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
4 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

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
6 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

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.
8 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

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
10 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

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
 1.3 The Semiconductor–Electrolyte Interface at Equilibrium 11

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.
12 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

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.)
 1.3 The Semiconductor–Electrolyte Interface at Equilibrium 13

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.
14 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

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.)
 1.3 The Semiconductor–Electrolyte Interface at Equilibrium 15

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
16 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

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.
18 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

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.
 1.4 Charge Transfer Processes in the Dark 19

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.
20 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

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.
 1.4 Charge Transfer Processes in the Dark 21

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.)
22 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

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.)
 1.4 Charge Transfer Processes in the Dark 23

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
24 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

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
26 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

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].
 1.5 Light Absorption by the Semiconductor Electrode and Carrier Collection 27

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.)
28 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

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
 1.5 Light Absorption by the Semiconductor Electrode and Carrier Collection 29

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
30 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

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.
 1.5 Light Absorption by the Semiconductor Electrode and Carrier Collection 31

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 de- pression 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].
32 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

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
 1.5 Light Absorption by the Semiconductor Electrode and Carrier Collection 33

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
34 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

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.
36 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

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
38 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

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,
40 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

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
42 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

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
44 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

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
46 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

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.
48 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

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.
50 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

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.
52 1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

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,
62 2 Experimental Techniques

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 .
 2.1 Photoelectrochemical Systems Characterization 63

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
64 2 Experimental Techniques

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 Photoelectrochemical Systems Characterization 65

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
66 2 Experimental Techniques

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
 2.1 Photoelectrochemical Systems Characterization 67

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
68 2 Experimental Techniques

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.
 2.1 Photoelectrochemical Systems Characterization 69

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. +
70 2 Experimental Techniques

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.
 2.1 Photoelectrochemical Systems Characterization 71

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,
72 2 Experimental Techniques

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
 2.1 Photoelectrochemical Systems Characterization 73

−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
74 2 Experimental Techniques

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
 2.1 Photoelectrochemical Systems Characterization 75

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
76 2 Experimental Techniques

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
 2.1 Photoelectrochemical Systems Characterization 77

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
78 2 Experimental Techniques

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
 2.1 Photoelectrochemical Systems Characterization 79

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
80 2 Experimental Techniques

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 inter- band 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
 2.1 Photoelectrochemical Systems Characterization 81

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
82 2 Experimental Techniques

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
 2.1 Photoelectrochemical Systems Characterization 83

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
84 2 Experimental Techniques

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
 2.1 Photoelectrochemical Systems Characterization 85

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
86 2 Experimental Techniques

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
 2.1 Photoelectrochemical Systems Characterization 87

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
Imaginary [ j(ω)/e Φ(ω)]

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
88 2 Experimental Techniques

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,
 2.1 Photoelectrochemical Systems Characterization 89

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].
90 2 Experimental Techniques

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 Photoelectrochemical Systems Characterization 91

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
92 2 Experimental Techniques

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
 2.1 Photoelectrochemical Systems Characterization 93

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
94 2 Experimental Techniques

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).
 2.1 Photoelectrochemical Systems Characterization 95

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).
96 2 Experimental Techniques

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.
 2.1 Photoelectrochemical Systems Characterization 97

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
98 2 Experimental Techniques

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)
 2.1 Photoelectrochemical Systems Characterization 99

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
100 2 Experimental Techniques

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
Imaginary [ j(ω)/e Φ(ω)]

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).
 2.1 Photoelectrochemical Systems Characterization 101

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.
102 2 Experimental Techniques

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.
 2.1 Photoelectrochemical Systems Characterization 103

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.
104 2 Experimental Techniques

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.
 2.1 Photoelectrochemical Systems Characterization 105

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.
106 2 Experimental Techniques

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
108 2 Experimental Techniques

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.
 2.2 Deposition of (Multiple Junction) Semiconductor Surfaces 109

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
110 2 Experimental Techniques

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
 2.2 Deposition of (Multiple Junction) Semiconductor Surfaces 111

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
112 2 Experimental Techniques

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.
 2.2 Deposition of (Multiple Junction) Semiconductor Surfaces 113

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
114 2 Experimental Techniques

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
 2.2 Deposition of (Multiple Junction) Semiconductor Surfaces 115

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
116 2 Experimental Techniques

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.
 2.2 Deposition of (Multiple Junction) Semiconductor Surfaces 117

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
118 2 Experimental Techniques

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.
 2.2 Deposition of (Multiple Junction) Semiconductor Surfaces 119

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
120 2 Experimental Techniques

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
 2.2 Deposition of (Multiple Junction) Semiconductor Surfaces 121

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
122 2 Experimental Techniques

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.
 2.2 Deposition of (Multiple Junction) Semiconductor Surfaces 123

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]
124 2 Experimental Techniques

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
 2.2 Deposition of (Multiple Junction) Semiconductor Surfaces 125

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.
126 2 Experimental Techniques

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
128 2 Experimental Techniques

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.
 2.3 Grafting Molecular Properties onto Semiconductor Surfaces 129

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 .
130 2 Experimental Techniques

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# .
 2.3 Grafting Molecular Properties onto Semiconductor Surfaces 131

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.
132 2 Experimental Techniques

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
 2.3 Grafting Molecular Properties onto Semiconductor Surfaces 133

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
134 2 Experimental Techniques

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)
 2.3 Grafting Molecular Properties onto Semiconductor Surfaces 135

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
136 2 Experimental Techniques

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
 2.3 Grafting Molecular Properties onto Semiconductor Surfaces 137

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
138 2 Experimental Techniques

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)
 2.3 Grafting Molecular Properties onto Semiconductor Surfaces 139

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
140 2 Experimental Techniques

(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
 2.3 Grafting Molecular Properties onto Semiconductor Surfaces 141

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 c