatment. illumination [110].
35 Fig. 15 Water contact angle as

a function of time under UV
30 1.1 mW cm−2 illumination for a polycrystalline
Contact angle for water
TiO2 film on glass.

25
20
[°]
15
10
0
0 20 40 60 80 100 120
UV irradiation time
[min]
The same fundamental phenomenon, is a very appealing one and in a functional

that is, the great increase in wettability on sense is very closely related to that of the
UV illumination, can be used in two broad photocatalytic self-cleaning surface. How-
areas of applications, that is, self-cleaning ever, the mechanisms are different. What
surfaces and antifogging or antibeading makes this situation particularly interest-
surfaces. Both of these areas are treated ing is that any particular TiO2 surface
briefly. possesses both types of properties simul-
taneously. Here the focus is on superhy-
6.2.5.1 Self-cleaning Surfaces drophilic properties.
The concept of the superhydrophilic self- The mechanism for superhydrophilic
cleaning surface or easily washable surface self-cleaning principally involves the very
Water
Oil is spontaneously
removed from the
surface.
Time
Superhydro-
phylic surface
Water
Normal surface
Fig. 16 Series of photographs showing how an oil droplet adhering to a polycrystalline
TiO2 film on glass is gradually removed from the surface by water as a result of a
UV-induced increase in hydrophilicity [25].
Fig. 17 Photograph showing Roof

alternating photocatalytic (A) and
nonphotocatalytic (B) exterior wall tiles
that have been placed so that water
from a corrugated metal roof runs off
B A B A
onto the wall [25].
2-1 3-1 1-1
A B A B
1-2 1-2 3-2
B A B A
A: Superhydrophilic treated tile

B: Normal tile
Fig. 18 Photograph showing heavy tarpaulin material, noncoated on the left and TiO2 -coated
on the right, after exposure to urban air for several months.
strong adsorption of water on the It again should be stressed that it is

photochemically activated TiO2 surface. possible for the same TiO2 surface to sup-
Organic compounds that are attached or port both the photocatalytic decomposition
absorbed on the surface are simply lifted reactions and the superhydrophilic reac-
off due to the fact that water is more tions. The proportion of the two pathways
strongly adsorbed. For example, if an oil can be controlled through the variation of
droplet is attached to the surface, wa- the film composition, for example, by SiO2
ter simply begins to undercut it until it addition to increase the proportion of the
dislodges (Fig. 16). This effect can be ex- hydrophilic pathway [112–114]. For given
tremely useful for construction materials applications and ambient conditions, for
to be used in the urban environment be- example, light intensity, humidity, and so
cause pollutants can be easily washed off forth, it may be necessary to fine-tune the
by rainwater or with a stream of water [25, proportion of the two pathways.
112, 113]. For example, siding material
that has been treated with TiO2 is resis- 6.2.5.2 Antifogging, Antibeading Surfaces
tant to staining from water that runs off Because one of the critical attributes of
a roof (Fig. 17). Another example is the glass is its extreme visual clarity, there
heavy cloth material (tarpaulin, Fig. 18) is a host of possible applications for
that is being used for temporary storage glass surfaces that do not fog or on
structures or shelters (Fig. 19). The corre- which water does not bead. Some of
sponding conventional materials are quite these are listed in Table 2, together with
difficult to clean because of their flexible those that are related to hydrophilic self-
nature. cleaning surfaces. Water, in the form
Fig. 19 Photograph of a temporary storage structure made from TiO2 -treated tarpaulin
(courtesy of Taiyo Kogyo Corp.).
Tab. 2 Applications of superhydrophilic technology
Property Category Application
Roads Tunnel lighting, tunnel walls, traffic signs,

and sound-proof walls
Houses Tiles on kitchen walls and bathrooms,
exterior tiles, roofs, and windows
Self-cleaning Buildings Aluminum panels, tiles, building stone,
crystallized glass, and glass film
Agriculture Plastic and glass greenhouses
Electric and electronic Computer displays and cover glass for solar
environment cells
Vehicles Paint work, coatings for exterior surfaces of
windows, and headlights
Daily necessities and Tableware, kitchenware, and spray-on
consumer products antifouling coatings
Paint General-purpose paints and coatings
Roads Road mirrors
Houses Mirrors for bathrooms and dressers
Stores Refrigerated showcases
Antifogging property Electric and electronic Heat exchangers for air conditioners,
environment high-voltage electric transmission
equipment
Vehicles Inside surfaces of windows, glass films,
rear-view mirrors, and windshields
Daily necessities and Spray-on antifogging coatings and films
consumer products
Paint General-purpose paints and coatings
Optical instruments Optical lenses
Biocompatibility Medical instruments and Contact lenses and catheters
supplies
of droplets ranging from micrometer to clarity. The first commercial application

millimeter size, tends either to scatter light of this phenomenon has been for au-
or simply to reflect or refract it randomly. tomobile side-view mirrors. The TOTO
In both cases, visual clarity is impaired Company in Japan currently is supply-
drastically. ing TiO2 -treated glass for such mirrors to
In a serendipitous discovery, Watanabe major automobile manufacturers. In addi-
and coworkers found that a TiO2 -SiO2 sur- tion, it is marketing adhesive TiO2 –coated
face could become extremely hydrophilic plastic films for the conversion of existing
under UV illumination [114]. The result of mirrors.
this property is that water spreads evenly As the properties of superhydrophilic
across the surface (Fig. 20). If the amount films continue to be improved, there will
of water is relatively small, the water layer be many other possible applications of an-
becomes very thin and evaporates quickly. tifogging, antibeading glass. For example,
If the amount of water is larger, it forms one of the principal challenges is to de-
a sheetlike layer that also has high visual velop films that can be activated at longer
Fig. 20 Photograph of the

UV effect of UV illumination on
water droplets covering a sheet
of TiO2 -treated glass [108].
Dark
(a) (b)
wavelengths, principally by decreasing the UV irradiation. Regarding the develop-

