Photoelectrochemistry

1

1

Fundamentals of Semiconductor
Electrochemistry and
Photoelectrochemistry
Krishnan Rajeshwar
The University of Texas at Arlington, Arlington, Texas

1.1
1.2
1.3
1.3.1
1.3.2
1.3.3
1.3.4
1.4
1.4.1
1.4.2
1.4.3
1.5
1.5.1
1.5.2
1.5.3
1.5.4
1.5.5
1.6
1.7
1.7.1
1.7.2
1.7.3
1.7.4
1.7.5

Introduction and Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Electron Energy Levels in Semiconductors and Energy Band Model .
The Semiconductor–Electrolyte Interface at Equilibrium . . . . . . . .
The Equilibration Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The Depletion Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Mapping of the Semiconductor Band-edge Positions Relative to
Solution Redox Levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Surface States and Other Complications . . . . . . . . . . . . . . . . . . .
Charge Transfer Processes in the Dark . . . . . . . . . . . . . . . . . . . .
Current-potential Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Dark Processes Mediated by Surface States or by Space Charge Layer
Recombination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Rate-limiting Steps in Charge Transfer Processes in the Dark . . . . .
Light Absorption by the Semiconductor Electrode and Carrier
Collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Light Absorption and Carrier Generation . . . . . . . . . . . . . . . . . . .
Carrier Collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Photocurrent-potential Behavior . . . . . . . . . . . . . . . . . . . . . . . . .
Dynamics of Photoinduced Charge Transfer . . . . . . . . . . . . . . . . .
Hot Carrier Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Multielectron Photoprocesses . . . . . . . . . . . . . . . . . . . . . . . . . .
Nanocrystalline Semiconductor Films and Size Quantization . . . . .
Introductory Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The Nanocrystalline Film–Electrolyte Interface and Charge Storage
Behavior in the Dark . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Photoexcitation and Carrier Collection: Steady State Behavior . . . . .
Photoexcitation and Carrier Collection: Dynamic Behavior . . . . . . .
Size Quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3
4
8
8
9
11
15
16
16
20
23
25
25
25
29
33
34
34
36
36
36
38
40
41

2

1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

1.8
1.8.1
1.8.2
1.9
1.10

Chemically Modiﬁed Semiconductor–Electrolyte Interfaces
Single Crystals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Nanocrystalline Semiconductor Films and Composites . . .
Types of Photoelectrochemical Devices . . . . . . . . . . . . . .
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

41
41
43
44
46
47
47

3

1.1

Introduction and Scope

The study of semiconductor–electrolyte interfaces has both fundamental and practical incentives. These interfaces have interesting similarities and differences with their semiconductor–metal (or metal oxide) and metal–electrolyte counterparts.
Thus, approaches to garnering a fundamental understanding of these interfaces have stemmed from both electrochemistry and solid-state physics perspectives and have proven to be equally fruitful. On the other hand, this knowledge base in turn impacts many technologies, including microelectronics, environmental remediation, sensors, solar cells, and energy storage. Some of these are discussed elsewhere in this volume. It is instructive to ﬁrst examine the historical evolution of this ﬁeld. Early work in the ﬁfties and sixties undoubtedly was motivated by application possibilities in electronics and came on the heels of discovery of the ﬁrst transistor. Electron transfer theories were also rapidly evolving during this period, starting from homogeneous systems to heterogeneous metal-electrolyte interfaces leading, in turn, to semiconductorelectrolyte junctions. The 1973 oil embargo

and the ensuing energy crisis caused a dramatic spurt in studies on semiconductor–electrolyte interfaces once the energy conversion possibilities of the latter were realized. Subsequent progress at both fundamental and applied levels in the late eighties and nineties has been more gradual and sustained. Much of this later research has been spurred by technological applicability in environmental remediation scenarios. Very recently, however, renewed interest in clean energy sources that are nonfossil in origin, has provided new impetus to the study of semiconductor–electrolyte interfaces. As we also learn to understand and manipulate these interfaces at an increasingly ﬁner (atomic) level, new microelectronics application possibilities may emerge, thus completing the cycle that ﬁrst began in the
1950s.
The ensuing discussion of the progress that has been made in this ﬁeld mainly hinges on studies that have appeared since about 1990. Several review articles and chapters have appeared since then that deal with semiconductor–electrolyte interfaces [1–10]; aspects related to electron transfer are featured in several of these.
This author is also aware of at least three books/monographs/proceedings volumes that have appeared since 1990 [11–13]. The

4

1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

reader is referred to the many authoritative accounts that exist before this time frame for a thorough coverage of details on semiconductor–electrolyte interfaces in general. Entry to this early literature may be found in the references cited earlier. In some instances, however, the discussion that follows necessarily delves into research dating back to the 1970s and
1980s.
To facilitate a self-contained description, we will start with well-established aspects related to the semiconductor energy band model and the electrostatics at semiconductor–electrolyte interfaces in the ‘‘dark’’. We shall then examine the processes of light absorption, electron-hole generation, and charge separation at these interfaces.
The steady state and dynamic aspects of charge transfer are then brieﬂy considered. Nanocrystalline semiconductor ﬁlms and size quantization are then discussed as are issues related to electron transfer across chemically modiﬁed semiconductor–electrolyte interfaces.
Finally, we shall introduce the various types of photoelectrochemical devices ranging from regenerative and photoelectrolysis cells to dye-sensitized solar cells. 1.2

Electron Energy Levels in Semiconductors and Energy Band Model

Unlike in molecular systems, semiconductor energy levels are so dense that they form, instead of discrete molecular orbital energy levels, broad energy bands.
Consider a solid composed of N atoms.
Its frontier band will have 2N energy eigenstates, each with an occupancy of two electrons of paired spin. Thus, a solid having atoms with odd number of valence

electrons (e.g. Al with [Ne]3s2 3p1 ) will have a partially occupied frontier band in which the electrons are delocalized. On the other hand, a solid with an even number of valence electrons (e.g. Si having an electron conﬁguration of [Ne]3s2 2p2 ) will have a fully occupied frontier band (termed a valence band, (VB)). The situation for Si is schematized in Fig. 1.
As Fig. 2 illustrates, the distinction between semiconductors and insulators is rather arbitrary and resides with the magnitude of the energy band gap (Eg ) between the ﬁlled and vacant bands.
Semiconductors typically have Eg in the
1 eV–4 eV range (Table 1). The vacant frontier band is termed a conduction band,
(CB) (Fig. 2). We shall see later that Eg has an important bearing on the optical response of a semiconductor.
For high density electron ensembles such as valence electrons in metals, Fermi statistics is applicable. In a thermodynamic sense, the Fermi level, EF (deﬁned at 0 K
Some elemental and compound semiconductors for photoelectrochemical applications Tab. 1

Semiconductor

Conductivity type(s) Si
GaAs
GaP
InP
CdS
CdSe
CdTe
TiO2

n, p n, p n, p n, p n n n, p n ZnO

n

a The

Optical band gap energy [eV]a
1.11
1.42
2.26
1.35
2.42
1.70
1.50
3.00(rutile)
3.2(anatase)
3.35

values quoted are for the bulk semiconductor. The gap energies increase in the size quantization regime (see Sect. 7).

1.2 Electron Energy Levels in Semiconductors and Energy Band Model
4 N states
0 electrons
6 N electrons
2 N states

Electron energy

3p

Eg
4 N states
4 N electrons

3s
2 N states
2 N electrons

ro

Distance
Schematic illustration of how energy bands in semiconductors evolve from discrete atomic states for the speciﬁc example of silicon.

Fig. 1

Relative disposition of the CB and VB for a semiconductor (a) and an insulator (b). Eg is the optical band gap energy.
Fig. 2

CB

CB

Eg

Eg

VB
(a)

as the energy at which the probability of ﬁnding an electron is 1/2) can be regarded as the electrochemical potential of the electron in a particular phase (in this case, a solid). Thus, all electronic energy levels below EF are occupied

VB
(b)

and those above EF are likely to be empty. Electrons in semiconductors may be regarded as low-density particle ensembles such that their occupancy in the valence and CBs may be approximated by the

5

6

1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

Boltzmann function [14, 15]: ne ≈ No exp −

Eo − EF kT (1)

Now we come to another important distinction between metals and semiconductors in that two types of electronic carriers are possible in the latter.
Consider the thermal excitation of an electron from VB to CB. This gives rise to a free electron in the CB and a vacancy or hole in the VB. A localized chemical picture for the case of Si shows that the hole may be regarded as a missing electron in a chemical bond (Fig. 3). There is a crude chemical analogy here with the dissociation of a solvent such as water into H3 O+ and OH− . In either case, equal numbers of oppositely charged species are produced.
Thus, Eq. (1) becomes: ni ≈ Nc exp −

EF − Ec kT (2)

pi ≈ Nv exp −

Ev − EF kT (3)

In Eqs. (2–3), Nc and Nv are the effective density of states (in cm−3 ) at the lower edge and top edge of CB and VB, respectively.
These expressions can be combined with the recognition that ni = pi to yield n2 ≈ No exp − i ≈ No exp −

Ec − Ev kT Eg kT (4)

To provide a numerical sense of the situation, Nc and Nv are typically both approximately 1019 cm−3 so that the constant
No (Nc Nv ) in Eq. (1) is about 1038 cm−3 .
For a semiconductor such as Si (with
Eg = 1.11 eV, Table 1), ni will be about
1010 cm−3 at 300 K according to Eq. (4).
This rough calculation lends credence to the original rationale for the use of Boltzmann statistics for the electron energy distribution in semiconductors (see preceding section).
The preceding case refers to the semiconductor in its intrinsic state with very low carrier concentrations under ambient conditions. The Fermi level, EF , in this case lies approximately in the middle of the energy band gap (Fig. 4a). This simply reﬂects the fact that the probability of electron occupancy is very high in VB and very low in CB and does not imply an occupiable energy level at EF itself.
In extrinsic semiconductors the carrier concentrations are perturbed such that n = p . Again the analogy with the addition of an acid or base to water is quite instructive here. Consider the case when donor impurities are added to a neutral semiconductor. Since the intrinsic carrier concentrations are so low (sub-parts per trillion), even additions in parts per billion levels can have a profound electrical effect. This process is known as doping of the semiconductor. In this particular case, the Fermi level shifts toward the CB edge (Fig. 4b). When the donor level is

e−
(+)

(−)
Si

e−

Si

h+

Si

h+
Si

Si

Si

Si

A localized picture of electron-hole pair generation
(see also Fig. 2a) in silicon.

Fig. 3

1.2 Electron Energy Levels in Semiconductors and Energy Band Model
Relative disposition of the Fermi level (EF ) for an intrinsic semiconductor
(a), for an n-type semiconductor (b), and a p-type semiconductor (c).

Fig. 4

EF

VB
(a)

EF

Energy

within a few kT in energy from the CB edge, appreciable electron concentrations are generated by the donor ionization process (at ambient temperatures) such that now n p . This is termed n-type doping, and the resultant (extrinsic) semiconductor is termed n-type. By analogy, p -type semiconductors have p
n. The terms minority and majority carriers now become appropriate in these cases. For a p -doped semiconductor case, the Fermi level now lies close to the VB edge (Fig. 4c).
The movement of EF with dopant concentration can also be rationalized via the
Nernst formalism [6].
Doping can be accomplished by adding altervalent impurities to the intrinsic semiconductor. For example, P (a Group
15 or VA element) will act as a donor in Si (a Group 14 or IVA element). This can be rationalized on chemical terms by noting that P needs only four valence electrons for tetrahedral bonding (as in the
Si lattice) – the ﬁfth electron is available for donation by each P atom. The donor density, ND nominally is approximately
1017 cm−3 . Thus, assuming that n ND
(complete ionization at 300 K), p will be only approximately 103 cm−3 [recall that the product ni pi is ∼1020 cm−6 (see preceding section)], bearing out the earlier qualitative assertion that n
p.
Impurity addition, however, is not the only doping mechanism. Nonstoichiometry in compound semiconductors such as
CdTe (Table 1) also gives rise to n- or p type behavior, depending on whether Cd or
Te is in slight excess, respectively. The defect chemistry in these solid chalcogenides controls their conductivity and doping in a

CB

(b)

EF
(c)

complex manner, a discussion of which is beyond the scope of this section. Excellent treatises are available on this topic and on the solid-state chemistry of semiconductors in general [16–22].
The foregoing discussion strictly refers to semiconductors in single-crystal form.
Amorphous and polycrystalline counterparts present other complications caused by the presence of defects, trap states, grain boundaries, and the like. For this reason we orient the subsequent discussion mainly toward single crystals, although comparisons with less ideal cases are made where appropriate. The distinction between metal and semiconductor electrodes is important when we consider the electrostatics across the corresponding solid–liquid interfaces; this distinction is made in the following section.

7

8

1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

1.3

The Semiconductor–Electrolyte Interface at
Equilibrium
1.3.1

The Equilibration Process

The electrochemical potential of electrons in a redox electrolyte is given by the Nernst expression Eredox = E o redox +

RT cox ln nF cred

(5)

In Eq. (5), cox and cred are the concentrations (roughly activities) of the oxidized and reduced species, respectively, in the redox couple. The parameter (Eredox = µe,redox ) as deﬁned by Eq. (5) can be identiﬁed with the Fermi level (EF,redox ) in the electrolyte. This was the topic of debate some years back [23], although this premise now appears to be well founded.
The task now is to relate the electron energy levels in the solid and liquid phases on a common basis.
The semiconductor solid-state physics community has adopted the electron energy in vacuum as reference, whereas electrochemists have traditionally used the standard hydrogen electrode (SHE) scale. While estimates vary [23–25], SHE appears to lie at −4.5 eV with respect to the vacuum level. We are now in a position to relate the redox potential Eredox (as deﬁned with reference to SHE) with the Fermi level EF,redox expressed versus the vacuum reference (Fig. 5a)
EF,redox = −4.5 eV − eo Eredox

(6)

When a semiconductor is immersed in this redox electrolyte, the electrochemical potential (Fermi level) is disparate across the interface. Equilibration of this interface thus necessitates the ﬂow of charge from one phase to the other and a ‘‘band bending’’ ensues within the semiconductor

phase. The situation before and after contact of the two phases is illustrated in
Fig. 5(b) and (c) for an n-type and p -type semiconductor, respectively. After contact, the net result of equilibration is that
EF = EF,redox and a ‘‘built-in’’ voltage, VSC develops within the semiconductor phase, as illustrated in the right hand frames of
Fig. 5(b) and (c).
It is instructive to further examine this equilibration process. Consider again an n-type semiconductor for illustrative purposes (Fig. 5b). The electronic charge needed for Fermi level equilibration in the semiconductor phase originates from the donor impurities (rather than from bonding electrons in the semiconductor lattice).
Thus, the depletion layer that arises as a consequence within the semiconductor contains positive charges from these ionized donors. The Fermi level in the semiconductor (EF,n ) moves ‘‘down’’ and the process stops when the Fermi level is the same on either side of the interface.
The rather substantial difference in the density of states on either side dictates that
EF,n moves farther than the corresponding level, EF,redox in the electrolyte. A particularly lucid account of this initial charge transfer is contained in Ref. 6.
The band-bending phenomenon, shown in Fig. 5(b) and (c), is by no means unique to the semiconductor–electrolyte interface. Analogous electrostatic adjustments occur whenever two dissimilar phases are in contact (e.g. semiconductorgas, semiconductor–metal). An important point of distinction from the corresponding metal case now becomes apparent.
For a metal, the charge, and thus the associated potential drop, is concentrated at the surface penetrating at most a
˚
few A into the interior. Stated differently, the high electronic conductivity of a metal cannot support an electric ﬁeld

1.3 The Semiconductor–Electrolyte Interface at Equilibrium
Vacuum reference

Energy

χ

∅
CB
EF

ox

λ o E redox

Eg red VB
(a)

Density of states

V SC

E CB
EF

E
E CB
E F, equil

E F, redox
E VB

E VB

(b)

E
E CB

E CB
E F, redox
EF
E VB

E F, equil
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 comes into contact with an electrolyte, almost all the potential drop at the interface occurs within the Helmholtz region in the electrolyte phase. On the other hand, the interfacial potential drop across a semiconductor-electrolyte junction (see following) is partitioned both as VSC and as VH leading to a simple equivalent circuit model comprising two capacitors (CSC and CH ) in series (Fig. 6). Further reﬁnements of the equivalent circuit description are given later but the point to note is the

rather variant behavior of a metal and a semiconductor at equilibrium with a redox electrolyte. 1.3.2

The Depletion Layer

There is a characteristic region within the semiconductor within which the charge would have been removed by the equilibration process. Beyond this boundary, the ionized donors (for a n-type semiconductor), have their compensating

9

1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry
Distance coordinate

VG

Gouy region VH
Semiconductor

V SC

Electrolyte
Potential

10

C SC

Helmholtz layer CH

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).

Fig. 6

charge (electrons), and the semiconductor as a whole is electrically neutral. This layer is the space charge region or the depletion layer, so termed because the layer is depleted of the majority carriers.
The potential distribution in this interfacial region can be quantiﬁed by relating the charge density and the electric ﬁeld strength as embodied by the Poisson equation [14, 26]. Under restrictive conditions, more fully discussed in Refs. 6 and 14, we obtain a particularly simple expression V (x) = −

eo ND
2εs

x2

(7)

× (o ≤ x ≤ W )
In Eq. (7), eo is the electronic charge and εs is the static dielectric constant of the semiconductor. The potential distribution is mapped in Fig. 6. We are now in a position to quantify the parameter VSC :

VSC = −

eo ND
2εs

W2

(8)

where W is the depletion layer width.
Further reﬂection shows how the magnitude of W should depend on the semiconductor parameter ND . Consider two cases of a semiconductor, one that is lightly doped (say ND ∼ 1016 cm−3 ) and another that is heavily doped (ND ∼ 1018 cm−3 ).
Obviously in the former case, the charge needed for Fermi level equilibration has to come from deeper into the solid and so the magnitude of W will be larger.
This suggests a strategy for chemical control of the electrostatics at the semiconductor–electrolyte interface [6]. Nominal dimensions of W are in the 10–1000 nm range. This may be compared with the corresponding Helmholtz layer width, typically 0.4–0.6 nm. With the capacitor-inseries model (see earlier section), we can see that the semiconductor space charge

1.3 The Semiconductor–Electrolyte Interface at Equilibrium

layer is usually the determinant factor in the total capacity of the interface. Once again, the contrast with the corresponding metal–electrolyte interface is striking.
Only when the semiconductor is degenerately doped (leading to rather large space charge layer charge, QSC and ‘‘thin’’ depletion layer widths) or when its surface is in accumulation does the situation become akin to the metal–electrolyte interface (see following). 1.3.3

Mapping of the Semiconductor Band-edge
Positions Relative to Solution Redox Levels

Considerations of interfacial electron transfer require knowledge of the relative positions of the participating energy levels in the two (semiconductor and solution) phases. Models for redox energy levels in solution have been exhaustively treated elsewhere [27, 28]. Besides the Fermi level of the redox system (Eq. 6), the thermal ﬂuctuation model [27, 28] leads to a Gaussian distribution of the energy levels for the occupied (reduced species) and the empty
(oxidized species) states, respectively, as illustrated in Fig. 5(a). The distribution functions for the states are given by
Dox = exp −

E − EF,redox − λ2
4 kT λ

(9a)

Dred = exp −

E − EF,redox + λ2
4 kT λ

(9b)

In Eqs. (9a) and (9b), λ is the solvent reorganization energy.
Now consider the relative disposition of these solution energy levels with respect to the semiconductor band edge positions at the interface. The total potential difference across this interface (Fig. 6) is given by
Vt = VSC + VH + VG

(10)

In Eq. (10), Vt is the potential as measured between an ohmic contact on the rear surface of the semiconductor electrode and the reference electrode (Fig. 6). The problematic factors in placing the semiconductor and solution energy levels on a common basis involve VH and VG . In other words, theoretical predictions of the magnitude of VSC (and how it changes as the redox couple is varied) are hampered by the lack of knowledge on the magnitude of VH and VG . A degree of simpliﬁcation is afforded by employing relatively concentrated electrolytes such that VG can be ignored. As with metals, the Helmholtz layer is developed by adsorption of ions or molecules on the semiconductor surface, by oriented dipoles, or especially in the case of oxides, by the formation of surface bonds between the solid surface and species in solution. Recourse to band edge placement can be sought through differential capacitance measurements on the semiconductor–redox electrolyte interface [29].
In the simplest case as more fully discussed elsewhere [14, 15, 29], one obtains the Mott-Schottky relation (for the speciﬁc instance of a n-type semiconductor) of the semiconductor depletion layer capacitance (CSC ), again by invoking the Poisson equation 2
1/CSC =

2 kT (V − Vfb ) −
ND eo εs eo (11)

In Eq. (11), Vfb is the so-called ﬂat band potential, that is the applied potential
(V ) at which the semiconductor energy bands are ‘‘ﬂat’’, leading up to the solution junction. Several points with respect to the applicability of Eq. (11) must be noted.
The Mott-Schottky regime spans about
1 V in applied bias potential for most semiconductor–electrolyte interfaces (i.e.

11

1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

in the region of depletion layer formation of the semiconductor space charge layer, see preceding section) [15]. The simple case considered here involves no mediator trap states or surface states at the interface such that the equivalent circuit of the interface essentially collapses to its most rudimentary form of CSC in series with the bulk resistance of the semiconductor.
Further, in all the earlier discussions, it is reiterated that the redox electrolyte is sufﬁciently concentrated that the potential drop across the Gouy (diffuse) layer in the solution can be neglected. Speciﬁc adsorption and other processes at the semiconductor–electrolyte interface will inﬂuence Vfb ; these are discussed elsewhere [29, 30] as are anomalies related to the measurement

8.0

process itself [31]. Figure 7 contains representative Mott-Schottky plots for both nand p -type GaAs electrodes in an ambient temperature molten salt electrolyte [32].
Once Vfb is known (from measurements), the Fermi level of the semiconductor at the surface is deﬁned. It is then a simple matter to place the energies corresponding to the conduction and
VBs at the surface (ECB and EVB , respectively) if the relevant doping levels are known. The difference between ECB and
EVB should approximately correspond to the semiconductor band gap energy, Eg
(see Figs. 4 and 7). Alternatively, if Vfb is measured for one given state of doping of the semiconductor (n- or p -doped), the other band edge position can be ﬁxed from
× 1015

× 1013

4.0

3.2

4.8

2.4

3.2

1.6

1.6

0.8

I C−2
[farad−2]

6.4

I C−2
[farad−2]

12

0
−0.5

−0.1

0.3

0.7

1.1

0
1.5

Potential
0/3 +

[V vs Al

]

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.)

Fig. 7

1.3 The Semiconductor–Electrolyte Interface at Equilibrium

knowledge of Eg . It is important to stress that the semiconductor surface band edge positions (as estimated from Vfb measurements) comprises all the terms in Eq. (10) and reﬂects the situation in situ for a given set of conditions (solution pH, redox concentration, etc.) of the semiconductorredox electrolyte. The situation obviously becomes complex when the charge distribution and mediation at the interface changes either via surface states and illumination or both. These complications are considered later. Figure 8 contains the relative dispositions of the surface band edges mapped for a number of semiconductors in aqueous media.
Having located the semiconductor band edge positions (relative either to the vacuum reference or a standard reference electrode), we can also place the Fermi

level of the redox system, EF,redox , on the same diagram. Energy diagrams such as those in Fig. 8 are important in considerations of charge transfer as we shall see later. In anticipation of this discussion, it is apparent that the three situations illustrated in Fig. 9 for an n-type semiconductor–electrolyte interface entail the participation of the semiconductor CB,
VB, and even states in its band gap in charge exchange with the solution species.
Here again is a point of departure from the metal case; viz., for a semiconductor, hole, electron, and surface state pathways must all be considered.
Let us return to the band bending process at the interface. For a given semiconductor, the expectation is that as the redox Fermi level is moved more positive (‘‘down’’ on the energy diagram),

Vacuum scale
SHE
[eV]
0
[V]
−3.0
−4.0

CdS

−1.5
−0.5
0

−4.5

1.4 eV

TiO2

2.4 eV

+1.0

Potential

−6.0

CdSe

2.3 eV

Energy

−5.0

GaAs GaP
Si

CdTe

1.1 eV ∼ 1.3 eV 1.5 eV

1.7 eV
∼ 3.0 eV

+2.0
−7.0
+3.0
−8.0

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.

Fig. 8

InP

13

1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

CB
EF

E redox

Energy

CB
EF

VB
(a)

CB
EF

o
E redox

E SS

VB
(b)

VB

o
E redox

(c)

Fig. 9

Three situations for a n-type semiconductor–electrolyte interface at equilibrium showing overlap of the redox energy levels with the semiconductor CB (a), with

surface states (b), and with the semiconductor
VB (c). A discrete energy level is assumed for the surface states as a ﬁrst approximation.

VSC should increase concomitantly. This is the ideal (band edge ‘‘pinned’’) situation.
In other words [23]

photopotential (see following). Figure 10 shows that this ideal situation indeed is realized for selected semiconductor–electrolyte interfaces [33]. As further discussed later, the analogy with the corresponding metal-semiconductor junctions
(Schottky barriers) is direct [5, 34–36].
Complications arise when there are surface states that mediate charge exchange at the interface. When their density is

d VSC
=1
dEredox

(12)

Equation (12) reﬂects the fact that the change in band bending faithfully tracks the redox potential change. A measure of the former is the open-circuit
1.2

I
0.8

Vph
[V]

14

slope = 1.0
0.4

0.0

−0.4

0.0

Vredox
[V vs. SCE]

+ 0.4

Fig. 10 Plot of the open-circuit photovoltage for amorphous
Si-methanol interfaces containing a series of one-electron redox couples.
(Reproduced with permission from Ref. 33.)

1.3 The Semiconductor–Electrolyte Interface at Equilibrium

high [37], they act as a ‘‘buffer’’, in that in the extreme case, carriers in the semiconductor energy bands are completely excluded from the equilibration process.
1.3.4

Surface States and Other Complications

Surface states arise because of the abrupt termination of the crystal lattice at the surface; obviously the bonding arrangement is different from that in the bulk.
Consider our prototype semiconductor, Si.
The tetrahedral bonding characteristic of the bulk gives way to coordinative unsaturation of the bonds for the Si surface atoms. This unsaturation is relieved either by surface reconstruction or bonds with extraneous (e.g. solvent) species [29,
38–40]. The surface bonding results in a localized electronic structure for the surface that is different from the bulk. The energies of these localized surface orbitals nominally lie in the forbidden band gap region. The corresponding states are thus able (depending on their energy location) to exchange charge with the conduction or VBs of the semiconductor and/or the redox electrolyte [29].
Unlike the case illustrated in Fig. 10, changes in the solution redox potential have been observed to cause no change in the magnitude of VSC . This situation is termed Fermi level pinning; in other words, the band edge positions are unpinned in these cases so that the movement of Eredox is accommodated by VH rather than by VSC . As mentioned earlier, it appears [37] that surface state densities as low as 1013 cm−2 (∼1% of a monolayer) sufﬁce to induce complete Fermi level pinning in certain cases. Of course, intermediate situations are possible. Thus, the ideal case manifests a slope of 1 in a plot of VSC (or an equivalent parameter)

versus Eredox (see Fig. 10). On the other hand, complete pinning results in a slope of zero. Intermediate cases of Fermi level exhibit slopes between 0 and 1 [41]. As stated earlier, there is a direct analogy here with the metal/semiconductor junction counterparts [42, 43]: φB = Sφm + const

(13)

In Eq. (13), φB is the so-called Schottky barrier height, φm is the metal work function, and S is a dimensionless parameter.
Attempts have been made to relate S to semiconductor properties [44–48].
To further complicate matters, the nonideal behavior of semiconductor–electrolyte interfaces as noted earlier is exacerbated when the latter are irradiated. Changes in the occupancy of these states cause further changes in VH , so that the semiconductor surface band edge positions are different in the dark and under illumination. These complications are considered later. The surface states as considered earlier are shallow (with respect to the band edge positions) and can essentially be considered as completely ionized at room temperature. However, for many oxide semiconductors, the trap states may be deep and thus are only partially ionized.
Speciﬁcally, they may be disposed with respect to the semiconductor Fermi level such that they are ionized only to a depth that is small relative to W [49]. The physical manifestation of such deep traps as observed in the AC impedance behavior of semiconductor–electrolyte interfaces has been discussed [14, 49].
Finally, within the Mott-Schottky approximation (Eq. 11), large values of εs or
ND can lead to the ratio VH /VSC becoming signiﬁcant. Figure 11 contains estimates of this ratio for several values of ND for a semiconductor with a large εs value (TiO2 , εs = 173) mapped as a function of the total

15

1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry
1.0
13
12

0.8

∆V H /∆VSC + ∆VH

16

11
10

0.6

9

0.4

8
7
6

0.2

5
2

3

4

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].
Clearly VH can become a sizable fraction of the total potential drop (approaching the situation for metals) under certain conditions. It has been shown [51] that the
Mott-Schottky plots will still be linear but the intercept on the potential axis is shifted from the Vfb value.
1.4

Charge Transfer Processes in the Dark
1.4.1

Current-potential Behavior

Let us return to the equilibrium situation of an n-type semiconductor in contact with a redox electrolyte and reconsider the situation in Fig. 9(a). This is shown again in Fig. 12(a) to underscore the fact that the interface is in a state of dynamic equilibrium. That is, the forward and

reverse (partial) currents exactly balance each other and there is no net current ﬂow across the interface. In fact, the situation here is similar to that occurring for a metal–redox electrolyte interface at the rest potential. We can write down expressions for the net current using a kinetics methodology as in Ref. 6 with some minor changes in notation: ic = −eo Aket cox (ns − nso )

(14)

In Eq. (14), ket is the rate constant for electron transfer, cox is the concentration of empty (acceptor) state in the redox electrolyte, ns and nso are the surface concentrations of electrons, the subscript ‘‘o’’ in the latter case denoting the equilibrium situation. Thus, as long as the semiconductor–electrolyte interface is not perturbed by an external (bias) potential, ns ≡ nso and the net current is zero. The voltage

1.4 Charge Transfer Processes in the Dark
E

E

E CB

E F, redox

Vbias

E VB

(a)

E

E CB

Vbias
E F, redox

E CB

E VB
(b)

E F, redox

E VB
(c)

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.

Fig. 12

dependence of the current is embodied in the ratio, ns /nso , which can be regarded as a measure of the extent to which the interface is driven away from equilibrium.
It must be noted that nso is not the bulk concentration of majority carriers (n) in the semiconductor because of the potential drop across the space charge layer [6, 35]. nso = n exp −

eo VSC kT similar to the metal case. The major difference resides in the vastly different state densities in the solid and the existence of an energy gap region. The two nonequilibrium situations are shown in Figs. 12(b) and 12(c), respectively. Away from equilibrium, we have the analogous Boltzmann expression counterpart to Eq. (15)

(15)

ns = n exp −

A few words about the units of the terms in Eq. (14) are in order at this juncture.
The term i /eo A may be regarded as a ﬂux (J ) in units of number of carriers crossing per unit area per second [1, 3,
8]. The concentration terms are in cm−3 ; thus ket has the dimensions of cm4 s−1 because of the second-order kinetics nature stemming from the two multiplied concentration terms in Eq. (14) [1, 3, 8].
Consider now the application of a bias potential to the interface. Intuitively when it is such that ns > nso , a reduction current (cathodic current) should ﬂow across the interface such that the oxidized redox species are converted to reduced species
(Ox → Red). On the other hand, when nso > ns , the current ﬂow direction is reversed and an anodic current should ﬂow.
Once again the situation here is somewhat

eo (VSC + V ) kT leading, in turn, to

(16)

ic = −eo Aket cox nso
× exp −

eo V kT −1

(17)

The assumption is inherent in the preceding discussion that all of the applied bias (V ) drops across the space charge layer such that we are modulating only the majority carrier population at the surface (and not the potential drop across the
Helmholtz layer). In other words, the band edge positions are pinned or there is no
Fermi level pinning (see Sect. 1.3.4).
Analogous expressions may be developed for majority carrier ﬂow for a p -type semiconductor in contact with a redox electrolyte, with the important caveat that the VB is involved in this process instead.

17

1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

readily rationalized. Finally, io will increase with doping because of the ‘‘thinness’’ of the resultant barrier at the surface.
When EF,redox is moved ‘‘down,’’ that is more positive, the band bending increases,
VSC increases and thus nso decreases. A similar alteration in cox affects EF,redox through the Nernst expression. In both instances, we are inﬂuencing the Fermi level at the interface at equilibrium. Thus, in a sense, io is a quantitative measure of the extent of rectiﬁcation of a given interface; that is, a smaller io value translates to better rectiﬁcation.
The reverse bias current remains at a very low value because of the lack of minority carriers (i.e. holes for n-type semiconductor) in the dark. Alternatively, injection of electrons from occupied redox levels (also an anodic current) has to thermally surmount the surface barrier [5, 34,
35]. Under extreme reverse bias, however, this barrier becomes ‘‘thin’’ and electrons can tunnel through it, leading to an abrupt increase in the anodic current. This process was studied even in the early days of

Equation (17) suggests that the cathodic current is exponentially dependent on potential for V < 0. This is the so-called forward-bias regime. On the other hand, when V > 0 (reverse-bias regime) the current is essentially independent of potential and, importantly, is of opposite sign. Simply put, in this case, the electron ﬂow direction (i.e. anodic) is from the occupied redox states into the semiconductor
CB (Fig. 12c). It should not, thus, be surprising that this process is independent of potential. Both bias regimes are contained in curve 1 in Fig. 13.
Of particular interest to this discussion is the ‘‘preexponential’’ term in Eq. (17): io = eo Aket cox nso

(18)

Analogous to the metal case, we can call this term the exchange current; it is the current that ﬂows at equilibrium when the partial cathodic and anodic components exactly balance one another. Of particular interest is the dependence of io on nso .
Also, variations in cox will affect the magnitude of io . Both these trends can be
Light, I 2 (> I 1)

Current

Anodic

Light, I 1

0

Dark

Cathodic

18

−

Voltage

+

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 latter case.

1.4 Charge Transfer Processes in the Dark

semiconductor electrochemistry [52] and a detailed discussion is found in a book chapter [14]. Ultimately the junction ‘‘breaks down’’ (at the so-called Zener limit). This dark current ﬂow is not shown in Fig. 13
(curve 1).
Returning to the forward-bias (cathodic) current ﬂow, Eq. (17) bears some analogy to the famous Tafel expressions in electrochemical kinetics. Thus, ignoring the unity term within the square brackets, Eq. (17) predicts a Tafel slope of 60 mV per decade at 298 K. In many instances [53, 54], such a slope indeed is observed. In many cases, however, the slopes are higher than the
‘‘ideal’’ value [14, 55–59].
The causes for this anomalous behavior are still not fully understood. It appears likely that many factors are involved: surface ﬁlm formation, varying potential drop across the Helmholtz region caused, for example, by surface state charging, and so on. Even crystallographic orientations appear to be important [59]. These aspects have been discussed by other authors [14,
55, 60].
We have so far considered only (majority carrier) processes involving the CB (again assuming for illustrative purposes a n-type semiconductor). Consider the interfacial situation depicted in Fig. 9(c). The energy states of the redox system now overlap with the VB of the semiconductor such that hole injection in the ‘‘dark’’ is possible. When the band bending is large, the injected holes remain at the surface and attack the semiconductor itself, causing the latter to undergo corrosion. If the bias potential is

such that the band bending is modest and the holes recombine with electrons (either via the surface states or in the space charge region itself), a cathodic current ﬂows that is carried by the majority carriers in the bulk. This recombination current pathway is schematized in Fig. 14 and is further discussed in the next section. Hole injection has been extensively studied especially for III–V (Group 13–15) semiconductors such as GaAs and GaP because of the relevance of this process to electroless etching and device fabrication technology. This topic has been reviewed [61–64].
The invokement of either the CB or the
VB of the semiconductor in charge exchange in the dark with solution redox species is not always straightforward. This is particularly true for multielectron redox processes to be discussed later. Movement of the semiconductor band edge positions
(i.e. band edge unpinning) relative to the redox energy levels also presents a complicating situation (see following). Some cases (e.g. Eu2+/3+ in contact with GaAs electrodes) are interesting in that the same

E
Diffusion

e−

e−

ECB e− e−

Et

E SS
Electrolyte
e−

Fig. 14 Hole injection into the VB of a n-type semiconductor from an oxidant
(e.g. Fe3+ ) and the injection or recombination pathway. Both surface state–mediated and depletion layer trap–mediated routes are shown for the recombination. E VB e− e−

h+

h+

19

20

1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

couple can interact with both bands [55,
65]. Thus, the oxidation of Eu2+ is a VB process (occurring at p -GaAs but not at nGaAs in the dark), whereas the reduction of Eu3+ (a facile process that reportedly occurs at rates close to the thermionic emission limit, Ref. 55) is mediated by CB electrons [65]. The [Cr(CN)6 ]3−/4− redox system behaves in a similar manner with respect to GaAs [66].
Electroluminescence, (EL), is a versatile probe for studying such carrier injection processes. Thus, hole injection into the VB of a n-type semiconductor leads to cathodic
EL, whereas electron injection into the CB of a p -type semiconductor leads to anodic
EL [67]. Examples of studies of cathodic
EL are commonplace [68–70]; however, anodic EL is not very common because the energy requirement for the redox couple has a very negative redox potential.
Nonetheless, anodic EL has been reported for the p -InP-[Cr(CN)6 ]4− interface [71].
Radical intermediates can also cause EL as discussed later for multielectron redox processes. EL is treated in more depth in another chapter.
This ﬁnally brings us to the comparability of the current-potential behavior

of n-type and p -type samples of a given semiconductor. It may be noted that for a redox process occurring via one of the bands (e.g. VB), the cathodic currents (electron transfer from VB to Ox) are expected to be equal for n- and p -type materials.
This idea has been pursued using the socalled quasi-Fermi level concept [55, 72,
73]. This model has been demonstrated quantitatively by studying the anodic decomposition of GaAs and the oxidation of redox species such as Cu+ and Fe2+ at nand p -type GaAs electrodes [72, 73].
1.4.2

Dark Processes Mediated by Surface States or by Space Charge Layer Recombination

Surface states were considered earlier
(Sect. 1.3.4) from an electrostatic perspective. Here we examine their dynamic consequences. There are two principal charge transfer routes involving surface states. Consider again an n-type semiconductor; the forward-bias current can either involve direct exchange of electrons between the semiconductor CB and Ox states in solution (Fig. 12b) or can be mediated by surface states (Fig. 15). The second

E e− ECB

EF

ESS

Electrolyte

EVB

Fig. 15 Surface state–mediated electron injection from the CB of a n-type semiconductor into the electrolyte. 1.4 Charge Transfer Processes in the Dark

route involves hole injection into the semiconductor VB again from Ox states in solution (Figs. 9c and 14). The recombination current is mediated either by surface states or via space charge layer recombination. We will consider ﬁrst the CB process.
Initial evidence for the intermediacy of surface states came from dark current measurements on n-TiO2 and n-SrTiO3 in the presence of oxidizing agents such as
[Fe(CN]6 3− , Fe3+ , and [IrCl6 ]2− [74, 75].
Similar early evidence that the charge transfer process was more complex than direct transfer of electrons from the

semiconductor CB also came from AC impedance spectroscopy measurements on n-ZnO, n-CdS, and n-CdSe in contact with [Fe(CN)6 ]3− species [76, 77].
The electrochemical impedance for surface state–mediated charge transfer has been computed recently [78]. The key results are summarized in Fig. 16.
Figure 16(a) contains the proposed equivalent circuit for the process and features a parallel connection of the impedance for the Faradaic process [ZF (ω)] (ω = angular frequency, 2πf ) and the capacitance of the semiconductor depletion layer, CSC . The
C1

C2

R1
∆

R2

(a)

Xs / Ω

−1500

−1000

−500

1 Hz
1 kHz

100 kHz

100 mHz
100 Hz

0

1 MHz
0

10 Hz

10 kHz
500

10 mHz
1000

1500

(b)
Fig. 16 Equivalent circuit (a) and a simulated
Nyquist plot (b) for the charge-transfer pathway illustrated in Fig. 15. The capacitance C1 represents that of the space charge layer and the

2000

2500

3000

3500

Rs / Ω parallel branch components represent the
Faradaic charge-transfer process. Refer to the original work for further details. (Reproduced with permission from Ref. 78.)

21

22

1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

former also involves a diffusion impedance
( ) of the Warburg type (see following).
The complex plane (Nyquist) plot predicted for the circuit is illustrated in Fig. 16(b).
The theoretically predicted AC impedance response was compared with experiments on n-GaAs rotating disk electrodes in sulfuric acid media [79]. The equivalent circuit in Fig. 16(a) was also compared with previous versions proposed by other authors [80–83]. These alternate versions differ in their assumption of no variations in potential drop across the Helmholtz region (i.e. inﬁnite CH ) and no concentration polarization in the electrolyte phase
(inﬁnite diffusion coefﬁcient for the redox species). Also discussed is the application of AC impedance spectroscopy for studying the kinetic reversibility of majority carrier charge transfer via the CB of a n-type semiconductor [82].
AC impedance spectroscopy also has seen extensive utility in the study of the hole injection or recombination process depicted in Fig. 14. An equivalent circuit for this process is illustrated in Fig. 17; it does resemble the circuit in Fig. 16(a), except for the Warburg component [84].
Early studies [85–88] utilized the recombination resistance parameter, Rr , that was extracted from model ﬁts of the measured
AC impedance data. This parameter was seen to be inversely related to the hole injection current, thus signifying that it is indeed related to the recombination process.
However, the challenge is to differentiate

whether recombination is mediated via surface states or whether it occurs in the depletion layer. Thus, the parameter Rr alone cannot afford this information and both the real and the imaginary parts of this additional impedance must be considered. Subsequent studies have addressed this aspect [85, 89–93]. The admittance corresponding to recombination at the surface [92] and in the space charge layer [93] was calculated from ﬁrst principles. These computations show that the recombination capacitance increases monotonically with decreasing band bending in the latter case, whereas it shows a peak in the former case as a function of potential.
Experimental studies [91] show that in the case of n-GaAs electrodes in contact with Ce4+ as the hole injection agent, surface recombination prevails. On the other hand, with n-GaP electrodes, recombination in the depletion layer must also be taken into account. Other discussions of the use of AC impedance spectroscopy for the study of hole injection or recombination are contained in Refs. 78 and 84.
The consequences of potential drop variations across the Helmholtz layer in the hole injection process have been examined by a variety of techniques [94, 95]. For example, chemical reaction of the GaAs surface with iodine results in a downward shift of the semiconductor band edge positions such that the reduction of iodine is mediated by CB electrons [95]. When sufﬁcient negative charge accumulates at

Csc

C
R1
R2

Fig. 17 Equivalent circuit representation of the injection or recombination process.
(Reproduced with permission from Ref. 84.)

1.4 Charge Transfer Processes in the Dark

the surface, the potential is redistributed between the semiconductor spacecharge layer and the Helmholtz region. Now iodine is reduced by hole injection as gauged by EL and AC impedance measurements [95].
1.4.3

Rate-limiting Steps in Charge Transfer
Processes in the Dark

The assumption is implicit in the discussion in Sect. 1.4.1 (leading to Eq. 18) that charge transfer kinetics at the semiconductor–electrolyte interface is the ratelimiting step. Fundamentally, we have to differentiate majority carrier capture and minority carrier injection processes in the dark. In the former case, transit through the semiconductor itself or charge exchange with the surface states can be potentially rate-limiting. In the latter case, there are three steps involved: hole injection into the semiconductor VB, charge exchange between the recombination center and the semiconductor CB, and diffusion of majority carriers (electrons) from the neutral region. Finally, mass transport processes in the electrolyte phase itself can be a limiting factor in the overall current ﬂow. We shall examine carrier capture and injection processes in turn.
The vast majority of outer-sphere, nonadsorbing redox systems to date have yielded values for ket in the 10−17 −
10−16 cm4 s−1 range [3, 8]. These include n-Si-CH3 OH [96, 97], n-InP-CH3 OH [98],
GaInP2 -coated n-GaAs-acetonitrile [99], and p -GaAs-HCl [100] interfaces. The redox couples in these studies have mostly involved metallocenes that show low proclivity to adsorb on the semiconductor surface. In these cases, the rate-determining step in the overall current ﬂow undoubtedly lies in the electron transfer event at

the interface itself. However, values for ket approximately three orders of magnitude higher have also been reported for similar interfaces, namely, n-GaAs-acetonitrilecontaining cobaltocenium [Co(Cp)2 + ] acceptors [99, 101]. Similarly, high values were reported for p -GaAs-acetonitrile interfaces with ferricenium and cobaltocenium redox species [102]. In these latter cases, alternative mechanisms (e.g. thermionic emission, see following) must be invoked in a rate-limiting role. Quartz crystal microbalance measurements have yielded evidence for adsorption of redox species (and consequently high local substrate concentration) in some of these
‘‘anomalous’’ instances [101].
In the majority carrier capture process, if the interfacial charge transfer kinetics are facile, the transport of majority carriers through the space charge region can play a rate-limiting role. The thermionic emission theory [34] assumes that every electron that reaches the semiconductor surface, and has the appropriate energy to overcome the potential barrier there, will cross the interface with a tunnel probability of unity. However, if the interfacial kinetics are sluggish, some of the electrons will be reﬂected at the interface. In this case, the exchange current io is no longer described by Eq. (18) but by Eq. (19) [34] io = AA∗

m∗ me T2

ns n (19)

In Eq. (19), A is the electrode area, A∗ is the
Richardson constant (120 A K−2 cm−2 ), m∗ /me is the relative effective electron mass in the CB, and T is the absolute temperature. In many of the reported instances [53,
55, 103], the current calculated from
Eq. (19) is much higher than that measured experimentally, signaling that interfacial charge-transfer kinetics are limiting

23

24

1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

the overall rate. On the other hand, in the nGaAs-acetonitrile-Co(Cp)2 + case [101], AC impedance spectroscopy data appear to support the assumption that thermionic emission is the current-limiting transport mechanism. Another factor that enters into this discussion is the mobility of the majority carriers. It has been argued [14] that in the case of low mobility materials (e.g. µn ∼
1 cm2 V−1 s−1 ), carrier transport from the semiconductor bulk to the interface itself can become limiting. Clearly a multitude of factors are important in majority carrier capture: ket , acceptor concentration in the electrolyte and carrier mobility.
What about the minority carrier injection process depicted in Fig. 14? Here, contrasting with the process considered earlier, the hole injection step is usually very fast (see following). Then the current is limited by diffusion or recombination described by the Shockley equation [104] io =

eo ADp n2 i nLp

(20a)

for bulk recombination, and io = 0.5 eo AW σ νth Nt ni

(20b)

for recombination within the semiconductor space charge region. In Eqs. (20a) and
(20b), Dp is the diffusion coefﬁcient for holes, Lp is the hole diffusion length, ni is the intrinsic carrier concentration, σ is an average capture cross-section for electrons and holes, νth is the thermal velocity of charge carriers, and Nt is the areal density of recombination (trap) centers in the middle of the energy gap (Fig. 14).
The diffusion or recombination mechanism results in considerable overpotential for (cathodic) current ﬂow in the dark
(again assuming a n-type semiconductor for illustration). Such a rate-limiting

process was found to describe the charge transfer at n-GaAs in 6 M HCl containing
Cu+ as the hole injecting species [55, 73].
Whatever the limiting mechanism, ultimately the current becomes limited by concentration polarization, that is, by the transport of redox species from the bulk electrolyte to the semiconductor surface.
The situation in this regard is no different from that at metal electrode–electrolyte interfaces. As in the latter case, hydrodynamic voltammetry is best suited to study mass transport. AC impedance spectroscopy can be another useful tool in this regard [105].
In the former case, the data can be processed via Levich plots of current vs. ω1/2 (ω = angular frequency). If processes other than solution mass transport become rate-limiting, then the plot will show a curvature and the current will even become independent of the electrode rotation rate.
In this case, inverse Levich (or KouteckyLevich) plots of 1/i vs. ω−1/2 can be used for further analyses. Such analyses have been done, for example, for n-GaAsacetonitrile-Co(Cp)2 + interfaces [101] and n- and p -GaAs electrodes in contact with the Cu+/2+ redox couple in HCl electrolyte [55, 73].
The diffusion impedance at semiconductor electrodes has been considered recently [105]. This author described the applicability of AC impedance spectroscopy for the study of electron capture and hole injection processes at n-GaAsH2 O/C2 H5 OH-methyl viologen, p -InPaq. KOH-Fe(CN)6 3 , n-GaAs-H2 SO4 -Ce4+ , and n-InP-aq. KOH-Fe(CN)6 3− interfaces.
In the case of electron capture processes, a
Randles-like equivalent circuit was found to be applicable [105]. On the other hand, no Warburg component was present in the hole injection case when the reverse

1.5 Light Absorption by the Semiconductor Electrode and Carrier Collection

reaction was negligible (Fig. 17). For a nonideal semiconductor–electrolyte contact
(see Sect. 1.3.4), a Warburg impedance appeared in the electrochemical impedance of an injection reaction as well, as exempliﬁed by the n-InP-Fe(CN)6 3− case [105].
1.5

Light Absorption by the Semiconductor
Electrode and Carrier Collection
1.5.1

Light Absorption and Carrier Generation

The optical band gap of the semiconductor (Sect. 2) is an important parameter in deﬁning its light absorption behavior.
In this quantized process, an electronhole pair is generated in the semiconductor when a photon of energy hν
(ν = frequency and hν > Eg ) is absorbed.
Optical excitation thus results in a delocalized electron in the CB, leaving behind a delocalized hole in the VB; this is the band-to-band transition. Such transitions are of two types: direct and indirect. In the former, momentum is conserved and the top of VB and the bottom of CB are both located at k = 0 (k is the electron wave vector). The absorption coefﬁcient (α ) for such transitions is given by [106] α = A (hν − Eg )1/2

(21)

In Eq. (21), A is a proportionality constant. Indirect transitions involve phonon modes; in this case Eq. (21) takes the form α = A (hν − Eg )2

(22)

A given material can exhibit a direct or indirect band-to-band transition depending on its crystal structure. For example, Si single crystals have an indirect transition located at 1.1 eV (Table 1). On the other hand, amorphous Si is characterized by a direct optical transition with a larger Eg

value (shorter wavelengths). Both types of transitions can also be seen in the same material, for example GaP [107].
Within the present context, the important point to note is that the absorption depths (given by 1/α ) are vastly different for direct and indirect transitions. While in the former case absorption depths span the
100–1000 nm range, they can be as large as 104 nm for an indirect transition [9].
Optical transitions in semiconductors can also involve localized states in the band gap. These become particularly important for semiconductors in nanocrystalline form (see following). Sub–band gap transitions can be probed with photons of energy below the threshold deﬁned by Eg .
1.5.2

Carrier Collection

The number of carriers collected (in an external circuit, for example) versus those optically generated deﬁnes the quantum yield ( ) – a parameter of considerable interest to photochemists. The difﬁculty here is to quantify the amount of light actually absorbed by the semiconductor as the cell walls, the electrolyte, and other components of the assembly are all capable of either absorbing or scattering some of the incident light. Unfortunately, this problem has not been comprehensively tackled, unlike in the situation with photocatalytic reactors involving semiconductor particulate suspensions, where such analyses are available [108–111]. Pending these, an effective quantum yield can still be deﬁned.
Returning to the carrier collection problem, consider Fig. 18 for an nsemiconductor–electrolyte interface. As can be seen, the electron-hole pairs are optically generated, both in the ﬁeld-free and in the space charge regions within the semiconductor. Recombination of these

25

26

1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry
1/α

LD + W
W
CB

hν

VB

Semiconductor

Redox 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 the space charge layer, and on the semiconductor surface (the latter in contact with the redox electrolyte). We are assuming here that light is incident from the electrolyte side. Rear illumination geometry can be proﬁtably employed and is considered 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 minority carriers (holes in this case) are swept to the surface and the electrons are driven to the rear ohmic contact. How fast the holes are drained away (by Faradaic reactions involving the redox electrolyte) will dictate how the Fermi levels compare with the equilibrium situation discussed earlier. The departure from equilibrium has been quantiﬁed in terms of the quasiFermi level concept discussed later.
The extent of collection of minority carriers from the region beyond the

depletion layer is dictated by the diffusion process. A diffusion length, L, can be deﬁned Lp =

Dp τp =

k T µp τp

(for n-type semiconductor)

(23)

The subscripts in Eq. (23) remind us that we are dealing with minority carrier collection; µp is the hole mobility and τp is the hole life-time. The characteristic length Lp deﬁnes the region within which electron-hole pair generation is fully effective. Pairs generated at depths longer than the Debye length, LD (LD = W + L) will simply recombine. Thus, the effective quantum yield for a given interface will depend on the relative magnitudes of
LD and the light penetration depth, 1/α
(Fig. 18) [112, 113].

1.5 Light Absorption by the Semiconductor Electrode and Carrier Collection

An expression for the ﬂux of photogenerated minority carriers arriving at the surface was originally given for a solid-state junction [114] and subsequently adapted to semiconductor-liquid junctions [115]. The major weakness of these early models hinges on their underlying assumption for the boundary condition that the surface concentration of minority carriers is zero. As pointed out elsewhere [14], this is a demanding condition necessitating very high magnitudes for the interfacial charge transfer rate constant, ket (see previous section). A modicum of improvement to the basic G¨rtner model was found [116] a by deﬁning a ﬂux rather than a concentration expression for the holes and a characteristic length where the bulk diffusion current transitions into a drift current.
However, this treatment still assumes that every hole entering the depletion layer edge exits this region and out into the electrolyte phase. The G¨rtner equation [114] can be writa ten in normalized form [113]
≡

jph
Io

=1−

exp(−αW )
1 + αLp

and recalling that W is proportional to
VSC 1/2 (Eq. 8), and VSC = V − Vfb , a test of the rudimentary model would lie in a plot of the LHS of Eq. (25) against (Vfb − V )1/2 .
Such plots are shown in Fig. 19 at four selected wavelengths for the p -GaP-H2 SO4 electrolyte interface [117].
While adherence to the G¨rtner model a is satisfactory for large values of VSC (i.e. large band bending, see Fig. 6), the model fails close to the ﬂat band situation. Interestingly, this problem is exacerbated as the semiconductor excitation wavelength becomes shorter (Fig. 19). Thus, another weakness of the basic G¨rtner model [114, a 115] is the neglect of surface recombination. At the ﬂat band situation, this model still predicts ﬁnite current ﬂow arising from the diffusive ﬂow of minority carriers toward and out of the interface (Fig. 20).
A variety of reﬁnements have been made to take into account the surface recombination effect [117–132]. The earliest of these [119, 120, 123] involve some simplifying assumptions:

(24)

In Eq. (24), is the effective quantum yield (see previous section), given by the ratio of the photocurrent density (iph /A), jph , to the incident light ﬂux, Io . Recasting
Eq. (24) in the form
) = ln(1 + αLp ) + αW

(25)

..
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 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 text.) (Reproduced with permission from Ref. 117.)

−ln (1 − Φ)

− ln(1 −

500

0.002

510
0.001
520
530

0

0.5

(Efb −

1.0

E )1/2/V 1/2

1.5

27

1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry
1.0
0.8

(a)

0.6

Current density
[mA cm2]

28

(c)

(b)

0.4
0.2

(d)

0

0.1

0.2

0.3

0.4

Voltage
[V]

−0.2

0.5

Fig. 20 Comparison of calculated current voltage proﬁles in the dark
(curve d) and under illumination
(curves a–c). Curve a is obtained from
..
the basic Gartner model. Curve (b) considers surface recombination and curve (c) considers both surface recombination and recombination in the space charge layer. These simulations are for a n-type semiconductor–electrolyte interface.
0.6 (Reproduced with permission from Ref. 138.)

−0.4

1. There is no recombination in the depletion layer. That is, all the holes optically generated in the bulk and within the depletion layer (Fig. 18) are swept to the surface without loss;
2. The steady state concentration of the optically generated minority carriers does not perturb the potential distribution in the dark (Fig. 6); and
3. There is a quasi-thermodynamic distribution of minority carriers within the depletion layer. This translates to a constant product term np across this region. Surface recombination in the vast majority of these treatments invoke the HallShockley-Read model [133, 134]. Deﬁning the G¨rtner limiting expression (Eq. 24) as a G , we obtain [14] ss =

G

1+

Dp exp VSC
LD (kt + St )

(26)

In Eq. (26), we have two new parameters, kt and St . These are the ﬁrst-order rate constant for hole transfer (units of cm s−1 , see following) and the surface recombination velocity, St . In the combined situation of high LD , high kt , and very low (or zero) St ,
Eq. (26) collapses to the G¨rtner limit. a At this juncture, it is worth noting that only one trap state at the surface has been assumed till now; further it is assumed that this surface state functions both as a carrier recombination site and as a charge-transfer pathway (Fig. 21). Both these assumptions are open to criticism.
An alternative model invoking two distinct types of surface states – one active in recombination and the other capable of mediating charge transfer – has been considered [135]. Nonetheless, the most serious ﬂaw in the above treatments lie in the neglect of carrier recombination in the depletion layer itself (as distinct from recombination at the surface). Reexamination of Eqs. (24) and (26) shows that the larger the Debye length, LD , and the depletion layer width, W , the higher the quantum yield. However, recombination in the space charge layer must become signiﬁcant at some value of W , thus providing a further limit to carrier collection.
Recombination within the space charge region is a nontrivial problem to treat from a computational perspective [14].
The methodology of Sah, Noyce, and
Shockley [136] has been used by several authors [126, 127, 128, 131, 138–140].
Figure 20 illustrates the sensitivity of the

1.5 Light Absorption by the Semiconductor Electrode and Carrier Collection
ECB
e−

ESS hν hν

E VB

h+

Surface state mediation of both minority carrier (i.e. hole) transfer and recombination for a n-type semiconductor electrolyte interface. Fig. 21

current-potential proﬁles at the semiconductor–electrolyte interface to this recombination mode [137].
Other models taking the above nonidealities to varying extent have been proposed; a detailed discussion of these lies beyond the scope of this section [141–147].
However, it is worth noting here that in some instances involving high-quality semiconductor–electrolyte interfaces the rate-determining recombination step does indeed appear to lie in the bulk semiconductor [1, 148]. Silicon photoelectrodes in methanolic media containing fast, oneelectron, outer-sphere redox couples were studied in these cases.
1.5.3

Photocurrent-potential Behavior

The current-voltage characteristics of an illuminated semiconductor electrode in contact with a redox electrolyte can be obtained by simply adding together the majority and minority current components. The majority carrier component is given by the diode equation (Eq. 17) while

the minority carrier current (iph ) is directly proportional to the photon ﬂux (Eq. 24).
Thus, the net current is given by: i = iph − io exp −

eo V kT −1

(27)

The minus sign in Eq. (27) underscores the fact that the majority carrier component ﬂows opposite (or ‘‘bucks’’) the minority carrier current ﬂow. This photocurrent component is shown as curves 2 and 3 in Fig. 13.
Equation (27) shows that the diode equation is offset by the iph term; this is exactly what is seen in Fig. 13. The plateau photocurrent is proportional to the photon ﬂux, Io , as illustrated for two different values of the incident light intensity in
Fig. 13. At the open-circuit condition of the interface, i = 0 (and neglecting the unity term within the square brackets relative to the exponential quantity), Eq. (27) leads to
Voc

iph kT ln eo io

(28)

Equation (28) underlines two important trends: First, Voc increases logarithmically

29

1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

with the photon ﬂux (with a slope of
∼60 mV at 298 K). Second, Voc decreases with an increase of io (again logarithmically). This underlines the importance of ensuring that the majority carriers do not
‘‘leak’’ through the interface. Because of the diode nature of the interface, from a device perspective, the semiconductor surface must be designed to have fast minority carrier transfer kinetics (and thus high iph ), but must be blocking to the ﬂow of majority carriers (from the CB for a n-type semiconductor) into the redox electrolyte. This challenge is similar to what the solid-state device physicists face, but relative to metals (with a high density of acceptor states), chemical control of redox electrolytes offers a powerful route to performance optimization of liquid-based interfaces as also pointed out by previous authors [1, 6, 8, 149–154].
Referring back to Fig. 13, the currentpotential curves under illumination of the semiconductor simply appear shifted
‘‘up’’ relative to the dark i − V counterpart. This, however, is the ideal scenario.
Anomalous photoeffects (APEs) are often observed that manifest as a crossover of light and dark current-voltage

curves, as illustrated in Fig. 22. Thus, the superposition principle [149–154] is not obeyed in this instance. The dashed line in Fig. 22 is produced by translating the photocurrent-voltage data by jSC , the short-circuit current density. If the superposition principle is held, this dashed curve would have overlaid the dark currentvoltage curve. Thus, this ‘‘excess’’ (forward bias) current embodies the APE, and the failure of superposition is quantiﬁed as the voltage difference ( V ) between the dark j − V data and the dashed line.
What are the ramiﬁcations of the crossover or the APE? First, mathematical modeling of carrier transport in a junctionbased solar photovoltaic system, according to
(27a)
j = jSC − jbk (V ) is not valid in the presence of this effect.
(In Eq. (27a), jbk is the ‘‘bucking’’ current density in the forward-bias regime, see previous section.) That is, a fully linearized system of differential equations and boundary conditions cannot be used to model the interface carrier transport.
Second, computation and modeling of the open-circuit voltage for such devices by

Light

Current density

30

∆V
Dark

Potential

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 the text.

1.5 Light Absorption by the Semiconductor Electrode and Carrier Collection

simply equating a constant photocurrent ﬂux, jph against the dark (recombination) current, jo is no longer possible
(see Eq. (28) and the accompanying earlier discussion). Third, and perhaps practically of most signiﬁcance, the V component represents a loss pathway in the photovoltage deliverable by the given device. Thus, the
(open-circuit) photovoltage is Voc instead of Voc + V in the ideal case in the absence of the APE. Therefore, it is important to quantify and understand the molecular and chemical origins of this effect. This has not been done so far, at least to this writer’s knowledge, for semiconductor–electrolyte interfaces. Of course, the cross-over effect is not conﬁned to such interfaces. It is interesting that a recent textbook [155], dealing primarily with solid-state solar cell devices, makes only a ﬂeeting reference to the underlying origin of the APE. Reference was made in this book to the cross-over of experimental dark and light j − V characteristics for a Cu2 S-CdS solidstate heterojunction solar cell but its origin was not explored. A light-induced junction modiﬁcation has also been reported for the
(Cd, Zn)S-CuInSe2 solid-state system [156,
157]. The cross-over effect appears to have been treated in even lesser depth in some classical textbooks on semiconductor devices [104, 155].
Probably the ﬁrst reported instance of observation of an APE was in 1977 for a n-TiO2 -NaOH electrolyte interface [158].
The APE was observed in the saturation region of cathodic current ﬂow and was induced by sub–band gap irradiation of the photoanode. A peak in the spectrum of the photoresponse at 800 nm (the corresponding photon energy being lower than the 3.0 eV band gap of TiO2 ) was used by the authors to invoke a surface

state–mediated electron transfer to O2 (in the electrolyte) as the origin of the photoeffect. Surface states were again invoked to explain a cathodic photoeffect at nCdS-aqueous polysulﬁde interfaces [159].
This photoeffect was only observed for the
(0001) single-crystal face of n-CdS but not for the (0001) orientation. A subsequent study of photoelectrochemical effects at selenium ﬁlms reported an anomalous anodic photocurrent at potentials positive of the ﬂat band location for the p -type ﬁlm [160]. This effect was assigned to a hole injection process via a tunneling mechanism. An increase of the tunneling probability under illumination was accommodated by a shrinking of the space charge layer at the interface. Photoenhancement of the forward current ﬂow was observed again for n-CdS, in this instance in contact with a [Fe(CN)6 ]3−/4− redox electrolyte [161]. This effect was observed only with a mechanically damaged surface, and disappeared after it had been etched with concentrated HCl.
Subsequent work [162] describes suppression of the cathodic photocurrent for n-CdS-[Fe(CN)6 ]3−/4− interfaces mechanical polishing of the electrode. As with an earlier study [163], the spectral dependence of the photocathodic effect implicated sub–band gap states. The suppression was explained by two alternative models involving a compensated insulating layer or by Fermi-level pinning.
Illumination was claimed to result in a dramatic increase of the (suppressed) cathodic current, which interestingly enough was observed only for n-CdS ﬁlms but not for crystals including n-CdTe, n-CdSe, nGaAs, n-ZnO, n-TiO2 and n-ZnSe. On the other hand, a subsequent paper describes a photocathodic effect for n-CdSe-sulﬁde interfaces in which an interfacial layer of selenium was implicated [163].

31

1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry
Fig. 23 Experimental data embodying
APE for the n-GaAs-AlCl3 /n-butylpyridinium chloride molten-salt electrolyte interface. Refer to the text for details.
(Reproduced with permission from
Ref. 167.)

5 × 100

5 × 10−1

Current
[mA cm−2]

32

5 × 10−2

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] and n-TiO2 photoanodes [166] implicate the formation of photoconductive layers in the APEs. Thus, the foregoing review suggests the following
1. APE is a very general phenomenon that has been observed for solid-state junctions for n- and p -type semiconductors alike, and for a wide variety of semiconductor materials.
2. The reported results and trends are often contradictory. It is quite possible that the experimental conditions in these studies were quite variant, thus precluding direct comparison of the results. 3. The mechanistic reasons given for the APE are possibly many, and generalizations may not be warranted.
Clearly, more research is needed on this topic.

Figure 23 contains an example of the
APE for the n-GaAs-AlCl3 /n-butylpyridinium chloride molten salt electrolyte interface [167]. The bottom curve in Fig. 23 is the measured dark current-voltage proﬁle.
The top curve was obtained from the photocurrent-voltage data (under irradiation of the semiconductor). Clearly, if the superposition principle held, the two curves would have coincided with one another. APEs have also been observed for nanocrystalline and chemically modiﬁed ﬁlms, as discussed in a subsequent section.
Light-induced changes in the electrostatics at the semiconductor–electrolyte interface are conveniently probed by capacitance-voltage measurements in the dark and under illumination of the semiconductor electrode. If charge trapping at the interface plays a decisive role
(whatever be the mechanism), the voltage

1.5 Light Absorption by the Semiconductor Electrode and Carrier Collection

drop across the illuminated interface is altered, and consequently the semiconductor band edge positions are shifted.
This, of course, is the Fermi-level pinning situation that was encountered earlier
(Sect. 1.3.4). Examples of studies addressing this aspect may be found, for example on p – GaAs [124] and CdTe [168,
169]–based aqueous electrolyte interfaces.
1.5.4

Dynamics of Photoinduced Charge Transfer

So far the discussion has centered on the steady state aspects of carrier generation and collection at semiconductor–electrolyte interfaces. As with their metal electrode counterparts, a wealth of information can be gleaned from perturbation-response type of measurements. An important difference, however, lies in the vastly different timescale windows that are accessible in the two cases.
The critical RC time-constant of the cell in a transient experiment is given by τcell = C(Rm + Rel )

(29)

In Eq. (29), Rm is the measurement resistor (across which the current or photocurrent is measured) and Rel is the electrolyte resistance. The term C is the capacitance, which, in the metal case, is the
Helmholtz layer capacitance, CH . (Once again, the Gouy region is ignored here.)
For semiconductor–electrolyte interfaces, we have seen that two layers are involved in a series circuit conﬁguration with corresponding capacitances of CSC and
CH (Fig. 6). Because CH
CSC , C
CSC . This assumption is usually justiﬁed because CH 10−5 F cm−2 and CSC =
10−8 − 10−0 F cm−2 . If the composite resistance (Rm + Rel ) is 100 ohm, then τcell for metal electrodes is ∼10−3 s and that for the semiconductor case is
10−6 − 10−7 s.

What are the processes important in a dynamic interrogation of the semiconductor–electrolyte interface?
1. Carrier generation within the semiconductor,
2. diffusion of minority carriers from the ﬁeld-free region to the space charge layer edge,
3. transit through this layer,
4. charge transfer across the interface, and
5. carrier recombination via surface states or via traps in the space charge layer.
Other phenomena such as thermalization also are important as discussed later in the context of hot carrier effects. The time constant (τcell ) of the cell and the measurement circuitry has complicated matters further and have caused some confusion in the interpretation of transient data. If a potentiostat is not used, then this time constant is given by Eq. (29).
One can envision three types of perturbation: an inﬁnitesimally narrow light pulse (a Dirac or δ -functional), a rectangular pulse (characteristic of chopped or interrupted irradiation), or periodic (usually sinusoidal) excitation. All three types of excitation and the corresponding responses have been treated on a common platform using the Laplace transform approach and transfer functions [170]. These perturbations refer to the temporal behavior adopted for the excitation light.
However, classical AC impedance spectroscopy methods employing periodic potential excitation can be combined with steady state irradiation (the so-called PEIS experiment). In the extreme case, both the light intensity and potential can be modulated (at different frequencies) and the (nonlinear) response can be measured at sum and difference frequencies.
The response parameters measured in all these cases are many but include

33

34

1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

the photocurrent, voltage, luminescence, or microwave conductivity. Clearly, semiconductor–electrolyte interfaces present a rich, albeit demanding landscape for probing non–steady state phenomena.
The dynamics of charge transfer across semiconductor–electrolyte interfaces are considered in more detail elsewhere in this volume.
1.5.5

Hot Carrier Transfer

Short wavelength photons (of energy much greater than Eg ) create ‘‘hot’’ carriers. If, somehow, thermalization of these carriers can be avoided, photoelectrochemical reactions that would otherwise be impossible with the ‘‘cooled’’ counterparts, that is, at very negative potentials for n-type semiconductors, would be an intriguing possibility. The key issue here is whether the rate of electron transfer across the interface can exceed the rate of hot electron cooling. The observation of hot carrier effects at semiconductor–electrolyte interfaces is a controversial matter [3, 7, 11, 171] and practical difﬁculties include problems with band edge movement at the interface and the like [4]. Under certain circumstances (e.g. quantum-well electrodes, oxide ﬁlm-covered metallic electrodes), it has been claimed that hot carrier transfer can indeed be sustained across the semiconductor–electrolyte interface [7, 172, 173].
1.6

Multielectron Photoprocesses

This section has thus far considered redox electrolytes comprising one electron oxidizing or reducing agents. Multielectron redox processes, however, are important in a variety of scenarios. Consider the reduction of protons to H2 (HER) – a

technologically important electrochemical process that has also been extensively studied from a mechanistic perspective on metallic electrodes.
Photoelectrolytic processes such as HER can be carried out on semiconductor electrodes. One can envision a HER mechanism on a p -type semiconductor of the sort: hν − → CB p − SC − − e− + h+
VB
−→ e− + S − − S−
CB
−

(30a)

+

−→
S +H −− S+H

•

−→ e− + H+ − − H•
CB
+

H +H

•

+ e−
CB
−

(30)

− − H2
−→

(30b)
(30c)
(30d)

−→
H+ + H• + S − − H2 + S

(30e)

h+
VB
h+
VB

(30f)

+H −− H
−→
•

+ S− − − S
−→

+

(30g)

In this above scheme, S denotes a surface state and both direct (Reactions
30c and 30d) and indirect (i.e. surface state–mediated) (Reactions 30b and 30e) radical and H2 -generating pathways are shown. Reactions 30f and 30g represent recombination routes involving the reaction intermediates. Admittedly, this scheme is daunting in its complexity, and the kinetics implications are as yet unclear. Early studies on p -GaP, p -GaAs, and other Group III–V
(13–15) semiconductors reported onset of cathodic photocurrents (attributable to HER) only at potentials far removed
(ca. 0.6 V) from Vfb [174]. This was attributed to Steps 30b and 30g in the preceding scheme. More recent work [175] has shown that the HER at illuminated p -InP-electrolyte contacts is accompanied by a photocorrosion reaction, leading to indium formation on the semiconductor surface. 1.6 Multielectron Photoprocesses

Interestingly, surface states themselves were chemically identiﬁed with H• (adads sorbed hydrogen atom intermediates) in the aforementioned study [166]. These species have also been implicated in accumulation layer formation and anodic
EL at n- and p -GaAs-electrolyte interfaces [176–178].
Another interesting characteristic of many multiequivalent redox systems is the phenomenon of photocurrent multiplication. This phenomenon may be illustrated for two systems utilizing illuminated n-type and p -type semiconductors respectively n-type hν n-SC − − e− + h+
−→

(31)

−→
HCOOH + h+ − − COOH• + H+
(31a)
−→
COOH• − − CO2 + H+ + e−

(31b)

p-type hν p-SC − − h+ + e−
−→
+

−

−→
O2 + H + e − − HO2

(31)
•

(31c)

−→
HO2 • + H+ − − H2 O2 + h+ (31d)
Thus, the key feature of photocurrent multiplication is a majority carrier injection step (Reactions 31b or 31d) from a reaction intermediate (usually a free radical) into the semiconductor CB or VB, respectively.
In the preceding examples, each photon generates two carriers in the external circuit, affording a quantum yield (in the ideal case) of 2. This is the ‘‘current-doubling’’ process. Practically, however, quantum yields somewhat lower than 2 are usually measured because Steps 31b or 31d compete with the further photooxidation or photoreduction of these intermediates, respectively. This is true especially at high

photon ﬂux values. Even multiplication factors as high as 4 are possible as in the photodissolution of n-Si in NH4 F media [179–182].
Photocurrent multiplication has been observed for a variety of semiconductors including Ge [180], Si [179–182],
ZnO [183–189], TiO2 [190–193], CdS [194,
195], GaP [196], InP [197], and GaAs [198–
200]. These studies have included both n- and p -type semiconductors, and have spanned a range of substrates, both organic and inorganic. As in the Si case, this phenomenon can also be caused by photodissolution reactions involving the semiconductor itself. The earlier studies have mainly employed voltammetry, particularly hydrodynamic voltammetry [193].
As more recent examples [2, 9, 10] reveal, intensity-modulated photocurrent spectroscopy, (IMPS), is also a powerful technique for the study of photocurrent multiplication. This leads us to another important category of multielectron photoprocesses involving the semiconductor itself. While photocorrosion is a nuisance from a device operation perspective, it is an important component of a device fabrication sequence in the microelectronics industry.
Two types of wet etching of semiconductors can be envisioned [201]. Both occur at open-circuit but one involves the action of chemical agents that cause the simultaneous rupture and formation of bonds.
Several aspects of photoetching have been reviewed [62–64, 202, 203] including reaction mechanisms, morphology of the etched surfaces, and etching kinetics in the dark and under illumination. General rules for the design of anisotropic photoetching systems have also been formulated [204]. Photoelectrochemical etching is considered in more detail elsewhere in this volume.

35

36

1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

1.7

Nanocrystalline Semiconductor Films and
Size Quantization
1.7.1

Introductory Remarks

From a materials perspective, the ﬁeld of semiconductor electrochemistry and photoelectrochemistry has evolved from the use of semiconductor single crystals to polycrystalline thin ﬁlms and, more recently, to nanocrystalline ﬁlms. The latter have been variously termed membranes, nanoporous or nanophase ﬁlms, mesoporous ﬁlms, nanostructured ﬁlms, and so on; they are distinguished from their polycrystalline electrode predecessors by the crystallite size (nm versus µm in the former) and by their permeability to the electrolyte phase. These ﬁlms are referred to as ‘‘nanocrystalline in the following sections. These features render three-dimensional geometry to nanocrystalline ﬁlms as opposed to the ‘‘ﬂat’’ or two-dimensional (planar) nature of single crystal or polycrystalline counterparts.
What are the virtues of these emerging photoelectrode materials? The ﬁrst is related to their enormous surface area.
Consider that the 3D structure is built up of close-packed spheres of radius, r . Then ignoring the void space, the speciﬁc area,
As (area/volume) is given by 3/r [205].
For r = 10 nm, As is on the order of
106 cm−1 , and for a 1 cm2 ﬁlm of 1 µm thickness, this value corresponds to an internal surface area of ∼100 cm2 (i.e. a ‘‘surface roughness factor’’ of 100).
Clearly, this becomes important if we want the electrolyte redox species to be adsorbed on the electrode surface (see following). Alternatively, a large amount of sensitization dye can be adsorbed onto the support semiconductor although this dye sensitization approach is not considered

in this chapter. By way of contrast, the amount of species that can be conﬁned in a monolayer on a corresponding ﬂat surface would be negligibly small.
As we shall see later, electron transport in nanocrystalline ﬁlms is necessarily accompanied by charge-compensating cations because the holes are rapidly injected into the ﬂooded electrolyte phase.
This provides opportunities for studying ion transport processes in mesoporous media that are coupled to electron motion. Ion insertion also has practical consequences as in energy storage device applications [206].
Surface state densities on the order of
∼1012 cm2 are commonplace for semiconductor electrodes of the sort considered in previous sections of this chapter.
These translate to equivalent volume densities of ∼1018 cm−3 for nanocrystalline ﬁlms. Such high densities enhance light absorption by trapped electrons in surface states, giving rise to photochromic and electrochromic effects [198–200] (see following). Unusually high photocurrent quantum yields are also observed with sub–band gap light with these photoelectrode materials. Corresponding sub–band gap phenomena are rather weak and difﬁcult to detect with single-crystal counterparts.
1.7.2

The Nanocrystalline Film–Electrolyte
Interface and Charge Storage Behavior in the Dark

Understanding of the electrostatics across nanocrystalline semiconductor ﬁlm-electrolyte junctions presents interesting challenges, particularly from a theoretical perspective. Concepts related to space charge layers, band bending, ﬂat band potential, and the like (Sect. 3) are not applicable here because the crystallite dimensions

1.7 Nanocrystalline Semiconductor Films and Size Quantization

comprising these layers are comparable to
(or even smaller than) nominal depletion layer widths.
The rather complete interpenetration of the electrode and electrolyte phases must mean that the Helmholtz double layer extends throughout the interior surface of the nanoporous network, much like a supercapacitor [9, 210] situation. Finally, unlike in the cases treated earlier, the semiconductor (especially if it is a metal oxide) is not heavily doped such that free majority carriers are not present in appreciable amounts.
This is indicated, for example, by the sensitivity of the conductivity of nanocrystalline
TiO2 layers to UV light – the conductivity is strongly enhanced on UV exposure, similar to a photoconductive effect. This effect has been interpreted, in terms of trap ﬁlling with recombination times considerably slower than the trapping processes under reverse bias [211, 212]. The light sensitivity also is diagnostic of the fact that the low electronic conductivity in the dark is not due to high interparticle resistances (i.e. in the ‘‘neck’’ regions), but rather is indicative of the low electron concentrations.
The electron concentration can be increased by forward-biasing the nanocrystalline electrode–electrolyte interface potentiostatically. The interface is driven thus into the accumulation regime for the majority carriers, and if a transparent rear contact (e.g. F-doped, SnO2 or Sn-doped indium oxide) is used, the resultant blue
(or bluish-black) coloration of the ﬁlm can be spectroscopically monitored [208,
209, 213]. Whether the optical response arises from CB electrons or from electrons trapped in surface states is not entirely clear. It has been claimed [214] that the absorption spectrum of the latter differs signiﬁcantly from CB electrons. Electrons in surface states can be chemically identiﬁed with Ti3+ defect sites that can be

detected, for example, by electron paramagnetic resonance spectroscopy [215, 216].
In either case, the resultant negative charge generated by electron accumulation at the internal surfaces has to be balanced by cations (from the electrolyte phase) for charge compensation. Such ion insertion reactions have been studied using techniques such as voltammetry, reﬂectance or absorption spectroscopy, chronoamperometry, and electrochemical quartz-crystal microgravimetry [213,
217–22]. Both aqueous and aprotic electrolytes have been deployed for these studies. Unlike in the single-crystal cases treated earlier, placement of the semiconductor energy band positions at the interface via Mott-Schottky analyses is not straightforward for nanocrystalline ﬁlms. Abrupt changes in slope and other nonidealities [215, 227, 223] have been observed, for example, in the Mott-Schottky plots for
TiO2 ﬁlms and attributed to the inﬂuence of the conductive glass that is normally employed to support these ﬁlms. This behavior is especially prevalent under reverse bias. The onset of majority carrier optical absorption (in the visible and near
IR range) under forward-bias instead has been proﬁtably employed to place the CB positions of TiO2 in aqueous media [208].
Impedance spectroscopy and electrochemical dye desorption experiments have been employed [224] to study the electrical characteristics of TiO2 nanocrystalline ﬁlms in the dark. This study as well as the others cited earlier demonstrate how the conductivity changes (as a result of electron injection from the support electrode) can cause the porous or nanocrystalline layer to manifest itself electrically, such that the active region moves away (i.e. outward) from the support as the forwardbias voltage is increased. The potential

37

38

1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

distribution has also been analyzed depending on whether the depletion layer width exceeds or is smaller than the typical dimension of the structural units in the nanocrystalline network [223].
1.7.3

Photoexcitation and Carrier Collection:
Steady State Behavior

Figure 24 contains a schematic representation of the nanocrystalline semiconductor ﬁlm–electrolyte interface at equilibrium
(Fig. 24a) and the corresponding situation under band gap irradiation of the semiconductor (Fig. 24b) [9]. Because the diffusion length of the photogenerated carriers is usually larger than the physical dimensions of the structural units, holes and electrons can reach the impregnated electrolyte phase before they are lost via bulk recombination. This contrasts the situation with the single-crystal cases discussed earlier. If, as is the case with TiO2 nanocrystalline ﬁlms, the holes are rapidly scavenged

by the electrolyte redox (speciﬁcally Red) species, collection of the photogenerated electrons at the rear contact becomes the determinant factor in the quantum yield. Thus, the quasi-Fermi level for holes remains close to EF,redox and that for electrons, EF,e moves ‘‘up’’, as depicted in Fig. 24b. Illumination thus induces an electron ﬂux, Jn (x) through the nanocrystalline phase. Under steady state conditions, the photocurrent density (jph ) is equal to eo Jn (x = d). The driving force for electron diffusion through the network of nanocrystallites has been calculated from ﬁrst principles [225]. It has been found that the driving force is approximately kT /eo divided by the thickness of the network.
Importantly, this free-energy gradient is found to be independent of the incident photon ﬂux.
It is important to reiterate that the charge separation in a nanocrystalline semiconductor–electrolyte interface does not depend on a built-in electric ﬁeld at the junction as in the single-crystal

E
EC
E F,redox

EV

Substrate
(a)

d

E

−

je(d )

0

x

−

−

E F,e

+
Red

Substrate

d
(b)

Ox
0

x

Schematic representation of a nanocrystalline semiconductor– electrolyte interface in the dark (a) and under illumination from the electrolyte side (b). Ec and Ev correspond to ECB and EVB in our notation. (Reproduced with permission from the authors of
Ref. 9.)

E F,redox Fig. 24

1.7 Nanocrystalline Semiconductor Films and Size Quantization

case. Instead, the differential kinetics for the reactions of photogenerated electrons and holes with electrolyte redox species account for the charge separation (and the generated photovoltage). The molecular factors underlying the sluggish scavenging of electrons at the nanocrystalline ﬁlmelectrolyte boundary (by the redox species) are as yet unclear. Clearly, the competition between surface recombination of these electrons (with the photogenerated holes) and collection at the rear contact dictates the magnitude of the quantum yield that is experimentally measured for a particular junction. Photocurrent losses have been recorded for electrolytes dosed with electron acceptors such as O2 and iodine [226]. Nanocrystalline TiO2 electrodes with thicknesses ranging from 2 µm to 38 µm were included in this study. In the presence of these electron-capture agents, electron collection (i.e. photocurrent) at the rear contact was seriously compromised. On the other hand, as high as 10% of the photons were converted to current for a 38 µm thick ﬁlm in a N2 -purged solution [226].
The result was obtained with front-side illumination geometry. As one would intuitively expect, carrier collection is most efﬁcient close to the rear contact. Indeed, marked differences have been observed for photoaction spectra with the two irradiation (i.e. through the electrolyte side vs. through the transparent rear contact) geometries for TiO2 , CdS, and CdSe nanocrystalline ﬁlms [227, 228].
Obviously, the relative magnitudes of the excitation wavelength and the ﬁlm thickness critically enter into this variant behavior. In the vast majority of cases, the iodide/triiodide redox couple has been employed (presumably because of its success in shuttling the photooxidized dye in the

sensitization experiments) although other redox electrolytes [e.g. SCN− /(SCN)2 − ;
228] have been employed as well. For the chalcogenide ﬁlms, sodium selenosulﬁte was employed [227]. It must be noted that, aside from losses caused by the surface recombination and back-reactions, an additional loss component from the increase in ﬁlm resistance must also be recognized, especially as the ﬁlm thickness is increased. The resistance loss manifests as a deterioration in the photovoltage and ﬁll factor.
In the discussion to this point, we have not considered trapping or release of the photogenerated electrons as they undergo transit to the rear contact. However, electrons trapped in localized interfacial states induce a countercharge in the Helmholtz double-layer, as discussed in the preceding discussion. The resultant voltage drop can introduce a nonnegligible ﬁeld component into the diffusional process. The time-dependence of the electron density, n(x, t) is given by [9]
∂ 2 n(x, t)
∂n(x, t)
= ηαIo e−αx + Dn
∂t
∂x 2 n(x, t) − no
−
(32) τ In Eq. (32), η is the electron injection efﬁciency, Dn is the diffusion coefﬁcient of electrons, and τ is the pseudo ﬁrstorder lifetime of electrons determined by back-reaction with Ox.
Even if the migration component is negligible (but see following), solution of
Eq. (32) presents difﬁculties because of the possible dependence of Dn on n and x .
Similarly, τ may depend on these two variables also. Nonetheless, the steady state solution of Eq. (32) has been obtained [229] by assuming that D and τ are constant and that η = 1. Under these conditions,

39

1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

the photocurrent is predicted to be independent of voltage – a rather physically implausible situation. In the forward-bias regime, η is expected to decrease and the back-reaction of the photogenerated electrons (with Ox) can no longer be neglected.
This brings us to the rear support-ﬁlm interface. What sort of barrier exists at this junction? Are the electron exchange kinetics voltage-dependent at this interface?
The effect of changing the work function of the substrate on the current-voltage curves
(in the dark and under illumination) has been investigated for TiO2 nanocrystalline ﬁlms [230]. The onset potential for the photocurrent is found to be the same regardless of whether SnO2 , Au, or Pt is used to support the ﬁlm. A Fermi level pinned rear interface was used to explain the results.
In general, the voltammograms for nanocrystalline electrodes are similar to what is observed for their single-crystal counterparts. An example of a photovoltammogram for CdS is contained in Fig. 25. The fact that a S-type proﬁle is observed culminating in a photon ﬂux-limited plateau regime is rather surprising given that (1) the ﬁlm is rather

insulating and (2) the electrolyte permeates into the network and possibly contacts the rear support electrode. The transition in the proﬁles from spiked at potentials near Von (photocurrent onset) to more rectangular at positive bias potentials (not shown in Fig. 25) must mean that the voltage does exert an effect on carrier transit through the network. No satisfactory explanation appears to exist at present to resolve this apparent anomaly.
1.7.4

Photoexcitation and Carrier Collection:
Dynamic Behavior

In this section, we brieﬂy consider the response of nanocrystalline semiconductor–electrolyte interfaces to either pulsed or periodic photoexcitation. Several points are noteworthy in this respect: (1) The photocurrent rise-time in response to an illumination step is nonlinear. Further, the response is faster when the light intensity is higher. (2) The decay proﬁles exhibit features on rather slow timescales extending up to several seconds. (3) The photocurrent decay transients exhibit a peaking behavior. The time at which this peak

2 µA

40

−0.8

−0.6

−0.4

−0.2

Potential
[V vs. Ag/AgCl]

0

Fig. 25 Photovoltammogram
0.2 under interrupted illumination

of a nanocrystalline CdS-sodium sulﬁte electrolyte interface in the reverse-bias regime.