magnitude of the band gap. Efforts along ment of photocatalytic water–purification
these lines were discussed briefly in the technology, some of the specific challenges
previous chapter. However, even with the are more complete mineralization, higher
band gap of unmodified anatase, the num- quantum yields, better regeneration of the
ber of possible applications is substantial, photocatalyst, and lower costs. Photocatal-
particularly those in which solar light can ysis may play a dominant role in the
be used, for example, for window glass purification of pretreated wastewater, and
for building exteriors or vehicles. These this may be more economical than treating
promise to be important applications. raw wastewater directly. In addition, do-
mestic photocatalytic water purifiers may
6.2.6 be an attractive new prospect for commer-
Conclusion cial development.
The ability of TiO2 surfaces to un-
Photocatalysis has become an increas- dergo photoinduced changes in wettability,
ingly promising technology for air and specifically, with large increases under
water purification. Although there are UV illumination, leads to a large num-
obstacles to overcome on the way to ber of possible applications, including
commercialization of this technology for self-cleaning construction materials, self-
water purification, commercial products cleaning glass, and antifogging, antibead-
for air purification already have appeared ing glass for windows and mirrors. These
on the market in Japan and recently applications are being commercialized at
in other countries. Photocatalytic sys-
present, and the range of applications
tems for the purification of room air
should continue to broaden as ways are
and car interiors already have been de-
found to enhance the ability to become
veloped and commercialized. TiO2 -coated
activated by visible light.
tiles have been installed in a number
of hospital operating rooms and have
successfully eliminated bacteria from the Acknowledgment
walls and from the air. TiO2 -coated con-
crete paving blocks have been installed The authors would like to express appre-
on busy roads in Osaka by the Mit- ciation to a number of coworkers who
subishi Company in Japan. The success have contributed heavily to this work, in-
of these developments is mainly because cluding Professors Kazuhito Hashimoto,
of the selection of environments where Toshiya Watanabe, and Rong Wang,
the target gas-phase pollutant levels are Dr. Yoshihisa Ohko, and Mr. Nobuyuki
low and thus the efficiency of TiO2 is Sakai. Dr. Bulusu V. Sarada helped in the
high under the low available levels of preparation of the figures. The authors
appreciate Dr. D.M. Blake for providing 16. A. L. Linsebigler, G. Lu, J. J. T. Yates, Chem.
copies of his comprehensive bibliography Rev. 1995, 95, 735–758.
17. K. Rajeshwar, J. Appl. Electrochem. 1995, 25,
on photocatalysis. 1067–1082.
18. Y. Parent, D. Blake, K. M. Bair et al., Solar
References Energy 1996, 56, 429–437.
19. U. Stafford, K. A. Gray, P. V. Kamat, Het-
1. D. M. Blake, in Bibliography of Work on erogen. Chem. Rev. 1996, 3, 77–104.
the Photocatalytic Removal of Hazardous 20. K. Vinodgopal, P. V. Kamat, CHEMTECH
Compounds from Water and Air, NREL/TP- 1996, 26, 18–22.
430-6084, National Renewable Energy Lab- 21. A. Fujishima, T. N. Rao, Proc. Indian Acad.
oratory, Golden, Colorado, 1994. Sci. (Chem. Sci.) 1997, 109, 471–486.
2. D. M. Blake, in Bibliography of Work on 22. A. Mills, S. LeHunte, J. Photochem. Photo-
biol. A: Chem. 1997, 108, 1–35.
the Photocatalytic Removal of Hazardous
23. J. Peral, X. Domenech, D. F. Ollis, J. Chem.
Compounds from Water and Air, NREL/TP-
Tech. Biotechnol. 1997, 70, 117–140.
473-20300, Update No. 1 (to June 1995),
24. N. Serpone, J. Photochem. Photobiol. A:
National Renewable Energy Laboratory,
Chem. 1997, 104, 1–12.
Golden, Colorado, 1995.
25. A. Fujishima, K. Hashimoto, T. Watanabe,
3. D. M. Blake, in Bibliography of Work on
in TiO2 Photocatalysis: Fundamentals and
the Photocatalytic Removal of Hazardous
Applications, BKC, Inc., Tokyo, Japan, 1999.
Compounds from Water and Air, NREL/TP-
26. J.-M. Hermann, Catal. Today 1999, 53,
430-22197, Update No. 2 (to October 1996),
115–129.
National Renewable Energy Laboratory,
27. M. I. Litter, Appl. Catal. B: Environ. 1999,
Golden, Colorado, 1996. 23, 89–114.
4. D. M. Blake, in Bibliography of Work on 28. M. Romero, J. Blanco, B. Sanchez et al.,
the Photocatalytic Removal of Hazardous Solar Energy 1999, 66, 169–182.
Compounds from Water and Air, NREL/TP- 29. O. M. Alfano, D. Bahnemann, A. E. Cassano
570-26797, Update No. 3 (to January 1999), et al., Catalysis Today 2000, 58, 199–230.
National Renewable Energy Laboratory, 30. A. Fujishima, T. N. Rao, D. A. Tryk, J. Pho-
Golden, Colorado, 1999, available at: tochem. Photobiol. C 2000, 1, 1–21.
http://www.ott.doe.gov/coolcar/pubs.html. 31. M. Tomkiewicz, Catal. Today 2000, 58,
5. D. M. Blake, P. Maness, Z. Huang et al., 115–123.
Sep. Purif. Methods. 1999, 28, 1–50. 32. D. A. Tryk, A. Fujishima, K. Honda, Elec-
6. M. A. Fox, M. T. Dulay, Chem. Rev. 1993, trochim. Acta 2000, 45, 2363–2376.
93, 341–357. 33. T. Matsunaga, R. Tomada, T. Nakajima
7. P. V. Kamat, Chem. Rev. 1993, 93, 267. et al., FEMS Microbiol. Lett. 1985, 29,
8. A. Mills, R. H. Davies, D. Worsley, Chem. 211–14.
Soc. Rev. 1993, 22, 417. 34. C. Wei, W.-Y. Lin, Z. Zainal et al., Environ
9. D. F. Ollis, H. Al-Ekabi, Eds. in Photocat- Sci. Technol. 1994, 28, 934.
alytic Purification and Treatment of Water 35. T. Matsunaga, M. Okochi, Environ. Sci.
and Air, Elsevier, Amsterdam, 1993. Technol. 1995, 29, 501–505.
10. H. Yoneyama, Crit. Rev. Solid State 1993, 36. R. J. Watts, S. Kong, M. P. Orr et al., Water
18, 69–111. Res. 1995, 29, 95–100.
11. W. A. Zeltner, C. G. Hill, M. A. Anderson, 37. M. Bekbolet, C. V. Araz, Chemosphere 1996,
CHEMTECH 1993, 23, 21–28. 32, 959–965.
12. N. Serpone, Res. Chem. Intermed. 1994, 20, 38. M. Bekbolet, Water Sci. Technol. 1997, 35,
953–992. 95–100.
13. K. I. Zamaraev, M. I. Khramov, V. N. 39. Y. Kikuchi, K. Sunada, T. Iyoda et al., J.
Parmon, Catal. Rev. 1994, 36, 617. Photochem. Photobiol. A: Chem. 1997, 106,
14. A. Heller, Acc. Chem. Res. 1995, 28, 51–56.
503–508. 40. W. A. Jacoby, B. C. Maness, E. J. Wolfrum
15. M. R. Hoffmann, S. T. Martin, W. Choi et al., Environ. Sci. Technol. 1998, 32,
et al., Chem. Rev. 1995, 95, 69–96. 2650–2653.
41. K. Sunada, Y. Kikuchi, K. Hashimoto et al., 62. M. L. Sauer, M. A. Hale, D. F. Ollis, J. Pho-
Environ. Sci. Technol. 1998, 32, 726–728. tochem. Photobiol. A: Chem. 1995, 88,
42. P.-C. Maness, S. Smolinski, D. M. Blake 169–178.
et al., Appl. Environ. Microbiol. 1999, 65, 63. R. M. Alberici, W. F. Jardim, Appl. Catal. B:
4094–4098. Environ. 1997, 14, 55–68.
43. A. Vidal, A. I. Diaz, A. E. Hraiki et al., Catal. 64. J. Stark, J. Rabani, J. Phys. Chem. B 1999,
Today 1999, 54, 283–290. 103, 8524–8531.
44. Z. Huang, P.-C. Maness, D. M. Blake et al., 65. W. A. Jacoby, M. R. Nimlos, D. M. Blake
J. Photochem. Photobiol. A: Chem. 2000, 130, et al., Environ. Sci. Technol. 1994, 28,
163–172. 1661–1668.
45. I. Sopyan, M. Watanabe, S. Murasawa et al., 66. Y. Luo, D. F. Ollis, J. Catal. 1996, 163, 1–11.
J. Photochem. Photobiol. A: Chem. 1996, 98, 67. O. Hennezel, D. F. Ollis, J. Catal. 1997, 167,
79–86. 118–126.
46. Y. Ohko, K. Hashimoto, A. Fujishima, J. 68. K. Wada, K. Yoshida, Y. Watanabe et al.,
Phys. Chem. A 1997, 101, 8057–8062. J. Chem. Soc., Faraday Trans. 1996, 92,
47. T. Watanabe, A. Kitamura, E. Kojima et al., 685–691.
U.S. Patent 5,874,701, February 23, 1999, in 69. S. A. Larson, J. A. Widegren, J. L. Falconer,
U. S. Patent & Trademark Office (Patent Full J. Catal. 1995, 157, 611–625.
Text and Image Database), Toto Co., Ltd., 70. N. Takeda, M. Ohtani, T. Torimoto et al., J.
USA, 1999. Phys. Chem. B 1997, 101, 2644–2649.
48. T. Watanabe, A. Kitamura, E. Kojima et al., 71. T. N. Obee, R. T. Brown, Environ. Sci. Tech-
U.S. Patent 6,139,803, October 31, 2000, in nol. 1995, 29, 1223–1231.
U. S. Patent & Trademark Office (Patent Full 72. N. N. Lichtin, M. Sedeghi, J. Photochem.
Photobiol. A: Chem. 1998, 113, 81–88.
Text and Image Database), Toto Co., Ltd.,
73. P. Kopf, E. Gilbert, S. H. Eberle, J. Pho-
USA, 2000.
tochem. Photobiol. A: Chem. 2000, 136,
49. K. Sunada, Y. Kikuchi, K. Hashimoto et al.,
163–168.
Environ. Sci. Technol. 1998, 32, 726–728.
74. J. Peral, D. F. Ollis, J. Catal. 1992, 136,
50. A. Fujishima, J. Ootsuki, T. Yamashita
554–565.
et al., Photomed. Photobiol. 1986, 8, 45–46.
75. S. Yamazaki, S. Tanaka, H. Tsukamoto, J.
51. R. Cai, K. Hashimoto, K. Itoh et al., Bull.
Photochem. Photobiol. A: Chem. 1999, 121,
Chem. Soc. Jpn. 1991, 64, 1268–1273. 55–61.
52. R. Cai, K. Hashimoto, Y. Kubota et al., 76. A. V. Vorontsov, E. N. Savinov, P. G.
Chem. Lett. 1992, 427–430. Smirniotis, Chem. Eng. Sci. 2000, 55,
53. R. Cai, H. Sakai, K. Hashimoto, Y. Kubota 5089–5098.
et al., Denki Kagaku 1992, 60, 314–321. 77. S. Sitkiewitz, A. Heller, Nouv. J. Chim. 1996,
54. H. Sakai, R. Baba, K. Hashimoto et al., 20, 233–241.
Chem. Lett. 1995, 185–186. 78. T. Minabe, D. A. Tryk, P. Sawunyama et al.,
55. R. Cai, Y. Kubota, T. Shuin et al., Cancer J. Photochem. Photobiol. A: Chem. 2000, 137,
Res. 1992, 52, 2346–2348. 53–62.
56. A. Fujishima, R. X. Cai, J. Otsuki et al., Elec- 79. T. Ibusuki, K. Takeuchi, J. Mol. Catal. 1994,
trochim. Acta 1993, 38, 153–157. 88, 93–102.
57. N. Huang, M. Xu, C. Yuan et al., J. Pho- 80. I. Nakamura, N. Negishi, S. Kutsuna et al.,
tochem. Photobiol. A: Chem 1997, 108, J. Mol. Catal. A: Chem. 2000, 162, 205–212.
229–233. 81. M. Anpo, Y. Ichihashi, M. Takeuchi et al.,
58. I. Sopyan, S. Murasawa, K. Hashimoto Res. Chem. Intermed. 1998, 24, 143–149.
et al., Chem. Lett. 1994, 723–726. 82. H. Q. Zhan, H. Tian, Dye Pigment. 1998, 37,
59. T. Noguchi, A. Fujishima, P. Sawunyama 231–239.
et al., Environ. Sci. Technol. 1998, 32, 83. S. G. D. Moraes, R. S. Freire, N. Duran,
3831–3833. Chemosphere 2000, 40, 369–373.
60. Y. Ohko, D. A. Tryk, K. Hashimoto et al., J. 84. K. Tanaka, K. Padermpole, T. Hisanaga,
Phys. Chem. B 1998, 102, 2699–2704. Wat. Res. 2000, 34, 327–333.
61. H. Matsubara, M. Takada, S. Koyama et al., 85. S. Malato, J. Blanco, C. Richter et al., Appl.
Chem. Lett. 1995, 767–768. Catal. B: Environ. 2000, 25, 31–38.
86. S. M. Rodriguez, C. Richter, J. B. Galvez 102. E. Pelizzetti, M. Borgarello, E. Borgarello

et al., Solar Energy 1996, 56, 401–410. et al., Chemosphere 1988, 17, 499–510.
87. E. Bessa, G. L. Sant’ Anna, M. Dezotti, J. 103. A. Sclafani, L. Palmisano, E. Davi, J. Pho-
Adv. Oxid. Technol. 1999, 4, 196–202. tochem. Photobiol. A: Chem. 1991, 56,
88. N. Z. Muradov, Solar Energy 1994, 52, 113–123.
283–288. 104. J. Sykora, Coord. Chem. Rev. 1997, 159,
89. R. Alnaizy, A. Akgerman, Water Res. 1999, 95–108.
33, 2021–2030. 105. H. Sakai, R. Baba, K. Hashimoto et al., J.
90. M. Hamerski, J. Grzechulska, A. W. Phys. Chem. 1995, 99, 11896–11900.
Morawski, Solar Energy 1999, 66, 395–399. 106. M. R. Prairie, L. R. Evens, B. M. Stange
91. H. Freudenhammer, D. Bahnemann, et al., Environ. Sci. Technol. 1993, 27,
L. Bousselmi et al., Wat. Sci. Technol. 1997, 1776–1782.
35, 149–156. 107. R. Baba, R. Honda, A. Fujishima et al.,
92. R. W. Matthews, in Photocatalytic Purifica- Chem. Lett. 1986, 1307.
tion and Treatment of Water and Air (Eds.: 108. R. Wang, K. Hashimoto, A. Fujishima et al.,
D. F. Ollis, H. Al-Ekabi), Elsevier, Amster- Nature 1997, 388, 431–432.
dam, 1993, pp. 121–138. 109. R. Wang, K. Hashimoto, A. Fujishima et al.,
93. R. W. Matthews, Water Res. 1986, 20, Adv. Mater. 1998, 10, 135–138.
569–578. 110. N. Sakai, R. Wang, A. Fujishima et al.,
94. A. Mills, A. Belghazi, D. Rodman, Water Langmuir 1998, 14, 5918–5920.
Res. 1996, 30, 1973–1978. 111. R. Wang, N. Sakai, A. Fujishima et al., J.
95. P. H. Chen, C. H. Jenq, Environ. Internat. Phys. Chem. B 1999, 103, 2188–2194.
1998, 24, 871–879. 112. M. Chikuni, M. Hayakawa, T. Watanabe
96. A. Vidal, Chemosphere 1998, 37, 387. et al., U. S. Patent 5,755,867, May 26, 1998,
97. G. S. Shephard, S. Stockenstrom, D. D. in U. S. Patent & Trademark Office (Patent
Villiers et al., Toxicon 1998, 36, 1895–1901. Full Text and Image Database), Shin-Etsu
98. A. R. Eggins, F. L. Palmer, J. A. Byrne, Wa- Chemical Co., Ltd. and Toto Ltd., USA,
ter Res. 1997, 31, 1223–1226. 1998.
99. Y. Ohko, I. Ando, C. Niwa et al., Environ. 113. M. Hayakawa, E. Kojima, K. Norimoto et al.,
Sci. Technol. 2001, 35, 2365–2368. U. S. Patent 6,013,372, January 11, 2000, in
100. Y. Kubota, C. Niwa, T. Iguchi et al., in U. S. Patent & Trademark Office (Patent Full
preparation. Text and Image Database), Toto, Ltd., USA,
101. K. Okamoto, Y. Yamamoto, H. Tanaka 2000.
et al., Bull. Chem. Soc. Jpn. 1985, 58, 114. A. Fujishima, D. A. Tryk, T. Watanabe et al.,
2023–2028. Internat. Glass Rev. 1998, 114–116.
6.3 further divided into three categories: area-

Silverless Photography – selective deposition onto semiconductor
Optical Image Recordings by Means electrodes, area-selective electrochemical
of Photoelectrochemical Processes reactions in the surface layer of semicon-
ductor electrodes, and area-selective reac-
Hiroshi Yoneyama
tions in electrochromic thin-film-covered
Anan National College of Technology,
Tokushima, Japan
semiconductor electrodes. The latter may
be classified into two groups: area-selective
6.3.1 reactions of electrochromic films contain-
Introduction ing dispersed semiconductor particles as
the photosensitizer and area-selective reac-
The term Silverless photography is used tions of dye films with use of a molecular
for photoimage-formation processes using photosensitizer.
a variety of photochemical reactions in-
cluding those using photoelectrochemical 6.3.2
systems [1]. Silverless photography pro- Area-selective Photoinduced Deposition
vides a photochemical means of image on Semiconductor Electrodes
processing [2]. Light-sensitive chelate coor-
dination compounds containing divalent 6.3.2.1 Deposition on Semiconductor
elements are often used for such pur- Electrodes with External Bias
poses [3]. For example, the photochemical The most significant feature of the semi-
cleavage of polymeric films of coordi- conductor electrode is that photosensitized
nation compounds such as [FeII (CN)6 - electrochemical reactions that are induced
PtIV (NH3 )4 ]n , followed by reactions of to occur using photogenerated minority
the resulting complexes with 3D tran- charge carriers take place only on the ir-
sition metal cations in solution, affords radiated electrode surfaces. Therefore, if
image formation in the polymeric films [4]. n-type semiconductor electrodes are irra-
Photoisomerization of molecules such diated under anodic bias, photosensitized
as azobenzene derivatives [5] and spiro- oxidation reactions are induced to take
oxazine dyes [6] are also useful in optical place at the irradiated electrode surfaces
image recordings. Furthermore, the use by the photogenerated valence band holes
of photochromic molecules with a combi- (as described in earlier chapters). Simi-
nation of liquid crystals can provide the larly, if p-type semiconductor electrodes
optical image storage [7, 8]. This article, are irradiated under cathodic bias, photo-
however, is restricted to electrochemical sensitized reduction reactions are induced
systems in which semiconductive materi- to take place on the irradiated electrode
als and/or redox dye molecules are used as surfaces with the use of photogener-
the photosensitizers. ated conduction band electrons. Oxidative
Photoelectrochemical imaging systems electrochemical deposition reactions are
may be roughly classified into two classes; then used in photoimage formations on
one is concerned with photoinduced elec- n-type semiconductor electrodes and elec-
trochemical reaction systems using var- trochemical reductive depositions are used
ious electrode configurations and the on p-type semiconductor electrodes. If
other with light-induced heterogeneous light scattering at the irradiated semicon-
reaction systems. The former may be ductor electrode surfaces were ignored, the
6.3 Silverless Photography – Optical Image Recordings by Means of Photoelectrochemical Processes 563
highest resolution of the area-selective de- were also reported [13]. The produced
position would be twice as large as the images are easily erased by applying the
mean diffusion length of the photogener- reverse bias to the electrodes in the dark.
ated minority charge carriers. Area-selective photooxidative deposition of
Area-selective photoinduced reductive polypyrrole from a pyrrole solution onto
depositions of copper (Cu), nickel (Ni), TiO2 was used to produce a conducting
palladium (Pd), and gold (Au) on cathod- polymer pattern [14, 15]. The photoelectro-
ically biased p-Si and p-GaAs [9] from chemical oxidation of water-soluble aro-
aqueous solutions containing the corre- matic compounds such as O-toluidine on
sponding metal ions were shown to take TiO2 electrodes resulted in hydrophobic
place at a resolution of less than 10 µm. deposits on the irradiated surface only,
Area-selective reductive deposition of hep- which was developed by applying oily
tylviologen radical cation salt (HV+• Br− ) color ink [16]. Photo-patterning was also
from a heptylviologen bromide solution reported with self-assembled monolayer
onto p-GaAs electrode under a cathodic (SAM) of bis[11-[(4-azidobenzoyl)oxy]-1-
bias was useful in photoinduced elec- undecyl]disulfide [17]. Irradiation of the
trochromic displays [10]. Optical images SAM-covered Au electrode through a
due to the deposited HV+. Br− were easily photomask in the presence of several
erased by applying the reverse bias (i.e. amines such as HNEt2 , HN(n-Bu)2 or
anodic bias) to the electrode. The area- HN(CH2 CH2 OH)2 resulted in the attach-
selective deposition was also reported for ments of the amine, and the amine-
photooxidative deposition of PbO2 from attached regions showed little activity
a Pb2+ solution onto n-type ZnO, TiO2 , for electrochemical deposition of polyani-
and SrTiO3 [11], and of Tl2 O3 on these line from an acidic aniline solution,
semiconductor electrodes from a Tl+ so- giving selective deposition of polyani-
lution [12]. Using these photodeposition line on the nonirradiated SAM surfaces
reactions, the photoelectrochemical imag- on Au. The area-selectivity in the pho-
ing was successfully achieved at the an- toelectrochemical reactions finds another
odically biased semiconductor electrodes. application – the micromachining of ZnO
One example of the photoimages produced surfaces by anodic photoetching [18]. Mi-
is given in Fig. 1. The deposited Tl2 O3 do cropatterns having 130-µm line width
not grow beyond the irradiated part of were successfully produced using this
the electrodes because the photoinduced technique.
oxidation reaction took place on the irradi-
ated surfaces only. For example, the rate of 6.3.2.2 Deposition on Semiconductor
the deposition reaction of Tl2 O3 is propor- Surfaces with no External Bias
tional to the irradiation intensities at the When a photooxidative deposition occurs
electrode surfaces. Here, irradiation of the on semiconductor surfaces without any
n-type semiconductor electrodes through applications of external bias, a reduction
a negative film of photography resulted in reaction takes place simultaneously as the
the deposition of the brown oxide of vari- counterpart reaction (see Chapter 8). Ir-
ous contrasts in reflection to the images of radiation of p-Si wafers in a zinc (Zn)
the negative film, allowing the formation or cadmium (Cd) salt solution resulted
of a portrait. Attempts to form photoim- in area-selective deposition of metals on
ages on n-Si using the Tl2 O3 deposition the irradiated part if the back surface of
Fig. 1 Image recorded by using photoinduced deposition of metallic oxides on

n-type semiconductor electrodes. This example shows Tl2 O3 deposition on ZnO
electrode from an aqueous solution containing a Tl(I) salt [12]. (With permission by
the Electrochemical Society, Inc.)
the wafers was coated by metallic films anodic dissolution of the coated metal in
of Zn or Cd [19]. The photoinduced re- the back surface. The metal film on the
action in that case is thought to consist back surface was useful in enhancing the
of the photoinduced cathodic deposition band bending with which the rate of the
of the metal in the irradiated surface and transfer of photogenerated electrons to the
solution phase was enhanced. Irradiation similar technique, polythiophene patterns

of TiO2 single crystal plates in a solution of high resolution were produced on n-Si
containing a Pd salt caused selective de- wafers [23] from a solution containing α-
position of Pd at surface flaws [20]. The terthienyl and Ag+ ions. Irradiation of a
surface flaws seem to provide electron- TiO2 -adhered alumina plate in an electro-
transfer channels from the TiO2 to the less Ni plating solution containing citric
electrolyte solution. In that case, the oxi- acid as the hole scavenger resulted in area-
dation reaction as the counterpart reaction selective deposition of Ni on the irradiated
is thought to be photoinduced oxidation of parts [24]. The mechanism for this event
water on the flaw-free irradiated surfaces. would be such that electrons and holes are
Such reaction selectivity of semiconduc- generated in the irradiated surface, and
tor surfaces was confirmed by observing the positive holes are quickly captured by
which part of the semiconductor surfaces, the hole scavenger, leaving the electrons at
flaws in the surfaces, or flaw-free well- the irradiated surface; then the reductive
etched surfaces is more favored for oxida- deposition of the metals takes place at the
tive or reductive deposition reactions [21]. irradiated surface.
There seems to be a general trend that
6.3.3
when n-type semiconductor such as TiO2
Image Formations Using Photoinduced
is used, photoinduced oxidation reactions Electrochromism
take place on the flaw-free well-etched sur-
faces, whereas reduction reactions occur 6.3.3.1 Electrochromic Materials
on the flaws especially in the dark. Sim- Electrochromic materials are often used
ilarly, if p-type semiconductors are used, as photoimage-recording materials. How-
photoinduced reduction reactions occur on ever, the process for making photoimages
the well-etched flaw-free surfaces, whereas in the electrochromic materials is different
oxidation reactions occur on the flaws espe- depending on whether the electrochromic
cially in the dark. Therefore, it may be said materials have semiconductivity or not.
that electrochemical reactions in which the The electrochromic materials having semi-
photogenerated minority charge carriers conductivity work both as the photosen-
are involved, take place on the irradiated sitizer and as the substrates in which
defect-free surfaces and the counterpart the photoimages are produced, whereas
reactions that are induced to occur by the those having metallic or insulating prop-
majority charge carriers take place on the erties need another photosensitizer to pro-
defects in the nonirradiated surfaces. duce photoimages in the electrochromic
Using such reaction selectivity of the substrates.
semiconductor surfaces, line-shaped pat- Table 1 gives popular electrochromic
terns of polypyrrole were successfully materials useful in photoinduced image
produced on well-etched n-type Si wafers, formations [25–28]. The list includes sev-
which had mechanically damaged back eral semiconductor electrochromic oxides
surfaces, by scanning He-Ne laser beam such as MoO3 , WO3 , and Nb2 O5 .
on the front surfaces in aqueous solution
containing pyrrole and Ag+ ions [22]. The 6.3.3.2 Photoinduced Electrochromism
deposition of silver (Ag) took place selec- of Semiconductive Oxides
tively on the whole damaged surfaces in the WO3 , MoO3 , and Nb2 O5 have n-type
dark as the counterpart reaction. Using a semiconductivity and are colored when
Tab. 1 Electrochromic materials are involved in hydrogen tungsten bronze