1.8 Chemically Modiﬁed Semiconductor–Electrolyte Interfaces

occurs varies with the square of the ﬁlm thickness, d . (4) A similar pattern is also seen in IMPS data where the transit time, τ is seen to be proportional to d 2 .
These observations have been interpreted within the framework of two distinct models, one involving trapping or detrapping of the photogenerated electrons [231, 232] and the other based on electron diffusion (or ﬁeld-assisted diffusion) not attenuated by electron localization [233, 234]. The millisecond transit times also mean that the transit times are very long compared to equilibration of majority carriers in a bulk semiconductor or electron-hole pair separation within the depletion layer of a ﬂat electrode. The slow transport is rationalized by a weak driving force and by invoking percolation effects [223].
It is interesting that the response patterns differ for different nanocrystalline electrodes [223]. For example, while trapping or detrapping effects appear to be relatively unimportant for GaP, the response for TiO2 , especially at low photon ﬂuxes, is governed by electron trapping or detrapping kinetics. This accounts for the faster response at higher photon ﬂuxes
(see preceding section).
1.7.5

Size Quantization

When electronic particles such as electrons and holes are constrained by potential barriers to regions of space that are comparable to or smaller than their de Broglie wavelength, the corresponding allowed energy states become discrete (i.e. quantized) rather than continuous. This manifests in the absorption (or emission) spectra as discrete lines that are reminiscent of atomic (line) transitions; these sharper features often appear superimposed on a broader envelope. Another manifestation

for semiconductors is that the energy band gap (Eg ) increases, or equivalently, the absorption threshold exhibits a blue shift.
The critical dimension for size quantization effects to appear in semiconductors depends on the effective mass (m∗ ) of the electronic charge carriers. For m∗ ∼ 0.05,
˚
the critical dimension is about 300 A; it decreases approximately linearly with increasing m∗ [7].
Size quantization effects and quantum dot photoelectrochemistry are discussed in more detail elsewhere in this volume.
1.8

Chemically Modiﬁed
Semiconductor–Electrolyte Interfaces
1.8.1

Single Crystals

Much of the research in the early
1980s on chemically modiﬁed semiconductor–electrolyte interfaces was directed toward protecting them from photocorrosion; this body of work has been reviewed [226]. Parallel efforts also went into improving minority carrier transfer at the interface by chemisorbing metal ions such as Ru3+ on the semiconductor surface. Chemical agents such as sulﬁde ions are known to passivate the semiconductor against surface recombination [6].
A study [22a] on electron transfer dynamics at p -GaAs-acetonitrile interfaces where the semiconductor surface was sulﬁdepassivated exempliﬁes this fact. In general, the mechanistic issue of whether these chemical agents improve minority carrier charge transfer by minimizing surface recombination or by a true catalytic action has not been completely resolved [1].
Yet another tactic involves perturbing the electrostatics at the semiconductor–electrolyte interface by adsorbing

41

42

1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry
A summary of approachesa to chemical modiﬁcation of semiconductor–electrolyte interfaces

Tab. 2

Modiﬁcation agent(s) Semiconductor(s)

Ru3+
Co3+
Os3+
Ag+
S2−
HS−
Thiolates
Dithiocarbamate
Lewis acids
Lewis bases
Cl−
Benzoic acid derivatives
Noble metals
Noble metals
RuO2 c
RuO2 c
Ptc
Ptc
Noble metals
Noble metalsd

Modiﬁcation objectiveb Sample
Reference(s)

n-GaAs n-GaAs n-GaAs p-InP p-GaAs
CdS, CdSe
CdS, CdSe
CdS, CdSe
CdS, CdSe
CdS, CdSe n-GaAs n+ -GaAs n-TiO2 p-InP n-CdS n-Si p-Si n-CdS n-CdS p-InP

A
A
A
A
A
B
B
B
B
B
B
B
C
C
C
C
C
C
C
C

239, 240
241
241
242
236
243
244
245, 246
247, 248
249
238
250
251
239, 252–254
255, 256
257
257, 258
259
254
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 donors or acceptors on the semiconductor surface [237]. In favorable cases, this increases the band bending at the interface by thus introducing a ﬁxed countercharge of opposite polarity (negative for a n-type semiconductor) at the junction. Chloride ion adsorption on the n-GaAs surface from ambient temperature AlCl3 /n-butylpyridinium chloride melts [30, 238] is a case in point; this process serves to improve the junction and the photovoltage that it delivers. Of

course, such ‘‘ﬁxed-charge’’ effects have long been known to the solid-state device physics and gas phase catalysis communities. Other agents that have been deployed for chemical tuning of the interfacial energetics at the semiconductor–electrolyte interface are listed in Table 2.
Native semiconductor surfaces are fairly inactive from a catalysis perspective.
Thus, noble metal or metal oxide islands have been implanted on photoelectrode surfaces as electron storage centers to drive multielectron redox processes such

1.8 Chemically Modiﬁed Semiconductor–Electrolyte Interfaces

as HER, photooxidation of H2 O, and photooxidation of HCl, HBr, or HI.
Examples of this sort of chemical modiﬁcation strategy are also contained in Table 2.
The advent of self-assembled monolayer (SAM) ﬁlms on electrode surfaces has rendered a high degree of molecular order and predictability to the chemical modiﬁcation approach. In particular, the use of these insulating, molecular spacers enables interrogation of critical issues in electron transfer such as the inﬂuence of chemical bonding and distance between the support electrode and the redox moieties on the rate constant for electron transfer. Many such studies on gold-conﬁned SAMs have appeared recently [261–263]. Corresponding studies on semiconductor surfaces (particularly
Group II–V compounds such as GaAs and
InP [264–266] and elemental semiconductors such as Si [267]) have also begun to appear. Alkanethiol-based or alkylsiloxane-based
SAMs have been proﬁtably employed in all these instances to probe the distance effect in electron transfer dynamics. The thiolbased SAMs have the virtue that the spacer length can be predictably altered simply by varying the number of methylene units in the chain. The distance dependence of ket is embodied in the parameter β , the decay coefﬁcient. For a critical discussion of the subtleties involved in the extraction and interpretation of this parameter, we refer to Ref. 262. A value of
0.49 ± 0.07 has been reported for this parameter for n-InP-alkanethiol-ferrocyanide interfaces [266]. This value is smaller than its counterpart for corresponding ﬁlms on gold surfaces, which range from ∼0.6 to
1.1 per methylene unit. The reason for this difference is not entirely clear, although several hypotheses were advanced by the authors [266].

1.8.2

Nanocrystalline Semiconductor Films and
Composites

Dye sensitization of nanocrystalline semiconductor ﬁlms certainly represents one popular approach to chemical modiﬁcation of the interface. However, this topic is covered in detail elsewhere in this volume. Other examples, from a non–dye sensitization perspective, are less common but two recent studies are noted [268,
269]. One utilizes the surface afﬁnity of TiO2 toward suitably derivatized viologens to construct chemically modiﬁed nanocrystalline ﬁlms suitable for displays, electrochromic (smart) windows, sensors, and the like [268]. In the other study [269], the TiO2 ﬁlm surface was modiﬁed with phosphotungstic acid (PWA). This compound belongs to a family of polyoxometallates that exhibit interesting electronand proton-transfer or storage properties and also high thermal stability [269]. Thus, these modiﬁed ﬁlms would be applicable in areas such as catalysis, sensors, electronics, and even medicine.
These TiO2 -PWA ﬁlms represent a logical bridge connecting single-phase semiconductor ﬁlms and multicomponent composite systems. Of course, highly evolved multicomponent assemblies occur in nature and there is no better example than the plant photosynthetic system.
The plant photosynthetic architecture contains synergistic components (e.g. lightharvesting antennae, membranes) each with a well-deﬁned and complementary function, to convert the incident photon energy, to move electrons vectorially, and to store the reaction products. The design and implementation of artiﬁcial analogs have proved to be a daunting task, both from a synthetic and characterization perspective. While this topic is covered

43

44

1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry

elsewhere in this series of volumes, we brieﬂy discuss in what follows, some simple multicomponent assemblies based on semiconductors.
Early examples in the 1980s were aimed at the design of composite systems for photoelectrolytic generation of H2 . Thus,
Naﬁon and SiO2 were used as supports for coprecipitated ZnS and CdS for photoassisted HER from aqueous sulﬁde media [270]. Subsequent work has addressed the mechanistic role of the support in the photoassisted HER [271]. Vectorial electron transfer was demonstrated in bipolar
TiO2 /Pt or CdSe/CoS photoelectrode panels arranged in series arrays for the photodecomposition of water to H2 and
O2 [272, 273].
More recently, matrix-semiconductor composites, that is, ﬁlms comprising of semiconductor particles that are dispersed in a nonphotoactive continuous matrix have been developed. Examples of matrix candidates are metals and polymers [274–279]. Occlusion electrosynthesis is a versatile method for preparing such composite ﬁlms as exempliﬁed by the
Ni/TiO2 and Ni/CdS family [280–282].
Matrix-semiconductor composite ﬁlms have two virtues from a photoelectrochemical perspective. First, their components can be separately chosen and optimized for a speciﬁc function. Thus, the matrix component can be chosen to have good adsorption tendency toward a targeted substrate. The semiconductor component then functions in the role of photogenerating charge carriers either for reducing or oxidizing this sequestered substrate. This photocatalytic strategy has been recently demonstrated both for organic substrates (methanol and formate ion) [283, 284] and an inorganic substrate
(sulﬁte) [285]. The net result in either case is an enhanced photocatalytic performance

of the composite because of the high local concentration of the substrate resulting from the matrix adsorption process. In principle, high surface area supports of the sort that are normally used in the gas-phase catalysis community can also be used in conjunction with TiO2 [286, 287]. These would include materials such as Al2 O3 ,
SiO2 , or diatomaceous earth. The resultant composite ﬁlms, however, cannot be used as electrodes because of their poor electronic conductivity.
The second important feature of a metal-semiconductor composite approach is that the metal can function as a template for chemical or electrochemical derivatization to afford a ﬁlm comprising molecular redox-semiconductor (or even semiconductor-semiconductor) contacts. Figure 26 generically illustrates the occlusion electrosynthesis approach for preparing M/TiO2 composite ﬁlms and a subsequent derivatization with ferri/ ferrocyanide to afford the corresponding metal hexacyanoferrate (MHCF)/TiO2 counterparts [288]. These chemically modiﬁed ﬁlms exhibit interesting ‘‘bipolar’’ photoelectrochemical behavior [289] and photoelectrochromic properties [290].
1.9

Types of Photoelectrochemical Devices

As Fig. 27 illustrates, there are basically three types of photoelectrochemical devices for solar energy conversion.
The ﬁrst type is regenerative in nature and the species that are photooxidized at the n-type semiconductor electrode are simply re-reduced at the counterelectrode (Fig. 27a). Instead of an electrocatalytic electrode [291, 292] where the counterelectrode reaction occurs in the dark (this is the situation schematized

1.9 Types of Photoelectrochemical Devices
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 photo-cathode may also be deployed in a tandem regenerative cell. In all these cases, the cells operate in the photovoltaic mode where the input photon energy is converted into electricity.
Interesting enough, it is the second type of device, namely a photoelectrolytic cell
(Fig. 27b), that ﬁrst caught the attention of a scientiﬁc and technological community in the 1970s that was searching

for alternative energy sources to fossilderived fuels. Thus in a landmark paper,
Fujishima and Honda [293] demonstrated that sunlight could be used to drive the photoelectrolysis of water using an n-TiO2 photoanode and a Pt counterelectrode.
Unfortunately, the requirements for efﬁciently splitting water are rather stringent, as discussed elsewhere in this volume.
In the third type of energy conversion device, the initial photoexcitation does not

45

46

1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry e− e−

Fig. 27 Types of photoelectrochemical devices for solar energy conversion. (a),
(b), and (c) depict regenerative, photoelectrolytic, and dye-sensitized conﬁgurations, respectively. As in the remainder of this chapter, an n-type semiconductor is assumed in these cases for speciﬁcity.

e−

hν

Ox
Red

(a)

h+
Semiconductor

Electrolyte

Metal

e− e− hν
Red

(b)

h+

e−

Ox′
Red′

Ox e− e−

D ∗/D +

e−
Ox

D +/D Red
D
Ox
(c)

D+

Red

occur in the semiconductor (unlike in the device counterparts in Figs. 27a and b) but occurs instead in a visible light-absorbing dye (Fig. 27c). Subsequent injection of an electron from the photoexcited dye into the semiconductor CB results in the ﬂow of a current in the external circuit. Sustained conversion of light energy is facilitated by regeneration of the reduced form of the dye via a reversible redox couple (e.g. iodide/triiodide) [294].
Therefore, this device, as its counterpart in
Fig. 27(a), also operates in a photovoltaic mode, or perhaps more appropriately, in a photogalvanic mode.

Other variants of the three types of device operation may be envisioned for semiconductor-liquid junctions. Thus, in the photoelectrolytic mode, the cell reaction clearly is driven (by light) in the contrathermodynamic direction, that is, G > 0.
However, there are many instances, involving, for example, the photooxidation of organic compounds in which light merely serves to accelerate the reaction rate. Thus these cells operate in the photocatalytic mode. In fact, aqueous suspensions comprising irradiated semiconductor particles may be considered to be an assemblage of short-circuited microelectrochemical cells operating in the photocatalytic mode.
Finally, a storage electrode may be incorporated even in a regenerative photoelectrochemical cell of the sort schematized in
Fig. 27(a). Thus, when the sun is shining, this storage electrode is ‘‘charged’’; in the dark, energy may be tapped (as from a battery) from this storage electrode [295–298].
Further details of these device types as well as nonenergy-related applications of photoelectrochemical cells (such as in environmental remediation) may be found in the chapters that follow in this volume.
1.10

Conclusion

In this introductory chapter, we have discussed the electrostatics of the semiconductor–liquid interface considering both single crystals as well as their nanocrystalline counterparts. The charge

1.10 Conclusion

transfer dynamics across both these types of interfaces have been described in the dark and under photoexcitation of the semiconductor. Finally, the various types of photoelectrochemical devices for solar energy conversion are introduced. Subsequent chapters in this volume provide further elaboration of some of these topics considered herein.
Acknowledgments

Research in the author’s laboratory on semiconductor electrochemistry and photoelectrochemistry since 1995 is funded, in part, by the Ofﬁce of Basic Energy Sciences, the US Department of Energy. A number of talented and dedicated coworkers and colleagues have been involved in collaborative research with the author over the past twenty years; their names appear in the publications cited from this laboratory. I also thank the University of Texas at Arlington for providing the facilities and infrastructure. I am grateful to Prof.
S. Licht for comments on an earlier version of the manuscript. Last but not least,
I thank Ms. Gloria Madden and Ms. Rita
Anderson for assistance in the preparation of this chapter.
References
1. N. S. Lewis, Acc. Chem. Res. 1990, 23, 176.
2. L. M. Peter, Chem. Rev. 1990, 90, 753.
3. N. S. Lewis, Annu. Rev. Phys. Chem. 1991,
42, 543.
4. C. A. Koval, J. N. Howard, Chem. Rev. 1992,
92, 411.
´
5. A. Kumar, W. C. A. Wilisch, N. S. Lewis,
Crit. Rev. Solid State Mater. Sci. 1993, 18,
327.
6. M. X. Tan, P. E. Laibinis, S. T. Nguyen et al.,
Prog. Inorg. Chem. 1994, 41, 21.
7. A. J. Nozik, R. Memming, J. Phys. Chem.
1996, 100, 13061.
8. N. S. Lewis, J. Phys. Chem. 1998, 102, 4843.

9. L. M. Peter, D. Vanmaekelbergh, Adv. Electrochem. Sci. Eng. 1999, 6, 77.
10. L. M. Peter in Comprehensive Chemical Kinetics, (Eds.: R. G. Compton, G. Hancock),
Elsevier, Amsterdam, 1999, pp. 223–280.
11. R. J. D. Miller, G. McLendon, A. J. Nozik et al., in Surface Electron-transfer Processes,
VCH Publishers, New York, 1995.
12. K. Rajeshwar, L. M. Peter, A. Fujishima et al., Eds., in Photoelectrochemistry Proc.,
The Electrochemical Society, Pennington,
NJ, 1997, Vol. 20.
13. N. Sato in Electrochemistry at Metal and
Semiconductor Electrodes, Elsevier, Amsterdam, 1998.
14. A. Hamnett in Comprehensive Chemical
Kinetics (Eds.: R. G. Compton), Elsevier,
Amsterdam, 1987, p. 61, Vol. 27.
15. J. S. Newman in Electrochemical Systems,
2nd edition, Prentice-Hall, Englewood
Cliffs, NJ, 1991, p. 496.
16. R. A. Smith in Semiconductors, Cambridge
University Press, Cambridge, 1964.
17. F. A. Kr¨ ger, H. J. Vink in Solid State Physics o (Eds.: F. Seitz, D. Turnbull), Academic
Press, New York, 1956, Vol. 3.
18. F. A. Kr¨ ger in The Chemistry of Imperfect o Crystals, North Holland, Amsterdam, 1964.
19. N. B. Hannay, Ed., in Semiconductors, Reinhold, New York, 1959.
20. A. R. West in Solid State Chemistry and Its
Applications, John Wiley, New York, 1984.
21. P. A. Cox in The Electronic Structure and
Chemistry of Solids, Oxford University Press,
Oxford, England, 1987.
22. R. Hoffmann in Solids and Surfaces, VCH,
New York and Weinheim, 1988.
23. K. Uosaki, H. Kita in Modern Aspects of
Electrochemistry (Eds.: R. E. White, J. O’M.
Bockris, B. E. Conway), Plenum, New York,
1986, pp. 1–60, Vol. 18.
24. S. Trasatti in The Absolute Electrode Potential:
An Explanatory Note, IUPAC Commission
I.3 (Electrochemistry), 1984.
25. H. Reiss, J. Electrochem. Soc. 1978, 125, 937.
26. Yu. V. Pleskov, Yu. Ya. Gurevich in Semiconductor Photoelectrochemistry, Consultants
Bureau, New York, 1986.
27. H. Gerischer, Adv. Electrochem. Eng. 1961,
1, 139.
28. H. Gerischer in Physical Chemistry: An Advanced Treatise (Eds.: H. Eyring, D. Henderson, W. Jost), Academic Press, New York,
1970, Vol. 9A.

47

48

1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry
29. S. R. Morrison in The Chemical Physics of
Surfaces, Plenum, New York and London,
1977.
30. P. Singh, R. Singh, R. Gale et al., J. Appl.
Phys. 1980, 51, 6286.
31. K. Rajeshwar in Molten Salt Techniques
(Eds.: R. J. Gale, D. G. Lovering), Plenum,
New York and London, 1984, pp. 221–252,
Vol. 2.
32. R. Thapar, K. Rajeshwar, Electrochim. Acta
1983, 28, 198.
33. C. M. Gronet, N. S. Lewis, G. W. Cogan et al., J. Electrochem. Soc. 1984, 131, 2873.
34. E. H. Rhoderick in Metal-Semiconductor
Contacts, Clarendon Press, Oxford, 1980.
35. S. Kar, K. Rajeshwar, P. Singh et al., Sol.
Energy 1979, 23, 129.
36. G. Horowitz, P. Allongue, H. Cachet,
J. Electrochem. Soc. 1984, 131, 2563.
37. A. J. Bard, A. B. Bocarsly, F. R. F. Fan et al.,
J. Am. Chem. Soc. 1980, 102, 3671.
38. A. Many, Y. Goldstein, N. B. Grover in
Semiconductor Surfaces, North Holland, Amsterdam, 1965.
39. G. A. Somorjai in Introduction to Surface
Chemistry and Catalysis, John Wiley, New
York, 1994.
40. A. W. Adamson, A. P. Gast in Physical
Chemistry of Surfaces, John Wiley, New York,
1997.
41. R. L. Van Meirhaeghe, F. Cardon, W. P.
Gomes, J. Electroanal. Chem. 1985, 188, 287.
42. J. Bardeen, Phys. Rev. 1947, 71, 717.
43. A. M. Cowley, S. M. Sze, J. Appl. Phys. 1965,
36, 3212.
44. S. Kurtin, T. C. McGill, C. A. Mead, Phys.
Rev. Lett. 1969, 22, 1433.
45. T. C. McGill, J. Vac. Sci. Technol. 1974, 11,
935.
46. L. J. Brillson, Phys. Rev. Lett. 1978, 40, 260.
47. L. J. Brillson, J. Vac. Sci. Technol. 1979, 16,
1137.
48. W. E. Spicer, P. W. Chye, P. R. Skeath et al.,
J. Vac. Sci. Technol. 1979, 16, 1422.
49. G. Nogami, J. Electrochem. Soc. 1982, 129,
2219.
50. K. Uosaki, H. Kita, J. Electrochem. Soc. 1983,
130, 895.
51. R. de Gruyse, W. P. Gomes, F. Cardon et al., J. Electrochem. Soc. 1975, 125, 711.
52. J. C. Tranchart, L. Hollan, R. Memming,
J. Electrochem. Soc. 1978, 125, 1185.
53. S. R. Morrison, Surf. Sci. 1969, 15, 363.

54. R. A. L. Vanden Berghe, F. Cardon, W. P.
Gomes, Surf. Sci. 1973, 39, 368.
55. D. Meissner, R. Memming, Electrochim.
Acta 1992, 37, 799.
56. H. Gerischer, Ber. Bunsen-Ges. Phys. Chem.
1965, 69, 578.
57. R. Memming, G. Schwandt, Electrochim.
Acta 1968, 13, 1299.
58. M. J. Madou, F. Cardon, W. P. Gomes, Ber.
Bunsen-Ges. Phys. Chem. 1977, 81, 1186.
59. H. H. Goossens, W. P. Gomes, J. Electrochem. Soc. 1991, 138, 1696.
60. D. Vanmaekelbergh, L. S. Yun, W. P.
Gomes et al., J. Electroanal. Chem. 1987,
221, 187.
61. J. J. Kelly, J. E. A. M. van den Meerakker,
P. H. L. Notten et al., Philips Tech. Rev. 1988,
44, 61.
62. P. H. L. Notten, J. E. A. M. van den Meerakker, J. J. Kelly in Etching of III-V Semiconductors: An Electrochemical Approach, Elsevier, Oxford, 1991.
63. H. H. Goossens, W. P. Gomes, Electrochim.
Acta 1992, 37, 811.
64. W. P. Gomes, H. H. Goossens in Adv. Electrochem. Sci. Eng. (Eds.: H. Gerischer,
C. W. Tobias), VCH, Weinheim, 1994,
Vol. 3.
65. D. Meissner, C. Sinn, R. Memming et al., in Homogeneous Photocatalysis (Eds.:
E. Pelizetti, N. Serpone, D. Reidel), Dordrecht, 1986.
66. J. E. A. M. van den Meerakker, J. Electroanal. Chem. 1988, 243, 161.
67. B. Pettinger, H-R. Sch¨ ppel, H. Gerischer, o Ber. Bunsen-Ges. Phys. Chem. 1976, 80, 849.
68. K. Uosaki, Trends Anal. Chem. 1990, 9, 98.
69. K. Uosaki, H. Kita, Ber. Bunsen-Ges. Phys.
Chem. 1984, 91, 447.
70. A. Manivannan, A. Fujishima, J. Lumin.
1988, 42, 43.
71. G. Oskam, E. A. Meulenkamp, J. Electroanal. Chem. 1992, 326, 213.
72. R. Reincke, R. Memming, J. Phys. Chem.
1992, 96, 1310.
73. R. Reineke, R. Memming, J. Phys. Chem.
1992, 96, 1317.
74. J. Vandermolen, W. P. Gomes, F. Cardon,
J. Electrochem. Soc. 1980, 127, 324.
75. P. Salvador, C. Gutierrez, J. Electrochem.
Soc. 1984, 131, 326.
76. V. A. Tyagai, G. Ya. Kolbasov, Surf. Sci.
1971, 28, 423.

1.10 Conclusion
77. W. P. Gomes, F. Cardon, Ber. Bunsen-Ges.
Phys. Chem. 1970,74, 431.
78. Z. Hens, J. Phys. Chem. B 1999, 103, 122.
79. Z. Hens, W. P. Gomes, J. Phys. Chem. B
1999, 103, 130.
80. F. Cardon, Physica 1972, 57, 390.
81. G. Horowitz, J. Electroanal. Chem. 1983,
159, 421.
82. D. Vanmaekelbergh, Electrochim. Acta 1997,
42, 1135.
83. Z. Hens, W. P. Gomes, Electrochim. Acta
1998, 43, 2577.
84. W. P. Gomes, D. Vanmaekelbergh, Electrochim. Acta 1996, 41, 967.
85. P. A. Allongue, H. Cachet, J. Electrochem.
Soc. 1985, 132, 45.
86. J. E. A. M. Van der Meerakker, J. J. Kelly,
P. H. L. Notten, J. Electrochem. Soc. 1985,
132, 638.
87. K. Schr¨ der, R. Memming, Ber. Bunsen-Ges. o Phys. Chem. 1985, 89, 385.
88. D. Vanmaekelbergh, W. P. Gomes,
F. Cardon, Ber. Bunsen-Ges. Phys. Chem.
1986, 90, 431.
89. D. Vanmaekelbergh, W. P. Gomes,
F. Cardon, Ber. Bunsen-Ges. Phys. Chem.
1985, 89, 994.
90. D. Vanmaekelbergh, W. P. Gomes,
F. Cardon, J. Electrochem. Soc. 1987, 134,
891.
91. D. Vanmaekelbergh, R. P. ter Heide,
W. Kruijt, Ber. Bunsen-Ges. Phys. Chem.
1989, 93, 1103.
92. D. Vanmaekelbergh, F. Cardon, J. Phys. D:
Appl. Phys. 1986, 19, 643.
93. D. Vanmaekelbergh, F. Cardon, Semicond.
Sci. Technol. 1988, 3, 124.
94. J. J. Kelly, P. H. L. Notten, Electrochim. Acta
1984, 29, 589.
95. J. E. A. M. van den Meerakker, Electrochim.
Acta 1985, 30, 435.
96. A. M. Fajardo, N. S. Lewis, Science 1996,
274, 969.
97. A. M. Fajardo, N. S. Lewis, J. Phys. Chem. B
1997, 101, 11136.
98. K. E. Pomykal, N. S. Lewis, J. Phys. Chem. B
1997, 101, 2476.
99. A. Meier, S. S. Kocha, M. C. Hanna et al.,
J. Phys. Chem. B 1997, 101, 7038.
100. I. Uhlendorf, R. Reincke-Koch, R. Memming, J. Phys. Chem. 1996, 100, 4930.
101. A. Meier, D. C. Selmarten, K. Siemoneit et al., J. Phys. Chem. B 1999, 103, 2122.

102. Y. Rosenwaks, B. R. Thacker, R. K. Ahrenkiel et al., J. Phys. Chem. 1992, 96, 10096.
103. H. Gerischer, J. Phys. Chem. 1991, 95, 1356.
104. S. M. Sze in Physics of Semiconductor Devices, John Wiley, New York, 1981.
105. Z. Hens, W. P. Gomes, J. Phys. Chem. B
1997, 101, 5814.
106. J. L. Pankove in Optical Processes in Semiconductors, Prentice Hall, Englewood Cliffs,
NJ, 1971.
107. D. Vanmaekelbergh, B. H. Erne, C. W. Cheung et al., Electrochim. Acta 1995, 40, 686.
108. I. Rosenberg, J. R. Brock, A. Heller, J. Mol.
Catal. 1992, 96, 3423.
109. J. Valladares, J. R. Bolton in Photocatalytic
Puriﬁcation and Treatment of Water and
Air (Eds.: D. F Ollis, H. Al-Ekabi), Elsevier,
Amsterdam, 1993, p. 111.
110. M. I. Cabrera, O. M. Alfano, A. E. Cassano,
J. Phys. Chem. 1996, 100, 20043.
111. C. A. Martin, M. A. Baltanas, A. E. Casano,
Environ. Sci. Technol. 1996, 30, 2355.
112. K. Rajeshwar, Spectroscopy 1993, 8, 16.
113. L. M. Peter in Comprehensive Chemical Kinetics (Eds.: R. G. Compton), Elsevier, Amsterdam, 1989, pp. 353–383, Vol. 29.
114. W. W. G¨rtner, Phys. Rev. 1959, 116, 84. a 115. M. A. Butler, J. Appl. Phys. 1977, 48, 1914.
116. V. A. Tyagai, Russ. J. Phys. Chem. 1965, 38,
1335.
117. J. Li, L. M. Peter, J. Electroanal. Chem. 1984,
165, 41.
118. L. M. Peter, J. Li, R. Peat, J. Electroanal.
Chem. 1984, 165, 29–41.
119. R. H. Wilson, CRC Crit. Rev. Solid State
Mater. Sci. 1980, 10, 1.
120. R. H. Wilson, J. Appl. Phys. 1977, 48, 4297.
121. K. Rajeshwar, J. Electrochem. Soc. 1980, 129,
1003.
122. J. F. McCann, D. Haneman, J. Electrochem.
Soc. 1982, 129, 1134.
123. R. Memming, Surf. Sci. 1964, 1, 88.
124. J. J. Kelly, R. Memming, J. Electrochem. Soc.
1982, 129, 730.
125. S. U. M. Khan, J. O’M. Bockris, J. Phys.
Chem. 1984, 88, 2504.
126. F. E. Guibaly, K. Colbow, B. L. Funt, J. Appl.
Phys. 1981, 52, 3480.
127. F. E. Guibaly, K. Colbow, J. Appl. Phys.
1982, 53, 1737.
128. F. E. Guibaly, K. Colbow, J. Appl. Phys.
1983, 54, 6488.
129. J-N. Chazalviel, J. Electrochem. Soc. 1982,
129, 963.

49

50

1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry
130. M. Nishida, J. Appl. Phys. 1980, 51, 1669.
131. F. E. Guibaly, K. Colbow, Can. J. Phys. 1981,
59, 1682.
132. D. L. Ullman, J. Electrochem. Soc. 1981, 128,
1269.
133. S. Ramakrishna, S. K. Rangarajan,
J. Electroanal. Chem. 1991, 308, 49.
134. W. Shockley, W. T. Read, Phys. Rev. 1952,
87, 835.
135. R. N. Hall, Phys. Rev. 1982, 87, 387.
136. S. J. Fonash in Solar Cell Device Physics, Academic Press, San Diego, 1981, Chapter 6, pp. 262–326.
137. C. T. Sah, R. N. Noyce, W. Shockley, Proc.
IRE 1957, 45, 1228.
138. J. Reichman, Appl. Phys. Lett. 1980, 36, 574.
139. W. I. Albery, P. N. Bartlett, A. Hamnett et al., J. Electrochem. Soc. 1981, 128, 1492.
140. W. J. Albery, P. N. Bartlett, J. Electrochem.
Soc. 1983, 130, 1699.
141. H. Reiss, J. Electrochem. Soc. 1978, 125, 937.
142. J. Thomchick, A. M. Buoncristiani, J. Appl.
Phys. 1980, 51, 6265.
143. W. Lorenz, M. Handschuh, J. Electroanal.
Chem. 1980, 111, 181.
144. W. Lorenz, C. Aegerter, M. Handschuh,
J. Electroanal. Chem. 1987, 221, 33.
145. M. Handschuh, W. Lorenz, J. Electroanal.
Chem. 1988, 251, 1.
146. P. Lemasson, A. Etcheberry, J. Gautron,
Electrochim. Acta 1982, 27, 607.
147. S. Ramakrishna, S. K. Rangarajan,
J. Electroanal. Chem. 1994, 369, 289.
148. M. L. Rosenbluth, C. M. Lieber, N. S. Lewis,
Appl. Phys. Lett. 1984, 45, 423.
149. S. Licht, Nature 1987, 330, 148.
150. S. Licht, D. Peramunage, Nature 1990, 345,
330.
151. S. Licht, N. Myung, J. Electrochem. Soc.
1995, 142, 840.
152. S. Licht, N. Myung, J. Electrochem. Soc.
1995, 142, 845.
153. S. Licht, F. Forouzan, J. Electrochem. Soc.
1995, 192, 1539.
154. S. Licht, D. Peramunage, J. Phys. Chem. B
1996, 100, 9082.
155. H. J. Hovel in Semiconductors and Semimetals, Solar Cells, Academic Press, New York,
1975, Vol. 11.
156. R. R. Potter, J. R. Sites, IEEE Trans. Electron
Devices 1984, ED-31, 571.
157. R. R. Potter, J. R. Sites, Appl. Phys. Lett.
1983, 43, 843.

158. H. Morisaki, M. Hariya, K. Yazawa, Appl.
Phys. Lett. 1977, 30, 7.
159. H. Minoura, M. Tsuiki, Chem. Lett. 1978,
205.
160. W. Gissler, J. Electrochem. Soc. 1980, 127,
1713.
161. B. Vainas, G. Hodes, J. DuBow, J. Electroanal. Chem. 1981, 130, 391.
162. N. M¨ ller, G. Hodes, B. Vainas, J. Elecu troanal. Chem. 1984, 172, 155.
163. R. Tenne, W. Giriat, J. Electroanal. Chem.
1985, 186, 127.
164. D. P. Amalnerkar, S. Radhakrishnan,
H. Minoura et al., Sol. Energy Mater. 1988,
18, 37.
165. D. P. Amalnerkar, S. Radhakrishnan,
H. Minoura et al., J. Electroanal. Chem.
1989, 260, 433.
166. S.-E. Lindquist, H. Vidarsson, J. Mol. Catal.
1986, 38, 131.
167. P. Singh, R. Singh, K. Rajeshwar et al.,
J. Electrochem. Soc. 1981, 128, 1145.
168. D. Lincot, J. Vedel, J. Electroanal. Chem.
1987, 220, 179.
169. D. Lincot, J. Vedel, J. Phys. Chem. 1988, 92,
4103.
170. L. M. Peter in Photocatalysis and the Environment (Ed.: M. Schiavello), Kluwer Academic Publishers, Dordrecht, 1988, p. 243,
Vol. 237.
171. B. Ba, H. Cachet, B. Fotouhi et al., Electrochim. Acta 1992, 37, 309.
172. Y-E. Sung, F. Galliard, A. J. Bard, J. Phys.
Chem. B 1998, 102, 9797.
173. Y-E. Sung, A. J. Bard, J. Phys. Chem. B 1998,
102, 9806.
174. M. P. Dare-Edwards, A. Hamnett, J. B.
Goodenough, J. Electroanal. Chem. 1981,
119, 109.
175. J. Schefold, J. Phys. Chem. 1992, 96, 8692.
176. K. Uosaki, H. Kita, J. Am. Chem. Soc. 1986,
108, 4294.
177. S. Kaneko, K. Uosaki, H. Kita, J. Phys. Chem.
1986, 90, 6654.
178. K. Uosaki, Y. Shigematsu, S. Kaneko et al.,
J. Phys. Chem. 1989, 93, 6521.
179. M. Matsumura, S. R. Morrison, J. Electroanal. Chem. 1983, 144, 113.
180. M. Matsumura, S. R. Morrison, J. Electroanal. Chem. 1983, 147, 157.
181. H. J. Lewerenz, J. Stumper, L. M. Peter,
Phys. Rev. Lett. 1988, 61, 1989.
182. L. M. Peter, A. N. Borazio, H. J. Lewerenz et al., J. Electroanal. Chem. 1990, 290, 229.

1.10 Conclusion
183. S. R. Morrison, T. Freund, J. Chem. Phys.
1967, 47, 1543.
184. W. P. Gomes, T. Freund, S. R. Morrison,
Surf. Sci. 1968, 13, 201.
185. W. P. Gomes, T. Freund, S. R. Morrison,
J. Electrochem. Soc. 1968, 115, 818.
186. K. Micka, H. Gerischer, J. Electroanal.
Chem. 1972, 38, 397.
187. J.-S. Lee, T. Kato, A. Fujishima et al., Bull.
Chem. Soc. Jpn. 1984, 57, 1179.
188. G. H. Schoenmakers, D. Vanmaekelbergh,
J. J. Kelly, J. Phys. Chem. 1996, 100, 3215.
189. G. H. Schoenmakers, D. Vanmaekelbergh,
J. J. Kelly, J. Chem. Soc., Faraday Trans.
1997, 93, 1127.
190. M. Miyake, H. Yoneyama, H. Tamura,
Chem. Lett. 1976, 635.
191. N. Hykaway, W. M. Sears, H. Morisaki et al., J. Phys. Chem. 1986, 90, 6663.
192. D. J. Fermin, E. A. Ponomarev, L. M. Peter, in Photoelectrochemistry, Proc., (Eds.: K.
Rajeshwar, L. M. Peter, A. Fujishima et al.),
The Electrochemical Society, Pennington,
NJ, 1997, Vol. 97–20.
193. G. Nogami, J. H. Kennedy, J. Electrochem.
Soc. 1989, 136, 2583.
194. W. J. Albery, N. L. Dias, C. P. Wilde, J. Electrochem. Soc. 1987, 134, 601.
195. P. Herrasti, L. M. Peter, J. Electroanal.
Chem. 1991, 305, 241.
196. J. Li, L. M. Peter, J. Electroanal. Chem. 1985,
182, 399.
197. B. H. Ern´ , D. Vanmaekelbergh, I. E. Vere meir, Electrochim. Acta 1993, 38, 2559.
198. J. Li, R. Peat, L. M. Peter, J. Electroanal.
Chem. 1986, 200, 333.
199. R. Peat, L. M. Peter, Electrochim. Acta 1986,
31, 731.
200. R. Peat, L. M. Peter, J. Electroanal. Chem.
1986, 209, 307.
201. D. Vanmaekelbergh, J. J. Kelly, J. Phys.
Chem. 1990, 94, 5406.
202. W. P. Gomes, S. Lingier, D. Vanmaekelbergh, J. Electroanal. Chem. 1989, 269, 237.
203. R. Tenne in Semiconductor Micromachining:
Fundamentals Electrochemistry and Physics
(Eds.: S. A. Campbell, H.-J. Lewerenz), John
Wiley, New York, 1994, pp. 139–175, Vol. 1.
204. J. van de Ven, H. J. P. Nabben, J. Electrochem. Soc. 1990, 137, 1603.
205. G. A. Somorjai in Chemistry in Two Dimensions: Surfaces, Cornell University Press,
Ithaca, NY, 1981.

206. S. Y. Huang, L. Kavan, I. Exnar et al., J. Electrochem. Soc. 1995, 142, L142.
207. J. G. Highﬁeld, M. Gr¨tzel, J. Phys. Chem. a 1988, 92, 464.
208. B. O’Regan, M. Gr¨tzel, D. Fitzmaurice, a Chem. Phys. Lett. 1991, 183, 89.
209. B. O’Regan, M. Gr¨tzel, D. Fitzmaurice, a J. Phys. Chem. 1991, 95, 10525.
210. B. E. Conway, J. Electrochem. Soc. 1991, 138,
1539.
211. R. K¨ nenkamp, R. Henninger, P. Hoyer, o J. Phys. Chem. 1993, 97, 7328.
212. R. K¨ nenkamp, R. Henninger, Appl. Phys. o A 1994, 58, 87.
213. I. Bedja, S. Hotchandani, P. V. Kamat,
J. Phys. Chem. 1996, 100, 19489.
214. G. Boschloo, D. Fitzmaurice, J. Phys. Chem.
B 1999, 103, 2228.
215. F. Cao, G. Oskam, P. C. Searson et al.,
J. Phys. Chem. 1995, 99, 11974.
216. R. F. Howe, M. Gr¨tzel, J. Phys. Chem. a 1985, 89, 4495.
217. L. A. Lyon, J. T. Hupp, J. Phys. Chem. 1995,
99, 15718.
218. H. Lindstr¨ m, S. S¨ dergren, A. Solbrand o o et al., J. Phys. Chem. B 1997, 101, 7710.
219. H. Lindstr¨ m, S. S¨ dergren, A. Solbrand o o et al., J. Phys. Chem. B 1997, 101, 7717.
220. S. Lunell, A. Stashans, L. Ojam¨e et al., a J. Am. Chem. Soc. 1997, 119, 7374.
221. L. A. Lyon, J. T. Hupp, J. Phys. Chem. B
1999, 103, 4623.
222. R. van de Krol, A. Goossens, J. Schoonman,
J. Electrochem. Soc. 1997, 144, 1723.
223. J. J. Kelly, D. Vanmaekelbergh, Electrochim.
Acta 1998, 43, 2773.
224. A. Zaban, A. Meier, B. A. Gregg, J. Phys.
Chem. B 1997, 101, 7985.
225. D. Vanmaekelbergh, P. E. de Jongh, J. Phys.
Chem. B 1999, 103, 747.
226. H. Rensmo, H. Lindstr¨ m, S. S¨ dergren o o et al., J. Electrochem. Soc. 1996, 143, 3173.
227. G. Hodes, I. D. J. Howell, L. M. Peter,
J. Electrochem. Soc. 1992, 139, 3136.
228. A. Hagfeldt, U. Bj¨ rkst´ n, S.-E. Lindquist, o e
Sol. Energy Mater. Sol. Cells 1992, 27, 293.
229. S. S¨ dergren, A. Hagfeldt, J. Olsson et al., o J. Phys. Chem. 1994, 98, 5552.
230. A. Shiga, A. Tsujiko, T. Ide et al., J. Phys.
Chem. B 1998, 102, 6049.
231. K. Schwarzburg, F. Willig, Appl. Phys. Lett.
1991, 58, 2520.
232. P. E. de Jongh, D. Vanmaekelbergh, Phys.
Rev. Lett. 1996, 77, 3427.

51

52

1 Fundamentals of Semiconductor Electrochemistry and Photoelectrochemistry
233. A. Solbrand, A. Henningson, S. S¨ dergren o et al., J. Phys. Chem. B 1999, 103, 1078.
234. A. Solbrand, H. Lindstr¨ m, H. Rensmo o et al., J. Phys. Chem. B 1997, 101, 2514.
235. K. Rajeshwar, J. Appl. Electrochem. 1985, 15,
1.
236. Y. Rosenwaks, B. R. Thacker, R. K. Ahrenkiel et al., J. Phys. Chem. 1992, 96, 10096.
237. S. Licht, V. Marcu, J. Electroanal Chem.
1986, 210, 197.
238. P. Singh, K. Rajeshwar, J. Electrochem. Soc.
1981, 128, 1724.
239. A. Heller, Acc. Chem. Res. 1981, 14, 154.
240. B. A. Parkinson, A. Heller, B. Miller, Appl.
Phys. Lett. 1978, 33, 521.
241. B. J. Tufts, I. L. Abrahams, L. G. Casagrande et al., J. Phys. Chem. 1989, 93, 3260.
242. A. Heller, H. J. Leamy, B. Miller et al.,
J. Phys. Chem. 1983, 87, 3239.
243. K. Uosaki, Y. Shigematsu, H. Kita et al.,
Anal. Chem. 1989, 61, 1980.
244. M. J. Natan, J. W. Thackeray, M. S.
Wrighton, J. Phys. Chem. 1986, 90, 4089.
245. J. W. Thackeray, M. J. Natan, P. Ng et al.,
J. Am. Chem. Soc. 1986, 108, 3570.
246. J. J. Hickman, M. S. Wrighton, J. Am.
Chem. Soc. 1991, 113, 4440.
247. C. J. Murphy, A. B. Ellis, J. Phys. Chem.
1990, 94, 3082.
248. J. Z. Zhang, A. B. Ellis, J. Phys. Chem. 1992,
96, 2700.
249. C. J. Murphy, G. C. Lisensky, L. K. Leung et al., J. Am. Chem. Soc. 1990, 112, 8344.
250. S. Bastide, R. Butruille, D. Cahen et al.,
J. Phys. Chem. B 1997, 101, 2678.
251. C -M. Wang, A. Heller, H. Gerischer, J. Am.
Chem. Soc. 1992, 114, 5230.
252. E. Aharon-Shalom, A. Heller, J. Electrochem.
Soc. 1982, 129, 2865.
253. A. Heller, E. Aharon-Shalom, W. A. Bonner et al., J. Am. Chem. Soc. 1982, 104, 6942.
254. D. E. Aspnes, A. Heller, J. Phys. Chem.
1983, 87, 4919.
255. K. Rajeshwar, M. Kaneko, A. Yamada et al.,
J. Phys. Chem. 1985, 89, 806.
256. K. Rajeshwar, M. Kaneko, J. Phys. Chem.
1985, 89, 3587.
257. L. Thompson, J. DuBow, K. Rajeshwar,
J. Electrochem. Soc. 1982, 129, 1934.
258. G. Hodes, L. Thompson, J. DuBow et al.,
J. Am. Chem. Soc. 1983, 105, 324.
259. A. W.-H. Mau, C-B. Huang, N. Kakuta et al.,
J. Am. Chem. Soc. 1984, 106, 6537.

260. C. Levy-Clement, A. Heller, W. A. Bonner et al., J. Electrochem. Soc. 1982, 129, 1701.
261. C. E. D. Chidsey, Science 1991, 251, 919.
262. J. F. Smalley, S. W. Feldberg, C. E. D. Chidsey et al., J. Phys. Chem. 1995, 99, 13141.
263. S. B. Sachs, S. P. Dudek, R. P. Hsung et al.,
J. Am. Chem. Soc. 1997, 119, 10563.
264. C. W. Sheen, J. X. Shi, J. Martensson et al.,
J. Am. Chem. Soc. 1992, 114, 1514.
265. Y. Gu, D. H. Waldeck, J. Phys. Chem. 1996,
100, 9573.
266. Y. Gu, D. H. Waldeck, J. Phys. Chem. B
1998, 102, 9015.
267. A. Haran, D. H. Waldeck, R. Naaman et al.,
Science 1994, 263, 948.
268. A. Hagfeldt, L. Walder, M. Gr¨tzel, Proc. a Soc. Photo-Opt. Instrum. Eng. 1995, 2531, 60.
269. H. Lindstrom, H. Rensmo, S.-E. Lindquist et al., Thin Solid Films 1998, 323, 141.
270. N. Kakuta, K. H. Park, M. F. Finlayson et al.,
J. Phys. Chem. 1985, 89, 732.
271. A. Ueno, N. Kakuta, K. H. Park et al.,
J. Phys. Chem. 1985, 89, 3828.
272. E. S. Smotkin, A. J. Bard, A. Campion et al.,
J. Phys. Chem. 1986, 90, 4604.
273. E. S. Smotkin, S. Cervera-March, A. J. Bard et al., J. Phys. Chem. 1987, 91, 6.
274. S. Kuwabata, N. Takahashi, S. Hirao et al.,
Chem. Mater. 1993, 5, 437.
275. C. S. C. Bose, K. Rajeshwar, J. Electroanal.
Chem. 1992, 333, 235.
276. Y. Son, N. R. de Tacconi, K. Rajeshwar,
J. Electroanal. Chem. 1993, 345, 135.
277. N. R. de Tacconi, Y. Son, K. Rajeshwar,
J. Phys. Chem. 1993, 97, 1042.
278. N. R. Avvaru, N. R. de Tacconi, K. Rajeshwar, Analyst 1998, 123, 113.
279. S. Ito, T. Deguchi, K. Imai et al., Electrochem. Solid State Lett. 1999, 2, 440.
280. M. Zhou, W-Y. Lin, N. R. de Tacconi et al.,
J. Electroanal. Chem. 1996, 402, 221.
281. M. Zhou, N. R. de Tacconi, K. Rajeshwar,
J. Electroanal. Chem. 1997, 421, 111.
282. N. R. de Tacconi, H. Wenren, K. Rajeshwar,
J. Electrochem. Soc. 1997, 144, 3159.
283. N. R. de Tacconi, H. Wenren, D. McChesney et al., Langmuir 1998, 14, 2933.
284. N. R. de Tacconi, J. Carmona, K. Rajeshwar, to be published.
285. N. R. de Tacconi, M. Mrkic, K. Rajeshwar,
Langmuir 2000, 16, 8426.
286. C. Anderson, A. J. Bard, J. Phys. Chem. B
1997, 101, 2611.

1.10 Conclusion
287. N. Takeda, M. Ohtani, T. Torimoto et al.,
J. Phys. Chem. B 1997, 101, 2644.
288. N. R. de Tacconi, K. Rajeshwar, R. O. Lezna,
Electrochim. Acta 2000, 45, 3403.
289. N. R. de Tacconi, J. Carmona, K. Rajeshwar,
J. Phys. Chem. B 1998, 101, 10151.
290. N. R. de Tacconi, J. Carmona, W. L. Balsam et al., Chem. Mater. 1998, 10, 25.
291. G. Hodes, J. Manassen, D. Cahen, J. Appl.
Electrochem. 1977, 7, 182.
292. G. Hodes, J. Manassen, D. Cahen, J. Electrochem. Soc. 1977, 127, 544.

293. A. Fujishima, K. Honda, Nature 1972, 238,
37.
294. B. O’Regan, M. Gr¨tzel, Nature 1991, 353, a 737.
295. J. Manassen, G. Hodes, D. Cahen,
J. Electrochem. Soc. 1977, 124, 532.
296. S. Licht, G. Hodes, R. Tenne et al., Nature
1987, 326, 863.
297. S. Licht, B. Wang, T. Soga et al., Appl. Phys.
Lett. 1999, 74, 4055.
298. S. Licht, J. Phys. Chem. B 2001, 105,
6281.

53

59

2.1

Photoelectrochemical Systems
Characterization
John J. Kelly, Zeger Hens and
Daniel Vanmaekelbergh
Utrecht University, Utrecht, The Netherlands
Zeger Hensalso
Laboratorium voor Fysische Chemie, Gent,
Belgium
2.1.1

Introduction

This chapter focuses on the characterization of photoelectrochemical systems.
Before the essential features of such systems are described in Sect. 2.1.2, the scope of the chapter is ﬁrst deﬁned on the basis of a brief history of the ﬁeld.
Scope of the Chapter
Semiconductor electrochemistry developed as a discipline with the development of semiconductor device technology [1–3].
Wet chemical processing including etching was important for the fabrication of early silicon and germanium devices.
With the increasing sophistication and miniaturization in silicon technology, wet processes were replaced by dry, mainly physical, methods. As a result, interest in semiconductor electrochemistry declined. 2.1.1.1

However, two developments in the 1970s led to a huge revival of interest and to the ‘‘birth’’ of photoelectrochemistry.
In 1971, Fujishima and Honda showed that it was possible to photolyze water in an electrochemical cell with a TiO2 photoanode and a Pt cathode without an external source [4]. Subsequently, a number of groups [5–8] described regenerative photoelectrochemical solar cells based on narrow band gap semiconductors and a redox couple. They showed that with the proper choice of reducing species it was possible to achieve stability of the photoanode during operation of the cell over longer periods. In further studies, Licht and coworkers developed efﬁcient systems that couple solar energy conversion with energy storage and show improved energy conversion and stability [9–14]. At the same time, the growth of the optoelectronics industry, based largely on III–V materials, revealed a need for new processing methods [15], including etching and metallization. Wet chemical and electrochemical methods proved very successful in this ﬁeld. In addition, electrochemical methods could be used in a simple way to make and characterize materials for a whole range of applications. The work described so far relates to single crystalline and, to a lesser extent, polycrystalline materials.

60

2 Experimental Techniques

Interest in porous photoelectrochemical systems was stimulated by the report in
1991 of O’Regan and Gr¨tzel [16] of a a novel solar cell based on a dye-sensitized nanoparticulate TiO2 photoanode. This work raised a whole range of interesting fundamental issues with regard to charge carrier dynamics, transport and interfacial transfer in porous semiconductor matrices permeated by an electrolyte solution. The discovery in 1992 [17] of the striking optical properties of nanoporous silicon, obtained by (photo)anodic etching, led to a reappraisal of the photoelectrochemistry of this material and to studies of other porous crystalline semiconductors [18].
Research on size quantization in colloidal systems was responsible for the development by Bard and coworkers of a new ﬁeld in photoelectrochemistry, that of nanodot electrodes [19, 20]. An ordered or disordered monolayer or sub-monolayer of nanometer-sized semiconductor particles is attached to a conducting substrate either directly or via a self-assembled organic monolayer. The monolayer acts as a spacer, allowing the distance between the dot and the substrate to be varied.
Absorption of light by the semiconductor dots gives rise to processes similar in many ways to those observed in bulk electrodes. However, because of the size quantization and the distinctive electrode geometry, striking new effects are found.
In this chapter, the characterization of the three types of systems described earlier have been considered: single crystal and polycrystalline bulk electrodes
(Sect. 2.1.3), dye-sensitized and quantum dot electrodes (Sect. 2.1.4), and macroporous and nanoporous semiconductors
(Sect. 2.1.5). Some essential features, common to all these systems, are introduced in the following subsection. The

obvious way to characterize a photoelectrochemical system is by electrochemical methods, and this approach will constitute the focus of the chapter. A brief and very general introduction to
(photo)electrochemical characterization is given in Sect. 2.1.2. This is not intended to be a conclusive review; for this a whole volume would be required. The application of these methods to the three systems is described in the subsequent sections. Where relevant, other nonelectrochemical methods of characterization are mentioned.
Special Features of
Photoelectrochemical Systems
Photoelectrochemical systems rely on the properties of a semiconductor electrode or of an electrode provided with a ‘‘sensitizer’’ layer consisting of, for example, dye molecules or quantum dots. This working electrode (WE) forms part of an electrochemical cell that also contains a counter electrode (CE) and, in many cases, a reference electrode (RE). Absorption of light by the semiconductor or the sensitizer layer gives rise to a photocurrent (PC) and/or a photovoltage, which can be measured in the external circuit. Conversely, the passage of current through the interface of the working electrode with the electrolyte solution can lead to light emission. In this section, some important aspects of such photoelectrochemical systems are reviewed. These topics are dealt with later in the chapter.
When a semiconductor is brought into electrical contact with an electrolyte solution containing a redox couple (Red/Ox+ ), equilibrium is established by exchange of charge between the two phases [21–23].
Figure 1(a) shows an example of an n-type semiconductor with a redox couple whose
Fermi energy (which is related to the redox potential VRed/Ox ) is located in the
2.1.1.2

2.1 Photoelectrochemical Systems Characterization

EF,n

Energy

e φSC

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 transferred from the conduction band (CB) of the semiconductor to solution creating a space charge layer (a depletion layer of width WSC ) within the semiconductor with a band bending eφSC . In equilibrium, the Fermi level is constant throughout the system. There are various ways in which equilibrium can be disturbed, for example, by illumination, by an externally applied potential, or by charge carrier injection from solution.
A photovoltaic (two-electrode) cell operates without an external voltage source [22,
23]. The working electrode is illuminated.
Two limiting cases can be distinguished.
If the electrodes are short-circuited, then the electrons and holes, generated by supra–band gap light, are separated by migration within the depletion layer and by diffusion. The holes react at the interface,

oxidizing the reduced species
−→
Red + hVB + − − Ox+

(1)

The electrons are detected as photocurrent in the external circuit. This ‘‘short-circuit photocurrent’’ depends on the competition between hole transfer across the interface
(the Faraday reaction) and electron-hole recombination in the bulk semiconductor and at the surface of the electrode. Under open-circuit conditions no photocurrent can ﬂow in the external circuit. The
Fermi level in the semiconductor is raised with respect to that in solution and a photovoltage is established. In the limiting case at high light intensity, the energy bands become ﬂat (Fig. 1b). The maximum photovoltage is determined by the difference in Fermi level under ﬂat band and equilibrium conditions. Clearly,

61

2 Experimental Techniques

such a system can supply electrical power
(Sect. 2.1.2.2).
Unlike photovoltaic cells, most photoelectrochemical systems do not operate without an external source. In a more general approach, the potential of the working electrode can be varied with respect to the equilibrium potential Veq (or to the potential of a RE) by means of an external voltage source (e.g. a potentiostat) connected between WE and CE [21, 24].
The current density (indicated here by j ) is measured both in the dark and under illumination as a function of the applied potential V . This is described in more detail in Sect. 2.1.2.2.
A strong oxidizing agent (OxS + ), that is one with electron acceptor levels corresponding to the valence band of the semiconductor, can extract electrons from the band thus creating holes
OxS + − − RedS + hVB +
−→

(2)

In an n-type semiconductor, these holes can recombine with majority carriers (electrons) from the CB [25]. Electroluminescence (EL) is expected from a semiconductor with a direct band gap for conditions in which the surface electron concentration is signiﬁcant. Similarly, electron injection

µh

Electron
Light
Hole conductor absorber conductor

into a p -type semiconductor can give rise to light emission [25].
Macroporous semiconductor electrodes resemble in many respects the bulk electrodes described earlier. However, there are clear differences between porous and nonporous systems that become even more pronounced in nanoporous electrodes. Porous systems are considered in
Sect. 2.1.5.
Recently, insulating nanocrystals have been attached to conducting substrates
(metal or indium tin oxide) by van der
Waals interactions or covalent bonding.
When such a system is used as a working electrode in a photoelectrochemical cell, a photocurrent (in the µA range) can be measured [19, 20, 26]. Clearly, a monolayer of nanocrystals can absorb a sufﬁcient fraction of the incident light to give to a measurable photocurrent. The absorption of a photon in a nanocrystal leads to a photoexcited state; an electron is promoted from a lower energy level (HOMO or valence band) to a higher energy level
(LUMO or CB) (Fig. 2). A photocurrent will be observed if the electron and hole are separated effectively before the system relaxes to the ground state. In
Fig. 2, the electron is transferred to the

Scheme representing the general principle of a photoelectrochemical system. Electrons are photoexcited in the absorber. The electron and hole are selectively transferred to an electron conductor (usually a metal or a semiconductor) and to a hole conductor
(a redox system in a liquid electrolyte).
The photovoltage is the difference between the electrochemical potentials in the electron and hole conducting phases; thus µe − µh .

Fig. 2

µe

Energy

62

2.1 Photoelectrochemical Systems Characterization

metal, the hole to an oxidizable species in solution. This leads to an upward shift of the electrochemical potential of the electrons µe and a downward shift of the electrochemical potential of the holes µh (and thus to a photovoltage Vphoto =
(µe − µh /e). Such a photoelectrochemical system shows a strong analogy with molecular (e.g. dye) systems adsorbed on metal or semiconductor electrodes.
The photoelectrochemical solar cell of
O’Regan and Gr¨tzel [16] combines the a special features of semiconductor porosity and sensitization described earlier.
This system consists of nanometer-sized
TiO2 crystallites interconnected to form a three-dimensional porous assembly. Dye molecules acting as the light absorber are chemisorbed on the internal surface. The efﬁciency of light absorbance is close to unity as a result of the fact that photons encounter an adsorbed dye molecule many times. The electrolyte solution, containing an I − /I3 − redox system, permeates the pores of the system. The TiO2 assembly acts as an electron conductor and the electrolyte as the ‘‘hole conductor’’. Because of the interpenetration of the electron and hole conducting phases on a nanometer scale, back transfer of the electron to I3 − is the main source of recombination.
The three systems described earlier have a number of features in common. To observe photocurrent, light obviously has to be absorbed by the system. The spatial separation of photogenerated electrons and holes must be more effective than their recombination. This requires efﬁcient kinetics for electron transfer across the solid/solution interface. These are topics that are addressed in the remainder of the chapter. Electrode reactions in photoelectrochemical systems may be quite complicated. To illustrate the general approach to the elucidation of reaction

mechanisms two complex systems are considered. Finally, charge carrier transport in porous semiconductor systems is a topic deserving special attention.
2.1.2

Photoelectrochemical Characterization
Methods

For scientiﬁc research on photoelectrochemical systems, the photoelectrode is generally the working electrode in a threeelectrode electrochemical cell. Using this set up, the system can be investigated by perturbing it and recording the system’s response. In general, such perturbationresponse methods may be classiﬁed according to the time dependence of the perturbation (steady state versus timeresolved) or to the physical nature of the perturbation and the response. In this section, the principles of steady state and time-resolved methods in general are ﬁrst discussed. Later on, two different perturbation-response techniques used to study semiconductor-based photoelectrochemical systems are discussed. Finally, methods in which illumination or minority carrier injection gives rise to light emission from the electrode are considered.
Steady State and Time-resolved
Methods
Steady state methods are basically simple: one applies a time-independent perturbation x to a system and records – if it exists – the time-independent response y(x) of the system. This may be repeated for different levels of the perturbation. In this way, a functional relation between perturbation and response is determined, which should provide information on the system studied. Consider, for example, an electrical resistor through which a current is passed (perturbation). Instantaneously, the system responds by maintaining a
2.1.2.1

63

64

2 Experimental Techniques

potential difference across the resistor. The linear relation between both quantities allows the characterization of the system.
A major drawback of steady state techniques is that no information is obtained on the dynamic properties of a system.
Take for instance an electronic circuit, which simulates a simple resistor except that the potential difference is established between its poles after some time delay.
Clearly, this circuit would emerge from a steady state analysis as a simple resistor: the system dynamics, which is a basic characteristic of any system, can be revealed only by recording the response from the moment the perturbation is applied (timeresolved measurements).
Many time-resolved methods do not record the transient response as outlined in the earlier example. In the case of linear systems, all information on the dynamics may be obtained by using sinusoidally varying perturbations x(t)
(harmonic modulation techniques) [27], a method far less sensitive to noise. In this section, the complex representation of sinusoidally varying signals is used, that √ x(t) = Re[X (ω) exp(iωt)], where is, i = −1. The quantity X(ω) contains the amplitude and the phase information of the sinusoidal signal, whereas the complex exponential exp(iωt) expresses the time dependence. A harmonically perturbed linear system has a response that is – after a certain transition time – also harmonic, differing from the perturbation only by its amplitude and phase (i.e. y(t) =
Re[Y (ω) exp(iωt)]). In this case, all the information on the dynamics of the system is contained in its transfer function H (ω), which is a complex function of the angular frequency, deﬁned as [27, 28]
H (ω) =

Y (ω)
X (ω)

(3a)

Time-resolved measurements on linear systems may be represented in many ways.
For example, in the Nyquist representation, the transfer function H (ω) is plotted as a point in a two-dimensional plane having coordinates [Re(H ) and Im(H )], for each frequency measured. In speciﬁc cases, the transfer function may be represented also by an equivalent electrical circuit. This is a combination of lumped circuit elements (resistor, capacitor, etc.) having the same perturbation-response behavior as the system studied.
In the case of a nonlinear system, a similar approach using harmonic perturbations is possible if a ‘‘small-signal’’ perturbation x(t) = Re[ X (ω) exp(iωt)], superimposed on a time-independent ‘‘bias’’ perturbation, is applied to the system. If the signal level of the perturbation is sufﬁciently small, a linear dependence of the response on the perturbation can be achieved (i.e. y(t) = Re[ Y (ω) exp(iωt)]).
Clearly, the transfer function deﬁned in
Eq. (3a) becomes a differential quantity:
H (ω) =

Y (ω)
X (ω)

(3b)

At low frequencies, H (ω) is equal to the slope ∂y/∂x of the steady state response y(x). The time-resolved electrochemical techniques discussed in Sects. 2.1.2.2 and
2.1.2.3 pertain to this class of small-signal modulation techniques.
In general, measuring the transfer function of a system under study using harmonic modulation techniques is straightforward. Interpreting the experimental data, however, is not. As will be demonstrated by experimental examples in Sects. 2.1.3 and 2.1.5, time-resolved methods become most powerful if the experimental impedance can be analyzed using a dynamic model for the system studied. 2.1 Photoelectrochemical Systems Characterization

Current Density Versus Potential
Techniques
Because most applications of (photo)electrochemical systems involve the transfer of electrons across an interface (Sect. 2.1.1), current density-potential techniques are commonly used in (photo)electrochemistry. In this case, the difference in electrochemical potential of electrons across the interface of interest (accessible via the working electrode – reference electrode potential difference) and the current density through this interface are used as the perturbation and the response (or vice versa). Two approaches can be distinguished. When (quasi) steady state signals are used, one speaks of current density versus potential measurements whereas harmonically modulated signals, superimposed on a bias, are involved in electrochemical impedance spectroscopy (EIS).
We introduce these two approaches on the basis of the kinetics of the simple system shown in Fig. 1.
If the potential of the n-type electrode, whose energy band diagram is shown in
Fig. 1(a), is made negative with respect to the equilibrium potential Veq , then, the band-bending eφSC decreases until ﬁnally ﬂat band condition (V = Vfb ) is reached
(Fig. 1b). If the potential is made more 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
1(b) is accompanied by an increase in the
2.1.2.2

(a) Schematic representation of the current-potential curves for an n-type semiconductor (see Fig. 1) in the dark (a) and under illumination with supra-band gap light (b). (b) The part of curve (b) relevant to photovoltaic applications. The open-circuit oc photovoltage VPH and the short-circuit sc are indicated, as is the photocurrent jPH m m rectangle deﬁning jPH and VPH .

electron concentration at the surface. As a result, a net cathodic current ﬂows across the interface due to the reduction of the oxidized species
Ox+ + eCB − − − Red
−→

(4)

Here, we assume that electron transfer only occurs via the CB and not via surface states. As in a Schottky diode, j generally increases exponentially with (decreasing) potential (Fig. 3a). The form of the dark current-potential curve, however, depends on the mechanism and kinetics of the charge-transfer reaction. At high overpotential, corresponding to a large deviation from equilibrium, the reaction expressed by Eq. (4) may become limited by mass transport in solution, that is, the cathodic current becomes potentialindependent (this is not shown in Fig. 3).
If the potential of the electrode of Fig. 1 is made more positive than Veq , the bandbending increases with respect to that at j b a V vs Veq

(a) sc j PH

Fig. 3

m j PH

V oc
PH

(b)

m
V PH

65

66

2 Experimental Techniques

equilibrium and strong depletion, or even inversion, may result. This corresponds to the blocking current range of the diode
(V > Veq ). A small potential-independent anodic current results from electron injection from the reduced species across the barrier into the conduction band.
Under illumination, a photocurrent is observed under depletion conditions if the photogenerated electron and hole are spatially separated before recombination can occur. The hole reacts at the surface
(Eq. 1) and the electron is collected at the counter electrode. For the simplest case in which no recombination occurs at the surface, G¨rtner [29] derived an expression a for the photocurrent density jPH taking into account the absorbed photon ﬂux
, the absorption coefﬁcient α , and the minority carrier diffusion length Lmin jPH = e

1−

exp(−αWSC )
1 + αLmin

(5a)

The potential dependence of the photocurrent is determined in the model by the dependence of the depletion layer thickness on the band-bending
WSC =

2εε0 φSC eND 1/2

(5b)

where ND is the donor density, ε is the dielectric constant of the semiconductor, and φSC equals V − Vfb (see Sect. 2.1.3.1).
In essence, this simple model states that an electron/hole pair will contribute to the photocurrent if it is generated within a distance from the electrolyte interface within which the electron and hole can be separated by migration and diffusion, that is, the penetration depth of the light 1/α is less than Lmin + WSC . Apart from the applied potential (which determines WSC ), the efﬁciency of charge separation depends on the quality of the semiconductor

(Lmin ) and on α (which is a function of the wavelength). Light absorption just above the band edge depends strongly on whether the semiconductor has a direct band gap (for which α is large and the penetration depth is small) or an indirect band gap (for which α is small and 1/α is large). The importance of such factors is illustrated in Sect. 2.1.5.2.
Recombination of electrons and holes via surface states competes with hole transfer to solution and thus reduces, and may even suppress, the photocurrent. To obtain a limiting photocurrent, a stronger band-bending is required. Obviously, more extended models are needed to describe the photocurrent-potential characteristics in this ‘‘less favorable’’ case [24].
A schematic curve for the dependence of the total current density (dark current +
PC) on the potential is shown in Fig. 3(a).
It is clear that the form of this curve on the anodic side will depend on the parameters as discussed earlier. The part of the curve relevant to photovoltaic applications is shown in Fig. 3(b). The short-circuit sc photocurrent density jPH and the openoc are indicated. It circuit photovoltage VPH should be noted that this part of the curve can be mapped without an external voltage source by measuring the photocurrent through and the voltage across a load resistor (between WE and CE) whose resistance is varied from zero to a very large value. The conversion efﬁciency η of a photovoltaic cell is deﬁned as η = Pm /Po where Po is the power of the incident radiation and Pm , the maximum power output of the cell, is given by m m
Pm = jPH × VPH

(6a)

m m The values of jPH and VPH depend on the shape of the current-potential curve under illumination (i.e. on both the dark current and photocurrent-potential

2.1 Photoelectrochemical Systems Characterization

characteristics). For a given temperature m m and light intensity jPH and VPH are chosen to give a rectangle of maximum area in the current-potential plot (see shaded area in
Fig. 3b). An important parameter for the characterization of a solar cell is the ﬁll factor (FF) deﬁned as
FF =

sc oc jPH VPH m m jPH VPH

(6b)

In steady state measurements, one generally applies a ﬁxed potential to the working electrode and measures the steady state current through the cell. Alternatively, the potential is scanned at a ﬁxed rate and the current is measured continuously. As in metal electrochemistry, information about the kinetics of surface reactions and the hydrodynamics of the system can be obtained by varying the potential scan rate.
The rotating disk electrode is a useful tool for studying the role of mass transport.
The rotating ring-disk electrode (RRDE) has two important applications in semiconductor electrochemistry. In studies of competitive reactions at the semiconductor disk, the reaction products can be analyzed at the ring of the RRDE (see Sect. 2.1.3.3).
Alternatively, reactive species can be generated at the metal disk of a RRDE and their electrochemistry studied at the semiconducting ring [25]. Important parameters in photoelectrochemistry are the intensity and the spectral distribution of the light used for photocurrent and photovoltage studies. In photovoltaics, steady state measurements can be used to obtain FF and solar energy conversion efﬁciencies [24].
Such results may give insight into reaction mechanisms. From the earlier discussion, it is clear that the j versus the V relation of an electrochemical system is nonlinear. Therefore, the electrochemical admittance (or its inverse, the electrochemical impedance) is

deﬁned as the transfer function relating a small signal variation of the working electrode potential and of the current density through the working electrode/electrolyte interface: YEC (ω) =

˜ j (ω)
V (ω)

(7)

In the case of a semiconductor-based photoelectrochemical system, the measurement of the electrochemical admittance serves two purposes. As is explained in
Sect. 2.1.3.1, it allows on the one hand the in situ determination of the energetics of the (bulk) semiconductor surface. On the other hand, it makes the dynamics of various (photo)electrochemical processes experimentally accessible. Clearly, EIS is also possible using an illuminated semiconductor, an experimental method sometimes referred to as PEIS. Finally, it should be noted that although the electrochemical admittance is determined experimentally
(the applied electrode potential is used as the perturbation), the electrochemical impedance is generally plotted as the result of an EIS measurement.
PC versus Light Intensity
Techniques
Typical to the study of photoelectrochemical systems are measurement techniques that use the light ﬂux incident on the working electrode as a perturbation and the resulting PC jPH as the system’s response [30]. In this case, the electrode potential can be used as an additional experimental variable. Clearly, the incident light ﬂux cannot be quantiﬁed directly.
Therefore, a reference signal proportional to this light ﬂux is generally used, for example, the voltage generated by the light ﬂux on a reference photodiode. This technique, which for steady state conditions has no particular name, has been applied several
2.1.2.3

67

2 Experimental Techniques

times with success, for example, to identify
(light-intensity dependent) photocurrent multiplication processes (see Sect. 2.1.3.3).
The optoelectrical transfer function relates an incident sinusoidally modulated light ﬂux and the resulting modulated photocurrent density. This quantity is deﬁned as [30]
YOE (ω) =

˜ jPH (ω) e (ω)

(8)

Again, it should be stressed that the optoelectrical transfer function is deﬁned using small-signal perturbations, superimposed on time-independent bias signals. Measurement of the optoelectrical impedance is often referred to as intensity modulated photocurrent spectroscopy (IMPS).
Luminescence-based Techniques
Photoluminescence (PL), like photocurrent, is a technique in which a light ﬂux incident on the working electrode acts as a perturbation. In this case, the response of the system is followed by measuring the intensity of the emitted light. As in photocurrent measurements, the electrode potential can be used as an additional variable.
As explained in Sect. 2.1.2.2, photocurrent is obtained when, under depletion conditions, electrons and holes created
2.1.2.4

Intensity /Current

68

PL

by illumination are effectively separated by migration and diffusion (Fig. 4). PL, on the other hand, is expected when the photogenerated carriers recombine radiatively, that is, at potentials approaching the ﬂat band potential Vfb (see Sect. 2.1.3.1).
Emission may result either from direct band-band recombination or from indirect recombination via a band gap state [25]. In the simplest case in which surface recombination can be disregarded, the potential dependence of the emission intensity IPL is described by the G¨rtner equation: a IPL = κ

exp(−αWSC )
1 + αLmin

where is the incident photon ﬂux and κ is the ratio of the rate of radiative recombination to the total recombination rate. In principle, it is possible to obtain values for
Lmin , WSC , and α from the potential dependence of IPL . The G¨rtner model has been a extended by Gerischer and coworkers [31,
32] to account for surface recombination.
In this case, surface recombination rates can be obtained from photoluminescence measurements. While photoluminescence has been mainly studied under steady state conditions or with transient techniques, harmonic-modulation measurements are possible. Beckmann and Memming [33]

PC

−

+

Potential
−

+

(9)

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

complicated potential-dependence points to changes in surface chemistry that inﬂuence nonradiative surface recombination.
The observation of electroluminescence in an electrochemical system clearly shows the involvement of minority carriers in the charge-transfer processes.
As far as we are aware, experiments on electroluminescence involving a sinusoidal perturbation of the applied potential have not been performed but large-signal potential step and pulse measurements have been reported. We shall return brieﬂy to these aspects in Sect. 2.1.3.2.5.
2.1.3

Bulk Systems

In this section, we consider some important aspects of photoelectrochemical systems based on single-crystal and polycrystalline electrodes. In Sect. 2.1.3.1, the use of EIS is described for the determination of the energetics of the semiconductor/electrolyte interface. Section 2.1.3.2 deals with the dynamics of electron/hole recombination at the semiconductor surface and the importance of electrochemical and optoelectrical impedance techniques for such studies. Finally, a somewhat
Intensity/Current

studied the photoluminescence of n-type
GaP by perturbing the system with a small sinusoidal modulation of the potential during a potential scan and measuring the response with lock-in techniques.
EL is observed when minority carriers, injected into a semiconductor from solution, recombine radiatively with majority carriers [25]. This is illustrated schematically in Fig. 5 for an n-type semiconductor.
At positive potentials, the surface electron concentration is very low. Holes injected by the oxidizing agent in solution are held at the electrode surface by the electric ﬁeld of the depletion layer. The surface holes generally cause oxidation and dissolution of the semiconductor. Clearly, in this potential range neither current nor light emission is observed. As the potential is made negative and the band-bending decreases, the injected holes recombine with electrons supplied via the external circuit.
This results in a cathodic current and, if recombination is radiative, in light emission.
In general, the cathodic current becomes potential-independent at negative potentials as a result of mass-transport limitations in solution. The emission intensity should, therefore, become constant. However, this is often not the case. A more

Schematic representation of the 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.

EL

Potential

Fig. 5

Current
−

+

+

69

2 Experimental Techniques

arbitrary choice of complex reactions is used to illustrate the general approach to the study of reaction mechanisms and reaction kinetics (see Sect. 2.1.3.3).
In Situ Energetics of
Semiconductor/Electrolyte Interfaces
An aspect of great importance to the electrochemical properties of a (bulk) semiconductor/electrolyte (s/e) interface is the energetic position of the upper edge of the valence band and the lower edge of the CB of the semiconductor at the interface. According to the Gerischermodel for charge transfer at bulk semiconductor electrodes, the energetics of the s/e interface plays an essential role in determining the rate and the mechanism of an electrode process [34–36].
Capacitance measurements – hence, electrochemical impedance measurements – constitute the most widely used in situ method to determine the energetics of s/e interfaces [37].
As described earlier, the s/e interface is electriﬁed, that is, charge is separated at the interface. According to Gauss’ law, the charge separation is accompanied by an inner-potential difference between the semiconductor and the electrolyte. Consequently, in the absence of any electrochemical process, the s/e interface acts as a capacitor, the capacitance of which may
2.1.3.1

C −2 / 1012 F−2 cm4

70

be calculated from the electrochemical admittance (YEC = iωC ).
Most simply, the system is modeled by assuming depletion or accumulation of free charge carriers at the semiconductor side of the interface. This charge is neutralized by an ionic counter-charge at the electrolyte side of the interface (Helmholtz layer) [36, 38]. From this picture, two (differential) capacitances may be deﬁned. The ﬁrst, CSC , relates a change of the charge
QSC accumulated in the semiconductor to a change of the potential drop φSC across the semiconductor (CSC = ∂QSC /∂φSC ).
The second, CEL , relates analogously the charge QEL in and the potential difference φEL across the electrolyte side of the interface (CEL = −∂QEL /∂φEL ). Assuming that the total potential drop φ across the interface equals the sum φSC + φEL , it may be easily shown that the interfacial capacitance C corresponds to the capacitance of the series connection of the capacitors CSC and CEL [36, 37, 39].
In Fig. 6, the inverse square of the capacitance of an n-InP|1.2 M HCl solution interface is plotted as a function of the potential applied to the n-InP electrode. The capacitance has been calculated by modeling the interface as a parallel connection of a capacitor and a resistor in series with the cell resistor. Clearly, both quantities are linearly related. This result can be

10

Vfb − kT/e

0
−500

Mott-Schottky curve determined at an n-InP electrode in a 1.2 M HCl aqueous solution. The interfacial capacitance is determined from the electrochemical impedance measured at
8.2 kHz using a parallel connection of a resistor and a capacitor in series with the cell resistor.

Fig. 6

5

0
V / mV vs. SCE

500

2.1 Photoelectrochemical Systems Characterization

understood if majority charge carriers are depleted from the semiconductor surface.
In that case, the capacitance CSC is given by the Mott-Schottky equation, which for an n-type semiconductor reads [36, 37, 39]:
CSC =

eND εε0
2

1/2

φSC −

kB T e −1/2

(10) where T is the absolute temperature and the constants e and kB indicate the elementary electrical charge and the
Boltzmann constant, respectively. The capacitance CSC can be identiﬁed directly with the interfacial capacitance if CSC
CEL . Although exact values of CEL are scarce in the literature, this inequality is generally assumed to be fulﬁlled for moderately doped semiconductors under depletion conditions [36, 37, 39, 40]. As a consequence, a change of the potential drop φ across the interface results mainly in a change of φSC (i.e. φEL is approximately constant). Hence, φSC may be written as the difference between the applied electrode potential V and the so-called ﬂat band potential Vfb [37]: φSC = V − Vfb

(11)

At the ﬂat band potential, there is no potential drop across the semiconductor and, hence, the semiconductor energy bands are ﬂat from the bulk up to the semiconductor surface (see Fig. 1b).
Moreover, because φEL is approximately constant, the energy bands are ﬁxed at the semiconductor surface. Hence, the position of the band edges at the surface may be calculated once the ﬂat band potential is known.
A problem often encountered when calculating the ﬂat band potential from
Mott-Schottky data is the frequency dependence of both the slope and the extrapolation point of the C −2 versus V curve.

Obviously, such so-called frequency dispersion hampers a proper determination of Vfb although reliable values are generally obtained at high measuring frequencies
(>10 kHz). The origin of this nonideal behavior is not well understood [37, 41].
However, to check the reliability of MottSchottky measurements, the capacitance should be measured in a broad frequency range [42–44].
Electron-hole Recombination
Dynamics at the s/e Interface

2.1.3.2

Introduction The dynamics of electron-hole recombination at the semiconductor surface has been extensively studied both at illuminated and at dark s/e interfaces [45–53]. For recombination, minority charge carriers should be present at the interface. For n-type semiconductors, holes may be supplied to the surface by illumination under depletion conditions using supra-band gap light (see
Sect. 2.1.2.2). Alternatively, holes may be injected into the valence band by a strong oxidizing agent in the electrolyte solution.
If a depletion layer exists, the injected holes accumulate at the semiconductor surface.
As it gives a nice and relatively simple illustration of the use of various characterization methods, we will discuss the subject of electron-hole recombination dynamics in the next sections, taking the n-GaAs photoanode as the main example.
More complicated topics – related to interfacial transfer of photogenerated charge carriers – are discussed in Section 2.1.3.3.

2.1.3.2.1

2.1.3.2.2 Current Density Versus Potential
Measurements Figure 7 shows a current density versus potential curve for an illuminated n-GaAs electrode (λ = 480 nm) in a 0.1 M H2 SO4 solution. As reported for various photoanodes in the literature,

71

2 Experimental Techniques
Current density versus potential curve, recorded at an illuminated n-GaAs electrode in a 0.1 M H2 SO4 aqueous solution. Indicated are the ﬂat band potential (as determined in the dark) and the different potential ranges
(see text).
Fig. 7

500

j
[µA cm−2]

72

A

B

C

250

Vfb
0
−1000 −750

−500

−250

0

V / mV vs. SCE

the j versus V plot shows – at potentials more positive than the ﬂat band potential – three different regions, indicated as
A, B, and C in Fig. 7 [45, 47, 49, 50, 54]. In region A, that is closest to Vfb , no current passes through the s/e interface, despite the minority charge carrier ﬂux towards the surface. On the other hand, a photocurrent plateau appears in region C – at the most positive potentials. The intermediate region B shows a transition between these two extreme situations. Clearly, the occurrence of a photocurrent plateau is in accordance with the G¨rtner-equation: a in this potential region, all photogenerated holes reach the semiconductor surface and participate in a charge-transfer reaction
(the oxidation of the semiconductor). On
WSC in the the other hand, since α −1 case of n-GaAs, the absence of photocurrent in region A is not accounted for by the G¨rtner-equation. This indicates a loss a of photogenerated holes, which can be attributed to electron-hole recombination at the semiconductor surface.
2.1.3.2.3 The Electrochemical Impedance of Surface Recombination Figure 8 shows the impedance spectrum of the illuminated n-GaAs|0.1 M H2 SO4 interface, as recorded in the potential region A (no steady state photocurrent). One can see a small semicircle at high frequencies.

For various illumination intensities, the diameter of the semicircle ﬁtting the data at high frequencies equals approximately kT /e|jPH | [45–47, 49]. In addition, it was shown that upon illumination, a capacitive peak appears in the C −2 versus V plot of the n-GaAs|0.1 M H2 SO4 interface [45, 46,
51]. The peak value proved to be a function of the frequency and the photocurrent density as measured in region C [51]. This behavior is markedly different from the purely capacitive impedance (vertical line in the Nyquist plane and straight MottSchottky plot) expected for a blocking s/e interface (see Sect. 2.1.3.1).
The appearance of both the semicircle and the capacitive peak were accounted for by Vanmaekelbergh and coworkers, by considering recombination of photogenerated holes with CB electrons at the semiconductor surface [51, 55–57]. The recombination mechanism assumed by these authors consists of the successive capture of an electron in an empty surface state and of a hole in an occupied surface state. Taking the rates of the electron
(hole) capture steps to be ﬁrst order in the CB electron density nS (valence band hole density pS ) and the density of empty
(ﬁlled) surface states, an electrochemical impedance corresponding to the equivalent circuit shown in Fig. 9 was calculated for this recombination mechanism. The

2.1 Photoelectrochemical Systems Characterization

Im(Z )
[Ω cm2]

−100

R1
−50

512 Hz

32 kHz

4 kHz
0
0

50

100

150

Re(Z )
[Ω cm2]
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. 8

Equivalent circuit obtained for surface recombination at a semiconductor/electrolyte interface. Explicit expressions of the different circuit elements are given in the text.

Fig. 9

C1

R1

impedance of the three different circuit elements read [51]: kT (12) e|jREC | kT βn nS
1
(13)
R2 = e|jREC | βp pS 1 − |jREC /jPH |
R1 =

C1 =

e|jREC | 1 kT βn nS

(14)

where jREC equals the current density associated with recombination and βn (βp ) denote the rate constant for electron (hole) capture. Because the photocurrent density adds to the current density that (dis)charges the

C2

space charge layer of the semiconductor, the overall impedance of the s/e interface consists of the parallel connection of the surface-recombination impedance and the space charge layer capacitor. At high frequencies, the recombination impedance reduces to the resistor R1 . Hence, the high-frequency impedance of the interface corresponds to the parallel connection of R1 and CSC . This parallel combination accounts for the experimentally observed high-frequency semicircle (note the correspondence between the experimental diameter and R1 ). In addition, the features of the capacitive peak in the Mott-Schottky plot could be qualitatively explained by

73

74

2 Experimental Techniques

considering the parallel connection of the recombination impedance and the space charge layer capacitor [51].
Clearly, electron-hole recombination is not limited to illuminated semiconductors.
Minority charge carriers also may be injected by an oxidizing or reducing agent in the electrolyte solution (see Sects. 2.1.1.2 and 2.1.2.4). Also in this case, the parallel connection of the surface recombination impedance and the space charge capacitor provides an accurate description of the experimental impedance [45]. For instance, the impedance spectrum of the nGaAs|Ce4+ system – Ce4+ is a well-known hole-injecting agent for n-GaAs – shows a capacitive semicircle with a diameter equal to kT /e|j | at high frequencies [45,
58]. It was also demonstrated that a capacitive peak, exhibiting the same functional dependence on frequency and current density as obtained with the n-GaAs photoanode, is present in a Mott-Schottky plot measured with the n-GaAs|Ce4+ system [45].
Both examples attribute convincingly the loss of photogenerated holes at the n-GaAs photoanode polarized in potential region
A to electron-hole recombination at the electrode surface. Because similar results have been obtained at the n-CdTe photoanode [50], one could think of the surfacerecombination impedance as a general ﬁngerprint of surface-recombination steps in an overall reaction mechanism. This is, however, not the case. For the n-InP photoanode, the features of the surface recombination are absent although recombination was shown to occur at the semiconductor surface [54]. In this particular case, this discrepancy was resolved by assuming that recombination does not occur at ﬁxed recombination centers but rather at intermediates of the anodic decomposition of the semiconductor [53]. In

addition, it was shown that the typical semicircle resulting from the parallel connection of the space charge layer capacitor and a resistor with resistance kT /e|j | is not uniquely related to surface-recombination processes. Any reaction step, the rate of which is proportional to the density of majority charge carriers at the interface, may contribute this feature to the overall impedance [59–61]. Hence, for direct transfer of majority charge carriers or for surface-state mediated transfer, this semicircle also may appear in the overall impedance spectrum [59]. These latter examples demonstrate the need for a reliable dynamic model of the chargetransfer reaction for the interpretation of impedance data.
2.1.3.2.4 The Optoelectrical Transfer Function of Surface Recombination If the nGaAs|0.1 M H2 SO4 system, polarized in potential region A, is suddenly exposed to a constant illumination, a photocurrent transient decaying from the value as measured in potential region C to zero is recorded [30]. This photocurrent decay is related to the increase from zero recombination at the moment of the exposure
(maximum photocurrent) to complete recombination (no photocurrent). The same picture arises from the optoelectrical transfer function (Fig. 10), which corresponds to a semicircle ranging from YOE = 1 at high frequencies to YOE = 0 at low frequencies [62]. As for the transient at short times, recombination is ruled out at high modulation frequencies leading to a maximum value of the modulated photocurrent, whereas at low frequencies recombination is fully operative, causing the disappearance of the photocurrent.
This impedance could be analyzed using the same dynamic model that describes the surface-recombination impedance. If the

2.1 Photoelectrochemical Systems Characterization
0.5

i˜R
)
i˜h

104 Hz

Im(1 −

ω max = 2π (4000) s−1

103 Hz

0.0
0.0

0.5

1.0

i˜
Re(1 − R ) i˜h 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).

Fig. 10

response time of the electrochemical cell is sufﬁciently fast, the resulting optoelectrical impedance, valid for potential region
A, reads according to Vanmaekelbergh and coworkers [62]:
YOE = 1 −

βn nS iω + βn nS

(15)

Apart from the ﬂattening of the semicircle, this transfer function accounts well for the experimental impedance. Moreover, from the characteristic frequency of the semicircle, the rate constant of electron capture may be calculated if the electron density at the semiconductor surface is known. From the example given in Fig. 10, a value of βn = 10−6 cm3 s−1 is obtained [62]. Analogous to the case of the n-GaAs photoanode, loss of photogenerated holes could be attributed to surface recombination by IMPS in the case of the illuminated n-CdS and n-InP electrodes [62, 54].
Summarizing, we can conclude that both examples demonstrate that impedance techniques are powerful tools for the

determination of reaction mechanisms, especially if they are combined with a mathematical model of the system studied.
Clearly, EIS and IMPS yield partly the same information. However, they are also complementary: IMPS enables the determination of the rate constant of majority charge carrier capture whereas the density of majority charge carriers at the interface is accessible only via EIS.
Moreover, EIS allows one to investigate analogous processes at dark electrodes.
2.1.3.2.5 Recombination Studied by Luminescence On normalizing the photoluminescence intensity IPL in Eq. (9) with respect to the maximum intensity Imax
(for WSC = 0) one obtains

IPL
= exp(−αWSC )
Imax

(16)

Because the thickness of the depletion layer can be obtained as a function of potential from impedance measurements
(see Sect. 2.1.3.1) Eq. (16) can be used to check the validity of the G¨rtner a 75

2 Experimental Techniques

1.0

−In(IPL/Imax)

3

0.8

0.6

IPL
[a.u.]

76

2
1
0

0.4

0

10

20

30

40

50

60

70

WSC
[nm]

0.2

0.0
−1.0

−0.5

0.0

0.5

1.0

1.5

2.0

V vs SCE
[V ]
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).

Fig. 11

approach and to determine the absorption coefﬁcient [63] (see Fig. 11). Alternatively, if α is known accurately then Eq. (16) gives information about the potential distribution within the semiconductor, as shown by Ellis and coworkers for GaAs [63,
64]. If α values are either available or determined for different wavelengths, then the minority carrier diffusion length [Lmin in Eq. 9] can be obtained.
In many cases, the photoluminescence intensity of n-type electrodes at negative potential does not attain a constant value as would be expected from Eq. (16)
[25]. Hysteresis is also frequently encountered in cyclic scanning experiments [25].
Such effects are due to potential-dependent changes in the surface chemistry. Under accumulation conditions in aqueous solution, hydrogen is evolved and may be incorporated into the electrode. In addition, the semiconductor may undergo

cathodic reduction. These processes introduce surface or near-surface states that provide pathways for nonradiative recombination. Gerischer and coworkers [31, 32] have studied such systems using an extended G¨rtner model. a There are numerous reports on the potential dependence of the electroluminescence for n-type electrodes under (quasi) steady state conditions and a limited number for p -type electrodes [25]. Such studies were mainly used for diagnostic purposes, that is, to investigate the role of minority carriers in electrode processes.
While there have been reports of timeresolved photoluminescence [65–68] and electroluminescence experiments [69–71] for the study of recombination, trapping and detrapping of photogenerated charge carriers, potential modulation has scarcely been used. An illustration of the possibilities of this approach is provided by

2.1 Photoelectrochemical Systems Characterization

work of Decker and coworkers [72] who used a frequency-dependent train of potential pulses to study the effect of surface chemistry on the radiative recombination of holes injected into n-type GaAs from
Fe(CN)6 3− in solution. An unexpected potential-dependence of the emission was attributed to a change in the surface termination from hydroxide to hydride coverage on going from positive to negative potentials. A time constant of 0.1 ms was estimated for this transition. This transformation was subsequently observed by
Ern´ and coworkers using in situ ine frared spectroscopy [73]. It is clear that potential-modulated luminescence techniques deserve wider attention.
Complex Charge-transfer
Processes
In this section we use two types of complex charge-transfer reaction to illustrate the general approach to the elucidation of reaction mechanisms at single crystal electrodes. These reactions are photocurrent doubling at n-type and p -type semiconductors and the (photo)anodic oxidation of the semiconductor itself.
2.1.3.3

2.1.3.3.1 PC-doubling Reactions PC doubling refers to a type of charge-transfer reaction in which both bands of the semiconductor are involved, thus emphasizing the distinctive features of semiconductor electrochemistry. The ﬁrst examples of such reactions relate to the photoanodic oxidation of species such as formate and tartrate at wide band gap n-type electrodes [74, 75]. A photon generates an electron-hole pair in the semiconductor.
The electron and hole are separated by the electric ﬁeld of the depletion layer. The electron is detected as photocurrent in the external circuit. The hole oxidizes a species from solution producing an intermediate

that is capable of injecting an electron into the CB. This electron also contributes to the photocurrent. As a result, for each photon absorbed in the system two charge carriers are registered as current. This corresponds to a ‘‘quantum efﬁciency’’ of two.
For formate oxidation at n-type ZnO, Morrison and coworkers [74, 75] suggested a mechanism of the type
•

HCOO− + hVB + − − H+ + COO −
−→
(17)
•

−→
COO − − − CO2 + eCB − (18)
Results obtained by Honda and coworkers [76] showed that the reaction is more complicated. The ZnO electrode is, in fact, dissolved during the current-doubling reaction. The stoichiometry found in their work is given by
ZnO + 2HCOO− + 2hVB + − −
−→
Zn2+ + H2 O + CO2 + 2eCB − (19)
This result calls into question mechanisms in which the reducing agent reacts directly with valence-band holes. Honda and coworkers proposed that formate is oxidized by atomic oxygen formed in the photoanodic oxidation of ZnO.
ZnO + 2hVB + − − Zn2+ + Oads (20)
−→
A subsequent study of the dependence of the kinetics of the photocurrent-doubling reaction on the light intensity supports this idea [77]. The following mechanism was suggested
•

•

Oads + HCOO− − − OH + COO −
−→
(21)
•

•

OH + HCOO− − − H2 O + COO −
−→
(22)
The stoichiometry of this scheme agrees with that reported by Honda and coworkers. This result could be checked by studying the kinetics of two types of competing

77

78

2 Experimental Techniques

process. Iodide ions present in solution compete for reactive intermediates and thus inﬂuence the quantum efﬁciency.
In the potential range between the onset and saturation of photocurrent, electronhole recombination competes with the charge-transfer reactions. The study of these competing processes by EIS and
IMPS supports the mechanism indicated by Eqs. (21) and (22) [77]. One of the problems encountered in the investigation of such complex mechanisms is the identiﬁcation of (reactive) intermediates. For the
ZnO/formate system Harbor and Hair [78] used ESR spin-trapping experiments to
•
show that the COO− radical anion is indeed formed.
Quantum efﬁciencies higher than one have been observed for a range of reducing agents at ZnO [74, 75]. The mechanisms have been studied less thoroughly. In the case of the oxidation of methanol, it is clear that the mechanism differs from that of formate (and tartrate) [77]. Photoanodic current doubling also has been observed for other n-type semiconductors such as TiO2 [79] and CdS [80]. Bogdanoff and Alonso-Vante [81] have described an interesting study of the competitive photoanodic oxidation of formic acid and water at TiO2 using differential electrochemical mass spectroscopy. In contrast to ZnO,
TiO2 is a stable photoanode. On the basis of on-line mass detection, the authors conclude that formic acid is oxidized by hydroxyl radicals produced by the photoanodic oxidation of water, a reaction somewhat similar to that of oxygen radicals at ZnO (see Eq. 21). In the case of formate oxidation on CdS, there is no evidence for corrosion of the semiconductor. A direct reaction of the current-doubling agent with a valence-band hole was suggested for the ﬁrst reaction step [80].

In 1969, Memming reported photocurrent doubling for the reduction of H2 O2 and S2 O8 2− at p -type GaP. In this case, two holes are detected as photocurrent for each photon absorbed. This result was explained by a two-step mechanism [82]. The ﬁrst step involves reduction of the oxidizing agent (e.g. H2 O2 ) by a photogenerated electron −→
H2 O2 + eCB − − − OH− + OH

•

(23)

and the hydroxyl radical intermediate injects a hole into the valence band
•

−→
OH − − OH− + hVB +

(24)

Since this ﬁrst report, photocurrent doubling has been found for a whole range of two-electron oxidizing agents at various semiconductors including Si, SiC, CdTe, and III–V materials [83]. A striking example of photocurrent ‘‘multiplication’’ is the reduction of iodate at p -GaAs [83]. In a wide range of light intensity, a quantum efﬁciency of three is observed, whereas at low light intensity there is evidence for an efﬁciency of six. This would mean that ﬁve of the six intermediates formed on reducing IO3 − to I− could inject a hole into the valence band of GaAs. In a number of these systems, the oxidizing agent causes chemical dissolution of the solid. In the case of GaAs, studies of etching and current doubling have led to the conclusion that the various processes involved (electron capture, hole injection, and chemical etching) are linked via a common intermediate formed by the chemisorption of the oxidizing agent on the surface [15, 84, 85].
Obviously, information about the mechanisms of such reactions can be obtained by studying the photocurrent as a function of system parameters (potential, light intensity, concentration, hydrodynamics, etc.). However, such measurements do not

2.1 Photoelectrochemical Systems Characterization

yield information about the reaction kinetics. Peter and coworkers [86, 87] were the ﬁrst to show that rate constants for majority carrier injection could be determined for such systems by IMPS. They studied current doubling for oxygen reduction at p -type GaAs and GaP. From a complexplane representation of the optoelectrical impedance, rate constants in the range
104 –105 s−1 were calculated. This result suggests that hole injection is a thermally
•
activated process. HO2 , postulated as the injecting species, gives rise to a surface energy level; in the case of GaAs, this is located about 0.4 eV above the valenceband edge.
Dissolution of Semiconductors
Another class of complex reactions that has been widely studied is the oxidative dissolution of elemental and compound semiconductors. There are a number of reasons for the interest in these systems.
The possible application of semiconductors in regenerative photoelectrochemical solar cells required the complete suppression of corrosion of the photoelectrode [5,
6, 22, 23]. On the other hand, with the development of the optoelectronics industry based on III–V materials there was a need for a more fundamental understanding of etching processes [15]. The revival of interest in porous semiconductors, triggered by the discovery of the unusual optical properties of porous Si, led to a general revival of interest in the mechanisms of porous etching [18].
Because the bonding states correspond to the valence band of a semiconductor, one expects holes to be important for the oxidation reaction [15, 24]. This is the case with most etching systems. The importance of holes is immediately clear from simple current-potential or cyclic voltammetric measurements. Generally,
2.1.3.3.2

the p -type semiconductor dissolves anodically in the dark showing an exponential increase of current with increasing potential, whereas the n-type semiconductor can only be oxidized if minority carriers are generated by light under depletions conditions (see Fig. 7). The anodic oxidation of most semiconductors is a complicated process. The reaction of GaAs, for example, which can be represented schematically by
GaAs + 6hVB + − − Ga(III) + As(III)
−→
(25) requires six valence-band holes to form trivalent gallium and arsenic species.
Chemical reactions are obviously also involved. The ﬁnal products depend on the nature of the electrolyte solution. The information that can be obtained from cyclic voltammetry is clearly rather limited.
Kinetic studies in which oxidation of a reducing agent present in solution competes with oxidation of the semiconductor have yielded a wealth of information, allowing quite detailed dissolution mechanisms to be proposed [88]. Generally, in these studies an RRDE is used; the products formed at the semiconductor disk are detected electrochemically at a noble metal ring.
Semiconductors can be etched under open-circuit conditions with an oxidizing agent that is capable of being reduced by extracting electrons from the valence band, that is, creating holes in the band (see Sect. 2.1.3.2.3) [15]. Information on hole injection can be obtained from electroluminescence studies on ntype electrodes. Electrochemical studies of such ‘‘electroless’’ etching systems have shown that, during dissolution, intermediates are formed with energy levels in the band gap [89]. From such states, electrons can be thermally excited into the CB.
If this occurs in an n-type semiconductor

79

80

2 Experimental Techniques

under depletion conditions then the injected electron can be detected as current in the external circuit. By measuring this anodic current as a function of injection rate and temperature, one gets extensive information about both chemical and electrochemical processes [88, 89]. Gomes and coworkers have shown that electron injection and RRDE competition studies are complementary [90]. The contribution of electron injection to the total anodic oxidation rate is, for most etching systems, small compared to that of holes.
Electron injection from reaction intermediates of the oxidation of n-type semiconductors can be observed as quantum efﬁciency larger than unity in photocurrentpotential measurements. There are two striking examples in the literature: the photoanodic dissolution of n-type silicon in HF solution [91, 92] and of n-type InP in HCl solution [54]. In these cases the quantum efﬁciency at low light intensity is exceptionally high, four for silicon and two for InP. In the case of Si, this means that only one photon (and thus one hole) is required to dissolve each silicon atom; three electrons are injected into the conduction band
Si + hVB + − − Si(IV) + 3eCB −
−→

(26)

At higher light intensity, the quantum efﬁciency drops indicating that reaction steps involving valence-band holes take over from electron injection steps. A quantum efﬁciency of two for InP means that three of the six oxidation steps require minority carriers and thus photons. As for silicon, the quantum efﬁciency decreases (from 2 to 1) as the light intensity is increased. As described in the previous section, IMPS can be used very effectively to study the mechanisms of ‘‘photocurrent multiplication’’ reactions. The method has proved particularly successful for the silicon and

InP systems, providing information about the sequence of the reaction steps and the magnitude of the rate constants for majority carrier injection [54, 92].
Because electron injection is detected during dissolution of n-type semiconductors under illumination, it seems likely that it should also occur during dissolution of p -type semiconductors in the dark. There are two ways in which this can be checked.
If the injected electrons recombine radiatively with majority carriers (the holes), light is emitted. EL corresponding to bandband recombination has been observed during anodic dissolution of p -type InP in
HCl solution (Fig. 12) [93]. The quantum efﬁciency for light emission, that is, the number of photons emitted per electron passed through the external circuit, was very low (approximately 10−6 ). This probably is due to a low rate of electron injection and a high rate of nonradiative recombination at the etching surface. Another approach that allows one to distinguish quantitatively between electron and hole contributions to the current is the p-n junction conﬁguration [94, 95]. This technique has been used for the study of the anodic dissolution of silicon in alkaline solution.
This is an unusual system showing a novel coupling of chemical and electrochemical steps [95]. Both p -type and n-type silicon can be oxidized electrochemically in the dark. In both cases, passivation occurs if a limiting current is exceeded. The anodic peak current observed as n-type silicon passivates is comparable to that of p -type silicon. Anodic current from an n-type electrode in the dark can, under normal circumstances, only result from electron injection into the CB. Because there is no electron donor present in solution, electron injection must occur from a surface species. This species very likely arises as a result of the chemical-etching reaction

2.1 Photoelectrochemical Systems Characterization
1.0

IEL
[a.u.]

0.8
0.6
0.4

1.0
0.8
0.6
0.4
0.2
0.0
800

900

1000

Wavelength
[nm]

0.2
0.0

j
[mA/cm2]

3.0

2.0

1.0

0.0
0.0

0.2

0.4

0.6

V vs SCE
[V]
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).

Fig. 12

of silicon by water. Anodic current with p type silicon can be due to a hole reaction or, as in the case of n-type silicon, to electron injection. Measurements with a p-n junction electrode showed the latter to be the case [95]; most of the anodic current can be attributed to minority carrier injection.
Electrical impedance measurements are essential in semiconductor dissolution studies. Under depletion conditions, a change in applied potential will usually give rise to a change in the potential drop within the semiconductor.
Impedance measurements combined with

Mott-Schottky theory can be used to determine the ﬂat band potential (see
Sect. 2.1.3.1). This then allows one to calculate the band-bending for any value of the applied potential, if the Helmholtz potential remains constant (i.e. the band edges are pinned). The band-bending determines the concentration of majority carriers at the surface (see Eqs. 25 and 26).
In a p -type electrode these are holes, which are essential for the dissolution reaction.
In an n-type electrode, the band-bending determines the surface electron concentration and thus, the rate of recombination

81

82

2 Experimental Techniques

with photogenerated or injected holes; this process competes with the oxidation of the semiconductor by holes.
The assumption of pinned band edges is very often not valid. Lincot and Vedel [50] in an early study of the photoanodic dissolution of n-type CdTe used EIS to show that the Fermi level becomes pinned in a wide potential range positive with respect to Vfb . This means that in this range the band-bending within the semiconductor remains constant while the potential changes across the Helmholtz layer. In their analysis, Lincot and Vedel consider the rate constants for change transfer to be exponentially dependent on the Helmholtz potential. In many early studies of anodic oxidation of p -type semiconductors it was tacitly assumed that the reaction was controlled solely by the surface hole concentration, that is, by the potential drop across the space charge layer of the solid. The role of the Helmholtz potential was neglected.
For p -type electrodes, anodic oxidation occurs at potentials close to the ﬂat band value and under accumulation conditions.
In addition, there will be a high density of surface states as a result of the breaking of surface bonds. In this case, a change in the applied potential is likely to be distributed partly or even completely over the Helmholtz layer. The potential dependence of the anodic current then is due to (in part) the changes in the rate constants for hole capture resulting from changes in the Helmholtz potential. Electrical impedance spectroscopy is necessary to decide whether etching is under space charge layer or Helmholtz-layer control.
The latter was shown to be the case for the dissolution of p -GaAs in acidic electrolytes [96]. Vanmaekelbergh and Searson in a study of the dissolution of p -type silicon in HF solution showed that EIS

also can be used to get information about electron injection processes [97].
Electrochemical measurements of the type described earlier give indirect evidence about dissolution processes. More direct chemical information can be obtained from in-situ spectroscopies, in particular from IR and Raman methods.
Chazalviel and coworkers have showed the power of this approach in studies on silicon and GaAs [73, 98, 99]. Electrochemical and spectroscopic techniques are macroscopic methods giving a view of the whole electrode surface. To study semiconductor dissolution at the microscopic (atomic) level, one needs techniques such as scanning tunneling microscopy (STM) and atomic force microscopy (AFM). The anodic and chemical dissolution of silicon has been studied in very elegant work by
Allongue and coworkers [100–102].
2.1.4

Sensitizer-based Photoelectrochemical
Systems

Dye molecules adsorbed on an electrode form, in principle, the simplest class of photoelectrochemical system. The basic scheme presented at the end of
Sect. 2.1.1.2 is directly applicable for understanding the mechanism of photocurrent generation. The dye molecules act as the light-absorbing species. If injection of a photogenerated charge carrier into the electrolyte or electrode competes effectively with relaxation of the excited state, a photocurrent might be observed in the external circuit. Singletto-triplet crossing often leads to a relatively long-lived excited state, which allows electron-hole separation to occur by an electron transfer process. The photochemistry and photoelectrochemistry of dyesensitized electrodes has been studied

2.1 Photoelectrochemical Systems Characterization

extensively, and the research is still continuing. Dye-sensitization of photochemical reactions is a key topic in photographic research [103, 104]. At present, there is a considerable effort being devoted to dye-sensitized porous photoelectrochemical solar cells [16, 105]. The fundamental research is focused on the chemical bonding between the light-absorbing molecules and the electrode surface, the mechanism of light absorption, and the dynamics of charge injection and recombination in the dyes. In this research, scanning probe methods [106, 107] and time-resolved optical spectroscopy [108, 109] are used together with photoelectrochemical characterization [24]. The reader is referred to a more specialized review for details [24].
Electrodes to which insulating lightabsorbing nanocrystals, instead of dyes, are attached form a relatively new class of system. With colloidal solution chemistry, a large variety of insulating nanocrystals can be prepared [110–113]. Well-known examples are II–VI compounds (CdS,
CdSe, and ZnO), III–V compounds (InP,
InAs, and GaN), and transition-metal oxides. In addition, there are various methods in solid-state chemistry and electrochemistry for preparing nanocrystals directly on surfaces. For instance,
CdS, CdSe, and PbS nanocrystals can be electrochemically deposited on gold electrodes [114]. Electrochemical oxidation of
Si leads to porous silicon that may contain a large number of Si nanocrystals [115].
Nanocrystalline colloidal systems have been extensively characterized with optical spectroscopy [110–113]. There are two effects that are essential for understanding the electronic and optical properties of insulating nanocrystals. First, quantum conﬁnement of the electron waves in the nanocrystal leads to discrete electron states at the top of the valence band

and bottom of the CB and to an increase of the band gap energy with respect to that of a macroscopic crystal. The energy-level spectrum shifts from that of a classical insulator to that of a molecule with a reduction of the dimensions in the 20–1-nm range. Second, a considerable fraction of the atoms of a nanocrystal lies at the surface. This leads to surface-electron states.
States of energy in the band gap can have a strong inﬂuence on the optical properties of nanocrystals. It is clear that the surface chemistry is extremely important.
Organic and inorganic molecules can passivate surface states, thus removing them from the optical gap. Functionalized capping molecules play an important role in providing stability against coagulation and in allowing the attachment of nanocrystals to solid (electrode) surfaces (see the following text).
Early reports have shown photoelectrochemical activity when, for example,
CdS and PbS nanocrystals are attached to a metal electrode in a sub-monolayer array [19, 20, 116–120]. Clearly, photoexcitation of the nanocrystals can lead to a long-lived state, which allows one of the charge carriers to be transferred from the nanocrystal before recombination occurs
(see Sect. 2.1.1.2).
A macroscopic PbS crystal has a band gap of 0.41 eV. Because of the small effective mass of the electrons and holes
(me,eff = mh,eff = 0.09 × me ), strong sizequantization occurs. The absorption spectrum of an aqueous suspension (of polyvinyl alcohol-capped) PbS nanocrystals, 6.5 nm in diameter (see TEM picture) is shown in Fig. 13. The HOMO-LUMO optical transition occurs at 2.1 eV, and two other absorption peaks are seen at 3.2 and
4.3 eV. When a gold electrode is immersed in this colloidal solution, PbS nanocrystals

83

2 Experimental Techniques
Fig. 13 Absorption spectrum of an aqueous suspension of nanocrystalline, size-quantized PbS particles (capped with polyvinylalcohol). A HR-TEM image of a typical PbS nanocrystal is shown in the insert; the diameter of the nanocrystals is about 6.5 nm (from
Ref. 122).

1.0
0.8

Absorbance
[a.u.]

84

a
0.6
0.4
0.2

10

[nm]

0

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, see Fig. 14 [121, 122].
The photoelectrochemical activity of a Au/Q-PbS electrode, illuminated with chopped green light, is shown in Fig. 14 for three different aqueous electrolyte solutions. With a 1 M KCl solution (case a), photocurrent transients are observed when the electrode potential is more positive then −0.4 V (versus SCE). Illumination gives a cathodic current that decreases to zero with increasing time; turning off the light induces an anodic current. Clearly, illumination leads ﬁrst to the transfer of an electron from gold to the photoexcited hole in the PbS nanocrystals (a
‘‘cathodic’’ transfer), followed by transfer of the photoexcited electron to empty states in the gold (an ‘‘anodic’’ transfer).
Energy relaxation in the gold due to electron transfer competes with electron-hole recombination in Q-PbS. For this to occur, the electrochemical potential in the gold electrode must be located between the energy levels corresponding to the photoexcited electron and hole. Recombination transients are suppressed when the
Fermi level is above 0.4 eV (versus SCE), which indicates that the long-lived excited

state contains an electron located at 0.4 eV versus SCE, thus considerably below the
LUMO of Q-PbS. The much faster kinetics of the photoexcited hole transfer indicates that the hole in the long-lived excited state is still delocalized, that is, it occupies the
HOMO. It was thus concluded that the excited state in Q-PbS leading to photoinduced electron transfer consists of a trapped electron and a delocalized hole in the HOMO (denoted as Q(etrap , hHOMO )).
The following scheme describes the decay dynamics of the long-lived state in Q-PbS: hν relaxation

−→
−−−
Q − − Q∗ − − − → relaxation − −→
Q(etrap , hHOMO ) − − − Q + heat
(27)
transfer from Au

− − − −→
Q(etrap , hHOMO ) − − − − − Q− (etrap )
(28)
transfer to Au

Q− (etrap ) − − − − → Q
−−−−

(29)

Interestingly, when tartrate is added as an electron donor to the solution, an anodic photocurrent is observed in the potential range positive with respect to −0.4 V

2.1 Photoelectrochemical Systems Characterization
1.0
0.0
−1.0

Light on

−2.0
−3.0

100

Light off

(a)

J
[µA /cm2]

50

Light off

1.0

0
0.5

0
Light on

50

100

150

[nm]

0.0
(b)

1.0
0.0
Light on

−1.0
(c)

Light off

−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 donation from tartrate can compete effectively with electron donation from the gold electrode [i.e. Eq. (28)]: transfer from tartrate

−−−−−−
Q(etrap , hHOMO ) + Red − − − − − − →
Q− (etrap ) + Ox+

(30)

In contrast, a cathodic photocurrent is observed when Fe(CN)6 3− is added to the solution (Fig. 14, case c). Thus, electron donation to the oxidizing species in solution competes effectively with electron donation to gold; it can be clearly seen in Fig. 14, case c that the electrochemical activity in the latter solution does not stop when the electrode Fermi level is increased

above 0.4 eV. This is logical, because the cathodic photocurrent corresponds to electron transfer to the oxidized species, not to the electrode.
Photoelectrodes, consisting of CdS nanocrystals that are directly attached to an electrode (gold or conducting oxide) or indirectly via a dithiol molecular linker have been studied extensively [19,
20, 116–120]. Here, we will focus on work that demonstrates the power of small-amplitude methods in photoelectrochemical characterization [26, 123, 124].
By measuring the photoelectrochemical activity of gold/Q-CdS electrodes, as the potential of the gold electrode was varied, it could be concluded that photoinduced electron transfer between the gold surface

85

86

2 Experimental Techniques

and the quantum dots is due to a longlived state in Q-CdS. This state, denoted as
[Q(eLUMO , htrap )], consists of an electron in the LUMO and a hole trapped in a level about 0.6 eV below the LUMO: hν relaxation

−→
− − −→
Q − − Q∗ − − − − relaxation − −→
Q(eLUMO , htrap ) − − − Q + heat
(31)
Time-resolved photobleaching experiments on Q-CdS colloidal solutions showed the existence of a long-lived state (lifetime 50 ms) that, very probably, corresponds to Q(eLUMO , htrap ) [125].
Figure 15 shows the optoelectrical transfer function (see Sect. 2.1.2.4), measured with near UV-light from an argon laser for a 1 M KCl solution (a, b) and a KOHtartrate solution (c, d). In a and c, the transfer function is plotted in the complex plane with the frequency as a parameter, in b and d, the modulus of the transfer function is plotted versus the modulation frequency ω. In the KOH-tartrate solution (c, d), an anodic vector is observed at sufﬁciently low frequency. This means that ∂jPH /e∂ > 0 (see Sect. 2.1.2.2). This agrees with the observation of a steady state anodic photocurrent, which increases with increasing light intensity. The anodic vector shrinks with increasing modulation frequency ω and, eventually, becomes zero for ω > 5 × 104 s−1 . Two semicircles with characteristic frequencies ωc = 6 s−1 and ωb = 700 s−1 are observed (measurements at 25 ◦ C). Measurements in the temperature range 0–60 ◦ C showed that ωb is temperature-independent, while ωc increases with increasing temperature.
Hence, ωc corresponds to the rate of electron donation from tartrate to the hole trapped in Q-CdS, while ωb is the rate of

photoinduced electron transfer from the
LUMO in Q-CdS to empty states in the gold electrode: ωb Q(eLUMO , htrap ) − − Q+ (htrap )
−→
ωc

−→
Q+ (htrap ) + Red − − Q + Ox+

(32)
(33)

With a 1 M KCl solution, the low-frequency limit ∂jPH /e∂ of the optoelectrical transfer function is zero; this agrees with the fact that there is no steady state photocurrent. The high-frequency limit of the optoelectrical transfer function is also zero. Starting from 105 s−1 , an anodic vector develops with decreasing modulation frequency. This corresponds to electron transfer from the Q-CdS LUMO to gold.
The rate ωb of this process cannot be clearly distinguished from the discharge frequency of the photoelectrochemical cell.
Electron transfer from the LUMO to the gold is probably faster than 3 × 104 s−1 .
At lower frequencies, the modulus of the optoelectrical transfer function shrinks to zero. This is because of the electron backtransfer from gold to the empty level in
Q-CdS following the faster LUMO-to-gold electron transfer: ωa −→
Q+ (htrap ) − − Q

(34)

The rate of this transfer (ωa ) is 6 × 103 s−1 and is independent of temperature. Thus, at sufﬁciently low modulation frequency, both the processes represented by Eqs. (32) and (34) are in phase with the modulated light intensity, and the resulting photocurrent is zero. When no redox system is present, relaxation of the long-lived excited state in Q-CdS, Q(eLUMO , htrap ), occurs via consecutive steps involving electron transfer from the LUMO to the gold, and from the gold to the trapped hole. This relaxation

2.1 Photoelectrochemical Systems Characterization
103 × 2 π rad/s ωa Imaginary [ j(ω)/e Φ(ω)]
[10−3]

1.0

102 × 2 π rad/s
0
10 × 2 π rad/s

−0.5

−1.0 104 × 2 π rad/s ωb

0

1.0

0.5

1.5

2.0

1.0
0.5

102

103

104

Frequency
[s−1]

(b)

1
10−1 × 2π rad/s
0

−1
103 × 2 π rad/s

−2
0

1

ωb
′

2

ωc
10 × 2 π rad/s 102 × 2 π rad/s

3

Modulus [ j(ω)/e Φ(ω)]
[10−2]

Imaginary [ j(ω)/e Φ(ω)]
[10−2]

1.5

2.5

Real [ j(ω) /e Φ(ω)]
[10−3]

(a)

(c)

Modulus [ j(ω)/e Φ(ω)]
[10−3]

0.5

3
2
1

100

4

Real [ j(ω) /e Φ(ω)]
[10−2]

4

(d)

101

102

103

104

Frequency
[s−1]

Fig. 15 The optoelectrical transfer function measured for a Q-CdS covered gold electrode in an electrochemical cell. The attached quantum dots have an average diameter of 5 nm. (a,b).
Results obtained in an aqueous solution
(pH = 7) with 1 M KCl (297 K). In (a), a plot of the transfer function is shown in the complex plane with the modulation frequency as a parameter. The meaning of the transfer function and of the characteristic frequencies (ωa , ωb ) is

discussed in the text. In (b), the modulus of the transfer function is plotted vs. the modulation frequency. (c,d) Results obtained in an aqueous solution (pH = 14) with 0.5 M tartrate as an electron donor. (c) Plot of the transfer function in the complex plane with the modulation frequency as a parameter. (d) Plot of the modulus of the transfer function versus the modulation frequency (from Ref. 123).

competes effectively with relaxation of the excited state in the particle.
It can be concluded that measurement of the optoelectrical transfer function in a photoelectrochemical cell is a powerful technique for studying the mechanism and kinetics of photoinduced electron transfer.
The technique has been further exploited

to determine the rates of photoinduced electron transfer in gold/spacer/Q-CdSe assemblies [121, 126, 127]. Cyclohexylidene disulﬁdes form well-ordered and rigid molecular monolayers on gold (111) surfaces because of strong Au-S bonding and intermolecular van der Waals interactions [121]. Q-CdSe quantum dots were

87

88

2 Experimental Techniques

linked covalently at the other end of the molecule via the S-termination. The rates of photoinduced electron transfer from the LUMO of Q-CdSe to gold (kL,Au ), and from gold to the trapped hole in
Q-CdSe (kAu,T ) were obtained from the optoelectrical transfer function. An exponential decrease of the electron transfer rate with increasing length of the spacer molecule is found, with a decay parameter
˚
β = 0.5 A−1 . This low value (the value in vacuum is typically four times larger) indicates a strong through-bond electronic coupling in the cyclohexylidene spacer molecules. This agrees with the result of quantum chemical calculations and with the strong S–S coupling shown by photoelectron spectroscopy [126, 127].
The measurement of changes in the optical properties of nanocrystals attached to an electrode caused by variations in the electrode potential forms a challenging but interesting characterization method. Recently, a bleaching of the HOMO-LUMO transition of CdS nanocrystals has been observed when the electrode Fermi level was in resonance with the LUMO [128]. This led the authors to conclude that the photoexcited state in CdS contains an electron in the LUMO (and a trapped hole). This is in agreement with the results obtained by photoelectrochemical characterization of gold/Q-CdS electrodes (see earlier).
2.1.5

Porous Photoelectrochemical Systems
Introduction
Here, we deﬁne a porous solid as a phase that contains empty spaces that are interconnected. Thus, there is a single-solid phase that can be permeated with a second phase. Porous metals have been extensively studied, and are widely applied in
2.1.5.1

heterogeneous catalysis, storage batteries, fuel cells, and super-capacitors [41].
Porous insulating or semiconducting phases have been considered as photochemical devices for the light-stimulated oxidation of organic waste components
[heterogeneous photocatalysis] [110–113,
129]. Porous photoelectrochemical systems have been studied extensively only in the last ten years [16, 105, 130–137]. Their fabrication is more demanding than that of photochemical systems, because electrical work is delivered in an external circuit upon illumination (see Sect. 2.1.2.2). In a porous photoelectrochemical system, the solid phase is insulating or semiconducting, whereas the permeated phase is an electrolyte solution. The current in the solid phase is due to the motion of electrons or holes; the current in the solution is due to the motion of an oxidized or reduced species.
A porous photoelectrochemical system can be prepared by the deposition of colloidal particles on a conducting substrate. It is essential that the particles are electrically connected and that there is electrical contact between the particles and the conducting substrate. Furthermore, the pore system should form a singlepermeated phase. A well-known example of such a system is the particulate TiO2 photoelectrode, which forms the basis of the photoelectrochemical solar cells proposed by O’Regan and Gr¨tzel and other a groups [16, 130–137]. A SEM picture of a TiO2 network, consisting of interconnected spheres with a diameter of 30 nm, is shown in Fig. 16(a); electrical contact between the particles and between the conducting substrate and the particles was achieved by slight sintering at 450 ◦ C.
An alternative route exists for the preparation of porous semiconductors. Many n-type single crystals, such as GaP, GaAs,

2.1 Photoelectrochemical Systems Characterization

Acc.V.
10.0 kV

Magn.
108639

200 nm

(a)

(b)

00018

3 µm

(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.

Fig. 16

Si, SiC, and TiO2 can be transformed into a porous network by anodic etching under conditions of severe band-bending [18,
115, 138–144]. Under such conditions, surface electrons located at the top of the valence band or in band gap states can tunnel through the gap into the CB; the surface localized holes generated in such a way are consumed in anodic dissolution of the material. The rate of interband tunneling (Zener breakdown [145]) is strongly dependent on the electric ﬁeld and the presence of surface defects. This

would explain why anodic dissolution is so strongly nonuniform over the surface and eventually leads to the formation of a porous semiconductor. However, many questions remain regarding porous photoelectrochemical etching. A well-studied example is macroporous GaP [139, 143].
When n-type GaP is subjected to a potential of 5 V or more (versus SCE) a highly porous network is formed (Fig. 16b) with dimensions of the structural units and the pores in the 100-nm range. The GaP network retains its crystallinity [146, 147].

89

90

2 Experimental Techniques

A ‘‘single-crystal GaP sponge’’ is different from a particulate network in that there are no grain boundaries and the connection between the network and the bulk
GaP matrix is ideal.
Special Properties of Porous
Photoelectrochemical Systems
Porous photoelectrochemical systems consist of an insulating or semiconducting solid network permeated with a conducting electrolyte solution; the dimensions of the solid structures and pores are in the 1–500-nm range. A typical semiconductor/electrolyte interface has a width of between 0.5 nm (the Helmholtz layer in a concentrated electrolyte solution) and 100 nm (typical depletion layer in a semiconductor). Thus, the width of the solid/electrolyte interfacial layer can be
2.1.5.2

much smaller but also larger than the dimensions of the solid structures and the pores. Therefore, porous semiconductors can show remarkable charge-storage properties. Other relevant length scales are the wavelength of visible light (400–700 nm) and the diffusion length of charge carriers before recombination. These length scales also can be in the same range as the dimensions in the porous structure; this leads to striking optical and electrodynamic properties. Extensive research is being performed in these ﬁelds, and a comprehensive review is beyond the scope of this section. Instead, the electrostatic, optical, and electrodynamic properties of porous semiconductor (electrodes) are brieﬂy discussed and the reader is referred to more detailed publications.

EF,n

Depletion

EF,n

n-type
Semiconductor

Solution
(a)

Depletion
Solution

200 nm

(b)

200 nm

EF,n
Depletion
Solution
(c)

20 nm

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

2.1.5.2.1 Penetration of the Interfacial
Layer in a Porous Semiconductor Electrode
In Fig. 17(a), the energetics of a typical interfacial region between an n-type semiconductor and an electrolyte solution is shown (see also Sect. 2.1.2.2). Electronic equilibrium exists between the semiconductor and a redox system present in the solution: the electrochemical potential of electrons µe in the solid is equal to that in the liquid phase, and does not change with the spatial coordinate x , perpendicular to the solid/liquid interface. The electrochemical potential of the electrons is also equal to the electron Fermi level, denoted as EF,n and can be written as ref EF,n (x) − EF = −e[ϕ(x) − ϕ ref ]

+ kT ln[n(x)/nref ]

(35)

where n(x) is the electron concentration in the CB of the semiconductor (i.e. the free-electron concentration), and −eϕ(x) the potential energy of an electron.
It is clear from Fig. 17(a) that a depletion layer for free electrons is present near the solid/solution interface. From Eq. (35) it follows that n(x) = nbulk e−[Ec (x)−Ec,bulk ]/kT

(36)

Here, EC (x) and EC,bulk is the energy of the CB edge at position x in the depletion layer and in the bulk, respectively; EC (x =
0) − EC,bulk is the band bending. The concentration of free electrons in the bulk, nbulk , is determined by the density of dopant atoms. The depletion layer of an n-type electrode is positively charged, because the ionized dopant atoms are not fully compensated by free electrons. The counter charge is located on the electrolyte side, very close to the interface (Helmholtz layer, width 0.5 nm).
Consider now electronic equilibrium for a macroporous electrode. The structural

units in macroporous GaP (doping density
1017 cm−3 ) have typical dimensions of about 150 nm [138, 139, 146, 147]. This means that, if the band bending of the electrode is smaller than one 1 eV, WSC
(50 nm, see Sect. 2.1.2.2) is smaller than half the width of a structure: the inside edge of the depletion layer is in the porous network; there is still a semiconducting bulk region in the porous solid (see
Fig. 17a). Therefore, the total surface area of the inside edge of the depletion layer is very large and so is the interfacial capacitance that is approximately equal to the internal surface area times CSC .
However, if the band-bending is more than
3 eV, WSC is larger than half the width of typical structures in macroporous GaP; the entire porous GaP structure is depleted of electrons, the inside edge of the depletion layer is located in the bulk substrate, outside the porous ﬁlm (Fig. 17b). The surface area of the inside edge of the depletion layer then corresponds to the macroscopic (geometrical) area of the electrode. Ern´ and coworkers measured the ine terfacial capacitance of macroporous GaP electrodes as a function of the electrode potential [138, 139]. It was found that the capacitance is large for sufﬁciently small band-bending (interfacial layer in the porous solid) and decreases to the value of a nonporous interface at larger band-bending. Similar effects have been found with macroporous SiC and Si electrodes [18, 141]. In fact, the interfacial capacitance is a measure for the surface area of the macroporous network, with the width of the depletion layer, WSC , as a measuring stick.
Finally, in nanoporous networks, the structural units have dimensions in the range 25–1 nm. For instance, the TiO2 particulate electrode forming the basis

91

92

2 Experimental Techniques

of a dye-sensitized solar cell consists of nanocrystals of 10–25 nm (Fig. 16b). It is clear that, under conditions of freeelectron depletion, band bending is nearly absent. (Fig. 17c). When electrons are supplied to the nanoporous system (by injection from a photoexcited dye or from a conducting substrate), the difference between EF and EC becomes smaller in the entire nanocrystal and the conductivity of the system increases.
Under conditions of electron accumulation, the interfacial capacitance C of a semiconductor/electrolyte contact tends to that of the Helmholtz layer (see
Sect. 2.1.3.1 with CSC ≥ CH ). The width of the interfacial double layer is reduced to about 0.5 nm; hence, it follows the internal surface of a porous electrode. As a result, the overall interfacial capacitance of a nanoporous system can be huge, being determined by the product of the total internal surface area of the system and the
Helmholtz-capacitance per unit geometric surface area [148, 149].
2.1.5.2.2 Charge Storage in a Porous Semiconductor Electrode In a bulk singlecrystal electrode with a ﬂat semiconductor/electrolyte interface, the electrochemical potential can be changed considerably by a relatively small change in the number of electrons present in the semiconductor.
This is due to the relatively small interfacial capacitance per unit of geometric surface area. In contrast, the entire threedimensional structure is interfacial in the case of macroporous and nanoporous systems interpenetrated with a conducting electrolyte. Consequently, the capacitance per geometric area can be very large, which means that a relatively large number of electrons are needed for a change in the electrochemical potential. This is exempliﬁed by the considerable variation

in microwave absorption induced by a given change in the Fermi level observed at macroporous GaP [150]. The large dynamic range in the total number of free electrons in a semiconducting network can be used as a tool for photoelectrochemical characterization. For instance, the optical absorbance by free carriers in nanoporous
TiO2 electrodes has been used to detect changes in the free carrier concentration in this system due to voltage modulation and modulation of the photoexcitation rate [151].
Clearly, these considerations are of importance for systems in which porous semiconductors or insulators are used
[super-capacitors, (chemical) sensors, and electrochromic devices] [152, 153].
2.1.5.2.3 Charge Storage in a Quantum Dot
System Hoyer and coworkers [154, 155] reported that the electrochemical potential of a porous particulate ZnO electrode
(with ZnO dots of 5-nm diameter) shifts to higher energy with increasing electron density n in a much more pronounced way than predicted from Eq. (35). This is caused by two physical phenomena that become important with very small particles (quantum dots). First, as a result of size-quantization, the energy levels of the CB (and valence band) become discrete and separated by considerable energy gaps (typically in the range of 0.1 eV).
Even in assemblies in which the nanocrystals are covalently linked, size-quantization may persist. Second, the charging energy per particle, e2 /Cparticle , (typically
0.01–0.1 eV) can form an important contribution to the electrochemical potential of nanometer-sized particles. Investigating nanoparticulate ZnO electrodes similar to those used by Hoyer, Meulenkamp [156] reported that the relationship between the electrochemical potential and the electron

2.1 Photoelectrochemical Systems Characterization

density depends on the nature of the electrolyte solution; this clearly shows the importance of the charging energy.
Size-quantization and single-dot charging energy play an important role in electron transport in metal–nanodot–metal double barrier tunnel junctions [157–161]. Study of electron transport in two-dimensional or three-dimensional assemblies consisting of nanometer-sized particles is still in its infancy; one may expect that single-dot charging (leading to Coulomb-blockade) and size-quantization will result in interesting and novel transport phenomena.
2.1.5.2.4 Light Scattering in Macroporous
Semiconductors The dimensions of the structural units and pores in macroporous semiconducting and insulating networks are often in the 100-nm range.
This is the same range as the wavelength of visible and UV light. Because of the structural variation of the refractive index on the wavelength scale, visible light can be strongly scattered in macroporous networks [162–165]. In macroporous GaP, for instance, the propagation of red (sub–band gap) light is strongly attenuated [146, 147]. The importance for photoelectrochemical systems lies in the fact that the effective absorption length of supra–band gap light in macroporous systems is reduced considerably with respect to that in single crystals. For example, the penetration depth 1/α of green light in a bulk GaP single crystal is about 10 µm, whereas macroporous GaP networks with a thickness of only 2 µm completely absorb green light [138, 139]. Visible and nearUV light is not scattered in nanoporous systems because the structural variation in the refractive index occurs on a scale much smaller than the wavelength of light.
In such a case, the effective absorption

coefﬁcient of the light can be estimated from effective medium theory.
2.1.5.2.5 Electron-hole Photoexcitation by
Sub-band Gap Light In a dye-sensitized porous photoelectrode, an electron from the dye is photoexcited into the CB by a photon of energy considerably below the band gap of the semiconductor. The dye molecules are anchored on the internal surface of the porous semiconductor. Light absorption is very effective because of multiple interactions of a single photon with the dye molecules. Similarly, a porous semiconductor without dye molecules may absorb sub-band gap light, and this may lead to photogeneration of free electrons and holes. The mechanisms of free carrier generation with sub-band gap light in macroporous GaP photoelectrodes have been investigated in detail [166, 167].
Surface-localized electrons involved in two-photon transitions and in a coupled optical-thermal transition were found to give rise to signiﬁcant sub-band gap photocurrent in this system.

Effective Electron-hole Separation
The ability of porous photoelectrochemical systems to separate effectively electrons and holes is widely known since the presentation of the dye-sensitized particulate
TiO2 solar cell [16, 105, 130–137]. In this system, the photocurrent quantum yield
(the number of electrons counted in the external circuit as photocurrent divided by the number of absorbed photons) is close to unity. This means that electron-hole pair recombination is essentially absent. Efﬁcient separation of photogenerated electrons and holes was demonstrated with several other photoelectrochemical systems [105, 130–137]. Photovoltaic devices based on permeated hole-conducting and
2.1.5.2.6

93

2 Experimental Techniques

electron-conducting polymer phases also show an enhanced photocurrent quantum yield [168–170]. The origin of this desirable feature can be demonstrated by comparing the photocurrent quantum yield of a GaP bulk single crystal with a macroporous GaP photoelectrode (see Fig. 18).
GaP absorbs light of energy between 2.2 and 2.7 eV by an indirect transition (hence, weakly, the absorption depth of green light is 10 µm). Because of the diffusion length
Lmin of minority carriers in n-GaP is relatively small (about 50 nm), the penetration depth 1/α is much larger than the width of the retrieval region (Lmin + WSC ). This results in a photocurrent quantum yield of about 0.01 [∼ α(Lmin + WSC )]; that is,
=
99% of the absorbed photons are converted into heat by recombination in the bulk. A macroporous GaP electrode, on the other hand, shows a photocurrent quantum yield of unity in a large potential range. The reason for this spectacular enhancement is illustrated in Fig. 17(b): in a macroporous GaP electrode, all minority carriers (holes in the valence band) are photogenerated within a distance from the interface that is smaller than the diffusion length Lmin . Minority carriers can thus

reach the surface without recombining.
If surface recombination is slower than transfer of the hole to the liquid electrolyte phase, the quantum yield will be close to one, in agreement with the experimental result. A considerable enhancement of the photocurrent quantum yield has been observed in several porous photoelectrochemical [16, 18, 105, 130–139] and photovoltaic systems [168–170].
2.1.5.2.7 Luminescence from Porous Electrodes As in bulk systems, photogeneration of charge carriers in a porous electrode or injection of minority carriers from solution can lead to light emission [25]. In a macroporous system in which the depletion layer can follow the contours of the porous matrix, one does not expect signiﬁcant differences between bulk and porous electrodes with regard to the potential dependence of the emission. If, however, the porosity is high and the dimensions of the structures become very small (e.g. 2.2 eV) (from Ref. 139).

2.1 Photoelectrochemical Systems Characterization

photoluminescence in the visible spectral range [171]. Of the various explanations given for this phenomenon, the most widely accepted is that of size quantization. As a result of the conﬁnement of charge carriers within nanometer-sized structures in the porous matrix, the effective band gap is widened while the ratio of radiative to nonradiative recombination is considerably enhanced [171].
In situ luminescence measurements provide information about the physical and chemical properties of porous silicon and about charge-transfer reactions at the silicon/solution interface [25].
In contrast to the photoluminescence from a single-crystal electrode (see Fig. 11), the emission from a porous n-type silicon electrode is constant at positive potentials and decreases only in the range negative

with respect to the ﬂat band potential
(Fig. 19) [172, 173]. Because of the absence of an electric ﬁeld in the porous layer and the strong conﬁnement of the carriers, the electron and hole are not separated in the potential range corresponding to depletion in a bulk electrode. As at a single-crystal electrode, hole injection from a strongly
•
oxidizing species such as SO4 − (generated electrochemically by the reduction of
S2 O8 2− at the electrode) gives rise to visible electroluminescence in porous n-type silicon [172, 173]. The emission increases in the range in which the photoluminescence decreases [172, 173] (see Fig. 19). An interesting aspect of the electroluminescence is the voltage tunability of the colour.
The emission maximum shifts to shorter wavelength as the potential is scanned to negative values, until ﬁnally the emission

1.0
EL
PL

Intensity [a.u.]

0.8

0.6

0.4

0.2

700 nm
750 nm
800 nm

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).

95

2 Experimental Techniques

is quenched. EL can be expected from a quantized structure if it is populated by an electron supplied from the bulk silicon.
Because of the larger band gap of porous silicon there is a mismatch of the CB edges.
As a result the Fermi energy of the bulk silicon must be raised to a level close to the conduction band edge of the quantized structure. This will occur for larger particles at less negative potential because the mismatch of the CB edges is small in this case. As a result, long wavelength emission will be turned on ﬁrst. Gradually as the Fermi level is raised further, that is, as the potential is made more negative, the smaller particles will participate and the emission maximum will shift to shorter wavelength. Quenching of EL and PL has been attributed to Auger recombination.
This can explain why the rise in EL on going to negative potentials is coupled to the quenching of PL. To give EL, an electron is required in a particle to create a hole via S2 O8 2− reduction (compare with
Eqs. 23 and 24). On the other hand, photoexcitation of a particle already occupied by an electron leads to Auger recombination; that is, the PL is quenched. At more negative potentials, the supply of an electron to a particle in which an electron-hole pair is present, leads to Auger quenching of the EL [172, 173].
EC

Energy

96

e

e

x

d

2.1.5.3 Electron Transport
2.1.5.3.1 Electron Diffusion,

Collection, and Recombination A unique feature of a porous photoelectrochemical system is the permeation of the solid semiconducting network by an electrolyte solution on a scale smaller than Lmin (Fig. 17b). As a result, one of the photogenerated charge carriers can be transferred to the solution. In the following, we will assume that the hole is transferred to the electrolyte solution and oxidizes a reduced species.
Photogenerated electrons are left in the

e

EF

EV

A wide range of luminescence effects have been reported for porous silicon; these have been reviewed by Kelly and coworkers [25]. During anodic oxidation, porous p -type silicon shows electroluminescence [174, 175], similar to that described for p -type InP in Sect. 2.1.3.3.2. In the case of porous silicon, however, a very strong light emission is observed in the visible spectral range (because of size quantization). On excitation of the bulk substrate with near-infrared light, porous n-type silicon can be photoanodically oxidized; this process is also accompanied by strong emission of visible light [176]. These results provide insight into the mechanism of anodic oxidation of the porous semiconductor.

h
1/ α 0

Fig. 20 Energy scheme for photogeneration and diffusion of electrons in a porous photoelectrode under steady state conditions. The light 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 the electrochemical potential EF
(dashed line). Transport is attenuated by multiple trapping/detrapping.

2.1 Photoelectrochemical Systems Characterization

solid network and diffuse, because of a gradient of the electrochemical potential, through the network over a considerable distance toward the metal contact, where they are collected (Fig. 20). The thickness, d, of a porous electrode and, hence, the length of the electron pathway is between
1 and 100 µm.
The transport of electrons in porous semiconductors is of interest to a wide audience, not only to electrochemists. On one hand, porous semiconducting and insulating networks can be considered as disordered systems, which show a strong resemblance to amorphous semiconductors [177, 178]. In porous semiconductors, diffusing electrons can be scattered not only by the lattice but also by the surface of the matrix and by grain boundaries. Scattering is a friction phenomenon reducing the free-electron mobility µ. In addition, porous networks have a large interfacial area. Therefore, a huge volume density of interfacial electronic states, distributed in the band gap, can be expected. The volume density of band gap states can be much larger than that of a macroscopic crystal and comparable to that of an amorphous semiconductor [177, 178]. A diffusing electron can be trapped in a state in the gap and hence, become temporarily localized. The electron is promoted back into the CB by thermal excitation (trapping/detrapping).
It is clear that electron scattering and multiple trapping/detrapping are different physical processes; scattering reduces the electron ﬂux in the system by reducing the mobility of the diffusing free electrons, whereas trapping decreases the electron ﬂux by reducing the density of free electrons.
Many recent experimental results show that electron transport through a porous semiconducting network is a slow process [179–183]. For instance, the average

time that electrons need to travel through the system before collection, that is, the transit time τtran (d) is in the millisecond to second range [179–183, 187, 188].
As a result, photogenerated electrons can be lost before collection, by transfer to the oxidized species in the solution, a process characterized by a time constant τrec . Electron back-transfer forms an important recombination process in dye-sensitized photoelectrochemical systems [16, 179–183].
2.1.5.3.2 Characterization of Electron Diffusion and Back-transfer by Light Intensity
Modulated Techniques In Sect. 2.1.2, it was shown that time-resolved methods are required to obtain information on the kinetics and dynamics of a system.
The measurement of the photocurrent response upon a small modulation of the light intensity is a very effective method to study electron dynamics in a porous system [184–186, 187, 188]. Neglecting trapping/detrapping of diffusing electrons and assuming the electrochemical potential gradient in the porous system to be independent of the spatial coordinate x (see
Fig. 20), Vanmaekelbergh and coworkers demonstrated that the optoelectrical transfer function may be written as a function of τtran en τrec [189, 190]:

˜ jPH (ω) e(1 − R) ˜ (ω)
=

1 − e−iωτtran (d) e−τtran (d)/τrec iωτtran (d) + [τtran (d)/τrec ]

(37)

Plots of the optoelectrical transfer function in the complex plane are presented in
Fig. 21. Attention is drawn to two limiting cases. If the transit time of photogenerated electrons τtran (d) through the porous network is much smaller than the electron lifetime τrec , photogenerated electrons will

97

2 Experimental Techniques
0.0

Im [ j(ω)/e Φ(ω )]

98

τrec / τd = 0.1

1/ τ rec

−0.2

τrec / τd = 1

−0.4

1/ τ d

−0.6

−0.2

0.0

0.2

τrec / τd = 100

0.4

0.6

0.8

1.0

Re [ j(ω)/e Φ(ω)]
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.
Fig. 21

reach the collecting contact without being lost; the collection efﬁciency is nearly one, and thus also the low-frequency limit of the transfer function. In this case (Fig. 21), the optoelectrical transfer function describes electron propagation. The shape of the function is determined by electron diffusion only and enters, surprisingly, the third quadrant of the complex plane at high frequencies. The average transit time through the porous network follows from the angular frequency ωm at which the imaginary part of the optoelectrical transfer function shows its ﬁrst minimum (characteristic frequency, starting from zero frequency): τtran (d) =

3 ωm (38)

It is easily inferred from Eq. (38) that the factor e−τtran (d)/τrec damps the oscillating function e−iωτtran (d) . As a result, the optoelectrical transfer function becomes increasingly semicircular if τtran (d)/τrec increases toward and above unity. When

recombination is dominant (τtran (d) τrec ), the optoelectrical transfer function reduces to
˜
τrec /τtran (d) j (ω)
=
(39)
˜ (ω) iωτrec + 1 e(1 − R)
This transfer function corresponds to a semicircle in the complex plane (Fig. 21).
The recombination lifetime follows from the characteristic frequency of the transfer function: 1/ωm = τrec ; the transit time can then be extracted from the low-frequency limit given by
∂jphoto
(1 − R)e ∂φ

=

τrec τtran (d)

(40)

Including trapping/detrapping in the model that leads to the transfer function expressed by Eq. (37) does not alter the optoelectrical transfer function itself.
However, the transit time that is obtained from ωm is increased by a factor
[1 + (EF,n )]. The value of the trapping parameter (EF,n ) (with respect to 1)

2.1 Photoelectrochemical Systems Characterization

determines to what extent multiple trapping or detrapping attenuates electron diffusion: electron transport is trap-limited if (EF,n )
1. It was found that the trapping parameter is given by the ratio of the overall volume density of states near the Fermi level and the average density of free electrons in the porous system for a given position of the electron Fermi level
[n(EF,n )]: sv,i (EF,n )

kT
=

i

n(EF,n )

(41)

In Eq. (41), i sv,i (E) denotes the volume density of band gap states per unit energy, at a given energy E . In a porous system, this quantity can be easily between
1018 and 1020 cm−3 eV−1 ; it is expected that is considerably above unity. This means that transport will be trap-limited.
In a porous photoelectrochemical system, n(EF,n ) increases linearly with increasing background light intensity . Hence, we predict that the trapping parameter and thus, the attenuation of electron transport caused by trapping will strongly decrease with increasing light intensity. This has been observed with several porous photoelectrochemical systems [184–186, 187,
188] (see following text). The fact that the transit time decreases with increasing background light intensity also shows that attenuation of electron transport is due to trapping, not to scattering. Furthermore, the experimental relationship between the transit time τtran (d) and the background light intensity can be used to map the density of state function kT i sv,i (E) in a certain energy range of the band gap.
2.1.5.3.3 Experimental Results on Electron Transport Electron transport and recombination (corresponding to electron back-transfer to the oxidized species I3 −

in the electrolyte) in nanoporous TiO2 photoelectrodes has been studied extensively [184–186, 187, 188, 191–194,
195]. It has been found that the transit time through such an electrode depends strongly on the background light intensity; several research groups found that τtran (d) ∝ −0.7 . In other words, the effective diffusion coefﬁcient Dn (EF,n ) is proportional to 0.7 . Values between 10−8 and
10−4 cm−2 s−1 were found for Dn (EF,n ) for typical light intensities between 1010 and 1015 cm−2 s−1 . These very low values and their dependence on the light intensity (hence, on the position of the electron Fermi level in the band gap) show that electron diffusion is strongly attenuated by trapping. In the framework presented earlier (trap-limited electron transport), the effective diffusion coefﬁcient reads
Dn (EF,n ) =

µ n(EF,n ) e (42)

sv,i (EF,n ) i where µ is free-electron mobility. A density-of-state function, which increases exponentially with increasing energy in the band gap has been derived from the earlier relationship. The photovoltage response upon modulation of the light intensity (with a given background light intensity) has been used to study the recombination kinetics in dye-sensitized
TiO2 solar cells [191–195]. It was found that the back-transfer of the electrons from the TiO2 network to I3 − is extremely slow.
This contrasts with the fast reduction of this species at Pt, where chemisorbed I3 − plays an essential role. A detailed study of the modulated photovoltage as a function of the background light intensity revealed that τrec ∝ 1/n(EF,n ) (this means that the rate of back-transfer is second order in the concentration of electrons) [196]. Peter

99

2 Experimental Techniques

Imaginary [ j(ω)/e Φ(ω)]

and coworkers proposed a mechanism for this multistep electron transfer process
•
in which the radical anion I2 − plays an essential role.
Electron diffusion through macroporous
GaP electrodes has also been studied.
0.0
− 0.1
− 0.2
− 0.3
− 0.4
− 0.5
− 0.6
− 0.7

Imaginary [ j(ω )/e Φ(ω)]

0.2

0.4

0.6

0.8

1.0

102

103

Real [ j(ω)/e Φ(ω)]

(a)
0.0
− 0.1
− 0.2
− 0.3
− 0.4
− 0.5
− 0.6
− 0.7

ωm

10−2

10−1

100

101

Modulation frequency
[Hz]

(b)

Electrolyte solution is 0.5M H2SO4 Fig. 22 (a) Complex-plane
200 Porosity is 25% representation of the measured

150
10 µm
20 µm
40 µm
100 µm
10 µm
20 µm
40 µm

100

50

0
300 400 500 600 700 800

(c)

Macroporous GaP is an ideal model system for several reasons: (1) the porous network is prepared by anodic etching of a single crystal; this gives a very good reproducibility in the preparation of the samples, (2) the porous network is a

ωm

− 0.2 0.0

Trapping level density: 2kBTΣS(EF)
[1015/cm3]

100

EF-EC
[meV]

optoelectrical transfer function (points) for a porous GaP electrode (40-µm thick) permeated with an aqueous
H2 SO4 solution. The background light intensity is 2 × 1014 cm−2 s−1 . (b) Plot of the imaginary part (points) as a function of the modulation frequency.
Full and dashed lines in (a) and (b) represent calculated plots: the dashed plot corresponds to Eq. (37); the full plot takes into account a Gaussian distribution around the average transit time. (c) Density of state function versus the energy in the band gap for macroporous GaP. The data are obtained from the trap-limited transit time (3/ωm ) measured with different background light intensities, (see text).

2.1 Photoelectrochemical Systems Characterization

single crystal without grain boundaries, and (3) the thickness of the porous ﬁlm and hence, the length of the electron pathway can be varied at will between 20 and 200 µm. This enables an additional check of the theoretical understanding of electron transport. The optoelectrical transfer function measured with a 40µm porous GaP photoelectrode is shown in Fig. 22 [188]. The transfer function is presented in the complex plane; the evolution of the imaginary component as a function of the modulation frequency is also shown. It is clear that the results are in agreement with the model presented earlier (Eq. 37) if a Gaussian distribution around the average transit time of the electrons is assumed. At a background light intensity of 2 × 1014 cm−2 s−1 , the transit time τtran (d) = 3/ωm is 1 s, that is, a factor 105 larger than that expected if one assumes that the mobility of the diffusing electrons is the same as in bulk GaP. Thus, electron transport is strongly attenuated by the effect of multiple trapping. The fact that τtran (d) decreases strongly with increasing background light intensity shows that trapping/detrapping in band gap states close to the electron Fermi level attenuates diffusion. A detailed analysis of the results obtained with porous GaP electrodes of different thickness led to a mapping of the density-of-state function in the upper half of the band gap (Fig. 22) [188].
References
1. A. Uhlir, Bell Systems Techn. J. 1956, 35, 333.
2. H. Gerischer, F. Beck, Z. Phys. Chem. N.F.
1957, 13, 389.
3. D. R. Turner, J. Electrochem. Soc. 1958, 105,
402.
4. A. Fujishima, K. Honda, Bull. Chem. Soc.
Jpn. 1971, 44, 1148.
5. A. B. Ellis, S. W. Kaiser, M. S. Wrighton,
J. Am. Chem. Soc. 1976, 98, 1635.

6. A. B. Ellis, J. M. Bolts, M. S. Wrighton,
J. Electrochem. Soc. 1977, 124, 1603.
7. G. Hodes, J. Manassen, D. Cahen, Nature
1976, 261, 403.
8. B. Miller, A. Heller, Nature 1976, 262, 680.
9. S. Licht, G. Hodes, R. Tenne et al., Nature
1987, 330, 148.
10. B. Wang, S. Licht, T. Soga et al., Solar Energy Mater. Sol. Cells 2000, 64, 311.
11. S. Licht, H. Tributsch, B. Wang et al.,
J. Phys. Chem. B 2000, 104, 8920.
12. S. Licht, J. Phys. Chem. 1986, 90, 1096.
13. S. Licht, D. Peramunage, Nature 1990, 345,
330.
14. S. Licht, O. Khaselev, T. Soga et al., Electrochem. Sol. State Lett. 1998, 1, 20.
15. P. H. L. Notten, J. E. A. M. van den Meerakker, J. J. Kelly, in Etching of III–V Semiconductors, An Electrochemical Approach, Elsevier Advanced Technology, Oxford, 1991.
16. B. O’Regan, M. Gr¨tzel, Nature 1991, 353, a 737.
17. L. T. Canham, Appl. Phys. Lett. 1990, 57,
1046.
18. J. J. Kelly, D. Vanmaekelbergh, Porousetched Semiconductors: Formation and
Characterization, in The Electrochemistry of Nanostructures: Preparation and Properties, Chap. 4 (Eds.: G. Hodes), Wiley-VCH,
Weinheim, 2001.
19. S. Ogawa, F. F. Ran, A. J. Bard, J. Phys.
Chem. B 1995, 99, 11182.
20. S. Ogawa, K. Hu, F. F. Ran et al., J. Phys.
Chem. B 1997, 101, 5707.
21. S. R. Morrison, in Electrochemistry at Semiconductor and Oxidized Metal Electrodes,
Plenum Press, New York, 1980.
22. J. Gobrecht, H. Gerischer, Solar Energy
Mater. 1979, 2, 131.
23. R. Memming, Electrochim. Acta 1980, 25,
77.
24. R. Memming, in Semiconductor Electrochemistry, Wiley-VHC, Weinheim, 2000.
25. J. J. Kelly, E. S. Kooij, E. A. Meulenkamp,
Electrochim. Acta 1999, 45, 561.
26. E. P. A. M. Bakkers, E. Reitsma, J. J. Kelly et al., J.Phys. Chem. B 1999, 103, 2781.
27. J. R. Macdonald in Impedance Spectroscopy,
Emphasizing Solid Materials and Systems,
John Wiley & Sons, New York, 1987.
28. R. Unbehauen in Systemtheorie 1, Allgemeine
Grundlagen, Signale und lineare Systeme im
Zeit- und Frequenzbereich R. Oldenbourg
Verlag, M¨ nchen, 1997. u 101

102

2 Experimental Techniques
29. W. G¨rtner, Phys. Rev. 1959, 116, 84. a 30. L. M. Peter, Chem. Rev. 1990, 90, 753.
31. B. Smandek, G. Chmiel, H. Gerischer, Ber.
Bunsen-Ges. Phys. Chem. 1989, 93, 1094.
32. G. Chmiel, H. Gerischer, J. Phys. Chem.
1990, 94, 1612.
33. K. H. Beckmann, R. Memming, J. Electrochem. Soc. 1969, 116, 368.
34. H. Gerischer, Z. Phys. Chem. 1960, 26, 325.
35. H. Gerischer, Z. Phys. Chem. 1961, 27, 48.
36. H. Gerischer, Semiconductor Electrochemistry, in Physical Chemistry, an Advanced
Treatise, Vol. IXA/Electrochemistry, Academic Press, New York, 1970.
37. W. P. Gomes, F. Cardon, in Progress in Surface Science (Eds.: S. G. Davison), Pergamon
Press, 1982, p. 155, Vol. 12.
38. J. O’ M. Bockris, A. K. N. Reddy in Modern
Electrochemistry, Plenum Press, New York,
1977, Vol. 2.
39. Y. V. Pleskov, Y. Y. Gurevich in Semiconductor Photoelectrochemistry, Consultants
Bureau, New York, 1986.
40. R. Degryse, W. P. Gomes, F. Cardon et al.,
J. Electrochem. Soc. 1975, 122, 711.
41. T. Pajkossy, J. Electroanal. Chem. 1994, 364,
111.
42. J. F. McCann, S. P. S. Baldwall, J. Electrochem. Soc. 1982, 129, 551.
43. Z. Hens, W. P. Gomes, Phys. Chem. Chem.
Phys. 1999, 1, 3607.
44. Z. Hens, W. P. Gomes, Phys. Chem. Chem.
Phys. 1999, 1, 3617.
45. K. Schr¨ der, R. Memming, Ber. Bunsen-Ges. o Phys. Chem. 1985, 89, 385.
46. D. Vanmaekelbergh, W. P. Gomes, Ber.
Bunsen-Ges. Phys. Chem. 1985, 89, 994.
47. J. E. A. M. van den Meerakker, J. J. Kelly,
P. H. L. Notten, J. Electrochem. Soc. 1985,
132, 638.
48. J. E. A. M. van den Meerakker, Electrochim.
Acta 1985, 30, 435.
49. D. Vanmaekelbergh, W. P. Gomes, F.
Cardon, Ber. Bunsen-Ges. Phys. Chem. 1986,
90, 431.
50. D. Lincot, J. Vedel, J. Electroanal. Chem.
1987, 220, 179.
51. D. Vanmaekelbergh, W. P. Gomes, F.
Cardon, J. Electrochem. Soc. 1987, 134, 891.
52. D. Vanmaekelbergh, R. P. ter Heide, W.
Kruijt, Ber. Bunsen-Ges. Phys. Chem. 1989,
93, 1103.
53. Z. Hens, W. P. Gomes, J. Phys. Chem. B
2000, 104, 7725.

54. B. H. Ern´ , D. Vanmaekelbergh, I. E. Vere meir, Electrochim. Acta 1993, 38, 2559.
55. D. Vanmaekelbergh, F. Cardon, J. Phys. D:
Appl. Phys. 1986, 19, 643.
56. D. Vanmaekelbergh, F. Cardon, Semicond.
Sci. Technol. 1988, 3, 124.
57. D. Vanmaekelbergh, F. Cardon, Electrochim.
Acta 1992, 37, 837.
58. Z. Hens, W. P. Gomes, J. Phys. Chem. B
1997, 101, 5814.
59. Z. Hens, W. P. Gomes, J. Electroanal. Chem.
1997, 437, 77.
60. Z. Hens, J. Phys. Chem. B 1999, 103, 122.
61. Z. Hens, W. P. Gomes, J. Phys. Chem. B
1999, 103, 130.
62. D. Vanmaekelbergh, A. R. de Wit, F. Cardon, J. Appl. Phys. 1993, 73, 5049.
63. A. A. Burk, P. B. Johnson, W. S. Hobson et al., J. Appl. Phys. 1986, 59, 1621.
64. A. B. Ellis, Luminescent Properties of Semiconductor Electrodes, in Chemistry and
Structure of Interfaces, New Laser and Optical Techniques, Chapter 6 (Eds.: R. B. Hall,
A. B. Ellis), VCH, Deerﬁeld Beach (FL),
1986.
65. B. A. Balko, G. L. Richmond, J. Phys. Chem.
1993, 97, 9002.
66. O. Kruger, Ch. Jung, Ber. Bunsen-Ges. Phys.
Chem. 1994, 98, 1022.
67. J. F. Kauffman,
C. S. Liu,
M. W. Karl,
J. Phys. Chem. 1998, 102, 6766.
68. G. N. Ryba, C. N. Kenyon, N. S. Lewis,
J. Phys. Chem. 1993, 97, 13814.
69. A. Manivannan, K. Hashimoto, A. Fujishima, J. Phys. Chem. 1992, 96, 3766.
70. A. Manivannan, K. Itoh, K. Hashimoto et al., J. Electrochem. Soc. 1990, 137, 3121.
71. J. Ouyang, F.-R. Fan, A. J. Bard, J. Electrochem. Soc. 1989, 136, 1033.
72. F. Decker, F. Prince, P. Motisuke, J. Appl.
Phys. 1985, 57, 2900.
73. B. H. Ern´ , F. Ozanam, J.-N. Chazalviel, e Phys. Rev. Lett. 1998, 80, 4337.
74. S. R. Morrison, T. Freund, J. Chem. Phys.
1967, 47, 1543.
75. W. P. Gomes, T. Freund, S. R. Morrison,
J. Electrochem. Soc. 1968, 115, 818.
76. A. Fujishima, T. Kato, E. Maekawa et al.,
Bull. Chem. Soc. Jpn. 1981, 54, 1671.
77. G. H. Schoenmakers, D. Vanmaekelbergh,
J. J. Kelly, J. Chem. Soc., Faraday Trans.
1997, 93, 1127.
78. J. R. Harbor, M. L. Hair, J. Phys. Chem.
1979, 83, 652.

2.1 Photoelectrochemical Systems Characterization
79. G. Nogami, J. H. Kennedy, J. Electrochem.
Soc. 1989, 136, 2583.
80. P. Herrasti, L. Peter, J. Electroanal. Chem.
1991, 305, 241.
81. P. Bogdanoff, N. Alonso-Vante, J. Electroanal. Chem. 1994, 379, 415.
82. R. Memming, J. Electrochem. Soc. 1969, 116,
785.
83. J. J. Kelly, B. P. Minks, N. A. M. Verhaegh et al., Electrochim. Acta 1992, 37, 909.
84. B. P. Minks, G. Oskam, D. Vanmaekelbergh et al., J. Electroanal. Chem. 1989, 273, 119.
85. B. P. Minks, D. Vanmaekelbergh, J. J. Kelly,
J. Electroanal. Chem. 1989, 273, 133.
86. R. Peat, L. M. Peter, Electrochim. Acta 1986,
31, 731.
87. R. Peat, L. M. Peter, J. Electroanal. Chem.
1986, 209, 307.
88. W. P. Gomes, H. H. Goossens, Electrochemistry of III–V Compound Semiconductors: Dissolution Kinetics and Etching in
Advances in Electrochemical Science and Engineering, (Eds.: H. Gerischer, C. W. Tobias),
VCH, Weinheim, 1994, Vol. 3.
89. D. Vanmaekelbergh, J. J. Kelly, J. Electrochem. Soc. 1989, 136, 108.
90. W. P. Gomes, S. Lingier, D. Vanmaekelbergh, J. Electroanal. Chem. 1989, 269, 237.
91. M. Matsumura, S. R. Morrison, J. Electroanal. Chem. 1983, 147, 159.
92. L. M.Peter, A. M. Borazio, H. J. Lewerenz et al., J. Electroanal. Chem. 1990, 290, 229.
93. G. H. Schoenmakers, R. Waagenaar, J. J.
Kelly, J. Electrochem. Soc. 1995, 142, L60.
94. S. Cattarin, L. M. Peter, D. J. Riley, J. Phys.
Chem. B 1997, 101, 4071.
95. X. Xia, C. M. A. Ashruf, P. J. French et al.,
J. Phys. Chem. B 2001, 105, 5722.
96. B. H. Ern´ , D. Vanmaekelbergh, J. Elece trochem. Soc. 1997, 144, 3385.
97. D. Vanmaekelbergh, P. C. Searson, J. Electrochem. Soc. 1994, 141, 697.
98. B. H. Ern´ , F. Ozanam, M. Stchakovsky e et al., J. Phys. Chem. B 2000, 104, 5961.
99. B. H. Ern´ , F. Ozanam, M. Stchakovsky e et al., J. Phys. Chem. B 2000, 104, 5974.
100. P. Allongue, H. Brune, H. Gerischer, Surf.
Sci. 1992, 275, 414.
101. P. Allongue, V. Costa Gieling, H. Gerischer,
J. Electrochem. Soc. 1993, 140, 1009.
102. P. Allongue, Phys. Rev. Lett. 1996, 77, 1986.
103. W. West, B. H. Carrol in The Theory of
Photographic Processes, (Eds.: C. E. K. Mees,
T. H. James), McMillan, New York, 1966.

104. R. W. Berriman, P. B. Gilman, Photogr. Sci.
Eng. 1973, 17, 235.
105. J. J. Kelly, D. Vanmaekelbergh, Electrochim.
Acta 1998, 43, 2773.
106. J. W. Gerritsen, E. J. G. Boon, G. Janssens et al., Appl. Phys. A 1998, 66, S79.
107 M. Kawasaki, H. Ishii, J. Imaging Sci. Technol. 1995, 39, 210.
108. T. Hannappel, B. Burfeindt, W. Storck et al., J. Phys. Chem. B 1997, 101, 6799.
109. Y. Tachibana, J. E. Moser, M. Graetzel et al., J. Phys. Chem. 1996, 100, 20056.
110. A. Henglein, Top. Curr. Chem. 1988, 143,
115.
111. H. Weller, Adv. Mater. 1993, 5, 88.
112. A. P. Alivisatos, J. Phys. Chem. 1996, 100,
13226.
113. A. N. Shipway, E. Katz, I. Willner, Chem.
Phys. Chem. 2000, 1, 18.
114. B. Alperson, H. Demange, I. Rubinstein et al., J. Phys. Chem. B 1999, 103, 4943.
115. R. L. Smith, S. D. Collins, J. Appl. Phys.
1992, 71, R1.
116. V. L. Colvin, A. N. Goldstein, A. P.
Alivisatos, J. Am. Chem. Soc. 1992, 114,
5221.
117. M. Miyake, H. Matsumoto, M. Nishizawa et al., Langmuir 1997, 13, 742.
118. T. Nakanishi, B. Ohtani, K. Uosaki, J. Phys.
Chem. B 1998, 102, 1571.
119. K. Hu, M. Brust, A. J. Bard, Chem. Mater.
1998, 10, 1160.
120. S. Drouard, S. G. Hickey, J. D. Riley, Chem.
Commun. 1999, 1, 67.
121. E. P. A. M. Bakkers in Charge transfer between semiconductor nanocrystals and a metal,
Ph. D. Thesis, University of Utrecht, 2000.
122. E. P. A. M. Bakkers, J. J. Kelly, D. Vanmaekelbergh, J. Electroanal. Chem. 2000, 482,
48.
123. E. P. A. M. Bakkers, E. Reitsma, J. J. Kelly,
D. Vanmaekelbergh, J. Phys. Chem. B 1999,
103, 2781.
124. S. G. Hickey, D. J. Riley, Electrochim. Acta
2000, 45, 3277.
125. W. J. Albery, G. T. Brown, J. R. Darwent et al., J. Chem. Soc. Faraday Trans. 1 1985,
81, 1999.
126. E. P. A. M. Bakkers, A. W. Marsman, L. W.
Jenneskens et al., Angew. Chem. Int. Ed.
2000, 39, 2297.
127. E. P. A. M. Bakkers, A. L. Roest, A. W.
Marsman et al., J. Phys. Chem. B 2000, 104,
7266.

103

104

2 Experimental Techniques
128. S. G. Hickey, J. D. Riley, E. Tull, J. Phys.
Chem. B 2000, 104, 7623.
129. M. R. Hoffmann, S. T. Martin, W. Choi et al., Chem. Rev. 1995, 95, 69.
130. G. Hodes, J. Electrochem. Soc. 1992, 139,
3136.
131. R. Gr¨ nwald, H. Tributsch, J. Phys. Chem. u B 1997, 101, 2564.
132. S. Y. Huang, J. Phys. Chem. B 1997, 101,
2576.
133. D. Liu, R. W. Fessenden, G. L. Hug et al.,
J. Phys. Chem. B 1997, 101, 2583.
134. R. Argazzi, C. A. Bignozzi, T. A. Heimer et al., J. Phys. Chem. B 1997, 101, 2591.
135. H. Rensmo, K. Keis, H. Lindstroem et al.,
J. Phys. Chem. B 1997, 101, 2598.
136. A. Wahl, J. Augustynski, J. Phys. Chem. B
1998, 102, 7820.
137. W. Kubo, K. Murakoshi, T. Kitamura et al.,
Chem. Lett. 1998, 12, 1241.
138. B. H. Ern´ , D. Vanmaekelbergh, J. J. Kelly, e Adv. Mater. 1995, 7, 739.
139. B. H. Ern´ , D. Vanmaekelbergh, J. J. Kelly, e J. Electrochem. Soc. 1996, 143, 305.
140. G. Oskam, A. Natarajan, P. C. Searson,
F. M. Ross, Appl. Surf. Sci. 1997, 119, 160.
141. J. van de Lagemaat, M. Plakman, D.
Vanmaekelbergh et al., Appl. Phys. Lett.
1996, 69, 2801.
142. T. Sugira, T. Yoshida, H. Minoura, Electrochem. Solid-State Lett. 1998, 1, 175.
143. A. O. Konstantinov, C. I. Harris, E. Janz´ n, e Appl. Phys. Lett. 1994, 65, 2699.
144. A. O. Konstantinov, A. Henry, C. I. Harris et al., Appl. Phys. Lett. 1995, 66, 2250.
145. C. Zener, Proc. R. Soc. London 1934, 145,
523.
146. F. J. P. Schuurmans, D. Vanmaekelbergh,
J. Van de Lagemaat et al., Science 1999, 284,
141.
147. F. J. P. Schuurmans, M. Megens, D.
Vanmaekelbergh et al., Phys. Rev. Lett. 1999,
83, 2183.
148. L. M. Peter, D. J. Riley, R. I. Wielgosz, Appl.
Phys. Lett. 1995, 66, 2355.
149. L. M. Peter, R. I. Wielgosz, Appl. Phys. Lett.
1996, 69, 806.
150. W. H. Lubberhuizen, D. Vanmaekelbergh,
E. Van Faassen, Journal of Porous Materials
2000, 7, 147.
151. B. Enright, B. G. Redmond, D. Fitzmaurice,
J. Phys. Chem. 1994, 98, 6195.
152. M. Gratzel, Nature 2001, 409, 575.

153. D. Cummins, G. Boschloo, M. Ryan et al.,
J. Phys. Chem. B 2000, 104, 11449.
154. P. Hoyer, R. Eichberger, H. Weller, Ber.
Bunsen-Ges. Phys. Chem. 1993, 97, 630.
155. P. Hoyer, H. Weller, J. Phys. Chem. 1995,
99, 14096.
156. E. A. Meulenkamp, J. Phys. Chem. B 1999,
103, 7831.
157. D. L. Klein, P. L. McEuen, J. E. B. Katari et al., Appl. Phys. Lett. 1996, 68, 2574.
158. D. L. Klein, R. Roth, A. K. L. Lim et al., Nature 1997, 389, 699.
159. U. Banin, Y. Cao, D. Katz et al., Nature
1999, 400, 542.
160. O. Millo, D. Katz, Y. Cao et al., Phys. Rev. B
2000, 61, 16773.
161. E. P. A. M. Bakkers, D. Vanmaekelbergh,
Phys. Rev. B 2000, 62, R7743.
162. J. D. Joannopoulos, P. R. Villeneuve, S. Fan,
Nature 1997, 386, 143.
163. U. Gr¨ ning, V. Lehmann, C. M. Engelhardt, u Appl. Phys. Lett. 1995, 66, 3254.
164. J. E. G. J. Wijnhoven, W. L. Vos, Science
1998, 281, 802.
165. Y. A. Vlasov, N. Yao, D. J. Norris, Adv.
Mater. 1999, 11, 165.
166. D. Vanmaekelbergh, L. van Pieterson, Phys.
Rev. Lett. 1998, 80, 82.
167. D. Vanmaekelbergh, M. A. Hamstra, L. van
Pieterson, J. Phys. Chem. B 1998, 102, 7997.
168. J. J. M. Halls, C. A. Walsh, N. C. Greenham et al., Nature 1995, 376, 498.
169. G. Yu, J. Gao, J. C. Hummelen et al., Science 1995, 270, 1789.
170. U. Bach, D. Lupo, P. Comte et al., Nature
1998, 395, 583.
171. A. G. Cullis, L. T. Canham, P. D. J. Calcott,
J. Appl. Phys. 1997, 82, 909.
172. E. A. Meulenkamp, L. M. Peter, D. J. Riley et al., J. Electroanal. Chem. 1995, 392, 97.
173. L. M. Peter, D. J. Riley, R. I. Wielgosz et al.,
Thin Solid Films 1996, 276, 123.
174. A. Bsiesy, F. Gaspard, R. Herino et al.,
J. Electrochem. Soc. 1991, 138, 3450.
175. M.Ligeon, F. Muller, R. Herino et al.,
J. Appl. Phys. 1993, 74, 1265.
176. E. S. Kooij, A. R. Rama, J. J. Kelly, Surf. Sci.
1997, 370, 125.
177. R. A. Street, Cambridge Solid State Science
Series, in Hydrogenated Amorphous Silicon,
(Eds.: R. W. Cahn, E. A. Davis, I. M. Ward),
1991.
178. H. Scher, M. F. Shlesinger, J. T. Bendler,
Phys. Today 1991, 44, 26.

2.1 Photoelectrochemical Systems Characterization
179. L. M. Peter, D. Vanmaekelbergh in Advances in Electrochemical Science and Engineering,
(Eds.: R. C. Alkire, D. M. Kolb), Wiley VCH,
Weinheim, 1999, Vol. 6.
180. K. Schwarzburg, F. Willig, Appl. Phys. Lett.
1991, 58, 2520.
181. P. Hoyer, H. Weller, J. Phys. Chem. 1995,
99, 14096.
182. A. Solbrand, H. Lindstroem, H. Rensmo et al., J. Phys. Chem. B 1997, 101, 2514.
183. F. Cao, G. Oskam, G. J. Meyer et al., J. Phys.
Chem. 1996, 100, 17021.
184. P. E. de Jongh, D. Vanmaekelbergh, Phys.
Rev. Lett. 1996, 77, 3427.
185. L. Dloczik, O. Ileperuma, I. Lauermann et al., J. Phys. Chem. B 1997, 101, 10281.
186. P. E. de Jongh, D. Vanmaekelbergh, J. Phys.
Chem. B 1997, 101, 2716.
187. D. Vanmaekelbergh, F. Iranzo Mar´n, J. van ı de Lagemaat, Ber. Bunsen-Ges. Phys. Chem.
1996, 100, 616.

188. A. L. Roest, D. Vanmaekelbergh, Phys. Rev.
B 2000, 62, 16926.
189. D. Vanmaekelbergh, P. E. de Jongh, Phys.
Rev. B 2000, 61, 4699.
190. P. E. de Jongh in Photoelectrochemistry of nanoporous semiconductor electrodes, Ph.D.
Thesis,. University of Utrecht, 1999.
191. A. C. Fisher, L. M. Peter, E. A. Ponomarev et al., J. Phys. Chem. B 2000, 104, 949.
192. J. van de Lagemaat, N. G. Park, A. J. Frank,
J. Phys. Chem. B 2000, 104, 2044.
193. N. Kopidakis, E. A. Schiff, N. G. Park et al.,
J. Phys. Chem. B 2000, 104, 3930.
194. J. van de Lagemaat, A. Frank., J. Phys. Chem.
B 2000, 104, 4292.
195. N. W. Duffy, L. M. Peter, R. M. G. Rajapakse et al., J. Phys. Chem. B 2000, 104,
8916.
196. G. Schlichth¨ rl, S. Y. Huang, J. Sprague o et al., J. Phys. Chem. B 1997, 101, 8141.

105

106

2 Experimental Techniques

2.2

Deposition of (Multiple Junction)
Semiconductor Surfaces
Tetsuo Soga
Nagoya Institute of Technology, Nagoya,
Japan
2.2.1

Introduction

A limited fraction of incident solar photons have sufﬁcient (greater than band gap) energy to initiate charge excitation and separation within a semiconductor.
Wide band gap semiconductor solar cells are capable of generating a high photovoltage but have a low limiting photocurrent caused by the low fraction of short wavelength light in the solar spectrum. Smaller band gap cells can utilize a larger fraction of the incident photons but generate lower photovoltage. Multijunction devices, also referred to as tandem, multiple window, split spectrum, and cascade solar cells, can overcome this limitation [1]. A variety of multiple semiconductor layers have been fabricated and explored for photovoltaic applications. These include GaInP on GaAs, GaAs on Si, GaAs on GaSb,
InP on GaInAs, GaAs on GaInAsP, etc [2].
Through a detailed speciﬁc example of one of these important multiple semiconductor systems consisting of various GaAs layers on various Si layers, this paper describes preparation and optimization of multijunction systems that are capable of efﬁcient photon absorption, charge separation, and charge transfer.
This section reviews the growth process of III–V compound semiconductor on Si substrate grown by metalorganic chemical vapor deposition (MOCVD). The nucleation of GaP and GaAs on Si substrate, the dislocation generation mechanism, antiphase domain (APD) structure of GaP on

Si substrate, the method to reduce dislocation density, the method to passivate the defects in GaAs on Si, and the application to photovoltaic solar cell are presented. Although highly mismatched systems such as GaAs-on-Si, AlGaAs-on-Si, and GaN-onSi are interesting for device applications, the study on the crystal growth of GaP on Si substrate is interesting for the fundamental understanding of the III–V compounds on Si substrate. It is expected that the understanding of the growth mechanism will lead us to achieve the growth of lowdislocation-density III–V compounds on
Si substrate.
The deposition of III–V compounds has recently been actively performed. Especially, the crystal growth on Si substrate has attracted attention since high-quality
GaAs layers were successfully grown on Si substrates in 1984 [3–7]. Although various devices such as lasers [8] and solar cells [9] have been fabricated on Si substrate, the device characteristics are not satisfactory because the existence of a high density of the threading dislocations in the epitaxial layer. The reduction of dislocation density is an important issue to obtain a high-performance compound semiconductor device on an Si substrate. The dislocations are generated due to lattice mismatch, thermal expansion mismatch, the crystal structure difference, the generation of APD, surface contamination, and so on.
The threading dislocations in the epitaxial layer on Si substrate are classiﬁed into (1) dislocations that originate from the dislocations in the Si substrate,
(2) dislocations generated by the coalescence of the islands at the initial stage,
(3) dislocations generated by the lattice mismatch, and (4) dislocations generated by the thermal stress during the cooling stage from the growth temperature.