formation. Because of the generation of
Material Color photogenerated positive holes and because
oxidized reduced the production of protons are restricted
to the irradiated surface only, the color
WO3 Trans.a Blue change occurs only at the irradiated sur-
IrOx Trans. Blue black faces. The colored images on the oxide
phosphotungstic White Blue films are easily erased by anodically polar-
acid
izing the oxide films in the image-forming
MoO3 Yellow Blue
Rh2 O3 Yellow Green solution.
V2 O5 Yellow Black Similar photoinduced electrochromism
Nb2 O5 Trans. Dark blue has been published on MoO3 films pre-
NiOx Trans. Dark bronze pared by electrodeposition [35] and vac-
CoOx Red purple Gray black
uum evaporation [36–39]. MoO3 needs
Prussian Blue White
Lutechium Green Red violet irradiation with near ultraviolet (UV) light
diphtalocyanine for its photoexcitation. However, it was
Heptylviologen Trans. Blue purple reported that a MoO3 film prepared by
Benzylviologen Trans. Blue vacuum evaporation has visible light sen-
Tetrathiafulvalene Yellow Purple
sitivity by cathodic prepolarization in a
Polypyrrole Brown black Yellow
Polyaniline Green Yellow propylene carbonate solution containing
Polythiophene Green red LiClO4 [36, 37, 39]. The cathodic prepo-
larization may form visible light-sensitive
a trans.: transparent. Lix MoO3 in the MoO3 lattice. Both UV-
and visible light–induced electrochromic
response are enhanced by coating the ox-
reduced. The photoinduced coloring of a ide surface with a thin porous Au or
WO3 film without any applied bias in air at- platinum (Pt) layer (20 nm) [39]. The for-
mosphere has been known since 1979 [25]. mation of a large Schottky barrier at the
WO3 particles suspended in water or in coated metal–MoO3 interface may assist
formic acid solution changes in color to in the separation of photogenerated posi-
blue with photon absorption [29]. The col- tive holes and electrons, resulting in the
oring occurs as a result of reduction of enhancement of the coloring of the ox-
WO3 to hydrogen tungsten bronze. The ide. The photoinduced coloration of MoO3
photooxidation of water or formic acid was found to be enhanced by mixing with
occurs simultaneously as the counterpart WO3 , and the highest enhancement was
reaction. The same is true for a WO3 film achieved on a film having 92% WO3 and
and WO3 -TiO2 composite film [30]. De- 8% MoO3 [40]. As in the case of WO3 ,
tailed studies of the photoinduced color the images formed in the MoO3 sur-
change of WO3 films caused by irradiation faces are erased by anodically polarizing
were carried out mostly in air atmosphere the oxide electrodes in the image-forming
containing aliphatic alcohol vapors such baths. A Nb2 O5 film prepared by vacuum
as ethanol and propanol as the hole scav- evaporation showed similar photoinduced
enger [31–34]. When the hole scavengers electrochromism if ethanol vapor was
are oxidized by the photogenerated pos- used as a hole scavenger in nitrogen
itive holes, protons are released, which atmosphere [41].
6.3.3.3 Combined Systems of film had very poor resolution. With an in-
Electrochromic Materials and crease in the irradiation time, the produced
Semiconductor Electrodes images easily spread beyond the irradiated
Attempts were made to form area-selective area because of high electrical conductivity
images on electrochromic thin films of polypyrrole.
coated on semiconductor electrodes. The Polyacrylamide gel films containing
first reports on this category were on Fe(III) dipyridinium complex as an elec-
Prussian blue film-coated TiO2 [42, 43] trochromic component were useful as
electrodes. Later, similar systems were the substrate for the formation of pho-
constructed using SrTiO3 films as the photographic images when coated on a CdSe
tosensitive substrate [44, 45]. The coating film [48]. Poly-Re(CO)3 (vbpy)Cl-coated n-
with Prussian blue films on semiconductor MoSe2 also allowed photoinduced elec-
electrodes was carried out by electrochemi- trochromic recordings in the coated poly-
cal synthesis of the Prussian blue films us- mer layer under anodic bias [49]. The
ing the semiconductor as the electrode. To images produced are easily erased by
produce area-selective photoimages, the applying the reverse bias to the polymer-
entire Prussian blue-coated film was first coated semiconductor electrodes.
reduced to Prussian white. The irradiation
was then carried out with a xenon lamp 6.3.3.4 Composites Made of
through a photomask under anodic bias to Electrochromic Materials and
allow area-selective photooxidation of the Semiconductor Powders
coated film, resulting in blue images in the Photoimages of high resolution were suc-
white film. The same principle was applied cessfully produced in polyaniline films
to poly(N -methylpyrrole) film–coated n- containing dispersed WO3 [50, 52], and
Si [46] and polypyrrole film–coated Si TiO2 [51, 53, 54] powder particles in the
electrodes of both n-type and p-type con- films. The polyaniline/semiconductor par-
ductivity [47]. These conducting films were ticles composite films were synthesized
deposited on the Si electrode surfaces by by electrochemical polymerization of ani-
electrochemical polymerization of the cor- line in aqueous acidic solution in the
responding monomer using the Si as the presence of the suspended semiconductor
electrode. When n-type Si electrodes are particles. The prepared composite films
used, the entire polypyrrole-coated film showed the color of polyaniline even if a
is first reduced by cathodic polarization high amount of the oxide particles was
to give yellow, and then the polypyrrole- used because WO3 and TiO2 are color-
coated Si electrodes are anodically po- less. One example of a photoimage formed
larized and irradiated through a mask. in a polyaniline/TiO2 composite film is
The area-selective photooxidation of the presented in Fig. 2. As shown in the
polypyrrole films is then induced to take figure, a portrait of high resolution was
place to give brown-black images in the obtained. Complex electrochemical reac-
irradiated surfaces. In the case of using p- tions are involved in the formation of such
type Si electrodes, the as-deposited brown good photoimages. Figure 3 shows the
black polypyrrole films changed in color reaction scheme for the photoimage for-
to yellow at the irradiated part under ca- mation in the composite film, derived from
thodic bias. However, it was found that the analyses of changes in absorption spec-
images produced in the coated polypyrrole tra of polyaniline films obtained during
Fig. 2 Photoimage formed on a composite film of polyaniline and

dispersed TiO2 particles by projecting the positive imaged on the film
that was immersed in 0.5 mol dm−3 phosphate buffer (pH 7) containing
20 wt% methanol. The illumination was carried out with a 55 Xe lamp
for 1 minute [49]. (With permission by the Royal Society of Chemistry).
A− A−
N N N+ N+ (B)
H H H H
Immersion in neutral aqueous solution

or −H + −A−
anodic polarization in nonaqueous solution
N N N N (A)
H H
+H + +A−
e−
A− A−
hν N N N+ N+ (B)
H H H H
CH3OH
h+ +e− −A−
HCHO + 2H +
N N N N (C)
H H H H
Fig. 3 Schematic illustration of compositional changes in polyaniline caused by
photoreduction. (A) deprotonated blue film, (B) as-grown and/or green film, and
(C) reduced yellow film. A− denotes a monovalent electrolyte anion [51]. (With
permission by American Chemical Society).
the course of the image formation [53] the electrochemically inactive insulating
and from studies using a quartz crystal polyaniline is changed into electrochemi-
microbalance [54]. When the composite cally active one at the irradiated parts only.
films are prepared, they exhibited vivid As shown in Fig. 3 for the reaction scheme,
green, which is characteristic of highly if the blue composite film is irradiated,
conducting polyaniline having high elec- electrons and positive holes are generated
trochemical activities. Polymer films of in the TiO2 particles in the film. The
high electrical conductivity and of high generated positive holes cause oxidation
electrochemical reactivity do not allow the of methanol at the irradiated TiO2 sur-
formation of photoimage of high reso- faces, resulting in release of protons there.
lution, as already described earlier for The produced protons are instantly picked
polypyrrole-coated Si electrodes. To pro- up by the deprotonated polyaniline, and
duce photoimage of high resolution, the the electrochemically inactive blue polyani-
electrical conductivity of polyaniline of the
line is then converted to electrochemically
composites has to be changed to be as
active green polyaniline, which is then re-
low as possible. Fortunately, polyaniline
duced by the photogenerated electrons in
films of high electrical conductivity are
the TiO2 to form yellow images. Because
easily changed into electrically insulating
the production of protons and their at-
ones either by anodically polarizing in
tachments to the deprotonated polyaniline
alcohol solutions such as methanol and
ethanol or in aqueous neutral solutions
or by immersion in concentrated alkaline
solutions [55]. With the use of either one Writing
of these procedures, the deprotonation oc- ITO Nafion PB TiO2
curs at polyaniline in an oxidized state, and
the color changes from vivid green to blue. ca. −0.4 V
The polyaniline/TiO2 composite films hav- 0.25 V
ing an electrically insulating blue film pre-
pared in such a manner is then immersed
Light
in ethanol solution, and irradiated under
MeOH
no applied bias through a photomask with
lights whose energy is high enough to
photoexcite the incorporated TiO2 . Yellow
images, which are characteristic of the
reduced polyaniline, are then formed in
Erasing
the blue films with high resolution. An
important phenomenon inherent in the
photoimage formation-processes is that
Fig. 4 Energy level diagram for electron-

transfer events at the bilayer electrode
consisting of a Nafion film containing
ferrocenylmethyltrimethylamnnonium ions
and a Prussian-blue film containing 14 wt%
TiO2 particles [55]. (With permission by the
Electrochemical Society, Inc.).
occurs only in the irradiated area, image of CdS in the strong acidic medium
formations of high resolution take place at in which the composite is prepared,
the irradiated part only. results obtained were not so good as
The same idea was used in the image those obtained with the polyaniline/TiO2
formation in polyaniline/CdS composite composites. The composite films of
films [56], but because of the poor stability Prussian blue and TiO2 fine particles
(a)
(b)
1
e−
E(3Ru(bpy)32+/Ru(bpy)33+)
hν
0 e−
E(MB/l −MB)
E vs. SCE
E Oxascrobic acid
[V]
1 E(Ru(bpy)32+/Ru(bpy)33+)
hν
E(3MB/l −MB)
2
Fig. 5 (a) Photoimage formed in polyaniline film containing methylene blue-