2.2 Deposition of (Multiple Junction) Semiconductor Surfaces

Among these four types of dislocations, we do not need to take into account item (1) because the dislocation density of Si is generally very low. In order to reduce the dislocation density according to items (2), (3), and (4), many efforts have been made. The efforts to change the initial growth mode from threedimensional (3D) to two-dimensional (2D) have been made to avoid the dislocations generated by the coalescence of the islands at the initial stage of the growth [10].
Although many methods using strained layer superlattice (SLS) buffer layer [3,
11, 12], rapid thermal annealing [13], thermal cycle annealing (TCA) [14, 15], and so on have been adopted to reduce the dislocation density of item (3), the dislocation density is still on the order of
106 cm−2 . The low temperature growth has been investigated to decrease the generation of the dislocation by the thermal stress [16, 17]. It is expected that the number of dislocations generated by the thermal stress is reduced when the growth temperature is low. Although the dislocation density of GaAs-on-Si on the order of 104 cm−2 has been obtained at the growth temperature of 350 ◦ C, thermal stability is the problem.

diethylzinc (DEZ), AsH3 and H2 Se, respectively. The Si substrate orientation was
(001) 2–4◦ tilted toward [110] direction. Si substrate was rinsed in organic solvents, followed by the repetition of the oxidation by H2 SO4 : H2 O = 4 : 1 and the removal of the oxides by 25% HF solution. After loading the substrate into the reactor, the substrate was heated at 1000 ◦ C for 10 min at hydrogen atmosphere. The V/III ratio was varied from 100 to 6400 with changing the PH3 ﬂow rate, keeping the TMG ﬂow rate constant for the growth of GaP on Si substrate. The gas pressure was varied from 76 to 380 torr. The epitaxial layer thickness was varied from 20 nm to
3.7 µm. The growth temperature was kept constant at 900 ◦ C. 3D growth was not observed under these growth conditions for the case of GaP growth on GaP substrate.
The samples were examined using Nomarski optical microscopy, cross-sectional transmission electron microscopy (TEM), and so on. GaAs was grown on Si substrate by the two-step growth method with
10-nm-thick GaAs buffer layer grown at
400 ◦ C.
2.2.3

Deposition of Gallium Phosphide on Silicon
Substrate

2.2.2

MOCVD

The epitaxial growth was performed using low-pressure or atmospheric pressure
MOCVD. The former consists of lampheated and the latter consists of rf-heated horizontal reactor with load lock chamber.
The substrate was put on the SiC-coated carbon susceptor and the temperature was controlled by the thermocouple inserted into the susceptor. Source gases for Al,
Ga, Zn, As and Se are trimethylalluminum (TMA), trimethylgallium (TMG),

Nucleation of GaP on Si Substrate
In general, the growth mode is divided into three categories, namely, 2D type,
Volmer-Weber (3D) type, and StranskiKrastanov (2D + 3D) type. In the case of the 2D mode, the dislocations are generated when the layer thickness exceeds the critical thickness [18]. On the other hand, the dislocations and stacking faults are generated at the coalescence of the islands formed at the initial stage of the growth in the case of the 3D mode [19]. Therefore, the defect density will be signiﬁcantly
2.2.3.1

107

2 Experimental Techniques
Material parameters of III–V compound semiconductors and Si

Tab. 1

GaP
˚
Lattice constant (A)
Linear thermal expansion coefﬁcient (×10−6 K−1 )
Crystal structure

GaAs

InP

Si

5.45
5.9

5.65
6.8

5.87
4.6

5.43
2.6

ZB

ZB

ZB

Diamond

Note: ZB: Zinc Blend.

reduced if a 2D growth can be realized from the beginning of the growth.
The 3D growth mode has origins not only in the basic material property differences between the epitaxial layer and the substrate (e.g. lattice constant mismatch, polar or nonpolar effect, etc.) but also in the growth conditions.
The material properties for various
III–V compounds and Si are shown in
Table 1. As shown in this table, the lattice constant of GaP is closest to that of Si.
Therefore, the effect of lattice mismatch

3200
Layer-type
1600

V/III ratio

108

800
Mixture-type
400
Island-type

200
100

0

76

152

Gas pressure
[Torr]

380

on the growth mode is expected to be minimized and other effects such as the surface migration effect, surface contamination, polar or nonpolar structure are emphasized. Figure 1 shows the growth mode of GaPon-Si for various gas pressure and the V/III ratio [20]. All the surface morphologies of samples were classiﬁed into three-types, namely, island-type, mixture-type (mixture of island and layer-type), and layer-type.
It is indicated that the growth mode of
GaP changes from island-type to layertype 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 the gas pressure of 76 torr [8]. Island-type growth is clearly demonstrated, and it is observed that the islands are not connected by a GaP layer between themselves at the edge of the island. This means that the growth mode of GaP on Si substrate under these growth conditions is not StranskiKrastanov type but Volmer-Weber type.
Faceting was observed at the boundary either on (111) or (211) type planes.
The island formation is interesting because these islands are not formed in
Growth mode of GaP on Si substrate for various V/III ratio and gas pressure. Fig. 1

2.2 Deposition of (Multiple Junction) Semiconductor Surfaces

the case of homoepitaxy; it is particular for the heteroepitaxy. In the case of the homoepitaxial growth, the source gases are usually incorporated into the step edge or terrace of the misoriented substrate. However, the island spacing is several orders of magnitude larger than that of the average step distance of the misoriented substrate. Therefore, the nucleation site is not governed by the substrate steps. Furthermore, the residual oxide or impurity is not the nucleation site because the GaP island density changes drastically with the growth conditions [21].
In the nucleation process for GaP on
Si substrate, the diffusion of the growth species through the boundary layer, the surface migration, and the nucleation at the nucleation sites should be taken into account. Comparing the heteroepitaxy of
GaP-on-Si and the homoepitaxy, there should be no difference in the diffusion process if the growth conditions are the same. Therefore, it is deduced that the difference between the homoepitaxy and the heteroepitaxy on Si is caused by the difference in the migration on the Si substrate. The migration length of the migrating species on Si is considered to be longer than for homoepitaxy. This is due to the weak interaction between Si and Ga or P atoms. An example of strong atomic interaction is the growth of a III–V compound semiconductor containing Al on Si substrate. It has been reported that AlGaP [22], AlGaAs [23], and AlAs [24] layers grown on Si substrate are ﬂat from the beginning of growth.
Before considering the migrating species for GaP on Si, the surface migration of
Ga is discussed. Usually, the migration species for the deposition of Ga ﬁlms on
Si substrate is the cluster of Ga atoms.
TMG is probably perfectly decomposed to Ga and a metal-radical at 900 ◦ C [25].

Therefore, in the case of the deposition of Ga on Si, the migrating species are supposed to be Gax -type clusters. It is expected that the molecular mass of a cluster increases during the surface migration. Because the island density increases gradually with increasing V/III ratio, the migration species would be a Gax Py -type cluster [21]. When the PH3 ﬂow rate is increased, the number of decomposed P atoms is increased, and a high density of
Gax Py -type cluster with large mass is easily formed. Clusters with large molecular mass are expected to migrate more slowly than those with smaller masses. When the cluster size exceeds the critical size, the clusters are deposited on the Si surface and the islands are formed.
Moreover, when the concentration of P atoms on the Si surface is high, P atoms absorbed on the Si surface are increased.
It results in the formation of the ﬂat layer because the P atoms absorbed on
Si capture the migrating species.
To explain the gas pressure dependence of the growth mode, other factors in the reactor must be considered. A possible factor affected by the gas pressure is the ﬂow velocity in the reactor. TMG is almost completely decomposed at the growth temperature [25]. In contrast, the decomposition of PH3 varies with the ﬂow velocity because the decomposition rate for PH3 is not as fast as TMG [26].
These results are qualitatively explained as follows. It is evident that the ﬂow velocity is increased when the pressure is decreased. The pyrolysis of PH3 takes place in the heated region in the reactor.
Therefore, the decrease of pressure makes the PH3 decomposition difﬁcult because the resident time for the source gases in the heated region becomes short, resulting in the atomic arrangement as shown in Fig. 2(a). So the higher V/III ratio is

109

110

2 Experimental Techniques
Ga
P
GaP

Si

Si
(a)

GaP

Si
(b)

Schematic atomic arrangement of 3D GaP on Si (a) and 2D
GaP on Si (b).

Fig. 2

necessary for lower growth pressure to produce a layer-type growth as shown in
Fig. 2(b).
In the growth of GaP, a simple reaction involving PH3 is considered to be the decomposition of PH3 and the production of phosphorus molecules, that is,
PH3 = P + 3H

(1)

Using simple reaction kinetics and solving the differential equations, that is, d[PH3 ]
(2)
= a [PH3 ] dt the partial pressures of PH3 in these cases are given by x pPH3 = pPH30 exp −a
(3)
v where pPH3 is the partial pressure of PH3 at the distance x from the entrance of the heated zone in the ﬂow direction with the ﬂow velocity v , pPH30 is the initial partial pressure of PH3 or the supplied partial pressure into the reactor, and a is a constant including the reaction rate constant, the cross-sectional area of the reactor and other parameters. The ﬂow

velocity v is proportional to the total ﬂow rate (F ) and inverse of the growth pressure
Pg , that is,
F
(4) v=b Pg and Eq. (3) is changed into pPH3 = pPH30 exp −c

Pg
F

(5)

where c is a modiﬁed constant including a , b, and x . On the other hand, using these notations, V/III ratio is given by
[V/III] =

pPH30 pTMGa (6)

The ratio of the phosphorus concentration([P]) to the gallium concentration
([Ga]), which is available to the growth on the substrate surface is expressed as,
[P/Ga] = [P]/[Ga]

(7)

where [P] and [Ga] are proportional to the concentration of the decomposed PH3 and TMGa, respectively. [Ga] is supposed to be constant on the growing surface for various growth pressures because the

2.2 Deposition of (Multiple Junction) Semiconductor Surfaces

TMGa is almost decomposed at the growth temperature. Thus, substituting the partial pressures and the relation for the V/III ratio into Eq. (6), the above ratio is given by
[P/Ga] = a

(pPH30 − pPH3 ) pTMGa × 1 − exp −c

= a [V/III]

Pg
F

(8)

where a is the proportionality constant.
Therefore, the initial growth mode of GaP, which is governed by [P]/[Ga] ratio is expressed using the V/III ratio, Pg and
F . From these results, it can be said that at lower growth pressure, a higher V/III ratio is required for the layer growth.
Generation of Dislocation
In the heteroepitaxial growth, the generation of misﬁt dislocation and the stress relaxation are related to each other. The misﬁt dislocation generation and the stress 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 the islands are not generated, the observed dislocations are generated after the layer thickness exceeds the critical thickness.
From the cross-sectional TEM micrograph, it can be seen that the GaP surface is very ﬂat, and defects such as dislocations or structural defects are not observed 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 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.
2.2.3.2

Atomic arrangement for GaP on
Si with A-type and B-type dislocation.

Fig. 3

In the TEM picture with the thickness of 3.7 µm, two kinds of dislocations with an extra-half plane in the Si substrate
(A-type) and the GaP layer (B-type) are observed [28]. The schematic models of atomic arrangements for these structures are shown in Fig. 3. In general, the dislocation generation should take place at random in the isotropic crystal [29], that is, four kinds of A-type dislocations with
Burgers’ vector of 1/2[1 0 1], 1/2[0−11],
1/2[0 1 1] and 1/2[−101] can be generated.
However, one salient feature came to be seen; instead of these four directional
A-type dislocation, only two directional Atype dislocations with Burgers vector of
1/2[1 0 1] or 1/2[0−11] are observed. A possible interpretation for this generation of A-type dislocation with only two-kinds of Burgers’ vectors instead of the four previously reported types of dislocations is the difference of the situation of the site, that is, dislocations are created at the step

GaP

Si
Si
P
Ga

A-type

GaP

Si

B-type

111

2 Experimental Techniques

Compressive

edge of the misoriented Si substrate rather than randomly isotropic generation. The direction of Burgers’ vector for the B-type dislocations is the reverse of that of the
A-type dislocation.
Assuming that the strain is completely accommodated at the growth temperature by 60◦ dislocations, the spacing between dislocations corresponds to 110 nm. However, the spacing between A-type dislocations is much smaller than the calculated value. This difference might be due to the uniformity of the Si substrate steps.
If the dislocations are generated by the lattice mismatch, an extra-half plane should occur in the Si substrate, which has a small lattice constant [29]. Because both the room temperature lattice constant and the thermal expansion coefﬁcient of
GaP are larger than those of Si, GaP should have the larger lattice constant than Si at the growth temperature. Hence, the lattice mismatch relaxation at the growth temperature is responsible for the generation of the A-type dislocation. The
B-type dislocations cannot be explained merely on the grounds of lattice mismatch.
Because no dislocation is generated at the initial stage of the growth, the B-type dislocations should be generated during the growth or cooling down process. If the

lattice strain of GaP is relaxed completely at the growth temperature by introducing
A-type misﬁt dislocations, the tensile stress is produced in the GaP layer during the cooling process because of the difference of the thermal expansion coefﬁcients of GaP and Si. The thermal expansion coefﬁcient of GaP is about 2.5 times larger than that of Si. In order to relax the tensile stress in the GaP layer, the dislocations with the extra-half plane in the GaP layer should be introduced. Accordingly, it is proved that the dislocations with the extra-half plane in the Si substrate are formed by the lattice mismatch and that those in the
GaP layer are formed during the cooling process to relax the thermal stress. This experiment supports the report that proved the generation of threading dislocations in GaP on Si during the cooling down process [30].
The stress applied to the GaP layer as a function of the thickness measured by
X-ray diffraction is shown in Fig. 4 [20].
The dotted line shows the stress value for the thermal stress calculated by the bimetal model. A thin GaP layer has a compressive stress, and the stress changes to tensile with increasing thickness. Considering the thermal stress between the growth temperature and room temperature, the GaP layer

1
0
1

Tensile

Stress
[× 109 dyn cm−2]

112

2
3
0

1

2

Thickness
[µm]

3

4

Stress of GaP layer on Si substrate as a function of the thickness. Fig. 4

2.2 Deposition of (Multiple Junction) Semiconductor Surfaces
APD structure and single domain structure at different step height. Fig. 5

Si
P
Ga

[001]

[110]

must have compressive stress at growth temperature when the layer thickness is thinner than the critical thickness. Because the strain energy, which is caused by the misﬁt, increases with the thickness, the misﬁt dislocations are generated with increasing 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 temperature. The strain is smaller than the calculated thermal stress because the thermal stress is partly relaxed by the generation of misﬁt dislocations with the extra half plane in the GaP layer.
Although the dislocation is not generated at the initial stage of the growth, the dislocations with the extra-half plane in the
Si substrate are generated at the growth temperature to relax the lattice mismatch.
During the cooling down process, the thermal stress is relaxed by generating the misﬁt dislocation with the extra-half plane in the GaP layer. Below the dislocation frozen temperature, the thermal expansion mismatch produces the tensile stress in the epitaxial layer without generating new dislocations, resulting in the large tensile stress at room temperature.
Annihilation of APD Structure
The problem of the APD occurs at the Si surface step as a result of the polar or nonpolar structure. As shown in Fig. 5, the APDs are generated when the Si has a single (or odd) atomic step, whereas the single domains are formed when the Si surface is double (or even) atomic steps.
The typical APD is shown in Fig. 6.
These were taken under dark ﬁeld
2.2.3.3

−
[110]

GaP

Si
Antiphase domain

GaP

Si
Single domain

¯ conditions using (002) (a) and (002) (b) reﬂections. In these ﬁgures, the contrasts of the domain region and matrix region are inverted by changing the reﬂection vector
¯
g from 002 to 002. The amplitudes of 002
¯
and 002 reﬂections have been calculated to unequal for most thicknesses in the case of zincblende structure [31, 32]. Therefore, it is proved that the antiphase boundary is normal to the (001) plane near the Si substrate. In most cases, the APD is annihilated on changing the orientation of the boundary from the (001) normal to higher index planes so as to minimize the total energy. The mechanisms for generation and annihilation of APDs are discussed. The
Si surface steps are usually composed of single and double atomic steps. The initial growth mode of GaP on Si is 2D as it is grown under a high V/III ratio. In the initial stage of growth, the Si substrate is covered with P under a high PH3 rate. Therefore, APD is introduced at the single atomic step position. In the case of

113

114

2 Experimental Techniques
Dark ﬁeld TEM micrograph of APD annihilated during the growth. The reﬂection vector is (002) (a) and (00-2) (b).

Fig. 6
002
GaP

Si

(a)
−
002
GaP

Si

(b)

0.1 µm

the growth on only (001) Si, the spacing between the steps is assumed to be large compared with the misoriented substrate.
Therefore, the size of the APD is large.
On the other hand, the size of the APD is small in the case of the misoriented substrate. Calculation shows that the antiphase boundary for the (211) and (110) antiphase boundaries are energetically more likely to form than those for the (111) or (100) planes [33]. Therefore, the appearance of the (110) antiphase boundary is in good agreement with the calculations. The total energy is increased with increasing thickness. In order to reduce the total energy, the antiphase boundary changes its orientation to the low-energy index plane. The higher index plane of the APD is estimated to be the (211) plane from the angle [31].
This result also supports the calculation that the energy of the antiphase boundary for (211) is smaller than that of (110).

If the size of APD introduced in the initial stage is small, the APD is 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 be annihilated. Therefore, APD remains at the surface. The observation that the
APD is observed in GaP on 2◦ off (001)
Si indicates the existence of single atomic steps. However, the density of single steps is lower than that of the exactly (001) Si substrate. On the other hand, all the steps 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 these samples.
2.2.4

Deposition of Gallium Arsenide on Silicon
Substrate
Nucleation
For lattice mismatched III–V semiconductors on Si, two kinds of misﬁt dislocations are observed; one is the pure-edge Lomer misﬁt dislocation, whose Burgers vector is parallel to the interface (type-I dislocation), and the other is the misﬁt dislocation, whose Burgers vector is 60◦ from the dislocation line (60◦ dislocation or typeII dislocation) [34]. Schematic illustrations of type-I and type-II dislocations are shown in Fig. 7(a) and 7(b), respectively.
A periodic array of misﬁt dislocations with an average spacing of about 8.1 nm is observed and the ratio of number of type-I to type-II dislocations is about 3 : 1 for GaAs on Si substrate. This means that the lattice mismatch is completely relaxed by the misﬁt dislocations. The majority of
2.2.4.1

2.2 Deposition of (Multiple Junction) Semiconductor Surfaces
Schematic illustration of type-I dislocation (a) and type-II dislocation (b).

Fig. 7

the dislocations are type-I dislocation, and this is also the case for materials grown by molecular beam epitaxy [35].
On the other hand, in the case of GaP on Si, most dislocations are also typeII dislocations. The HRTEM micrographs show that most dislocations in GaAs on
GaP substrates are type II.
This section discusses the generation mechanisms for type-I and type-II dislocations. Although the lattice mismatch for GaAs/Si and GaAs/GaP are almost the same, the types of dislocations are different. Therefore, the type of dislocation cannot be explained alone by lattice mismatch. Furthermore, the thermal stress in the epitaxial layer cannot explain the difference of the dislocation type because the thermal stress for GaAs/Si and GaP/Si is almost the same although the types of dislocations are different.
Some mechanisms have been proposed to explain the generation of type-I and type-II dislocations. The reported mechanisms for the generation of type-II dislocation are (1) the bending of the threading dislocation in the substrate parallel to the interface [36] and (2) the glide of dislocation from the surface, which forms a half-loop [29]. The reported mechanisms for the generation of type-I dislocation are (1) the reaction of two type-II dislocations [37] and (2) the dislocation climb of the pure edge dislocation from the surface [29].
Misﬁt dislocation generation can be explained by the earlier-mentioned mechanisms when the growth mode is 2D and the epitaxial layer is ﬂat. On the other hand, no dislocation generation mechanism has been reported when 3D islands are formed at the initial stage of growth and the lattice

GaAs

Si
(a)

Type I

Si
As
Ga

GaAs

Si

(b)

Type II

mismatch is large. Therefore, another dislocation generation mechanism should be considered because the dislocation generation is modiﬁed by island formation. For example, misﬁt dislocations are generated in GaAs/Si heterostructures at the beginning of growth and before the coverage of Si with a GaAs layer has been completed [38, 39]. Therefore, it is deduced that the misﬁt dislocation generation is greatly affected by the initial growth mode.
The type of dominant dislocation, the lattice mismatch and the growth mode for
GaAs on Si, GaP on Si, GaAs on GaP, and AlGaAs on Si are summarized in
Table 2. The table shows that the type of dislocation is affected by the growth mode rather than the lattice mismatch; type-I dislocations are dominant for material systems with a 3D growth mode, and the type-II dislocations are dominant for the material systems with a 2D growth mode.
In the cases of GaP/Si and GaAs/GaP, the type-II dislocation generation probably

115

116

2 Experimental Techniques
Dominant type of dislocation, lattice mismatch, and growth modes

Tab. 2

Dislocation

GaAs on Si
AlGaAs on Si
GaP on Si
GaAs on GaP

Type-I
Type-I
Type-II
Type-II

Lattice Growth mismatch mode
(%)
4.1
4.2
0.37
3.7

3D
3D
2D
2D

is due to glide from the surface because
2D growth occurs at the beginning of the growth, and the dislocation density of the substrate is extremely low compared with that of the epitaxial layer. On the other hand, in the case of GaAs/Si and
AlGaAs/Si, the dislocation generation cannot be explained solely by the mechanisms reported until now. If the type-I dislocations are generated after the reaction of two 60◦ dislocations, the Burgers vectors should satisfy the condition:
¯ −→ a/2[011] + a/2[101] − − a/2[110]
(a : lattice constant)
(type II)
(type II)
(type I)
However, the probability that all the type-II dislocations are changed to type-I dislocations is very low. Moreover, it is impossible to explain the generation of type-I dislocations by climb from the surface.

The generation of type-I dislocations by the climb process is enhanced by increasing the point defect density. Therefore, the type-I dislocation should be observed in
GaP on Si as the density of point defects is high in GaP grown on Si because of the high growth temperature. But, type-I dislocations are rarely observed in GaP on
Si. Therefore, another mechanism should be considered to explain the generation of type-I dislocations.
If the 3D islands are formed at the initial stage of the growth, the size of the island increases with growth time, although the island density is constant. This means that the misﬁt dislocations are generated while the island size is increasing because the spacing between islands is much larger than that of the misﬁt dislocations. The type-I dislocation generation mechanism for 3D growth is shown in Fig. 8 in the case of GaAs on Si. When the island size is smaller than the critical size, the misﬁt strain is accommodated by elastic strain. The strain energy of the island is raised when the island size is increased. In order to relieve the lattice mismatch, misﬁt dislocations are generated at the edge of the island when the critical island size has been exceeded as shown in Fig. 8. Therefore, type-I dislocation generation is preferentially enhanced when the initial growth mode is

GaAs

T
Si

Fig. 8 Dislocation generation mechanism for GaAs on Si substrate. 2.2 Deposition of (Multiple Junction) Semiconductor Surfaces

3D, and type-II dislocations are dominant when the initial growth mode is 2D.
Effect of TCA
In0.1 Ga0.9 As/GaAs SLS was inserted in the
GaAs layer to reduce the dislocation density. The individual layer thickness and the total layer numbers of SLS are 20 nm and
10, respectively. The TCA was performed just before the SLS growth (TCA1) and after the SLS growth (TCA2). The upper and lower temperatures of TCA are 900 and 300 ◦ C, respectively. Numbers of TCA were changed in this study. The intentional doping was not performed in GaAs on Si.
The sample structure is shown in Fig. 9.
2.2.4.2

Figure 10 shows the dark spot defect
(DSD) density as a function of the number of TCA1 for various number of TCA2. The
DSD density of GaAs on Si without SLS is also shown. The DSD density decreases with increasing the number of TCA1 and
TCA2. The lowest DSD density obtained in this study is 3.8 × 106 cm−2 . It is indicated that the TCA both before and after the
SLS growth is more effective in reducing the DSD density. Even if the dislocation density is reduced by TCA1, dislocations are generated in the SLS by the lattice mismatch of SLS and GaAs. Although the individual layer thickness of SLS is thinner than the critical thickness, the total

GaAs (1.8 µm)
TCA 2
TCA 1

InGaAs (20 nm)/GaAs (20 nm) SLS

GaAs (1.0 µm)

Si

Fig. 9 GaAs on Si substrate using InGaAs/GaAs SLS buffer layer. 108

DSD Density
[cm−2]

Without SLS
With SLS (TCA2 = 0)
107

With SLS (TCA2 = 1)

DSD density of GaAs on Si as a function of TCA1 for various number of TCA2.
Fig. 10

106

0

1

2

3

4

Number of TCA1

5

6

117

2 Experimental Techniques

SLS thickness is thicker than the critical thickness [12]. It is suggested from the experimental results that the dislocations generated at SLS are bended by TCA2, resulting in the low dislocation density.
Until now, the low etch pit density on the order of 106 cm−2 has been obtained using
SLS and TCA for the total epitaxial layer thickness of more than 3.5 µm [40–43].
Few papers have been reported on the growth of GaAs on Si, with the dislocation density of 106 cm−2 at the epitaxial layer thickness of less than 3 µm.
The crystal quality of GaAs with 900 ◦ C
TCA and SLS is inferior to that on a GaAs substrate because a high density of dislocations is generated in the epitaxial layer, which degrades the minority carrier lifetime. The TCA temperature was optimized to improve the minority carrier lifetime.
In this experiment, SLS buffer later was not used.
Figure 11 shows the DSD density of
GaAs grown on Si substrate revealed by electron beam–induced current (EBIC) measurement for various TCA temperature. The DSD density decreases with increasing the TCA temperature gradually and is on the order of 106 cm−2 at
1000 ◦ C. The crystal quality improvement by the relatively high TCA temperature

is due to the enhancement of dislocation movement, large compressive stress, and the generation of point defects at higher temperature, which in turn reduce the dislocation density effectively.
Figure 12 shows the minority carrier lifetime of GaAs, Al0.15 Ga0.85 As and
Al0.22 Ga0.78 As grown on Si for various
TCA temperatures. The minority carrier lifetime is also improved with increasing the TCA temperature, which is supported by the decrease of DSD at high temperature. The minority carrier lifetime of
GaAs grown on Si with 1000 ◦ C TCA is
3.36 ns. Those for GaAs and AlGaAs for various Al compositions, are shown in
Fig. 13. The lifetime of GaAs grown on
GaAs substrate is also plotted for comparison [44–48]. Although it is impossible to compare the lifetime because the carrier concentration is not the same for all the samples, it is estimated that the lifetime of
GaAs and AlGaAs on Si is approximately one order shorter than those grown on
GaAs substrate.
Effect of Hydrogenation
Hydrogenation was carried out in a quartz tube, where a hydrogen plasma was excited by rf power via a copper coil encircling the quartz tube. The
2.2.4.3

108

Dark spot density
[cm−2]

118

107

106
900

950

TCA Temperature
[°C]

1000

Fig. 11 DSD density of GaAs on Si as a function of TCA temperature. 2.2 Deposition of (Multiple Junction) Semiconductor Surfaces
Fig. 12 Minority carrier lifetime of GaAs and AlGaAs for various
TCA temperature.

Minority carrier lifetime
[ns]

4

x=0

3

2

x = 0.15
1
x = 0.22

0
850

900

950

1000

TCA Temperature
[°C]

Minority carrier lifetime τ
[ns]

100

Ahrenkiel et al. (1988)
'tHooft et al. (1981) on GaAs
Timmons et al. (1990)
Zarem et al. (1989)
Ahrenkiel et al. (1991)
GaAs and AlGaAs on Si with 1000 °C TCA

10

1

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 substrate temperature during the plasma treatment were 90 W, 2 hours, and 250 ◦ C, respectively. In order to recover shallow level passivation and the damage induced by the plasma treatment, post annealing was performed in an AsH3 + H2 ambient

at various temperatures ranging from
350 ◦ C to 450 ◦ C for 10 minutes. The TCA temperature was 900 ◦ C and the SLS buffer layer was not used.
Carrier concentration proﬁles of unintentionally doped GaAs grown on Si substrates, before and after hydrogenation, are

119

2 Experimental Techniques
1018

Carrier concentration
[cm−3 ]

120

As-grown
1017

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 n-type (1 × 1017 cm−3 ) because of Si autodoping during the growth [49]. For the hydrogenated sample, the carrier concentration is reduced to about 3 × 1016 cm−3 at depth exceeding 1 µm. This is due to the electrical passivation of the shallow levels.
Because the major donor in GaAs grown on Si substrates is Si via auto-doping from the substrate, passivation will occur by the formation of SiH0 complexes via the reaction
Si+ + H0 + e− − − SiH0
−→
The SiH0 complex will then be dissociated by heat or applied electric ﬁelds. There is a kink in the concentration proﬁle curve at a depth of nearly 0.8 µm, which corresponds to the plasma-induced damage [50], as the knee goes deeper with increased plasma treatment time. After a 10-minute annealing at 450 ◦ C in AsH3 + H2 ambient, the donor electrical activities were completely restored to their initial levels.

Figure 15 shows the 4.2 K PL spectra of GaAs on Si for as-grown sample and hydrogenated sample, hydrogenated sample annealed at 450 ◦ C. The major peaks are peak B corresponding to the heavy hole-associated free exciton and peak C corresponding to the carbon impuritybound exciton. After the sample is treated by the hydrogen plasma, the full width at a half maximum (FWHM) of peak B narrows from 4.49 meV to 3.83 meV. This narrowing is due to the passivation of localized states. With 450 ◦ C annealing where the shallow level is completely recovered, the FWHM is a little narrower than that of the as-grown sample.
Figure 16 shows the minority carrier lifetime derived from the time resolved photoluminescence decay curve. The minority carrier lifetime increases from 1.66 ns
(as-grown) to 4.66 ns after the plasma treatment, and gradually decreases with increasing annealing temperature. It is difﬁcult to judge the crystal quality only by the

2.2 Deposition of (Multiple Junction) Semiconductor Surfaces

4.2 K

A

B

C

D

E

PL Intensity
[a.u.]

4.19 meV
Hydrogenated + annealed (450 °C)

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 hydrogenated GaAs on Si with various postannealing temperature.

Minority carrier lifetime
[ns]

minority carrier lifetime because the lifetime is also affected by the shallow carrier concentration. In Fig. 14, the shallow carrier concentration of the 450 ◦ C annealed sample is the same as that of the asgrown sample because with passivation the shallow level is completely restored. The minority carrier lifetime after hydrogen 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 the defects generated by lattice mismatch and thermal expansion mismatch are electrically and optically passivated. The

5

4

3

2
As-grown
1

0

Hydrogenated

350

450

Annealing temperature
[°C]

121

122

2 Experimental Techniques

longer minority carrier lifetime of the hydrogenated sample (before annealing) is due to defect passivation and shallow level passivation. Species in the plasma include free radicals, ions, and electrons. Among these species, free radicals can effectively passivate the defects. Furthermore, it is well known that ions can damage the semiconductor surface. Therefore, during the hydrogenation process, defect passivation and damage formation take place at the same time. The minority carrier lifetime of the hydrogenated sample annealed at
450 ◦ C is longer than that of the as-grown sample because the defects generated during the plasma treatment are passivated.
The DLTS spectra show that the peak of the Si-defect related level becomes smaller after hydrogen plasma treatment. It suggests that the Si-related defect level is

passivated by hydrogenation. The passivation effect remains even after 450 ◦ C annealing, where the shallow level is completely restored.
From these experiments it can be concluded that (1) the shallow level that has been passivated by hydrogenation is completely recovered by annealing at
450 ◦ C in AsH3 + H2 ambient, (2) the deep levels are still passivated by hydrogen after annealing at 450 ◦ C in AsH3 + H2 ambient, and (3) hydrogenation followed by 450 ◦ C annealing produces a longer minority carrier lifetime at the same shallow carrier concentration.
Application to Photovoltaic Device
Figure 17 shows the schematic crosssectional view of a GaAs/Si tandem solar cell. It consists of n+ -GaAs buffer layer,
2.2.4.4

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

n -GaAs

0.6 µm

n +-GaAs

1.5 µm

n +-Si

1.1 µm

p -Si (350 µm)

p +-Si

Top cell

Au/AuSb

Bottom cell

0.5 µm
Au

Schematic cross-sectional view of three-terminal GaAs/Si tandem solar cell.
Fig. 17

2.2 Deposition of (Multiple Junction) Semiconductor Surfaces

n-GaAs layer, p + -GaAs, p + -Alx Ga1−x .
As graded band emitter layer (x : 0 →
0.29), p + -Al0.8 Ga0.2 . As window layer and p -GaAs cap layer. After the epitaxial growth, electrodes of Au-Zn/Au,
Au-Sb/Au, and Au were formed on the p + GaAs layer, n+ -Si and p + -Si, respectively.
The surface of the cell was coated with a double-layer antireﬂection coating (ARC) using ZnS (49 nm) and MgF2 (71 nm). The total area and active area of the solar cell are 25 mm2 and 22.05 mm2 , respectively.
The photovoltaic measurements were performed under AM0 and 1sun conditions at 27 ◦ C.
The open-circuit voltage (Voc ) of GaAs solar cell fabricated on Si substrate as a function of TCA temperature is shown in Fig. 18 [51]. Voc of GaAs solar cell fabricated on GaAs substrate is also indicated. It is known that Voc is very sensitive to the minority carrier lifetime.
Voc of solar cell grown on Si is improved with increasing the TCA temperature, but
0.1–0.12 V smaller than that fabricated on
GaAs substrate.
Table 3 shows short-circuit current (Jsc ),
Voc , ﬁll factor (FF) and conversion efﬁciency (η) of the solar cell with 1000 ◦ C
TCA and 900 ◦ C TCA (shown in the parentheses). A total conversion efﬁciency of
22.1% has been achieved by combining the

Photovoltaic properties of GaAs/Si tandem solar cell with 1000 ◦ C TCA

Tab. 3

(mA/cm2 )
Jsc
Top cell
Bottom cell

(V)
Voc

34.8
(32.6)
15.0
(14.8)

0.90
(0.88)
0.52
(0.52)

(%)
FF

(%) η 76.8
17.7
(75.7) (16.1)
77.2
4.4
(78.3) (4.5)

Total

22.1
(20.6)

Note: Values in ( ) show those with 900 ◦ C TCA.

GaAs top cell (η = 17.7%) and Si bottom cell (η = 4.4%) in a three-terminal conﬁguration. This is the highest efﬁciency for the GaAs/Si monolithic tandem solar cell ever reported.
Although the conversion efﬁciency of the top cell is improved with increasing the
TCA temperature from 900 ◦ C to 1000 ◦ C, the conversion efﬁciency of the Si bottom cell is reduced. This would be due to the degradation of minority carrier lifetime and the formation of the deep junction by the high-temperature heat treatment.
A two-terminal Al0.15 Ga0.85 As/Si tandem solar cell was fabricated in order to attain the photocurrent matching between the top cell and the bottom cell.
The structure and the current-voltage

Open-circuit voltage
[V]

1.1

Open-circuit voltage of
GaAs solar cell on Si as a function of TCA temperature.
Fig. 18

1.0

on GaAs on Si

0.9

0.8
900

950

1000

TCA Temperature
[°C]

123

2 Experimental Techniques
Au-Zn/Au
p+-GaAs p+-Al0.8Ga0.2As 1 × 1018 cm−3

p+-Alx Ga1−x As

1×

cm−3

0.3 µm

n -Al0.15Ga0.85As

2 × 1017 cm−3

1.0 µm

n+-Al0.15Ga0.85As

1×

MgF2 /ZnS

(x : 0.15 → 0.30)

1018

0.05 µm

Top cell
1018

cm−3

2.5 µm

n+-GaAs

20 nm

p+-Si

1.0 µm

n -Si

350 µm

n+-Si

0.8 µm

Bottom cell

Au-Sb/Au

Fig. 19 Schematic cross-sectional view of two-terminal
Al0.15 Ga0.85 As/Si tandem solar cell.

Fig. 20 Current-voltage characteristics of two-terminal
Al0.15 Ga0.85 As/Si tandem solar cell. 30
25

Current density
[mA cm−2 ]

124

20
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
20, respectively. It is observed that the short-circuit current of the top cell is perfectly matched to that of the bottom cell.
A conversion efﬁciency of 21.2% utilizing

the two-terminal conﬁguration has been obtained [52].
It has been pointed out that a 2-terminal tandem solar cell with an efﬁciency higher than 30% can be obtained by using a

2.2 Deposition of (Multiple Junction) Semiconductor Surfaces

top cell material with a band gap energy of 1.7–1.8 eV over an Si bottom cell.
This structure resulted in photocurrent matching between the top cell and the bottom cell. However, in our study, the current matching was obtained by using the Al0.15 Ga0.85 As top cell, of which the band gap energy is 1.61 eV. This is because the short-circuit current of the top cell is inferior to the ideal one. The main reason for the degradation is the short minority carrier lifetime caused by a high density of dislocation in the AlGaAs layer on Si. If it becomes possible to grow an AlGaAs layer on an Si substrate with a long minority carrier lifetime, comparable to that grown on GaAs substrate, a higher efﬁciency tandem solar cells can be obtained by increasing the Al composition so that the photocurrent matching between the top cell and the bottom cell is retained.
The improvement of the Si bottom cell is also important for the increase of the total conversion efﬁciency. The main problem is that the conversion efﬁciency is degraded after the crystal growth process [53]. The junction depth becomes deeper and the As atoms diffuse into the Si substrate during the growth. A low temperature growth process is necessary for the improvement of the bottom cell. Therefore, in the future, the technology to grow an AlGaAs layer with long minority carrier lifetime at low temperature should be investigated to increase the efﬁciency of the tandem cell.
2.2.5

Summary

The nucleation, dislocation generation, stress relaxation, and the annihilation of
APD in the GaP/Si heteroepitaxial growth have been reviewed. The growth mode appears to be island-type for the low values of V/III ratio and the low gas

pressure. When the V/III ratio or the gas pressure is increased, the growth mode changes from island-type to layer-type for thin GaP layer thickness. The two-type of misﬁt dislocations, which are generated by the lattice mismatch and the thermal expansion mismatch, are observed. A high density of APDs that propagate to the surface and are annihilated during growth has been observed. The nucleation of GaAs on Si and the effects of SLS buffer layer,
TCA, and hydrogenation are described.
The crystal quality has been improved by using SLS buffer layer, increasing the TCA temperature and using hydrogen plasma treatment. The GaAs layer grown on Si substrate has been applied to photovoltaic devices. References
1. J. P. Benner, J. M. Olson, T. J. Coutts, in
Advances in Solar Energy, (Ed.: K. W. Boer),
American Solar Energy Society, Inc., 1992,
p. 125, Vol. 7, Chapter 4.
2. M. A. Green, K. Emery, K. Bucher et al.,
Progr. Photovolt. 1999, 11, 31.
3. T. Soga, S. Hattori, S. Sakai et al., Electron.
Lett. 1984, 20, 916–918.
4. M. Akiyama, Y. Kawarada, K. Kaminishi,
Jpn. J. Appl. Phys. 1984, 23, L843–L845.
5. W. I. Wang, Appl. Phys. Lett. 1984, 44,
1149–1150.
6. B.-Y. Tsaur, G. M. Metze, Appl. Phys. Lett.
1984, 45, 535–537.
7. W. T. Masselink, T. Henderson, J. Klem et al., Appl. Phys. Lett. 1984, 45, 1309–1311.
8. T. Egawa, H. Tada, Y. Kobayashi et al., Appl.
Phys. Lett. 1990, 57, 1179–1181.
9. T. Soga, T. Kato, M. Yang et al., J. Appl. Phys.
1995, 78, 4196–4199.
10. T. Soga, T. George, T. Suzuki et al., Appl.
Phys. Lett. 1991, 58, 2108–2110.
11. R. Fischer, D. Neuman, H. Zabel et al., Appl.
Phys. Lett. 1986, 48, 1223–1225.
12. Y. Watanabe, Y. Kadota, H. Okamoto et al.,
J. Cryst. Growth 1998, 93, 459–463.
13. N. Chand, R. People, F. A. Baicocchi et al.,
Appl. Phys. Lett. 1986, 49, 815–817.

125

126

2 Experimental Techniques
14. M. Yamaguchi, A. Yamamoto, M. Tachikawa et al., Appl. Phys. Lett. 1988, 53, 2293–2295.
15. M. Yamaguchi, J. Mater. Res. 1991, 6,
376–384.
16. K. Nozawa, Y. Horikoshi, Jpn. J. Appl. Phys.
1991, 30, L668–L670.
17. H. Shimomura, Y. Okada, M. Kawabe, Jpn.
J. Appl. Phys. 1992, 31, L628–L630.
18. F. R. N. Nabarro (Ed.), in Dislocations in
Solids, North-Holland, Amsterdam, 1979, pp. 461–545.
19. K. Tamamura, K. Akimoto, Y. Mori, J. Cryst.
Growth 1998, 94, 821–825.
20. T. Soga, T. Suzuki, M. Mori et al., J. Cryst.
Growth 1993, 132, 134–140.
21. T. Soga, T. George, T. Jimbo et al., Jpn.
J. Appl. Phys. 1991, 30, 3471–3474.
22. N. Noto, S. Nozaki, T. Egawa et al., Mater.
Res. Soc. Symp. Proc. 1989, 48, 247–252.
23. T. Soga, T. George, T. Jimbo et al., Appl.
Phys. Lett. 1991, 58, 1170–1172.
24. O. Ueda, K. Kitahara, N. Ohtsuka et al.,
Mater. Res. Soc. Symp. Proc. 1991, 221,
393–398.
25. J. Nishizawa, T. Kurabayashi, J. Cryst. Growth
1988, 93, 98–102.
26. G. B. Stringfellow, J. Cryst. Growth 1991, 115,
418–420.
27. T. Soga, T. Jimbo, M. Umeno, Jpn. J. Appl.
Phys. 1993, 32, L767–L769.
28. T. Soga, T. Jimbo, M. Umeno, Appl. Phys.
Lett. 1993, 63, 2543–2545.
29. J. W. Matthews, in Epitaxial Growth Part B,
Academic, New York, 1975, Chapter 8.
30. M. Tachikawa, H. Mori, Appl. Phys. Lett.
1990, 56, 2225–2227.
31. T. Soga, H. Nishikawa, T. Jimbo et al., Jpn. J.
Appl. Phys. 1993, 32, 4912–4915.
32. O. Ueda, T. Soga, T. Jimbo et al.,
Defect Control in Semiconductor, (Ed.:
K. Sumino), Elsevier Science, Amsterdam,
1990, pp. 1141–1146.
33. P. M. Petroff, J. Vac. Sci. Technol. 1986, B4,
874–876.
34. T. Soga, T. Jimbo, M. Umeno, J. Cryst.
Growth 1994, 145, 358–362.

35. N. Otsuka, C. Choi, Y. Nakamura et al.,
Mater. Res. Soc. Symp. Proc. 1985, 67,
85–90.
36. J. W. Matthews, A. E. Blakeslee, J. Cryst.
Growth 1974, 27, 118–123.
37. K. Rajan, M. Denhoff, J. Appl. Phys. 1987, 62,
1710–1715.
38. D. Gerthsen, D. K. Biegelsen, F. A. Ponce et al., J. Cryst. Growth 1990, 106, 157–163.
39. T. Yao, H. Kakao, H. Kawanami et al.,
J. Cryst. Growth 1989, 95, 107–111.
40. T. Soga, T. Jimbo, M. Umeno, Appl. Phys.
Lett. 1990, 56, 1443–1445.
41. T. Soga, S. Nozaki, N. Noto et al., Jpn. J. Appl.
Phys. 1989, 28, 2441–2445.
42. N. Hayafuji, S. Ochi, M. Miyashita et al.,
J. Cryst. Growth 1989, 28, 2441–2445.
43. T. Soga, H. Nishikawa, T. Jimbo et al.,
J. Cryst. Growth 1991, 107, 479–483.
44. R. K. Ahrenkiel, D. J. Dunlavy, T. Hanak,
Sol. Cells 1998, 24, 339–347.
45. R. K. Ahrenkiel, B. M. Keyes, T. C. Shen et al., J. Appl. Phys. 1991, 69, 3094–3099.
46. H. A. Zarem, J. A. Lebens, K. B. Nordstrom et al., Appl. Phys. Lett. 1989, 55, 2622–2624.
47. M. L. Timmons, T. S. Colpitts, R. Venkaasubramanian et al., Appl. Phys. Lett. 1990, 56,
1850–1852.
48. G. W. ’tHooft, C. Van Opdorp, H. Veenvliet et al., J. Cryst. Growth 1981, 55, 173–179.
49. S. Nozaki, J. J. Murray, A. T. Wu et al., Appl.
Phys. Lett. 1989, 55, 1674–1676.
50. N. Das Gupta, R. Riemenschneider, H. L.
Hartnagel, J. Electrochem. Soc. 1993, 140,
2038–2043.
51. T. Soga, M. Kawai, K. Otsuka et al., 2nd
World Conference and Exhibition on Photovoltaic Solar Energy Conversion, European
Commission, Joint Research Centre, Ispra,
Italy, 1998, pp. 3737–3740.
52. T. Soga, K. Baskar, T. Kato et al., J. Cryst.
Growth 1997, 174, 579–584.
53. M. Yang, T. Soga, T. Egawa et al., Sol. Energy
Mater. Sol. Cells 1994, 35, 45–51.

2.3 Grafting Molecular Properties onto Semiconductor Surfaces

2.3

Grafting Molecular Properties onto
Semiconductor Surfaces
Rami Cohen, Gonen Ashkenasy,
Abraham Shanzer, David Cahen
Weizmann Institute of Science, Rehovot, Israel
2.3.1

Surface Electronic Properties

The chemistry of the crystal surface is very different from that of the bulk. Although the atoms are arranged in a well-ordered structure in the bulk, on the surface they may adopt a different position from that corresponding to the perfect crystal lattice because of the absence of part of the neighboring atoms. In that case, the surface is said to be relaxed or reconstructed [1].
A second difference between nearly every surface and the bulk is associated with the chemical environment of the bonding atoms. Under ambient conditions, the surface is covered by one or more atomic layers of adsorbates such as oxygen (mostly as oxide or hydroxide), carbon or hydrocarbons and is not atomically clean (i.e. the atomic composition is not the same as that in the bulk). Thus, it is not surprising that nearly always the ‘‘real’’ surfaces are disordered and lack the periodicity of the bulk. Correspondingly, the positions and distribution of energy levels and their occupation by electrons at the surface generally differ from those associated with the bulk.
The importance of this situation becomes obvious when we try to use these crystals, especially semiconductor crystals, in a device. In this case, surface defects and imperfections play a dominant role in the electron transport in and out of the device and in this way inﬂuence its performance.
It is to be remembered that all contacts of semiconductors to the outside

world and the junction between two different semiconductors involve their surfaces.
Even one of the simplest of electronic devices, the homojunction diode, requires outside metal–semiconductor contacts.
Therefore, understanding and control over the electronic properties of semiconductor surfaces are essential for constructing devices and for ﬁne-tuning their performance. Figure 1 shows the energy-band diagram of an n-type semiconductor and the electrical properties associated with the surface. This chapter mostly deals with n-type semiconductors and deﬁnitions are given only for this type. The difference between the energetics of the bulk and that of the surface is demonstrated by the relatively high density of energy levels inside the semiconductor band gap. These levels are referred to as surface states and their importance stems from the fact that they are involved in most electronic loss mechanisms such as charge-trapping and recombination. It was shown [4], using quantum mechanical calculations, that surface states can localize electrical charges in contrast to bulk states where electrons are delocalized.
Because of this localization the surface becomes charged with respect to the bulk and an electric potential difference is created between the surface and the bulk, the so-called built-in potential (Vs ). This is shown in the band diagrams as bending of the valence and conduction bands, a feature called ‘‘surface band-bending’’. In relation to the bulk, the surface states of an n-type semiconductor localize negative charges and the opposite is true for p -type material. Because it is easier to measure the band-bending than the surface states, in many cases the surface band-bending is used as an indicator for changes in the density of surface states. Because the capture

127

2 Experimental Techniques
Vacuum level

Electron affinity

Φ
Work
function

χ

CB

Band bending
Vs

Empty surface states
Occupied
surface states EF
Eg

Surface resonances Energy

VB

Density of states

Distance
Space charge region (SCR)

Distance

128

Distance

Distance

(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.
Fig. 1

2.3 Grafting Molecular Properties onto Semiconductor Surfaces

cross-sections of electrons and holes are higher on the surface than in the bulk, surface states also serve as fast recombination centers of electrons and holes. This phenomenon can critically affect the electrontransfer efﬁciency in electronic devices and thus their performance. The recombination rate of electrons and holes at the surface is expressed in units of velocity
(cm s−1 ) and is called surface recombination velocity (SRV or s ). It reﬂects the rate at which minority carriers (holes in n-type materials) are consumed at the surface. In many cases, SRV is determined by the surface states close to the middle of the band gap where the chances for electron-hole recombination are the highest [5].
Other surface electrical properties indicated in Fig. 1 are the work function,
, and the electron afﬁnity, χ . In semiconductors, electron afﬁnity is deﬁned as the difference in energy between the local vacuum level [6] and the bottom of the conduction band (CB) at the surface. The
(local) work function is the minimum energy required for an electron to escape into vacuum (just outside the range of the crystal potential) from the Fermi level. Although this deﬁnition is straightforward for metals, for semiconductors it should be borne in mind that there are mostly no real energy levels at the Fermi level. The work function is determined by the electron afﬁnity, the band-bending, and the energy difference between the Fermi level and the conduction/valence band in the bulk. 2.3.2

Requirements of Molecular Surface
Treatments

Because the surface charge and dipole are determined by the surface chemistry, it is the surface chemistry that to a large extent

determines the surface electronic properties. Therefore, to control the surface electrical properties and to make the semiconductor useful for electronic devices, we look for treatments that will interact with the surface and modify it chemically.
The requirements of molecular surface treatments, which are also named passivation treatments, are the following:
1. Remove surface and interface states or at least eliminate them in the energy interval of the band gap. Because the presence of surface states is conducive for charge-trapping and recombination effects, moving them out of the band gap is necessary for the construction of high-speed semiconductor devices like the ones that are based on GaAs and InP.
2. Tuning the surface electron afﬁnity.
Control over the electron afﬁnity is essential for band edge engineering of interfaces such as those in photovoltaic solar cells and for controlling the barrier height for electron loss to the surroundings. The second effect is demonstrated for the GaAs/(Al,Ga)As system where a band offset of 0.4 eV [7] did not block electron loss from the
GaAs into the passivating (Al,Ga)As layer. 3. Strong bonding that can withstand device-processing. Ideally, the molecules will chemisorb on the surface.
It should be noted that the binding groups themselves, while interacting with the surface, could modify the surface chemistry and thus the surface energetics on binding.
4. Chemical protection from adsorbates, especially atmospheric adsorbates. In those cases where surface oxidation is detrimental, there is the daunting task of protection from O2 .

129

130

2 Experimental Techniques

2.3.3

Strategies to Control Surface Electronic
Properties

Strategies for tuning surface electronic properties, work function, electron afﬁnity, band-bending, and SRV have to take into account the origin of chemical and physical properties. Use of the versatility of molecular chemistry can help control these properties in a predetermined fashion.
2.3.3.1 Controlling the Band-bending (Vs ) and Surface Recombination Velocity (SRV)
As noted earlier, band-bending near the surface results from charge localization on the surface states. Thus, control over the band-bending requires a mechanism that will control the density and occupation of the surface states inside the band gap.
Vacuum level

Figure 2 indicates a method to modify the band-bending by shifting surface-state energies with respect to the band edges. If the surface states are shifted out of the band gap, above the CB minimum or below the valence band (VB) maximum
(i.e. converted into surface resonances), they will couple to the band continuum and their charge will be delocalized in the band. As a result, Vs will be modiﬁed.
Another way to change the Vs is by changing the surface-state energy inside the band gap. The states are occupied at energies below the Fermi level and are empty at those above it (this holds, strictly speaking, only at 0 ◦ K). Therefore by modifying the surface states so that their energies will change from above (below) to below (above) the Fermi level, the surface states can be occupied (emptied) and the
Vacuum level

CB

CB

Surface treatment EF

EF
Occupied
surface states Vs′

Vs
VB

VB

SCR

SCR

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 .

Fig. 2

2.3 Grafting Molecular Properties onto Semiconductor Surfaces

from a change in the chemical environment caused by adsorbed extrinsic atoms.

surface charge changed. Change in the distribution of energies of the surface states can also affect the SRV. Because the SRV depends on the density of surface states with energies around the midgap, shifting of the surface-state energies away from the midgap can lead to reduction in the probability of charge recombination.
The strategy to modify surface states, and thus the Vs and SRV, is based on interaction of chemically grafted molecules with these states. The key is to ﬁnd molecules that will modify the semiconductor surface chemistry in a way that involves the surface states. In this respect, the origin of the surface states should be considered. Intrinsic surface states originate from the termination of the crystal bulk and the breaking of chemical bonds at the surface, whereas extrinsic surface states originate from crystal imperfections, such as missing surface atoms, line defects, or

Controlling the Electron Afﬁnity, χ
By deﬁnition, χ depends on the energy difference between the vacuum level and the bottom of the CB at the surface.
Therefore, any treatment that inﬂuences the surface potential will modify χ .
Modiﬁcation of the surface potential can be achieved by utilizing polar molecules that will bind to the surface and change its potential. Figure 3 shows schematically the manner in which a polar molecule can modify the surface χ . The surface potential will be reduced if the molecular dipole is pointing toward the surface and will increase if the dipole is directed in the opposite direction. This approach of using dipolar molecules was applied to tune the χ of metals [8–11] and semiconductors [8, 12–14]. It should be mentioned
2.3.3.2

A

(a)
A

+

d

q m −

Binding group Decrease of c d B

(b)
B

+d

q m −d

Increase of c

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.

Fig. 3

131

132

2 Experimental Techniques

that because short-range atomic forces determine the energy positions of the surface states inside the band gap, they need not be affected by the presence of the macropotential. Therefore, looking at
Fig. 1, a polar molecule modiﬁes the energy of the bands with respect to the vacuum level, but not necessarily those of the surface states with respect to the band edges or the surface-state (and band) energy level densities.
The change in the electron afﬁnity caused by the molecule’s dipole moment can be described in terms of a parallel plate capacitor, using the well-known
Helmholtz equation
V =N ×µ×

cos θ εεo (1)

where V is the potential drop caused by the dipole layer, µ is the dipole moment, N the dipole density per unit area, θ the angle between the dipole and the surface normal, ε is the relative dielectric permittivity of the ﬁlm, and εo the permittivity of free space [15]. If values for N , cos θ , and ε are known, V (and thus χ ) can be calculated. The equation also demonstrates that controlling χ can be achieved by tuning the dipole direction
(toward/from the surface), its magnitude and its orientation with respect to the surface normal. It has to be noted that the concept of a dielectric constant for a monolayer is problematic. It can be related to the molecular polarizability using the
Clausius-Mosotti relation (cf. discussion in [14]).
Controlling the Work Function,
As shown in Fig. 1, the work function is the energy difference between the Fermi and vacuum levels and depends on the electron afﬁnity and band-bending. Thus, by controlling either the electron afﬁnity

or band-bending, we can tune the work function =

χ±

Vs

(2)

with + for n-type and − for p -type.
2.3.4

Strategy for Molecule Selection

The chemical strategy to control the surface energetics using organic molecules is illustrated schematically in Fig. 4. The idea is to incorporate several molecular properties simultaneously in one molecule and to allow systematic modiﬁcation of one speciﬁc property, independent of others. This approach provides
1. a simple tool to investigate the relation of the macroscopic properties of the semiconductors and the molecular properties of the adsorbed molecules; 2. a simple tool to enable the development of models for surface engineering.
A different approach, which is more suitable for nonmolecular, extended bonded, electrically conducting inorganic materials, is to use different molecular layers, each of which has a different function.
For example, the ﬁrst layer can electrically passivate the surface and a second layer can give long-term chemical protection [7].
2.3.5

Organic and Inorganic Molecular Surface
Treatments

2.3.3.3

In general, we can classify the chemical surface treatments into two classes, organic and inorganic. Table 1 summarizes the advantages and disadvantages of each group. Although, as noted in the

2.3 Grafting Molecular Properties onto Semiconductor Surfaces
Auxiliary groups
Polar group
Light-sensitive group
Hydrophobic group

Binding groups covalent bond slightly polar bond
Alkyl chain/ ionic bond π system [benzene]

Molecular properties A

Synthesis
A

A

A

Adsorption

A

A

A
Macroscopic
properties
Φ, χ, Vs

Chemical strategy to control the surface electrical properties using multifunctional molecules.

Fig. 4

table, the chemical stability of organic treatments is generally lower than that of inorganic ones, Dorsten and coworkers [16] demonstrated that sulﬁdization of
GaAs using octadecylthiol gave rise to a sulﬁde layer that was as stable as that which could be obtained with corresponding inorganic treatments. Using the ﬂexibility of organic synthesis, one can incorporate several functional groups in a single molecule, each performing a different role and ideally independent of each other. The versatility of the organic treatments can also be used to study the relationship between modiﬁcation of the surface chemistry and surface electronic effects.

Inorganic Surface Treatments
Table 2 presents several examples of inorganic surface treatments used to modify the surface electronic properties.
The ﬁrst is oxidation of silicon (Si), a cornerstone of today’s electronic device industry [17]. The treatment moves most of the surface states out of the Si band gap [2, 18] and enables the worldwide use of silicon for device manufacturing. On
III–V and II–VI group semiconductors, the chemistry of the semiconductor’s native oxides is more complicated than on silicon, and plain oxidation of the surface cannot lead to chemically stable, nearly defect-free surfaces [7]. For example, on
2.3.5.1

133

134

2 Experimental Techniques
Tab. 1

Advantages and disadvantages of organic and inorganic treatments
Organic treatments

Advantages

Disadvantages

* Structural versatility and ﬂexibility * Can incorporate several properties simultaneously (polarity, hydrophobicity, light sensitivity) * Allow systematic modiﬁcation of one speciﬁc property, often without effect on other properties * Limited chemical and thermal stability

Inorganic treatments
* Chemical stability
* Thermal stability

* Strong binding/interaction

* Limited chemical ﬂexibility

Examples of substrates for which inorganic chemical surface modiﬁcations have been developed Tab. 2

Semiconductor

Treatment

References

Silicon

Oxygen
Hydrogen
Bromine
As

2, 18
18, 28
29
30

GaAs

Sulﬁde and Selenide
Ruthenium
Phosphine (PH3 )
P2 O5 /NH4 OH
Chlorine
Cesium
Sb
H2 S
Iodine

InP

CdS
Ruthenium

46
47, 48

CdTe (polycrystalline)

Hydrogen plasma
Ruthenium
Reactive metal interlayer
Oxygen anneal

49
50, 51
52
53

(Hg, Cd)Te

Sulﬁde
Hydrogen

54
55, 56

CuInSe2
(polycrystalline)

Oxygen anneal

53, 57

20, 24, 27, 31, 32, 33–37
38
39
40, 41
42
43
44, 24
26
45

2.3 Grafting Molecular Properties onto Semiconductor Surfaces

GaAs(100) and (110) [2] surfaces in the ambient, high surface state densities pin the Fermi level and generate a high SRV.
Covering GaAs(100) with (Al,Ga)As was found to be effective in terms of reducing surface electron-hole recombination, but could not block loss of electrons to the passivating (Al,Ga)As layer. Another approach to modify the surface chemistry of GaAs is the use of inorganic sulﬁdes.
These treatments were found to remove part of the surface oxide and form sulﬁdes of Ga and As [19]. Studies of metal-insulator semiconductor (MIS) and
Schottky diode structures of GaAs, treated with (NH4 )2 S and Na2 S [20, 21], revealed that aqueous sulﬁde treatments induce only minor changes in the net surface state density. Their main effect is reduction of the density of the trap states, which are farther in energy from the band edges than the shallow, doping levels.
This means that these treatments modify the positions of the surface-state energy levels [19, 22]. This observation agrees with results from other studies where sulﬁde treatment was claimed to repin the surface
Fermi level at a different energy [22–25].
A different approach was introduced by
Shen and coworkers [26] who used plasma
H2 S treatment. In contrast to the aqueous sulﬁde treatments, this treatment leads to signiﬁcant reduction of the surface-state density (up to three orders of magnitude) on sulﬁdization. The main drawback of the inorganic sulﬁde treatments is the instability of the sulfur-GaAs bond in the ambient (vs oxygen attack, in particular) [19, 27].
According to Lunt [58], the efﬁcacy of a selenide treatment applied to GaAs is expected to be stronger than that of a sulﬁde one because of the electrondeﬁcient nature of the GaAs binding site. Indeed, several studies [59, 60, 31]

support this hypothesis and reveal a strong reduction in the band-bending of
GaAs on treatment with Se-containing reagents. The stable phase was found to be Ga2 Se3 [32], which has a close lattice match to GaAs, and therefore creates an almost strain-free layer.
Another class of surface treatments is based on halogens. Halogens are mostly used for surface etching and were found to dissociate upon adsorption [19]. The efﬁcacy of halogen treatment for modifying the surface electronic properties was found to depend on the morphology and composition of the surface. In the case of GaAs, halogens showed high reactivity toward the
Ga-rich surface [61, 62].
Parkinson and coworkers demonstrated improvement in the open-circuit voltage
(Voc ) and ﬁll factor of n-GaAs/K2 Se-K2 Se2 KOH/C photoelectrochemical solar cells on treatment with Ru (III) [38]. The improvement in the cell performance was ascribed to the shift of the surface-state energies, which was thought to lead to a reduction in the main power-loss mechanism, electron-tunneling through the surface states. Treatment with Ru was also found to improve the performance, specially the stability, of InP-based photoelectrochemical solar cells [47]. This improvement was related to an increased barrier height [48].
Organic Surface Treatments
Table 3 summarizes examples of organic surface treatments used for changing the surface electronic properties. As a ﬁrst example we consider self-assembly of organo-silanes on oxidized silicon. These self-assembled monolayers are known to be stable and well ordered [63, 64]. Modiﬁcation of the monolayer properties can be achieved in a predetermined fashion by changing the chemical structure of the
2.3.5.2

135

136

2 Experimental Techniques
Tab. 3

Several organic chemical treatments used for surface modiﬁcations

Semiconductor

Treatments

References

Silicon, single crystal and porous

Self assembly of organosilanes
Acids and bases
Amines
Common organic solvents
Other organic adsorbates

GaAs

Organic disulﬁdes
Porphyrins for NO Detection
Dicarboxylic acids
Sulﬁdes and thiols
Dithiocarbamate
CH3 CSNH2
Benzoic acids
Alcohols
Ethyl iodide
Dimethylcadmium
Dimethylzinc
Trimethylgallium
Diethylzinc
HS(CH2 )Si(OCH3 )

InP

Dicarboxylic acids
Alkanethiol
Various other molecules

76
89
90

CdTe

Benzoic acids
Dicarboxylic acids

12
75

CdSe

Dicarboxylic acids
Dithiocarbamate
Thiols
Anionic sulfur donor
Ethylenediaminetetraacetic acid
Dialkyl chalcogenides
Oleﬁns
Aniline derivatives
Amines
Boranes
Fullerenes
Carbonyl compounds
TCNQ derivatives
Benzoic acids
Silapentanes and chlorinated Silanes
Cyanide
TOPO
Porphyrins for O2 Detection

TiO2

Benzoic acids

ITO

Benzoic and other carboxylic acids

CuInSe2

Dicarboxylic acids
Benzoic and hydroxamic acids
Disulﬁdes

65
69
66, 70
71
67, 128
72
73, 129
72, 74, 75–77
58, 68, 78, 79, 80
81
81, 82
14, 83, 130
84
85
86
86
86
87
88

76
91
92
93
94
95
96
97
98, 99
100
101
102
103
13
104
105
106
131, 132
107
108–110
111
12
8

2.3 Grafting Molecular Properties onto Semiconductor Surfaces

self-assembling molecules. On the basis of that, a systematic modiﬁcation of the surface electron afﬁnity and band-bending of n-type silicon was demonstrated by using self-assembly of substituted quinoline chromophores [65]. The chemical scheme included two steps:
• self-assembly of organo-silanes that add organic functionality to the surface, and
• grafting of a series of substituted quinoline chromophores that add the variable (in this case polar) functionality.
On porous silicon, which has a high surface-to-volume ratio and thus a large number of accessible surface states, a substantial change in PL was recorded on exposure to amines and different solvents [66, 67]. On GaAs, organic sulﬁdes and thiols were reported to induce PL changes that were attributed to changes in both the SRV and the Vs [58, 68].
Adsorbing a series of benzoic acids with systematically varying dipole moment led to a different response. Changes in contact potential differences (CPD) were correlated to changes in the electron afﬁnity without observable effect on the surface band-bending [14]. Substituted dicarboxylic acids, which were found to bind by a two-site mechanism, rather than a one-site one, such as the benzoic acids [112], modiﬁed both the bandbending and the electron afﬁnity of the
GaAs [72, 74]. This difference in the electrical effect on treatment with carboxylic and dicarboxylic acids can be explained by the different binding group and the change in the molecular energy levels [75].
On CdTe, CPD measurements showed that organic benzoic and dicarboxylic acids could modify the electron afﬁnity and/or the surface band-bending [12, 75]. On the wider band gap semiconductor, CdSe, PL

data that were obtained after adsorption of several series of organic molecules (including aniline and carbonyl compounds) were interpreted in terms of changes in the band-bending [96–99, 102], whereas CPD changes induced by adsorption of benzoic acids and aniline derivatives were rationalized in terms of changes in electron afﬁnity [13]. The different conclusions from luminescence intensity measurements and
CPD measurements were ascribed to the fact that the PL measurements were done in solution, whereas the CPD measurements were performed in air. This difference also reﬂects the effects of interaction with the surrounding medium.
To provide information on moleculesurface interaction and on the feasibility of chemical treatments for use in electronic devices, like sensors, attempts have been made to establish a correlation between molecular properties and changes in surface properties. Table 4 summarizes works in which such a correlation was found.
2.3.5.3.1 Correlation of Molecular Parameter with Changes in the Surface Electron
Afﬁnity As mentioned earlier, control of the electron afﬁnity or surface potential can be achieved by modifying the molecular dipole moments of the adsorbed molecules. Figure 5 shows the change in the electron afﬁnity of CdTe, CdSe, and GaAs, which was deduced from CPD measured by a Kelvin probe, upon grafting a series of substituted benzoic acids onto the semiconductor surface. The linear relation between the change in the electron afﬁnity and the dipole moment of the substituted benzoic acid is clearly observed. The dipole moments of the substituted phenyl groups reﬂect the electron withdrawing or donating power of the substituents. Bruening et al. used electrochemical and CPD measurements in

137

138

2 Experimental Techniques
Tab. 4

Transfer of molecular properties to the macroscopic properties of semiconductors

Type of correlation

Sample

Change in electron afﬁnity, χ , with dipole moment of the benzoyl substituent and cammett substituent constants

Silicon (single crystal)
GaAs
CdTe
CdSe
CuInSe2
TiO2
ITO
Silicon
CdTe
CdSe
CdSe

Change in built-in potential, Vs , with ionization potential, proton afﬁnity, electron-accepting power; LUMO energy of the molecules
Change in photoluminescence, PL, with electrochemical redox potential
Change in surface recombination velocity
(SRV), s, with ionization potential
Change in PL with Hammett parameters
Change in barrier height with number of carbon atoms in chain
Change in barrier height with molecular dipole moment combination with ellipsometry and FTIR data to study the effects of binding a series of cyclic disulﬁdes with systematically varying dipole moment and different degrees of hydrophobicity to Au and CuInSe2 [8].
They found that the magnitude and direction of the change in electron afﬁnity depend on the surface coverage, the orientation of the molecular dipole relative to the surface normal, and the mode of binding. Knowing the molecular surface coverage and tilt angle and thus the residual molecular dipole moment, as well as the atomic electronegativity difference at the molecule-surface interface allowed theoretical estimates of the experimentally observed change in the electron afﬁnity, that is, it gave predictive power to Eq. (1).
In further work Wu et al. tested Eqn. (1) and used a series of molecules similar to that shown in Fig. 7, but completely conjugated. The main effect was found to be a smaller inﬂuence of the dipoles which

References
65
14, 77, 130
12, 75
13
12, 8
107
109
65, 113
75
96, 97, 114, 99
103

CdSe

115

CdSe, CdS
GaAs (thiols)

102
80

GaAs, Si

77, 128, 130

can be understood from the increase in effective dielectric constant of the molecular layer [130].
2.3.5.3.2 Correlation of Molecular Parameters with Changes in the Band-Bending In general, the change in Vs can be viewed by two mechanisms,

• molecule-induced surface oxidation/ reduction (generalized acid–base reaction) where the molecules either accept or donate electrons [117] according to the difference in the oxidation/reduction potentials, or alternatively by
• frontier orbital interaction mechanism, which involves the energy levels of the molecular frontier orbitals and of the semiconductor surface states [118, 119].
In the second mechanism, we consider the interaction between the highest occupied molecular orbital (HOMO)

2.3 Grafting Molecular Properties onto Semiconductor Surfaces
600
R

R = CH3O, H, F, Br, NO2

400

∆χ
[mV]

200

COOH
Nitro
Bromo
Fluoro

Hydro
0
Methoxy

CdSe
GaAs

−200

CdTe
−400

−2

0

2

4

Dipole moment
[Debye]
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].

Fig. 5

level of the surface or the molecule and the lowest unoccupied molecular orbital
(LUMO) level of the other. It is an extension of the well-known frontier orbital interaction between the energy levels of two molecules forming a complex, as shown in Fig. 6(a). On interaction, the
HOMO level is stabilized to lower energy, whereas the LUMO level is destabilized and pushed up in energy.
Table 4 summarizes studies in which the molecular ionization potential, which is related to the HOMO level, and the molecular electron afﬁnity, related to the molecular LUMO level, are correlated to the changes in Vs . This correlation ﬁts the orbital interaction mechanism.
Figure 6 shows several scenarios for the interaction of a given molecular
LUMO level with surface states at different energies. In the ﬁrst case, shown

in Fig. 6(b), the molecule’s LUMO level interacts with occupied surface states, which are below the Fermi level and therefore assume the role of the HOMO level.
On interaction, the surface-state energies increase (with respect to vacuum), that is, they are ‘‘pushed’’ down in energy toward the VB, whereas the molecular LUMO level is ‘‘pushed’’ up in energy toward the CB.
As a result, electrons that were formerly localized on the surface states may now occupy surface resonances with only partial localization at the surface. Therefore, they do not anymore (or little) contribute to the net surface charge. Reduction of Vs is thus expected. A different result is expected if the surface states are close to the semiconductor midgap, as shown in the second case (Fig. 6c). Because in this case the surface states are well removed from the band edge, only part of them will turn into

139

140

2 Experimental Techniques
(a) LUMO
HOMO

CB

EF

(b) LUMO

VB

e-

CB

(c) LUMO

EF

VB

h+

(d)

CB

EF
LUMO

VB

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.

Fig. 6

surface resonances and the net effect on the total surface state density, and thus on Vs , is expected to be moderate, compared to the ﬁrst case (Fig. 6b). The SRV, on the other hand, is expected to change signiﬁcantly because it depends critically

on the density of the midgap (deep) surface states [5].
In the third scenario, shown in Fig. 6(d), we consider the interaction of a molecular
LUMO level with empty surface states of a p -type semiconductor. Because the LUMO

2.3 Grafting Molecular Properties onto Semiconductor Surfaces

level of the molecule is well below the
Fermi level, electron transfer from the surface to the molecule is expected and the molecule plays the role of the HOMO level in the interaction. As in the other cases, the surface states that were pushed in energies to more than the CB minimum will turn into surface resonances that will lead to a reduction in Vs .
The energy levels of the molecule after interaction should also be considered.
These levels are modiﬁed (i.e. changed in energy, relative to the band edges) on interaction with the surface states, as was demonstrated [120, 133] for the interaction of thiophenol derivatives with Cu(111).
If, on interaction, the molecular LUMO level is at an energy that corresponds to somewhere within the band gap and below
EF , the state will act as an occupied surface state and will increase (decrease) the net surface charge of n-(p -)type surfaces. On the other hand, if the molecular LUMO level is pushed below the VB maximum, then the state acts as a surface resonance and only a moderate effect on the bandbending is expected. It should be noted that although our discussion in this article is limited to the molecular LUMO level, analogous interactions of molecular
HOMO levels with surface states can also be energetically favorable [119] and lead to modiﬁcation of surface states and/or SRV.
Figure 7 shows an example of the change in Vs of an etched n-CdTe single crystal as a function of the LUMO energy level of a series of dicarboxylic acid derivatives [75]. The systematic modiﬁcation of the molecular LUMO level was achieved by changing the molecule’s substituents from electron-donating to electron-accepting groups. The change in Vs was suggested to stem from two contributions: • a constant change of 170 mV because of the binding group and
• a second contribution that could be correlated with the molecules’ LUMO energy level and that increased with decreasing energy separation between the molecule’s LUMO state- and the surface state-energy level [75].
This second contribution is denoted schematically in Fig. 7 and can be attributed to extended coupling of the molecules’ energy level and the semiconductor surface states, as the molecules’
LUMO energies become closer in energy to those of the surface states. The predictive power of the frontier orbital interaction scheme, noted earlier, was demonstrated for n-CdSe, n-GaAs, n-InP, and p -GaAs crystals when changes in the Vs and in the
SRV of the crystals on interaction with a given dicarboxylic acid molecule (DCDC and DHDC) could be explained [112]. On the basis of the studies mentioned earlier, it was found that the ability of the chemical treatment to modify the electronic properties of semiconductor surfaces depends on the following parameters:
1. The molecule’s frontier orbital energy level and the difference in energy between that level and those of the interacting surface states. The smaller the energy distance, the stronger is the molecule-surface coupling and the larger can be the induced change in the surface electronic properties.
2. The surface-states’ energy levels and densities. In surfaces where the dominant surface states are close to the band edges, the main effect on interaction is a change in band-bending. On surfaces with dominant states close to midgap, the main effect on interaction

141

2 Experimental Techniques
(a)

X

X

600 o 500

X
( )B
∞∞

∞∞

o

O

O

O
HO

O

O

O
HO

OH

VI

OH

I. X = OMe; DMDC II. X = OMe; PMDC
III. X = H; DHDC
IV. X = CN; PCDC
V. X = CF3; DFDC
VI. X = CN; DCDC

400

∆V S
[mV]

V

IV
300

200
I
100

III

II
−3.5

−4

−4.5

−5

−5.5

(b)

I
−4

II

−5

III
IV

−4

CdTe

EF

−5

VI
−6

−6

−7

CdTe electron energy levels
[eV]

LUMO energy
[eV]

LUMO energy
[eV]

142

−7

(a) Change in Vs of n-CdTe upon adsorption of dicarboxylic acid derivatives as a function of the benzoyl substituents’ LUMO energy [75]. Vs is the change in Vs relative to the etched surface.
(b) Energy diagram of bare CdTe and of LUMO of isolated molecule before adsorption. The CdTe energy levels were experimentally measured by CPD, and the Fermi level was calculated from the doping density. Fig. 7

2.3 Grafting Molecular Properties onto Semiconductor Surfaces

is a change in surface recombination velocity.
According to this approach, designing molecular surface treatments should start by mapping the surface-state energies and then choosing a molecule with energy levels that are close to the target surface states and that can successfully interact with them. In practical cases, semiconductor surfaces may possess several populations of surface states and thus molecular selectivity will be required.
Further progress in this direction depends on
• better knowledge of molecular energy levels on an absolute scale (to allow comparison with semiconductor energy levels);
• consideration of other factors that might affect molecule-surface coupling, such as, for example, the shape of the orbitals, that is, symmetry considerations.
2.3.6

Examples of Molecular Control over
Optoelectronic Devices

The progress in device-processing techniques and the ability to design new multifunctional molecules, whose properties can be modiﬁed systematically, opens new directions for construction of molecular devices. Given below are several examples of the application of organic molecules for controlling the performance of optoelectronic devices based on inorganic, nonmolecular materials.
The ﬁrst system we consider is a Schottky diode and its derivatives. The electrical characteristics of this diode depend on the interface states and dipoles and can thus reﬂect changes induced by adsorbed molecules. Petty, Roberts, and coworkers
[121, 122] used the Langmuir-Blodgett (LB)

technique to prepare a Schottky diode of
Au/Cd-stearate/CdTe on both n- and p CdTe. Such a molecular ﬁlm increased both the barrier height and the opencircuit voltage of this MIS (metal-insulator semiconductor) solar cell structure. The ﬁll factor was found to depend on the thickness of the LB chain and was the highest for one monolayer (2.5 nm). These ﬁndings are in accordance with the optimum insulator thickness found for silicon MIS solar cells of 1 to 2 nm. Allara and coworkers [80] studied the effect of self-assembled monolayer (SAM) alkane thiols with different chain lengths on the Schottky barrier height of the Au/n-GaAs system. They found that while the barrier height increased only very slightly, the increase correlated linearly with the number of the carbon atoms in the alkane chain. They attributed this change to an increase of the negative charge at the binding group
(i.e. sulfur atom) as the thickness of the
SAM increased. However, an opposite effect on the barrier height was noted for the thiol-passivated Cu/n-GaAs system, which suggests that other mechanisms should also be considered.
Although tuning the Schottky barrier properties by way of changing the length of the molecules is essentially similar to the effect of changing the thickness of any dielectric ﬁlm between the metal and semiconductor, recent examples [77,
107–109, 123, 128, 130] show that true molecular effects can be obtained.
Vilan et al. demonstrated that molecules adsorbed at the Au/n-GaAs interface can tune the energetics of this metal–semiconductor diode, because of their chemical character, namely, by systematic substitution on aromatic rings [77]. Using a series of dicarboxylic acid derivatives whose dipole was varied systematically, they produced diodes in which the effective

143

144

2 Experimental Techniques

barrier height is tuned by the molecular dipole moment. Qualitatively, compared to the unmodiﬁed junction, molecules with a dipole pointing toward the surface increased the forward current in the current–voltage (I – V ) curve, whereas those with a dipole pointing away from the surface decreased it. The interesting thing here is that the molecules do not form a perfect monolayer and there is high probability of pinholes in the layer. Despite this, the molecules strongly affect the diode characteristics. These results reveal that it is not necessary to use molecules that conduct current through them to affect device electrical characteristics and that molecules that electrostatically interact with the surface can also be applied.
The results of Wu et al. [130], who prepared and measured Au/molecule/nGaAs diodes with the earlier mentioned series of conjugated molecules agree with these conclusions.
Selzer and Cahen [128] showed that a complementary conﬁguration can be used, as well. They adsorbed ligands on the metal rather than on the semiconductor side of the junction. In that case the metal work function, rather than the semiconductor’s electron afﬁnity is changed.
They used the molecules of Ref. [8], which are similar to those shown in Fig. 7(a), but with cyclic disulﬁde, instead of dicarboxylic acid binding groups, so as to allow chemisorption onto Au. The molecularly modiﬁed metal was then used to prepare
Au/molecule/SiOx /Si diodes. Results are naturally opposite to those obtained with
GaAS/Au diodes [77], because the substituted phenyl groups point in opposite directions in the two cases.
Molecular layers will probably also ﬁnd their way in optoelectronic devices. Krueger et al. used a series of benzoic acids [12–14] to modify the

current–voltage characteristics of the inorganic–organic TiO2 /spiro-MeOTAD
(an amorphous organic hole conductor) heterojunction [107]. These changes were correlated with the changes in the work function of the modiﬁed TiO2 and were attributed to variations in the built-in voltage at the TiO2 /spiro-MeOTAD interface.
A rough correlation between the change of the work function and the current–voltage characteristics was demonstrated also by
Cambell and colleagues [123], who used a self-assembled layer of conjugated thiol molecules to manipulate the Schottky barrier height between a Cu electrode and an organic electronic material used for light-emitting diode (LED) systems.
Zuppiroli, Nuesch, and coworkers demonstrated [108, 109, 134] that pretreatment of the transparent conductor indium tin oxide (ITO) with a few derivatized organic molecules can signiﬁcantly improve (decrease) the turn-on characteristic (ﬁeld) that denotes the ﬁeld above which light emission is observed in the ITO/poly(paraphenylene)/Al LED system. The turn-on ﬁeld was reduced from 200 MV m−1 to 100 MV m−1 after molecule adsorption. The decrease in the turn-on ﬁeld was correlated with the dipole moment of the carboxylic group.
The molecules modiﬁed (increased) the work function of the ITO. In this way they increased the hole injection efﬁciency in the device. Moreover, the device durability was increased. The stability under DC operation was increased from 15 minutes to
2.5 hours after the molecular treatment.
Friend and coworkers [124] showed that molecular surface treatments could also be applied in the new ﬁeld of organic-based optoelectronic circuits. Improvement in charge mobility of a ﬁeld effect transistor (FET)-LED device was demonstrated by using hexamethyldisilazane before

2.3 Grafting Molecular Properties onto Semiconductor Surfaces
Window layer
CdS
microcrystals

Functional molecules CuIn(Ga)Se2
Back contact

Schematic view (not to scale!) of molecular control over
Cu(In,Ga)Se2 /CdS heterojunction thin ﬁlm, polycrystalline solar cell.
The molecules are at the interface between the CuInSe2 ﬁlm and the CdS microcrystals. Fig. 8

deposition of the conjugated polymer.
The molecular treatment was suggested to promote phase segregation, that is, structural order at the interface. The earlier-mentioned studies reveal the potential of molecular treatments for tuning the charge transfer and charge transport across (semi)conductor/(semi)conductor interfaces. A different potential application of a molecular treatment is their use as a blocking layer against charge leakage. For example, Langmuir-Blodgett (LB) ﬁlms of
22-tricosanoic, CH2 =CH(CH2 )20 −COOH
[67] and hexadecanol [68] were shown to have charge-blocking properties on GaAsbased devices that are superior to those shown with them on (Al,Ga)As ones.
Using capacitance measurements of the metal-(molecular insulator)-semiconductor (MIS) structure indicated reduction of the interface trap density [83] and of the barrier characteristics [84].
Solar cells are another ﬁeld of applications for functional molecules. This idea is schematically presented in Fig. 8 for a heterojunction solar cell, where the molecules are located at the interface between the

n- and p -type surface. Gal et al. [111] used an experimental scheme similar to Fig. 8 to show the effect of organic acids on the electrical characteristics of polycrystalline
CuIn(Ga)Se2 (CIGSe)/CdS solar cells (see
Fig. 9). The changes in the I – V characteristics could be correlated with the molecules’ dipole moments that modify the band line-up at the interface, rather than as a direct effect on the surface Vs .
Specially in view of the later work [77, 107,
128, 130], these results indicate that it is not essential to form an ordered layer in order to modify device characteristics. However, what is required is a molecule that will modify the energetics at the interface so that any charge carrier passing from one side of the junction to the other will be inﬂuenced by it.
2.3.7

Summary

Fine-tuning of semiconductor surface electrical properties can be achieved by grafting multifunctional organic molecules onto the surface. In such molecules, one function takes care of the binding

145

2 Experimental Techniques

4

a b 2

Current
[mA]

146

c

0

−2

−0.8

−0.4

0

0.4

0.8

Voltage
[V]
Current–voltage characteristics of molecular, treated (a and b) and untreated (c) Au/CdS/Cu(In,Ga)Se2 /Mo solar cells in dark and under illumination (light intensity: ∼1.5 AM). The molecules used, shown schematically in Fig. 7, are DMDC and DCDC for a and b, respectively. The molecules were deposited on the polycrystalline Cu(In,Ga)Se2 surface before the wet chemical deposition of the CdS [111].

Fig. 9

to the semiconductor surface, whereas another one conveys a desired property to that surface and thus to the semiconductor. The way in which this can happen is conveniently studied by using a series of molecules in which the desired property is varied in a systematic fashion.
Implementation of molecular tools provides new opportunities for hybrid organic–inorganic systems. The increasing interest in, and importance of, nanoscale technologies and the use of nanocrystals, where surface-related effects often dominate bulk properties [106, 125–127], adds to the importance of controlling and engineering surface properties. Correlating the molecular properties of the grafted molecules to the changes in the macroscopic semiconductor properties can extend our understanding of molecule-surface interaction mechanisms

and help design molecular tools for controlling semiconductor surfaces. Recent work shows that this can be carried over to interfaces of semiconductors in devices where molecules will directly or indirectly inﬂuence active device characteristics. In this way, even small modiﬁcations in the molecular structure can induce signiﬁcant and predictable changes in device behavior. Acknowledgments

For our own contributions mentioned in this chapter we thank Leeor Kronik, Ellen Moons, Merlin Bruening,
Stephane Bastide, Dori Gal, Ayelet Vilan, Tamar Moav and Rahel Lazar, as well Ron Naaman and his group, for fruitful collaborations. We acknowledge

2.3 Grafting Molecular Properties onto Semiconductor Surfaces

the pivotal contributions of the late
Dr. Jacqueline Libman to the early parts of our work and Art Ellis (Univ. of Wisconsin,
Madison) for collaboration and inspiration for the molecule-surface interaction mechanism. Partial support from the USIsrael Binational- and the Israel Science
Foundations, the Minerva Foundation
(Munich), the Fussfeld Fund and the German BMBF, through its bilateral Energy
Research Program with the Israel Ministry of Science is gratefully acknowledged.
References
1. E. H. Rhoderick, R. H. Williams, MetalSemiconductor Contacts, 2nd edition, Oxford, Clarendon, 1988.
2. H. L¨ th, Surfaces and Interfaces of Solid u Materials, 3rd edition, Springer, Germany,
1998.
3. G. V. Hansson, R. I. G. Uhrberg, Photoelectron Spectroscopy of Surface States on Semiconductor Surfaces, Sur. Sci. Rep.
1988, 9, 197–292.
4. A. Many, Y. Goldstein, N. B. Grover, Semiconductor Surfaces, Amsterdam, NorthHolland, 1965.
5. D. V. Lang, C. H. Henry, Nonradiative Recombination at Deep Levels in GaAs and GaP by Lattice-Relaxation Multiphonon Emission, Phys. Rev. Lett. 1975, 35,
1525–1528.
6. H. Ishii et al., Energy Level Alignment and Interfacial Electronic Structures at
Organic/Metal and Organic/Organic Interfaces, Adv. Mater. 1999, 11(8), 605–625.
7. A. M. Green, W. E. Spicer, Do We Need a
New Methodology for GaAS Passivation?
J.Vac. Sci. Technol. A 1993, 11, 1061–1069.
8. M. Bruening et al., Simultaneous Control of Surface Potential and Wetting of
Solids with Chemisorbed Multifunctional
Ligands, J. Am. Chem. Soc. 1997, 119,
5720–5728.
9. S. D. Evans et al., Self-Assembled Monolayers of Alkanethiols Containing a Polar
Aromatic Group: Effect of the Dipole Position on Molecular Packing, Orientation, and
Surface Wetting Properties, J. Am. Chem.
Soc. 1991, 113, 4121–4131.

10. S. D. Evans, A. Ulman, Surface Potential
Studies of Alkyl-Thiol Monolayers Adsorbed on Gold, Chem. Phys. Lett. 1990,
170(5, 6), 462–466.
11. I. H. Campbell et al., Controlling Schottky
Energy Barriers in Organic Electronic
Devices using Self-assembled Monolayers,
Phys. Rev. B 1996, 54, R1432–R14 324.
12. M. Bruening et al., Polar Ligand Adsorption
Controls Semiconductor Surface Potentials,
J. Am. Chem. Soc. 1994, 116, 2972–2977.
13. M. Bruening et al., Controlling The Work
Function of CdSe by Chemisorption of Benzoic Acid Derivatives and Chemical Etching,
J. Phys. Chem. 1995, 99, 8368–8373.
14. S. Bastide et al., Controlling the Work
Function of GaAs by Chemisorption of
Benzoic Acid Derivatives, J. Phys. Chem.,
1997, 101, 2678–2684.
15. D. M. Taylor, O. N. Oliveira, H. Morgan,
Models for Interpreting Surface Potential
Measurements and their Application to
Phospholipid Monolayers, J. Coll. Int. Sci.
1990, 139(2), 508–518.
16. J. F. Dorsten, J. E. Maslar, P. W. Bohn, NearSurface Electronic Structure in GaAs (100)
Modiﬁed with Self-assembled Monolayers of Octadecylthiol, Appl. Phys. Lett. 1995, 66,
1755–1757.
17. S. M. Sze, Physics of Semiconductor Devices,
7th edition, 2nd Chapter, John Wiley &
Sons, USA, 1981.
18. W. M¨ nch, Semiconductor Surfaces and Ino terfaces, 2nd edition, in Surface Science
(Ed.: E Gerhard), Springer, Berlin, 1995, pp. 274–284.
19. F. Seker et al., Surface Chemistry of Prototypical Bulk II–VI and III–V Semiconductors and Implications for Chemical Sensing,
Chem. Rev. 2000, 100, 2505–2536.
20. J.-F. Fan, H. Oigawa, Y. Nannichi, The
Effect of (NH4 )2 S Treatment on the
Interface Characteristics of GaAs MIS
Structures, Jap. J. Appl. Phys. 1988, 27,
L1331–L1333.
21. J.-F. Fan, H. Oigawa, Y. Nannichi, MetalDependent Schottky Height with the
(NH4 )2 Sx Treated GaAs. Jap. J. Appl. Phys.
1988, 27(11), L2125–L2127.
22. D. Liu et al., Deep level transient spectroscopy study of GaAs surface states treated with inorganic sulﬁdes, Appl. Phys. Lett.
1988, 53, 1059–1061.

147

148

2 Experimental Techniques
23. H. Hasegawa et al., Control of Fermi Level
Pinning and Recombination Processes at
GaAs Surfaces by Chemical and Photochemical Treatments, J. Vac. Sci. Technol.
B 1988, 6, 1184–1192.
24. S. Hildebrandt et al., Optical Study of BandBending and Interface Recombination at Sb, S, and Se Covered Gallium
Arsenide Surfaces, Appl. Surf. Sci. 1993, 63,
153–157.
25. R. S. Besser, C. R. Helms, Effect of Sodium
Sulﬁde Treatment on Band Bending in
GaAs, Appl. Phys. Lett. 1988, 52, 1707–1709.
26. H. Shen, W. Zhou, J. Pamulapati, Photoreﬂectance Study of H2 S Plasma-Passivated
GaAs Surfaces, Appl. Phys. Lett. 1999, 74,
1430–1432.
27. V. L. Berkovits et al., Chemical and Photochemical Processes in Sulﬁde Passivation of
GaAs(100), In Situ Optical Study and Photoemission Analysis, J. Vac. Sci. Technol. A
1998, 16, 2528–2538.
28. M. Stutzmann, C. P. Herrero, Hydrogen
Passivation of Shallow Acceptors in Silicon,
Phys. Scripta. 1989, T25, 276–282.
29. M. J. Bedzyk, W. M. Gibson, J. A. Golovchenko, X-ray Standing Wave Analysis for
Bromine Chemisorbed on Silicon, J. Vac.
Sci. Technol. 1982, 20, 634–637.
30. R. I. G. Uhrberg et al., Electronic Structure,
Atomic Structure, and the Passivated Nature of the Arsenic-Terminated Si (111)
Surface, Phys. Rev. B 1987, 35, 3945–3951.
31. B. A. Kuruvilla et al., Passivation of GaAs
(100) Using Selenium Sulﬁde, J. Appl. Phys.
1993, 73, 4384–4387.
32. J. Sun et al., Chemical Bonding and
Electronic Properties of SeS2 -Treated
GaAs(100), J. Appl. Phys. 1999, 85,
969–977.
33. E. Yablonovitch et al., Band-Bending, Fermi
Level Pinning and Surface Fixed Charge on
Chemically Prepared GaAs Surfaces, Appl.
Phys. Lett. 1989, 54, 555–557.
34. E. Yablonovitch et al., Nearly Ideal Electronic Properties of Sulﬁde-Coated GaAs
Surfaces, Appl. Phys. Lett. 1987, 51(6),
439–441.
35. C. J. Sandroff et al., Dramatic Enhancement in the Gain of a GaAs/AlGaAs
Heterostructure Bipolar Transistor by Surface Chemical Passivation, Appl. Phys. Lett.
1987, 51(1), 33–35.

36. C. J. Sandroff et al., Electronic Passivation of GaAs Surfaces Through the Formation of Arsenic – Sulfur Bonds, Appl. Phys. Lett.
1989, 54(4), 362–364.
37. J. S. Herman, F. L. Terry, Hydrogen Sulﬁde
Plasma Passivation of Gallium Arsenide,
Appl. Phys. Lett. 1992, 60, 716–717.
38. B. A. Parkinson, A. Heller, B. Miller, Effects of Cations on the Performance of the Photoanode in the N-GaAs/K2 Se-K2 Se2 KOH/C Semiconductor Liquid Junction
Solar Cell, J. Electrochem. Soc. 1979, 126(6),
954–960.
39. P. Viktorovitch et al., Improved Electronic
Properties of GaAs Surfaces Stabilized with
Phosphorous, Appl. Phys. Lett. 1991, 58(21),
2387–2389.
40. H. H. Lee, R. J. Racicot, S. H. Lee, Surface
Passivation of GaAs, Appl. Phys. Lett. 1989,
54(8), 724–726.
41. L. Ferrari et al., Optical and Spectroscopic
Characterization of GaAs Passivated Surfaces, Surf. Sci. 1995, 331–333, 447–452.
42. D. Troost et al., Pinning of the Fermi level
Close to the Valence-Band Top by Chlorine
Adsorbed on Cleaved GaAs (110) Surfaces,
J. Vac. Sci. Technol. 1987, B5(4), 1119–1124.
43. H. J. Clemens, J. Von Wienskowski,
W. M¨ nch, On the Interaction of Cesium o with GaAs(110) and Ge(111) Surfaces: Work
Function Measurements and Adsorption
Site Model, Surf. Sci. 1978, 78, 648–666.
44. P. Skeath et al., Bonding of Antimony on
GaAs (110): A Prototypical System for
Adsorption of Column-V Elements on
III–V Compounds, Phys. Rev. B 1983, 27,
6246.
45. K. Jacobi, G. Steinert, W. Ranke, Iodine
Etching of the GaAs(-1-1-1) Surface Studied by LEED, AES, and Mass Spectroscopy, Surf.
Sci. 1976, 57, 571–579.
46. K. Vaccaro et al., Indium Phosphide Passivation Using Thin Layers of Cadmium Sulﬁde, Appl. Phys. Lett. 1995, 67(4), 527–529.
47. D. N. Bose, Y. Ramprakash, S. Basu, Characterization of n-InP Surfaces Before and
After Surface Modiﬁcation, Mater. Lett.
1989, 88(9), 364–368.
48. D. N. Bose, J. N. Roy, S. Basu, Improved
Schottky Barrier on n-InP by Surface Modiﬁcation, Mater. Lett. 1984, 2(5B), 455–457.
49. A. J. Nelson, S. P. Frigo, R. A. Rosenberg,
Chemistry and Electronic Structure of the

2.3 Grafting Molecular Properties onto Semiconductor Surfaces

50.

51.

52.

53.

54.

55.

56.

57.

58.

59.

60.

61.

62.

H2 Plasma Passivated Surface of CdTe,
J. Appl. Phys. 1994, 73(3), 1632–1637.
D. N. Bose, S. Basu, K. C. Mandal, Characterization of Chemically Modiﬁed CdTe
Surfaces, Thin Solid Films 1988, 164, 13–19.
D. N. Bose et al., XPS Investigation of
CdTe Surfaces: Effect of Ru Modiﬁcation,
Semicon. Technol. 1989, 4, 866–870.
J. L. Shaw et al., Chemically Controlled
Deep Level Formation and Band-Bending at Metal-CdTe Interfaces, Appl. Phys. Lett.
1988, 53(18), 1723–1725.
D. Cahen, R. Nouﬁ, Surface Passivation of
Polycrystalline, Chalcogenide Based Photovoltaic Cells, Solar Cells 1991, 30(1–4),
53–59.
Y. Nemirovsky, D. Rosenfeld, Surface Passivation and 1/f Noise Phenomena in
HgCdTe Photodiodes, J. Vac. Sci. Technol.
1990, A8(2), 1159–1167.
Y. F. Chen et al., Hydrogen Passivation in
Cd1−x Znx Te Studied by Photoluminescence, Appl. Phys. Lett. 1991, 58(5), 493–495.
Y. F. Chen, W. S. Chen, Inﬂuence of Hydrogen Passivation on the Infrared Spectra of Hg0.8 Cd0.2 Te, Appl. Phys. Lett. 1991,
59(6), 703–705.
D. Cahen, R. Nouﬁ, Defect Chemical Explanation for the Effect of Air Anneal on
CdS/CuInSe2 Solar Cell Performance, Appl.
Phys. Lett. 1989, 54(6), 558–560.
S. R. Lunt et al., Chemical Studies of the
Passivation of GaAs Surface Recombination
Using Sulﬁdes and Thiols, J. Appl. Phys.
1991, 70(12), 7449–7467.
C. J. Sandorff et al., Enhanced Electronic
Properties of GaAs Surfaces Chemically
Passivated by Selenium Reactions, J. Appl.
Phys. 1990, 67(1), 586–588.
S. A. Chambers, V. S. Sundaram, Structure, Chemistry, and Band-Bending at SePassivated GaAs(001) Surfaces, Appl. Phys.
Lett. 1990, 57, 2342–2344.
W. C. Simpson, J. A. Yarmoff, Fundamental studies of halogen reactions with III–V semiconductor surfaces, Ann. Rev. Phys.
Chem. 1996, 47, 527–554.
Y. Liu, A. J. Komrowski, A. C. Kummel,
Side–selective reaction of Br2 with second layer Ga atoms on the As-rich GaAs(001)
2 × 4 surface, Phys. Rev. Lett. 1998, 81,
413–416.

63. R. Maoz, J. Sagiv, On the Formation and
Structure of Self-Assembling Monolayers,
J. Coll. Int. Sci. 1984, 100(2), 465–496.
64. A. Ulman, An Introduction to Ultrathin
Organic Films, Academic Press, New York,
1991.
65. R. Cohen et al., Molecular Electronic Tuning of Si Surfaces, Chem. Phys. Lett. 1997,
279, 270–274.
66. B. Sweryda et al., A Comparison of Porous
Silicon and Silicon Nanocrystallite Photoluminescence Quenching with Amines,
J. Phys. Chem. 1996, 100(32), 13 776–13 780, and reference therein.
67. J. M. Lauerhaas, M. J. Sailor, Chemical
Modiﬁcation of the Photoluminescence
Quenching of Porous Silicon, Science 1993,
261, 1567–1568.
68. S. R. Lunt, P. G. Santangelo, N. S. Lewis,
Passivation of GaAs Surface Recombination with Organic Thiols, J. Vac. Sci. Technol. B
1991, 9(4), 2333–2336.
69. J. K. M. Chun et al., Proton Gated Emission from Porous Silicon, J. Am. Chem. Soc.
1993, 115, 3024–3025.
70. R. R. Chandler et al., Steric Considerations in the Amine-Induced Quenching of Luminescent Porous Silicon, J. Phys. Chem. 1995,
99(21), 8851–8855.
71. J. M. Lauerhaas et al., Reversible Luminescence Quenching of Porous Si by Solvents, J. Amer. Chem. Soc. 1992, 114(5),
1911–1912.
72. A. Vilan, Chemical Modiﬁcation of the electronic Properties of GaAs Surface, M.Sc thesis, Weizmann Institute of Science,
Rehovot, 1996.
73. D. G. Wu et al., Novel NO biosensor based on surface derivatization of GaAs by
‘‘hinged’’ Fe porphyrin, Angew. Chem. Int.
Ed. 2000, 39, 4496–4500.
74. R. Cohen et al., Frontier Orbital Model of
Semiconductor Surface Passivation: Dicarboxylic Acids on n- and p-GaAs, Adv. Mater.
2000, 12, 33–37.
75. R. Cohen et al., Controlling Electronic Properties of CdTe by Adsorption of Dicarboxylic
Acid Derivatives: Relating Molecular Parameters to Band-Bending and Electron
Afﬁnity Changes, Adv. Mater. 1997, 9,
746–749.
76. R. Cohen et al., Molecular Control over
Semiconductor Surface Electronic Properties: Dicarboxylic Acids on CdSe, CdTe,

149

150

2 Experimental Techniques

77.

78.

79.

80.

81.

82.

83.

84.

85.

86.

87.

88.

89.

GaAs and InP, J. Am. Chem. Soc. 1999, 121,
10 545–10 553.
A. Vilan, A. Shanzer, D. Cahen, Molecular
Control over Au/GaAs Diodes, Nature 2000,
404, 166–168.
C. D. Bain, A New Class of Self-Assembled
Monolayers: Organic Thiols on Gallium
Arsenide, Adv. Mater. 1992, 4(9), 591–594.
W. C. Sheen et al., A New Class of Organized Self-Assembled Monolayers: Alkane
Thiols on GaAs (100), J. Am. Chem. Soc.
1992, 114, 1514–1515.
O. S. Nakagawa et al., GaAs Interfaces with
Octadecyl Thiol Self-Assembled Monolayer:
Structural and Electrical Properties, Jap.
J. Appl. Phys. 1991, 30, 3759–376.
K. Asai et al., Electronic Passivation of
GaAs Surfaces by Electrodeposition of
Organic Molecules Containing Reactive
Sulfur, J. Appl. Phys. 1995, 77, 1582–1586.
E. D. Lu et al., A Sulfur Passivation of
GaAs Surface by an Organic Molecular,
CH3 CSNH2 Treatment, Appl. Phys. Lett.
1996, 69, 2282–2284.
M. Tabib-Azar, A. S. Dewa, W. H. Ko,
Langmuir-Blodgett Film Passivation of
Unpinned n-Type Gallium Arsenide
Surfaces, Appl. Phys. Lett. 1988, 52,
206–208.
V. J. Rao, V. S. Kulkarni, Interface Barrier
Height in n-GaAs(100)/Langmuir-Blodgett
Film Structures, Thin Solid Films 1991, 198,
357–362.
N. K. Singh et al., Etching Reactions of
C2 H5 I on GaAs(100), Sur. Sci. 1998, 409,
272–282.
P. J. Lasky et al., The Adsorption and
Thermal Reaction of Dimethylcadmim,
Dimethylzinc and Trimethylgallium on
GaAs(110), Surf. Sci. 1996, 364, 312–324.
H. T. Lam, N. Venkateswaran, J. M. Vohs,
The Reactions of Diethylzinc on GalliumRich and Arsenic-Rich Reconstructions of
GaAs(100), Surf. Sci. 1998, 401, 34–46.
T. Hou et al., Passivation of GaAs (100) with an Adhesion Promoting Self-Assembled Monolayer, Chem. Mater. 1997, 9,
3181–3186.
Y. Gu, D. H. Waldeck, Electron Tunneling at the Semiconductor-Insulator-Electrolyte
Interface. Photocurrent Studies of the n-InP-Alkanethiol-Ferrocyanide System,
J. Phys. Chem. B 1998, 102, 9015–9028.

90. M. Sturzenegger, N. S. Lewis, An X-Ray
Photoelectron Spectroscopic and Chemical
Reactivity Study of Routes to Functionalization of Etched InP Surfaces, J. Am. Chem.
Soc. 1996, 118, 3045–3046.
91. J. W. Thackeray et al., Interactions of Diethyldithiocarbamate with n-type Cadmium
Sulﬁde and Cadmium Selenide: Efﬁcient Photoelectrochemical Oxidation to the
Disulﬁde and Flat-Band Potential of the
Semiconductor as a Function of Adsorbate
Concentration, J. Am. Chem. Soc. 1986, 108,
3570–3577.
92. M. J. Natan, J. W. Thackeray, M. S. Wrighton, Interaction of Thiols with n-Type
Cadmium Sulﬁde and n-Type Cadmium
Selenide in Aqueous Solutions: Adsorption of Thiolate Anion and Efﬁcient Photoelectrochemical Oxidation to Disulﬁdes, J. Phys.
Chem. 1986, 90, 4089–4098.
93. J. J. Hickman, M. S. Wrighton, Face-Speciﬁc Interactions of Anionic Sulfur Donors with Oriented Crystals of (0001) CdX (X =
Se, S) and Correlation with Electrochemical
Properties, J. Am. Chem. Soc. 1991, 113,
4440–4448.
94. T. Uchihara et al., Effect of Ethylenediaminetetraacetic Acid on the Photocatalytic
Activities and Flat-Band Potentials of Cadmium Sulﬁde and Cadmium Selenide,
J. Phys. Chem. 1990, 94, 415–418.
95. J. K. Lorenz, T. F. Kuech, A. B. Ellis, Cadmium Selenide Photoluminescence as a
Probe for the Surface Adsorption of Dialkyl Chalcogenides, Langmuir 1998, 14,
1680–1683.
96. G. J. Meyer et al., Semiconductor-Oleﬁn
Adducts. Photoluminescent Properties of
Cadmium Sulﬁde and Cadmium Selenide in the Presence of Butenes, J. Am. Chem.
Soc. 1989, 111, 5146–5148.
97. C. J. Murphy et al., PhotoluminescenceBased Correlation of Semiconductor Electric Field Thickness with Adsorbate Hammett Substituent Constants. Adsorption of Aniline Derivatives onto Cadmium
Selenide, J. Am. Chem. Soc. 1990, 112,
8344–8348.
98. G. C. Lisensky et al., Electro-Optical Evidence for Chelate Effect at Semiconductor
Surfaces, Science 1990, 248, 840–843.
99. G. J. Meyer, G. C. Lisensky, A. B. Ellis, Evidence for Adduct Formation at the

2.3 Grafting Molecular Properties onto Semiconductor Surfaces

100.

101.

102.

103.

104.

105.

106.

107.

108.

Semiconductor-Gas Interface. Photoluminescent Properties of Cadmium Selenide in the Presence of Amines, J. Am. Chem. Soc.
1988, 110, 4914–4918.
D. R. Neu, J. A. Olson, A. B. Ellis, Photoluminescence as a probe of the Adsorption of Gaseous Boranes onto the Surface of
Cadmium Selenide Crystals, J. Phys. Chem.
1993, 97, 5713–5716.
J. Z. Zhang, M. J. Geselbracht, A. B. Ellis,
Binding of Fullerenes to Cadmium Sulﬁde and Cadmium Selenide Surfaces, Photoluminescence as a Probe of Strong, Lewis
Acidity-Driven, Surface Adduct Formation,
J. Am. Chem. Soc. 1993, 115, 7789–7793.
K. D. Kepler et al., Surface-Bound Carbonyl
Compounds as Lewis Acids. Photoluminescence as a Probe for the Binding of Ketones and Aldehydes to Cadmium Sulﬁde and
Cadmium Selenide Surfaces, J. Phys. Chem.
1995, 99(43), 16 011–16 017.
J. Z. Zhang, A. B. Ellis, Adsorption of TCNQ
Derivatives onto the Surface of Cadmium
Selenide Single Crystals. Quenching of
Semiconductor Photoluminescence by a
Family of Strong-Acids, J. Phys. Chem. 1992,
96, 2700–2704.
R. J. Brainard et al., Modulation of Cadmium Selenide Photoluminescence Intensity by Adsorption of Silapentanes and
Chlorinated Silanes, J. Phys. Chem. B 1997,
101, 11 180–11 184.
S. Licht, D. Peramunage, Rational Electrolyte Modiﬁcation of n-CdSe/([KFe(CN)6 ]3/2-) Photoelectrochemistry, J. Electrochem.
Soc. 1992, 139, L23–L26.
J. K. Lorenz, A. B. Ellis, Surfactant-Semiconductor Interfaces: Perturbation of the
Photoluminescence of Bulk Cadmium Selenide by Adsorption of Tri-n-octylphosphine Oxide as a Probe of Solution Aggregation with Relevance to Nanocrystal
Stabilization, J. Am. Chem. Soc. 1998, 120,
10 970–10 975.
J. Kr¨ ger, U. Bach, M. Gr¨tzel, Modiﬁcau a tion of TiO2 Heterojunctions with Benzoic Acid Derivatives in Hybrid Molecular
Solid-State Devices, Adv. Mater. 2000, 12,
447–451.
F. Nuesch et al., Derivatized Electrodes in the Construction of Organic Light Emitting
Diodes, Adv. Mater. 1997, 9, 222–225.

109. L. Zuppiroli et al., Self-Assembled Monolayers as Interfaces for Organic Optoelectronic Devices, Eur. Phys. J. B 1999, 11,
505–512.
110. F. Nuesch et al., Chemical Potential Shifts at Organic Device Electrodes Induced by
Grafted Monolayers, Chem. Phys. Lett. 1998,
288, 861–867.
111. D. Gal et al., Engineering the Interface
Energetics of Solar Cells by Grafting
Molecular Properties onto Semiconductors,
Proc. Indian Acad. Sci. 1997, 109, 487–496.
112. A. Vilan et al., Real-Time Electronic Monitoring of Adsorption Kinetics: Evidence for
Two-site Adsorption Mechanism of Dicarboxylic Acids on GaAs(100), J. Phys. Chem.
B 1998, 102, 3307–3309.
113. Y. Kanemitsu, Slow Decay Dynamics of
Visible Luminescence in Porous Silicon:
Hopping of Carriers Conﬁned on a Shell
Region in Nanometer-size Si Crystallities,
Phys. Rev. 1993, B48, 12 357.
114. J. K. Lorenz, T. F. Kuech, A. B. Ellis, Cadmium Selenide Photoluminescence as a
Probe for the Surface Adsorption of Dialkyl Chalcogenides, Langmuir 1998, 14,
1680–1683, and reference therein.
115. W. J. Dollard, M. L. Shumaker, D. H. Waldeck, Time-Resolved Studies of Charge
Carrier Relaxation in Chemically Modiﬁed
Semiconductor Electrodes: n-CdSe/Silane
Interfaces, J. Phys. Chem. 1993, 97,
4141–4148.
116. A. L. McClellan, Tables of Experimental
Dipole Moments, Freeman, San Francisco,
1963.
117. J. E. Huheey, Inorganic Chemistry, 3rd edition, Harper & Row, New-York, 1983, pp. 312–325.
118. G. L. Miessler, D. L. Tarr, Inorganic Chemistry, Prentice Hall, Englewood Cliffs, 1991.
119. R. Hoffman, Solids and Surfaces: A Chemist’s
View of Bonding in Extended Structures, VCH
Publishers, New York, 1988, pp. 68–71.
120. T. Vondrak, C. J. Cramer, X.-Y. Zhu, The
Nature of Electronic Contact in SelfAssembled Monolayers for Molecular Electronics: Evidence for Strong Coupling,
J. Phys. Chem. B 1999, 103, 8915–8919.
121. M. C. Petty, G. G. Roberts, CdTe/LangmuirFilm MIS Structures, Electron. Lett. 1979,
15(12), 335–336.
122. I. M. Dharmadasa, G. G. Roberts, M. C.
Petty, Cadmium Telluride/Langmuir Film

151

152

2 Experimental Techniques

123.

124.

125.

126.
127.

128.

129.

Photovoltaic Structures, Electron. Lett. 1980,
16(6), 201–202.
I. H. Campbell et al., Controlling Charge
Injection in Organic Electronic Devices
Using Self-Assembled Monolayers, Appl.
Phys. Lett. 1997, 71, 3528–3530.
H. Sirringhaus, N. Tessler, R. H. Friend,
Integrated Optoelectronic Devices Based on
Conjugated Polymers, Science 1998, 280,
1741–1744.
L. Kronik et al., Surface States and Photovoltaic Effects in CdSe Quantum Dot Films,
J. Electrochem. Soc. 1998, 145, 1748–1755.
A. Hagfeldt, M. Gr¨tzel, Molecular Photoa voltaics, Acc. Chem. Res. 2000, 33, 269–277.
D. Cahen et al., Nature of Photovoltaic Action in Dye-Sensitized Solar Cells, J. Phys.
Chem. B 2000, 104, 2053–2059.
Y. Selzer, D. Cahen, Fine-tuning of
Au/SiO2 /Si diodes by varying interfacial dipoles using molecular monolayers, Adv.
Mater. 2001, 13, 508–511.
D. G. Wu et al., Direct Detection of Low
Concentration NO in Physiological Solu-

130.

131.

132.

133.

134.

tions by a New GaAs-Based Sensor, Chemistry Eur. J. 2001, 7, 1743–1749.
D. G. Wu et al., Tuning of Au/n-GaAs
Diodes with Highly Conjugated Molecules,
J. Phys. Chem. B 2001, in press.
G. Ashkenasy et al.,
Assemblies of
‘‘Hinged’’ Iron-Porphyrins as Potential
Oxygen Sensors, J. Amer. Chem. Soc. 2000,
122, 1116–1122.
A. Ivanisevic et al., Linker-enhanced Binding of Metalloporphyrins to CdSe and
Implications for O2 detection, Langmuir
2000, 16, 7852–7858.
T. Vondrak et al., Interfacial Electronic
Structure in Thiolate Self-assembled Monolayers; Implications for Molecular Electronics, J. Amer. Chem. Soc. 2000, 122,
4700–4707.
M. Carrara, F. N¨ esch, L. Zuppiroli, Caru boxylic acid anchoring groups for the construction of self-assembled monolayers on organic device electrodes, Synth. Metals
2001, 121, 1633–1634.

2.4 Capacitance, Luminescence, and Related Optical Techniques

2.4

Capacitance, Luminescence, and Related
Optical Techniques
Yoshihiro Nakato
Osaka University, Osaka, Japan
2.4.1

Capacitance
Space Charge Layer at a
Semiconductor–Electrolyte Interface
An electrical double layer is formed at a semiconductor–electrolyte interface, similar to a metal–electrolyte interface [1–3].
The difference, however, is in the charge distribution between the two interfaces. At a metal–electrolyte interface, the charge on the metal side is localized just at the metal surface, whereas, at a semiconductor–electrolyte interface, the charge on the semiconductor side is distributed deep in the interior of the semiconductor, forming a wide space charge layer. In concentrated electrolyte solutions (∼0.5 M and higher), the charge on the electrolyte side is localized at the (outer) Helmholtz layer for interfaces with either metals or semiconductors.
Figure 1 schematically illustrates (1) the charge distribution, (2) the charge-density distribution, (3) the electric ﬁeld distribution, (4) the potential distribution, and
(5) the band bending at the semiconductor–electrolyte interface, under an ideal condition that no surface charge nor surface dipole is present at the semiconductor.
The semiconductor is assumed to be ntype, but a similar diagram can be drawn for a p -type semiconductor with the sign of electric charges inverted. The band bending in the semiconductor is simply caused by the potential distribution in it.
The n-type semiconductor has an electron donor in the crystal, doped as an impurity, such as phosphor atoms in n-Si,
2.4.1.1

whereas the p -type semiconductor has an electron acceptor such as boron atoms in p -Si. The space charge in the semiconductor is composed of an ionized electron donor or acceptor. The space charge layer of a ﬁnite thickness is formed in the semiconductor when the density of the electron donor, ND , or the density of the electron acceptor, NA , is low, in a range from 1014 to
1018 cm−3 . The principle of electrical neutrality at the electrical double layer leads to an extended distribution of charges deep toward the interior of the semiconductor.
The situation is rather similar to the formation of an extended charge distribution, called a Gouy layer, in the electrolyte in cases where the electrolyte concentration is low. The difference is that the space charges in the Gouy layer are composed of electrolyte ions and are mobile, whereas those in the semiconductor are composed of ionized donors or acceptors and therefore are spatially ﬁxed and immobile.
Differential Capacitance of a
Semiconductor–Electrolyte Interface
The potential distribution in the space charge layer of a semiconductor can be given by solving the Poisson equation for a given charge distribution [1–3]. For a semiconductor–electrolyte interface such as the one shown in Fig. 1, if the density of an electron donor, ND , is constant throughout the semiconductor, the potential, φ(x), at a distance, x , from the surface is given as follows [1, 2]:
2.4.1.2

φ(x) =

qND ε0 εs

1
W x − x 2 + φ(0)
2

(0 ≤ x ≤ W )

(1)

where q is the elementary charge, ε0 the permittivity of vacuum, εs the dielectric constant of semiconductor, φ (0) the potential at the surface (x = 0), and W the

153

2 Experimental Techniques
−
−
−
−
−
Ions at the
Helmholtz layer −
−
−

+
+

+
+

Charge density

0

0

+

Ionized donor
+
n -type semiconductor

n
Region
X

0

0

Fig. 1 Schematic illustration of
(a) charge distribution,
(b) charge-density distribution,
(c) electric ﬁeld distribution,
(d) potential distribution, and (e) band bending at an n-type semiconductor–electrolyte interface under an ideal condition that neither surface charge nor surface dipole is present at the semiconductor.

+

Depletion region Electric field

(b)

+

Electrolyte

(a)

W

X

Potential

(c)

X

0

(d)
− eUfb
Energy

154

EC
EF = − eU

Eredox

EV

(e)

width of the space charge layer. The φ (0) and W are given by φ(0) =

qND W δ ε0 εs

2ε0 εs
W=
qND

given by
Qs = qND W = [(2qε0 εs ND )
× (U − Ufb − kT /q)]1/2

(2) kT U − Ufb − q 1/2

(3)

where δ is the thickness of the (outer)
Helmholtz layer, U is the electrode potential, and Ufb the ﬂat band potential.
The space charge, Qs , per unit area is

(4)

Thus, the differential capacitance of the space charge layer, Cd , per unit area is given as follows:
Cd =

∂Qs
2
=
∂U
qε0 εs ND
×

U − Ufb −

kT q −1/2

(5)

2.4 Capacitance, Luminescence, and Related Optical Techniques

This equation can be rewritten as follows:
2
1
=
2 qε0 εs ND
Cd

U − Ufb −

kT q (6)

2
The plot of 1/Cd measured against U is called the Mott–Schottky plot, which can be used to determine Ufb and ND (or NA ), as discussed later.
It has been assumed thus far that no surface charge nor surface dipole is present at the semiconductor. In general cases, both surface charges and surface dipoles are present in the semiconductor owing to adsorption equilibria of various ions between the electrolyte and the semiconductor surface as well as formation of polar bonds at the semiconductor surface. Such surface

charges and surface dipoles cause a shift of the semiconductor band positions at the surface, as shown schematically in Fig. 2.
The shift is expressed as a change in φ (0) or
Ufb in the foregoing equations. However, the preceding equations themselves can be applied to such cases with the changed φ (0) or Ufb .
Measurement of Differential
Capacitance
The capacitance of the semiconductor–electrolyte interface can be measured by use of a semiconductor electrode, in which the front side of the semiconductor is in contact with the electrolyte and the rear side is electrically connected with a metallic leading wire via an ohmic contact.
2.4.1.3

Ions at the
Helmholtz layer
Surface charge

Charge density

Electrolyte
(a)

−
−
+
+− +
+
−
+
+− +
−
+
Ionized donor
−+
+
+−
n -type semiconductor

Depletion region 0

n
Region

X

0

Potential

(b)

0

X

(c)

Schematic illustration of
(a) charge distribution,
(b) charge-density distribution,
(c) potential distribution, and (d) band bending at the semiconductor–electrolyte interface under a condition that negative surface charges are present.

Fig. 2

Energy

− eUfb

(d)

Eredox

EC
EF = − eU
EV

155

156

2 Experimental Techniques

Impedance analyzer

Potentiostat

Recorder or computer
Function
generator

Fig. 3 Experimental setup for measurements of capacitance at the semiconductor–electrolyte interface. W.E.: working electrode (semiconductor electrode), C.E.: counter electrode, and R.E.: reference electrode. R.E.

C.E.

W.E.

The experimental setup for capacitance measurements is schematically shown in
Fig. 3. The electrode potential U is regulated with a potentiostat. The differential capacitance is measured by superimposing an AC voltage with a small amplitude of about 10 mV and a frequency of a few Hz to 1 MHz on the electrode potential. One can use a commercial impedance analyzer to measure the differential capacitance, together with a personal computer to analyze obtained data automatically.
Recently, scanning capacitance microscopy (SCM) using a ﬁne tip has been developed to investigate two-dimensional dopant proﬁling of semiconductor surfaces [4].
Mott–Schottky Plots and Flat Band
Potentials
2
Equation 6 indicates that a plot of 1/Cd against U gives a straight line with a slope of (2/qε0 εs ND ), which is termed the
Mott–Schottky plot, as mentioned earlier.
The extrapolation of the straight line to
2
1/Cd = 0 gives (Ufb + kT /q ). Therefore, the plot can be used to determine the ﬂat band potential Ufb . The donor density
ND (or the acceptor density NA ) can also be determined from the slopes of the plots. Figure 4 shows examples of Mott–Schottky plots, obtained for n-Si(111) and n-Si(100) electrodes in 7.1 M
2.4.1.4

hydrogen iodide [5]. It should be noted that the slope of the straight line depends not only on ND (or NA ) and εs but also on the true surface area (or surface roughness) of the semiconductor electrode.
The Ufb is one of the most important quantities for semiconductor electrodes because it determines the band edge positions at the semiconductor–electrolyte interface, which in turn, determine the energies of conduction-band electrons and valence-band holes reacting with the electrolyte solution. It is known that Ufb for most semiconductors, such as n- and p -GaAs, GaP, InP, n-ZnO, n-TiO2 , and n-SnO2 , in aqueous electrolytes is solely determined by the solution pH and shifts in proportion to pH with a ratio of
−0.059 V/pH [1, 2]. This is explained by an adsorption equilibrium for H+ or OH− at the semiconductor–electrolyte interface, for example,
−−
−
Ss -OH + Haq + −− − Ss -OH2 +

(7)

where Ss -OH refers to surface OH group at the semiconductor.
The Ufb for n- and p -Si [6] and metal calcogenide semiconductors such as nCdS, n-CdSe, and CdTe [2, 7] does not obey the foregoing rule, remaining nearly constant in a range of pH lower than about
6 for Si and about 10 for n-CdS. This is most probably because the semiconductor

2.4 Capacitance, Luminescence, and Related Optical Techniques
Examples of
Mott–Schottky plots obtained for n-Si(111) and n-Si(100) electrodes in aqueous 7.1 M HI.

Fig. 4

6

(1/C2) × 1017
[F−2cm4]

5

n- Si (111)
10.0 ∼ 15.0 Ωcm

4
3
2

n- Si (100)
0.8 ∼ 1.16 Ωcm

1
0

−0.8 −0.4 0

0.4 0.8 1.2 1.6

Potential versus SCE
[V]

surface has no OH group in this pH range and no adsorption equilibrium for H+ or
OH− is attained. For metal calcogenide semiconductors such as n-CdS, the Ufb shifts by adsorption of HS− and Cd2+ ions. It is also known that Ufb for some semiconductor electrodes shifts by a change in the surface termination bond [8,
9], as well as electrode illumination [10,
11] and the presence of a redox couple [10,
12]. The Ufb in nonaqueous electrolytes has been reported for some semiconductor electrodes [13, 14].
2.4.2

Luminescence from Semiconductor
Electrodes
Photoluminescence and
Electroluminescence
Illumination of a semiconductor electrode generates excited electrons in the conduction band and holes in the valence band. Some of them recombine with each other radiatively, resulting in emission of luminescence called photoluminescence
(PL). The radiative recombination may occur directly between the conduction and valence bands (inter-band transition) or via certain impurity or defect levels within the band gap at which either electrons or holes, or both are trapped, as schematically illustrated in Fig. 5(a). Also,
2.4.2.1

the radiative recombination may occur either in the semiconductor bulk or at the semiconductor surface (Fig. 5a).
On the other hand, EL is emitted when an n-type semiconductor electrode, for example, is negatively biased in an electrolyte solution containing a strong oxidant. Under this condition, holes are injected into the valence band by the oxidant, some of which recombine with electrons in the conduction band existing as the majority carrier in the n-type semiconductor, resulting in emission of luminescence called EL, as illustrated in Fig. 5(b). Similarly, EL is emitted when a p -type semiconductor electrode is positively biased in an electrolyte solution containing a strong reductant because some of the electrons, injected into the conduction band by the reductant, recombine with holes existing as the majority carrier in the p -type semiconductor. In both cases, EL may be emitted either via direct recombination between the conduction and valence bands or via an impurity level(s), and either via recombination in the semiconductor bulk or at the semiconductor surface. The situation is quite the same as that in PL emission (Fig. 5).
Figure 6 shows examples of PL and
EL spectra from some semiconductor electrodes [15]. It is to be noted, however, that such spectra strongly depend on

157

158

2 Experimental Techniques
Conduction band
−
−

−

−

−

−

Surface luminescence Illumination

Bulk luminescence
Surface
states
+

+

+
Valence band
(a)
Conduction band
−

−

−

−
Surface luminescence

Bulk luminescence
Surface
states
+

(b)

Valence band

+
+

e

Oxidant
•
•
(Fe3 +, HO , or SO4 −)

Mechanisms of emission of (a) PL and (b) electroluminescence
(EL) from a semiconductor electrode in an electrolyte solution.

Fig. 5

semiconductor materials used (that is, the kind and the amount of impurities and/or defects) and the kind of surface treatments made.
Measurements of Luminescence from Semiconductor Electrodes

2.4.2.2

2.4.2.2.1 DC Methods The luminescence
(PL or EL) from a semiconductor electrode can be measured by a simple conventional

DC method as shown in Fig. 7, although no illumination equipment is necessary for
EL measurements. The spectra in Fig. 6 are measured by this conventional DC method. In the DC method, the luminescence from a semiconductor electrode is usually collected in a normal direction by use of a lens and led to a slit of a monochromator, followed by detection with an appropriate photomultiplier. The electrode potential of

2.4 Capacitance, Luminescence, and Related Optical Techniques
Examples of PL and EL spectra of semiconductor electrodes. Eg : the band gap energy, hν : photon energy, λ: wavelength. Fig. 6

λ
[nm]
400

500

600

700 800

1000

n -GaP

Luminescence intensity

Eg

n -CdS
Eg

n -ZnO
Eg

n -TiO2
Eg

25

20

3.0

∼ v [103 cm−1]

2.5

2.0

15

10

1.5

hν
[eV]
Monochromater
Mirror
Hg lamp
Filter

Monochromater
Photomultiplier

Pt
Semiconductor
SCE
Potentiostat

Potential sweeper

V

X-Y recorder

Fig. 7 Experimental setup for simple DC measurements of PL and
EL from a semiconductor electrode.

159

160

2 Experimental Techniques

the semiconductor electrode is regulated with a potentiostat. By this way, one can measure PL or EL spectra at constant electrode potentials, and the luminescence intensity versus potential curves as compared with simultaneously measured current versus potential curves. The use of a personal computer is convenient for data processing and storage.
For PL measurements, the semiconductor electrode is illuminated by monochromatic light, using a high-pressure mercury lamp, a nitrogen laser, an argon laser, and so forth, as the light source. Much care should be taken to minimize a contribution of scattered stray light from the illumination light to luminescence spectra. The detection limit of weak PL is in most cases determined by how much the intensity of the scattered stray light is reduced. High-intensity illumination with high-quality monochromatic light by use of a laser is effective to detect weak PL, but very high intensity illumination generates very high density electrons and holes at the semiconductor surface, which may alter the essential mechanism of carrier dynamics in the semiconductor. Very high intensity illumination may also generate much heat (temperature increase) at the semiconductor surface.
For EL measurements, the detection limit of weak EL is mainly determined by the detection sensitivity of an experimental apparatus used because no illumination is needed in this case. One can use various high-sensitivity detection methods such as a lock-in ampliﬁer [16, 17], boxcar integration [17], and photon counting [17].
Care should, however, be taken when the time-averaging method is used because the luminescence intensity often changes with time even at a constant electrode potential, probably because the surface chemical structure of the semiconductor electrode

may change. Recently, space-resolved twodimensional detection of EL has been developed by use of a CCD camera [18].
Pulsed Techniques Time-resolved PL (and EL) measurements give much information on carrier dynamics in the semiconductor bulk and at the semiconductor surface. Figure 8 schematically illustrates a convenient method, called a time-correlated single photon counting method or a picosecond single photon timing method [19, 20, 21], by which weak PL can be measured at a high sensitivity with a time resolution on the order of picoseconds. A picosecond or femtosecond pulse laser with a pulse width of a few ps is used for excitation of the sample (semiconductor electrode). The luminescence from the sample is usually detected with a microchannel plate photomultiplier
(MCP-PMT) having a small transient time spread. A part of the excitation laser pulse is split from the main beam and detected with a photodiode for use as the starting pulse. The luminescence signal, detected with the MCP-PMT, is used as the stopping pulse. This means that the sample was excited at the time of the starting pulse and a luminescent photon was emitted at the time of the stopping pulse. The starting and stopping pulses are led to a time-to-ampliﬁer converter (TAC) to measure the time separation between them.
Repeated measurements give a histogram of the number of luminescent photons versus time, which should agree with a luminescence decay curve.
PL measurements with a higher time resolution of several tens of femtoseconds can be achieved by means of femtosecond up-conversion spectroscopy [22].
The experimental setup is illustrated in
Fig. 9. In this method, the luminescence intensity is measured by making use

2.4.2.2.2

2.4 Capacitance, Luminescence, and Related Optical Techniques
Multichannel
pulse height analyzer Photodiode
Laser
Start

A/D Conv.

TAC

Monochromater
Stop

Time difference
Sample

Voltage change

Counts

MCP-PMT

Time
Picosecond single-photon timing spectroscopy

Experimental setup for a time-correlated single-photon counting method or picosecond single-photon timing method, used for time-resolved luminescence measurements in a range of picoseconds.

Fig. 8

T
PM

Pulses from 800 nm,
Ti:sapphire laser.
82 MHz, 50 fs
10 nJ/pulse.
SHG beam:
∼ 400 nm

f = 150 mm
Spectrometer
BBO
SFG
Sample

BBO
SHG

400 nm,
< 0.5 nJ

Fundamental beam: ∼ 800 nm

Scanning optical delay,
850 ps time window

λ /2 plate
Dichroic
beamsplitter

Experimental setup for femtosecond up-conversion spectroscopy used for time-resolved luminescence measurements in a range of several tens femtoseconds.

Fig. 9

161

2 Experimental Techniques

of ‘‘gating’’ by another laser pulse. A femtosecond laser, such as a Ti-sapphire laser, is normally used as the light source for excitation. The emitted luminescence with a frequency ω1 is mixed with a gating laser pulse having a frequency ω2 in a nonlinear crystal such as BBO (β barium borate) to give an up-conversion light pulse with a frequency ω3 = ω1 + ω2 , which is measured with a photomultiplier.
The up-conversion light pulse with ω3 is generated only when the emitted luminescence and the gating laser pulse come to BBO simultaneously. Also, the intensity of the generated up-conversion light is in proportion to the product of the intensity of the emitted luminescence and that of the gating laser pulse. Thus, the luminescence decay curve in relative intensity is measured by giving various delays to the gating laser pulse.
Figure 10 shows examples of luminescence decays obtained by the femtosecond

Counts

162

up-conversion spectroscopy [23]. An aqueous or methanol suspension of nanocrystalline TiO2 particles, on which a cumarin dye is adsorbed, is used as the sample.
The luminescence is emitted from the adsorbed dye, and its decay represents the rate of electron transfer from excited adsorbed dye to the conduction band of TiO2 .
Bulk and Surface Luminescence
The radiative recombination of electrons and holes can occur either in the semiconductor bulk (bulk PL or EL) or at the semiconductor surface (surface PL or
EL), as mentioned earlier. In semiconductor electrochemistry, it is important to distinguish whether observed PL or
EL is emitted from the semiconductor bulk or surface because surface PL and
EL give more information on surface structures and processes at semiconductor electrodes, such as surface states, surface reaction intermediates, and so forth.
2.4.2.3

C343/MeOH
D-1421/MeOH

0
−0.5

0

0.5

( τ ET > 1.5 psec)

D-1421/water
C343/water

1000

( τ ET = 30 fsec)
( τ ET < 20 fsec)

1.0

1.5

Time
[ps]
Fig. 10 Examples of luminescence decays obtained by femtosecond up-conversion spectroscopy. The luminescence is emitted from a cumarin dye adsorbed on nanocrystalline
TiO2 particles suspended in water or methanol (MeOH). C343: cumarin 343 and D-1421:
7-diethylaminocumarin-3-carboxylic acid.

2.4 Capacitance, Luminescence, and Related Optical Techniques

Naturally, the intensity of surface PL is strongly inﬂuenced by the chemical composition of the electrolyte, in particular, by whether oxidants or reductants are included in it because electrons or holes trapped at surface states can easily react with oxidants or reductants in the electrolyte, resulting in luminescence quenching. The intensity of surface PL is also strongly dependent on surface pretreatments. However, it is to be noted that bulk PL is also inﬂuenced by these factors because photo-generated electrons and/or holes in the semiconductor bulk can diffuse to the surface and react with an oxidant or reductant in the electrolyte and also surface species (surface states) before they recombine. This leads to a decrease in the densities of electrons and holes in the semiconductor bulk and hence to a decrease in the intensity of bulk PL.
One effective way to distinguish experimentally the bulk and surface luminescence is to measure the electrode-potential

−1.0

Conduction band
−

U versus SCE/V

0.0

dependence of the luminescence intensity.
The surface luminescence is observed only under a forward bias with weak band bending or in the presence of an accumulation layer, namely, the surface luminescence is observed only when the surface densities of both electrons and holes are sufﬁciently high. On the other hand, the bulk luminescence is observed even in the presence of large band bending under which the surface density of the majority carrier is very low, if the penetration depth of incident illumination light is considerably larger than the width of the space charge layer.
The arguments made thus far show that both bulk and surface PL can be used to investigate surface processes at semiconductor electrodes. The measurements of bulk PL indicate that most of semiconductors emit bulk PL. They can be used to detect some surface processes and reactions [24]. The measurement of surface PL, if measured, is very powerful because its behavior gives direct information

−

−

−

Ef

ε0

O2
X1.47

1.0

Luminescence
(840 nm)

e

hν
(Illumination)

I−/I3−
H2Q/Q
Br − / Br2
CI− / CI2

2.0
OH−
3.0

+
Valence band

n-TiO2

Electrolyte (pH 1.0)

Fig. 11 Energy band diagram for explaining a PL band at 840 nm, emitted via a surface state connected with a surface reaction intermediate (X1.47 ) of water photooxidation on n-TiO2 .

163

164

2 Experimental Techniques

on surface structures and processes. The most important is to measure surface PL that is emitted via surface states connected with surface intermediates of electrode reactions. An example of such luminescence is a luminescence band peaked at 840 nm
(Fig. 6), observed for a rutile-type n-TiO2 electrode, which was activated beforehand by photoetching in aqueous H2 SO4 under anodic bias. In situ PL measurements, combined with in situ photocurrent measurements, can lead to detailed studies on molecular mechanisms of electrode reactions. It is concluded [25] that the 840-nm band for rutile n-TiO2 is emitted via a transition of an electron in the conduction band to a vacant level of a surface

reaction intermediate (X1.47 , probably vacant 2p-level of surface Ti-O· radical) of water photooxidation, as shown in Fig. 11.
2.4.3

Other Optical Techniques
2.4.3.1

Time-resolved Laser Spectroscopy

2.4.3.1.1 Transient Absorption Spectroscopy Measurements of transient absorption spectra are much more difﬁcult than the measurements of luminescence
(PL) described thus far, especially for solid samples. However, time-resolved transient absorption spectroscopy has been used to study carrier dynamics in the

Femtosecond Ti:sapphire laser system
780 nm, 3 mJ/pulse, 170 fs (10 Hz)
SHG
2ω ω H2O
(in 1 cm cell)

Optical delay fs white-light continuum Powder sample
(in 2 mm cell)

Iris

MonochroMCPD1 mator Monochro- MCPD2 mator ; Mirror
; Dichroic mirror
; Filter

Pump pulse
(390 nm)

Fig. 12 Experimental setup for femtosecond (pump probe) transient absorption spectroscopy.

2.4 Capacitance, Luminescence, and Related Optical Techniques

conduction band of TiO2 particles. The rate of photoelectron transfer from adsorbed dyes to the conduction band of TiO2 in vacuum and solution is also determined by use of measurements of transient absorption of injected electrons. It is reported that the photoelectron transfer occurs in a time range of 150–25 femtoseconds for chemically bound, strongly interacting dyes [28–31].

semiconductor bulk and at the semiconductor surface. Figure 12 shows an example of experimental setup for femtosecond
(pump probe) transient absorption spectroscopy [26]. The principle is to generate an excited state of the sample (semiconductor electrode) by a laser pulse (pump pulse) and then measure an absorption spectrum of the excited state by another laser pulse
(probe pulse), with varied delays in time from the pump pulse. A femtosecond laser, such as a Ti-sapphire laser, is used as the light source. In Fig. 12, a laser pulse with a frequency ω from the light source is converted with BBO to a light pulse with
2ω for use as the pump pulse. The original pulse with ω is separated with a mirror and used for the probe pulse. The latter pulse is led through an optical delay circuit to an H2 O cell to be converted into white light continuum for measurements of absorption spectra. The spectral intensity of the probe pulse is measured with a multichannel photodiode (MCPD).
Figure 13 shows an example of transient absorption spectra observed by direct excitation of nanocrystalline TiO2 particles suspended in vacuum [27]. The broad absorption band in the red to near infrared region is assigned to light absorption of photo-generated electrons in the

Transient Grating Spectroscopy
Transient grating spectroscopy is relatively easily handled compared with the transient absorption spectroscopy, and is often used to study carrier dynamics at semiconductor electrodes [32]. Figure 14 schematically shows the principle of transient grating spectroscopy. A femtosecond laser pulse for sample excitation is split into two beams, which are crossed again at the semiconductor surface to produce an optical striped interference pattern. The interference pattern produces a striped pattern of the densities of photo-generated electrons and holes near the semiconductor surface. The latter striped pattern gives rise to a striped pattern of optical refractive index near the semiconductor surface, which is monitored by measuring a diffraction pattern of a second probe laser
2.4.3.1.2

% Absorption
[%]

20

Transient absorption spectra of nanocrystalline TiO2 particles in vacuum after excitation by a femtosecond laser pulse (390 nm,
170 fs, 2 mJ cm−2 ). The delay time after excitation is indicated in the ﬁgure.
Fig. 13

15

0.2 ps
2 ps
10 ps
100 ps
1 ns
5 ns

10
5
0
400

450

500

550 600
Wavelength
[nm]

650

700

750

165

2 Experimental Techniques
Pump

Sample

White light continuum Diffraction spectrum Wavelength / nm

θ

Probe

Transient grating

Reflection

∼
∆n

Λ = λ ex / 2sin(θ /2)

Λ

Transmission

A schematic view of a sample region in transient grating spectroscopy to explain its principle.
Fig. 14

Fig. 15 Examples of transient grating signal decays, observed for an n-TiO2
(100) electrode by excitation at 360 nm and probing at 670 nm. The excitation pulse intensity in a µW unit is (a) 1.4,
(b) 4.1, (c) 6.1, (d) 9.7, and (e) 42.

(100)

Diffraction intensity

166

e d c b a
0

10

20

30

40

50

Time
[ps]

pulse (white light continuum). The carrier dynamics near the semiconductor surface is observed as rise and decay curves for the diffraction pattern of the second laser probe pulse.
Figure 15 shows examples of decays of transient grating signal (the intensity of the diffraction pattern) observed for an n-TiO2
(100) electrode by excitation at 360 nm and probing at 670 nm [33]. The decays are related with the rate of electron-hole recombination near the n-TiO2 surface.
In Situ Spectroscopic Investigation of Semiconductor Surfaces
Various spectroscopic techniques have been used for in situ investigations of
2.4.3.2

surface structures at semiconductor electrodes. In situ multiple internal reﬂectance
Fourier Transform Infrared (FTIR) spectroscopy is widely used to investigate surface termination bonds and their reactions. Figure 16 shows a semiconductor
(Si) wafer and an electrochemical cell for in situ multiple internal reﬂectance spectroscopy. By this method, various hydrogen termination bonds (Si−H, SiH2 , and SiH3 ) at terraces and steps for hydrogen ﬂuoride–etched or ammonium ﬂuoride–etched Si (111) and (100) are clearly detected with good spectral resolution [34, 35].
In situ investigations of semiconductor surfaces are also done by means

2.4 Capacitance, Luminescence, and Related Optical Techniques
Fig. 16 A silicon wafer and an electrochemical cell for in situ multiple internal reﬂectance
FTIR spectroscopy.

Si

IR light

Solution
In situ cell

IR

Si prism

of second harmonic generation (SHG) spectroscopy, ultraviolet and infrared light mixing spectroscopy, (surface enhanced) laser Raman spectroscopy, X-ray diffraction (XRD) method, electron paramagnetic resonance (EPR) spectroscopy, and so forth. Space-resolved investigations are also done by means of surface near-ﬁeld optical microscopy (SNOM), photoconductive atomic force microscopy, and so forth.
Other Miscellaneous Techniques
The electrolyte electroreﬂectance (EER) method is successfully used [36, 37] to determine the ﬂat band potential (Ufb ).
In this method, the optical reﬂectance at the semiconductor electrode surface is measured under modulation of the electrode potential by superimposition of a small AC voltage. The modulation of the electrode potential causes modulation of the density of majority carriers near the semiconductor surface, which in turn, causes modulation in light reﬂectance.
In situ microwave photoconductivity measurements, combined with photocurrent measurements, are also successfully used [38] to investigate dynamics of photo-generated minority carriers and interfacial kinetics at semiconductor electrodes. The microwave signal is closely
2.4.3.3

related with the density of free carriers, and modulated illumination enables us to investigate minority carrier dynamics.
Surface photovoltage (SPV) measurements have been used to investigate electronic structures at the semiconductor surfaces for semiconductor–vacuum and semiconductor–gas interfaces. This method is applied to semiconductor– liquid interfaces in case of insulating liquids [39].
2.4.4

Summary

Various methods to measure the capacitance and luminescence (PL and EL) at the semiconductor–electrolyte interface have been reviewed together with related optical techniques to investigate surface structures, optical properties, and carrier dynamics at the interfaces. Recent rapid progress in laser spectroscopy has enabled us to investigate very fast interfacial processes with a high time resolution of a few ten femtoseconds.
Moreover, rapid progress in scanning probe microscopy, combined with optical techniques, has enabled us to investigate interfacial structures and processes with a high space resolution of an atomic
(subnanometer) scale. Progress in these

167

168

2 Experimental Techniques

ﬁelds continues and it is thus highly probable that other new powerful methods will appear in the near future.
References
1. S. R. Morrison in Electrochemistry at Semiconductor and Oxidized Metal Electrodes, Plenum
Press, New York, 1980.
2. Y. V. Pleskow, Y. Y. Gurevich, in Semiconductor Photoelectrochemistry (Translated by
P. N. Bartlett), Consultants Bureau, New
York, 1986.
3. S. M. Sze in Physics of Semiconductor Devices,
2nd edition, John Wiley & Sons, New York,
1981.
4. V. V. Zavyalov, J. S. McMurray, C. C.
Williams, Rev. Sci. Instrum. 1999, 70,
158–164.
5. Unpublished result in a laboratory of
Prof. Y. Nakato, Osaka University.
6. Y. Nakato, T. Ueda, Y. Egi et al., J. Electrochem. Soc. 1987, 134, 353–358.
7. H. Minoura, T. Watanabe, T. Oki et al., Jpn.
J. Appl. Phys. 1977, 16, 865.
8. J. N. Chazalviel, J. Electroanal. Chem. 1987,
233, 37.
9. M. Fujitani, R. Hinogami, J. G. Jia et al.,
Chem. Lett. 1997, 1041–1042.
10. A. J. Nozik, R. Memming, J. Phys. Chem.
1996, 100, 13061–13078.
11. A. J. McEvoy, M. Etman, R. Memming, J.
Electroanal. Chem. 1985, 190, 225.
12. Y. Nakato, A. Tsumura, H. Tsubomura, J.
Electrochem. Soc. 1981, 128, 1300–1304.
13. B. L. Wheeler, A. J. Bard, J. Electrochem. Soc.
1983, 130, 1680.
14. W. Jaegermann in Modern Aspects of Electrochemistry (Eds.: R. E. White, B. E. Conway,
J. O’M. Bockris), Plenum, New York, 1996, pp. 136.
15. Y. Nakato, A. Tsumura, H. Tsubomura,
Chem. Phys. Lett. 1982, 85, 387–390.
16. ftp://ftp.batnet.com/pub/wombats/srsys/ ftp/lia.pdf 17. ftp://ftp.batnet.com/pub/wombats/srsys/ ftp/sr.pdf 18. M. Oyama, M. Mitani, M. Washida et al., J.
Electroanal. Chem. 1999, 473, 166–172.
19. D. V. O’Connor, D. Phillips in Time-correlated Single Photon Counting, Academic Press,
New York, 1984.
20. I. Yamazaki, N. Tamai, H. Kume et al., Rev.
Sci. Instrum. 1985, 56, 1187–1194.
21. N. Tamai, M. Ishikawa, N. Kitamura et al.,
Chem. Phys. Lett. 1991, 184, 398–403.
22. G. S. Beddard, T. Dout, G. Porter, Chem.
Phys. 1981, 61, 17.
23. K. Murakoshi, S. Yanagida, M. Capel et al.,
ACS Symp. Series 679 (Nanostructured
Materials-Clusters, Thin Films, and Composites), Amer. Chem. Soc. 1997, Chap. 17, pp. 221–238.
24. K. Mecker, A. B. Ellis, J. Phys. Chem. B 1999,
103, 995–1001.
25. Y. Nakato, H. Akanuma, Y. Magari et al., J.
Phys. Chem. B 1997, 101, 4934–4939.
26. T. Asahi, A. Furube, M. Ichikawa et al., Rev.
Sci. Instrum. 1998, 69, 361.
27. A. Furube, T. Asahi, H. Masuhara, J. Phys.
Chem. B 1999, 103, 3120.
28. Y. Tachibana, J. E. Moser, M. Graetzel et al.,
J. Phys. Chem. 1996, 100, 20056–20062.
29. T. Hannappel, B. Burfeindt, W. Storck et al.,
J. Phys. Chem. B 1997, 101, 6799–6802.
30. R. J. Ellingson, J. B. Asbury, S. Ferrere et al.,
J. Phys. Chem. B 1998, 102, 6455–6458.
31. J. B. Asbury, R. J. Ellingson, H. N. Ghosh et al., J. Phys. Chem. B 1999, 103, 3110–3119.
32. J. J. Kasinski, L. A. Comez-Jahn, K. J. Faran et al., J. Chem. Phys. 1989, 90, 1253–1269.
33. Unpublished result in a laboratory of
Prof. N. Tamai, Kwansei-Gakuin University.
34. P. Jakob, Y. J. Chabal, J. Chem. Phys. 1991,
95, 2897–2909.
35. M. Niwano, T. Miura, N. Miyamoto, J. Electrochem. Soc. 1998, 145, 659–661.
36. M. Pujadas, P. Salvador, J. Electrochem. Soc.
1989, 136, 716–723.
37. L. J. Ferrer, H. Muraki, P. Salvador, J. Phys.
Chem. 1986, 90, 2805–2807.
38. A. M. Chaparro, H. Tributsch, J. Phys. Chem.
B 1997, 101, 7428–7434.
39. S. Bastide, D. Gal, D. Cahen, Rev. Sci. Instrum. 1999, 70, 4032–4036.

173

3.1

Preparation of Nanocrystalline
Semiconductor Materials
Gary Hodes, Yitzhak Mastai
Weizmann Institute of Science, Rehovot, Israel
3.1.1

Introduction

The ﬁeld of nanophase materials is one that covers a wide and active area of research. The various properties of these materials, including mechanical, optical, electrical, and structural, are often very different from the same materials in the ‘‘bulk’’ phase. An example of this difference is the case of quantum dots
(QDs) – nanoparticles that are sufﬁciently small that their electronic energy structure is changed from that of the bulk material.
The size regime of semiconductor QDs varies from about 1 nm up to tens of nm in size, depending on the material properties. Considerable effort has been made in recent years to synthesize nanocrystalline semiconductors with control over the size and size distribution and to characterize their properties. As a result of these studies, it is now possible to synthesize a variety of nanocrystalline semiconductors with varying degrees of control over the size, shape, and a narrow size distribution.

Different strategies are used to prepare and keep the particles small and involve control of nucleation and growth of the crystallites. This can be achieved in a number of different ways and is described in the following section.
Electrodeposition normally leads to small particle size, largely because it is a low-temperature technique (molten salt electrolysis being an exception), thereby minimizing grain growth. Other factors also contribute to limitations of grain size and are discussed in this chapter. The low temperatures also minimize interdiffusion between electrodeposit and substrate or between different electrodeposited layers. In addition, electrodeposition allows a very high degree of control over the amount of material deposited through Faraday’s law, which relates the amount of material deposited to the deposition charge. This feature is particularly desirable when isolated nanocrystals are to be deposited on a substrate. Several different methods of electrodeposition, such as reduction of (chalcogen) oxide or zero valent (chalcogen) compounds in aqueous electrolytes, deposition from nonaqueous solvents, anodization, deposition from molten salts, and deposition by a multistage process, have been used. Electrodeposition offers a wide range of control of the nanocrystalline ﬁlm owing

174

3 Semiconductor Nanostructure

to the many parameters involved (potential or current density, pulsed electrodeposition, solution composition, convection, and temperature).
Although this chapter is limited to electrodeposition of semiconductors, it is only fair to mention, even if brieﬂy, some examples of electrodeposition of metal nanostructures. This is important because the principles and techniques used in electrodepositing metals are essentially the same as those used for depositing semiconductors – the main difference is that almost all studies on electrodeposition of nanocrystalline semiconductors involve compound semiconductors, with the added complications this entails.
Examples include pulsed electrodeposition of metal multilayers [1, 2], porous membrane-templated electrodeposition of gold nanotubes [3], and Ni nanowires [4].
3.1.2

Electrochemical/Chemical Hybrid
Deposition

A method for the synthesis of semiconductor QDs, called the Electrochemical/Chemical method (E/C), has been developed by Penner and his associates [5].
The E/C synthesis of semiconductors nanocrystals involves three steps:
1. In the ﬁrst step, metal nanocrystals M0
(e.g. Cd, Cu, Zn) are electrochemically deposited onto basal plane–oriented graphite electrode surfaces from aqueous solution of metal ions Mn+ .
2. Next, the metal nanocrystals are oxidized, either chemically or electrochemically, to the metal oxide or hydroxide.
3. Finally, the oxide (hydroxide) is displaced with gaseous hydrogen chalcogenide or chalcogenide anion (e.g. S−2 ) in solution to form nanocrystals of MX.

Step I – Electrodeposition of metal nanoparticles.
In the E/C synthesis, the ﬁrst step in which metal nanocrystals are deposited on the substrate is critical. The semiconductor particles grow from the metal particles on a particle-by-particle basis (i.e. each metal particle is chemically transformed to the corresponding semiconductor). The size and size distribution, therefore, in the ﬁrst step determine the ﬁnal size of the semiconductor particles. For that reason, it is important to achieve an understanding of the growth mechanism of the metal nanocrystal deposition. Penner and associates studied the electrodeposition of various metal nanoparticles (Ag, Pt, Zn,
Cu, Cd) mainly onto basal plane–oriented graphite and also onto Si electrodes [6–11].
The depositions were carried out from dilute aqueous solutions of metal ions using a potentiostatic pulse regime. A short (typically tens of ms) potential pulse was applied followed by open-circuit conditions.
It was demonstrated that the deposition of metal nanocrystals occurs via a VolmerWeber (three-dimensional growth) mode.
For example, in the deposition of platinum on graphite, platinum particles with a mean diameter of 5.2 nm were formed following the deposition of quantities equal to only 0.039 monolayers of platinum [10].
Volmer-Weber growth is favored at surfaces that are characterized by low interfacial free energy (like the graphite basal plane). Transmission electron microscopy
(TEM) and selected area electron diffraction (ED) of the graphite surfaces revealed that the deposited Pt particles are not epitaxially oriented with the graphite surfaces.
On the basis of electron microscopy and electrochemical studies, a two-step mechanism for the growth of Pt nanoparticles was proposed: ‘‘instantaneous’’ nucleation that takes part within a few ms, followed by

3.1 Preparation of Nanocrystalline Semiconductor Materials

diffusion-controlled growth from solution.
A similar mechanism was observed in the deposition of other metal nanoparticles.
For short pulse durations (approximately
10 ms), particles of a few nm with a narrow size distribution were obtained.
The particle size and size distribution both increased for increasingly greater pulse durations. In general for pulse plating, a high overpotential (or high current density) and short pulse duration favor reduced grain size caused by increased nucleation density and less time for growth. This increasing size distribution with growth time contrasted with three-dimensional nucleation and growth of colloidal particles, where instantaneous nucleation followed by diffusion-limited growth leads to a narrow size distribution. This difference was explained by a nonhomogeneous distribution of initially nucleated particles on the substrate [12, 13]. Where many nuclei were formed close together, their diffusion layers interacted, resulting in a reduced rate of diffusion and slower growth. On the other hand, particles that nucleated relatively far from other nuclei retained independent and thinner diffusion layers and therefore grew faster. In other words, nanoparticles that nucleate close together end up being smaller than particles that nucleate far from neighbors.
The size distribution could be narrowed for greater amounts of deposit by employing a train of short deposition pulses
(≤10 ms) followed by a much longer off-time (at open circuit), rather than a single longer deposition pulse [14]. These shorter deposition pulses diffusionally decouple the growth of densely nucleated regions from sparsely nucleated regions, and by this means, they narrow the size distribution of the nanocrystallites.

Steps II and III – The conversion of metal nanoparticles to semiconductor nanoparticles. The conversion procedure of metal nanoparticles to the corresponding semiconductor particles involves one or two steps. If the desired semiconductor is an oxide, such as ZnO [8], the conversion procedure involves one step, namely, the oxidation of the metal nanoparticles.
Other anions require an additional step in which the oxide (or hydroxide) is displaced by other ions such as in the case of CdS [14–16] and CuI [6]. (Although the formation of an oxide is not essential to obtain the ﬁnal semiconductor, oxide formation will normally occur unless steps are taken to prevent it, and it is more logical to deliberately oxidize the metal electrochemically.) For CdS, the Cd nanoparticles are spontaneously oxidized in the plating solution (at pH = 6 and open circuit) to
Cd(OH)2 . The displacement of OH by S to yield CdS can be carried out in two different ways: immersion of the Cd(OH)2 in an aqueous sulﬁde solution (Na2 S at pH = 10) [15], or by exposure of the
Cd(OH)2 to gas phase H2 S at 300 ◦ C [16].
In the synthesis from aqueous S2− , CdS particles with a mean crystal diameter between 2 and 8 nm were prepared. The nanocrystals were oriented with the c-axis of the CdS-wurtzite unit perpendicular to the electrode surface (i.e. graphite basal plane). In the gas-phase conversion, CdS/S core-shell structures were obtained by decomposition of the H2 S to S on the
CdS surface. β -CuI was formed by ﬁrst converting the Cu nanocrystals to Cu2 O followed by treatment in aqueous KI [6].
The various nanocrystalline semiconductors exhibited strong, size-dependent room temperature luminescence, although

175

176

3 Semiconductor Nanostructure

they were deposited directly on graphite [6,
8, 14]. The graphite, in contrast to metals, apparently does not efﬁciently quench the luminescence. Of particular importance is the observation that, under suitable preparation conditions, all the semiconductors emitted essentially band gap luminescence at room temperature with little or no deep sub–band gap band response that is normally characteristic of these materials. This implies that the nanocrystals were of very high quality.
3.1.3

Electrodeposition of Nanocrystals from
Nonaqueous Solution

One of the simplest techniques used to electrodeposit semiconductors is cathodic deposition from nonaqueous solutions containing elemental chalcogen (S, Se) and a metal salt, ﬁrst described by Baranski and
Fawcett [17]. Two main mechanisms have been considered: deposition of metal (e.g.
Cd) followed by chemical reaction with elemental chalcogen in solution and reduction of chalcogen to (poly)chalcogenide followed by ionic reaction between chalcogenide and metal cations. Which mechanism will dominate depends on the speciﬁc system (substrate, semiconductor, deposition conditions) and may change in the same deposition.
This technique was found to give nanocrystalline semiconductors which exhibited quantum size effects with typical lateral crystal size of 5 nm [18, 19]. The height of the crystals [measured by X-ray diffraction (XRD)] was often up to several times larger than the lateral dimensions
(measured by TEM). In fact, the increased transparency to shorter wavelengths because of size quantization, together with good (photo)conductivity expected from

‘‘wires’’, was used to improve thin ﬁlm
CdS/CuInS2 photovoltaic cells [20].
The nature of the anion of the metal salt is important in this technique. Salts of many anions are either insoluble or unstable in hot dimethyl sulfoxide (DMSO) containing dissolved chalcogen. Perchlorate and chloride are the most commonly used anions although other halides, methylsulfonate, and boroﬂuoride have also been used. However, another reason that the anion is important, particularly relevant to the present discussion, is that it affects the crystal size and therefore the band gap. Films deposited from Cl− solutions exhibit band gaps between 0.1 and
0.2 eV higher than those of the same compound deposited from ClO4 − [21]. More recent and detailed studies indicate that that this difference is due to adsorption of the more strongly adsorbed Cl− on the growing CdX surface, preventing further crystal growth [22].
Although both CdSe and CdS can be deposited in nanocrystalline form by this nonaqueous deposition, the essentially total insolubility of Te in DMSO prevents the use of the method for deposition of nanocrystalline tellurides (a small amount of Te can be codissolved with Se and mixed selenide-tellurides with small amounts of
Te can be deposited; see following section).
However, a related method to deposit
CdTe has been described by Cocivera and associates [23, 24]. They reacted elemental
Te with tri-n-butyl phosphine (TBP), which reacts with Te to form TBP telluride.
This compound, together with a Cd salt dissolved in propylene carbonate, allowed cathodic electrodeposition of CdTe. The as-deposited ﬁlms were reported to be Xray amorphous, a fact that suggested that they might in fact be nanocrystalline [26].
(Cd,Hg)Te ﬁlms grown by the same

3.1 Preparation of Nanocrystalline Semiconductor Materials

technique exhibited a crystal size of about
5 nm [25].
This technique was modiﬁed and simpliﬁed, in particular, by using a onestep technique to prepare the solution in
DMSO [26]. The CdTe was indeed found to be nanocrystalline with a wide size distribution varying from several nm up to tens of nm. In addition, the ﬁlms were generally nonstoichiometric with excess Te or
Cd, depending on the deposition potential.
It was difﬁcult to deposit close to stoichiometric ﬁlm. To improve the stoichiometry, reverse pulse deposition was used to strip excess Cd or Te during the anodic pulse.
The pulse regime also decreased the crystal size to an extent depending on the pulse parameters (crystal size typically several nm) and improved the size distribution.
The main factors limiting crystal growth were short pulse on-times and capping of the crystals with strongly adsorbing phosphine, mainly during the pulse off-time.
3.1.4

Size Control Using Semiconductorsubstrate Lattice Mismatch

An important factor that can inﬂuence crystal size, particularly for the ﬁrst layer of crystals, is the nature of the substrate.
Deposition of CdSe from the above DMSO electrolyte using Cd(ClO4 )2 onto ﬁlms of evaporated Au on glass or mica resulted in crystals of about 4–5 nm in size.
The distribution of the nanocrystals on the Au depended on current density and deposition temperature: high currents and low temperatures favored isolated crystals, whereas increasing aggregation occurred with higher temperature and decreased current [27].
ED [27] and HTEM [28] showed that the CdSe crystals were epitaxially deposited on the Au in a {111}Au ||{00.2}Cdse

and {110}Au ||{11.0}Cdse orientation relationship. The epitaxy arises because of the good lattice match between the CdSe and Au in a 2 : 3 relationship (−0.6% mismatch). Beyond the ﬁrst layer of crystals, the epitaxy is gradually lost and the crystal size grows for the perchlorate bath; for a
CdCl2 bath, the crystal size does not grow much because of capping, as discussed earlier. The mismatch strain, which gradually increases as the crystal grows, eventually leads to termination of growth. This can explain the relatively narrow size distribution obtained. This hypothesis of strain-determined growth termination suggested that crystal size should be controllable by choice of the semiconductor and substrate lattice parameters: the larger the mismatch, the smaller the crystal size, and vice versa. A range of different electrodeposited semiconductor-substrate combinations has been investigated to test this.
The lattice parameter of the CdSe was varied with (assumed) little change in chemical interaction with the substrate by depositing an alloy of CdSex Te1−x [29].
From Vegard’s law, a value of x = 0.88
(12% Te) should result in an increased lattice with a perfect match to Au. Although
Te could not be dissolved in DMSO, small amounts of Te could be dissolved in the presence of dissolved Se. The concentration of Te in the electrolyte was not known (it was almost certainly
106 -cm for Cu-poor ones. In particular, at a Cu concentration
>9.8 vol%, the resistivity decreased suddenly, typical of a percolation process. The resistivity in the perpendicular direction was much larger than that in the lateral one for reasonably Cu-rich samples (ratios of >109 ). No less important is the fact that

3.1 Preparation of Nanocrystalline Semiconductor Materials

although the transport in the lateral direction was always ohmic, the transport in the vertical direction exhibited pronounced negative differential resistance, attributed to resonant tunneling from Cu into hole states in Cu2 O [45, 46].
3.1.6

Template-directed Electrodeposition

There are a number of studies on electrodeposition of semiconductors in the pores of various membranes forming nanowires of the semiconductor. Klein and associates electrodeposited CdSe and CdTe into the pores of Anopore membranes [47]. These membranes are 50-µm thick with closely spaced 200-nm-sized pores. One side of the membrane was sputter-coated with Au followed by Ni electrodeposition, which partially ﬁlled the pores, as a substrate.
CdSe and CdTe were electrodeposited from acidic aqueous solutions containing
CdSO4 and SeO2 or TeO2 . The semiconductors grow as compact wires that are many µm long and 200–300 nm in diameter. By electrodepositing ﬁrst CdSe and then CdTe, nanowires were formed, which were compositionally graded over their length.
Chakarvarti and Vetter electrodeposited
Se in the pores of nuclear track membranes [48]. These pores were relatively large (2.5 µm).
Using porous anodic aluminium oxide ﬁlms, Routkevitchand associates electrodeposited very thin CdS nanowires [49].
The alumina membranes were typically
1–3-µm thick with pore sizes ranging from
9 to 35 nm. The CdS was electrodeposited, using ac electrodeposition, from a DMSO solution containing CdCl2 and elemental
S. The porous alumina, as anodized, was separated from the Al by a dense oxide

forming a rectifying contact. AC deposition then resulted in formation of CdS only in the pores and not in cracks or defects in the ﬁlm. The CdS wire dimensions closely followed those of the pores. For the smaller pores, the wires were apparently made up of single crystals joined in the axial direction, whereas for the larger pores, the coherence length, measured by XRD, was less than the pore diameter, suggesting that the wires were composed of several crystals also in the radial direction (or possibly, defected single crystals). It should be noted that even for ﬁlms of CdS deposited by this technique the crystal size is very small (see preceding section). The CdS was hexagonal (wurtzite) with its c-axis aligned predominantly along the membrane thickness. HTEM studies of these nanowires conﬁrmed the well-ordered and basal plane-textured CdS crystallinity [50].
Small quantum size effects were seen for the smallest wires, with band gaps varying from 2.36 to 2.42 eV [51].
The same technique to form CdS nanowires, only employing dc electrodeposition, has been used by Xu and associates [52]. In this case, the Al was etched away and a silver ﬁlm evaporated on the membrane. The wires in this case were textured with the (10.1) plane parallel to the membrane surface [in contrast to the ac deposition where the (00.1) plane grew parallel to the surface]. This technique was extended to CdSe and CdTe nanowires [53].
This group also deposited CdS nanowires in porous alumina templates using an aqueous solution containing CdCl2 and thioacetamide at pH = 4.6 [54]. This deposition is actually an electrochemically induced chemical solution deposition initiated by local pH changes caused by electrolysis [55]. In this case, the wires

181

182

3 Semiconductor Nanostructure

grew with the (00.1) plane parallel to the membrane surface.
3.1.7

Occlusion Electrodeposition of Composites

Apart from the layered Cu/Cu2 O composites described earlier, composites of nanocrystalline semiconductors with nonsemiconductors (metals or polymers) have been electrodeposited by incorporation of the semiconductor phase from solution into the electrodepositing metal
(also known as occlusion). WO3 particles suspended in solution can be incorporated in electrodeposited polypyrrole [56] or polyaniline [57] ﬁlms, and the resulting ﬁlms exhibit electrochromism, both of the polymer and of the WO3 . Although this technique worked only for oxides with very low isoelectric points, the isoelectric point of oxides with relatively high points of zero charge could be increased by adsorption of anions, such as sulfate or iodide, from solution. In that manner TiO2 was incorporated into electrodeposited polypyrrole ﬁlms, and the resulting ﬁlms showed a photoanodic photocurrent response because of the incorporated TiO2 [58]. Very high concentrations of WO3 – up to 53 wt% – were incorporated into electrodeposited polypyrrole [59]. High stirring rates and low current densities of plating favored high rates of incorporation. Incorporation of TiO2 into polypyrrole was attributed to mechanical entrapment of the
TiO2 particles in the rough, soft polypyrrole. In addition, the electric ﬁeld at the surface of the growing ﬁlm was also ascribed a role [60].
In a similar manner, TiO2 particles in suspension were incorporated into growing electrodeposited Ni ﬁlms and the resulting ﬁlms were photoelectrochemically active [61]. The morphology (and

photoelectrochemical response) of TiO2 was dependent on the electrodeposited metal matrix: deposition of Ni, Cu, Ag, and
In matrices resulted in different dispersion geometries of the TiO2 particles [62].
In the case of Cu, the photoelectrochemical response at a high pH showed both n-type and p-type polarities because of the formation of p-Cu2 O. The incorporation of alumina into electrodeposited Ni ﬁlms was shown to be controlled by diffusion of the alumina particles to the Ni surface and their residence time at the surface [63].
CdS, as well as TiO2 particles occluded into electrodeposited Ni, were shown to exhibit increased photoelectrochemical activity after potential cycling owing to passivation of the Ni matrix [64].
3.1.8

Sonoelectrochemical Formation of
Nanocrystalline Semiconductors

Sonoelectrochemical synthesis has recently been used for the preparation of semiconductor nanocrystalline powders.
In the sonoelectrochemical method, the ultrasound horn acts as both cathode and ultrasound emitter. This technique was used for preparing metal powders [65] and was extended to CdTe, although details of the CdTe particle size were not given [66]. CdSe nanocrystalline powders have been prepared by pulsed sonoelectrochemical reduction from an aqueous selenosulfate solution. The crystal size could be varied from X-ray amorphous up to 9 nm (sphalerite phase) by controlling the various electrodeposition and sonic parameters [67]. Crystal size was smaller for lower preparation temperatures, higher ultrasound intensity, and shorter current pulse width. These dependencies could be explained based on a pulse of electric

3.1 Preparation of Nanocrystalline Semiconductor Materials

current producing a high density of ﬁne particles on the sonic tip, followed by a burst of ultrasonic energy, which removes the particles from the cathode into the solution and prevents them from growing.
Similarly, PbSe nanocrystalline powders with a crystal size of 10 to 16 nm were prepared in the same manner from a similar solution using salts of Pb instead of
Cd [68].
MoS2 can be electrodeposited from a thiomolybdate solution [69, 70]. However, the deposit is apparently amorphous and requires annealing to crystallize. We reasoned that the high temperatures in and around ultrasonic cavitation bubbles might crystallize the electrodeposited
MoS2 . Sonoelectrochemical formation of
MoS2 powders using the above solution indeed resulted in partial crystallization of the as-deposited MoS2 , but in the form of multiple closed shell fullerene-like structures [71]. The structures were polyhedra of typically several tens of nm in size comprising layers (typically about 10) of crystallized MoS2 surrounding what was probably amorphous MoS2 . In some cases, nanotubes of MoS2 were also obtained. Although the cause of formation of these structures is not known, it is likely to be related to the buildup and collapse of the cavitation bubbles.
References
1. J. Yahalom, Surf. Coat. Technol. 1998, 105,
VII–VIII.
2. T. Cohen, J. Yahalom, W. D. Kaplan, Rev.
Anal. Chem. 1999, 18, 279.
3. C. R. Martin, D. T. Mitchell in Electroanalytical Chemistry (Eds.: A. J. Bard and I. Rubinstein), Marcel Dekker, Inc., New York, Basel
1999, p. 1, Vol. 21.
4. L. Sun, P. C. Searson, C. L. Chien, Appl.
Phys. Lett. 1999, 74, 2803.
5. R. M. Penner, Acc. Chem. Res. 2000, 33, 78.

6. G. S. Hsiao, M. G. Anderson, S. Gorer et al.,
J. Am. Chem. Soc. 1997, 119, 1439.
7. S. Gorer, G. S. Hsiao, M. G. Anderson et al.,
Electrochim. Acta 1998, 43, 2799.
8. R. M. Nyffenegger, B. Craft, M. Shaaban et al., Chem. Mat. 1998, 10, 1120.
9. J. V. Zoval, R. M. Stiger, P. R. Biernacki et al.,
J. Phys. Chem. 1996, 100, 837.
10. J. V. Zoval, J. Lee, S. Gorer et al., J. Phys.
Chem. B 1998, 102, 1166.
11. R. M. Stiger, S. Gorer, B. Craft et al., Langmuir 1999, 15, 790.
12. J. L. Fransaer, R. M. Penner, J. Phys. Chem.
B 1999, 103, 7643.
13. H. Liu, R. M. Penner, J. Phys. Chem. B 2000,
104, 9131.
14. S. Gorer, R. M. Penner, J. Phys. Chem. B
1999, 103, 5750.
15. M. A. Anderson, S. Gorer, R. M. Penner, J.
Phys. Chem. B 1997, 101, 5895.
16. S. Gorer, J. A. Ganske, J. C. Hemminger et al., J. Am. Chem. Soc. 1998, 120, 9584.
17. A. S. Baranski, W. R. Fawcett, J. Electrochem.
Soc. 1980, 127, 766.
18. G. Hodes, A. Albu-Yaron, Optical, Structural and Photoelectrochemical Properties of
Quantum Box Semiconductor Films, Proc.
Electrochem. Soc. 1988, 88–14, Vol. 298.
19. G. Hodes, Isr. J. Chem. 1993, 33, 95.
20. D. Gal, G. Hodes, D. Hariskos et al., Appl.
Phys. Lett. 1998, 73, 3135.
21. G. Hodes, T. Engelhard, A. Albu-Yaron et al.,
Mat. Res. Soc. Symp. 1990, 164, 81.
22. Y. Mastai, D. Gal, G. Hodes, J. Electrochem.
Soc. 2000, 147, 1435.
23. A. Darkowski, M. Cocivera, J. Electrochem.
Soc. 1985, 132, 2768.
24. J. von Windheim, M. Cocivera, J. Electrochem. Soc. 1987, 134, 440.
25. C. L. Colyer, M. Cocivera, J. Electrochem. Soc.
1992, 139, 406.
26. Y. Mastai, G. Hodes, J. Phys. Chem. B 1997,
101, 2685.
27. Y. Golan, L. Margulis, I. Rubinstein et al.,
Langmuir 1992, 8, 749.
28. Y. Golan, L. Margulis, G. Hodes et al., Surf.
Sci. 1994, 311, L633.
29. Y. Golan, J. L. Hutchison, I. Rubinstein et al.,
Adv. Mater. 1996, 8, 631.
30. G. Hodes, Y. Golan, D. Behar et al., in
Nanoparticles and Nanostructured Films,
(Ed.: J. H. Fendler), Wiley-VCH, Germany
1998, 1.

183

184

3 Semiconductor Nanostructure
31. D. Behar, I. Rubinstein, G. Hodes et al., Superlattices Microstruct. 1999, 25, 601.
32. Y. Golan, E. Ter-Ovanesyan, Y. Manassen et al., Surf. Sci. 1996, 350, 277.
33. Y. Golan, G. Hodes, I. Rubinstein, J. Phys.
Chem. 1996, 100, 2220.
34. Y. Zhang, G. Hodes, I. Rubinstein et al., Adv.
Mater. 1999, 11, 1437.
35. J. A. Switzer, M. J. Shane, R. J. Phillips, Science 1990, 247, 444.
36. J. A. Switzer, R. J. Phillips, T. D. Golden,
Appl. Phys. Lett. 1995, 66, 819.
37. J. A. Switzer, R. P. Raffaelle, R. J. Phillips et al., Science 1992, 258, 1918.
38. T. D. Golden, R. P. Raffaelle, J. A. Switzer,
Appl. Phys. Lett. 1993, 63, 1501.
39. R. J. Phillips, T. D. Golden, M. G. Shumsky et al., Chem. Mat. 1997, 9, 1670.
40. J. A. Switzer, C. J. Hung, B. E. Breyfogle et al., Science 1994, 264, 1573.
41. R. A. Vanleeuwen, C. J. Hung, D. R. Kammler et al., J. Phys. Chem. 1995, 99, 15 247.
42. J. A. Switzer, C. J. Hung, E. W. Bohannan et al., Adv. Mat. 1997, 9, 334.
43. J. A. Switzer, C. J. Hung, L. Y. Huang et al.,
J. Mat. Res. 1998, 13, 909.
44. E. W. Bohannan, L. Y. Huang, F. S. Miller et al., Langmuir 1999, 15, 813.
45. J. A. Switzer, C. J. Hung, L. Y. Huang et al.,
J. Am. Chem. Soc. 1998, 120, 3530.
46. J. A. Switzer, B. M. Maune, E. R. Raub et al.,
J. Phys. Chem. B 1999, 103, 395.
47. J. D. Klein, R. D. Herrick, D. Palmer et al.,
Chem. Mat. 1993, 5, 902.
48. S. K. Chakarvarti, J. Vetter, J. Micromech.
Microeng. 1993, 3, 57.
49. D. Routkevitch, T. Bigioni, M. Moskovits et al., J. Phys. Chem. 1996, 100, 14 037.
50. J. L. Hutchison, D. Routkevitch, A. AlbuYaron et al., Microscopy of Semiconducting
Materials 1997, 1997, 389.
51. D. Routkevitch, T. L. Haslett, L. Ryan et al.,
Chem. Phys. 1996, 210, 343.

52. D. S. Xu, Y. J. Xu, D. P. Chen et al., Chem.
Phys. Lett. 2000, 325, 340.
53. D. S. Xu, D. P. Chen, Y. J. Xu et al., Pure
Appl. Chem. 2000, 72, 127.
54. D. S. Xu, Y. J. Xu, D. P. Chen et al., Adv.
Mater. 2000, 12, 520.
55. K. Yamaguchi, T. Yoshida, T. Sugiura et al.,
J. Phys. Chem. B 1998, 102, 9677.
56. H. Yoneyama, Y. Shoji, J. Electrochem. Soc.
1990, 137, 3826.
57. H. Yoneyama, S. Hirao, S. Kuwabata, J. Electrochem. Soc. 1992, 139, 3141.
58. K. Kawai, N. Mihara, S. Kuwabata et al., J.
Electrochem. Soc. 1990, 137, 1793.
59. F. Beck, M. Dahlhaus, J. Appl. Electrochem.
1993, 23, 781.
60. F. Beck, M. Dahlhaus, N. Zahedi, Electrochim. Acta 1992, 37, 1265.
61. M. Zhou, N. R. deTacconi, K. Rajeshwar, J.
Electroanal. Chem. 1997, 421, 111.
62. N. R. de Tacconi, C. A. Boyles, K. Rajeshwar,
Langmuir 2000, 16, 5665.
63. P. M. Vereecken, I. Shao, P. C. Searson, J.
Electrochem. Soc. 2000, 147, 2572.
64. N. R. de Tacconi, H. Wenren, K. Rajeshwar, J. Electrochem. Soc. 1997, 144,
3159.
65. J. Reisse, H. Francois, J. Vandercammen et al., Electrochim. Acta 1994, 39,
37.
66. J. Reisse, T. Caulier, C. Deckerheer et al.,
Ultrason. Sonochem. 1996, 3, S147.
67. Y. Mastai, R. Polsky, Y. Koltypin et al., J. Am.
Chem. Soc. 1999, 121, 10 047.
68. J. J. Zhu, S. T. Aruna, Y. Koltypin et al.,
Chem. Mat. 2000, 12, 143.
69. H. J. Byker, A. E. Austin, J. Electrochem. Soc.
1981, 128, 381 (CA no. 536).
70. E. A. Ponomarev, M. NeumannSpallart,
G. Hodes et al., Thin Solid Films 1996, 280,
86.
71. Y. Mastai, M. Homyonfer, A. Gedanken et al., Adv. Mat. 1999, 11, 1010.