and Ru(bpy)3 2+ -bound Nafion by illumination with a 500-W Xenon lamp for
30 seconds in 1 mol dm−3 HCl containing 0.1 mol dm−3 ascorbic acid, and
(b) Energetic correlation of photoexcited methylene blue and Ru(bpy)3 2+ to
ascorbic acid used as an electron donor. (With permission by the
Electrochemical Society, Inc.).
prepared on ITO electrodes also showed constructed using a composite made of a

good properties for photoimage forma- cation exchange polymer and polyethylene
tions with high resolution [57]. However, glycol as a solid electrolyte [60].
it was found that the photoimages pro-
duced grew beyond the irradiated area 6.3.4
with increasing the irradiation time. The Summary
electron transport from the produced
Prussian white to the surrounding Prus- So far, various optical image recordings
with the use of photoelectrochemical
sian blue through the conducting ITO
reactions have been published, as de-
was found to be responsible. This seri-
scribed in the preceding sections. Many
ous problem was solved by inserting a
systems use a couple of a photosensitizer
thin Nafion film containing ferrocenyl-
and an electrochromic material. The key
methyltrimethylammonium (FA) as an
for construction of recording systems in
electron relay between the ITO substrate
these cases is to appropriately choose the
and the Prussian blue/TiO2 composite
photosensitizer and the electrochromic
film. The energy scheme for the image
material to obtain good photoimages of
formation is given in Fig. 4.
high resolution. Because there are a lot
6.3.3.5 Image Formation with the Use of a of electrochromic materials and photosen-
Dye as the Photosensitizer sitizers available for photoinduced elec-
Photoimages could be formed if the trochromic reactions, the development of
semiconductor powder dispersed in the novel recording systems with the use of a
electrochromic film were replaced by a re- variety of combinations would still be of
dox dye. The utility of the idea was demon- active concern.
strated by immobilizing in electrically
conductive polyaniline film, with redox ac- References
tive methylene blue used as an image-
forming molecule and Ru(bpy)3 3+/4+ as 1. Yu. A. Cherkasov, J. Opt. Technol. 1997, 64,
872–879.
a photosensitizer [58]. If aniline is electro- 2. M. R. V. Sahyun, J. Chem. Educ. 1973, 50,
chemically polymerized in the presence 88–93.
of methylene blue and Nafion, methy- 3. O. V. Mikhailov, J. Coord. Chem. 1999, 47,
lene blue–bound Nafion is incorporated 31–58.
4. Y. Wu, B. W. Pfenning, A. Bocarsly et al.,
in the resulting polyaniline films, and the
Inorg. Chem. 1995, 34, 4262–4267.
film exhibits electrochromic properties as 5. Z.-F. Liu, B. H. Loo, R. Baba et al., Chem.
a result of the redox reaction of methy- Lett. 1990, 1023–1026.
lene blue [59]. The polyaniline worked 6. W. Du, H. Y. Tam, Optics Commun. 1996,
as a conducting matrix in that case. A 126, 223–229.
7. T. Ikeda, O. Tsutumi, T. Sasaki, Synth. Met.
photosensitive film with an ability for
1996, 81, 289–296.
recording photoimages was prepared by 8. M. V. Kozlovsky, V. P. Shibaev, A. I. Stak-
electrochemical polymerization of aniline hanovs et al., Liq. Cryst. 1998, 24, 759–767.
in the presence of methylene blue–bound 9. R. H. Micheels, A. D. Darrow, II, R. D.
Nafion and Ru(bpy)3 3+/4+ . The reaction Rauch, Appl. Phys. Lett. 1981, 39, 418–420.
10. B. Reichman, F.-R. F. Fan, A. J. Bard,
scheme of the image-forming processes J. Elecrochem. Soc. 1980, 127, 333–338.
is illustrated in Fig. 5. A solid-state de- 11. T. Inoue, A. Fujishima, K. Honda, Jpn. J.
vice using the same components was Appl. Phys. 1979, 18, 2177–2178.
12. T. Inoue, A. Fujishima, K. Honda, J. Elec- 38. K. Ajito, L. A. Nagahara, D. A. Tryk et al.,
trochem. Soc. 1980, 127, 1582–1588. J. Phys. Chem. 1995, 99, 16383–16388.
13. R. J. Phillips, M. J. Shane, J. A. Switzer, 39. J. N. Yao, Y. A. Yang, B. H. Loo, J. Phys.
J. Mater. Res. 1989, 4, 923–929. Chem. B 1998, 102, 1856–1860.
14. M. Okano, K. Itoh, A. Fujishima et al., Chem. 40. J. N. Yao, B. H. Loo, K. Hashimoto et al.,
Lett. 1986, 469–472. Ber. Bunsen-Ges. Phys.Chem. 1991, 95,
15. M. Okano, K. Itoh, A. Fujishima et al., 554–556.
J. Electrochem. Soc. 1987, 134, 837–841. 41. J. N. Yao, B. H. Loo, K. Hashimoto et al.,
16. H. Masuda, N. Shimidzu, S. Ohno, Chem. Ber. Bunsen-Ges. Phys. Chem. 1992, 96,
Lett. 1984, 1701–1704. 699–701.
17. L. F. Rozsnyai, M. S. Wrighton, J. Am. Chem. 42. D. W. DeBerry, A. Viehbeck, J. Electrochem.
Soc. 1994, 116, 5993–5994. Soc. 1983, 130, 249–251.
18. M. Futsuhara, K. Yoshioka, Y. Ishida et al., 43. K. Itaya, I. Uchida, S. Toshima et al., J. Elec-
J.Electrochem. Soc. 1996, 143, 3743–3746. trochem. Soc. 1984, 131, 2086–2091.
19. T. L. Rose, D. H. Longendorfer, R. D. Rauh, 44. J. P. Ziegler, J. C. Hemminger, J. Elec-
Appl. Phys. Lett. 1983, 42, 193–195. trochem. Soc. 1988, 134, 358–363.
20. H. Yoneyama, H. Shiotani, N. Nishimura 45. J. P. Ziegler, E. K. Lesniewski, J. C.
et al., Chem. Lett. 1981, 157–160. Hemminger, J. Appl. Phys. 1987, 61,
21. T. Kobayashi, Y. Taniguchi, H. Yoneyama 3099–3104.
et al., J. Phys. Chem. 1983, 87, 768–775. 46. O. Inganas, I. Lundstrom, J. Electrochem. Soc.
22. H. Yoneyama, M. Kitayama, Chem. Lett. 1986, 1984, 131, 1129–1132.
657–660. 47. H. Yoneyama, K. Wakamoto, H. Tamura,
J. Electrochem. Soc. 1985, 132, 2414–2417.
23. H. Yoneyama, K. Kawai, S. Kuwabata,
48. A. A. Nekrasov, V. F. Ivano, A. V. Vannikov,
J. Electrochem. Soc. 1988, 135, 1699–1702.
J. Photogr. Sci. 1993, 41, 43–47.
24. S. Morishita, K. Suzuki, Bull. Chem. Soc. Jpn.
49. H.-T. Zhang, P. Subaramanian, O. Fussary-
1994, 67, 843–846.
del et al., Sol. Energy Mater. 1992, 25,
25. R. J. Colton, A. M. Guzman, J. W. Rabalais,
315–325.
Acc. Chem. Res. 1978, 11, 170–176.
50. H. Yoneyama, S. Hirao, S. Kuwabata, J. Elec-
26. C. M. Lampert, Sol. Energy Mater. 1984, 11,
trochem. Soc. 1992, 139, 3141–3146.
1–27.
51. H. Yoneyama, N. Takahashi, S. Kuwabata,
27. K. Hyodo, Electrochim. Acta 1994, 39, J. Chem. Soc. Chem. Commun. 1992,
265–272. 716–717.
28. R. J. Mortimer, Electrochim. Acta 1999, 44, 52. S. Kuwabata, S. Hirao, H. Yoneyama, Denki
2971–2981. Kagaku oyobi kogyo Butsuri Kagaku 1992, 60,
29. M. Fujii, T. Kawai, H. Nakamatsu et al., 1097–1100.
J. Chem. Soc. Chem. Commun. 1983, 53. S. Kuwabata, N. Takahashi, S. Hirao et al.,
1428–1429. Chem. Mater. 1993, 5, 437–441.
30. B. Ohtani, T. Atsumi, S. Nishimoto et al., 54. S. Kuwabata, A. Kishimoto, H. Yoneyama,
Chem. Lett. 1988, 295–298. J. Electroanal. Chem. 1994, 377, 261–268.
31. E. Kikuchi, K. Iida, A. Fujishima, J. Elec- 55. S. Kuwabata, N. Kihira, H. Yoneyama, Chem.
troanal. Chem. 1993, 351, 105–114. Mater. 1993, 5, 716–719.
32. B. H. Loo, J. N. Yao, H. D. Coble et al., Appl. 56. H. Yoneyama, M. Tokuda, S. Kuwabata, Elec-
Surf. Sci. 1994, 81, 175–181. trochim. Acta 1994, 39, 1315–1320.
33. E. Kikuchi, N. Hirota, A. Fujishima et al., 57. M. Nishizawa, S. Kuwabata, H. Yone-
J. Electroanal. Chem. 1995, 381, 15–19. yama, J. Electrochem. Soc. 1996, 143,
34. Y. A. Yang, Y. W. Cao, P. Chen et al., J. Phys. 3462–3465.
Chem. Solids 1998, 59, 1667–1670. 58. S. Kuwabata, K. Mitsui, H. Yoneyama,
35. J. N. Yao, B. H. Loo, K. Hashimoto et al., J. Electrochem. Soc. 1992, 139, 1824–1830.
J. Electroanal. Chem. 1990, 290, 263–267. 59. S. Kuwabata, K. Mitsui, H. Yoneyama,
36. J. N. Yao, B. H. Loo, A. Fujishima, Ber. J. Electroanal. Chem. 1990, 281, 97–107.
Bunsen-Ges. Phys. Chem. 1990, 94, 13–17. 60. N. Kobayashi, T. Yano, K. Teshima et al.,
37. J. N. Yao, B. H. Loo, A. Fujishima, Nature Electrochim. Acta 1998, 43, 1645–1649.
1992, 355, 624–626.
6.4 Photoelectrochemical Etching 573
6.4 In the 1970s, the idea of photoelec-