3.2 Macroporous Microstructures Including Silicon

3.2

Macroporous Microstructures Including
Silicon
C. L´vy-Cl´ment e e
LCMTR, Thiais, France
3.2.1

Introduction

Porous materials are generally obtained by localized electrochemical corrosion under anodic polarization (anodization). Some metals such as aluminum or stainless steel can be made porous. Comparatively, a far larger number of semiconductors can be made porous, leading to important changes in the physical properties. The best and ﬁrst known example is that of porous silicon (Si), which exhibits strong photoluminescence in the visible range instead of the weak luminescence of bulk Si.
Since its discovery more than forty years ago [1–3] and recognition of its porous nature ten years later [4], it has attracted considerable research interest because of its morphological, electrical, and optical properties, which are different from those of bulk silicon [5]. Interest in porous Si during the 1970s and 1980s was because of the fact that it could easily be converted into thick oxide ﬁlms and be used for dielectric (trench) isolation of active Si devices and in the full isolation by porous oxidized
Si (FIPOS) process, resulting in less than
200 published papers, spanning a period of 35 years [5]. During that period of time, investigation of the localized corrosion of other semiconductors (groups IV, III–V, and II–VI semiconductors) raised limited interest, until the discovery in 1990 of the strong luminescence of porous silicon arising from quantum-conﬁnement effects [6].
Since then, research work was mainly focussed on various aspects of porous Si, particularly on the rich variety of porous

Si structures that can form in hydroﬂuoric acid (HF) – based solutions [5, 7–11], models evoked for explaining them [5, 12,
3], elucidation of the structure of luminescent material and the origin of its efﬁcient luminescence [11, 16], and numerous technological applications of porous Si [17,
18]. Tremendous quantity of results (several thousand publications) have been analyzed all along the 1990s in various conference proceedings [19–26] and review articles [5, 8, 9, 11, 14–18, 27–30], as well as edited books [32–36]. Following the work on luminescent porous Si, a number of studies have been undertaken to render other semiconductors porous. This chapter will focus on the electrochemical aspects of porous semiconductors including porous Si.
3.2.2

Basic Principles
Thermodynamic Considerations
The general concept of the anodic dissolution process and reaction pathway was ﬁrst explored with group-IV semiconductors (i.e. silicon and germanium) (1–3).
3.2.2.1

Si + 6F− + (4 − n)h+ − −
−→
SiF6 2− + ne−

(1)

h+ and e− are a valence band hole and a conduction band electron, respectively. Quadrivalent dissolution to Si(IV) is assumed and n is a number between
0 and 4. This work led Gerischer to develop a general model of semiconductor/electrolyte interfaces, particularly for
III–V materials [37]. This model combines semiconductor band theory and electrochemistry. It assumes isoenergetic electron transfer across the semiconductor/solution interface, with the rate of

185

186

3 Semiconductor Nanostructures

decomposition potential is below the conduction band minimum (Ec < n Ed ) i.e. more positive than the conduction band.
A similar statement can be given for the holes. The semiconductor electrode is unstable against anodic (photo)corrosion if its decomposition potential is higher (more negative) than the valence band maximum
(Ev > p Ed ) of the semiconductor. In this case, the hole arriving at the interface carries out an oxidation reaction with the semiconductor or undergoes fast recombination with the electrons. The conditions of corrosion in an electrolyte are illustrated in Fig. 1.
Generally, the semiconductors are protected against cathodic (photo)corrosion in indifferent electrolytes because of the formation of a thin metallic ﬁlm produced at the ﬁrst instant of (photo)corrosion by the (photo)electrons, which for GaAs, for example, can be written as

reaction dependent on the overlap of the density of states.
When photoelectrochemical solar cells became popular in the 1970s, many reports appeared concerning the stability, dissolution, and ﬂat-band potential of semiconductors in solutions. These papers investigated parameters such as the energy level of the band edges, which is critical for the thermodynamic stability of the semiconductor and how to determine the potential for the onset of the (photo)electrochemical etching [38–40]. The criterion for thermodynamic stability of a semiconductor electrode in an electrolyte solution is determined by the position of the Fermi level EF with respect to the decomposition potential of the electrode with either the conduction band electrons n Ed or valence band holes p E . Under illumination, the quasi-Fermi d level replaces the Fermi level. The Fermi level is usually found within the band gap of the semiconductor and its position is not easily evaluated (especially the quasiFermi level of minority carriers). Therefore it was found more practical to use the conduction band minimum (Ec ) and valence band maximum (Ev ) as criteria for electrode corrosion. Thus, a semiconductor will be corroded in a certain electrolyte by the conduction band electrons if its nE d

E

n

E
Ec
nE d Ec

p

Ev

Ed n Ed
Ev

(b)

(c)

(d)

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

Fig. 1

Ec
Ed

Ev pE d

pE d (a)

(2)

The situation is different in the anodic regime. Gerischer [38] and Heller [42] pointed out that the majority of semiconductors (in the dark for p -type or under illumination for n-type) are unstable against corrosion with valence band
(photo)holes, as it is illustrated in Figs. 1(c)

Ec

pE d Ev

−→
GaAs + 3e− − − Ga0 + As3−

3.2 Macroporous Microstructures Including Silicon
Schematic drawings of the band diagram representing decomposition of semiconductor under anodization.
(a) photogenerated holes in an n-type semiconductor, (b) bias-generated majority carriers in a p-type semiconductor (from Ref. 41).

Fig. 2

Ec

Ec

e−

EF

p

pE

Ed h+ hv

Ev

d

h+
EF

Ev h+ h+
(a)

and 2. The oxidation reaction induced by the holes in the valence band is the oxidation of the semiconductor itself, which for
GaAs, for example, can be written as:
GaAs + 6h+ − − Ga3+ + As3+
−→

(3)

Depending on the composition of the electrolyte, the ions from the semiconductor are either soluble in the electrolyte or are insoluble, and a new phase can form, generally, an oxide. By choosing an appropriate electrolyte, potential bias, and anodization time, the corrosion can be controlled. Electrochemical Reactions
As for many other electrochemical reactions, two kinetic regimes are distinguishable in the anodic dissolution reaction [43]:
(1) reaction kinetics is rate-limiting (kinetic control) and (2) the rate-limiting step is the diffusion of one of the reactants to (or products from) the electrode surface. The morphology and properties of the semiconductor surface are totally different in the two regimes. Generally, if the electrochemical corrosion reaction is carried out at sufﬁciently low current densities, the electrode kinetics is rate-limiting, in
3.2.2.2

(b)

which case a rough or porous morphology is developed at the semiconductor surface. Under high current densities, a solid oxide is formed at the semiconductor surface because of mass transport of a limited supply of reactants or products. The reaction is diffusion-controlled and a smooth surface is consequently obtained. This is best demonstrated for Si in the presence of a solution of hydroﬂuoric acid (HF) [44, 45]. Current–voltage studies
(Fig. 3) performed on Si-HF electrolyte junctions show that the main requirement for electrochemical porous Si formation is that anodic current densities below a critical value, icrit , must be used. For a lowdoped n-type semiconductor, light must be supplied. The exact dissolution chemistry of silicon is still a matter of debate; however, ignoring the intermediate steps involved in the dissolution mechanism, the overall dissolution reaction for porous Si formation is as follows:
=⇒
Si + 6HF + (2 − λ)h+ = = SiF6 2− + H2
+ 4H+ + λe− ,

with λ ≤ 1 (4)

Porous Si formation is characterized by the fact that two charges are required to

187

3 Semiconductor Nanostructures
14
Electropolishing
12
10

I
[mA cm–2 ]

188

PS formation

icrit

8
6
Anodic dissolution
4
2
0

−1

0

1

2

E
[V/SCE]
Fig. 3

Typical current–potential curve for p-type Si in the dark in HF.

dissolve one Si atom from the electrode,
(i.e. the dissolution valence number equals
2) [1, 2, 46a,b] and H2 is evolved. When current density is larger than icrit , mass transfer to the solution limits the reaction and electropolishing occurs with a uniform dissolution of all the Si atoms at the surface of the wafer (dissolution valence number is 4). It is assumed that an anodic oxide is formed and is in turn chemically dissolved in HF. The overall dissolution reaction (Eq. (1) can then be better written as [46a,b])
=⇒
Si + 2H2 O + 4h+ = = SiO2 + 4H+
(5a)
=⇒
SiO2 + 6HF = = SiF6 2− + 3H2 O
(5b)
icrit value depends on HF concentration.
For low HF concentration (70% porosity) exhibit narrow undulating Si columns with

3.2 Macroporous Microstructures Including Silicon
Tab. 1

Different varieties of porous Si.

Type

Doping density

p

1015 cm−3

p+ and n+

1017 –1019 cm−3

p

1014 –1015 cm−3

HF concentration 25% aqueous
HF
25% aqueous
HF

n

1015 –1018 cm−3

30% aqueous
HF
8% HF in DMF or DMSO
5% aqueous

n, (100) oriented 2.1015 cm−3

6% HF aqueous

n

2.1015 cm−3

5% HF aqueous

1016 –1017 cm−3

3.1018 cm−3

Current density/ voltage Morphology

50 mA cm−2

Microporous [6]

3–300 mA cm−2 breakdown tunneling
40 mA cm−2

Mesoporous, [86, 89]

Macroporous [102]

10 mA cm−2

Idem [80]

15 mA cm−2 front-side illumination
10 mA cm−2 back-side illumination

Duplex layer
(nanoporous + macroporous) [70]
Array of vertical macropores developed on prestructured nuclei [72]
20 µm deep channels [67]
Duplex layer
(nanoporous + macroporous) [93]

10 V
2–15 mA cm−2

Fig. 7 TEM image and diffraction pattern from microporous Si formed on low-doped p-type Si (from Ref. 11).

195

3 Semiconductor Nanostructures
TEM image (bright-ﬁeld, under focus) of thin, high-porosity luminescent microporous. Columnar Si structures are arrowed (from
Ref. 7).

Fig. 8

diameter size down to 10 cm) substrates. It can be formed in HF contained aqueous and organic solutions (Fig. 16a).
Using speciﬁc HF/organic electrolytes

(b)

(a)

20 nm

35 nm

Fig. 15 HRTEM micrographs showing details of the nanometer-sized ﬁbers (plan-view): (a) (100) oriented, Nd = 1015 cm−3 ; (b) (111) oriented, Nd = 1018 cm−3 (from Refs. 91, 92).

a
Fig. 16 Cross-sectional SEM images of macroporous Si formed on (100) p-type Si. (a) Na = 1015 cm−3 in 10% HF/dimethylsulfoxide;
(b) Na = 1017 cm−3 in 8% HF/dimethylformamide.

b

3.2 Macroporous Microstructures Including Silicon

[dimethylformamide (DMF) and dimethylsulfoxide (DMSO)] permitted macropore formation on (0.2 cm) medium-doped
Si (Fig. 16b) [80]. The macropores diameter and depth depend on the doping density. The diameter varies from a few microns to 0.2 µ. The macropore depth can reach several hundred microns when organic electrolytes are used [114].
Models of Porous Si Formation
Several qualitative models have been proposed to explain porous Si formation but none of them allow full explanation of the rich variety of morphology exhibited by porous Si and, in particular, the formation of the duplex layers
(nano + macroporous). In addition, they possess very little predictive power. A majority of the models focussed on the pore propagation, whereas the mechanism of pore initiation received very little attention.
A comprehensive review of the various models proposed to explain pore formation is found in excellent review articles by
Smith and Collins [5], Parkhutik [12], and
Chazalviel and coworkers [13]. Two main categories of models have been proposed.
The ﬁrst one is basically electrostatic in nature, based on the consideration that physical effects associated with the SCR play a major role in the pore-formation mechanism. The second category is based on computer simulations.
3.2.3.6

Pore Initiation The few proposed models are related to the surface chemistry of Si during pore initiation. In general, defects are invoked as active sites for pore initiation. However, scanning tunneling microscopy (STM) observations have clearly shown that in initial stages the dissolution is isotropic and uniform on the nanometer scale, meaning that it is independent of defects [115].

3.2.3.6.1

Kinetics of pore nucleation on (100) p -type (0.05 cm) and n-type (1–4 cm) was observed as small steps in fastcurrent and potential pulse transients.
It was shown that the formation of porous Si could be better explained and correlated with experimental results if a silicon oxide intermediate was considered [116].
From the following observations that
(1) Si dissolves with formation of molecular H2 and protons, (2) only Si forms microporous layers with a single crystalline skeleton, and (3) hydrogen (Deuterium) species penetrate into the Si substrate upon porous Si formation, a model was proposed to explain pore initiation. It was assumed that hydrogen incorporation induces structural defects in Si, which may act as active sites for the localized Si dissolution. The model relates pore initiation to the selective dissolution of the hydrogeninduced structural defects at the surface of bulk silicon [117, 118].
Microetchpits One model developed to explain etchpit formation on II–VI semiconductors [119–121] has been extended to porous Si. This model attempts to address the initiation of the n-type macroporous morphology through nonuniformities in the photoetching current caused by nonuniform microﬁelds existing around the dopant atoms nearest to the semiconductor–electrolyte interface [100], or alternatively near the semiconductor surface because of a trapped (localized) positive charge formed in the ﬁrst instant of the photocorrosion at the interface. A calculation of the relevant parameters as a function of the doping density of the silicon wafer shows that up to Nd = 1017 cm−3 , the Debye screening length is larger than the mean distance of the ﬁrst donor atoms.
This means that up to this doping value

201

202

3 Semiconductor Nanostructures

the donor charges are not shielded and that the donors are expected to have an inﬂuence on the etchpit pattern. The ﬁrst layer of dopant (with a mean distance
1/3
a = 1/ 2N d from the semiconductor surface) contributes to a nonuniform photocurrent, which is high near the dopant atoms and can be ascribed to a photoinduced avalanche effect with a multiplication factor M(Ea ) where Ea is the electric ﬁeld near the dopant atom [100, 122]. An important consequence of this model is that the surface layer becomes dopant-poor after PEC-etching. This assumption was supported by the reduced surface recombination velocity, as indicated by photovoltaic measurements, which show substantial gain in the short wavelength part of the photocurrent spectrum and increased ﬁll factor, following PEC etching [123]. Once the initiation process sets in, other mechanisms, such as photocurrent focussing in the tip, may take place. A redistribution process starts. Some pores stop growing and terminate, whereas other pores continue to grow with increasing diameters.
Lehmann and F¨ ll [71, 72] showed that o pore initiation at the surface is determined by the doping density of the Si substrate.
From the initial 1010 cm−2 PEC-etched pits formed, only one in a hundred or one in a thousand survive and become pore tips. The density of the resulting pore tips is determined by the SCR width and therefore by the doping density.
Pore Propagation
Models Based on the Role of the Space
Charge Region Two fundamental aspects of porous-etching are the electrostatics and electrochemical kinetics. Electrostatics is crucial in determining the sites wherein semiconductor valence band holes are available for the dissolution reactions (e.g.
3.2.3.6.2

at the etchpit initiation sites or at the bottom of the pores). Electrochemical kinetics determines how fast holes react with the surface at sites when they are available.
Speciﬁc surface chemistries can make the reactivity for holes strongly different from one semiconductor–electrolyte system to another. The disadvantages of the most popular models proposed to explain porous Si formation are that they do not take into account the chemical aspects that may inﬂuence the kinetics at the
Si/electrolyte interface.
For pores to propagate, the pore walls have to be passivated and the pore tips to be active in the dissolution reaction.
Consequently, a surface, which is depleted of holes, is passivated to porous-etching, which means that anodization is selflimiting.
One must emphasize the distinct role of the SCR for n- and p -type Si. In the case of p -Si, it acts as a barrier for the holes, and only near the pore tips is the ﬁeld sufﬁciently large for opening gaps at the top of the barrier. The reaction may proceed entirely over the valence band (n = 2). In the case of n-type Si, the SCR rather acts as a sink channeling the holes to the pore tips. The porous Si may be formed on n-type Si under front and back illumination or under strong polarization (avalanche breakdown). On heavily doped n- and p -type Si, porous
Si may be obtained by hole-tunneling.
Eventual presence of intermediate Si oxide may favor the electron-tunneling process through the conduction band.
In summary, the effects responsible for pore wall passivation (top row) and for passivation breakdown at the pore tip
(middle row) as well as the resulting kind of porous Si structure together with substrate doping type (bottom row) are presented in
Fig. 17 [86].

3.2 Macroporous Microstructures Including Silicon
Field effect
(SCR)

(a)

(b)

(c)

Tunneling

Mesoporous Si on p + and n

Diffusion

Macroporous Si on p −

Avalanche breakdown Minority carrier collection Macroetchpits on n −

Macroporous Si on n −

Fig. 17 Field effect is proposed to be responsible for pore wall passivation
(a). Effects responsible for passivation breakdown at the pore tip (b) and the resulting porous silicon structure together with substrate doping type (c)
(adapted from Ref. 86).

Models Relevant to Micropore Formation
The mechanism of microporous Si formation from p -type is controversial. A puzzling question remains as to the unusually high stability of thin ﬁbers or nanoparticles of Si to electrolytic attack or, in other words, what prevents nanometerthick ﬁbers/particles from being dissolved altogether as the etching front moves microns into the crystal bulk.
The ﬁrst model, proposed by Beale and coworkers [85, 89], assumes that at potentials corresponding to pore growth the
Fermi level of Si is pinned near the middle of the band gap due to of a high number of surface states. This results in a potential drop at the interface, which restricts the hole transfer from the substrate. Anodic currents may ﬂow as a result of thermionic emission across the potential barrier. The local current ﬂow is then determined by the potential height, which in turn is dependent on the local electric ﬁeld. High electric ﬁeld favors hole transfer. Hence, the electric ﬁeld strength being greatest at the base pore, this results in dissolution only at the pore tips. Therefore, following pore initiation, the pores propagate as a result of the focussing effect of the electrical

ﬁeld. As a pore grows, the radius of curvature increases and the local ﬁeld decreases and the pore growth decreases. To maintain the total current density, new pores start to grow. As soon as the ﬁber/particle diameter in microporous Si gets thin, its resistivity is sharply increased by the Fermi level pinning to surface states. Because the electrochemical dissolution is essentially a current-controlled process, further etching is impossible and a thin ﬁber/nanoparticle is stable. This model is often disregarded because Fermi level pinning by surface states must be assumed. Unfortunately electron paramagnetic resonance (EPR) measurements indicate that there is a low density of surface states in the porous layer [124, 125]. This has been conﬁrmed by Searson and coworkers, using in situ impedance spectroscopy to measure interface states at [(111), Nd = 1013 cm−3 ] Si surfaces in ﬂuoride solutions [126, 127].
They found, at pH between 3 and 6, a very low density of electrically active surface states equal to 2 × 1010 cm−2 corresponding to about one in every 105 surface states.
Nevertheless, the idea of barrier-lowering may remain relevant. For example, if part of the applied potential appears in

203

204

3 Semiconductor Nanostructures

the Helmholtz layer [83, 84], the barrier will be decreased by the corresponding amount. If the Helmholtz potential drop is proportional to the surface electric ﬁeld, the barrier lowering will be inversely proportional to the SCR thickness.
A very often cited model is the so-called
‘‘quantum wire model’’ [62]. The aim of the model developed by Lehmann and
Gos¨ le was to explain the morphology obe served for porous Si, which luminesces as a result of quantum conﬁnement. As the dimensions of the particle decrease, the band gap increases (from 1.1 to 1.5 eV).
Because of quantum conﬁnement, the energy of valence band states is lowered in the walls between the micropores.
The holes are transferred to the interface only at the pore bottoms. The quantum conﬁnement model may be regarded as chemical passivation of pore wall dissolution, where the nanocrystalline silicon particles are the passivating species. However, this model seems to be contradictory to the well-established properties of isotype heterojunctions. A barrier height of a few hundred meV is known to be largely insufﬁcient for blocking majority carrier transport in a heterojunction at room temperature for both the directions of electric

current [128]. In addition, the ‘‘quantum wire model’’ has nothing to say about the formation of all other types of pores.
Models Relevant to Mesopore Formation
The Beale and coworkers’ model has been extended to explain mesopore formation on degenerately doped p - and n-type Si (in the dark) for which hole-tunneling is the major charge-transfer mechanism [85, 89].
The probability of a hole-tunneling across the Si-electrolyte interface depends on both the potential barrier height and SCR width.
The total electric ﬁeld at the pore tip results in both a lowering of the potential barrier and a decrease of the SCR width. Hence, as mentioned earlier, the propagation of pores is favored. As the pores grow, the depletion layers of adjacent pores may overlap. This results in the probability of tunneling becoming negligible at the pore walls; that is, the reaction is passivated at the pore walls. Overlap of the depletion layers prevents branching and explains the observed columnar structure.
Zhang [66] and Searson [9] conﬁrmed that model and evoked a ‘‘ﬁeld-enhancement effect’’ at the tips of the pores wherein the current ﬂow is controlled by a tunneling mechanism (Fig. 18). Their

Field strength (a.u.)

l

Fig. 18 Field strength distribution around an elliptically shaped pore tip (from Ref. 8).

3.2 Macroporous Microstructures Including Silicon

model predicts the formation of highly oriented parallel pores for (100)-oriented substrates and gives a reasonable explanation for the regular pore distribution and spacing by considering the depletion layer width. The calculation of the current distribution around the pore front using a two-dimensional solution to Poisson’s equation shows that the current is considerably greater at the pore tip as compared with the pore wall [9], accounting for the unidirectional growth of the pores. In the case of multiple adjacent pores, the equipotential lines in the region between the pore walls are shifted further away from the surface and into the bulk, effectively depleting the pore wall regions. The pore spacing is determined by the distance at which the region between the pore walls becomes completely depleted and is comparable to the calculated SCR width at a planar surface.
In 1972, Theunissen [67] observed formation of wide-etched channels for donor concentrations of less than Nd = 2 ×
1017 cm−3 under high voltage (>10 V), in
2.5–5% HF, when anodization was performed in the dark. He proposed that at some pore tips the electrical ﬁeld strength was sufﬁcient to allow avalanche breakdown at the depletion layer, which then generated the necessary carriers for further pore growth. This was corroborated recently by Lehmann et al. [86] showing experimentally and by simulations based on the electrical ﬁeld distribution present at the pore tip and pore walls that the mesopore formation is dominated by charge carrier tunneling, whereas avalanche breakdown is responsible for the formation of large etchpits.
Models Relevant to Macropore Formation
Macropores can be formed under illumination on n- and n+ -Si and in the dark

on n+ -Si as well as on low- and mediumdoped p -type Si.
1. n-type porous Si under illumination:
There is a reasonable agreement on the formation of macroporous Si from ntype Si [71, 72]. The formation mechanism of n-type macroporous Si is ruled by the reversed-biased SCR at the solution interface. Front-side illumination: The condition at the pore tip of an illuminated n-type electrode is different from that in the dark, because the presence of a breakdown ﬁeld is not necessary to generate charge carriers. Every depression or pit in the surface of the n-type silicon anode bends the electric ﬁeld in the SCR in a way that the concave surface regions become more efﬁcient in collecting holes than the convex ones. Concave regions are etched preferentially and the pores start to grow, consequently enhancing the local current density [68].
After initiation of microetchpits, macropore formation occurs. The density of the resulting pore tips is determined by the
SCR width and therefore by the doping density. The thickness of the remaining silicon walls between the pores is two times the SCR width and as a result the pore walls are charge carrier–depleted (Fig. 19). This ﬁnding has been conﬁrmed by impedance spectroscopy studies [104].
Until now, no model has been proposed to explain the simultaneous formation of the duplex nanoporous–macroporous structure, when the n-Si and n+ -type Si are illuminated from the front side.
The striking property of the orientation dependence of the nanoporous and macroporous structures may be correlated with the observation that the critical current icrit is greatest for the (100)-oriented substrate [72]. One possible explanation considers the process in which one Si−H

205

206

3 Semiconductor Nanostructures
Fig. 19 Schema showing the effective width of the SCR. (a) when the wall thickness, d, is smaller than twice the
SCR thickness, W, (b) when d is larger than twice W (from Ref. 66).

bond is substituted by a Si−F bond and may well be an indication that the surface chemistry is an extremely sensitive function of the silicon crystallographic orientation. For instance, the unreconstructed Si
(111) surface is ideally terminated by one
H atom per Si, forming a monohydride
(≡SiH) [129], and the Si (100) surface by two H atoms per Si, forming a dihydride
(=SiH2 ) [130]. It has been proposed that because of the presence of the two H atoms per Si the dihydride-terminated (100) surface is sterically hindered [77, 78, 31]. The steric hindrance causes bond strain, enhances the chemical reactivity, and the dissolution will occur faster along the (100) planes [5, 72].
Rear illumination: When anodization is performed under illumination from the backside, holes move to the interface by diffusion until they are captured by the
SCR, which accelerates them toward the pore bottoms. Lehmann [72] has clearly demonstrated the validity of the macropore propagation model by studying, in aqueous HF solution, the formation of a regular orthogonal pattern of cylindrical macropores on a rear-illuminated substrate. He initiated the pores by using standard lithography to produce a predetermined homogeneous pattern of pits and subsequently developed the pores by alkaline etching. Illumination with a

suitable wavelength from the rear of the
Si wafer favors the collection of the photogenerated holes in the bulk and their migration toward the tip of the initiated pits. Such an experiment is only possible for lightly doped n-type Si, because the diffusion length of the holes is comparable to the sample thickness. Stable macropore growth occurs when the current density is limited by hole generation and not by the applied bias. In addition, if the local current density at the pore tip, itip , exceeds the critical current density (itip ≥ icrit ), no nanoporous silicon will form at the pore tip because of the presence of the electropolishing regime. For a given initiation pattern, the pore diameter, dp , and wall thickness, W , are determined by the ratio between the applied current, i , and the critical current, icrit [Eqs. (8) and (9)]. dp = p

i

0.5

(8)

icrit

and w =p 1−

i icrit 0.5

(9)

The diameter of the macropores etched at different current densities, i , is a linear function of the square root of i /icrit following Eq. (8). It was found that the wall thickness could be up to a factor 10 smaller or wider than the SCR width. The rate of

3.2 Macroporous Microstructures Including Silicon

pore growth is a function only of the critical current density icrit (which is a function of the HF concentration and temperature).
The stable macropore formation obtained for an applied bias is sufﬁcient to generate the critical current density
(about 1 V). This understanding of the formation mechanism has allowed good control of the geometrical parameters of

macropore arrays (Fig. 20). It has been shown recently that 100-µm wide pores with 2-µm wall thickness and 200-µm depth (Fig. 21) can be formed on highly resistive (2000–5000 cm) n-type Si [132].
The demonstration that there is no restriction concerning the wall spacing opens the route to form vertical structure by
(photo)electrochemistry.

Cross-sectional SEM images and 45◦ bevel of n-type Si samples (Nd = 1015 cm−3 ) showing the predetermined patterns of macropores. (a) orthogonal array, (b) hexagonal

array. Pore growth was induced by regular patterns of etchpits produced by standard lithography and subsequent alkaline etching
(inset upper right) (from Ref. 72).

Fig. 20

Fig. 21 Cross-sectional SEM image of macropores obtained from a prepatterned n-type Si
(2000–5000 cm) in 3% HF electrolyte with backside illumination (from Ref. 132).

10 µm

207

208

3 Semiconductor Nanostructures

A recent systematic study of macropore formation performed on various doped ntype Si substrates with rear illumination, by F¨ ll and coworkers [106] showed that a o strong inﬂuence of the SCR on the average macropore density is indeed observed in accordance with the Lehmann model [72]
(i.e. an increased anodic bias decreases the density of pores), except for highly doped
Si. It was observed that an increasing anodic bias increases the pore density, in contrast to the prediction. The pore growth seems to be dominated by the chemicaltransfer rate and most likely calls for a chemical passivation mechanism of the macropore walls.
2. p -type Si: Formation of macroporous
Si is surprising at ﬁrst sight because in contrast to n-type Si, p -type electrodes are under forward bias conditions. An extended SCR is therefore not expected for a p -type electrode in the anodic regime. To explain macropore formation process on p -type Si, Lehmann and R¨ nnebeck [102] o put forward the key role of diffusion across the space charge region. Especially, the ﬁeld enhancement and the associated narrowing of the SCR at the pore tips are assumed to be responsible for the pore tip dissolution and pore wall passivation. This model predicts macropore formation up to Na = 1017 cm−3 doping level, but fails to explain why macroporous formation on moderately doped p -type
Si (1–0.2 cm) is only observed when
HF-contained organic protophilic solvents
(DMF and DMSO) are used [80].
Another model proposed by Kohl and coworkers [77, 78] is based upon the strain – induced preferential etching described earlier. The model accounts for the formation of macropores and highly branched micropores when the silicon is rendered porous in either nonaqueous or aqueous HF solutions, respectively.

They suggest that the contrast between aqueous and nonaqueous etching can be attributed to two factors, the competition of OH− with F− for complexing Si, and the kinetically slow dissolution of oxide
(or hydroxide) species formed in aqueous solutions. Computer Simulations These have been attempted to obtain morphologies similar to those observed in porous Si. The models proposed to explain porous Si formation are similar in spirit to those previously used to understand the complex crystalgrowth phenomena. Although porous Si formation is a dissolution process, similarity is found with the growth phenomenon.
The models fall in two categories. The ﬁrst type corresponds to the popular diffusionlimited aggregation (DLA) model, which is based on the diffusion of an electrostatic species such as hole (electron) to (from) the interface [133]. The second type is the
Mullins-Sekerta instability model [134]. It consists of analyzing the linear instability of the Si/electrolyte interface by taking into account both the transport of holes in the semiconductor and ions in the electrolyte together with the surface tension of Si. The
DLA model has ﬁrst been used by Smith and coworkers to explain pore formation as a result of the diffusion of holes to the interface [135, 136]. Yan and Hu adopted a different approach by modeling the interfacial dynamics governing the formation of porous Si using a two-dimensional twocomponent resistor network model [137].
The ratio of the resistances of the two networks is used as a control parameter to simulate the advance of the interface according to stochastic dynamics, which involves local current. John and Singh developed a diffusion-induced nucleation model for the formation of porous Si based on two primary processes [138]. The

3.2 Macroporous Microstructures Including Silicon

diffusion of holes from the bulk to the surface is controlled mainly by (1) the SCR width (w ) and (2) the drift-diffusion length
(l ) of holes inside the lattice. The theoretical models permit obtaining the porous Si morphology, which looks very similar to what is observed in practice but generally lack physical substantiation and serve as an illustrative facility rather than an analytical tool relevant for both scientiﬁc and practical applications.
The concept of the instability model used to explain porous Si formation was ﬁrst introduced by Kang and Jorn´ [139, e 140]. The model considers the pore nucleation at the Si surface as a mathematical problem of the instability of a planar interface toward small perturbations. The interface can be destabilized for an optimal deformation wavelength as a result of a competition between the destabilizing effect of hole diffusion and the stabilizing ones because of ion diffusion in the electrolyte and surface tension of Si. The optimal wavelength is expected to give an order of magnitude of the interpore spacing. The analytical and numerical stability analysis of the Si/electrolyte interface for
PEC-etching of n-type Si was performed by Valance [141, 142]. This model allowed expression of the dissolution speed and derivation of the scaling laws for interpore spacing as a function of the doping level of Si and applied potential. Another important result obtained by Kang and Jorn´ e was the relationship between the intensity of the rear illumination and the pore diameter: the higher the intensity, the larger the photocurrent per pore and the larger the pore diameter. Chazalviel and coworkers recently addressed the linear stability analysis of the interface to the case of p -type Si [13, 143]. It allowed an understanding of the observed changes in the

distribution of structure sizes (from microporous to macroporous) as the layer thickened, and the dependence of the pore sizes on the resistivity of the starting material. 3.2.4

Porous Semiconductors: A Review
PEC and EC Etching (Table 2)
The ﬁrst observation of intentional localized corrosion on semiconductors, other than Si, was purposely performed to corrugate the semiconductor surface, decrease the reﬂectivity, improve the optoelectronic properties, and consequently increase the performance of photoelectrochemical solar cells.
3.2.4.1

Group II–VI Semiconductors
Beside the porous Si, II–VI semiconductors are those for which the origin of deep microrelief (micro-etchpit) formation was studied ﬁrst. Long ago, n-type
CdS [144], CdSe [145, 146, 47], n- and p -type CdTe [119, 120, 148, 149], n-type
CdSe0.65 Te0.35 [150], ZnSe [121], and CdHgTe [151] have been found to undergo an extreme surface roughening under EC and PEC etching. The II–VI electrodes were exposed in aqua regia to a reverse bias of over 1 V/SCE accompanied by highintensity illumination. The duration of the
PEC-etching generally did not exceed 5 s.
The anodic dissolution valency is two, which implies the formation of elemental tellurium forming an insulating layer that hampers further transfer of photogenerated holes.
3.2.4.1.1

−→
CdTe + 2h+ − − Cd(aq) 2+ + Te0 (10)
After removal of the insulating (photo)corrosion products obtained by dissolution in a suitable medium, the surface contains

209

Band gap size (eV)

2.42

1.72

1.44

2.7

1.35

1.22
1.37

Type

d

d

d

d

i

i d CdS [144]

CdSe [146,
147]

CdTe (111)
[119, 120,
148, 149]

ZnSe [121]

InP (149)

WSe2 [158]

PEC-etched semiconductors.

Semiconductor

Tab. 2

n

n,
1–2 × 1018 cm−3

1016 –1017 cm−3
(5–0.05 cm) n, 4 × 1017 cm−3

1016 –1017 cm
(1 cm) n, 1000 cm

n , 2 cm
(4.5 × 1010 cm−2 donors) n, 10 cm

Conduction type 0.4–1 M
HClO4

1 M HCl

HNO3 , HCl,
H2 O (1/4/20)

HNO3 , HCl,
H2 O (1/4/20)

HNO3 , HCl,
H2 O (1/4/20)
1 M Na2 SO3
0.2 M FeSO4 +
0.1 M H2 SO4

HNO3 , HCl,
H2 O (1/4/20)

Electrolyte

10 mA cm−2 , light, 1–120 min.
+1 V, light

+1 V, light

+1 V, light,
3–4.5 s

+1 V, light,
3–4.5 s, 12 mA

Potential
(V/SCE)/current
density λ laser (nm)
457.9
514

Microetchpits on van der Waals planes on initiated punctures

Etchpits λ laser (nm)
1010 cm−2
457.9
5 × 108 cm−2
472.7
0
488
Elongated etchpits µm. Groove structure along (011) axis

Idem 0.15 µm

1 µm triangular etchpits

0.2 µm etchpits

Etchpits
2 × 109 cm−2
2 × 108 cm−2
1-µm etchpits

Morphology

210

3 Semiconductor Nanostructures

3.2 Macroporous Microstructures Including Silicon

(a)

(b)

Plan view SEM images of photoetched (111) n-type CdSe. Inﬂuence of the doping concentration on the density of the micro-etchpits.
(a) Nd = 1016 cm−3 ; (b) Nd = 1017 cm−3 (from Refs. 145, 146).

Fig. 22

a pattern of highly dense microetchpits
(>109 cm−2 ) whose density and shape vary with the doping density and crystallographic orientation of the material
(Figs. 22a and 22b). The size of the microetchpits observed under scanning electron microscopy (SEM) is on the order of a few hundred nanometers. On (110) and
(111) surfaces, the pits have parallelogram and triangular shapes, respectively. The
PEC-etched (texturized) surfaces have not been described in terms of the formation of a porous layer, especially as the depth of the pores is not known. However, there is a striking similarity with some of the macroporous silicon morphologies. It has been suggested that the texturized surface arises from the preferential etching of surface defects in the vicinity of dopants [121].
The characteristic size (200 nm–1 µm) of the observed structures and its decrease with increasing sample doping were given as arguments in favor of this model.
Group III–V Semiconductors
General trends observed with II–VI semiconductors are also observed on
III–V semiconductors. Formation of etchpits was observed on n-InP in the presence of HCl solution, under illumination [47, 152–154]. The main
3.2.4.1.2

difference with the II–VI compounds is that the corrosion products do not lead to a passivation layer and a long duration PEC etching renders the InP surface porous.
This will be discussed in the next section.
Layered Semiconductors Layered semiconductors such as InSe [155],
WX2 (X = S, Se) [156–158] and SnS2
(X = S, Se) [158] have been photoelectrochemically etched in a manner similar to the II–VI semiconductors. A review of this work is given in Ref. 158. The shape of the etchpits depended strongly of the crystallographic orientation of the etched surface. Once again, the purpose of this work was to improve the photoelectrochemical characteristics of electrodes. In the case of the van der Waals surface, known to be free of surface states (dangling bonds), no photoetchpits are formed upon PEC etching. On the contrary, the surface perpendicular to the van der Waals surface that has a signiﬁcant number of dangling bonds is strongly corroded. Micrometer photo-etchpits can be initiated on WSe2 van der Waals planes after creating defects by microscopic punctures at selected places [159] (Fig. 23). No porous layer has been observed.
3.2.4.1.3

211

212

3 Semiconductor Nanostructures

Fig. 23 Plan view SEM image of photo-etchpits initiated at the van der
Waals surface of WS2 , after creating defects by microscopic puncture at selected places and upon PEC etching (adapted from Ref. 41).

Porous-etching (Table 3)
Following the work on luminescent porous
Si, a number of studies have been undertaken to render other semiconductors nanoporous. The motivation of such studies was based on the fact that if they can exhibit tunable luminescence in a similar way to nanoporous Si, then common features or differences might reveal the mechanisms involved. Studies concern indirect band gap semiconductors such as those of column IV (Ge, Si1−x Gex , SiC) and GaP, as well as direct band gap III–V alloys (GaAs and InP) and II–VI compounds (CdTe, Cd0.95 Zn0.05 Te, CdTe, and
ZnTe).
3.2.4.2

3.2.4.2.1 Porous Group-IV Semiconductors, Other Than Si Various column-IV semiconductors have been made microporous using recipes inspired from those used in porous silicon: p -type
Ge (0.66 eV), Six−1 Gex (x = 0.04–0.4) alloys, n- and p -type SiC (2.8–3.2 eV), p -type hydrogenated amorphous Si (a-SiH), and
Si1−x Cx alloys have been made porous in ethanolic HF, either in the dark or under

illumination. Depending on the experimental conditions (n- or p -type, in the dark or under illumination), the reported structure sizes generally range from 10 nm to
1 µm.
In the case of a-SiH and Si1−x Cx , the thickness of the nanoporous layer is limited by the formation of macropores, an instability of the growth front attributed to the high resistivity of the starting material [160].
1. p -type Ge: Porous Ge [161–166] formed in 10% ethanolic HF on
6–10 cm (100) p -type Ge under galvanostatic conditions (50 mA cm−2 ) most probably has a similar nanostructure (∼2 nm) to porous Si obtained on low-doped Si as suggested by extended
X-ray absorption ﬁne structure (EXAFS) measurements [164]. One main difference is that the porous Ge surface is much rougher than the porous
Si surface. A 10-min. porous-etching produces a roughness greater than
5 µm, which is similar to that obtained from porous-etching Si for an hour under similar conditions. The

0.66

i

i

i

Ge [161–166]

Si0.8 Ge0.2
[166–175]

6H-SiC
[176–195]

3

Band gap size (eV)

Type

Porous semiconductors other than Si

Semiconductor

Tab. 3

cm

Diluted HF

2.5 M NH4 F

n,
1–3 × 1018 cm−3

HF (12.5–24%)/
C2 H5 OH/H2 O

(10%) HF

Electrolyte

p,
2.2 × 1018 cm−3

p, 1019 cm−3
1017 cm−3

p, 6–10

Conduction type

2.5 mA cm−2
0.5–1 h + light

50 mA cm−2

10–40 mA cm−2

50 mA cm−2 ,
10 min

Potential (V/SCE) or current density

(continued overleaf)

1–10 nm Interpore spacing, Crystallite morphology undetermined i.e. spongelike crystallite network, or wire-like

Porous + roughened morphology and shallow depression
(5 nm), these suboxide nanoparticles react with
H2 S gas, which also diffuses through the nozzles (c) into the inner reactor (a). In this reactor, the mass difference between the two gases, which determines their diffusivity through the nozzles c, is exploited to separate the reaction zones for the reduction and sulﬁdization. The sulﬁde coated oxide nanoparticles are carried by the carrier gas outside the reactor a. Because these nanoparticles are surface passivated they land on the ceramic ﬁlter (d) and the oxide to sulﬁde conversion reaction continues within the core without coalescence of the nanoparticles. This process yields a

245

246

3 Semiconductor Nanostructures
Scheme of the experimental GPR setup.

Flow controller
1

Flow controller
2,3,4

Flow controller
2,3,4
N2

Nozzles, headon view:
(a)
Exit
N2 /H2
H2S

N2 /H2
14 mm

H2S

Vapor
MoO3

Upper heating element

To

Upper heating element

Quartz rod H
(c)

Tn

(b)

28 mm
Lower
heating element 55 mm

Lower heating element

IF-MoS2

(d)

(a) Schematic representation of the experimental vertical gas-phase reactor [16]: a-inner tube; b-middle tube; c-nozzles; d-external tube; (b) blow up of the inner reactor-a. The quartz rod is used to collect the various reaction products for analysis. Fig. 3

3.3 Inorganic Nanoparticles with Fullerene-like Structure and Nanotubes
Scheme of the suboxide nanoclusters formation.

N2

N2

MoO3 powder N2

MoO3 powder Molecular clusters
MoO3
MoO3–x
Nanosize clusters
MoO3–x

K1
H2

MoO3–x

H2

MoS2

K2
H2 / N2
H2S

H2 / N2
H2S

K3
K4 H

H2

2

H2S

Fig. 3

H2S

(Continued)

pure IF-MoS2 phase, and the control over the size and shape of the nanoparticles is quite good.
The synthesis of a pure phase of
WS2 nanotubes, 2–10-mµ long and with diameters in the range of 20–30 nm, has been recently reported [7]. Here, short tungsten oxide nanowhiskers (50–300-nm long) are prepared by heating a tungsten ﬁlament in the presence of water vapor.
These short nanowhiskers are reacted with
H2 S under mild reducing conditions. The length of the resulting nanotubes is determined by the interplay between three reactions: tip growth of the oxide nanowhisker, reduction, and sulﬁdization. In a strong reducing atmosphere (5% hydrogen in the gas mixture), relatively short nanotubes are obtained. On the other hand, long (up to 10 µm) and highly crystalline nanotubes

are obtained by slowing down the rate of the reduction and sulﬁdization reactions.
A typical assemblage of such nanotubes is shown at two magniﬁcations in Fig. 4.
One can visualize the growth process of the encapsulated nanowhisker as follows. The short oxide nanowhisker reacts with H2 S and forms a protective tungsten disulﬁde monomolecular layer, which covers the entire surface of the growing nanowhisker, except for its tip, which continues to grow uninterrupted. This WS2 monomolecular skin prohibits coalescence of the nanoparticle with neighboring oxide nanoparticles, which therefore drastically slows their coarsening. Simultaneous condensation of
(WO3 )n or (WO3−x •H2 O)n clusters on the uncovered (sulfur-free) nanowhisker tip, and their immediate reduction by hydrogen gas, lead to a reduced volatility of

247

248

3 Semiconductor Nanostructures

(a)

(b)

SEM micrograph of a mat of long WS2 nanotubes at two different magniﬁcations [8].

Fig. 4

these clusters and therefore to the tip growth. This concerted mechanism leads to a fast growth of the sulﬁde-coated oxide nanowhisker. Once the tungsten oxide

source is depleted, the vapor pressure of the tungsten oxide in the gas phase decreases and the rate of the tip growth slowdown. It is believed that the source

3.3 Inorganic Nanoparticles with Fullerene-like Structure and Nanotubes

of the oxide clusters in the vapor phase stems from very small (20 nm) remains amorphous. Because the transformation of the amorphous core into a crystalline structure involves a slow outdiffusion of sulfur atoms; the core of the nanoparticles is unable to crystallize during the short (ns-µs) pulses. It is likely that the sonochemical formation of IF-MoS2 nanoparticles can be attributed also to the self-propagating selfextinguishing process described earlier.
In a related study, scroll-like structures were prepared from the layered compound GaOOH by sonicating an aqueous solution of GaCl3 [32]. This study again shows the preponderance of nanoparticles with a rolled-up structure from layered

compounds. It has been pointed out [77] that in the case of nonvolatile compounds, the sonochemical reaction takes place on the interfacial layer between the liquid solution and the gas bubble. It is hard to envisage a gas bubble of such an asymmetric shape as the GaOOH scroll-like structure. Alternatively, it is proposed that a monomolecular layer of GaOOH is formed on the bubble’s envelope, which rolls into a scroll-like shape once the bubble is collapsed. Rolled-up scroll-like structures were obtained by sonication of InCl3 ,
TlCl3, and AlCl3 , which demonstrates the generality of this process [32]. Hydrolysis of this group of compounds (MCl3 ) results in the formation of the layered compounds
MOOH, which, on crystallization, prefer the fullerene-like structure. However, the hydrolysis of MCl3 compounds may also lead to MOCl compounds, also having a layered structure, and its formation during the sonication of MCl3 solutions has not been convincingly excluded. Figure 6(a) shows a nested structure of Ni(OH)2 obtained by the sonochemical reaction of
NiCl2 solution (courtesy of Y. Rosenfeld
Hacohen, Weizmann Institute). The magniﬁed core of the same nanoparticle is shown in Fig. 6(b), which clearly reveals its nested fullerene-like structure with the hollow center. Nanoscale tubules with scroll-like structure have been obtained from potassium hexaniobate (K4 Nb6 O17 ) by acid exchange and careful exfoliation in basic solution [78, 79]. The exfoliation process results in monomolecular layers, which are unstable against folding even at room temperature and consequently form the more stable scroll-like structures.
More recently, the formation of fullerene-like Tl2 O nanoparticles by the sonochemical reaction of TlCl3 in aqueous solution was reported [80]. Tl2 O possesses the anti-CdCl2 structure, with the oxygen

3.3 Inorganic Nanoparticles with Fullerene-like Structure and Nanotubes

layer sandwiched between two thallium layers. Currently, the yield of the IF-Tl2 O product is not very high (ca. 10%), but puriﬁcation of this phase could be obtained by heating of the sample to 300 ◦ C. The size and shape control of the fullerene-like particles is not easy in these types of reactions.
Nonetheless, the fact that this is a room temperature process is rather promising, and future developments will hopefully permit better control of the reaction products. Perhaps, the most important aspect of this work is the fact that bulk Tl2 O is not stable in the ambient atmosphere, but the fullerene-like structure is found to be rather stable in these conditions. This fact stems from the closed (seamless) nature of

TEM view of a nested fullerene-like Ni(OH)2 nanoparticle. The layer to layer distance is 0.46 nm.
Fig. 6

the cage. In contrast to that, macroscopic platelets of this compound are unstable in the ambient because facile water and oxygen intercalation occurs through the edges into the van der Waals gap separating the molecular layers. This process disrupts the stacking of the molecular layers and leads to their exfoliation and rapid oxidation.
This extra stability of IF nanostructures in ambient conditions is discussed in greater detail in the following section. It was discussed before in the context of the synthesis of IF-VS2 , which does not have a stable 2D macroscopic analog [81].
NbS2 is a compound with a layered structure, exhibiting a transition to superconductivity at 6 K. Figure 7 shows an

253

254

3 Semiconductor Nanostructures

Fig. 6

(Continued)

‘‘onion’’-shaped NbS2 nanoparticle, produced by the reaction of Nb2 O5 with H2 S
(courtesy of Dr. M. Homyonfer, Weizmann Institute). The size and curvature dependence of the superconducting transition temperature (and the critical ﬁelds) is of fundamental importance in this case, and serves as a stimulus for these studies. The electronic structure of NbS2 nanotubes was calculated recently [82].
Optimized nanotube structures of these compounds are found to be stable and possess metallic behavior.
NiCl2 is a layered compound with CdCl2 structure, where the Ni layer is sandwiched between two chlorine layers and six Cl atoms surround each Ni atom in an octahedral arrangement. Strong ferromagnetic interactions occur between the Ni

atoms, orienting the magnetic dipoles in the a -b plane of the layer (⊥ c). Weak antiferromagnetic coupling between the Ni dipoles of adjacent layers lead to the antiferromagnetic coupling in this material with N´ el temperature of 51 K. Spherical e and polyhedral nanoparticles of NiCl2 and nanotubes thereof have been reported [83].
Such nanostructures cannot be antiferromagnetic because there is no atomic layer with the same number of atoms in these structures. Furthermore, closed polyhedral structures with an odd number of layers (1 and 3) have been synthesized, which cannot be antiferromagnetic. Unfortunately, the synthesis of large amounts of these nanostructures proved to be rather difﬁcult, mainly because of the hygroscopic nature of the compound.

3.3 Inorganic Nanoparticles with Fullerene-like Structure and Nanotubes

TEM micrograph of NbS2 nanoparticle with nested
(‘‘onion’’) fullerene-like structure (courtesy of
Dr. M. Homyonfer, Weizmann Institute). The layer to layer distance is 0.59 nm.

Fig. 7

CdCl2 and its ﬁrst hydrate CdCl2 •nH2 O have a layered structure. CdCl2 has a hexagonal structure with two CdCl2 layers in the unit cell, whereas the second is rhombohedral with three CdCl2 •nH2 O layers in the unit cell. Because of its hygroscopisity, the hydrate-free compound is not stable in ambient conditions. Careful drying of the powder produces the relatively stable CdCl2 •H2 O compound, although further hydration of the compound is inevitable in ambient conditions. Irradiation of this powder by the electron beam of a TEM leads to the loss of the water molecules and recrystallization to waterfree CdCl2 nanoparticles with closedcage polyhedral structures [84]. Similarly, nested CdCl2 structures were obtained by annealing CdCl2 •nH2 O in the oven at
750 ◦ C under argon ﬂow. Figure 8 shows

a typical hexagonal closed structure of
CdCl2 obtained by e-beam irradiation of the CdCl2 •H2 O precursor powder. Admittedly, the detailed structure of CdCl2 cages has not been unraveled as yet. Because bulk CdCl2 is extremely hygroscopic, it is impossible to handle the bulk material in the ambient atmosphere. In contrast, the fullerene-like structures are perfectly stable in the ambient, which again is a manifestation of the kinetic stabilization of the closed-cage structure.
BN and Bx Cy Nz nanotubes and fullerene-like structures have been synthesized by various laboratories in recent years. The most popular methods are the plasma arc and laser ablation techniques. The ﬁrst report on the synthesis of BN nanotubes, using the arc-discharge technique, was by the Zettl group [85, 86]. Because

255

256

3 Semiconductor Nanostructures
TEM micrograph showing a CdCl2 cage structure with four layers in the shell and hexagonal symmetry. The layer to layer distance is 0.58 nm [84].

Fig. 8

BN is an insulator, a composite anode was prepared from a tungsten rod with an empty bore in the center, which was stuffed with a pressed hexagonal BN powder. For the cathode, water-cooled Cu rod was used. The collected gray soot contained a limited amount of multiwall BN nanotubes. It is possible that in this case the tungsten also serves as a catalyst. By perfecting this method, macroscopic amounts of double-wall BN nanotubes of a uniform diameter (2 nm) were obtained in large amounts [14]. An alternative route employed HfB2 electrodes in nitrogen atmosphere [87]. This route led to the synthesis of BN nanotubes with varying number of walls, from a single-wall to multiple-wall nanotubes. The Hf was not incorporated into the tube and probably played the role of a catalyst. Using Ta instead of W as the metal anode, BN nanotubes with ﬂat heads, alluding to the existence of three
B2 N2 squares in the cap have been observed [88]. In another synthetic approach, pyrolysis of CH3 CNBCl3 complex in the presence of a Co catalyst provided Bx Cy Nz nanotubes and nanoﬁbers [89]. Clear evidence in support of the cap containing three B2 N2 squares was obtained. This is to be contrasted with carbon nanotube caps,

which contain six pentagons. Recently, long and quite perfect BN nanotubes were obtained by focusing a continuous CO2 laser onto hexagonal boron nitride powder in N2 gas atmosphere [90]. Furthermore, ropes consisting of hexagonal array of
BN nanotubes have been observed in this study, which is indicative of the uniform diameter (2 nm) of the nanotubes. The synthesis of Bx Cy Nz nanotubes and BN-cage structure has been intensively pursued by
Bando’s group (see for example Refs. 91,
92). Thus, a successful strategy for the synthesis of single- and multiple-wall Bx Cy Nz nanotubes through chemical substitution of carbon nanotubes has been demonstrated [91]. Here, single-wall carbon nanotube bundles were thermally treated with
BO3 at 1523–1623 K under a nitrogen gas ﬂow. The resulting nanotubes had diameters of 1.2–1.4 nm, which is similar to the precursor nanotubes. Electron beam irradiation of hexagonal boron nitride resulted in octahedral BN onions [92]. These structures are characterized by the presence of six B2 N2 squares embedded in the hexagonal BN network.
Concentric multiwall BN/C nanotubes were prepared. Here, a spontaneous segregation of nanotubes of a different chemical

3.3 Inorganic Nanoparticles with Fullerene-like Structure and Nanotubes

composition was observed. Thus, concentric carbon nanotubes followed by BN and again carbon nanotubes were identiﬁed [93]. The segregation of the carbon nanotubes in the inner end outer surfaces of the concentric nanotubes structure is attributed to the lower surface energy of graphite as compared with hexagonal
BN. Concentric BC2 N and carbon nanotubes have also been reported. Using laser ablation method, concentric CN and carbon nanotubes with SiO2 core have been prepared [94]. Built-in semiconducting junctions can be fabricated using the composite BN−C nanotubes. Silver nanoparticles encapsulated within boron nitride nanocages were produced by mixing boric acid, urea, and silver nitrate and reduction at 700 ◦ C under hydrogen atmosphere [95]. Generalization of this method to the encapsulation of other metallic nanoparticles was discussed.
As discussed earlier brieﬂy, semicrystalline or amorphous nanotubes can be obtained from 3D compounds and metals by depositing a precursor on a nanotube template intermediately, and subsequently removing the template by calcination. If the template molecules are not removed and they are able to effectively passivate the dangling bonds of the compound, a perfectly crystalline nanotube composite can be obtained. However, after hightemperature calcination, the organic scaffold is removed and the inorganic oxide remains. Because a nanotube is the rolledup structure of a 2D-molecular sheet, there is no way that all the chemical bonds of the 3D-inorganic compound will be fully satisﬁed on the nanotube inner and outer surfaces. Furthermore, the number of molecules increases with the diameter, and hence a full commensuration between the various molecular layers is not possible. Therefore, nanotubes of 3D

compounds cannot form a perfectly crystalline structure, and the nanotube surface is not going to be inert. Nonetheless, there are certain applications such as in catalysis, where such a high-surface area pattern with reactive surface sites, that is unsaturated bonds, is highly desirable. The ﬁrst report of SiO2 nanotubes [96] was observed serendipitously during the synthesis of spherical silica particles by the hydrolysis of tetraethylorthosilicate in a mixture of water, ammonia, ethanol, and tartaric acid. More recently, nanotubes of
SiO2 [59, 60], TiO2 [5, 97, 98], Al2 O3 , and
ZrO2 [5, 99], and so forth have been prepared by the self-assembly of molecular moieties on preprepared templates, which instigate uniaxial growth mode. One can distinguish between solid templates, such as carbon nanotubes, porous alumina, and soft templates, such as elongated micelles.
In fact, there is almost no limitation on the type of inorganic compound that can be
‘‘molded’’ into this shape using this strategy. A related synthesis of tubular β -Ag2 Se crystals has been described [100]. Here, hydrothermal reaction between AgCl, Se, and NaOH lead to the formation of tubular structures with a hexagonal cross section.
We note in passing that although the production of carbon nanotubes does not lend itself to an easy scale-up, the tunability of the carbon nanotube radii and the perfection of its structure could be important for their use as a template for the growth of inorganic nanotubes with a controlled radius. This property can be rather important for the selective catalysis of certain reactions, where either the reaction precursor or the product must diffuse through the (inorganic) nanotube inner core.
The rational synthesis of peptide-based nanotubes by self-assembling of polypeptides into a supramolecular structures was demonstrated. This self-organization leads

257

258

3 Semiconductor Nanostructures

to peptide nanotubes, having channels of
0.8 nm in diameter and a few hundred nm long [101]. The connectivity of the proteins in these nanotubes is provided by weak bonds such as hydrogen bonds. These structures beneﬁt from the relative ﬂexibility of the protein backbone, which does not exist in nanotubes of covalently bonded inorganic compounds.
3.3.4

Thermodynamic, Structural, and
Topological Considerations

The thermodynamic stability of the fullerene-like materials is rather intricate and far from being fully understood. IF structures are not expected to be globally stable, but they are probably the stable phase of a layered compound, when the nanoparticles are not allowed to grow beyond, say a fraction of a micron. Therefore, a narrow domain of conditions, where nanophases of this kind exist, is assumed.
The existence zone of the IF phase on the binary-phase diagram must be very close to the existence zone of the layered compound itself. This idea is supported by a number of observations. For example, the W-S phase diagram provides a very convenient pathway for the synthesis of IF-WS2 . The compound WS3 , which is stable below 850 ◦ C under excess of sulfur, is amorphous. This compound will therefore lose sulfur atoms and crystallize into the compound WS2 , which has a layered structure, on heating or when sulfur is denied from its environment. If isolated nanoparticles of WS3 are prepared and they are allowed to crystallize under the condition that no crystallite can grow beyond 0.2 mµ, fullerene-like WS2 (MoS2 ) particles and nanotubes will become the favored phase. This principle serves as a principal guideline for the synthesis of

bulk amounts of the IF-WS2 phase [63], and WS2 nanotubes in particular [7, 8].
Unfortunately, in most cases the situation is not as favorable and more work is needed to clarify the existence zone of the IF phase in the phase diagram (in the vicinity of the layered compound).
Another very important implication of the formation of nanoparticles with IF structures is that in several cases it has been shown that the IF nanoparticles are stable, but the bulk form of the layered compound is either very difﬁcult to synthesize or is totally unstable. The reason for this surprising observation is probably related to the fact that the IF structure is always closed and hence it does not expose reactive edges and interacts only very weakly with the ambient, which in many cases is hostile to the layered compound. For example, Na intercalated MoS2 is unstable in a moistured ambient because water is sucked between the layers and into the van der Waals gap of the platelet and exfoliates it. In contrast, Na intercalated IF-MoS2 has been produced and was found to be stable in the ambient or even as alcoholic suspensions [81]. Coaxial nanotubes of MoS2 and WS2 intercalated with Ag and Au atoms were recently reported [102, 103]. The analogous phase in the bulk material has not been reported so far, which is another demonstration for this point. Chalcogenides of the ﬁrst row of transition metals, such as CrSe2 and
VS2 , are not stable in the layered structure.
However, Na intercalation endows extra stability to the layered structure because of the charge transfer of electrons from the metal into the partially empty valence band of the host [104]. Thus, for example,
NaCrSe2 and LiVS2 form a superlattice in which the alkali metal layer and the transition metal layer alternate. The structure of this compound can be visualized akin to

3.3 Inorganic Nanoparticles with Fullerene-like Structure and Nanotubes

the layered structure of CrSe2 , in which the octahedral sites in the van der Waals gap between adjacent layers are fully occupied by the Na (Li) atoms. Notwithstanding, VS2 nanoparticles with a fullerene-like structure, that is, consisting of layered VS2 were found to be stable [81]. The unexpected extra stability of this structure emanates from the closed seamless structure of the
IF, which does not expose the chemically reactive sites to the hostile environment.
Similarly, γ -In2 S3 , which is unstable as a layered structure in the bulk (platelets), was also found to be stable in the IF form [81]. More recently, nanotubes of InS were obtained in a low temperature reaction between tributyl indium and H2 S in the presence of thiobenzene catalyst [105].
Until this work, the layered structure of
InS was not known. This work emphasizes again on the relative stability of the nanotubular structure in the presence of moisture and oxygen compared to the instability of the macroscopic 2D-crystalline form. Many compounds such as GaN come in more than one crystalline structure, of which the layered structure may be one [106]. Although the layered polymorph is not stable in ambient conditions, this phase can nevertheless be synthesized under extreme conditions, and subsequently rapidly quenched to ambient conditions, where sluggish kinetics will slow its transformation into the stable phase (wurzite).
On the same token, it is possible to assume that nanotubes and fullerene-like structures of GaN can be formed, for example by using similar strategies to the ones used for the BN synthesis. This idea opens new avenues for the synthesis of layered compounds, which could not be previously obtained or could not be exposed to the ambient, and therefore could only be studied to a limited extent. On the

other hand, this concept provides a vehicle for the study of nanotubular structures with interesting properties, which could not be anticipated before.
Many layered compounds come in more than one stacking polytype [107]. For example, the two most abundant polytypes of MoS2 are the 2H and 3R. The 2H polytype stands for a hexagonal structure with two S−Mo−S layers in the unit cell (AbA · · · BaB · · · AbA · · · BaB, and so forth). The 3R polytype has a rhombohedral unit cell of three repeating layers
(AbA · · · BcB · · · CaC · · · AbA · · · BcB · · ·
CaC, and so forth). In the case of MoS2 , the most common polytype is the 2H form but the 3R polytype was found, for example, in thin MoS2 ﬁlms prepared by sputtering [108]. The nanotubes grown by the gas-phase reaction between MoO3 and
H2 S at 850 ◦ C were found to belong to the 2H polytype [4, 107]. The same is true for WS2 nanotubes obtained from WO3 and H2 S [7, 8]. The appearance of the 3R polytype in such nanotubes can probably be associated with strain. For example, a ‘‘superlattice’’ of 2H and 3R polytypes was found to exist in MoS2 nanotubes grown by chemical vapor transport [50].
Strain effects are invoked in explaining the preference of the rhombohedral polytype in both MoS2 and WS2 microtubes grown in the same way [49]. These observations indicate that the growth kinetics of the nanotubes and of thin ﬁlms inﬂuence the strain-relief mechanism, and therefore different polytypes can be adopted by the nanotubes. The trigonal prismatic structure of MoS2 alludes to the possibility of forming stable point defects consisting of a triangle or a rhombus [58]. In the past, evidence in support of the existence of ‘‘buckytetrahedra’’ [110] and ‘‘bucky-cubes’’ [111], which have four triangles and six rhombi

259

260

3 Semiconductor Nanostructures

in their corners respectively, were found.
However, the most compelling evidence in support of this idea was obtained in nanoparticles collected from the soot of laser ablated MoS2 [22]. Detailed theoretical calculations indicate that rectangular and even octahedral elements are inherently stable in the nanotube tip [19].
Figure 9 shows the caps of MoS2 nanotubes with zigzag (a) and armchair (b) structure [19]. These drawings show that only small distortions of the Mo−S bond and the S−Mo−S dihedral angles are necessary to close the cap by three rectangles or octahedron (and four rectangles). Sharp cusps and even a rectangular apex were noticed in WS2 nanotubes [7, 8]. These features are probably a manifestation of the inherent stability of elements of symmetry

Caps of MoS2 Nanotubes

(a)

(b)

lower than pentagons, such as triangles and squares, in the structure of MoS2 , and so forth. Point defects of this symmetry were generally not observed in carbon fullerenes [112], most likely because the sp2 bonding of carbon atoms in graphite is not favorable for such topological elements. These examples and others illustrate the inﬂuence of the lattice structure of the layered compound on the detailed topology of the fullerene-like nanoparticle or of the nanotube cap obtained from such compounds. All Bx Cy Nz nanotubes are made of a hexagonal network of sp2 bonded atoms, with three nearest neighbors to each atom [23, 85, 86, 113–116]. In the case of BC2 N nanotubes, two different arrangement of the sheet are possible, leading to two isomers with different structure and distinct in their electrical properties [114].
The chemical composition of the IF phase deviates only very slightly, if at all, from the composition of the bulk layered compound. Deviations from stoichiometry can only occur in the cap of the nanotube.
In fact, even the most modern analytical techniques such as scanning probe techniques and high-resolution (spatial) electron energy loss spectroscopy are unable to resolve such tiny deviations from the stoichiometry, like the excess or absence of a single Mo (W) or S (Se) atom in the nanotube cap.
XRD studies have shown an expansion of 2–4% in the c-axis of multiwall IF structures (including inorganic nanotubes) [4,
21, 45, 46]. The shift to lower angles (larger
Computer-generated pictures of
(a) zigzag and (b) armchair nanotube caps [19]. Figure 9(c) and (d) are two arrangements of the MoS2 molecules in the caps of zigzag and armchair nanotubes, respectively.

Fig. 9

(c)

(d)

3.3 Inorganic Nanoparticles with Fullerene-like Structure and Nanotubes

c-axis spacing) for the IF phase compared to the bulk material is a clear distinction of this phase and serves as a quality measure for the synthetic process. The average size of the nanoparticles can be calculated from the peak width. Note that full commensuration between the upper and lower layer of the multiwall IF nanoparticle or nanotubes is not possible because the number of atoms in the upper layer is always larger than in the underlying layer. However, the structural relationship between the different layers, which is typical for the different polytypes, such as 2H and 3R, is preserved.
3.3.5

Physical Properties
Band Structure Calculations
Earlier, a few groups used powerful theoretical tools to calculate the stability, bandstructure, and other physical properties of Bx Cy Nz nanotubes and fullerene-like nanoparticles [23, 113–118]. A few striking conclusions emerged from these studies.
First, it was found that B−B and N−N nearest neighbors do not lead to stable polyhedral structures. Instead, dissimilar
B−N pair of atoms are thermodynamically preferred. This observation implies that B2 N2 squares, rather than the ﬁvemember rings found in carbon fullerenes and nanotubes, are preferred in the case of BN polyhedra and nanotubes. Experimental veriﬁcation for this hypothesis has been obtained in the work of a few groups [88, 119]. Secondly, in contrast to carbon nanotubes, which can be metallic or semiconducting depending on their chirality, all BN nanotubes were found to be semiconductors (insulators), independent of their chirality. Thirdly, whereas the smallest forbidden gap of zigzag (n, 0) nanotubes was found to be a direct ( ) one, an indirect band gap ( - ) is
3.3.5.1

calculated for the armchair (n, n) nanotubes. Bulk BN material has an indirect band gap of 5.8 eV. This is to be contrasted with carbon nanotubes, which are either metallic or semiconducting, depending on their (n, m) values. The fourth point to be noted is that the strain in the nanotubes scales as 1/D 2 , where D is the nanotube diameter. The strain effect is predominant for nanotubes with small diameters and therefore overwhelmingly, the band gap of inorganic nanotubes was found to decrease with a decreasing diameter of the (inorganic) nanotubes. In contrast to that, the band gap of semiconducting carbon nanotubes increases with a shrinking diameter of the cage. It should be furthermore emphasized that generically, the band gap of semiconducting nanoparticles increases with a decrease in the particle diameter, which is attributed to the quantum size conﬁnement of the electron wave function [120–122].
As mentioned in the previous section,
BC2 N nanotubes have two isomers with distinctly different structure and properties. One of them with alternating carbon and B−N chains was predicted to be a semiconductor. In the armchair (n, n) conﬁguration, the alternating conducting carbon and insulating B−N chains form a solenoid, which on proper doping can become a nanosize coil [113]. The electrical properties of BC3 nanotubes are largely inﬂuenced by their packing arrangements.
Theory shows [114] that concentric multiwall nanotubes of this kind are metallic, whereas isolated single wall nanotubes are a semiconductor.
Further work was carried out on nanotubes of the semiconducting layered compound GaSe [18]. In this compound, each atomic layer consists of a Ga−Ga dimer sandwiched between two outer selenium atoms in a hexagonal arrangement. This

261

3 Semiconductor Nanostructures

work indicated that some of the early observations made for BN and boro-carbonitride nanotubes are not unique to these layered compounds, and are valid for a much wider group of nanotubes. First, it was found that like the bulk material, GaSe nanotubes are semiconductors. Furthermore, the strain energy in the nanotube was shown to increase, and consequently the band gap was found to shrink as the nanotube diameter becomes smaller.
Recent theoretical work on (single wall) MoS2 and WS2 nanotubes [19,
123] conﬁrmed these earlier observations.
Although the lowest band gap of the armchair (n, n) nanotubes was found to be indirect, a direct transition was predicted for the zigzag (n, 0) nanotubes (Fig. 10).
Additionally, an 1/D 2 dependence of the strain energy versus diameter was observed for these nanotubes. The strain is about one order of magnitude larger in these nanotubes as compared to carbon or

BN nanotubes of the same diameter. The reason for this effect is the bulky nature of the triple S−M−S layer (vide infra) as compared to the one layer of carbon or BN in the respective nanotubes. In contrast to carbon nanotubes and with semiconductor nanoparticles in general, the band gap energy shrinks with a decreasing nanotube diameter. These ﬁndings, which can be attributed to both strain effects and Brillouin zone folding of the energy bands of the nanotubes suggest a new mechanism for optical tuning through strain effects in the hollow nanocrystalline structures of layered compounds. The existence of a direct gap in zigzag nanotubes is rather important because it suggests that such nanostructures may exhibit strong (electro) luminescence, which has not been observed for the bulk material.
The potential of synthesizing p-n or
Schottky junctions, which are a builtin unit of the nanotubes, is probably a

3

2

1

Energy
[eV]

262

0

−1

−2
Γ

Calculated band structure for WS2 (22,0) nanotube [82]. (Courtesy of
Prof. G. Seifert, Paderborn University.)

Fig. 10

X

3.3 Inorganic Nanoparticles with Fullerene-like Structure and Nanotubes

realistic proposition. There are various alternative ways to accomplish this task. One way is to convert the nanotube surface or tip into a metal. This can be achieved by various surface modiﬁcations (treatments), such as electrochemical deposition or doping. Another way is to attach a short organic moiety with a thiol head group and a metal at the rear, pointing outward. The self-assembly of the monolayer with loosely spaced metal atoms on the nanotube surface will endow a metallic conductivity to the surface ﬁlm, which is in an intimate contact with the WS2 nanotube, leading therefore to a Schottky junction. Another possible way to establish a junction is to synthesize the WS2 nanotube on top of a metallic carbon nanotube. Similarly, synthesizing composite WS2 (semiconductor)/NbS2 (metal) nanotubes, is feasible. The electronic structure of NbS2 nanotubes show that they possess a high density of states in the
Fermi level, irrespective of their chirality, which is indicative of their high electrical conductivity [82]. Therefore, the synthesis of complex nanotubes with, for example, semiconducting WS2 (MoS2 ) core and metallic NbS2 shell, is believed to be doable and is also rather intriguing. MoS2 monolayers were shown to exhibit a ﬁrst-order phase transformation from 2H into 1T polytype, where each Mo atom is surrounded by six S atoms in an octahedral fashion [124–126]. The 1T polytype has half-ﬁlled Mo d level and is consequently a metal. There are already some indications that the MoS2 octahedra produced by laser ablation [22] acquire the 1T (octahedral) arrangement of the S atoms around each
Mo atom, rather than the trigonal prismatic structure of bulk MoS2 [127]. Recent work also indicates that the tip of WS2 and
MoS2 nanotubes consist of half octahedra, in which case it is likely that the nanotube

body is a semiconductor, whereas the tip is essentially a metal. In this case, a built-in semiconductor (Schottky) junction could have been established inadvertently.
Black phosphorous (b-P) and arsine (As) crystallize in a layered structure, in which each atom is bound to three neighbors like atoms. In contrast to the ﬂat hexagonal network of carbon atoms in graphite, the
P-sheet (As) forms a puckered hexagonal honeycomb. To minimize the repulsive energy between the lone pairs of the two neighboring phosphorous atoms, they are arranged in opposite direction to each other in an alternating fashion. Using density functional tight-binding theory, the structure, electronic structure, and the mechanical properties of b-P nanotubes were derived [128–130]. The strain energy was found to scale as 1/D 2 , but is larger than that of carbon nanotubes of the same diameter D . The larger strain energy was attributed to the repulsion between the electron lone pairs of next nearestneighbor atoms. Six ﬁve-member rings were found to establish the most stable apex for the nanotubes. Here too, the energy gap was found to shrink with a decreasing D .
The electronic structure of GaN nanotubes was also calculated [106] and was essentially in accordance with the band structure calculations of the other inorganic nanotubes. The band gap of nanotubes with diameter >2 nm is above 4 eV, and it shrinks with the nanotube diameter.
Zigzag nanotubes are found to have a direct transition, which suggests that they could serve as an ultrasmall blue lightemitting source. The structure and stability of CaSi2 nanotubes have been investigated but few details are currently available [129,
130].
The transport properties of inorganic nanotubes have not been reported so far.

263

264

3 Semiconductor Nanostructures

A wealth of information exists for the transport properties of the bulk 2D-layered materials, which is summarized in a few review articles (see Refs. 107, 131).
Optical Studies in the UV and
Visible
Measurements of the optical properties in this range of wavelengths can probe the fundamental electronic transitions in these nanostructures. Some of the aforementioned effects have in fact been experimentally revealed in this series of experiments [132]. As mentioned earlier, the IF nanoparticles in this study were prepared by a careful sulﬁdization of oxide nanoparticles. Brieﬂy, the reaction starts on the surface of the oxide nanoparticle and proceeds inward, and hence the number of closed (fullerene-like) sulﬁde layers can be controlled quite accurately during the reaction. Also, the deeper is the sulﬁde layer in the nanoparticle, the smaller is its radius and the larger is the strain in the nanostructure. Once available in sufﬁcient quantities, the absorption spectra of thin ﬁlms of the fullerene-like particles and nanotubes were measured at various temperatures (4–300 K). The excitonic nature of the absorption of the nanoparticles was established, which is a manifestation of the semiconducting nature of the material (Fig. 11). In addition to the previous report [132], which included a detailed analysis of the A and
B excitons, the present spectrum reveals also the C exciton. Therefore, in accordance with the theoretical analysis [19, 82], the present work shows the semiconductor behavior of the fullerene-like particles.
This suggests that these new phases might be very suitable for photoelectrochemical and photocatalytic applications. The redshift in the exciton energy, which increased with the number of sulﬁde layers
3.3.5.2

of the nanoparticles, was established [132].
The temperature dependence of the exciton energy was not very different from the behavior of the exciton in the bulk material.
This observation indicates that the redshift in the exciton energy cannot be attributed to defects or dislocations in the IF material, but rather it is a genuine property of the inorganic fullerene-like and nanotube structures. In contrast to previous observations, IF phases with less than ﬁve layers of sulﬁde revealed a clear blueshift in the excitonic transition energy, which was associated with the quantum size effect. Figure 12 summarizes this series of experiments and the two effects. The redshift of the exciton peak in the absorption measurements, which is the result of the strain in the bent layer on one hand, and the blueshift for IF structures with very few layers and large diameter (minimum strain) on the other hand, can be discerned.
The WS2 and MoS2 nanotubes and the nested fullerene-like structures used for the experiments described in Figs. 11 and 12 had relatively large diameters
(>20 nm). Therefore, the strain energy is not particularly large in the ﬁrst few closed layers of the sulﬁde, but it nevertheless increases as the oxide core is progressively converted into sulﬁde, that is, closed sulﬁde layers of smaller and smaller diameter are formed. This unique experimental opportunity permitted a clear distinction to be made between the strain effect and the quantum size effect of the electronic wave function. In the early stages of the reaction, the strain is not very large and therefore the conﬁnement of the exciton along the c-axis is evident from the blueshift in the exciton peak. The closed and therefore seamless nature of the MS2 layer is analogous to an inﬁnite crystal in the a -b plane and hence quantum size effects in this plane can be ruled out.

3.3 Inorganic Nanoparticles with Fullerene-like Structure and Nanotubes
Absorption of IF-MoS2 (273 K)
5

C

4.5
4

Absorption

3.5
B

3

A

2.5

Series1

2
1.5
1
0.5
0
400

500

600

700

800

Wavelength
[nm]

(a)

Absorption of IF-MoS2 (176 K)
5
4.5
4

Absorption

3.5
3
2.5
Series1

2
1.5
1
0.5
0
400

(b)

500

600

700

800

Wavelength
[nm]

Absorption spectra of IF-MoS2 nanoparticles at 300 and 176 K. The position of the A,B, and
C excitons of the bulk 2H-MoS2 are marked with arrows.

Fig. 11

However, there is a clear conﬁnement effect observable perpendicular to the a b plane, that is, in the c-direction. The quantum size effect in layered compounds was studied in the past [133, 134]. The energy shift because of this effect ( Eg ) can be expressed as:

Eg =

h2
Lz 2
4µ

(1)

Here, µ is the exciton effective mass parallel to the c-axis and Lz is the
(average) thickness of the WS2 nested structure (Lz = n × 0.6.2 nm, where n

265

3 Semiconductor Nanostructures
100

50

Energy shift
[meV]

266

0

−50

−100

−150

0

5

10

15

20

25

Distribution of the number of sulfide layers
[n]
Fig. 12 The dependence of the A exciton shifts on the number of layers in the IF structure [132]. The error bar represents the distribution of the number of layers determined with TEM for each sample. The y-axis error bar is
±10 meV.

is the number of WS2 layers) in the nanoparticle. In a previous study of ultrathin ﬁlms of 2H-WSe2 , Eg of the
A exciton was found to obey Eq. (1) over a limited thickness range.
Eg
exhibited a linear dependence on 1/Lz 2 for Lz in the range of 4–7 nm and became asymptotically constant for Lz >
8 nm [133]. A similar trend is observed for IF-WS2 and MoS2 [132]. Therefore, the quantum size effect is observed for IF structures with a very small number of
WS2 layers (n < 5) and large diameter.
Note that in the current measurements, IF ﬁlms 150-nm thick were used, but because each IF structure is isolated and the exciton wave function cannot diffuse from one nanoparticle to the other, the quantum size effect can be distinguished in this case.

Note also, that because of the (residual) strain effect, the energy for both the A and
B excitons is smaller than for their bulk counterparts. The corresponding redshift in the absorption spectrum has also been found for MoS2 nanotubes [51].
These studies suggest a new kind of optical tunability. Combined with the observation that achiral inorganic nanotubes are predicted to exhibit direct optical transitions [19, 23, 82, 106], new opportunities for optical device technology, based on
GaN, or MoS2 nanotubes as light-emitting diodes and lasers, could emerge in the future. The importance of strong light sources a few nm in size in nanotechnology can be appreciated from the need to miniaturize current submicron light sources for lithography.

3.3 Inorganic Nanoparticles with Fullerene-like Structure and Nanotubes

Raman Spectroscopy
Raman and resonance Raman (RR) measurements of fullerene-like particles of
MoS2 have been carried out recently [135].
Using 488-nm excitation from Ar ion laser light source, the two strongest Raman features in the Raman spectrum of the crystalline particles, at 383 and 408 cm−1 , which correspond to the E2g 1 and A1g modes respectively (Table 1), were found to be dominant also in IF-MoS2 and in MoS2 platelets of a very small size.
A distinct broadening of these two features could be discerned as the size of the nanoparticles was reduced. In analogy to the models describing quantum conﬁnement in electronic transitions, it was assumed that quantum conﬁnement leads to contributions of modes from the edge of the Brillouin-zone. Thus, phonon modes with a high density of states in the edge of the Brillouin-zone are expected to have signiﬁcant contribution to the Raman spectra.
Lineshape analysis of the peaks led to the conclusion that the phonons are conﬁned
3.3.5.3

in coherent domains of about 10 nm in size within the IF nanoparticles. Such domains could be associated with the faceting of the polyhedral IF structures.
RR spectra were obtained by using the 632.8-nm (1,96 eV) line of a He-Ne laser. Figure 13 shows the RR spectra of a few MoS2 samples. Table 1 lists the peak positions and the assignments of the various peaks for the room temperature spectra. A few second-order Raman transitions were also identiﬁed. The intensity of the 226 cm−1 peak did not vary much by lowering the temperature, and therefore it cannot be assigned to a second-order transition. This peak was therefore attributed to a zone-boundary phonon, activated by the relaxation of the q = 0 selection rule in the nanoparticles. Lineshape analysis of the intense 460 cm−1 mode revealed that it is a superposition of two peaks at 456 and 465 cm−1 . The lower frequency peak is assigned to a second order 2LA(M) process, whereas the higher energy peak is associated with the A2u mode, which is

Raman peaks observed in MoS2 nanoparticle spectra at room temperature and the corresponding assignments. All peak positions are in cm−1

Tab. 1

Bulk
MoS2 12

PL-MoS2
˚
5000 A

PL-MoS2
˚
30 × 50 A

IF-MoS2
˚
800 A

IF-MoS2
˚
200 A

Symmetry assignment 177

179

180
226

180
227
248

A1g (M) − LA(M)
LA(M)

382
407
42110
465

384
409
419
460

381
408

378
407
Weak
452
495

179
226
248
283
378
406
Weak
452
496

526

529

572
599
641

572
601
644

545
565
591
633

543
563
593
633

455
498
545
∼557
595
635

E1g ( )
E2g 1 ( )
A1g ( )
2 × LA(M)
Edge phonon
E1g (M) + LA(M)
2 × E1g ( )
E2g 1 (M) + LA(M)
A1g (M) + LA(M)

267

268

3 Semiconductor Nanostructures

a

a

b b c

c

d d e

e

200
(a)

300

400

500

600

700

800

900

Raman shift
[cm−1]

200 300 400 500 600 700
(b)

800 900

Raman Shift
[cm−1]

Fig. 13 RR spectra excited by the
632.8 nm (1.96 eV) laser line at room temperature (a) and 125 K (b), showing second order Raman bands for several MoS2

nanoparticle samples [135, 136]: a. IF-MoS2
(20 nm); b. IF-MoS2 (80 nm); c. MoS2 platelets (5 × 30 nm2 ); 2H-MoS2 (500 nm);
e. 2H-MoS2 bulk.

Raman inactive in crystalline MoS2 , but is activated by the strong resonance Raman effect in the nanoparticles.

carbon nanotubes. The Young’s modulus of the b-P nanotubes was calculated [128].
The observed value, 300 Gpa, is 25% of the Young modulus of carbon nanotubes.
The Poisson ratio of b-P nanotubes was calculated to be 0.25 in this work.
An elastic continuum model, which takes into account the energy of bending; the dislocation energy, and the surface energy, was used to describe the mechanical properties of multilayer cage structures to a ﬁrst approximation [137].

Mechanical Properties
The mechanical properties of the inorganic nanotubes have only been investigated to a relatively small extent. The Young’s modulus of multiwall BN nanotubes was measured within a TEM [25], and was found to be about 1.2 TPa, which is comparable to the values measured for
3.3.5.4

3.3 Inorganic Nanoparticles with Fullerene-like Structure and Nanotubes

A ﬁrst-order phase transition from an evenly curved (quasi-spherical) structure into a polyhedral cage was predicted for nested fullerenes with shell thickness larger than about one-tenth of the nanotube radius. Indeed, such a transition was observed during the synthesis of IFWS2 particles [45, 46]. Initially the oxide nanoparticles enfolded with a few WS2 molecular layers were found to be quasispherical. They transformed into a very faceted structure, when the thickness of the sulﬁde shell in the nanoparticles exceeded a few nm. Further theoretical analysis of the mechanical and elastic properties of IF-MS2 onions has been undertaken recently [138]. First, by summing (integration) the interaction between the nanoparticles and the underlying substrate, the adhesion energy per unit area reads: u =
−A/12πd 2 (=100 erg/cm2 ) with A the
Hamaker constant (of the order 10−12 erg), and d - an atomic cutoff for the van der
Waals (vdW) interaction (0.165 nm). Considering a spherical fullerene-like particle of a radius R , the total adhesion energy of the nanoparticle with the substrate is: EA = −AR/6d . Taking R as 60 nm yields EA = 6.10−11 erg = 1400 kT. The adhesion energy between two particles of this radius is smaller by a factor of two only. This calculation alludes to the appreciable adhesion of the nanoparticles to the underlying substrate surface or to each other. The adhesion between the two MS2 layers in the ‘‘onion’’ can be calculated likewise and amounts to 106 kT. Indeed, it was found that IF-MoS2 nanoparticles form small clusters, which cannot be easily separated into isolated nanoparticles by ultrasonic treatment in various solvents. However, high resolution imaging of the nanoparticles by scanning probe microscopy techniques have eluded the experimenters. This fact was attributed to

an easy tip induced sliding and rolling of the nanoparticles on the underlying substrates. This fact cannot be easily reconciled with the calculated high-adhesion energies of the nanoparticle to the underlying substrate, or to the observed tendency of the IF nanoparticles to form stable clusters. Obviously, the IF nanoparticles can be solvated by water from the ambient, which could lead to a signiﬁcant reduction in the adhesion energy. This effect was not considered in the theoretical analysis [138].
This study has further indicated that deformation of the nanoparticles because of the adhesion or because of shearing forces of the ﬂuid is small and consequently delamination of the nanoparticles is not likely. However, this study also showed that, whereas small pressure leads to reversible deformations of the nanoparticles, strong pressure brings about an (buckling) instability, which will eventually result in their delamination. Clearly, more work is needed to understand the physical properties of the IF nanoparticles better.
3.3.6

Electrochemistry and
Photoelectrochemistry Using IF-Materials and Inorganic Nanotubes

The study of the electrochemical and photoelectrochemical behavior of the IF nanoparticles could not be done before sufﬁcient amounts of the nanoparticles were available. Recently, a few groups reported the successful synthesis of macroscopic amounts of these nanoparticles [16, 45, 46,
31], which makes such studies feasible.
Electrodeposition of IF-MS2
(M = W,Mo)Nanoparticles and their
Photoelectrochemical Properties
Sodium intercalated IF-MS2 powder was synthesized from sodium doped MO3
3.3.6.1

269

270

3 Semiconductor Nanostructures

oxide precursors [81]. Sonication of the products in alcoholic solutions led to the formation of stable suspensions. On the other hand, nonintercalated IF-MS2 powders did not form stable suspensions even after prolonged sonication, and precipitated after a short while [81]. These results indicate that the intercalation of alkali metal atoms in the van der Waals gap of the
IF particles led to a partial charge transfer from the alkali metal atom to the host lattice, which increased the polarizability of the nanostructures, enabling them to disperse in polar solvents. The transparency of the suspensions and their stability increased with the amount of alkali metal intercalated into the IF structures. Alcoholic suspensions prepared from IF powder (both fullerene-like particles and nanotubes), which contained large amount of intercalant (>5%), were found to be virtually indeﬁnitely stable. The optical absorption of the IF suspensions, measured in the solution, was found to be very similar to that of thin ﬁlms of the same nanoparticles [132].
In the next step, thin ﬁlms of IF nanoparticles were deposited onto a gold substrate by electophoretic deposition. Given the chemical afﬁnity of sulfur to gold, it is not surprising that electrophoretic deposition led to relatively well adhering IF ﬁlms. Furthermore, some selectivity with respect to the IF sizes and the number of MS2 layers in the ﬁlms was achieved by varying the potential of the electrode.
The thickness of the ﬁlm was controlled by varying the electrophoresis time. Because the nonintercalated IF particles do not form stable suspensions, ﬁlms of such material were obtained by electrophoresis from vigorously sonicated dispersions of the IF powder. Alternatively, ﬁlms of intercalated IF nanoparticles were prepared by evaporating the solvent from a dip-coated

metal substrate; this method, however, resulted in poorly adhering ﬁlms.
The optical absorption spectra of intercalated 2H-MoS2 did not show appreciable changes on alkali metal intercalation up to concentrations of 30%, where a transition into a metallic phase at room temperature and a further transition into a superconductor at about 3–7 K was reported [139,
140]. Because the concentration of the intercalating metal atoms in the IF nanoparticles did not exceed 10%, no changes in the optical transmission spectra were anticipated nor were they found to occur.
Also, the intercalation of alkali atoms in the IF particles induces n-type conductivity of the host.
The prevalence of dangling bonds on the prismatic faces of 2H-MS2 crystallites leads to a rapid recombination of the photoexcited carriers. Consequently, the performance of thin ﬁlm photovoltaic devices made of layered compounds has been disappointing. The absence of dangling bonds in IF material suggests that this problem could be alleviated here. Therefore, the photocurrent response of IF-WS2 ﬁlms in selenosulfate solutions (0.2 M Se and
0.4 M Na2 SO3 ) was examined and compared to that of 2H-WS2 ﬁlms. The response was found to be very sensitive to the density of dislocations in the ﬁlm.
Figure 14 shows the quantum efﬁciency
(number of collected charges and number of incident photons) of a typical IF-WS2 ﬁlm with a low density of dislocations, as a function of the excitation wavelength.
On the other hand, ﬁlms having nested fullerene-like particles with substantial amounts of dislocations exhibited a poor photoresponse and substantial losses at short wavelengths, which indicates that the dislocations impair the lifetime of excited carriers in the ﬁlm [141]. Films of

3.3 Inorganic Nanoparticles with Fullerene-like Structure and Nanotubes

10

Absorbance

Quantum efficiency
[%]

12

8
6

2
1.6
1.2
0.8
400

4

600

700

800

Wavelength
[nm]

2
0
400

500

500

600

700

800

900

Wavelength
[nm]
Photoresponse spectra of thin ﬁlms of IF-WS2 deposited on a gold substrate. The three transitions caused by excitons A, B, and
C are obvious in the ﬁgure. The absorption of a similar IF-WS2 ﬁlm deposited on a conductive (ITO) glass is shown in the inset [143].

Fig. 14

nanotubes also showed substantial photoresponse. Finally, ﬁlms made of 2H-WS2 platelets (each about 1 µm in size), which are known to have many recombination
¯
centers on the prismatic (1010) face, did not exhibit any measurable photoresponse under comparable conditions. The photocurrent decreased with negative bias, reaching zero at −1.0 V versus the Pt foil counterelectrode, thus afﬁrming the n-type conductivity of the alkali-metal intercalated
IF particles. The photoresponse of the IF ﬁlms did not show any degradation after
48 hours of continuous illumination. Electrolyte electrotransmission (EET) spectra of the IF-WS2 ﬁlms were also recorded and were in accordance with previous studies of 2H-WS2 crystals [142]. The inset of
Fig. 14 shows the absorption spectrum of the ﬁlms, which clearly reveals the direct excitonic transitions of the ﬁlm at 2.02 (A exciton), 2.4 eV (B exciton), and 2.9 eV (C exciton), respectively [143].
Limited photoelectrochemical measurements were done also on ﬁlms of SnS2 nanoparticles with fullerene-like structure.

Figure 15 shows the spectral response of ﬁlms consisting of IF-SnS2 nanoparticles with an oxide core, which were prepared by incomplete conversion of SnO2 nanoparticles into the respective sulﬁde [143].
An indirect transition of approximately
2.03 eV was obtained by extrapolation of the photocurrent1/2 versus photon energy.
This band gap is found to be in accordance with published data for bulk SnS2 [144].
The long wavelength tail extends to over
800 nm, beyond the bulk indirect band gap value, and is clearly a sub–band gap component. Sub–band gap photoresponse has been observed previously in semiconductor photoelectrodes, and was attributed to absorption in intra–band gap states.
The modest photoelectrochemical performance of the IF-based ﬁlms can be ascribed to a number of factors. Perhaps the most critical issue is the charge transfer across the nanoparticles boundaries.
The other issue is the charge transfer to the back contact itself. These issues will have to be studied in greater detail to identify the loss mechanisms and improve the

271

3 Semiconductor Nanostructures
0.03
0.2

(Q.E. %)1/2

0.025

Quantum efficiency
[%]

272

0.02

0.15
0.1
0.05

0.015
0
400

500

0.01

600

700

800

900

Wavelength
(nm)

0.005

0
400

500

600

700

800

900

Wavelength
[nm]
Photocurrent spectrum of a fullerene-like SnS2 ﬁlm on a Au substrate. Inset:
Plot of the (quantum efﬁciency)1/2 versus wavelength. The extrapolation to zero signal yields a band gap of 2.03 eV [143].

Fig. 15

light-induced charge collection efﬁciency of the ﬁlms.
Electrochemical Studies with V2 O5 and Metallic Nanotubes
The synthesis of V2 O2.45 (alkylamine) nanotubes, by a sol-gel reaction using vanadium oxide triisopropoxide in the presence of hexadecyl amine template and hydrothermal treatment was described [6, 31]. These nanotubes consist of concentric shells with alternating
V2 O2.45 (alkylamine) layers. Most of them are obtained as scrolls. Further studies [61] showed that the nanotube material could be used as a Li insertion electrode. To obtain the electrode material, the nanotubes were reﬂuxed, ﬁrst, for 24 hours in ethanolic solution of NaCl, in order to remove the hexadecylamine template.
Sodium ions were intercalated between
3.3.6.2

the atomic layers of the nanotubes. The reﬂuxed nanotubes were mixed (50%) with teﬂonized carbon black, and the mixture was pressed onto a titanium foil, which served as current collector. The electrochemical studies were carried out in a hermetically sealed three-electrode cell.
The anode and cathode compartments were separated by a glass ﬁber separator and the electrolyte was 1 M LiClO4 solution of propylene carbonate (PC).
The distance between the atomic layers in V2 O2.45 (TEMP), where TEMP =
C16 H33 NH2 , was found to be about 3 nm before the reﬂux. It decreased to only
0.65 nm after the amine was soaked out.
The surface area of the anode material was determined to be 35 m2 g−1 . Figure 16 shows the cyclic voltammogram of the electrodes with and without the template against Li+ /Li reference electrode. The

3.3 Inorganic Nanoparticles with Fullerene-like Structure and Nanotubes

cathodic current observed in potentials more negative than 2 V corresponds to
Li insertion into the template-containing electrode, which can be described according to the following reaction:
−→
XLi+ + x e− + VO2.45 (TEMP)0.34 − −
Lix VO2.45 (TEMP)0.34

(2)

The oxidation step starting in potential more anodic than 3 V corresponds to the dissolution of the Li ions from the nanotube material. The large potential difference between the charge and discharge potentials is indicative of sluggish kinetics or insertion and dissolution reactions. This phenomena stems probably from the slowness of the diffusion of the Li ion into and out of the conﬁned space between the nanotube walls. A speciﬁc charge and discharge of 120 mAhg−1 was obtained for the ﬁrst ﬁve cycles. However, the speciﬁc charge decreased with the number of cycles, probably because of an irreversible reaction of the amine, and amorphization of the nanotubes. When the

template was removed, the speciﬁc charge increased to 180 mAhg−1 . Furthermore, the process was quite reversible with relatively a small kinetic barrier, as indicated by the small potential difference between the cathodic and anodic peaks. However, the speciﬁc charge decreased relatively fast in the ﬁrst few cycles. The charge capacity leveled off after a few cycles and it reached the value of 100 mAhg−1 after 10 cycles.
Further work is needed to elaborate the mechanism of the insertion and discharge processes and to further improve the electrode stability.
Metallic nanotubes can be synthesized using hard or soft template. Notwithstanding their incomplete crystallinity, their high surface area can be exploited for various electrochemical reactions. The large surface area of the nanotube material may lead to improved kinetics of the electrochemical reactions and to a decrease in the losses due to overvoltage of the reaction. A novel utilization of this principle is demonstrated in Ref. 145. Here, palladium nanotubes, a few microns long were

60

Specific current
[mA g−1]

40
20
0
−20
With template
Template-free

−40
−60

1.5

2.0

2.5

3.0

3.5

Potential vs. Li/Li+
[V]
Cyclic voltammograms of the VO2.45 (TEMP)0.34 and template-free nanotubes in 1 M LiClO4 /polycarbonate electrolyte [61].

Fig. 16

4.0

273

274

3 Semiconductor Nanostructures

synthesized by ﬁrst electrodepositing Cu nanotubes on a perforated polycarbonate template. Following this step, electroless deposition of the Pd, which displaced the Cu nanotubes into solution, and dissolution of the polycarbonate led to a suspension of Pd nanotubes. These nanotubes were subsequently mixed with Ni slurry and used as the negative electrode in nickel-metal hydride battery. The Pd nanotubes contributed only 1% to the weight of the negative electrode. Several charge–discharge cycles of the battery were carried out and compared to the reference battery, which did not contain the
Pd nanotube in the cathode. The nanotube containing battery showed a substantially higher capacity compared to the reference battery. In another study [146], Co and Fe nanowires and nanotubes were directly electrodeposited inside the pores of a perforated polycarbonate template.
The nanotubes were 5–6 micron long and their diameter varied between 10–80 nm.
These studies indicate that metallic nanotubes may play an important role in future electrochemical devices of different kinds. 3.3.7

Applications

The spherical shape of the fullerene-like nanoparticles and their inert sulfurterminated surface suggests that MoS2 particles could be used as a solid-lubricant additive in lubrication ﬂuids, greases, and even in solid matrices. Applications of a pure IF-MoS2 powder could be envisioned in high-vacuum and microelectronic equipment, where organic residues with high vapor pressure can lead to severe contamination problems [147, 148].
Because the MoS2 layers are held together by weak van der Waals forces; they can

provide easy shear between two close metal surfaces, which slide past each other. At the same time, bulk MoS2 particles, which come in the form of platelets, serve as spacers, eliminating contact between the two metal surfaces and minimizing the metal wear. Therefore, MoS2 powder is used as a ubiquitous solid lubricant in various systems, especially under heavy loads, where ﬂuid lubricants cannot support the load and are squeezed out of the contact region between the two metal surfaces. Unfortunately, MoS2 platelets tend to adhere to the metal surfaces through their reactive pris¯ matic (1010) edges, in which conﬁguration they tend to ‘‘glue’’ the two metal surfaces together, rather than serve as a solid lubricant. During the mechanical action of the engine parts, abrasion and burnishing of the solid lubricant produces smaller and smaller platelets, increasing their tendency to stick to the metal surfaces through their reactive prismatic edges. Furthermore, the exposed prismatic edges are reactive sites, which facilitate chemical oxidation of the platelets. These phenomena adversely affect the tribological beneﬁts of the solid lubricant and lead to a relatively rapid disappearance of their beneﬁcial effects.
In contrast, the spherical IF-MS2 nanoparticles are expected to behave like nano–ball bearings and under mechanical stress they would slowly exfoliate or mechanically deform to a rugby-shape ball as indicated in
Ref. 138, but would not lose their tribological beneﬁts until they are completely gone, or oxidize. To test this hypothesis, various mixtures of the solid powder and lubrication ﬂuids were prepared and tested under standard conditions [149]. The beneﬁcial effect of IF powder as a solid lubricant additive has been thus conﬁrmed through a long series of experiments [149–152]; some of them are summarized in Table 2.
It has to be emphasized that the IF-solid

3.3 Inorganic Nanoparticles with Fullerene-like Structure and Nanotubes

lubricant seem to be particularly effective under heavy loads. The use of a pure lubricating ﬂuid without any solid lubricant, under these conditions, leads to a rapid wear and deterioration of the mating metal surfaces.
More recently, a number of studies have indicated that the IF material can serve as a dry solid lubricant [21, 153, 154].
In one study, IF-MoS2 were produced by the arc-discharge technique and collected on a Ti foil forming a thin ﬁlm of this material [21]. The IF-based ﬁlm exhibited very low friction coefﬁcients (>0.01), even under 45% humidity. Under similar conditions, a sputtered MoS2 ﬁlm exhibited a friction coefﬁcient greater than 0.1 and rapid wear.

The mechanism of the action of the IF nanoparticles as additives in lubrication ﬂuids is more complicated than was initially thought. First, it is clear that the more spherical the nanoparticles and the fewest structural defects they contain, the better is their performance as solid lubricant additives [152]. Three main mechanisms responsible for the onset of failure of the nanoparticles in tribological tests have been clearly identiﬁed. They include exfoliation of the nanoparticles; deformation into a rugby ball shape, and explosion. The partially damaged
(deformed) nanoparticles are left with reactive edges and dislocations, which can undergo further oxidation, and a complete loss of their tribological action. Recent

Wear (W) and friction (µ) coefﬁcients of a steel block in contact with a steel disk for four kinds of solids lubricants mixed with mineral oil. Average particle size shown in parentheses at the head of each column

Tab. 2

Exp.

Velocity
(m s−1 )

Load
(N)

Conc.
(wt%)

0.44

300

5

µ
W
Ra

1800

20
45
60
60

W
W
µ µ 60

W
T ◦C

a

b

0.22
600
300

c

0.11 to
0.44

300 to
3000

Coeff.

Pure oil 2H-MoS2
(4 µm)

0.07
1.6•10−8
0.83

2H-WS2
(0.5 µm)

2H-WS2
(4 µm)

IF-WS2
(120 nm)

0.07
1.3•10−8
0.75

0.05
1.5•10−8
1.18

0.03
0.7•10−8
0.53

2.3•10−1
0.043
0.042

5.1•10−1
1.3•10−1
0.034
0.028

4.6•10−4 2.8•10−4
88

1.9•10−4
72

7.9•10−1
3.9•10−1
0.10
0.067

Note: (a) In this experiment the sliding track length was 1.27•104 m (8 hr). A commercial oil for transmission systems (Delcol) was used. W is given in mm3 mm−1 N−1 . Proﬁles of the wear track were measured by stylus proﬁlometry. The results of the experiment with the powder of 2H-MoS2 with 0.5-µm grain size was very similar to the results of the 4-µm 2H-MoS2 powder and is therefore not reported in detail here. The average roughness (Ra ) of the area of the wear-track was determined after the experiment and is also included. Ra is given in µm.
(b) In this experiment, the sliding track length was 2.38•103 m (6 hr).
(c) In this experiment, the load on the disk was increased from 300 N to 3000 N in steps of 300 N, each step lasting 300s. At each load, the velocity was increased from 0.11 to 0.44 m sec−1 in steps of
0.11 m sec−1 (3.5 hr per experiment). The temperature of the ﬂat block was determined during the experiment by contacting a thermocouple to the block. Reported temperature is average after the initial run-in.

275

276

3 Semiconductor Nanostructures

nanotribological experiments using the surface force apparatus with the IFWS2 lubricant mixed with tetradecane between two perpendicular mica surfaces revealed that material transfer from the IF nanoparticles onto the mica surface plays a major role in reducing the friction between two mica surfaces [150, 151]. No evidence in support of a rolling friction mechanism could be obtained in these studies. On the other hand, 2H-WS2 platelets of a similar size exhibited poor tribological properties in similar experiments, and furthermore, no evidence for material transfer could be obtained in this case.
It was argued [138] that the shear forces provided by the surface force apparatus are below the threshold necessary to onset the rolling of the nanoparticles. However, the shear rates of the tribological tests reported in Refs. 149, 152 are sufﬁcient to induce rolling friction, according to the calculations. These experiments and many others carried out over the last few years, suggest an important application for these nanoparticles as an additive in lubrication ﬂuids or greases.
Self-lubrication of mechanical parts can alleviate some of the technological complexities involved in the lubrication by ﬂuids of mechanical systems, and the environmental impact of this technology.
For example, the use of ﬂuid lubricants in the automotive industry adds some
2–3% to the overall weight of the cars.
Furthermore, the technological complexities of these intricate systems lessens their reliability. The used ﬂuid lubricant must be processed or buried in special depots, which adds to the adverse effects of the automotive systems on the environment.
Recently, self lubricating bronze-graphite sliding bearings impregnated with 3 to
5% IF-WS2 nanoparticles, have been prepared and tested [153]. Figure 17 shows the

effect of the IF impregnation on the loading capacity of these bearings. Whereas the unlubricated bearings could not withstand loads larger than about 35 kg, the
IF impregnated bearings supported loads of over 90 kg before seizure. In addition, long-term tests showed that the lifetime of the self-lubricating bearing impregnated with the IF material can be extended by one to two orders of magnitudes.
The remarkable effect of the IF material has been attributed to the slow release of the IF nanoparticles, which reside in the porous matrix of the self-lubricating bearings, as demonstrated in Fig. 18. 2HWS2 (MoS2 ) platelets, which adhere to the metal surface, contribute very little to the self-lubrication. In contrast, the IF nanoparticles, which reside in the porous metal matrix, are slowly released to the surface and provide a very efﬁcient means of self-lubrication mechanism. Similarly, iron-graphite self-lubricating bearings impregnated with IF-WS2 nanoparticles were fabricated and tested [154]. These bearings are much harder and could carry much higher loads. Although the effect of the IF material was somewhat inferior in this case, substantial improvements in the loading capacity and the lifetime of the
IF-impregnated bearings was observed.
This series of studies suggest numerous applications for the IF materials in selflubricating systems.
Another important ﬁeld where inorganic nanotubes can be useful is as tips in scanning probe microscopy [24]. Here applications in the inspection of microelectronics circuitry have been demonstrated and potential applications in nanolithography are being contemplated. A comparison between a WS2 nanotube tip and a microfabricated Si tip indicates that while the microfabricated conical-shaped Si tip is unable to probe the bottom of deep and narrow

3.3 Inorganic Nanoparticles with Fullerene-like Structure and Nanotubes
Wear sensor
Load
Bronze-graphite
0.35

250

200

Steel disc

1
0.25

1′
150

0.2
2′

Temperature
[ °C ]

Friction coefficient

0.3

2
3′

0.15

0.1

100

3
50

0.05

0

30

39

48

57

66

75

84

93

0

Load
[kg]
Fig. 17 Friction coefﬁcient (1, 2, 3) and temperature (1 , 2 , 3 ) versus load (in kg) of porous bronze-graphite block against hardened steel disk (HRC 52). In these experiments, after a run-in period of 10–30 hours, the samples were tested under a load of 30 kg and sliding velocity of 1 m sec−1 for 11 hours. Subsequently, the

loads were increased from 30 kg with an increment of 9 kg and remained one hour under each load. (1, 1 ) bronze-graphite sample without added solid-lubricant; (2, 2 ) bronze graphite sample with 2H−WS2 (6%); (3, 3 ) the same sample with (5%) hollow (IF) WS2 nanoparticles [153].

Fig. 18 Schematic illustration of the wear mechanism of porous metal matrices impregnated by solid lubricant particles. The shaded areas are representative of the metal grains; the concentric circles represent the fullerene-like nanoparticles.

(a)

(b)

277

278

3 Semiconductor Nanostructures

grooves, the slender and inert nanotube tip can go down at least 1 µm deep and image the bottom of the groove faithfully [24].
This particular tip has been tested for a few months with no signs of deterioration, which is indicative of its resilience and passive surface. Other kinds of tips have been in use for high-resolution imaging using scanning probe microscopy in recent years. However, the present tips are rather stiff and inert and consequently they are likely to serve in high-resolution imaging of rough surfaces having features with large aspect ratio. Furthermore, inorganic nanotubes exhibit strong absorption of light in the visible part of the spectrum and their electrical conductivity can be varied over many orders of magnitude by doping and intercalation. This suggests numerous applications, in areas such as nanolithography, photocatalysis, and others.
3.3.8

Conclusion

Inorganic fullerene-like structures and inorganic nanotubes, in particular, are a generic structure of nanoparticles of inorganic layered (2D) compounds. Various synthetic approaches to produce these nanostructures are presented. In some cases, such as WS2 , MoS2 , BN, and
V2 O5 ,both fullerene-like nanoparticles and nanotubes are produced in gross amounts.
However, size and shape control is still at its infancy. Study of these novel nanostructures has led to the observation of a few interesting properties and some potential applications in tribology, high-energy density batteries, and nanoelectronics.
Acknowledgment

I am indebted to Dr. Ronit Popovitz-Biro for the assistance with some of the TEM

images and to Dr. Rita Rosentsveig for the synthesis of the IF-WS2 nanoparticles. This work was supported in part by the following agencies: Israeli Ministry of Science (Tashtiot program); USA-Israel
Binational Science Foundation; Israel science Foundation; Krupp von Bohlen and Halbach Stiftung (Germany); FranceIsrael R&D (AFIRST) Foundation; Israeli
Academy of Sciences (First program).
References
1. H. W. Kroto, J. R. Heath, S. C. O’Brein et al., Nature 1985, 318, 162–163.
2. S. Iijima, Nature 1991, 354, 56–58.
3. R. Tenne, L. Margulis, M. Genut et al., Nature 1992, 360, 444–445.
4. Y. Feldman, E. Wasserman, D. J. Srolovitz et al., Science 1995, 267, 222–225.
5. B. C. Satishkumar, A. Govindaraj, E. M.
Vogel et al., J. Mater. Res. 1997, 12,
604–606.
6. M. E. Spahr, P. Bitterli, R. Nesper et al.,
Angew. Chem. Int. Ed. 1998, 37, 1263–1265.
7. A. Rothschild, G. L. Frey, M. Homyonfer et al., Mater.Res.Innov. 1999, 3, 145–149.
8. A. Rothschild, J. Sloan, R. Tenne, J. Am.
Chem. Soc. 2000, 122, 5169–5179.
9. A. Rothschild, R. Popovitz-Biro, R. Tenne,
J. Phys. Chem. B 2000, 104, 8976–8981.
10. M. Remskar, Z. Skraba, M. Regula et al.,
Adv. Mater. 1998, 10, 246–249.
11. Y. Q. Zhu, W. K. Hsu, N. Grobert et al.,
Chem. Mater. 2000, 12, 1190–1194.
12. Y. Q. Zhy, W.K.Hasu, H.Terrones et al.,
J. Mater. Chem. 2000, 10, 2570–2577.
13. W. K. Hsu, B. H. Chang, Y. Q. Zhu et al.,
J. Am. Chem. Soc. 2000, 122, 10155–10158.
14. J. Cumings, A. Zettl, Chem. Phys. Lett. 2000,
316, 211–216.
15. J. Cumings, A. Zettl, Chem. Phys. Lett. 2000,
318, 497.
16. A. Zak, Y. Feldman, V. Alperovich et al.,
J. Am. Chem. Soc. 2000, 122, 11 108–11 116.
17. P. W. Fowler, D. E. Manolopoulos in An
Atlas of Fullerenes, Oxford University Press,
Cambridge, 1995.
18. M. Cote, M. L. Cohen, D. J. Chadi, Phys.
Rev. B 1998, 58, R4277–R4280.

3.3 Inorganic Nanoparticles with Fullerene-like Structure and Nanotubes
19. G. Seifert, H. Terrones, M. Terrones et al.,
Phys. Rev. Lett. 2000, 85, 146–149.
20. M. Remskar, private communication.
21. M. Chhowalla, G. A. J. Amaratunga, Nature
2000, 407, 164–167.
22. P. A. Parilla, A. C. Dillon, K. M. Jones et al.,
Nature 1999, 397, 114.
23. A. Rubio, J. L. Corkill, M. L. Cohen, Phys.
Rev. B 1994, 49, 5081–5084.
24. A. Rothschild, S. R. Cohen, R. Tenne, Appl.
Phys. Lett. 1999, 75, 4025–4027.
25. N. G. Chopra, A. Zettl, Solid State Commun.
1998, 105, 297–300.
26. L. Pauling, Proc. Natl. Acad Sci. 1930, 16,
578–582.
27. K. Kato, I. Kawada, T. Takahashi, Acta Crystallogr. 1977, B33, 3437–3443.
28. L. Guemas, P. Rabu, A. Meerschut et al.,
Mater. Res. Bull. 1988, 23, 1061–1069.
29. G. A. Wiegers, A. Meerschut, J. Alloys
Compd. 1992, 178, 351–368.
30. K. Suzuki, T. Enoki, K. Imaeda, Solid State
Commun. 1991, 78, 73–77.
31. F. Krumeich, H. -J. Muhr, M. Niederberger et al., J. Am. Chem. Soc. 1999, 121,
8324–8331.
32. S. Avivi, Y. Mastai, G. Hodes et al., J. Am.
Chem. Soc. 1999, 121, 4196–4199.
33. J. S. Kasper, P. Hagenmuller, M.
Pouchard et al., Science 1965, 150, 1713.
34. C. Cros, M. Pouchard, E. P. Hagenmuller,
J. Solid State Chem. 1970, 2, 570.
35. H. Kawaji, H. Horie, S. Yamanaka et al.,
Phys. Rev. Lett. 1995, 74, 1427–1429.
36. S. Saito, A. Oshiyama, Phys. Rev. B 1995, 51,
2628–2631.
37. E. Kaxiras, K. Jackson, Phys. Rev. Lett. 1993,
71, 727–730.
38. L. Zeger, E. Kaxiras, Phys. Rev. Lett. 1993,
70, 2920–2923.
39. D. Wichmann, K. Jug, J. Phys. Chem. B
1999, 103, 10087–10091.
40. N. N. Grennwood, A. Earnshaw in Chemistry of the Elements, Pergamon Press, Oxford, 1984.
41. R. Tenne, Adv. Mater. 1995, 7, 795–965.
42. Micelles, Membranes, Microemulsions and
Monolayers, (Eds.: W. M. Gelbart, A. BenShaul, D. Roux), Springer, New York, 1994.
43. W. Shenton, T. Douglas, M. Young et al.,
Adv. Mater. 1999, 11, 253–256.
44. S. W. Guo, L. Konopny, R. PopovitzBiro et al., Adv. Mater. 2000, 12, 302–306.

45. Y. Feldman, G. L. Frey, M. Homyonfer et al., J. Am. Chem. Soc. 1996, 118,
5362–5367.
46. Y. Feldman, V. Lyakhovitskaya, R. Tenne,
J. Am. Chem. Soc. 1998, 120, 4176–4183.
47. M. Remskar, Z. Skraba, F. Cl´ ton et al., e Appl. Phys. Lett. 1996, 69, 351–353.
48. M. Remskar, Z. Skraba, F. Cl´ ton et al., e Surf. Rev. Lett. 1998, 5, 423–426.
49. M. Remskar, Z. Skraba, C. Ballif et al., Surf.
Sci. 1999, 435, 637–641.
50. M. Remskar, Z. Skraba, R. Sanjin´ s et al., e Appl. Phys. Lett. 1999, 74, 3633–3635.
51. C. M. Zelenski, P. K. Dorhout, J. Am. Chem.
Soc. 1998, 120, 734–742.
52. C. R. Martin, Acc. Chem. Res. 1995, 28,
61–68.
53. H. Masuda, K. Fukuda, Science 1995, 268,
1466–1468.
54. P. Santiago, D. Mendoza, E. Espinosa et al., submitted. 55. J. Sloan, J. L. Hutchison, R. Tenne et al.,
J. Solid State Chem. 1999, 144, 100–117.
56. E. Iguchi, J. Solid State Chem. 1978, 23,
231–239.
57. W. B. Hu, Y. Q. Zhu, W. K. Hsu et al., Appl.
Phys. A. 2000, 70, 231–233.
58. L. Margulis, G. Salitra, R. Tenne et al., Nature 1993, 365, 113–114.
59. H. Nakamura, Y. Matsui, J. Am. Chem. Soc.
1995, 117, 2651–2652.
60. H. P. Lin, C. Y Mou, S. B Liu, Adv. Mater.
2000, 12, 103–106.
61. M. E. Spahr, P. Stoschitzki-Bitterli, R.
Nesper et al., J. Electrochem. Soc. 1999, 146,
2780–2783.
62. Y. Li, M. Sui, Y. Ding et al., Adv. Mater.
2000, 12, 818–821.
63. M. Harada, M. Adachi, Adv. Mater. 2000,
12, 839–841.
64. P. M. Ajayan, O. Stephan, P. Redlich et al.,
Nature 1995, 375, 564–567.
65. A. Leist, S. Stauf, S. L¨ ken et al., J. Mater. o Chem. 1998, 8, 241–244.
66. P. Afanasiev, C. Geantet, C. Thomazeau et al., Chem. Commun. 2000, 1001–1002.
67. M. Jos´ -Yacam´ n, H. Lopez, P. Santiago e a et al., Appl. Phys. Lett. 1996, 69, 1065–1067.
68. D. Ugarte, Nature 1992, 359, 707–709.
69. D. Vollath, D. V. Szabo, Mater. Lett. 1998,
35, 236–244.
70. D. Vollath, D. V. Szab´ , Acta Materialia o 2000, 48, 953–967.

279

280

3 Semiconductor Nanostructures
71. M. Homyonfer, Y. Mastai, M. Hershﬁnkel et al., J. Am. Chem. Soc. 1996, 118,
7804–7808.
72. Y. Mastai, M. Homyonfer, A. Gedanken et al., Adv. Mater. 1999, 11, 1010–1013.
73. K. S. Suslick, S. -B. Choe, A. A. Cichovlas et al., Nature 1991, 353, 414–416.
74. T. J. Mason, J. P. Lorimer, D. J. Walton, Ultrasonics 1990, 28, 333–337.
75. A. Durant, J. L. Deplancke, R. Winand et al.,
Tetrahedron Lett. 1995, 36, 4257–4260.
76. R. Tenne, M. Homyonfer, Y. Feldman,
Chem. Mater. 1998, 10, 3225–3238.
77. K. S. Suslick, D. A. Hammerton, R. E.
Cline, J. Am. Chem. Soc. 1986, 108,
5641–5642.
78. G. B. Saupe, C. C. Waraksa, H -N Kim et al.,
Chem. Mater. 2000, 12, 1556–1562.
79. R. Abe, K. Shinohara, A. Tanaka et al.,
Chem. Mater. 1997, 9, 2179–2184.
80. S. Avivi, Y. Mastai, A. Gedanken, J. Am.
Chem. Soc. 2000, 122, 4331–4334.
81. M. Homyonfer, B. Alperson, Yu. Rosenberg et al., J. Am. Chem. Soc. 1997, 119,
2693–2698.
82. G. Seifert, H. Terrones, M. Terrones et al.,
Solid State Commun. 2000, 115, 635–638.
83. Y. Rosenfeld Hacohen, E. Grunbaum, R.
Tenne et al., Nature 1998, 395, 336.
84. R. Popovitz-Biro, A. Twerski, Y. Rosenfeld
Hacohen et al., Isr. J. Chem., in press.
85. N. G. Chopra, J. Luyken, K. Cherry et al.,
Science 1995, 269, 966–967.
86. Z. W. Sieh, K. Cherrey, N. G. Chopra et al.,
Phys. Rev. B 1995, 51, 11229–11232.
87. A. Loiseau, F. Willaime, N. Demoncy et al.,
Phys. Rev. Lett. 1996, 76, 4737–4740.
88. M. Terrones, W. K. Hsu, H. Terrones et al.,
Chem. Phys. Lett. 1996, 259, 568–573.
89. M. Terrones, A. M. Benito, C. MantecaDiego et al., Chem. Phys. Lett. 1996, 257,
576–582.
90. T. Laude, A. Marraud, Y. Matsui et al., Appl.
Phys. Lett. 2000, 76, 3239–3241.
91. D. Golberg, Y. Bando, W. Han et al., Chem.
Phys. Lett. 1999, 308, 337–342.
92. D. Golberg, Y. Bando, O. St´ phan et al., e Appl. Phys. Lett. 1998, 73, 2441–2443.
93. K. Suenaga, C. Colliex, N. Demoncy et al.,
Science 1997, 278, 653–655.
94. Y. Zhang, K. Suenaga, S. Iijima, Science
1998, 281, 973–975.
95. T. Oku, T. Kusunose, K. Niihara et al.,
J. Mater. Chem. 2000, 10, 255–257.

96. W. St¨ ber, A. Fink, E. Bohn, J. Colloid o Interface Sci. 1968, 26, 62–69.
97. T. Kasuga, M. Hiramatsu, A. Hoson et al.,
Langmuir 1998, 14, 3160–3163.
98. T. Kasuga, M. Hiramatsu, A. Hoson, et al.,
Adv. Mater. 1999, 11, 1307–1311.
99. C. N. R. Rao, B. C. Satishkumar, A. Govindaraj, Chem. Commun. 1997, 1581–1582.
100. J. Hu, B. Deng, Q. Lu et al., Chem. Commun. 2000, 715–716.
101. M. R. Ghadiri, J. R. Granja, R. A.
Milligan et al., Nature 1993, 366, 324–327.
102. M. Remskar, Z. Skraba, R. Sanjines et al.,
Surf. Rev Lett. 1999, 6, 1283–1287.
103. M. Remskar, Z. Skraba, P. Stadelmann et al., Adv. Mater. 2000, 12, 814–818.
104. J. M. Vandenberg-Voorhoeve, Physics and chemistry of materials with layered structures, Optical and Electrical Properties, (Eds.:
P. A. Lee), D. Reidel Publishing Company,
Dordecht-Holland, 1976, p. 447, Vol. 4,
Ch. 8.
105. J. A. Hollingsworth, D. M. Poojary, A.
Clearﬁeld et al., J. Am. Chem. Soc. 2000,
122, 3562–3563.
106. S. M. Lee, Y. H. Lee, Y. G. Hwang et al.,
Phys. Rev. B 1999, 60, 7788–7791.
107. J. A. Wilson, A. D. Yoffe, Adv. Phys. 1969,
18, 193–335.
108. J. Moser, F. L´ vy, F. Bussy, J. Vac. Sci. e Technol. A 1994, 12, 494–500.
109. L. Margulis, P. Dluzewski, Y. Feldman et al., J. Microscopy 1996, 181, 68–71.
110. L. Margulis, R. Tenne, S. Iijima, Microsc.
Microanal. Microstruct. 1996, 7, 87–89.
111. R. Tenne, Adv. Mater. 1995, 7, 965–995.
112. W. Qian, M. D. Bartberger, S. J.
Pastor et al., J. Am. Chem. Soc. 2000, 122,
8333–8334.
113. Y. Miyamoto, A. Rubio, M. L. Cohen et al.,
Phys. Rev. B 1994, 50, 4976–4979.
114. Y. Miyamoto, A. Rubio, S. G. Louie et al.,
Phys. Rev. B 1994, 50, 18360–18366.
115. F. Jensen, H. Toftlund, Chem. Phys. Lett.
1993, 201, 95–98.
116. G. Seifert, P. W. Fowler, D. Mitchell et al.,
Chem. Phys. Lett. 1997, 268, 352–358.
117. K. Kobayashi, N. Kurita, Phys. Rev. Lett.
1993, 70, 3542–3544.
118. X. Xia, D. A. Jelski, J. R. Bowser et al., J. Am.
Chem. Soc. 1992, 114, 6493–6496.
119. O. St´ phan, Y. Bando, A. Loiseau et al., e Appl. Phys. A 1998, 67, 107–111.

3.3 Inorganic Nanoparticles with Fullerene-like Structure and Nanotubes
120. A. D. Yoffe, K. J. Howlett, P. M. Williams,
Cathodoluminescence studies in the SEM,
Scanning Electron Microscopy/1973 (Part II),
1973, 301–308.
121. D. A. B. Miller, D. S. Chemla, S. SchmittRink, in Optical Nonlinearities and Instabilities in Semiconductors (Eds.: H. Haug),
Academic Press, Orlando, FL, 1988, pp. 325.
122. M. L. Steigerwald, L. E. Brus, Annu. Rev.
Mater. Sci. 1989, 19, 471–495.
123. G. Seifert, H. Terrones M. Terrones et al.,
Solid State Commun. 2000, 114, 245–248.
124. X. R. Qin, D. Yang, R. F. Frindt et al., Phys.
Rev. B 1991, 44, 3490–3493.
125. S. J. Sandoval, D. Yang, R. F. Frindt et al.,
Phys. Rev. B 1991, 44, 3955–3962.
126. D. Yang, R. F. Frindt, Mol. Cryst. Liq. Cryst.
Sci. Technol., Sect. A 1994, 244, 355–360.
127. M. J. Hebben, B.A. Parkinson, private communication. 128. G. Seifert, E. Hern´ ndez, Chem. Phys. Lett. a 2000, 318, 355–360.
129. G. Seifert, T. Frauenheim, J. Korean Phys.
Soc. 2000, 37, 89–92.
130. G. Seifert, E. Hernandez, to be published.
131. E. Bucher, Photovoltaic properties of solid state junctions of layered semiconductors,
Physics and Chemistry of Materials with
Layered Structures, (Eds.: A. Aruchamy),
Kluwer Academic, Dordrecht, 1992, pp. 1–81, Vol. 14.
132. G. L. Frey, S. Elani, M. Homyonfer et al.,
Phys. Rev. B 1998, 57, 6666–6671.
133. F. Consadori, R. F. Frindt, Phys. Rev. B
1970, 2, 4893–4896.
134. M. W. Peterson, A. J. Nozik, Quantum size effects in layered semiconductor colloids,
Physics and Chemistry of Materials with
Layered Structures, (Eds.: A Aruchamy),
Kluwer Academic, Dordrecht, 1992,
297–317, Vol. 14.
135. G. L. Frey, R. Tenne, M. J. Matthews et al.,
J. Mater. Res. 1998, 13, 2412–2417.
136. G. L. Frey, R. Tenne, M. J. Matthews et al.,
Phys. Rev. B 1999, 60, 2883–2893.

137. D. J. Srolovitz, S. A. Safran, M. Homyonfer et al., Phys. Rev. Lett. 1995, 74, 1779–1782.
138. U. S. Schwarz, S. Komura, S. A. Safran, Europhys. Lett. 2000, 50, 762–768.
139. L. T. Chadderton, D. Fink, Y. Gamaly et al.,
Nucl. Instr. Methods Phys. Res. B 1994, 91,
71–77.
140. R. B. Somoano, J. A. Woollam, Intercalated
Layered Materials, (Eds.: F. L´ vy), D. Reidel e Publishing Company, Dordrecht, 1979, pp. 307–319.
141. H. J. Lewerenz, A. Heller, F. DiSalvo, J. Am.
Chem. Soc. 1980, 102, 1877–1880.
142. J. Bordas, Optical and Electrical Properties,
Physics and Chemistry of Materials with layered Structures, (Eds.: P. A. Lee), D. Reidel
Publishing Company, Dordrecht, 1976, pp. 145–230.
143. B. Alperson, M. Homyonfer, R. Tenne,
J. Electroanal. Chem. 1999, 473, 186–191.
144. G. Domingo, R. S. Itoga, C. R. Kannewurf,
Phys. Rev. 1966, 143, 143.
145. V. Badri, A. M. Hermann, Intl. J. Hydrogen
Energy 2000, 25, 249–253.
146. G. Tourillon, L. Pontonnier, J. P. Levy et al., Electrochem. Solid-State Lett. 2000, 3,
20–23.
147. I. L. Singer in Fundamentals of Friction:
Macroscopic and Microscopic Processes, (Eds.:
I. L. Singer H. M. Pollock), Kluwer, Dordrecht, 1992, p. 237.
148. F. P. Bowden, D. Tabor, in Anchor, Garden
City, New York, 1973.
149. L. Rapoport, Yu. Bilik, Y. Feldman et al.,
Nature 1997, 387, 791–793.
150. Y. Golan, C. Drummond, M. Homyonfer et al., Adv. Mater. 1999, 11, 934–937.
151. Y. Golan, C. Drummond, J. Israelachvili et al., Wear 2000, 245, 190–195.
152. L. Rapoport, Y. Feldman, M. Homyonfer et al., Wear 1999, 225-229, 975–982.
153. L. Rapoport, M. Lvovsky, I. Lepsker et al.
Adv. Eng. Mater. 2001, 3, 71–75.
154. V. Leshchinsky, E. Alyoshina, M. Lvovsky et al., submitted.

281

287

4.1

The Photoelectrochemistry of
Semiconductor/Electrolyte Solar Cells
Maheshwar Sharon
Indian Institute of Technology, Bombay, India
4.1.1

Introduction

Statistical assessment suggests that
12 000 kg of coal per capita per year is being used by developed countries, whereas
150 kg is being used by developing countries. It is estimated that in the last century, man has consumed the energy equivalent to 4 × 1021 J and in the next century would need energy equivalent to 100 × 1021 J.
The rate of use of fossil fuels (e.g. oil, gas, coal, etc.) has linearly increased with respect to utilization time. However, it cannot follow the same trend indeﬁnitely. To avoid an undesirable situation arising due to the shortage of fossil fuels, scientists all over the world, are trying to take advantage of renewable energy sources such as solar, wind, ocean, and so on.
The sun is the cleanest and most abundant energy source. It is estimated that out of the total solar energy (3.8 × 1020 MW), earth receives about 1.7 × 106 MW. Therefore, in recent years, attention has been paid to use solar energy for terrestrial

applications. One of the most important aspects in using solar energy is its conversion from solar radiation into electrical energy.
This is achieved by using semiconductorbased photovoltaic (PV) cells. Today’s PV market is 151 MW per year corresponding to a value of about 0.7 − 1 billion
US$. This is a remarkable market but still far away from being a noticeable contribution to the world energy consumption. The major reason for the low penetration of PV today is the high cost.
Broadly speaking, there are three types of
PV cells (1) n : p junction cell, (2) metalsemiconductor cell also known as MetalSchottky junction, and (3) semiconductorelectrolyte junction. This section discusses the last type of cell.
In 1839, Becquerel [1] ﬁrst discovered the PV phenomena in electrochemical systems. Brattain and Garret [2, 3] were pioneers explaining aspects of the properties of semiconductor–electrolyte interfaces.
Fujishima and Honda [4] reported the ﬁrst indication of a practical application of a photoelectrochemical (PEC) system in 1972. This paper sparked off a wave of investigations all over the world. It would be appropriate, however, to suggest that the interest in photoelectrochemistry of semiconductor blossomed only after the pioneering work of Gerischer [5] and
Myamlin and Pleskov [6]. These studies

288

4 Solar Energy Conversion without Dye Sensitization

led to the discovery of wet PV solar cell.
This is popularly known as a PEC cell. It was believed that a PEC cell might be more economically viable as compared with the solid-state PV cell. Simplicity in fabricating the cell was the main reason for such an optimistic view. Moreover, as discussed in a later chapter, it has been possible to make a rechargeable battery with in situ storage capability by using a PEC cell [7–9]. The wet type PEC cell suffers from the instability of semiconductor in aqueous media. It has been realized that in spite of its simplicity and economical viability, the wet-type PEC cell cannot easily replace a silicon PV solar cell, unless we discover photoelectrochemically stable semiconductor materials possessing band gap approximately 1.4 eV. Till such time, future of (unsensitized) wet-type PEC solar cell appears gloomy. However, research utilizing a standard PEC cell conﬁguration is not complex, facilitating testing of the photoactivity of new semiconductor materials. Current PEC cell research trends have also been directed toward sensitized semiconductors and applications in the ﬁeld of decontamination of water from pathogenic bacteria [10], detoxiﬁcation of water from toxic organic/inorganic materials, photography [11] or in any other ﬁeld where instability of the semiconductor does not pose a real problem.
The present section is devoted to various aspects that are essential for the development of a PEC cell.
4.1.2

Description of a PEC Cell

A PEC cell consists of a photoactive semiconductor electrode (either n- or p -type) and a metal counter-electrode. Both these electrodes are immersed in a suitable redox electrolyte. The semiconductor is

protected by insulation, so that only one of its surfaces is exposed to the redox electrolyte. In a regenerative PEC solar cell, the metal counter-electrode is expected to perform an electrochemical reaction that is the opposite of the process occurring at the semiconductor electrode. The counter-electrode should be electrochemically stable. It would be desirable that work function of counterelectrode is compatible with that of the
Fermi level of the semiconductor. The matching of work function may not be logical with metallic electrodes such as platinum, but may be important with semiconducting electrode such as SnO2 .
On illumination of the semiconductorelectrolyte junction with a light having energy greater than the band gap of the semiconductor, photogenerated electron/holes are separated in the space charge region. The photogenerated minority carriers (holes for n-type or electron for p -type semiconductor) arrive at the interface of the semiconductor–electrolyte.
Photogenerated majority carriers (i.e. electrons for n-type and holes for p -type semiconductor) accumulate at the backside of the semiconductor (i.e. the side that is not illuminated and not in contact with the redox electrolyte). With the help of a connecting wire, photogenerated majority carriers are transported to counter-electrode via a load (Fig. 1). These carriers at the counter-electrode electrochemically react with the redox electrolyte
(i.e. reduction of redox electrolyte occurs with n-type or oxidation of redox electrolyte with p -type semiconductor).
Conversely, the photogenerated minority carriers generated at the interface of the semiconductor perform the opposite reaction, which occurred at the metal counter-electrode. Thus, photogenerated carriers are consumed. Because of these

4.1 The Photoelectrochemistry of Semiconductor/Electrolyte Solar Cells
Load
A

(Semiconductor)
Working electrode
Light

Counter electrode

hν
Redox electrolyte

Quartz window

Fig. 1

A schematic diagram of a PEC cell.

two reactions, there is no net change in the electrolyte, and electrical power is produced in the load.
Details of the electrochemical reactions with respect to a PEC cell utilizing a p -type semiconductor dipped in
FeSO4 /Fe2 (SO4 )3 electrolyte are given as hν + p -semiconductor − − e− /h+
−→
(1)

Separation of charge carrier (e− /h+ ) caused by junction formation
(e− /h+ )depletion region − − e− + h+
− → surf bulk (2) h+ is transported to the counterbulk electrode via a load and oxidizes redox electrolyte, that is,
−→
h+ + Fe2+ − − Fe3+ bulk (3)

The corresponding chemical reaction is
2FeSO4 + 2h+ − − Fe2 (SO4 )3 , bulk − →

e− + Fe3+ − − Fe2+
−→
surf

(5)

Corresponding chemical reaction is
Fe2 (SO4 )3 + 2e− − − 2FeSO4 , (6)
−→
surf

At the p -semiconductor electrode
× (photogenerated carriers)

electrolyte to maintain the neutrality of the solution. At the counter electrode e− surf reduces the redox electrolyte, that is,

(4)

that is, near this electrode, one additional
−
SO4 − ion is required to be transported from the counter-electrode or the bulk

that is, at the counter-electrode, one
−
additional SO4 − ion is formed that needs to be transported to the semiconductor or to the bulk solution to maintain its neutrality. The reaction in this regenerative PEC is the conversion of photon energy into electrical energy without destroying chemical composition of the redox electrolyte or counter-electrode or semiconductor electrode, that is, hν (photon energy) → electrical energy.
Since in such a PEC cell, both electrodes are immersed in the same electrolyte and speciﬁc reactions occur only at the semiconductor and the metal, there is no need for a separator. Hence, charge balance due to oxidation and reduction processes is maintained.

289

290

4 Solar Energy Conversion without Dye Sensitization

4.1.3

Types of PEC Cells

Figure 2 shows various possible PEC [12] cells. In a PEC cell (or electrochemical
PV cell), optical energy is converted into electrical energy, with zero change in the free energy of redox electrolyte. In this cell, the electrochemical reaction occurring at the counter-electrode is opposite to the photoassisted reaction occurring at the semiconductor electrode (Fig. 3). Thus, the light energy is converted into electrical energy, with no change in the solution composition or electrode material. This is the case described earlier. In a photoelectrosynthetic cell, optical energy is converted into chemical energy with nonzero free energy change in the electrolyte. In this case, a net chemical change occurs

on illumination. This type of cell can be further classiﬁed as photoelectrolytic cell
( G > 0, Fig. 3) and photocatalytic cell
( G < 0, Fig. 3), depending on relative location of the potentials of the two redox couples (e.g. O/R and O /R ). In the latter type of cells, anodic and cathodic compartments need to be separated to prevent mixing of the two redox couples.
4.1.4

Aim of this Section

In this section, we shall be concerned with developing a regenerative PEC solar cell, where G = 0. From previous discussions, we understand how a PEC cell is fabricated. However, if we are to develop new materials for making an economically viable PEC cell, it is essential that we

Electrochemical photovoltaic cells
(Optical energy converted into electrical energy)

∆G = 0
Photoelectrolytic cells
(Optical energy stored as chemical energy in endorgic reactions
e.g. H2O
H2 + 1/2 O2)

Photoelectrochemical cells ∆G ≠ 0

∆G > 0

Photoelectrosynthetic cells
(Optical energy used to affect chemical reactions)
∆G < 0

Photocatalytic cells
(Optical energy provides activation energy for exorgic reactions
e.g. N2 + 3H2
2NH3)
Fig. 2

Classiﬁcation of PEC cells [8].

4.1 The Photoelectrochemistry of Semiconductor/Electrolyte Solar Cells e e
−

− e e

e

O
R

Electrochemical photovoltaic cell

O
R

∆G = 0

e
(a)

n -SC

+

soln

M

p -SC

e

O′
R′

(b)

−

e

soln

M

p -SC

soln M e −

O
R

Photocatalytic cell

e
O′
R′

O′
R′

O
R

e

∆G < 0

+

+

Fig. 3

∆G > 0

+

−

n -SC

Photoelectrolytic cell O
R

e

(c)

e
O′
R′

O
R
e

n -SC

M

e

−

+

soln

+

soln

M

p -SC

soln

M

A schematic representation of different types of PEC cells.

understand the principle, which controls the conversion of light energy into electrical energy. This knowledge will help us to design a desirable semiconductor. It is also necessary to realize that there are some constraints while employing a semiconductor and a redox electrolyte to make a
PEC cell. Therefore, knowledge regarding potential energy of redox electrolyte and its relation with the band positions of semiconductor is beneﬁcial. To accomplish this we also need to perceive a simple

technique to procure information about the band position of the semiconductor.
Finally, a simple experimental technique to maneuver the photocharacteristics of semiconductor-redox electrolyte is desirable. Unless, we have this information, it would be an extremely difﬁcult task to launch production of new materials for developing an economically viable and photoelectrochemically stable PEC cell. In the forthcoming sections, efforts are made to provide this information as brieﬂy as

291

292

4 Solar Energy Conversion without Dye Sensitization

possible because discussions on each of these items are dealt extensively in other sections of this book.
4.1.5

Semiconductor-electrolyte Junction

When a semiconductor comes in contact with another material of Fermi level, different from the semiconductor, a junction is formed. This can be formed between n-type and p -type semiconductors, or a metal and a semiconductor, or a semiconductor and a redox electrolyte.
The semiconductor-electrolyte junction should not be confused with a metalelectrolyte junction. In the metal-electrolyte junction, the potential drop occurs entirely on the solution side and practically nil on the metal side, whereas for semiconductor–electrolyte interface, the potential drop occurs on the semiconductor side as well as the solution side.
For clariﬁcation of the type of junctions formed at the semiconductor-electrolyte, let us take an example of n-type semiconductor. In addition to possessing free electrons (referred to as the majority carrier), n-type semiconductor also possesses holes (referred to as the minority carrier).
The concentration of holes is temperaturedependent and is equivalent to the intrinsic concentration of the carrier (which is related to the concentration of Frankel defects). It can be shown mathematically that the Fermi level of minority carrier lies at almost half the band gap position. On the other hand, the concentration of majority carriers as well as the Fermi level depends on doping concentration. Thus, the Fermi level of the majority carrier can lie anywhere between the conduction band edge and the intrinsic Fermi level that is situated at ∼ 1 Eg .
2
Similarly, the redox electrolyte can also be viewed as a material possessing

a speciﬁc Fermi level. For example, reduction of Fe3+ and Ce4+ requires 0.77 and 1.44 volts versus hydrogen electrode, respectively. Alternatively, we can say that the energy of electrons present in the reduced product of these materials must be equivalent to this potential. Hence, these potentials can be referred to as the work function (or Fermi level) of the redox electrolyte. Let us imagine a situation where an n-type semiconductor is brought in contact with a redox electrolyte (Fig. 4a). Let us also assume that the electrochemical potential, that is, its redox potential, is almost equal to the intrinsic Fermi level of the semiconductor. This situation forces the electrons (majority carriers) to ﬂow from the semiconductor to the electrolyte.
This migration continues until the two
Fermi levels achieve an equilibrium position (Fig. 4b). At this condition, no further migration of electron occurs and a dynamic equilibrium is established. In this situation, instead of electrochemically reacting with the redox electrolyte, these majority carriers accumulate at the interface of semiconductor–electrolyte to maintain neutrality of the material.
When electrons move toward the interface, lattice sites from where electrons are originated become positively charged
(Fig. 4b). Creation of positively charged carriers occurs randomly in the semiconductor (shown by a curved vertical line in
Fig. 4b), but are situated not very far from the interface. Any mathematical treatment for such random distribution or even approximated exponential distribution of charged ions becomes a complicated system to deal with.
Hence, a simpliﬁed mathematical model is visualized. It is assumed that all positively charged species instead of being scattered are situated in one hypothetical

4.1 The Photoelectrochemistry of Semiconductor/Electrolyte Solar Cells
Vacuum level before forming junction
−Ve

EA
+e
+e
+e

(a)

+e
+e
+e

+e +e
+e +e
+e +e

+e
+e
+e

+e
EF
+e
+e

electrolyte

m

Semiconductor

Va

O
−Ve

(b)

V

E Fredox

cu

level after forming junction

u

+
++
+ e e+ + +
+
e e+
+e +e +e + + e e +e + + + +
+e +e +e + +
+ ++ e e +e
+e +e + +

EA

E Fredox

+ wn

e +e +
+
+e +e +
+
e+ e +e +e +e + +
+
e+ e +e + +
+e +e +

E Fredox

+e

(c)

wn
++
+
+
+
Lp +
+

e+
+e +e +e + +
+e +e +e ee +e + +
(d)

+e
+e

−Lp −Wn

E redox

x
O

Schematic diagrams showing the formation of band bending at a semiconductor–electrolyte interface. (a) A semiconductor possessing its charge homogeneously distributed. +/e refers to a situation in which lattice sites are neutral. A horizontal line of vacuum scale showing no variation in energy of electrons of semiconductor.
(b) The contact of semiconductor with redox electrolyte disturbs the neutrality, and nonhomogeneous charge separation occurs. Vacuum scale takes the same shape as that of semiconductor to maintain its electron afﬁnity. (c) A hypothetical model showing charges retained in two parallels planes separated by a distance w. (d) Creation of diffusion region as a result of accumulation of minority carriers at plane situated at distance x = −w.

Fig. 4

293

294

4 Solar Energy Conversion without Dye Sensitization

plane at a distance ‘‘w ’’ from the interface
(Fig. 4c). We can get a better idea of this hypothetical plane by considering various positive charges q1 , q2 , q3 , . . . qn situated at x1 , x2 , x3 , . . . xn distances from the interface of the semiconductor. The total electrical ﬁeld created between the negative charges situated at the interface
(i.e. at x = 0) and positive charges situated at various distances would be sum of the product of charges divided by their respective distances.
That is, total electrical ﬁeld created
(F) =
≈

q1 e − q2 e− qn e −
+
+ ··· + x1 x2 xn Q+ Q− total total

(7) w where Q+ is total positive charge of total minority carriers, equivalent to the total charge of majority carriers accumulated at the interface (Q− ) and ‘‘w ’’ is the total distance of the hypothetical plane from the interface (x = 0), which contains Q+ total total charge, such that (Q+ Q− )/w = total total
F. It is worth noting that ‘‘w ’’ is not equal to the sum of all distances (i.e. w = x1 + x2 + x3 + · · · + xn ).
This hypothetical visualization of charge distribution near the interface (i.e. its presence in one plane) reveals the possibility of formation of charges in two parallel planes.
One plane appears at the interface (which is populated with the majority carriers at x = 0) and another at a depth w (which is populated with charges equivalent to total charges developed at x = −w ). The depth between these two planes, (Fig. 4c), is called the space charge width designated as wn (for n-type) and wp (for p -type). It should be understood that concentration of minority carriers at x = −wn would be much greater than the intrinsic concentration of hole at x ≈ −α . In this mathematical model, it is assumed that

all charges present within the distance of x = 0 and x = −w are regular and fully ionized. Moreover, within the space charge region, there are no free carriers. In other words, entire lattice sites falling within the space charge region are fully ionized.
It is also assumed that electrostatic ﬁeld becomes zero beyond x = −wn .
Should the changes in distribution of electrons (majority carriers) in n-type be abrupt at the interface as compared to the bulk or should it change linearly or exponentially? The most reasonable assumption would be that the change in concentration of majority carrier between the bulk and interface is exponential. This situation can be illustrated by an exponential bending in the conduction band (Fig. 4).
Similarly, the valence band must also bend in a similar fashion to maintain the difference between the conduction band and valence band, which is equal to the band gap of the semiconductor. Since difference between the Fermi level and conduction band is related to the magnitude of dopant’s concentration, this difference must be maintained even after the formation of band bending. However, it is difﬁcult to present it pictorially. Hence, we compromise by depicting it as a horizontal broken line.
The position of the conduction band on the vacuum scale also reﬂects the magnitude of electron afﬁnity of the semiconductor. This value is not altered because the semiconductor has come in contact with a redox electrolyte. Assuming the vacuum level to follow a similar bend as the conduction band solves this problem (Fig. 4). Considering all these factors, the entire energy diagram of the semiconductor after making its contact with redox electrolyte may be expressed as shown in Fig. 5. The variation in charges and potential within the space charge region are shown in Fig. 6.

4.1 The Photoelectrochemistry of Semiconductor/Electrolyte Solar Cells
Vacuum scale
−Ve
Dark

Illuminated

E

Ec
EF

nv

∆E N

Ev
Ev
Redox
(a) n -type semicond electrolyte

Light

N nE F

w
Eredox

E

−

Ec
E
N pE F

+

Eredox

+

Redox n -type semicond electrolyte

E
Ec

EF
Ev

E
−

Ec w Eredox

∆E
EF

−

N nE F

Eredox

N pE F

Ev

Vacuum scale

Light
+

Redox p -type semicond electrolyte

Redox
(b) p -type semicond electrolyte

Nature of band bending in the dark and under illumination with
(a) an n-type semiconductor-electrolyte and (b) a p-type semiconductor–electrolyte interface. Vacuum scale changes its shape to match the shape of band bending. Variation in Fermi level for minority carriers is shown by broken line (for illuminated semiconductor).

Fig. 5

+

Space charge layer Charge
−
Gouy layer

(a)

Schematic diagrams showing charge and potential distribution in an n-semiconductor–electrolyte interface.
(a) Variation in charge distribution within the space charge region and in
Gouy layer. (b) Variation in potential within the space charge region, and electrolyte. In Helmholtz layer it changes linearly, whereas in Gouy layer it decreases exponentially.

Fig. 6

Potential x −wn O

(b)

n -semiconductor Helmholtz layer Electrolyte

295

296

4 Solar Energy Conversion without Dye Sensitization

4.1.6

What Kind of Force Exists in this Region?

As per this model, minority carrier concentration at x = −wn would be larger than the intrinsic carrier concentration.
Moreover, it is also reasonable to assume that intrinsic carriers are homogeneously distributed in the bulk region. Therefore, the total magnitude of charge present at x = −wn must also decrease exponentially until its concentration becomes equal to the intrinsic equilibrium concentration
(Fig. 4d). Let us assume that at x = −Lp we establish such a position, then the depth between x = −wn and x = −Lp is called the diffusion region. In the diffusion region, thus we have an exponential concentration gradient established (Fig. 4d).
What are possible reasons that prevent majority carrier to increase to an inﬁnite value at x = 0? There must be an opposing force operating to establish an equilibrium condition. What is that force? When a semiconductor comes in contact with redox electrolyte, the majority carrier tends to move toward the interface. When these carriers move toward the interface, they create similar number of oppositely charged carriers (which can be termed as minority carriers). These minority carriers oppose the ﬂow of majority carriers toward the interface. Since no mobile carriers can be stationed in the space charge region, the concentration of minority carriers has to be established in the diffusion region only. Thus, we see that the force that allows establishment of an equilibrium condition (or the force that opposes the ﬂow of majority carrier toward the interface) is the force that allows the ﬂow of freshly generated minority carriers toward x = −α . The ﬂow of minority carrier is also referred to as drift ﬂow. When these two confronting forces become equal

in magnitude, an equilibrium condition is established. This condition limits the concentration of majority carrier at the interface. 4.1.7

Magnitude of Potential Developed at the
Interface

What would be the magnitude of potential developed between the charges accumulated at the interface (x = 0) and at plane x = −wn (or for p -type semiconductor at x = −wp )? The driving force for these carriers depends on the magnitude of the potential difference between the Fermi level of the semiconductor and the redox potential of the electrolyte (i.e. EF − EF(redox) ).
This potential is known as the contact potential (θ ). Can we fabricate a PEC cell, which gives a contact potential equal to the band gap value of the semiconductor? In other words, can we form a PEC cell with a semiconductor (whose Ec ≈ EF ) and redox electrolyte (whose Eredox ≈ Ev ), such that the contact potential (θ) = Eg ? The approximate Fermi level of the semiconductor (i.e. the intrinsic semiconductor) is approximately equal to half the band gap of the semiconductor (i.e. 1 Eg ). Therefore,
2
redox electrolyte cannot lower the Fermi level of n-type semiconductor beyond 1 Eg .
2
This condition puts a restriction to the maximum achievable contact potential (θ ), and is equal to 1 Eg value. This also sug2 gests that for a given semiconductor, the most suitable electrolyte would be the one that has a redox potential that is almost equal to the intrinsic Fermi level of the semiconductor. In conclusion, we see that when a semiconductor comes in contact with a redox electrolyte, two types of regions are formed: a space charge region of width
‘‘w ’’ and diffusion region of width ‘‘L,’’

4.1 The Photoelectrochemistry of Semiconductor/Electrolyte Solar Cells

in addition to its bulk region. In the bulk region, there is no accumulation of charge.
4.1.8

Effect of Junction Illumination

We shall now examine the effect of illumination of these three regions separately.
When photon of energy greater than band gap (known as band gap light) falls in the space charge region, electrons from the valence band get excited to the conduction band, leaving behind a positive charge (i.e. hole) in the valence band. These carriers experience an electrostatic potential in the space charge region. The magnitude of the electrical ﬁeld is very high in this region.
For example, with a contact potential of
˚
1 V and space charge width of 1000 A, an
5 V/cm is electrostatic ﬁeld of around 10 formed. This ﬁeld is strong enough to prevent the recombination of photogenerated electron/holes pairs in the space charge region. As a result, photogenerated holes migrate toward the interface and electrons migrate toward the diffusion region
(Fig. 5). When illumination is conducted in the diffusion region, electron/hole pairs also may be formed. In this region, there is only equilibrium drift current (established due to ﬂow of minority carrier toward the bulk of semiconductor) and no electrostatic ﬁeld. Therefore, photogenerated electrons/holes of diffusion region experience only the drift current as a driving force to prevent their recombination.
To maintain the equilibrium condition, electrons are directed to move toward the bulk and holes toward the interface. Drift current being not very strong, the majority photogenerated carriers in this region undergo recombination. Finally, if bulk region is illuminated, photogenerated electrons/holes would undergo 100% recombination, as this region has no force of any

kind to prevent them from recombining.
Thus, the net result of these illuminations is that photogenerated minority carriers accumulate at the interface and photogenerated majority carriers accumulate at the bulk of the semiconductor.
While photogenerated holes at the interface perform oxidation of the redox electrolyte, photogenerated electrons move via the load toward the counter-electrode to perform reduction of the redox electrolyte
(Fig. 1). Thus, electrical energy is created from light energy.
4.1.9

Efﬁciency of Conversion of Light Energy into Electrical Energy

After constructing a PEC cell, it is necessary to ﬁnd the efﬁciency of conversion of light energy into electrical energy. For this purpose, a PEC cell is connected with a resistance (as a load), variable
D.C. power source and an ammeter in series (Fig. 7a). A current-voltage characteristic is measured by applying different potentials to the PEC cell and measuring the corresponding current. This is called dark current (Fig. 7b). The PEC cell is then illuminated with a constant source of light. The current-voltage characteristic is again measured as before. This current is the photocurrent (Fig. 7b). It is observed that the dark current is constant in the reverse bias condition. This current is known as the saturation current
(I o ). The magnitude of the saturation curo rent (Iph ) increases when the PEC cell is illuminated. The difference between the two saturation currents is the photocurrent (Iph ) that is produced from the PEC cell. However, the efﬁciency of power conversion by the PEC cell is calculated from the current-voltage characteristics, plotted in the fourth quadrant. It is observed that

297

4 Solar Energy Conversion without Dye Sensitization
Schematic diagram for
(a) measuring current-voltage characteristics of a PEC connected in series with a load (R) and a variable dc power source and an ammeter (A).
(b) current-voltage characteristics of a
PEC cell under dark and illuminated condition (c) current-voltage characteristics (Fig. 7b) of a PEC cell replotted assuming its photocurrent to be a positive quantity.

Fig. 7

R

PEC
Cell
A

−Ve

inati on I

Illum

(a)

Da rk dc Power source V max

Io

V
I Ph

V oc

Imax

P max

Io
Ph
−I

(b)

I sc

I sc
FF = 1

P max

I max

Vmax
Photocurrent

298

Voc
(c)

Photovoltage

the maximum photopotential is obtained when photocurrent is zero. This potential is known as the open-circuit potential
(Voc ). Similarly, a maximum photocurrent is observed when photopotential is zero.
This photocurrent is known as short-circuit current (Isc ). One would get zero power by working at these values. To ﬁnd the maximum power, which can be drawn

from the PEC cell, a square or rectangle with maximum area is drawn (Fig. 7b,c).
The intercept of the square at the currentvoltage curve is the magnitude of maximum power (Pmax ) that can be drawn from the PEC cell. Similarly, the intercept on the potential axis is the magnitude of the maximum photopotential (Vmax ) and intercept on current axis is the value of maximum

4.1 The Photoelectrochemistry of Semiconductor/Electrolyte Solar Cells

photocurrent (Imax ) that can be drawn from the PEC cell. These characteristics are obtained when the PEC cell is connected in series with a load of resistance
(R ). To determine the maximum power load that can be connected to the PEC cell, the current-voltage curve is drawn for various resistances, and the corresponding
Pmax is determined. The resistance that gives the maximum Pmax is taken as the maximum load resistance (Rmax ) that can be connected in series with a PEC to drive the maximum power. The power efﬁciency of the PEC cell is calculated from the ratio of
Pmax to the power input (Pin ) to the cell.
The nature of the junction formed between the semiconductor and the electrolyte can also be obtained from the current-voltage characteristic (Fig. 7b,c). This is obtained from a factor known as ﬁll factor (FF), which is given by:
FF =

Imax Vmax
Isc Voc

(8)

FF decides how well the curve approximates a rectangle. In an ideal condition
FF = 1. Generally, current-voltage curve instead of plotting in fourth quadrant is plotted in the ﬁrst quadrant (Fig. 7c). Note, that the photocurrent, although shown as a positive current in ﬁgure, depends whether an n-type (i.e. negative current) or p -type
(i.e. positive current) semiconductor is utilized.
4.1.10

Factors Considered in Selecting a
Semiconductor

From the previous discussions, it is clear that material for a PEC cell must be able to absorb band gap light in the space charge region. In other words, the reciprocal of the absorption coefﬁcient (i.e. 1/α ) of band gap light preferably should be equal to the

space charge width (w ). While the material should have large space charge width
(≈1/α ) to ensure entire absorption of band gap light in the space charge region, the diffusion region should also be large. This is because the magnitude of photocurrent ﬂowing through the semiconductor depends on the width of the diffusion region.
It will be seen in the forthcoming section that concentration of dopant as well as the redox electrolyte forming the junction can control the width of the space charge region. Moreover, the FF should also approach to unity. Both the space charge width and the FF depend mainly on quality of semiconductor-electrolyte junction.
However, width of the diffusion region is the speciﬁc property of the material. Furthermore, the contribution of redox electrolyte in giving a better efﬁciency of a PEC cell is signiﬁcant and is later discussed.
4.1.11

Energy Levels in Redox Couples
4.1.11.1 How Do We Compare Redox
Potential with Fermi Level of a
Semiconductor?
At the gas-solid interface, the vacuum level is taken as the reference energy level; the energy of an electron in vacuum is taken as zero (i.e. Evaccum = 0). The energy of a bound electron in the solid is referred as a negative value in this reference scale. In electrochemistry, the standard zero level of the energy is the potential of hydrogen ions (at unit activity), which is in equilibrium with hydrogen gas at one atmosphere pressure. The electrode potential of any other redox electrolyte is referenced to this potential.
With this hydrogen scale, the value of redox potential can be either positive or negative. Conventionally, positive redox potentials are shown below zero, whereas

299

300

4 Solar Energy Conversion without Dye Sensitization
(ev)
Vacuum

0.0
−0.2
−0.2
−4.5

0.0 (NHE)
+0.2

(SCE)

+0.4
−5.1

Ec

+0.6
Fe+2/Fe+3
+1.0

−5.5

+1.2

Ev
(a)

n -semiconductor

Electrolyte

E (ev)
Density of state
+1.0

W

re

d

(E

)

+0.5

E redox

E ox (Unoccupied level)

2λ

0.0

E red (occupied level)

−0.5 x Wo

)
(E

−1.0

(b)

Semiconductor

Electrolyte

Comparison of energy levels of redox electrolyte with that of semiconductor.
(a) Band diagram of a semiconductor (in vacuum scale) and redox potential of electrolyte
(in normal hydrogen scale (NHE)). Redox potential is shown to possess one value for its oxidized/reduced species. (b) Energy levels of redox electrolyte showing its distribution and variation in energy for oxidized and reduced levels.

Fig. 8

4.1 The Photoelectrochemistry of Semiconductor/Electrolyte Solar Cells

negative redox potentials are shown above the zero (Fig. 8a).
When dealing with semiconductor-electrolyte junction, we have to consider both of these energy scales. Therefore, there is a need to ﬁnd out the relationship between them, so that one can draw useful conclusion from the band diagram of the semiconductor–electrolyte interface (Fig. 8a).
Considering the free energy changes associated with the reduction of silver, one can derive the relationship between
Evacuum and the NHE. The energy involved with various reactions can be presented by considering the following four steps [8]:
Ag+ ) + e−
(g
Ag(g)
Ag+ )
(aq
Ag(s) + H+ )
(aq

= Ag(g)
= Ag(s)
= Ag+ )
(g
= Ag+ ) + 1 H2
(aq
2

H+ + e−

=

1
2

−7.64 eV
−2.60 eV
+4.96 eV
+0.80 eV

−4.48 eV
(9)
In this analysis, ‘‘g’’ indicates the gas phase, ‘‘s’’ the solid, and ‘‘aq’’ the aqueous solution phase. Thus, the relation between the two scales, namely, Evacuum and NHE can be represented as:
H2

o
EF(redox) (eV) = Eredox(NHE) (volt)

−4.48 (eV)
(10)
and EF(redox) are the potentials of a redox couple on the electrochemical
NHE scale and vacuum scale, respectively. o Eredox(NHE)

4.1.11.2 Do the Energy Levels for
Reduction and Oxidation Process Possess the Same Value?
The energy level of the redox electrolyte shown in Fig. 8(a) is an oversimpliﬁed model because it indicates that the energy levels for the reduced and oxidized species of the redox couple, Fe2+/3+ , are the same.
This also means that if electrons were to be transferred from electrolyte to the

semiconductor and back, it would come back to the same energy level of the electrolyte. Since the charge density on ions is different, the extent of hydration differs. Therefore, it has been proposed that the energy states of reduced and oxidized species should not be discrete but are broadened into thermal distribution of states caused by thermal ﬂuctuation, interaction with solvents, and interaction with directly bonded ligands. The position of the localized energy states in a polar solvent depends greatly on whether the state is occupied or not because of the strong interaction with the solvent.
Fluctuation in the structure of hydrated ions thus extends the energy of the electronic quantum states in the redox couple around two most probable values.
The energy change caused by the variation in solvation shell can be represented in terms of harmonic oscillators. On the basis of the concepts proposed by Levich [13] and the distribution function proposed by
Gerischer [14–16] a model for the energy levels in solution is discussed in this section. In this model, the ground state vibration level of only the ﬁrst electronic state is considered relevant to electron transfer processes at the electrode. The energy change caused by vibrations in the solvent shell is represented in terms of harmonic oscillations. The potential energy is considered to vary around the ground state to an equal degree.
The resulting distribution functions for oxidized and reduced species are given as:
Wox (E) =

Wred (E) =

o
[exp{−(E − Eox )2 /4πkT λ}]
(4πkT λ)1/2
(11)
o
[exp{−(E − Ered )2 /4πkT λ}]

(4πkT λ)1/2

(12)

301

302

4 Solar Energy Conversion without Dye Sensitization o o where Eox and Ered are the most probable energies for the unoccupied and occupied quantum states in solution, respectively.
The parameter λ represents the activation energy for the process of transferring the solvation shell structure from equilibrium condition of one species to the most probable structure of the other. This is also called reorientation energy. The density of states for reduced and oxidized species at a particular energy are given as

Dred (E) = Cred × Wred (E)

(13)

Dox (E) = Cox × Wox (E)

(14)

where Cred and Cox are the concentrations of the reduced and oxidized species, respectively. The energy distribution is a Gaussian bell shaped curve (Fig. 8b).
In the harmonic oscillator approximation, the reorientation energy of reduced and oxidized species are equal and are given by o o o Eredox = Eox − λ = Ered + λ

···
(15)

or, o o
Eox − Ered = 2λ

···

(16)

Bockris and Khan [17], however, have suggested that energy is not symmetrically distributed around the central point (i.e. the ground state) but in a Maxwell type distribution. The most interesting concept derived from these views is that like a semiconductor, the oxidized and reduced species of redox electrolyte are linked with a conduction band (unoccupied electronic state) and a valence band (occupied electronic state). The energy necessary to transfer an electron from reduced to oxidized state is
2λ, analogous to the band gap of a semiconductor Eg .
For an effective charge transfer to occur, the energy level of the semiconductor

must match with energy level of the redox electrolyte. For example, an electron transferred from the valence band to redox electrolyte would be facilitated if the level of the former matches the oxidized level of the later. Alternatively, if electrons are to be transferred to the conduction band from the redox electrolyte, the conduction band must match the reduced level of the redox electrolyte. Thus, knowledge of such distribution functions of the redox electrolyte helps in understanding the I–V characteristics of a semiconductor, especially when the λ value of the electrolyte approaches the material’s band gap. Under such a condition, the rectifying character of semiconductor-electrolyte junction does not control the nature of anodic and cathodic current, that is, an n-type semiconductor would give a cathodic current in addition to the anodic current, and the reverse would be true for a p -type semiconductor. With this type of condition, one either has to select an electrolyte whose redox potential (i.e. its Fermi level) is such that only one of the two levels of the redox o o electrolyte (Eox or Ered as the requirement may be) matches the concerned band edge level of the semiconductor. Alternatively, bias would have to be applied to the electrode to stop the unwanted current.
4.1.11.3 Effect of Charge Distribution in the Electrolyte
What is the role of the constituents of the redox electrolyte present at the interface of the semiconductor-electrolyte junction? Choice and optimization of the electrolyte has a substantial effect on PEC solar cells [18–19]. Before a semiconductor is immersed in an electrolyte, anions and cations freely and randomly move in the solution. As a result of this movement, no speciﬁc spatial accumulation of ions occurs in the solution. This situation alters

4.1 The Photoelectrochemistry of Semiconductor/Electrolyte Solar Cells

as soon as we insert a semiconductor into the electrolyte. Because of the difference between the redox potential of the electrolyte and the Fermi level of the semiconductor, accumulations of the majority carriers occur at the interface of the semiconductor. These majority carriers attract
(hydrated) cations toward the interface.
Since they may be hydrated, instead of directly coming in contact with the semiconductor, they may approach the semiconductor spaced by water molecule. This situation creates a double layer at the interface: a negative charge at the interface of an n-type semiconductor and a positive charge
(i.e. cation) separated by a water molecule, in the solution. Interface of the semiconductor that contains the majority carrier can be assumed as one plane possessing these charges. At a water molecule distance from the interface, we have another plane carrying the positive charge of cations. This plane is termed as Helmholtz plane. This double layer (known as Helmholtz double layer ) thus develops an electrostatic potential. Because of formation of the double layer, as compared to the bulk solution, a large concentration of cations thus gets accumulated at the interface. In a dilute solution, however, it is observed that the concentration of cations (i.e. after the double layer) decreases exponentially to a point where there is no accumulation of charge and the solution behaves almost like the bulk solution. The region between the Helmhotlz plane and the plane where there is no accumulation of charge is called the Guoy-Chapman diffusion region (Fig. 6a). The variation in potential in
Helmholtz region is expected to be linear, but it is exponential in Guoy-Chapman region (Fig. 6b). Thus, as with a semiconductor, we can visualize the formation of three charge regions in the solution: the
Helmholtz region, Gouy-Chapman region,

and the bulk region, (where there is no accumulation of charge).
4.1.12

Capacitance of the Space Charge Layer

What is the role of the various regions formed on the property of the semiconductor-electrolyte junction? Can we take some advantages from them to procure some useful parameters of the semiconductor-electrolyte junction? We shall examine this brieﬂy.
It has been seen earlier that knowledge of band edge positions of the semiconductor and distribution of energy in the redox system are helpful in selecting a suitable redox electrolyte to get an efﬁcient charge transfer across the semiconductor–electrolyte interface. The band edge positions of a semiconductor can be determined if we know the value of the ﬂat band potential (i.e. Fermi level position) and the band gap of the semiconductor. The ﬂat band position is equivalent to the magnitude of the biasing potential applied to the semiconductor so as to make the space charge region (i.e. depletion region) width
(wn for n-type semiconductor) zero. This potential is equivalent to Fermi level of the semiconductor. One of the easiest methods to determine the ﬂat band potential is by studying a differential capacitance of the semiconductor–electrolyte interface.
We have seen that a semiconductorelectrolyte junction possesses three regions in the semiconductor and three in the electrolyte. These layers act as parallelplate capacitors in series, so that the resultant reciprocal capacitance is the sum of the reciprocal of the capacitance for all the capacitors.
1/C(total) = 1/CSC + 1/Cdiff
+ 1/CH + 1/CG

(17)

303

304

4 Solar Energy Conversion without Dye Sensitization

Capacitance due to the bulk region in an electrolyte and a semiconductor are overlooked because there is no charge accumulation in these two regions. It has also been experimentally observed that magnitude of the capacitance of the diffusion layer in semiconductor (Cdiff ), GuoyChapman layer (CG ), and the Helmhotlz layer (CH ) are very high compared to that of the space charge layer (CSC ). Under this condition, the reciprocal of capacitance of the space charge layer becomes equivalent to reciprocal of total capacitance of the semiconductor-electrolyte junction (i.e. 1/C(total) = 1/CSC ). Hence, if we can measure the total capacitance formed with the semiconductor-electrolyte junction, understanding of the capacitance of space charge layer can be obtained.
Measurement of the potential drop across the space charge layer yields information about the capacitance formed with the semiconductor-electrolyte junction; this capacitance and its measurement are described in a separate chapter.
4.1.13

Semiconductor-electrolyte Junction Under
Illumination

In Sect. 4.4, we examined the effect of the illumination of semiconductor-electrolyte junctions, using a light source of photon energy greater than the band gap of semiconductor. We shall now try to develop a model to get a quantitative estimation of the photocurrent from the magnitude of the concentration of photogenerated carriers. ..
G artner’s Model
Wolfgang G¨rtner [20] derived a relation a for photocurrent generated in space charge region and diffusion region after illuminating a semiconductor-metal
4.1.13.1

junction. Since, semiconductor–metal interface behaves similar to semiconductorelectrolyte junction, G¨rtner’s model has a been successfully applied to PEC cell, as well [21–22]. The photocurrent (Jscl ) due to carriers generated in the space charge region is given as
Jscl = qI o (1 − e−α w ),

···

···

(18)

where q = charge of the electron; I o = incident photon ﬂux (number of photons s−1 cm−2 ); α = absorption coefﬁcient for photon of energy hν (cm−1 ); w = space charge width (cm). This is designated as wn or wp for n-type and p -type semiconductor, respectively; Jscl = photocurrent density (A cm−2 ).
Similarly, an equation for photocurrent originated from the diffusion region is given as:
Jdiff = −qI o
− qImin

Lmin e−α w
1 + Lmin
Dmin
Lmin

···

(19)

where Lmin = width of diffusion layer
(cm). In the n-type semiconductor, it will be due to holes (designated as
Lp ) where as in p -type due to electron
(designated as Ln ). Imin = equilibrium concentration of minority carriers (cm−3 ), which is designated as po and no in n-type and p -type semiconductors, respectively.
Dmin = Diffusion coefﬁcient for minority carrier (cm2 s−1 ). For holes in n-type it will be designated as Dp and for electrons in p -type semiconductor it will be designated as Dn . Jdiff = photocurrent due to photogenerated carriers in the diffusion region.
Total photocurrent passing through the junction would be sum of these two components. 4.1 The Photoelectrochemistry of Semiconductor/Electrolyte Solar Cells

Jtotal = {qI o (1 − e−α w )}
+ −qI o

useful to examine Eq. (22) by keeping one of these variables constant.

α Lmin
1 + αLmin

× e−α w − qImin

Dmin
Lmin

For a large band gap semiconductor, Imin ,
Dmin , and Lmin would be very small; hence, their contribution can be neglected.
Thus, the Eq. (20) simpliﬁes to
Jtotal = −qI o

1 − e−α w
(1 + αLmin )

· · · (21)

It is important to remember that the polarity of photocurrent is negative, as one would get experimentally (Fig. 7b).
..
4.1.13.1.1 Application of G artner Model
Equation (21) can be slightly modiﬁed to include a term ‘‘quantum yield (φ )’’ to get information about some useful parameters of semiconductors, for example, band gap, diffusion length, space charge width, and so on. Quantum yield is deﬁned as the ratio of photocurrent (Jphoto ) to the total ﬂux of light (I o ) used to illuminate the semiconductor: that is, quantum yield
(φ) = Jphoto /I o .
Equation (21) can thus be rewritten as

(1 − φ) =

When α is Constant during Photocurrent Measurement This condition is achieved when the junction is illuminated with a monochromatic light and the photocurrent is measured as a function of applied potential (V ), using a potential close to the ﬂat band potential. Substituting the value w = [(2εεo /eNA )(V − Vfb )]1/2 to
Eq. (22) we have

4.1.13.1.2

(20)

e−α w
1 + αLmin

Taking the logarithm of this equation, we have
− ln(1 − φ) = αw + ln(1 + αLmin ) (22)
In Eq. (22), there are two variables, α (absorption coefﬁcient of the semiconductor), which depends on the wavelength of light, and w , (the space charge width), which is related to the dopant’s concentration and magnitude of the bias. Therefore, it is

− ln(1 − φ) = α

2εo ε qNA 1/2

(V − Vfb )1/2

+ ln(1 + αLmin )

(23)

Equation (23) suggests that if − ln(1 − φ) is plotted against (V )1/2 , a linear graph is obtained. Intercept of the linear plot on the potential axis, when − ln(1 − φ) = 0, would give the value of the ﬂat band potential (Vfb ) because none of the terms of α(2εo ε/qNA )1/2 (V − Vfb )1/2 except (V −
Vfb )1/2 can be zero. A word of caution is necessary in this type of calculation. This equation is valid only under the condition of band bending. Therefore, it is always better to extrapolate the linear plot from the region where (V − Vfb ) > 0, that is, from the region where forward bias is much greater than the Vfb values.
Slope of the linear region where (V −
Vfb ) > 0 gives the value of α(2εo ε/qNA )1/2 for the wavelength of light used during illumination. If NA is known from Hall measurements or from capacitance measurement, α can be obtained.
If the experiment is repeated at different wavelengths of light, one can get the value of α for different wavelengths also. This information is extremely useful because
1/α is almost equal to the penetration depth of the corresponding wavelength of light. It is desirable to know the magnitude

305

306

4 Solar Energy Conversion without Dye Sensitization

of (1/α) for band gap light, so that efforts could be made to select a suitable redox electrolyte such that (1/α) ∼ w .
In addition, we can also calculate Ln from magnitude of the intercept (i.e. ln(1 + αLn )) of the linear plot on y -axis.
From Ln the lifetime of the carrier (τ ) can be calculated by using the following equation
Ln =

µn kT τ q 1/2

(24)

µn the mobility of the carrier, can be obtained from Hall measurement.
From these measurements, we can thus evaluate the ﬂat band potential
(which is equal to EF for heavily doped semiconductor), absorption coefﬁcient (α ) for band gap light, diffusion length (L) and lifetime of the carrier (τ ).
4.1.13.1.3 When ‘w’ is Constant During
Using α =
Photocurrent Measurement
{A(hν − Eg )n/2 }/ hν and keeping w constant, Eq. (22) becomes

− ln(1 − φ) =

A(hν − Eg )n/2 hν × w + ln(1 + αLmin ) (25)

When (αLmin ) ranged to get

1, Eq. (25) can be rear-

−[{ln(1 − φ)}hν ]2/n = Aw(hν − Eg )
(26)
In order to utilize Eq. (26), a semiconductor-electrolyte junction is illuminated by different wavelengths of light, keeping the applied potential to a PEC cell constant
(i.e. under a constant forward potential).
A plot of [{ln(1 − φ)}hν ]2/n versus hν would thus result in a linear graph. Let us assume that the value of the intercept on the x -axis at [{ln(1 − φ)}hν ]2/n = 0 is
‘s ’. At this condition, none of the terms

of Aw(hν − Eg ) can be made zero except
(hν − Eg ). This would mean that the value of (hν) (i.e. the intercept on (hν )-axis ‘‘s ’’) should be equal to Eg . If we get the linear plot with n = 1, the band gap is a direct type and if n = 4, it is an indirect one. Some semiconductors may give linear plot for both values of n, indicating simultaneous presence of both types of band gap.
Thus, application of the G¨rtner model a to data obtained from photocurrent versus potential measurement yields several important parameters of a semiconductorelectrolyte junctions (e.g. Eg , Vfb , Lmin , τ , and w ).
4.1.14

Effect of Counter-Electrode

Besides the considerations on the parameters of the semiconductor and the redox electrolyte, performance of the counterelectrode also matters for designing an efﬁcient PEC cell. The counter-electrode in a PEC cell is a charge carrier collecting material, and catalyzes the electron transfer, but is not chemically involved in the electrochemical reaction.
The counter-electrode material should be electrically conducting, economical, and chemically stable in an electrolyte.
It should minimize polarization losses during exchange of electrons at the interface. In other words, polarizability of the counter-electrode should be small in a given redox electrolyte, indicating that the electrode reaction occurring at the counterelectrode is reversible and fast. This maximizes the potential obtained from the semiconductor–electrolyte interface.
For many redox reactions, noble metals such as Pt, Au have been preferred as counter-electrode by electrochemists.
However, those are expensive for terrestrial application, and materials such

4.1 The Photoelectrochemistry of Semiconductor/Electrolyte Solar Cells

as carbon (graphite, glassy carbon), and degenerated transparent [23] semiconductors (SnO2 or ITO) have attracted the attention of scientists for their use as counter-electrodes. For aqueous polysulﬁde redox reactions, Hodes and coworkers [24] have shown that the use cobalt sulﬁde as an electrocatalyst gives much better performance than high surface area platinum or carbon electrodes.
When a counter-electrode other than inert platinum electrode is used, it is necessary to conﬁrm whether the electrochemical reaction occurring at the counter-electrode is as reversible as a platinum counter-electrode [25]. Sharon and coworkers [26] has studied the efﬁciency of three types of counterelectrodes (ITO, Pt and Platinum-coated
ITO) in a PEC cell of conﬁguration:

α -PbO/Fe (CN)6 4− (0.1M), Fe(CN)6 3−
(0.01M) pH 9.2/(counter-electrode). They observed that while platinum counterelectrode gives the best performance
(Fig. 9a), surface modiﬁcation of ITO counter-electrode with platinum (Fig. 9b) gives almost similar performance to that obtained with platinum. However, untreated highly conducting ITO electrode gives considerably low power characteristics (Fig. 9c). Transparent counterelectrode, in addition, avoids the problem of light absorption by the redox electrolyte [25]. With a transparent counterelectrode, it is possible to place the semiconductor very close to the counterelectrode, which also serves as a window for the PEC cell. This backwall conﬁguration is especially helpful when very concentrated electrolytes are used [27].