Photoelectrochemical Etching trochemical cells for light to electrical
and/or chemical energy conversion at-
Hideki Minoura and Takashi Sugiura tracted scientists in various fields as a
Gifu University, Yanagido, Japan
novel energy conversion device. Photoetch-
6.4.1 ing has been widely used to modify the
Introduction semiconductor surface and consequently
improve the cell efficiency [1, 2]. This im-
Since the earliest stage of the study provement is attributed to several effects,
of semiconductor photoelectrochemistry, such as a decrease in reflection losses [3],
much attention has been drawn to the an increase in effective electrode area, a
instability of semiconductors toward their removal of surface defects acting as re-
photocorrosion in solution (see Chapter 1). combination centers [4–7], and change in
However, this photocorrosion reaction has the chemical composition of surface [8, 9].
also been effectively utilized as a unique The photoetching works well for typically
technique for semiconductor surface pro- photocorrosive semiconductors, such as
CdS, CdSe, and CdTe. The photocorrosion
cessing, which is called ‘‘photoelectrochem-
of other compound semiconductors have
ical etching’’ or simply ‘‘photoetching.’’
been also examined, for example, CuIn5 S8 ,
This term is analogous to that used for
some other ternary semiconductors, InSe,
the metal corrosion applied to surface
MoSe2 , and WSe2 , and so forth [10–23].
treatment, which is often called chemical
The significant effect of photoetching is
etching. In chemical etching processes, the
usually seen with a polycrystalline sam-
etching reaction is caused by an oxidiz-
ple, which has a high defect density at the
ing agent in solution. In photoetching,
surface. Photoetching can be effectively
the presence of a photogenerated hole in
applied, often better than chemical etch-
an n-type semiconductor at the electrolyte ing to the surface processing of thin film
interface is analogous to a broken chemi- samples because of its better controllabil-
cal bond and allows the ionic dissolution ity of etching rate and etching quantity by
of the semiconductor itself. Photoetch- measuring the photocurrent.
ing technique has several unique features. Photoetching is known to be a powerful
The etching rate is usually proportional technique for the characterization of semi-
to the rate of photogeneration of a mi- conductor materials, especially III–V com-
nority carrier, that is, light intensity. This pound semiconductors [24–37], which is
can be monitored by measuring the pho- useful not only to reveal and decorate dis-
tocurrent flowing throughout the external locations, defects, and precipitates but also
circuit, and likewise, the etching quantity to determine the local conductivity type at
can be monitored during the etching pro- the surface.
cess. One more important feature is spatial Photoetching has been also applied to
selectivity at the etching site. Etching pro- fabricate microstructures on the surface of
ceeds only at the irradiated site of the semiconductor. Because the photoetching
surface. It also depends on the crystallo- reaction occurs selectively at the irradi-
graphic orientation, the presence of lattice ated part of the surface, one can etch the
defects, and the energy band structure at semiconductor to form three-dimensional
semiconductor–electrolyte interface. structures by spatial modulation of the
light intensity or by providing physical Preferential dissolution of specified sites

masking. An example is the production of on the surface of semiconductors results in
gratings on the surface of several semicon- the generation of a unique etching pattern.
ductors, for example, CdS [38], GaAs [39, In what follows, we shall describe some ex-
40], and InP [41–43], by the use of a amples that we have found in recent years,
holographic set-up. This technique allows especially focusing on the surface tailor-
fast generation of gratings with submi- ing of cadmium chalcogenide and TiO2
cron meter size without using photoresist semiconductors.
mask. An excellent review article covers
the photoelectrochemical surface process- 6.4.2
ing of III–V semiconductors [44]. Surface-tailoring of Semiconductor
Another example of photoetching appli-
cation is the preparation of monodisperse 6.4.2.1 CdX(X = S, Se)
CdS nanoparticles by the use of size- In this section, we review the character-
selective photoetching technique, which ization of photoetching patterns evolved
has been recently reported by Yoneyama on the surface of polycrystalline CdX(X =
and his coworkers [45–49]. Meijerink and S, Se), which are one of the most
his coworkers applied this technique to extensively studied semiconductors as
other compound semiconductor particles, photoelectrodes in photoelectrochemical
such as ZnS, PbS, and ZnO [50, 51]. It cells, photoconductors, and other opto-
is based on the fact that as the band gap electronic usages. Anodic photocurrent
of size-quantized semiconductor nanopar- flowing through the electrode–electrolyte
ticles increases with a decrease in their interface is entirely due to the oxidative dis-
size, larger CdS nanoparticles can be se- solution of the semiconductors themselves
lectively photoexcited and photocorroded to form Cd2+ ion and elemental chalco-
under irradiation by light. gen, except in certain redox solutions,
Photoetching has also been applied such as aqueous polysulfide, polyselenide,
to make three-dimensional processing of and alkanolamine solutions. For exam-
silicon surfaces. Porous structured silicon ple, in an aqueous NaCl solution, CdSe
is of great interest for electronic and is photoanodically decomposed to form
micromechanical devices (see Chapter 10). Cd2+ ion and elemental Se. After the
Photoanodic dissolution does not pro- photoelectrolysis of a CdSe electrode under
ceed homogeneously over the surface. potentiostatic conditions and a subsequent
(a)
(b)
Fig. 1 SEM photograph of a

polycrystalline CdSe surface
after photoetching at +0.5 V
versus saturated calomel
(c) electrode (SCE) in NaCl
1 µm
solution.
dipping in a polysulfide solution to remove CdSe, photoanodized at +0.5 V versus SCE

surface selenium layer, a fine etching pat- in NaCl solution. It is apparent that the
tern appears on the surface [52–56]. A grain boundary is selectively dissolved and
scanning electron microscope (SEM) pho- characteristic etching patterns, which de-
tograph shown in Fig. 1 is a typical etching pend on the crystallographic orientation,
pattern of the surface of a polycrystalline appear on the surface of each grain.
(a)
1 µm
(b)
1 µm
(c)
1 µm
Fig. 2 SEM photographs of single crystal CdSe surfaces with
different crystallographic plane exposed after photoetching.
Figure 2 shows SEM photographs of An evolution of such a unique skeleton

single-crystal CdSe surface with differ- structure, consisting of grain boundaries
ent crystallographic planes exposed after should be noted in the following points;
photoetching under the same condition one is that such a structure does not evolve
as Fig. 1. By comparing these SEM pho- by other techniques and the other is that
tographs one can determine the crystal- grain boundaries thus left undissolved are
lographic orientation of each grain of subject to closer examinations of grain
polycrystalline samples. It turns out that an boundary.
exposed surface of the grain (a), which has Such a potential-dependent etching site
terrace-like etching pattern, corresponds selectivity is explained by using an energy
to (0002) face, terminated with Cd atoms. band model. Figure 4 is a schematic energy
Likewise, an exposed surface of the grain band model of the polycrystalline n-type
(b), which has layered etching pattern, semiconductor with a grain boundary in
corresponds to the face parallel to c-axis, contact with an electrolyte. It is well known
such as (101̄0) or (112̄0), and that of the that there is a buildup of a potential barrier
grain (c), which has fine pit pattern, cor- because of the formation of electron ac-
responds to (0002̄) face, terminated with ceptor levels at the grain boundary. The
Se atoms. II–VI compound crystals with space charge layers are formed not only
wurtzite structure lack the inversion sym- at the semiconductor–electrolyte interface
metry along their polar axis (c-axis) and but also on both sides of the grain bound-
many workers have reported the difference ary. In the case of weak anodic polarization,
in their properties [57, 58]. The photoetch- namely, near the flat band potential, al-
ing site selectivity also shows a significant most all photogenerated holes that reach
difference between the two polar surfaces. the grain boundary are likely to recombine
One of the important findings in our with the trapped electrons, resulting in lit-
studies on photoetching pattern obser- tle or no photocurrent flowing at the grain
vation is that the photoetching pattern boundary. It means that the grain bound-
differs by varying the photoetching po- ary has less tendency to be photoanodically
tential. Quite interestingly, under weak dissolved. More anodic polarization brings
anodic polarization, the grain boundaries about the following two effects: one is to
are left undissolved and the bulk of each reduce the number of trapped electrons by
grain are selectively dissolved, as shown lowering the Fermi level and the other is to
in Fig. 3, which is in contrast to Fig. 1. generate stronger electric field in the space
Fig. 3 SEM photograph of a

polycrystalline CdSe surface
2 µm after photoetching at −0.3 V
versus SCE in NaCl solution.
Fig. 4 Schematic energy band model of the polycrystalline

n-type semiconductor, having grain boundary at the
solid–electrolyte interface.
charge layer, especially at a grain boundary, it promises a broad range of uses as a

where an additional electric field exists due photocatalyst to decompose harmful or-
to the grain boundary potential. This situ- ganic compounds in the environment (see
ation explains the preferential dissolution Chapters 21, 22). The TiO2 -based dye-
of grain boundary. sensitized photoelectrochemical cell is also
CdS, CdSe-CdS mixed crystals, and receiving great attention as a possible can-
CdSx Se1−x solid solution electrodes also didate for converting solar energy into
exhibit the same etching site selectivity, electricity on a large scale (see Chap-
depending on the crystallographic orien- ter 16–19). TiO2 with a large surface area
tation and the etching potential [54]. It and a high crystallinity is preferable for
is well known that the grain boundaries these applications. It has been found that
play an important role in the electroce- the photoetching helps satisfy both of these
ramic semiconductor devices, such as a requirements [63–65]. Because the pho-
photoelectrochemical cell, a varistor, and toanodic dissolution of TiO2 takes place
a positive temperature coefficient (PTC) only in an aqueous solution of sulfuric
thermistor [59–62]. More detailed studies acid in competition with water oxida-
tion [66–70], the photoetching should be
on the microstructure of the grain bound-
carried out in this solution.
ary by examining the species, thus formed,
Figure 5 shows the surface morphology
and separated from semiconductor pellets,
of a polycrystalline TiO2 after photoetching
would contribute further understanding of
at +1.0 V versus SCE, which is about 1.2 V
the grain boundary and will be described
anodic with respect to photocurrent onset
in Sect. 6.4.3.
at the electrode. The electrode used is a
pellet prepared by sintering at 1300 ◦ C
6.4.2.2 TiO2 for six hours in N2 atmosphere and
TiO2 has been playing a leading role in subsequently reduced at 700 ◦ C for four
active recent research for the utilization hours in 10% H2 /N2 gas to get n-type
of photoenergy. Because of its high ox- semiconductivity. These SEM pictures
idative power, stability, and nontoxicity, show that the characteristic etching pattern
(a)
1 µm
(b)
1 µm
Fig. 5 SEM photographs of the TiO2 surface after

photoelectrochemical etching under strong anodic
polarization (+1.0 V versus SCE).
appeared on the surface. The platelets TiO2 + SO4 2− + 2h+

are regularly ordered and interconnected
−−−→ TiO · SO4 + 12 O2 (1)
thin walls of TiO2 , creating quadrangular
cells with uniform size. This suggests A transmission electron microscope
that the etching reaction proceeds along a (TEM) photograph (a) and its comple-
specific crystallographic orientation. This mentary selected area electron diffraction
morphology was named as the nano- (SAD) pattern (b) of the thin walls of
honeycomb structure. High magnification TiO2, collected from the photoetched sur-
observation (b) reveals that the thickness face are shown in Fig. 6. These are thin
of the platelets is only 20–30 nm and the and long platelets, having a uniform thick-
surface is very smooth. Faradic efficiency ness of several tens of nanometers. The
for photoelectrochemical dissolution of well-regulated lattice fringe of 0.25 nm
TiO2 was determined to be approximately is observed in the TEM photograph (a)
3% by assuming this reaction to be a and it corresponds to the spacing of (101)
two-hole process as expressed by Eq. (1). planes (0.2487 nm). The SAD pattern (b) of
Fig. 6 TEM photograph (a) and its (a)

complementary SAD pattern (b) of
specimens collected from the
0.2487 nm
photoetched TiO2 surface. The
high-magnification image of lattice
fringe is inserted in (a).
the same part of the specimen has a (100)

zone axis of TiO2 with a rutile structure.
The direction of (101̄) diffraction spot is
perpendicular to that of the lattice fringe 500 nm
(b)
shown in the inset of (a), indicating that the
platelet has its c-axis parallel with its long
side. This result shows that the photoetch-
ing reaction proceeds along the c-axis to
construct the deep square honeycomb seen (200)
in Fig. 5. Detailed analysis of many platelet
samples revealed that the surface of all (101)
the walls that construct nano-haneycomb
structure has (100) crystallographic face. (002) Zone axis
It is interesting to note that the surface of [100]
the platelet is very smooth and the rough-
ness is less than several nanometers, as
is seen in atomic force microscopy (AFM) reducing temperature. The size of the hon-
photograph (Fig. 7a). Some irregularities eycomb of the TiO2 sample with higher
in this picture are probably caused by phys- donor density (B in Fig. 8) is obviously
ically attached particles of dusts. Figure 7 smaller than that of a sample with lower
suggests that this platelet is 200 nm in donor density (A in Fig. 8). Figure 9 shows
width, 1 µm in length, and 25 nm in thick- a SEM photograph of single crystal ru-
ness. These observations are in good agree- tile TiO2 surface ((001) face) (Nd = 1020
ment with the SEM pictures (Fig. 5). It is cm−3 ) after photoetching at +1.0 V versus
likely that the photoetching proceeds by SCE. Regularly ordered submicron porous
the layer-by-layer dissolution mechanism structure, consisting of a thin wall with
to create atomically flat surfaces of TiO2 . (100) crystal faces, evolved over the sur-
To find useful applications of the nanos- face. Such a structure apparently satisfies
tructured TiO2 , the size of each hon- the requirements for a large specific sur-
eycomb should be controlled. According face area and high crystallinity.
to our recent experimental results, the In analogy with cadmium chalco-
size is independent of the photoetch- genides, mentioned in the previous
ing conditions and is varied by changing section, the formation of a ‘‘skeleton’’
donor density (Nd ) of TiO2 . Figure 8 structure, consisting of only grain bound-
shows SEM photographs of polycrystalline aries is also possible by anodizing a poly-
TiO2 , having different donor densities af- crystalline TiO2 electrode at weak anodic
ter photoetching at +1.0 V versus SCE. polarization under illumination. A typi-
Donor density was increased by raising the cal SEM photograph of the TiO2 surface
(a) 1.5
[µm]
0.5
0
0 0.5 1 1.5
[µm]
(b) 69.02
[nm]
0.0
0 1282.701
[nm]
Fig. 7AFM image and it’s cross-sectional analysis of TiO2
specimen collected from the photoetched surface.
photoetched at +0.1 V versus SCE is is shown in Fig. 11. A grain boundary has
shown in Fig. 10. Interestingly, the grains a uniform thickness of about several tens
are selectively dissolved, whereas the grain of nanometers. Detailed analyses of grain
boundaries were left undissolved. The boundary is described as follows. Figure 12
height of the grain boundary wall is several shows its high-resolution TEM photograph
microns and the thickness is 10 ∼ 20 nm. and corresponding SAD pattern. The in-
Such a potential dependence of the pho- cident direction of the electron beam is
toetching pattern can be explained in the parallel to the (110) zone axis. The well-
same manner shown in Fig. 4. regulated lattice fringe with a distance of
0.32 nm, which corresponds to the inter-
6.4.3 planar spacing of (110) plane (0.3247 nm).
Analyses of Grain Boundaries This result shows that the grain boundary
is not amorphous but crystalline.
Now that one can separate and isolate A TEM photograph and its correspond-
grain boundaries from a photoetched poly- ing SAD pattern of another part of the
crystalline TiO2 surface, these samples same specimen are shown in Fig. 13. The
are subjected to closer characterization. A moiré pattern of the 0.7 nm width is seen
TEM photograph of a grain boundary taken in the TEM image (a). Such a moiré pattern
out from the surface of a photoetched TiO2 is produced by the interference of crystal
Fig. 8 SEM photographs of

polycrystalline TiO2 having different (a)
donor density after photoetching at
+1.0 V versus SCE. Samples reduced at
(a) 700 ◦ C, four hours and (b) 1000 ◦ C,
one hour, respectively, in the stream of
10% H2 /N2 gas.
100 nm
(b)
100 nm
Fig. 9 SEM photograph of

single crystal rutile TiO2 surface
((001) face) after photoetching
at +1.0 V versus SCE.
1 µm
Fig. 10 SEM photograph of

skeleton-structured TiO2
surface after photoetching at
+0.1 V versus SCE.
1 µm
Fig. 11 TEM photograph of the

grain boundary specimen
picked up from the TiO2 surface
after photoetching.
1 µm
Fig. 12 High-resolution TEM

(110) photograph and SAD pattern of
the grain boundary.
3.247 Å
10 nm
lattices of two twisted thin crystals. In the diffraction (spot A) arises from the crys-
SAD pattern (b) an unexpected diffraction tal (a) and is superimposed with a twist
spot, C, emerges in addition to two sets of 18◦ . When an incident electron beam
of (111) lattice reflections (A & B) and is reflected two times by each (111) lat-
(110) lattice reflections, D, of a rutile struc- tice planes of crystals (a) and (b), it causes
ture. It was interpreted that the spot C double diffraction of the electron beam to
originates from double electron diffrac- create spot C. Overlapping of two (111)
tion at the grain boundary, as illustrated in lattice planes twisted with a small angle
Fig. 14. Considering that the grain bound- produces a moiré fringe. Its spacing is
ary has a sandwich structure that consists calculated by Eq. (2).
of two thin crystals (a, b) of TiO2, hav-
ing different crystal orientations, is quite d1 d2
dM = (2)
reasonable. One of the (111) lattice diffrac- (d12 + d22 − 2d1 d2 cos α)1/2
tion (spot B in Fig. 13b) and (110) lattice
diffraction (spot D in Fig. 13b) are assigned where dM is the spacing of moiré fringe,
to the diffraction spots, arising from the d1 , d2 refers to the spacing of each lattice,
crystal (b) with the (110) zone axis parallel and α is the angle between the two lattices.
to the electron beam. Another (111) lattice By substituting 0.2188 nm (lattice spacing
Fig. 13 TEM photograph (a)

(a)
and its complementary SAD
pattern (b) of the grain 7Å
boundary.
500 nm
(b)
a A
(111)
B (111)
c b
C
D (110)
Fig. 14 Schematic illustration Electron beam

of double-electron diffraction at
the grain boundary.
Grain
boundary
a
b
C′
A′ B′ A
C B
of (111) plane) for d1 and d2 , and 18◦ perpendicular to the vector c of the electron
(the angle between two electron diffraction diffraction spot C, which is the sum of the
spots, A and B) for α, dM is calculated vectors a and b of diffraction spots of two
to be 0.7 nm, which is consistent with (111) planes (A and B).
the spacing of moiré fringe in Fig. 13(a). On the basis of the above information
The direction of the moiré fringe is the structure of the polycrystalline TiO2
Fig. 15 Schematic illustration

of the grain boundary model of
Grain boundary Grain semiconductor.
Space charge layer
Grain
Grain
boundary
Grain
surface can be schematically illustrated grain boundary by using this technique

as Fig. 15. The grain boundaries are rep- must contribute to further understanding
resented by a solid line and the space of the grain boundary of ceramics.
charge layer that is formed near the grain
boundary is represented by a dotted line. 6.4.4
At the grain boundary region in contact Conclusion and Prospect
with an electrolyte, the photogenerated
In conclusion, the photoetching technique
holes arrives at the surface recombine
has been found to have an advantage
with the electrons trapped in the grain
of unique site-selective dissolution and
boundary region, whereas the dissolution
to contribute not only to microtailoring
of the grain bulk in contact with the elec-
of semiconductor surface but also to
trolyte takes place in competition with
further characterization of grain structure.
the water oxidation. Consequently, it is
Particularly, realization of TiO2 with a
understandable that the grain boundary
large specific surface area and high
thus left undissolved is a thin plate,
crystallinity seems to be very effective for
consisting of two platelets with different
improving sensor devices, a photocatalyst,
crystallographic orientations attached to
and a dye-sensitized solar cell.
each other. It seems that the thickness of
the grain boundary thus left undissolved,
References
depends on the thickness of the space
charge layer formed at the grain bound- 1. G. Hodes, J. Manassen, D. Cahen, US
ary. This model is consistent with our Patent 4 1983, 386, 142.
results of the grain boundary observation 2. R. N. Pandey, K. S. Babu, K. S. Chandra,
O. N. Srivastava, Prog. Surf. Sci. 1996, 52,
of CdX (X = S, Se) [54, 55], where SAD pat-
125–192.
terns of the grain boundary obtained from 3. D. Cahen, G. Hodes, J. Manassen, R. Tenne,
the photoetched polycrystalline CdSe and in Photoeffects at Semiconductor – Electrolyte
CdSe–CdS mixed crystal has diffraction Interface (Eds.: A. J. Nozik), The American
spots having multiple zone axes. Interest- Chemical Society, Washington, D.C., 1981,
ingly, diffraction spots by both CdS and pp. 369–385.
4. K. Matsushita, R. Chiba, S. Okuyama et al.,
CdSe are also observed for grain bound- J. Electrochem. Soc. 1993, 140, 2097–2100.
ary samples of a photoetched CdSe–CdS 5. C. R. Elliott, J. C. Regnault, J. Electrochem.
mixed crystal. More detailed analyses of Soc. 1980, 127, 1557–1562.
6. T. Saitoh, S. Matsubara, S. Minagawa, 33. J. L. Weyher, Int. Phys. Conf. Ser. 1995, 146,
J. Electrochem. Soc. 1975, 122, 670–674. 399–408.
7. M. M. Faktor, J. L. Stevenson, J. Electrochem. 34. D. Rotter, J. Uffmann, J. Ackermann et al.,
Soc. 1978, 125, 621–629. Mater. Res. Soc. Symp. Proc. 1998, 482,
8. E. Galun, G. Hodes, M. Peisach et al., 1003–1008.
J. Cryst. Growth 1992, 117, 666–671. 35. C. Youtsey, G. Bulman, I. Adesida, Mater.
9. R. Khare, D. B. Young, E. L. Hu, J. Elec- Res. Soc. Symp. Proc. 1997, 468, 349–354.
trochem. Soc. 1993, 140, L117–L118. 36. C. Youtsey, I. Adesida, G. Bulman, Electron.
10. H. H. Streckert, J. Tong, A. B. Ellis, J. Am. Lett. 1997, 33, 245–246.
Chem. Soc. 1982, 104, 581–588. 37. D. Soltz, M. A. D. Paoli, L. Cescato, J. Vac.
11. R. Tenne, Appl. Phys. Lett. 1983, 43, 201–203. Sci. Technol. B 1996, 14, 1784–1790.
12. G. Hodes, D. Cahen, H. J. Leamy, J. Appl. 38. V. A. Tyagai, V. A. Sterligov, Ya. Kolbasov,
Phys. 1983, 54, 4676–4678. Electrochim. Acta 1977, 22, 819.
13. M. A. Russak, C. Creter, J. Electrochem. Soc. 39. D. V. Podlesnik, H. H. Gilgen, R. M. Osgood
1985, 132, 1741–1745. et al., Appl. Phys. Lett. 1983, 43, 1083–1085.
14. G. Hodes, J. Manassen, D. Cahen, J. Elec- 40. R. M. Lum, A. M. Glass, F. W. Ostermayer
trochem. Soc. 1981, 128, 2325–2330. et al., J. Appl. Phys. 1985, 57, 39–44.
15. R. Tenne, G. Hodes, Appl. Phys. Lett. 1980, 41. R. Matz, J. Zirrgiebel, J. Appl. Phys. 1988, 64,
37, 428–430. 3402–3406.
16. N. Mueller, R. Tenne, Appl. Phys. Lett. 1981, 42. R. M. Lum, F. W. Ostermayer, P. A. Kohl
39, 283–285. et al., Appl. Phys. Lett. 1985, 47, 269–271.
17. G. Dagan, S. Endo, G. Hodes et al., Sol. 43. F. Decker, D. A. Soltz, L. Cescato, Electro-
Energy. Mater. 1984, 11, 57–74. chim. Acta 1993, 38, 95–99.
18. G. Dagan, D. Cahen, J. Electrochem. Soc. 44. R. David Rauh, in Electrochemistry of
1987, 134, 592–600. Semiconductor and Electronics (Eds.: J. Mc-
19. R. Tenne, B. Theys, J. Rioux et al., J. Appl. Hardy, F. Ludwig), Park Ridge, N.J., 1992,
Phys. 1985, 57, 141–145. pp. 177–216.
20. R. Tenne, A. Wold, Appl. Phys. Lett. 1985, 47, 45. H. Matsumoto, T. Sakata, H. Mori et al.,
707–709. Chem. Lett. 1995, 595–596.
21. A. M. Chaparro, P. Salvador, A. Mir, J. Elec- 46. H. Matsumoto, T. Sakata, H. Mori et al.,
troanal. Chem. 1997, 422, 35–44. J. Phys.Chem. 1996, 100, 13781.
22. V. D. Das, L. Damodare, J. Appl. Phys. 1997, 47. T. Torimoto, H. Nishiyama, T. Sakata et al.,
81, 1522–1530. J. Electrochem. Soc. 1998, 145, 1964–1968.
23. H. Minoura, Y. Ueno, H. Kaigawa et al., 48. M. Miyake, T. Torimoto, T. Sakata et al.,
J. Electrochem. Soc. 1989, 136, 1392–1395. Langmuir 1999, 15, 1503–1507.
24. L. Hollan, J. C. Tranchart, R. Memming, 49. T. Torimoto, H. Kontani, T. Sakata et al.,
J. Electrochem. Soc. 1979, 126, 855–859. Chem. Lett. 1999, 379–380.
25. A. Yamamoto, S. Yano, J. Electrochem. Soc. 50. A. van Dijken, D. Vanmaekelbergh,
1975, 122, 260–267. A. Meijerink, Chem. Phys. Lett. 1997, 269,
26. A. Yamamoto, S. Tohno, C. Umemura, 494–499.
J. Electrochem. Soc. 1981, 128, 1095–1100. 51. A. van Dijken, A. H. Janssen, M. H. P.
27. T. Saito, S. Matsubara, S. Minagawa, J. Elec- Smitsmans et al., Chem. Mater. 1998, 10,
trochem. Soc. 1975, 122, 670–674. 3513–3522.
28. F. Kuhn-Kuhnenfeld, J. Electrochem. Soc. 52. T. Sugiura, H. Minoura, Y. Ueno, Denki Ka-
1972, 119, 1063–1068. gaku 1989, 57, 1133–1136.
29. K. Takahashi, Jap. J. Appl. Phys. 1979, 18, 53. T. Sugiura, H. Tomimatsu, H. Minoura
1741. et al., Denki Kagaku 1993, 61, 661–665.
30. P. A. Kohl, IBM J. Res. Dev. 1998, 42, 54. T. Sugiura, R. Tenpaku, H. Minoura et al.,
629–637. Electrochim. Acta 1993, 38, 593–569.
31. E. A. Miller, G. L. Richmond, J. Phys. Chem. 55. T. Sugiura, M. Hida, H. Minoura et al., Appl.
B 1997, 101, 2669–2677. Phys. Lett. 1990, 56, 1954–1956.
32. S. Mueller, J. L. Weyher, K. Koehler et al., 56. T. Sugiura, M. Hida, M. Ohnishi et al., Elec-
Int. Phys. Conf. Ser. 1996, 145, 279–284. trochim. Acta 1992, 37, 1429–1432.
57. H. Iwanaga, T. Yoshiie, T. Yamaguchi et al., 63. T. Sugiura, T. Yoshida, H. Minoura, Elec-
J. Crystal Growth 1979, 47, 703–711. trochem. and Solid-state Lett. 1998, 1, 175–177.
58. T. Yoshiie, H. Iwanaga, T. Yamaguchi et al., 64. T. Sugiura, S. Itoh, T. Ooi et al., J. Elec-
J. Crystal Growth 1981, 51, 624–626. troanal. Chem. 1999, 473, 204–208.
59. L. M. Levinson, S. Hirano, Eds., Grain 65. T. Sugiura, T. Yoshida, H. Minoura, Electro-
Boundaries and Interfacial Phenomena in chemistry 1999, 67, 1234–1236.
Electro Ceramics, in Ceramic Transactions, 66. L. A. Harris, R. H. Wilson, J. Electrochem.
The American Ceramic Soc., Westerville, Soc. 1976, 123, 1010–1015.
Ohio, 1993, Vol. 41. 67. L. A. Harris, D. R. Cross, M. E. Gerstner,
60. T. R. Gupta, J. Am. Ceram. Soc. 1990, 73, J. Electrochem. Soc. 1977, 124, 839–844.
1817–1840. 68. M. E. Gerstner, J. Electrochem. Soc. 1979, 126,
61. J. A. S. Ikeda, Y.-M. Chiang, J. Am. Ceram. 944–949.
Soc. 1993, 76, 2437–2446. 69. Y. Nakato, H. Akanuma, J. Shimizu et al.,
62. J. A. S. Ikeda, Y.-M. Chiang, A. J. Garratt- J. Electroanal. Chem. 1995, 396, 35–39.
Reed et al., J. Am. Ceram. Soc. 1993, 76, 70. Y. Nakato, H. Akanuma, Y. Magari et al.,
2447–2459. J. Phys. Chem. B 1997, 101, 4934.

Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

Uploaded by

Copyright:

Available Formats

1

1.1 Introduction and Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.8 Chemically Modified Semiconductor–Electrolyte Interfaces . . . . . . 41

1.1 and the ensuing energy crisis caused

Fig. 2 Relative disposition of

Boltzmann function [14, 15]: To provide a numerical sense of the

Fig. 4 Relative disposition of the Fermi

within a few kT in energy from the CB VB

1.3 phase. The situation before and after con-

(a) Density of states

knowledge of Eg . It is important to stress level of the redox system, EF,redox , on

Fig. 8 Relative dispositions of various semiconductor band edge positions

VSC should increase concomitantly. This is photopotential (see following). Figure 10

Surface states arise because of the abrupt φB = Sφm + const (13)

(a) (b) (c)

Fig. 13 Current-potential curves for a

0 500 1000 1500 2000 2500 3000 3500

former also involves a diffusion impedance whether recombination is mediated via

C Fig. 17 Equivalent circuit

An expression for the flux of photo- and recalling that W is proportional to

In Eq. (24), + is the effective quantum

− ln(1 − +) = ln(1 + αLp ) + αW (25) 510

1.0 Fig. 20 Comparison of calculated

curve (c) considers both surface

recombination and recombination in the

1. There is no recombination in the At this juncture, it is worth noting that

simply equating a constant photocurrent state–mediated electron transfer to O2 (in

5 × 100 Fig. 23 Experimental data embodying

the photocurrent, voltage, luminescence, technologically important electrochemical

Interestingly, surface states themselves photon flux values. Even multiplication

1.7 in this chapter. By way of contrast, the

Tab. 2 A summary of approachesa to chemical modification of

Modification Semiconductor(s) Modification Sample

Ru3+ n-GaAs A 239, 240

a Approaches to photoanode stabilization based on polymer films containing

B: electrostatic modification; C: catalysis of multielectron photoprocesses

polymeric or indium tin oxide.

n-MoSe2 (or n-WSe2 ) in a two-photoelectrode cell configuration.

as HER, photooxidation of H2 O, and 1.8.2

Fig. 26 Schematic illustration of the occlusion electrosynthesis approach for the

in Fig. 27a), a p-type semiconductor for alternative energy sources to fossil-

e− Fig. 27 Types of photoelectrochemical

29. S. R. Morrison in The Chemical Physics of 54. R. A. L. Vanden Berghe, F. Cardon, W. P.

77. W. P. Gomes, F. Cardon, Ber. Bunsen-Ges. 102. Y. Rosenwaks, B. R. Thacker, R. K. Ahren-

233. A. Solbrand, A. Henningson, S. Södergren 260. C. Levy-Clement, A. Heller, W. A. Bonner

2.1 However, two developments in the 1970s

Interest in porous photoelectrochemical obvious way to characterize a photoelec-

µe Fig. 2 Scheme representing the general

the absorber. The electron and hole are

potential difference across the resistor. The Time-resolved measurements on linear

2.1.2.2 Current Density Versus Potential electron concentration at the surface. As a

equilibrium and strong depletion, or even (Lmin ) and on α (which is a function

characteristics). For a given temperature defined as the transfer function relating a

Potential Fig. 4 Schematic representation of the

studied the photoluminescence of n-type complicated potential-dependence points

Fig. 5 Schematic representation of the Current

arbitrary choice of complex reactions is be calculated from the electrochemical ad-

Fig. 7 Current density versus potential

band potential (as determined in the

Fig. 9 Equivalent circuit C1

response time of the electrochemical cell determination of reaction mechanisms,

process. Iodide ions present in solution In 1969, Memming reported photocur-

deed formed. Since this first report, photocurrent dou-