0.006
(a)

0.005

(b)
0.004

Pmax
(c)

I
[Amp]

0.003
0.002
0.001
0.000
−0.001
−0.002
−0.2

0.0

0.2

0.4

0.6

0.8

1.0

V
[Volts]
Current-voltage characteristic of a PEC cell (α -PbO/Fe (CN)6 4− (0.1M), Fe(CN)6 3−
(0.01M) pH 9.2/(counter-electrode). Three types of counter-electrodes were used:
(a) conducting ITO electrode, (b) surface modiﬁed ITO electrode using platinum and
(c) platinum electrode.
Fig. 9

307

308

4 Solar Energy Conversion without Dye Sensitization

Effect of Surface States
It is important to appreciate that even if a material possess desirable characteristics, as discussed in the previous sections, it may not exhibit the desired efﬁciency of light energy conversion into electrical energy. For example, in spite of semiconductor and redox electrolyte possessing desired Fermi levels experimentally determined contact potential might be much less than the anticipated value. Presence of surface state is one of the factors responsible to show these deﬁciencies. Hence, efforts are required to minimize the adverse effects of surface states.
Surface states mainly originate from the lack of balance in the valence states of atoms/ions present at the surface as compared to atoms/ions present beneath the surface of thin ﬁlm or single crystal.
Consequently, atoms/ions present at the surface possess either more or less number of electrons as compared with those present in the bulk material. This situation creates surface states of various types. All of them behave like a donor or an acceptor sites on the surface. Thus, irrespective of the type of surface states, they tend to trap photo-generated carriers at the surface of the semiconductor. These trapped carriers are released at a slow rate. Moreover, these surface states can also cause pinning of the Fermi level. Pinning of the Fermi level takes away the freedom of Fermi level to adjust its position while attaining the equilibrium condition. Thus, pinning of the Fermi level decreases the magnitude of anticipated contact potential. Hence, for an efﬁcient PEC cell, the semiconductor should be free from surface states. It is impossible to fully eliminate surface states.
However, this effect can be minimized.
Since these surface states behave either as a site possessing positive or negative
4.1.14.1

charge, it is possible to neutralize their charges by controlling pH of the solution.
Sharon and coworkers [28] have developed a simple technique to ﬁnd the pH at which the effect of surface states is minimum. Their model also gives information regarding the magnitude of other parameters such as ﬂat band potential and band bending potential. The experiment is based on the fact that concentration of a majority carrier at the surface of the semiconductor (in dark condition) is controlled by the extent of band bending formed in the space charge region. Similarly, freedom of majority carrier to move freely over the surface also depends on the nature and magnitude of the surface states present over the surface. Therefore, the magnitude of surface states present on the surface inﬂuences the surface conductivity. The activation energy for the surface conductivity should thus also be related to the nature and magnitude of the surface states. For surface conductivity measurement, two ends of a thin ﬁlm of semiconductor are connected with platinum wires. The entire surface except the top surface of the ﬁlm is covered with an insulating paint.
This ﬁlm is then immersed in a solution of known pH whose temperature is controlled very accurately. Surface resistance is measured versus temperature (normally in the range of −10 ◦ C to +25 ◦ C). This experiment is repeated at different pH values. Log of the surface conductivity is plotted versus reciprocal of the absolute temperature to get the activation energy.
This calculation is carried out for each pH. Finally, a plot of activation energy versus pH is plotted, which shows either a minimum with n-type semiconductor or a maximum with p -type semiconductor.
These pH values correspond to point of zero charge (pzc) of the material.

4.1 The Photoelectrochemistry of Semiconductor/Electrolyte Solar Cells
−8.4

In σ

−7.6
−7.2
−6.8

In conductivity × 10−2

30°C
21°C
12°C
4°C
2°C
−9°C

−8.0

−6.4
−6.0
−5.6

630
610
590
570
550
530
510

5.4 5.8 6.2 6.6 7.0 7.4 7.6 8.2 8.6 9.0 9.4

Barrier height (Activation energy)
[eV]

0.6
0.5
0.4
0.3
0.2
0.1
0.0
0123456789

pH

0

1

2

3

(d)

4

5

6

7

pH

(b)

Barrier height (Activation energy)
[eV]

pH

(a)

(c)

0°C
5°
10°
15°
20°
25°

0.16
0.14
0.12
0.10
0.08
0.06
0.04
0.02
5

6

7

8

9 10

pH

Surface conductivity measurements of thin ﬁlms of α -Fe2 O3 and p-SnSe21 . Log of surface conductivity versus pH at different temperatures for
(a) n-α -Fe2 O3 (b) p-SnSe. Activation energy of surface conductivity versus pH for (c) n-α -Fe2 O3 and (d) p-SnSe.
Fig. 10

In Fig. 10, typical results of thin ﬁlms nα -Fe2 O3 (a,c) and p -SnSe (b,d) are shown. n-α -Fe2 O3 gives a minimum at pH 4.1, which is equivalent to its pzc. A reverse behavior is obtained with p -SnS showing a maximum at pH 7.5, which is equivalent to its pzc. This is the only nondestructive method available to get information about pH that corresponds to pzc for large area thin-ﬁlm semiconductor material.
A PEC cell operating at the pH equivalent to the pzc will experience the least hindrance caused by the surface states, and may be able to show its anticipated PEC behavior. PEC operating at pzc would be able to show its predicted contact potential

because the surface of the semiconductor would have no excess charge for pinning the Fermi level. Moreover, photogenerated carriers will also not be trapped by the surface states as their charges are neutralized by pH of the solution.
4.1.15

Photodecomposition of Semiconductor

In the dark many semiconductors are stable in contact with a given electrolyte, but start decomposing on illumination of the junction. This and energy conversion efﬁciency are the most serious problems faced in PEC devices. If the photo-generated

309

310

4 Solar Energy Conversion without Dye Sensitization

minority carriers are not rapidly used in oxidation or reduction by the electrolyte immediately, they have a tendency to react with the semiconductor material. The semiconductor electrode therefore starts corroding. PEC etching is being dealt in a greater depth in a separate chapter. However, the easiest way to establish
PEC stability of the semiconductor is by measuring the photocurrent of a PEC cell over a longer period under continuous illumination. If semiconductor is stable, the photocurrent will show a constant value over a period of illumination.

Unfortunately, most of the low band gap materials studied so far have been found to undergo such corrosions. Therefore, there is a need to develop newer materials for developing a photoelectrochemically stable semiconductor. Materials for PEC Cell
Because of the limitation of space, it is difﬁcult to discuss details of materials studied for various types of PEC cell. Some of the materials and the redox electrolytes used for studying PEC cells are tabulated in Tables 1 and 2. Nevertheless, Sharon
4.1.15.1

Some of the chalcogenide materials studied for developing PEC cell in various redox electrolytes Tab. 1

Material

Electrolyte system

Reference

n-CdS

NaOH,S,Na2 S; NaOH,Na2 Te;
NaOH,Na2 (SCH2 COO)S; KCl,
KFe(CN)6 3−/4− ; NaI, I2 in CH3 CN

31–33, 35, 39

n-CdSe

NaOH, Na2 S, S; NaOH, Na2 Se; NaOH,
Na2 Te; NaOH, Na2 S,S, Se;
NaOH,Na2 (SCH2 COO)S, Se;
KOH,Na2 S,Sn ; Fe(CN)6 3−/4− pH > 8,
Cesium polysulphides, S,CuSO4

31–33, 36–38, 40–42

n-CdTe/p-CdTe

NaOH-Na2 Se, NaOH-Na2 Te, K2 Se-Se-KOH

34, 35, 40, 43

n-CdInS2 ,
CuInSe2 , CuInS2

NaOH-Na2 S-S

44–45,100

n-Re6 Se8 Cl2

I3 − -I− -H2 SO4

46

n-CdIn2 Se4

KOH-S-Na2 S

47

n-CuIn5 S8

Sulﬁde, Polysulﬁde,

48

n-WS2

Halide solution,

49

p-WS2

CH3 CN/(n-Bu4 N)ClO4 ,

50

n-MoS2

Et2 N-halide ion in CH3 CN, NaI-I2 (aq),
Fe2+ /Fe3+ ,

51, 55, 56

n-MoSe2

I/I2 , Fe2+ /Fe3+ , (Et4 N)X/2 in CH3 CN;
X = halogen, HBr/Br2 pH = 1

52–53, 58

n-WSe2

Fe(CN)3−/4− , NaI/I− -Na2 SO4 − ;
3
H2 SO4 ,pH = 0

52, 80

p-WSe2

Fe2+/3+ -H2 SO4

59

n-SnS

Ce4+ /Ce3+ , 0.5 M; H2 SO4

70

n-Bi2 S3

NaOH-Na2 S-S

60–61

4.1 The Photoelectrochemistry of Semiconductor/Electrolyte Solar Cells
Tab. 2 Some of the phosphide/oxide and other materials studied for developing PEC cell in various redox electrolytes

Compound

Electrolyte

Reference

n-InP

NaOH/Fe(CN)6 3−/4− ; NaOH-Na2 Te,
VCl2 -VCl5 -HCl

62

p-InP

NaOH/Fe(CN)6 3−/4− ; NaOH-Na2 Te,
VCl2 -VCl5 -HCl

63

p-GaP

VCl2 -VCl5-Na2 SO4 ; pH = 2.5

64

ZnP2 , Zn3 P2

NaOH/Fe(CN)6

3−/4−

65

Fe2 O3

NaOH/Fe(CN)6 3−/4−

66

YFeo3

Fe(CN)6 3−/4− , pH = 7 & 9

67

Pb3 O4

IO3 − /I−

68

BaTiO3

Ce3+ /Ce4+ , Fe2+ /Fe3+ ; pH = 2

69

PbO(2−x )

Fe(CN)6 3−/4− , pH = 9

71

n-GaAs

Fe2+/3+ ; I− /I3 − ; S2− /S2− ; Fe(CN)6 3−/4− n 43, 74–76, 79–80

p-GaAs

I− /I3 − pH = 0; KCl-VCl2 /VCl; Eu3+

57, 78, 81–83

p-Si

HCl-VCl2 /VCl3

84–87

n-Si

FeCp2 ; n-Bu4 N-ClO4 in ethanol;
FeCp2 = ferrocene; I− /I3 − -FeCl2 /FeCl3

54, 87, 89–93

GaAs0.72 P0.28

Nonaqueous electrolyte

78, 81

Chronological improvements in solar to electrical conversion efﬁciencies of some regenerative PEC solar cells

Tab. 3

Year

Solid/dissolved salts 1998
1991

AlGaAs/Si/V2+/3+
Nano TiO2 (dye sensitized)/I3 − n-CdSe/KFe(CN)6 3− n-GaAs/Se2− (Os3+ ) n-Cd(Se,Te)/S2− n-CuInSe2 /I− n-WSe2 /I− n-MoSe2 /I− p-InP/V3+ n-GaAs/Se2 -(Ru2+ )

1990
1987
1985
1982
1982
1981
1981
1978

% Efﬁciency

Reference

∼19
∼7

96
97

∼16
∼15
∼13
∼9
∼10
∼9
∼11
∼12

98
99
100
101
102
103
104
105

and coworkers [29, 30] have extensively reviewed these materials.
Examples of chronological improvements in high solar to electrical conversion efﬁciency PEC’s are summarized in

Table 3. Highest reported conversion efﬁciencies include more than 9% for dye sensitized thin ﬁlm of n-TiO2 (nanostructure) in polyiodide electrolyte [97] and more than 16% for single-crystal n-CdSe

311

312

4 Solar Energy Conversion without Dye Sensitization

in cyanide modiﬁed K3 Fe(CN)6 3− electrolyte [98]. The synthesis of single-crystal semiconductor material is not cost effective, being both time intensive and limited to formation of smaller surface area of materials. A variety of techniques have been developed to provide polycrystalline or amorphous thin-ﬁlm semiconductors.
A particular advantage of electrodeposition
(plating) techniques is a one-step chemical synthesis of the semiconductor. PV, as well as PEC solar cells requires the use of thin-ﬁlm semiconductors to be cost competitive. Each of these techniques consumes relatively small amounts of material and can be applied to develop large surface area electrodes. Modes of ﬁlm preparation include slurry techniques, chemical vapor deposition, wackercast, rf sputtering, vacuum coevaporation, sol gel, chemical bath deposition, hot pressing, and electroplating. These techniques have been used to prepare thin ﬁlms of n-CdS, n-CdSe, n-CdTe, n-GaAs, p -InP, n-Si and p -Si, nTiO2 and used in PEC solar cells [97, 105].
Advantageously, thin ﬁlm PEC’s have exhibited photocurrent lifetimes signiﬁcantly greater than single-crystal PEC’s.
This phenomenon has been attributed to the high microscopic surface area, and resulting lower effective microscopic photocurrent density, passing through the thin-ﬁlm devices. The longest lifetime PEC device to date, slurry deposited thin-ﬁlm n-Cd (Se, Te) electrodes immersed in modiﬁed polysulﬁde solutions, has demonstrated outdoor operation for approximately one year. In this system, photoelectrode and electrolyte stability was achieved by optimization of the solution alkalinity, using alternative cations, and modifying the ratio of dissolved sulfur to sulﬁde in solution [106]. The recent advance in dyesensitized PEC devices requires future

efforts to extend the lifetimes of the requisite dyes [97].
But unfortunately none of these materials show a long-term PEC stability to warrant making of a commercially viable
PEC cell. Interestingly, it has been hinted that phosphides of Co, Fe, Mo, Ni, V, and W (band gap in vicinity of 1 eV) may be photoelectrochemically stable. Among the oxide materials, PbO2 (Eg = 1.6 eV) is worth exploring because it is highly stable in acidic medium. Unfortunately, due to high oxygen vapor pressure, PbO2 exits as a degenerated semiconductor (i.e. it behaves like a metal). If suitable doping could open the band gap, then PbO2 can become one of the best materials for making commercially viable PEC cell. Semiconducting carbon is another class of material that needs to be extensively studied for this purpose. Earlier, diamond was the only known semiconductor of carbon. But in the recent past, it has been possible to synthesize semiconducting carbon with a band gap in the vicinity of 1.0 to 2.8 eV. Sharon and coworkers [107, 108] has recently reported a PEC cell from a semiconducting carbon made from the pyrolysis of camphor. In addition, efforts should be made to reduce the band gap of stable oxides such as TiO2 into a low band gap material by manipulating to obtain their mixed oxides. Mixed oxides of Mo & W also need similar attention. Sharon and coworkers [109, 110] have tried to develop a theoretical model to predict the variation in band gap by varying the composition of components in mixed oxides. In addition to this, there is a need to develop a theoretical model, to predict PEC stability of materials. Such models will go a long way to minimize the number of experiments needed in developing materials for making a commercially viable PEC cell.
It is necessary to emphasize that in spite of the simplicity in fabrication, the future

4.1 The Photoelectrochemistry of Semiconductor/Electrolyte Solar Cells

of economically viable PEC depends entirely on the development of suitable low band gap, photoelectrochemically stable materials. 4.1.16

Laser Scanning Technique for a Large Area
Electrode

4.1.17

Summary

In this section, efforts are made to explain the basic principles of a regenerative PEC solar cell. Greater emphasis is made on

2.8

0

(a)

0.0

0

7.
30
4.
87

0

5.6

0

2.8

0

(b)

2.
43

0

8.4

2.
43

5.6

0.
00

0

4.
87

8.4

x

0.0

0

0.
00

7.
30

x

I ph
[a.u.]

I ph
[a.u.]

y

y

There are materials, which can form varieties of nonstoichiometric compounds.
For example, PbO2 can coexist as PbOx where x can have values from 0 to 2.
Among these oxides, most photoactive oxide is PbO0.8 . If PbO0.8 oxide is to be prepared by anodic oxidation of lead, it is possible that surface of the lead ﬁlm might contain other forms of oxides as well. Thus, a PEC cell prepared by anodization of lead may give a low photoresponse, not because PbO0.8 is an inappropriate material but because anodized surface may be contaminated with other forms of oxides as well. There seems to have no nondestructive technique available to conﬁrm the uniformity of surface with photoactive species, especially of a large area electrode. Sharon and coworkers [111] has recently developed a laser scanning system to overcome this problem. A laser

beam is scanned over the entire surface of the ﬁlm, and photocurrent is measured at each illuminated point. A 3D plot of photocurrent versus (x, y ) distances is made. If the composition of materials formed over the large surface area electrode contains photoactive oxide, a uniform distribution of photocurrent would be observed. Failing to get uniform photocurrent would be an indication of nonuniformity of the surface with the desired material. Though this technique is not speciﬁc about the composition, it is a very useful tool to give a guideline for improving the preparation technique especially for a large area electrode. The resolution of this experiment, however, depends on the scanning size of the laser beam. For example, the 3D laser scanning map of an anodized lead sheet
(2.5 cm2 ) shows that the photocurrent is not as uniform (Fig. 11a) as obtained with ﬁlm anodized by condition shown in
Fig. 11(b).

3D photocurrent map obtained with scanning of He/Ne laser (3 mW) over the surface of lead sheet anodized in the potential range (a) −0.32 to +0.08 V and (b) −0.22 to +0.08 V versus SCE [26].

Fig. 11

313

314

4 Solar Energy Conversion without Dye Sensitization

criteria for selection of a semiconductor and a redox electrolyte to make a PEC cell operating at its maximum performance.
Effect of light on illuminating a junction formed with semiconductor–electrolyte interface is discussed with a view to characterize various parameters inﬂuencing the efﬁciency of a PEC cell. The role of counterelectrode and minimizing of the surface states to achieve a desirable performance of a PEC cell are also discussed. Whenever, a large surface area electrode of semiconductor of Ax By type is prepared especially by an electrochemical method, uniform deposition of material over the entire surface with similar composition is always a problem to establish. A laser scanning technique has been discussed for this purpose. Although PEC cells with efﬁciency as high as 16% have been developed, because of its instability problem, it has not been possible to commercialize. A section is devoted to discuss the material aspect of the PEC and a future direction of research needed to fabricate a commercially viable PEC. A chronological improvement made in PEC are also shown in a tabular form.
Acknowledgment

I am extremely grateful to all my students who have contributed a lot via their Ph.D. thesis and research work, which enabled me to perceive the complexity of PEC cell. Their interactions enriched me to evolve a simpler way to comprehend the intricacy of a PEC cell. This has immensely helped me in penning this section. I am also grateful to Dr. Michael Neumann-Spallart, CNRS,
Meudon, France, for going through the manuscript and for his valuable suggestions, which helped me in formulating this chapter.

References
1. A. E. Becquerel, C.R. Acad. Sci. 1839, 9, 561.
2. W. H. Brattain, C. G. B. Garret, Semiconductor surface phenomena in Semiconducting materials Proc. Conf. Univ. Reading, England, 1951, 37.
3. W. H. Brattain, C. G. B. Garrett, Phys. Rev.
1955, 99, 376.
4. A. Fujishima, K. Honda, Nature 1972, 238,
37.
5. H. Gerischer in Semiconductor electrochemistry: in Physical Chemistry: An advanced treatise (Eds.: H. Eyring, D. Henderson,
W. Jost), 9A, Academic Press, New York,
1970, p. 463.
6. V. A. Myamlin, Yu. V. Pleskov, Electrochemistry of Semiconductors, Plenum Press, New
York, 1967.
7. M. Sharon, P. Veluchamy, C. Natrajan,
C. D. Kumar, Electrochim. Acta 1991, 36(7),
1107.
8. S. Licht, G. Hodes, R. Tenne et al., Nature
1987, 326, 863.
9. S. Licht, B. Wang, T. Soga et al., Appl. Phys.
Lett. 1999, 74, 4055.
10. B. Pal, M. Sharon, D. Kamat, Chem. Environ. Res. 1996, 5(1–4), 51.
11. H. Yoneyama, N. Takahashi, K. Susumu et al., J. Chem. Soc. Chem. Commun. 1992,
9, 716.
12. A. J. Nozik in Photoelectrochemical devices for solar energy conversion, NATO advance study institute series, Series B; Physics, Plenum press, New York, 1981, p. 280, Vol .69.
13. V. G. Levich in Physical Chemistry – An
Advanced Treatise, (Eds.: H. Eyring, D. Henderson, W. Jost), Academic Press, New
York, 1970, Vol. 9B, Chapter 12.
14. H. Gerischer, Surf. Sci. 1969, 13, 265.
15. H. Gerischer, in Special Topics in Electrochemistry, (Ed.: P. G. Rock), Elsevier, Amsterdam, 1977.
16. H. Gerischer, Surf. Sci. 1969, 18, 97.
´
17. J.O.M Bockris, S. U. M. Khan, J. Electrochem. Soc. 1985, 132, 2648.
18. S. Licht, Nature 1987, 330, 148.
19. S. Licht, Sol. Energy Mater. Sol. Cells 1995,
38, 353.
20. W. W. G¨rtner, Phys. Rev. 1959, 116, 84. a 21. I. Mukhopadhyay, M. Sharon, Electrochim.
Acta. 1997, 42(1), 67–72.
22. I. Mukhopadhyay, M. Sharon, Sol. Energy
Mater. Sol. Cells 1997, 45, 141.

4.1 The Photoelectrochemistry of Semiconductor/Electrolyte Solar Cells
23. J. F. Gibbons, G. W. Cogan, C. M. Gronet et al., Appl. Phys. Lett. 1984, 44, 539.
24. G. Hodes, J. Manassen, D. Cahen, J. Electrochem. Soc. 1980, 127, 544.
25. R. Tenne, M. Peisach, C. A. Rabe et al.,
J. Electroanal. Chem. 1989, 269, 389.
26. M. Sharon, S. Ghosh, J. Solid State Electrochem. 1999, 4, 52.
27. S. Licht, F. Forouzan, J. Electrochem. Soc.
1995, 142, 1539.
28. M. Sharon, B. M. Prasad, K. Basavaswaran,
Indian J. Chem. 1989, 28A, 935.
29. M. Sharon, K. Basavaswaran, N. P. Sathe,
J. Sci. Ind. Res. 1985, 44, 593.
30. M. Sharon, G. Tamizhmani, J. Mater. Sci.
1986, 21, 2193.
31. R. N. Nouﬁ, P. A. Kohl, A. J. Bard, J. Electrochem. Soc. 1978, 125, 375.
32. M. S. Wrighton, A. B. Ellis, S. W. Kaiser,
J. Am. Chem. Soc. 1976, 98, 1635.
33. S. Chandra, S. K. Pandey, R. C. Agrawal,
Sol. Cell 1980, 1, 36.
34. W. J. Danaher, L. E. Lyons, Nature 1978,
271, 139.
35. J. M. Bolts, A. B. Ellis, K. D. Legg et al.,
J. Am. Chem. Soc. 1977, 99, 4826.
36. R. Nouﬁ, D. Tench, L. F. Warren, J. Electrochem. Soc. 1980, 127, 2709.
37. A. Aruchamy, A. Venkatarathnam, G. V.
Subba Rao et al., Electrochem. Acta. 1982,
27, 701.
38. H. Gerischer, J. Electroanal. Chem. 1975, 58,
263.
39. M. Tsuiki, H. Minoura, T. Nakamura et al.,
J. Appl. Electrochem. 1978, 8, 523.
40. A. B. Ellis, S. W. Kaiser, M. S. Wrighton et al., J. Am. Chem. Soc. 1977, 99, 2839.
41. R. Tenne, G. Hodes, Appl. Phys. Lett. 1980,
37, 428.
42. K. Rajeshwar, J. Appl. Electrochem. 1985, 15,
1.
43. A. Heller, K. C. Chang, B. Miller, J.Am.
Chem. Soc. 1978, 100, 684.
44. L. Fornarini, M. Lazzari, L. P. Bicelli et al.,
J. Power Source 1981, 6, 371.
45. Y. Mirovsky, D. Cahen, Appl. Phys. Lett.
1982, 40, 727.
46. N. LeNagard, A. Perrin, M. Sergent et al.,
Mater. Res. Bull. 1985, 20, 835.
47. R. Tenne, Y. Mirovsky, Y. Greenstein et al.,
J. Electrochem. Soc. 1982, 129, 1506.
48. B. Scrosati, L. Fornarini, G. Razzini et al.,
J. Electrochem. Soc. 1985, 132, 593.

49. J. A. Baglio, G. S. Calabrese, E. Kamieniecki et al., J. Electrochem. Soc. 1982, 129, 1461.
50. J. A. Baglio, G. S. Calabrese, D. J. Harrison et al., J. Am. Chem. Soc. 1983, 105, 2246.
51. H. Tributsch, Ber. Bunsen-Ges Phys. Chem.
1977, 81, 361.
52. H. Tributsch, H. Gerischer, C. Clemen et al., Ber. Bunsen-Ges Phys. Chem. 1979,
83, 655.
53. L. F. Schneemeyer, M. S. Wrighton, J. Am.
Chem. Soc. 1980, 102, 6964.
54. T. Skotheim, L. G. Peterson, O. Ingan¨s a et al., J. Electrochem. Soc. 1982, 129, 1737.
55. G. Djemal, N. Muller, U. Lachish et al., Sol.
Energy Mater. 1981, 5, 403.
56. R. Audas, J. C. Irwin, J. Appl. Phys. 1981, 52,
6954.
57. A. J. Bard, F.-R.-F. Fan, J. Am. Chem. Soc.
1980, 102, 3677.
58. H. Tributsch, Ber. Bunsen-Ges Phys. Chem.
1978, 82, 169.
59. H. Gerischer, W. Kautek, Ber. Bunsen-Ges
Phys. Chem. 1980, 84, 645.
60. B. Miller, A. Heller, Nature 1976, 262, 680.
61. P. K. Mahapatra, C. B. Roy, Sol. Cells
1982–1983, 7, 225.
62. M. S. Wrighton, A. B. Ellis, J. M. Bolts,
J. Electrochem. Soc. 1977, 124, 1603.
63. Y. Ramprakash, S. Basu, D. N. Bose,
J. Indian Chem. Soc. 1981, LVIII, 153.
64. M. Tomkiewicz, J. M. Woodall, Science
1977, 196, 990.
65. H. von K¨nel, L. Gantert, R. Hauger et al., a Hydrogen Energy Prog. 1984, 5, 969.
66. M. Sharon, B. M. Prasad, Sol. Energy Mater.
1983, 8, 457.
67. M. Sharon, B. M. Prasad, Electrochim. Acta
1985, 30(3), 331.
68. M. Sharon, S. Kumar, S. R. Jawalekar, Bull.
Mater. Sci. 1986, 8(3), 415.
69. M. Sharon, A. Sinha, Sol. Energy Mater.
1984, 9, 391.
70. M. Sharon, K. Basavaswaran, Sol. Cells
1988, 25, 97.
71. P. Veluchamy, M. Sharon, M. Shimizu et al., J. Electroanal. Chem. 1994, 371, 205.
72. A. Heller, B. Miller, S. S. Chu et al., J. Am.
Chem. Soc. 1979, 101, 7633.
73. A. J. Bard, F.-R.-F. Fan, B. Reichman, J. Am.
Chem. Soc. 1979, 101, 7633.
74. K. Rajeshwar, M. Kaneko, A. Yamado,
J. Electrochem. Soc. 1983, 130, 38.
75. K. Rajeshwar, P. Singh, J. Du Bow, Trans.
ASME J. Sol. Energy Eng. 1982, 104, 146.

315

316

4 Solar Energy Conversion without Dye Sensitization
76. A. Barrasse, H. Cachat, G. Horowitz et al.,
Rev. Phys. Appl. 1982, 17, 801.
77. D. S. Ginley, R. M. Bie Feld, B. A. Parkinson et al., J. Electrochem. Soc. 1982, 129, 145.
78. W. S. Hobson, A. B. Ellis, Appl. Phys. Lett.
1982, 41, 891.
79. S. Gourgaud, D. Elliott, J. Electrochem. Soc.
1977, 124, 102.
80. A. Heller, B. Miller, B. A. Parkinson,
J. Electrochem. Soc. 1979, 126, 954.
81. M. G. Chris, S. L. Nathan, Nature 1982, 300,
733.
82. R. Memming, J. Electrochem. Soc. 1978, 125,
117.
83. A. J. Bard, R. E. Malpas, K. Itaya, J. Am.
Chem. Soc. 1979, 101, 2535.
84. H. J. Lawerenz, M. Lubke, Appl. Phys. Lett.
1981, 39, 798.
85. W. M. Ayers, J. Electrochem. Soc. 1982, 129,
1644.
86. A. J. Bard, F.-R.-F. Fan, G. A. Hope,
J. Electrochem. Soc. 1982, 129, 1647.
87. T. S. Jayadevaiah, Appl. Phys. Lett. 1974, 25,
399.
88. A. J. Bard, F.-R.-F. Fan, H. S. White et al.,
J. Am. Chem. Soc. 1980, 102, 5142.
89. M. S. Wrighton, A. B. Bocarsly, E. G. Walton et al., J. Electroanal. Chem. 1979, 100,
283.
90. M. S. Wrighton, K. D. Legg, A. B. Ellis et al.,
Proc. Natln. Acad. Sci. USA 1977, 74, 4116.
91. M. S. Wrighton, R. G. Austin, A. B. Bocarsly et al., J. Am. Chem. Soc. 1978, 100, 1602.
92. M. S. Wrighton, J. M. Bolts, A. B. Bocarsly et al., J. Am. Chem. Soc. 1979, 101, 1378.
93. R. M. Candea, M. Kastner, R. Goodman et al., J. Appl. Phys. 1976, 47, 2724.

94. A. J. Nozik, Appl. Phys. Lett. 1976, 29, 150.
95. A. Heller, M. Robbins, K. J. Bachmann et al., J. Electrochem. Soc., 1978, 125, 831.
96. S. Licht, O. Khaselev, T. Soga et al., Electrochem. Soc. State Lett. 1998, 1, 20.
97. B. O’Regan, M. Gr¨tzel, Nature 1991, 353, a 737.
98. S. Licht, D. Peramunage, Nature 1990, 345,
330.
99. S. Licht, D. Peramunage, Nature 1991, 354,
440.
100. B. J. Tufts, I. L. Abrahams, P. G. Santangelo et al., Nature 1987, 326, 861.
101. S. Licht, R. Tenne, G. Dagan et al., Appl.
Phys. Lett. 1985, 46, 608.
102. S. Menezes, H. J. Lewerenz, K. J. Bachmann, Nature 1983, 305, 615.
103. G. Kline, K. Kam Keung, R. Ziegler et al.,
Sol. Energy. Mater. 1982, 6, 337.
104. G. Kline, K. Kam, D. Canﬁeld et al., Sol.
Energy. Mater. 1981, 4, 301.
105. A. Heller, B. Miller, Appl. Phys. Lett. 1981,
38, 282.
106. B. A. Parkinson, A. Heller, B. Miller, Appl.
Phys. Lett. 1978, 33, 521.
107. S. Licht, J. Phys. Chem. 1986, 90, 1096.
108. K. Mukhopadhyay, I. Mukhopadhyay,
M. Sharon et al., Carbon, 1997, 35, 863.
109. M. Sharon, I. Mukhopadhyay, K. Mukhopadhyay, Sol. Energy Mater. Sol. Cells 1997,
45(1), 35.
110. K. Murali Krishna, M. Sharon, M. K.
Mishra, J. Phys. Solids 1996, 57(5), 615.
111. K. Murali Krishna, M. Sharon, M. K. Mishra et al., Electrochem. Acta 1996, 41(13), 1999.
112. M. Sharon, I. Mukhopadhyay, S. Ghosh,
J. Solid State Electrochem. 1999, 3, 141.

4.2 Photoelectrochemical Solar Energy Storage Cells

4.2

Photoelectrochemical Solar Energy
Storage Cells
Stuart Licht
Technion – Israel Institute of Technology,
Haifa, Israel
4.2.1

Introduction

Although society’s electrical needs are largely continuous, clouds and darkness dictate that photovoltaic solar cells have an intermittent output. A photoelectrochemical solar cell (PEC) can generate not only electrical but also electrochemical energy and provide the basis for a system with an energy-storage component
(PECS). Sufﬁciently energetic insolation incident on semiconductors can drive electrochemical oxidation/reduction and generate chemical, electrical, or electrochemical energy. Aspects include efﬁcient dye-sensitized or direct solar to electrical energy conversion, photoetching, photoelectrochemical water-splitting, environmental cleanup, and solar energy storage cells. This chapter focuses on photoenergy storage concepts based on photoelectrochemical processes, but includes a necessary comparison to other methods proposed for the conversion and storage of solar energy. The PEC uses light to carry out a chemical reaction, converting light to chemical energy. This fundamental difference of the photovolatic solar cell’s (PV) solid–solid interface, and the PEC’s solid–liquid interface has several ramiﬁcations in cell function and application. Energetic constraints imposed by single band gap semiconductors have limited the demonstrated values of photoelectrochemical solar to electrical energy conversion efﬁciency to 16%, and multiple band gap cells can lead to signiﬁcantly

higher conversion efﬁciencies [1, 2a,b].
Photoelectrochemical systems may facilitate not only solar to electrical energy conversion but have also led to investigations in photoelectrochemical synthesis, photoelectrochemical production of fuels, and photoelectrochemical detoxiﬁcation of pollutants as discussed in other chapters in this volume.
Regenerative Photoelectrochemical
Conversion
In illuminated semiconductor systems, the absorption of photons generates excited electronic states. These excited states have lifetimes of limited duration. Without a mechanism of charge separation their intrinsic energy would be lost through relaxation (recombination). Several distinct mechanisms of charge separation have been considered in designing efﬁcient photoelectrochemical systems. At illuminated semiconductor–liquid interfaces, an electric ﬁeld (the space charge layer) occurs concurrent with charge–ion redistribution at the interface. On photogeneration of electron-hole pairs, this electric ﬁeld impedes recombinative processes by oppositely accelerating and separating these charges, resulting in minority carrier injection into the electrolytic redox couple. This concept of carrier generation is illustrated in Fig. 1(a) (for an n-type PEC) and has been the theoretical basis for several efﬁcient semiconductor-redox couple
PEC cells. Illumination of the electrode surface with light, whose photon energy is greater than the band gap, promotes electrons into the conduction band leaving holes in the valance band. In the case of a photoanode, band-bending in the depletion region drives any electron that is promoted into the conduction band into the interior of the semiconductor and holes
4.2.1.1

317

318

4 Solar Energy Conversion without Dye Sensitization
Load

Load

E

E
Light

EC
EF

−

−

Red

EV

+

Ox+

−

Energy gain +

EC
EF

−

−

Light

D∗

D

Red

Ox+

−

Energy gain EV

n -type semiconductor Redox
Counter
electrolyte electrode

n -type
Dye
semiconductor

(a)

Redox
Counter
electrolyte electrode
(b)

Carrier generation under illumination arising at (a) the semiconductor–liquid interface and
(b) the semiconductor–dye sensitizer–liquid interface.

Fig. 1

in the valance band toward the electrolyte, where they participate in an oxidation reaction. Electrons through the bulk drive an external load before they reach the counter electrode or storage electrode, where they participate in a reduction process. Under illumination and open circuit, a negative potential is created in a photoanode, and as a result the fermi level for the photoanode shifts in the negative direction, thus reducing the band-bending. Under illumination with increasing intensity, the semiconductor fermi level shifts continually toward negative potentials until the band-bending effectively reduces to zero, which corresponds to the ﬂat band condition. At this point, a photoanode exhibits its maximum photovoltage, which is equal to the barrier height. Excitation can also occur in molecules directly adsorbed and acting as a mediator at the semiconductor interface. In this dye sensitization mode, the function of light absorption is separated from charge carrier transport. Photoexcitation occurs at the dye and photogenerated

charge is then injected into a wide band gap semiconductor. This alternative carrier generation mode can also lead to effective charge separation as illustrated in Fig. 1(b). The ﬁrst high solar to electric conversion efﬁciency example of such a device was demonstrated in 1991 [3] through the use of a novel high surface area
(nanostructured thin ﬁlm) n-TiO2 , coated with a well-matched trimeric ruthenium complex dye immersed in an aqueous polyiodide electrolyte. The unusually high surface area of the transparent semiconductor coupled to the well-matched spectral characteristics of the dye leads to a device that harvests a high proportion of insolation.
Photoelectrochemical Storage
PECs can generate not only electrical but also electrochemical energy. Figure 2 presents one conﬁguration of a PEC combining in situ electrochemical storage and solar-conversion capabilities; providing continuous output insensitive to daily variations in illumination. A high solar to
4.2.1.2

4.2 Photoelectrochemical Solar Energy Storage Cells hn S

P=0

P

S

POLYSULFIDE
OXIDATION

L

LOAD
(a)

LOAD

1.8 m CsHS
1.8 m CsOH

Sn/SnS

TIN
OXIDATION

MEMBRANE

POLYSULFIDE
REDUCTION

CoS

0.8 m CsHS
0.8 m CsOS
1.0 m Cs2S4

1.8 m CsHS
1.8 m CsOH

Sn/SnS

MEMBRANE

n -Cd(Se,Te)
TIN SULFIDE
REDUCTION

CoS

POLYSULFIDE
REDUCTION

0.8 m CsHS
0.8 m CsOS
1.0 m Cs2S4

n -Cd(Se,Te)

L
(b)

Schematic of a photoelectrochemical solar cell combining both solar conversion and storage capabilities. (a) Under illumination; (b) in the dark.

Fig. 2

electric conversion efﬁciency cell conﬁguration of this type was demonstrated in
1987 and used a Cd(Se,Te)/Sx conversion half cell and a Sn/SnS storage system, resulting in a solar cell with a continuous output [4]. Under illumination, as seen in Fig. 2(a), the photocurrent drives an external load. Simultaneously, a portion of the photocurrent is used in the direct electrochemical reduction of metal cations in the device storage half-cell. In darkness or below a certain level of light, the storage compartment spontaneously delivers power by metal oxidation, as seen in Fig. 2(b).
4.2.2

Comparative Solar-Storage Processes
Thermal Conversion and Storage
Solar insolation can be used to directly activate a variety of thermal processes; the enthalpy is stored physically or chemically and then either directly utilized or released
4.2.2.1

upon reversal of the storage process.
To date, the predominant nonbiologic utilization of solar energy is to heat a working ﬂuid that is maintained in an insulated enclosure, storing a portion of the incident solar radiation for future use.
Limitations of this approach include the low energy available per unit mass of the storage medium, and low efﬁciencies of thermal to mechanical and thermal to electrical energy conversion. A variety of passive and dynamic optical concentrators have been studied to compensate for these limitations. The high temperatures generated by concentrated solar power have been utilized to drive highly endothermic reactions. The reverse reaction releases the chemically stored energy as thermal energy. Various systems have been investigated such as in 1985 [5]:
SO3 − − SO2 + 1 O2
−→
2

H = 98.94
(1)

319

320

4 Solar Energy Conversion without Dye Sensitization

Photochemical Storage
In photochemical processes, photon absorption creates a molecular excited state or stimulates an interband electronic transition in a semiconductor that induces a molecular change. Comprehensive reviews, including those by
Gratzel [6], Kalyanasundaram [7], and Harriman [8], have discussed various aspects of photochemical energy conversion. A photoactivated molecular excited state can drive either (1) photodissociation
(2) photoisomerisation or (3) photoredox reactions. Processes based on semiconductors may involve photovoltaic or photoelectrochemical systems.
A substantial photochemical effort research has centered on the photoredox storage of solar energy. Molecular photoredox processes use electron transfer from photoinduced excited states:
4.2.2.2

A + hν − − A∗
−→

(2)

The electron transfer may be either direct A∗ − − A+ + e−
−→
(3) or a variety of indirect processes as exempliﬁed by either
A∗ − − B;
−→
followed by
B − − B+ + e −
−→
or

(4)

−→
A∗ + B − − AB;

followed by
AB − − AB+ + e− (5)
−→
Solar-activated photodissociation processes generally involve cleavage of a simple molecule into several energetic products. Limitations of this approach include the limited absorption of solar energy by the molecule, low quantum

yield, rapid back reaction, and difﬁculties separating the product species. An example of storage by such photodissociation processes is exempliﬁed by
2NOCl − − 2NO + Cl2
−→

(6)

In photoisomerization, an absorbed photon activates molecular rearrangement and conversion of organic molecules into strained isomers. The products are stored.
Despite attractive features including a high heat storage capacity and good thermal stability, most systems tested have poor efﬁciencies. These systems necessitate transformation to strained conformations at high energies; energies consistent with wavelengths below 450 nm. This excludes much of the energy inherent in the Air
Mass 1 (AM1) solar spectra. The stored thermal energy is released on catalyticinduced reversion to the starting components. An example has been presented by
Kutal for the photocatalytic transformation of norbornadiene to quadricyclane [9].
Photochemical redox reactions can generate fuel formation, including H2 , CH3 , and CH3 OH. Because of its availability, the splitting of water to produce H2 has been the focus of particular attention. H2 O is transparent to near UV or visible radiation, and therefore sensitization is required to drive the water-splitting process. In early attempts on photoredox-splitting of water by Heidt and McMillian, the process was sensitized by using solution redox species such as Ce3+/4+ [10].
Ce3+ (aq) + H+ − − Ce4+ (aq) + 1 H2
−→
2
(7)
2Ce4+ (aq) + H2 O − − 2Ce3+ (aq)
−→
+ 2H+ + 1 O2
2

(8)

These processes have displayed poor quantum yields. As in photoisomerization

4.2 Photoelectrochemical Solar Energy Storage Cells nν Decomposition
∗TEOA

∗Ru(bpy) 2+
3

Ru(bpy)3 2+

Mv++
3
Ru(bpy)3 +

TEOA

Mv+•

TEOA

CH3 — N

H2O

N·(CH3·CH2·OH)3

Mv++

Pt

+

N — CH3

H2

+

Scheme 1

processes, these reactions are also generally driven only by high-energy radiation
(short wavelength) and cannot efﬁciently convert incident AM1 solar radiation.
H2 or O2 generation from water is a multielectron process. Optimization of photoredox-splitting of water necessitates the presence of a catalyst to mediate this complex multielectron transfer. In one such process, a sacriﬁcial reagent triethanolamine (TEOA) is consumed irreversibly in the process, as denoted in
Sch. 1 [9]:
Direct multielectron processes are rare and instead incorporate one or more radical intermediate steps. These reactive intermediates are susceptible to unfavorable side reactions, resulting in substantial losses in the energy-conversion process.
Kinetically favored back reactions further reduce the overall conversion efﬁciency.
The engineering of these complex molecular organizations provides a substantial scientiﬁc challenge and have generally resulted in systems with low conversion efﬁciencies. Semiconductor Photoredox
Storage
Semiconductor surfaces have been used as sensitizers to drive photochemical conversion and storage of solar energy. In
4.2.2.3

principle, this should lead to a higher level of photon absorption and more effective charge separation. Both effects can substantially increase solar photochemicalconversion efﬁciency, but these systems have not yet displayed high efﬁciencies of fuel generation or long-term stabilities.
Photoredox processes at semiconductor electrodes generating fuels or products other than hydrogen, including methanol and ammonia, have been attempted with low overall yields. The photoelectrolysis of
HI into H2 and I3 − at p -InP electrodes has been described [11]. These H2 and I3 − photogenerated products are prime candidates for a fuel cell. Analogous advanced systems, in which the photoelectrochemically generated fuels have been successfully recombined to generate electrical energy, are discussed later in this section.
A system exemplifying photoelectrochemical synthesis to generate hydrogen is water photoelectrolysis. An early demonstration of water photoelectrolysis used TiO2 (band gap 3.0 eV) and was capable of photoelectrolysis at ∼0.1% solar to chemical energy–conversion efﬁciency [12]. The semiconductor SrTiO3 was demonstrated to successfully split water in a direct photon-driven process by
Bolts and Wrighton (1976), albeit at low solar energy–conversion efﬁciencies [13].

321

322

4 Solar Energy Conversion without Dye Sensitization

The high SrTiO3 band gap, Eg , of 3.2 eV creates sufﬁcient energetic charge to drive the photoredox process. This excludes the longer wavelength photons and corresponds to only a small fraction of incident solar radiation. To improve the solar response, Eg has to be lowered; in a single band gap system, an optimum efﬁciency can be expected around 1.4 eV.
In photoelectrochemical water-splitting systems, corrosion of the semiconductor photoelectrodes can pose a signiﬁcant problem. Most surface-stabilizing redox reactions compete with oxygen and hydrogen generation and must be excluded from these systems. To enhance the solar response of high band gap materials, techniques such as dye sensitization and impurity sensitization have been attempted, although with little improvement [14]. Semiconductor surfaces have been modiﬁed to protect low band gap materials against photocorrosion [15, 16]. A self-driven photoelectrochemical cell consisting of Ptcoated p -InP and Mn-oxide-coated n-GaAs has been demonstrated to operate at 8.2% maximum efﬁciency to generate H2 and
O2 under simulated sunlight [17], and more recently a two band gap cell in a tandem arrangement has been used to split water at 12% efﬁciency [18]. A multijunction GaAs, Si cell has been recently used to drive water-splitting at over 18% solar to electrical-conversion efﬁciency [19].
Colloids and suspensions of semiconductors have been used for the photoredoxsplitting of water. The principle advantage of a ﬁne suspension is the large active surface area available. Reaction rates of H2 and O2 generation have been enhanced by loading the particles with small deposit of precious metals, and although signiﬁcant progress was made in this direction, a practical system is yet to be demonstrated [11].

4.2.3

Modes of Photoelectrochemical Storage

Conversion of a regenerative PEC to a photoelectrochemical storage solar cell (PECS) can incorporate several increasingly sophisticated solar energy conversion and storage conﬁgurations.
Two-Electrode Conﬁgurations
A variety of two-electrode conﬁgurations have been investigated as PECS systems.
Important variations of these photoelectrochemical conversion and storage conﬁgurations are summarized in Table 1. In each case, and as summarized in Fig. 3 for the simplest conﬁgurations, exposure to light drives separate redox couples and a current through the external load. There is a net chemical change in the system, with an overall increase in free energy.
In the absence of illumination, the generated chemical change drives a spontaneous discharge reaction. The electrochemical discharge induces a reverse current. In each case in Table 1, exposure to light drives separate redox couples and current through the external load.
Consistent with Fig. 1, in a regenerative
PEC, illumination drives work through an external load without inducing a net change in the chemical composition of the system. This compares with the twoelectrode PECS conﬁgurations shown in
Fig. 3(a) and (b). Unlike a regenerative system, there is a net chemical change in the system, with an overall increase in free energy. In the absence of illumination, the generated chemical change drives a spontaneous discharge reaction.
The electrochemical discharge induces a reverse current. Utilizing two quasireversible chemical processes, changes taking place in the system during illumination can be reversed in the dark. Similar to
4.2.3.1

4.2 Photoelectrochemical Solar Energy Storage Cells
Tab. 1 Important two-electrode photoelectrochemical conversion and storage conﬁgurations Scheme

Electrode 1

Electrode 2

SPE
SPE
SPE
SPE-Redox ASPE
SPE

I
II
III
IV
V

Electrolyte(s)
| Redox A Redox B |
| Redox A-membrane-Redox B |
| Redox A |
| Redox B |
| Redox A-membrane-Redox B |

CE
CE
Redox BCE -CE
CE
SPE

Note: Components of these systems include a semiconductor photoelectrode (SPE) and a counter electrode (CE). At the electrode–electrolyte interface, redox couples ‘‘A’’ or ‘‘B’’ are either in solution (| Redox |), counter electrode–conﬁned (| Redox BCE -CE) or conﬁned to the semiconductor photoelectrode (SPE-Redox ASPE |). e− R′

R hν n-type semiconductor R′

O hν x−
O

L

e

L

x−
R

O′

n-type semiconductor Storage electrode O′

Membrane

Counter electrode Schematic diagram of a two-electrode storage cell. On the left the storage electrode contains an insoluble redox couple, and on the right a soluble redox couple, with storage represented as

Fig. 3

hν

−→
R + O −− −− O + R
←−

a secondary battery, the system discharges producing an electric ﬂow in the opposite direction and the system gradually returns to the same original chemical state.
Each of the cells shown in Fig. 3 has some disadvantages. For both bound
(Fig. 3left) and soluble (Fig. 3right) redox couples, the redox species may chemically react with and impair the active materials of the photoelectrode. Furthermore, during the discharge process, the

photoelectrode is kinetically unsuited to perform as a counter electrode. In the absence of illumination, the photoelectrode P, in this case a photoanode, now assumes the role of a counter electrode by supporting a reduction process. For the photoanode to perform efﬁciently during illumination (charging), this very same reduction process should be inhibited to minimize photooxidation back reaction losses. Hence, the same photoelectrode

323

324

4 Solar Energy Conversion without Dye Sensitization

cannot efﬁciently fulﬁll the duel role of being kinetically sluggish to reduction during illumination and yet being kinetically facile to the same reduction during dark discharge. The conﬁguration represented in Fig. 3 has another disadvantage, the disparity between the small surface area needed to minimize photocurrent dark current losses and the large surface area necessary to minimize storage polarization losses to maximize storage capacity [20].

conﬁguration as shown in Fig. 4. In Fig. 4, the switches E and F are generally alternated during charge and discharge. During the charging, only switch E may be closed, facilitating the storage process, and during discharge, E is kept open while F is closed. In this case, chemical changes that took place during the storage phase are reversed, and a current ﬂow is maintained from the storage electrode to a third
(counter) electrode that is kept in the ﬁrst compartment. To minimize polarization losses during the discharge, this third electrode should be kinetically fast to the redox couple used in the ﬁrst compartment.
Still an improved situation would be to

Three-Electrode Conﬁgurations
Several of the two-electrode conﬁguration disadvantages can be overcome by considering a three-electrode storage cell
4.2.3.2

E
••
L
••
F

R′

R

R hν EL

O′

O

O

P
First compartment

A

M

S

Second compartment

Schematic diagram for a storage system with a third electrode (counter electrode) in the photoelectrode compartment. P = Photoelectrode,
A = Counter electrode, M = Membrane, S = Storage electrode,
EL = Electrolyte, E, F = Electrical switch, and L = Load.
Fig. 4

4.2 Photoelectrochemical Solar Energy Storage Cells

have both switches closed all the time. In this case, electric current ﬂows from the photoelectrode to both counter and storage electrodes. The system is energetically tuned such that when insolation is available, a signiﬁcant fraction of the converted energy ﬂows to the storage electrode. In the dark or diminished insolation, the storage electrode begins to discharge, driving continued current through the load.
In this system, a proper balance should be maintained between the potential of the solar energy–conversion process and the electrochemical potential of the storage process. There may be residual electric ﬂow through the photoelectrode during dark cell discharge, as the photoelectrode is sluggish, but not entirely passive, to a reduction process. This can be corrected by inserting a diode between the photoelectrode and the outer circuit.
4.2.4

Optimization of Photoelectrochemical
Storage

The power obtained is the product of voltage and current, and consideration of the photocurrent is as important as the photovoltage. If the band-bending is sufﬁciently large, then the minority carrier redox reaction, which is essential to maintain the photocurrent, can compete effectively with the recombination of photogenerated electron hole pairs. This recombination represents a loss of absorbed photo energy. Therefore, an objective is to maintain a high band-bending and at the same time a signiﬁcant photovoltage. A photoanode creates a negative photovoltage under illumination, which results in reducing the band-bending. In principal, one way to accomplish high band-bending is to choose a very positive redox couple in the electrolyte.
The converse is true for a photoanode.

Improvements relating to its stability and conversion efﬁciency are of paramount importance. Improvements of the
Photoresponse of a Photoelectrode
To improve the solar response of a photoelectrode, a proper match between the solar spectrum and the band gap of the semiconductor should be maintained.
When a single band gap semiconductor is used, a band gap in the vicinity of 1.4 eV is most desirable from the standpoint of optimum solar-conversion efﬁciency. An important criterion is that the minority carrier that is driven toward the semiconductor–electrolyte interface should not participate in a photocorrosion reaction that is detrimental to the long-term stability of the photoelectrode. Photocorrosion can be viewed in terms of either kinetic or thermodynamic considerations and the real cause may be a mixture of both. From thermodynamic perspective, a photoanode is susceptible to corrosion if the fermi level for holes is at a positive potential with respective to the semiconductor corrosion potential [21]. The corrosion can be prevented or at least inhibited by choosing a redox couple that has its Eredox more negative than that for the corrosion process [22,
23]. The kinetic approach has been to allow another desired redox process to occur at a much faster rate than the photocorrosion reaction [13]. Other attempts to minimize the photocorrosion has been to coat the photoelectrode surface with layers such as
Se [24] and protective conductive polymer ﬁlms [25], and to search for alternate low band gap semiconductors [26]. Extensive reviews on the performance and stability of cadmium chalcogenides include those by Cahen and coworkers, 1980 [27] and
Hodes, 1983 [28]. Etching of photoelectrode surface has been recognized and
4.2.4.1

325

326

4 Solar Energy Conversion without Dye Sensitization

widely used as an important treatment to achieve high-conversion efﬁciency [29].
This effect is mostly attributed to removal of surface states that may act as trapping centers for photogenerated carriers.
A related procedure called photoetching, initially developed for CdS and then applied to a wide variety of semiconductors, improves the photoelectrode performance and preferentially removes the surface defects acting as recombination centers [30].
In addition to the variety of etching procedures, several other surface treatments have been used to improve photoelectrode performance. Examples include a Ga3+ ion dip on CdSe [31], ZnCl2 dip on thin ﬁlm
CdSe [32], Ru on GaAs [33], Ru on InP [34], and Cu on CdSe [34]. Reasons explaining the effectiveness of these dips range from a decrease of dark current to electrocatalysis by surface-deposited metal atoms. Solution phase chemistry of the electrolyte is an important parameter that has been shown to dramatically inﬂuence photoeffects. The equilibrium position of the redox couple will affect equilibrium band-bending.
A photoanodic system with a solution containing a more positive potential redox couple causes a greater band-bending, which in turn leads to a higher photovoltage and efﬁcient carrier separation under normal experimental conditions.
Effect of the Electrolyte
Semiconductor photoeffects in a complex redox electrolyte are largely affected by the solution properties such as solution redox level, interfacial kinetics (adsorption), conductivity, viscosity, overall ionic activity, solution stability, and transparency within a crucial wavelength region. Redox electrolytes are known to inhibit unfavorable phenomena such as surface recombination and trapping [35]. In addition, the solution redox couple may induce
4.2.4.2

a favorable inﬂuence on the PEC system by improved charge-transfer kinetics leading to improved stability of the photoelectrode [36]. Additives incorporated in redox electrolytes are known to enhance the performance of PEC systems. Addition of small concentrations of Se in polysulﬁde electrolyte is known to improve the stability of CdSe (single crystal)/polysulﬁde system [22]. In this case, Se improves the
PEC performance by reducing S/Se exchange and by increasing the dissolution of the photooxidized product S, which is the rate-determining step in the oxidation of sulﬁde at the anode. Addition of Cu2+ into the I− /I3 − electrolyte is known to improve the stability of CuInX2 photoelectrode considerably [37], and in the same electrolyte, tungsten and molybdenum dichalcogenide photoelectrochemistry can be substantially improved by addition of
Ag+ , or other metal cations, and shift of the I− /I3 − Eredox [38].
In the case of CdSe/polysulﬁde system, solution activity, conductivity, efﬁciency of the photoanode (ﬁll factor), chargetransfer kinetics at the interface, and the stability of the photoelectrode are known to exhibit improvements in the trend
Li > Na > K > Cs > for alkali polysulﬁde electrolyte. This trend is explained in terms of the secondary cation effect on electrochemical anion oxidation in concentrated aqueous polysulﬁde electrolytes [39]. In the case of Cd(Se,Te)/polysulﬁde system, the efﬁciency of light energy conversion is improved by using a polysulﬁde electrolyte without added hydroxide because of the combined effect of increasing the solution transparency, relative increase of
S4 2− , and decrease in S3 2− in solution.
For the same photoelectrode–electrolyte system, an optimum photoeffect was observed for a solution containing a sulfur–sulﬁde ratio of 1.5 : 2.1 with l : 2 molal

4.2 Photoelectrochemical Solar Energy Storage Cells

potassium sulﬁde concentrations because of the combined effect of optimized solution viscosity, transparency, activity, and shift in solution redox level [40]. Stability of the polysulﬁde redox electrolyte, which is another parameter that determines the long-term performance of a
PEC cell, has been shown to increase with sulfur and alkali metal sulﬁde concentration and to decrease with either increasing
−OH− concentration or at high ratio of added sulfur to alkali metal sulﬁde [39,
41]. The combined polysulﬁde electrolyte optimization can substantially enhance cadmium chalcogenide photoelectrochemical conversion.
Chemical composition of the electrolyte is a particularly important parameter in PEC systems based on complex electrolytes, such as polysulﬁde or ferro/ ferricyanide. In the latter redox couple, replacement of a single hexacyano ligand strongly changes the photoelectrochemical response of illuminated n-CdSe [42], and addition of the KCN to the electrolyte can increase n-CdSe and n-CdTe photovoltage by 200 mV [43].
Effect of the Counter Electrode
In a photoanodic system, even at moderate current densities, the occurrence of sluggish counter electrode kinetics for the cathodic process will cause signiﬁcant polarization losses and diminish the photovoltage. Minimization of these kinetic limitations necessitates a counter electrode with good catalytic properties. For example, as shown by [34], CoS on stainless steel or brass electrodes exhibits electrocatalytic properties toward polysulﬁde reduction and overpotentials as low as
1 mV cm2 mA−1 has been realized. Composition of a particular redox electrolyte may have a bearing on the extent of counter electrode polarization [32, 39].

In PEC systems, a compromise is maintained to simultaneously optimize the photoelectrode efﬁciency, stability, and electrolytic properties of the electrolyte.
Practical PEC systems often require large working and counter electrodes and their geometric conﬁguration within the PEC system will effect mass transport and effective cell current. In some cases, advantageous use has been made of selective sluggish counter electrode kinetics toward certain cathodic processes. For example, carbon is a poor cathode for H2 evolution compared to Pt, and the direct hydrogenation of anthraquinone at a PEC cathode has been avoided by using a carbon anode [44]. In this case, such hydrogenation represents an undesirable side reaction.
Combined Optimization of Storage and Photoconversion
An efﬁcient photoelectrochemical conversion and storage system requires not only an efﬁcient functional performance of the separate cell components but also a system compatibility. In the combined photoelectrochemical storage system, simultaneous parameters to be optimized include
4.2.4.4

4.2.4.3

1. minimization of light losses reaching the photoelectrode,
2. high photoelectrode–conversion efﬁciency of solar energy,
3. close potential match between the photopotential and the required storagecharting potential,
4. high current and potential efﬁciency of the redox storage process,
5. high energy capacity of the redox storage, 6. reversibility (large number of charge– discharge cycles of the redox storage),
7. stability of the photoelectrode,
8. stability of the electrolyte,
9. stability of the counter electrode,

327

328

4 Solar Energy Conversion without Dye Sensitization

10. economy and cost effectiveness, and
11. reduced toxicity and utilization of environmentally benign materials.
A photoelectrochemical solar cell implicitly contains an electrolytic medium. In the majority of laboratory PEC conﬁgurations, incident light travels through the electrolytic medium before illuminating the photoelectrode. The resultant light absorbance by the electrolyte is a signiﬁcant loss, which is avoided by use of a back cell conﬁguration. For example, the substantial absorptivity of dissolved polysenide species has been avoided in a n-GaAs photoelectrochemistry through the use of the back wall cell conﬁguration presented in
Fig. 5 [45].
The photoelectrochemical system shown in Fig. 4 is a combination of a photoelectrode, electrolyte, membrane, storage, and

a counter electrode. As an example of challenges that may arise in the combined photoconversion and storage system, consider an n−CdSe/polysulﬁde/tin sulﬁde version of Fig. 4 and consisting of Cell 1.
Cell: 1. CdSe | HS− , OH− , Sx 2− |
Membrane | HS− , OH− | SnS | Sn
With illumination, the cell exhibits simultaneous photoelectrode, counter electrode and storage reactions, and equilibria including Photoanode:
−→
HS− + OH− − − S + 2e− + H2 O
(9)
Photocompartment equilibria:
−→
S + S2− − − S( x + 1)2− x Counter electrode:
−→
S + 2e− − − HS− + OH−
Photoelectrode
contact

Sapphire window (10)

Counter electrode contact

Cu wire
Mounted
thin-film
PEC

hν hν hν

Ag paste
Au grid

1.9 µm thick n -GaAs

Photoelectrode contact Polyselenide
(aq)

Window
Epoxy

Fig. 5

(11)

A back wall n-GaAs/aqueous polyselenide photoelectrochemical cell.

4.2 Photoelectrochemical Solar Energy Storage Cells

Storage electrode:
−

−→
SnS + H2 O + 2e − −
Sn + HS− + OH−

(12)

Unlike the case of the analogous regenerative PEC system, in the preceding equations, sulfur formed at the photoanode (and dissolved as polysulﬁde species,
S(x+1) 2− for x = 1 to 4) is not balanced by the reduction that is taking place at the counter electrode because of the simultaneous reduction process taking place at the storage electrode. As a result, sulfur is accumulated in the photoelectrode compartment and is removed only in the subsequent discharge process. This dynamic

variation in electrolyte composition may have a profound inﬂuence on the stability of the photoelectrode and electrolyte and on cell potential. Hence, to minimize these effects, either excess polysulﬁde must be included in the photocompartment or a limit must be set to the maximum depth of cell charge and discharge.
Another important consideration is the energy compatibility between the photoconversion and the storage processes. This compatibility is referred to as the voltage optimization. Figure 6 presents the combined IV characteristics for an ideal photoelectrode and current–voltage curves for two alternative redox processes; process
A and process B. Vph is the maximum

Redox process A

Redox process B

P: IVphoto

A:

ISC

Current

B:

Vmax
EB

VPh

EA

Current–voltage curves for electrochemical storage processes, A or B.
Process A may be charged by the photodriven current–voltage curve P, whereas process B may not. In the photodriven IV curve P, Vmax is the voltage corresponding to the point of maximum

Fig. 6

Potential

power, Pmax , and Isc and Vph are the short-circuit current and open-circuit photopotentials, respectively. EA and EB refer to the redox potentials for redox processes, A and B, respectively. 329

330

4 Solar Energy Conversion without Dye Sensitization

photovoltage that can be generated. Isc is the short-circuit photocurrent corresponding to maximum band-bending. In Fig. 6, consider the electrochemical process represented by curve B. This process is located outside the region of potentials generated by the photoelectrode; it does not represent a potential storage system to be driven by a single photoelectrode. In such a case, a serial combination of more than one photoelectrode would be necessary. For a redox process to be a potential candidate for a redox storage system, the storage and photodriven current–voltage curves should intersect. Whereas Vph and
Isc correspond to zero power, the point
Pmax shown in Fig. 6 corresponds to the maximum power point. Solar energy conversion is accomplished at its maximum efﬁciency only during operation in the potential vicinity of Pmax .
By adjusting the electrical load L, shown in Fig. 4, the system can be constrained to operate near its maximum power efﬁciency. In this case, if the counter electrode is not polarized, the potential difference between photoelectrode and the counter electrode will be close to Vmax . If one chooses a facile redox process for the storage electrode, as indicated by the sharply rising IV curve for process A in Fig. 6, with Eredox in the vicinity of Vmax , then the potential during charge and discharge of the storage process will remain near Vmax .
As a result, the potential will be a highly invariant current variation through the load L, regardless of insolation intensity.
This situation represents an ideal match between solar energy conversion and storage processes within a PECS. Nonideality occurs with poor voltage-matching or kinetic limitations and polarization losses associated with the counter, storage, or photoelectrodes. Ideally, the membrane used to separate the two-cell compartments, as indicated in Fig. 4, must be permeable only to ions that will transport charge, but that will not chemically react or otherwise impair any electrode. The permeability of membranes is generally less than ideal. Different membranes permit other ions and water to permeate to a varying degree [46, 47]. Gross mixing of active materials across the membrane causes them to combine chemically and in the process lose energy. Favorable qualities that a membrane should exhibit are low permeability toward chemically reactive ions, low resistivity, mechanical integrity, and cost effectiveness.
4.2.5

High Efﬁciency Solar Cells with Storage
Multiple Band Gap Cells with
Storage
A limited fraction of incident solar photons have sufﬁcient (greater than band gap) energy to initiate charge excitation within a semiconductor. Because of the low fraction of short wavelength solar light, wide band gap solar cells generate a high photovoltage but have low photocurrent. Smaller band gap cells can use a larger fraction of the incident photons, but generate lower photovoltage. Multiple band gap devices can overcome these limitations. In stacked multijunction systems, the topmost cell absorbs (and converts) energetic photons, but it is transparent to lower energy photons. Subsequent layer(s) absorb the lower energy photons. Conversion efﬁciencies can be enhanced, and calculations predict that a 1.64-eV and 0.96-eV two–band gap system has an ideal efﬁciency of 38% and
50%, light of 1 and 1000 suns intensity, respectively. The ideal efﬁciency increases to a limit of 72% for a 36–band gap solar cell [48].
4.2.5.1

4.2 Photoelectrochemical Solar Energy Storage Cells

Recently, high solar conversion and storage efﬁciencies have been attained with a system that combines efﬁcient multiple band gap semiconductors, with a simultaneous high capacity electrochemical storage [49, 50]. The energy diagram for one of several multiple band gap cells is presented in Fig. 7, and several other conﬁgurations are also feasible [1, 2a, b].
In the ﬁgure, storage occurs at a potential of Eredox = EA+/A − EB/B+ . On illumination, two photons generate each electron, a fraction of which drives a load, whereas the remainder (1/xe− ) charges the storage redox couple. Without light – the potential falls below Eredox – the storage couple spontaneously discharges. This dark discharge is directed through the load rather than through the multijunction semiconductor’s high dark resistance.
Cell: 2. In Fig. 8, an operational form of the solar conversion is presented and a

storage cell described by the Fig. 7 energy diagram. The single cell contains both multiple band gap and electrochemical storage, which unlike conventional photovoltaics, provides a nearly constant energetic output in illuminated or dark conditions. The cell combines bipolar AlGaAs (Eg = 1.6 eV) and Si (Eg = 1.0 eV) and AB5 metal hydride/NiOOH storage.
Appropriate lattice-matching between AlGaAs and Si is critical to minimize dark current, provide ohmic contact without absorption loss, and maximize cell efﬁciency.
The NiOOH/MH metal hydride storage process is near ideal for the AlGaAs/Si because of the excellent match of the storage and photocharging potentials. The electrochemical storage processes utilizes MH oxidation and nickel oxyhydroxide reduction:
MH + OH− − − M + H2 O + e− ;
−→
EM/MH = −0.8 V vs SHE

(13)

1/xe–

Rload

e–

ECs

EFermi(ns)

–

e

Vs

FCw

hν E
Gs

EFermi(nw=ps)
+
EGw > hν > EGs ⇒ h

Vw hν EGw

Semiconductor

E Vs
Electrocatalyst

Electrocatalyst anode

V = Vw + Vo

Electrocatalyst cathode

(1–1/x)e–
1/xA → 1/xA+ ηA EAredox

1/xB → 1/xB+
E Bredox

ηB

Electrolyte

EFermi(pw)
Storage bipolar MPEC hν > EGW ⇒ h

+

E VW

p n Wide gap

Ohmic junction p n Small gap

Energy diagram for a bipolar band gap indirect ohmic storage multiple band gap photoelectrochemical solar cell (MBPEC).

Fig. 7

331

4 Solar Energy Conversion without Dye Sensitization
1/xe–
Storage current Nickel cathode

NiOOH + H2O + e–
Ni(OH)2 + OH–

x

Separator with KOH electrolyte

Load current N+-Si

P

N-Si

Au-Sb/Au

350 µm
(4 × 1019cm–3) 800 nm

1.0 µm
(1 × 10

(8 × 1015cm–3)

10 nm

20 nm

19cm–3)
+-Si

GaAs(buffer layer)

N-GaAs

1.0 µm

(1 × 1018cm–3)
N+-Al(0.15)Ga(0.85)As

1.7 nm

300 nm

(2 × 1017cm–3)

50 nm
(1 × 1018cm–3)

N-Al(0.15)Ga(0.85)As

P+-Al

P+-Al0.8Ga0.2As

AR coating [ZnS(50nm)/MgF2(70nm)]

18
–3
(0.3-0.15)Ga(0.7-0.85)As (1 × 10 cm )

e–

Metal hydride anode
MHx–1 + H2O + e–
MH + OH–

Photo current Fig. 8

(1−1/x)e-

Rload

P+ -GaAs

Au-Zn/Au

2hν
⇒
Illumination

332

The bipolar AlGaAs/Si/MH/NiOOH MBPEC solar cell.

NiOOH + H2 O + e− − −
−→
Ni(OH)2 + OH− ;
ENiOOH/Ni(OH)2 = 0.4 V vs SHE (14)
As reported [49] and as shown in
Fig. 9, the cell generates a light variation insensitive potential of 1.2–1.3 V at total (including storage losses) solar–electrical energy conversion efﬁciency of over 18%.
A long-term indoor cycling experiment was conducted to probe the stability of the AlGaAs/Si metal hydride storage solar cell [50]. Unlike the variable insolation of
Fig. 9, in each 24-hour cycle, a constant simulated AM0 (135.3 mW cm−2 ) illumination was applied for 12 hours, followed by 12 hours of darkness, and the cell potential, and storage (charge and discharge) currents monitored as a function of time

over approximately an eight-month period.
Figure 10 presents representative results for two-day periods occurring 0, 40, 140, and 240 days into the experiment. As summarized in the lower curves of the ﬁgure, the load potential is again nearly constant, despite a 100% variation in illumination
(AM0/dark) conditions. Over a 24-hour period, the load potential increases by ∼2% as the cell charges with illumination, followed by a similar decrease in potential as stored energy is spontaneously released in the dark. The cycles exhibited in Fig. 10 are representative, and as observed exhibit little variation on the order of weeks, and exhibit a variation of ∼1% over a period of months. In this ﬁgure, photopower is determined as the product of the measured cell potential and measured photocurrent.
Power over load is determined as the

4.2 Photoelectrochemical Solar Energy Storage Cells

Load potential
[mA]

1.5
1.0

Generated VCell insensitive to illumination variation

0.5
AlGa/Si MH multiple band gap storage solar cell

I Photo
[mA]

4
Area = 0.22 cm2

2
0

Dark

Illumination

Dark

Illumination

I Storage,
[mA]

2
Discharge

0

Discharge

Charge
−2

Charge insitu AB5/NiOOH
Metal hydride storage

0

6

12

18

24

30

36

42

48

Time,
[h]
Two days measured conversion and storage characteristics of the
AlGaAs/Si/MH/NiOOH MBPEC solar cell.

Fig. 9

product of the measured cell potential and measured load current.
Under constant 12-hour (AM0) illumination, the long-term indoor cycling cell generated a nearly constant photocurrent density of 21.2 (constant to within 1% or
±0.2 mA cm−2 ), and as seen in the top curves of Fig. 10, a photopower that varied by ±3%. The cell’s storage component exhibits the expected increase in charging potential with cumulative charging, which moves the system to a higher photopotential. The observed increase in photopower during 12 hours of illumination is because of this increase in photopotential with cumulative charging. A majority of the photogenerated power drives the redox

cell, and the remainder consists of the power over load during illumination, as illustrated in Fig. 10. In the dark, inclusive of storage losses, the stored energy is spontaneously released and this power over load during both 12-hour illumination and
12-hour dark periods is also summarized.
The cell is a single physical–chemical device generating load current without any external switching.
PECS Driving an External Fuel Cell
In the early 1980s, Texas Instruments,
Inc. developed an innovative program based on a hybrid photovoltaic storage that used imbedded multilayer photoanode and photocathode silicon spheres and was
4.2.5.2

333

4 Solar Energy Conversion without Dye Sensitization
7

7

6

6

5

5
Potential over load
Photopower
Power over load

4

Illumination cycle:
12 hour 135.5 mW cm−2 & 12 hour dark
Area: 0.22 cm2

3

3

AlGaAs/Si MH bipolar gap storage solar C
2/3 year summary

2

2

1

1

0

4
Power
[mW]

Load potential
[V]

334

0

2

40

42

140

142

240

0
242

Time, days
Fig. 10 Eight-months photopotential and power characteristics of the AlGaAs/Si/MH/NiOOH solar cell under fully charged AM0 conditions. Each day, the cell is illuminated for 12 hours and is in the dark for 12 hours.

designed to provide a close match between their maximum power point voltage and a solution phase bromine oxidation process in acidic solution. The program was discontinued, but the system has several attractive features.
Cell: 3 [51]. In the bromine-imbedded Si sphere system, energy stored as bromine is recovered in an external hydrogen bromine fuel cell. The conversion and storage reaction and cell conﬁguration are summarized by
−→
2HBr + H2 O + hν − −
H2 + Br2 + H2 O

(15)

Photo anode:
Contact metal | Ohmic contact | n-Si | p -Si | Surface metal | Solution

Photo cathode:
Contact metal | Ohmic contact | p -Si | n-Si | Surface metal | Solution
A description of the Texas Instrument cell action is provided in Fig. 11.
4.2.6

Other Examples of Photoelectrochemical
Storage Cells

Photoelectrochemical storage cells of different conﬁgurations have been suggested, designed, and tested for their performance, under sunlight or artiﬁcial illumination.
Although none of these conﬁgurations has attained the high solar to electrical conversion and storage efﬁciency of the system in the earlier sections, they

4.2 Photoelectrochemical Solar Energy Storage Cells
Glass panel cover Separator

H2

Hydrogen bromide cathode
1/2 H2 e− H e pn

np

e

h

Br2

HBr anode Br h h Glass p n h− e 1/2 Br2
Metal

pn

Conductive plane

Fuel cell converter Storage and heat exchanger Electrical energy

Hydrogen storage Thermal energy

Schematic diagram of the Texas Instruments Solar Energy
System. Illumination occurs through the electrolyte to produce hydrogen gas that can be stored as a metal hydride and bromine that can be stored as aqueous tribromide. A hydrogen-bromine fuel cell is used to convert chemical to electrical energy and regenerate the hydrogen bromide electrolyte in a closed loop cyclic system. Thermal energy can be extracted through a heat exchanger.

Fig. 11

are of signiﬁcant scientiﬁc interest and form a solid basis for further development toward future systems. The following three sections, covering PECS with either a solution, solid, or intercalation storage redox processes, and provides a brief summary of many of these investigations, with a particular emphasis on their performance.
PECS Cells with Solution Phase
Storage
Cell: 4 [52]. This is an example of the use of photoexcitable absorbers to promote a redox process using the following reaction sequence: 4.2.6.1

TH+ + 2Fe2+ + 3H+ − − TH4 2+ + 2Fe3+
−→
(TH+ = Thionene acetate) (16)

The back reaction between Leucothionene (TH4 2+ ) and Fe3+ is slow. Leucothionene is oxidized at the SnO2 electrode.
TH4 2+ − − TH+ + 2e− + 3H+
−→

(17)

Ferric cation is reduced at a Pt electrode in a second compartment.
As a result, the concentration ratio of
Fe3+ /Fe2+ is increased in the ﬁrst compartment and decreases in the second compartment, which is equivalent to a difference in chemical potential. The system returns to its original uncharged state by discharging in the dark. Only 60 mV of potential difference is equivalent to a decade change of the Fe3+ /Fe2+ concentration ratio. Hence, this concentration cell does not generate a signiﬁcant potential and the power density is low. The cell has the conﬁguration 335

336

4 Solar Energy Conversion without Dye Sensitization

SnO2 | 0.0001 M Thionene Acetate,
0.01 M FeSO4 , 0.001 M Fe2 (SO4 )3
H2 SO4 , pH = 1.7–2 | Ion-exchange membrane | 0.01 M FeSO4 0.001 M
Fe2 (SO4 )3 | Pt
Studied cell characteristics: Polycrystalline photoelectrode, Pt counter electrode, Eredox of Fe3+ /Fe2+ = 0.77 V vs
SHE, standard hydrogen electrode illumination 40–50 mW cm−2 by a tungsten lamp, initial current during discharge =
9.1 µA, and initial voltage = 10.9 mV.
Cell: 5 [53]. The photoelectrode used in this investigation was a powder-pressed sintered pellet. The high band gap of the semiconductor used (n-Pb3 O4 ; Eg =
2.1 eV) and electron hole recombinations at grain boundaries have contributed to an observed low-conversion efﬁciency, which drives the overall cell reaction:
−

−

I + 6OH + 6Fe

3+

Dark

−−→
−−
←−−
Light

IO3 − + 3H2 O + 6Fe2+

(18)

A saturated salt bridge is used between the cell compartments to minimize membrane IR loss, but it allows the active redox species to mix and chemically combine across the junction. The cell has the form n-Pb3 O4 | 0.1 M Fe3+ , Saturated
Fe2+ | Salt bridge | 0.1 M IO3 − ,
0.1 M I− | Pt
Studied cell characteristics: Polycrystalline photoelectrode, Pt counter electrodes, three-electrode conﬁguration as in
Fig. 4., Eredox of Fe3+ /Fe2+ = 0.77 V vs
SHE, Eredox of IO3 − /I− = −0 − 26 V vs
SHE, Illumination 60 mW cm−2 quartz halogen lamp, conversion efﬁciency =
0.09%,
FF = 0.38, photopotential =
172 mV, Vmax = 172 mV, charge efﬁciency of the battery = 74%, potential

difference between Pt electrodes before charging = 720 mV, and after charging =
840 mV.
Cell: 6 [54]. In this cell, a wide gap semiconductor (Eg = 3.3), BaTiO3 is used, capable of absorbing only near UV radiation and comprising less than 5% of available solar energy. This limits its practical use for solar energy conversion. The following storage couple is used:
Light

−−
Ce3+ + Fe3+ −−→ Ce4+ + Fe2+
←−−

(19)

Dark

The storage system operates well away from the maximum power point of the semiconductor device, and therefore storage and discharge efﬁciency is poor. The cell uses a salt bridge between the two compartments and is of the form
Single crystal | n-BaTiO3 | 0.1 M Ce2
(SO4 )3 , 0.005 M Ce(SO4 )2 | Salt bridge |
0.1 M Fe2 (SO4 )3 , 0.005 M FeSO4 | Pt
Studied cell characteristics: Pt counter electrodes, three-electrode conﬁguration as in Fig. 4, Eredox of Ce4+ /Ce3+ =
1.45 V vs SHE, Eredox of Fe3+ /Fe2+ =
0.77 V vs SHE, illumination sunlight, conversion efﬁciency = 0.01%, FF = 0.26, photopotential = 730 mV, Vmax = 0.33 V, charge efﬁciency 15%, potential across two
Pt terminals of the charged cell = 0.60, and short-circuit current = 0.12 mA.
Cell: 7 [55]. In this study, attempts have been made to improve the behavior of a
MoS2 electrode in HBr electrolyte. Equilibration between the electrode and HBr has been improved by subjecting the electrode to a dark anodic potential. A ratio of 10 between the areas of counter and working electrode is another favorable feature in this study to minimize the polarization resistance at the counter electrode.
Naﬁon-315 membrane contributes only a

4.2 Photoelectrochemical Solar Energy Storage Cells

moderate resistance of 20 ohm cm2 . The storage reaction and cell conﬁguration are
Light

−−→
−−
2Br− + 3I2 ←−− Br2 + 2I3 − (20)
Dark

n-MoSe2 | 0.1 M HBr, 0.01 M Br2 |
Naﬁon 315 | 1 M KI, 0.18 M I2 | Pt
Studied cell characteristics: Single crystal photoelectrode, Pt counter electrodes, cell conﬁguration as in Fig. 4,
Eredox of Br/Br− = 1.087 vs SHE, Eredox of I3 − /I− = 0.534 V vs SHE. Illumination
200 mW cm−2 Xe lamp, conversion efﬁciency = 6.2%, potential across two Pt terminals of the charged cell = 0.49, and short-circuit current = 0.5 mA.
Cell: 8 [55]. The next cell also uses a
Naﬁon membrane, but makes use of nCdSe to drive a polysulﬁde–polyselenide storage couple. Low-output power density is the biggest drawback in this cell. The storage reaction and cell conﬁguration are
Dark

−−
S0 + Se2− −−→ S2− + Se0
←−−

(21)

Light

n-CdSe | 1 M in Na2 S, S, NaOH | Naﬁon
315 | 1 M in M Na2 Se, Se, NaOH | Pt
Studied cell characteristics: Polycrystalline photoelectrode, Pt counter electrodes, cell conﬁguration as in
Fig. 4, Eredox of polysulﬁde electrolyte =
−700 mV vs SCE, saturated calomel electrode, Eredox of Se2 2− /Se2− = −800 mV vs
SCE, illumination 100 mW cm−2 Xenon lamp, conversion efﬁciency 4%, FF =
0.45, photovoltage = −400 mV, charged cell has an open-circuit voltage of 60 mV, and initial current across a 100 ohm across
Pt electrode = 0.5 mA.
Cell: 9 [44]. This study uses organic redox species for energy-storage purposes.
Stability of the n-WSe2 photoanode in iodine electrolyte and the stability of

anthraquinone redox couple have been demonstrated in this study. Any H2 evolution would carry out direct hydrogenation of AQ and associated side reactions, and therefore a carbon electrode has been selected because of the H2 over potential on this electrode. The cell underwent several deep charge and discharge cycles with reproducible performance. The storage reaction and cell form are
Light

−−
2I− + AQ + 2H+ −−→ AQH2 + I2 (22)
←−−
Dark

n-WSe2 | 1 M KI, 0.1 M Na2 SO4 , 0.5 M
H2 SO4 | Saturated KCl bridge | 5 × 10−2
M AQ, 0.5 M H2 SO4 | C
Studied cell characteristics: Single-crystal photoelectrode, C counter electrode during charging, Pt during discharging. The cell conﬁguration is similar to that in Fig. 4. Eredox of I3− /I− =
0.534 V vs SHE, Eredox of AQ/AQH2 , illumination 150 mW cm−2 He–Ne laser
(632.8 nm), conversion efﬁciency = 9%, discharge across a 10 ohm load produces a current of 1 mA cm−2 , and open-circuit voltage 200 mV.
Cell: 10 [44]. This study uses a p -WSe2 photocathode rather than n-WSe2 . During the cell discharge, oxidation of AQH2 at the surface of p -WSe2 indicates that the electrode has the duel role of being a cathode during the charging and being the anode during the discharge. As discussed earlier, this limits the activity and lowcurrent densities were observed. The storage reaction and cell conﬁguration are
Light

−−
AQ + 2H+ + 2I −−→ AQH2 + I2 (23)
←−−
Dark

p -WSe2 | 5 × 10−2 M AQ | Saturated |
1 M KI, 0.5 M H2 SO4 | Pt
Single crystal | 0.5 M H2 SO4 | Salt bridge | 0.5 M Na2 SO4 |

337

338

4 Solar Energy Conversion without Dye Sensitization

Cell: 11 [55]. The theoretical band gap of
WSe2 provides a near ideal single band gap match for the solar spectrum. But the following cell has some disadvantages. These include the low solubility of the storage redox couple employed, MV2+ and MV+• and the possibility of undesirable side reactions of the radical ion MV+• . Using dual (n-type and p -type) photoelectrodes expands the potential regime one can access for the redox-storage couple. The storage reaction and cell conﬁguration are
Light

−−
2I− + 2MV2+ −−→ 2MV+• + I2 (24)
←−−
Dark

−

n-WSe2 | I | MV2+ | p -WSe2
PECS Cells Including a Solid
Phase–Storage Couple
The earlier experimental investigations,
Cells 2–9, use only solution phase redox couples. However, as indicated in the following examples, a solid phase–storage couple may also be employed, which in principle tends to increase the cell’s storage capacity.
Cell: 12 [57]. Having at least one component in insoluble form may add compactness into the cell conﬁguration, although low conductivity of the insoluble active component may cause signiﬁcant polarization losses associated with the storage electrode, as exempliﬁed by the low conductivity of silver (chloride) in one of the next cells. The next four cells use a TiO2 polycrystalline photoelectrode. In the ﬁrst cell, the storage reaction and cell conﬁguration are
4.2.6.2

Studied cell characteristics: Polycrystalline photoelectrode, Pt counter electrodes, cell conﬁguration is similar to Fig. 4, Eredox of O2 , H+ /H2 O couple = 1.23 V vs NHE at pH = 1,
Eredox of Ag/Ag+ = 0.80 V vs NHE, normal hydrogen electrode illumination
500 W Hg lamp, conversion efﬁciency =
1%, photopotential = 0.28 V vs NHE, open-circuit voltage of the charged cell = 0.28 V, and short-circuit current =
0.3 mA cm−2 .
Cell: 13 [57]. Of the four TiO2 Cells
10–13, the following cell exhibited the highest short-circuit discharge current and voltage. However, during the charging process, a stationary concentration of
Ce4+ was observed in the photoanode compartment. This suggest the existence of competing process that consumes the oxidized species Ce4+ . The later is known to participate in photochemical reactions under illumination [10]. Considering the low concentration of the reduced form of active materials used with the photoanode, there is a possibility that the water oxidation becomes the dominant process during charging. In this study, it was observed that with a passage of a charge of 10 coulomb during charging, Ce4+ present in the photoanode compartment accounted for only 22% of the charge. In this second TiO2 photoelectrode cell, the storage reaction and cell conﬁguration are
Light

−−
Ce3+ + Ag+ −−→ Ce4+ + Ag (26)
←−−
Dark

−−
2H2 O + 4Ag+ −−→ 4H+ + 4Ag + O2
←−−

TiO2 | 1 M HNO3 , 0.05 M Ce2 (SO4 )3 ,
0.1 M Ce(SO4 )2 | Anion Speciﬁc Membrane | 1 M AgNO3 , 1 M KNO3 | Ag

(25)
TiO2 | 1 M HNO3 , 1 M KNO3 | Anion
Speciﬁc Membrane | 1 M AgNO3 , 1 M
KNO3 | Ag

Studied cell characteristics: Polycrystalline photoelectrode, Pt counter electrode, cell conﬁguration illumination etc. are similar to the earlier cell, Eredox

Light
Dark

4.2 Photoelectrochemical Solar Energy Storage Cells

of Ce4+ /Ce3+ vs SHE, charge efﬁciency of the cell without stirring =
18%, open-circuit voltage of the charged cell = 0.76, and initial short-circuit current 1.3 mA cm−2 .
Cell: 14 [57]. In this third TiO2 photoelectrode cell, the storage reaction and cell conﬁguration are
Light

−−
2Fe2+ + Cu2+ −−→ 2Fe3+ + Cu (27)
←−−
Dark

TiO2 | 1 M KNO3 , 0.01 M FeSO4 | Anion
Speciﬁc Membrane | 0.025 M CuSO4 , 1 M
KNO3 | Cu
Studied cell characteristics: Polycrystalline photoelectrode, Eredox of
Cu2+ /Cu = 0.34 V vs NHE, Eredox of
Fe3+ /Fe2+ = 0.77 vs NHE, open-circuit voltage of the charged cell = 0.3, and shortcircuit current 1.5 mA cm−2 .
Cell: 15 [57]. The wide band gap of
TiO2 is not an appropriate match to the solar spectrum. In this fourth TiO2 photoelectrode cell, the storage reaction and cell conﬁguration are
Light

−−→
−−
Fe2+ + AgCl ←−− Fe3+ + Ag + Cl−
Dark

(28)

TiO2 | 0.2 M KCl, 0.01 M FeCl2 |Anion
Speciﬁc Membrane | 0.2 M KCl | AgCl | Pt
Studied cell characteristics: Polycrystalline photoelectrode, cell conﬁguration and illumination are the same as in the previous cell, Eredox Fe3+ /Fe2+ = 0.77 V vs NHE, Eredox of AgCl/Ag+ = 0.22 V vs
NHE, open-circuit voltage of the charged cell = 0.39 V, and short-circuit current =
0.4 mA cm−2 .
Cell: 16 [58]. In this next cell, Ni is deposited during charge at 80% charge efﬁciency. Losses may be because of the

competing reaction of H2 evolution. Cell voltage of the charged cell is higher than the photovoltage available, which indicates the possible inﬂuence of another redox couple Ni(OH)2 /NiOH− occurring at a higher redox potential. Only about 55% of the charge stored can be recovered during discharge. The possibility of self-discharge reactions because of imperfect permeability of the membrane has been cited as a possible cause, and is further complicated by the complex ferro/ferricyanide equilibria that is known to occur (Licht, 1995).
In this cell, the storage reaction and cell conﬁguration are
Light

−−→
−−
2Fe(CN)6 4− + Ni2+ ←−− Fe(CN)6 3− + Ni
Dark

(29)

n-GaP | 0.2 M K2 SO4 , pH = 6.7, 0.05 M
0.05 M K4 Fe(CN)6 | Anion
K3 Fe(CN)6 ,
Speciﬁc Membrane | 0.05 M K2 SO4 , 0.2 M
NiSO4 , 0.06 M NiCl2 | Pt
Studied cell characteristics: Single crystal photoelectrode, Pt counter electrodes, cell conﬁguration is similar to Fig. 4.
Eredox of Fe(CN)6 3− /Fe(CN)6 4+ is 0.36 V vs NHE, Eredox of Ni2+ /Ni = −0.25 V vs NHE, illumination 500 W Hg lamp, conversion efﬁciency 13% for 450–540nm region, photovoltage = 0.63 V, opencircuit voltage of the charged cell = 0.75 V, short-circuit current = 4.3 mA cm−2 , and charge efﬁciency = 55%.
Cell: 17 [55]. The conversion efﬁciency data in the following cell reﬂect the poor quality of the GaAs material that was used, although in other studies, there has been higher efﬁciency GaAs
PEC (without storage). In this study, signiﬁcant polarization was observed and performance data of the storage cell was not reported. The storage reaction and cell

339

340

4 Solar Energy Conversion without Dye Sensitization

conﬁguration are
Dark

−−→
−−
Cd + Se2 2− + 2OH− ←−−
Light

Cd(OH)2 + 2Se2− (30) n-GaAs | 0.1 M Na2 Se, 0.1 M Se, 1 M
NaOH | Naﬁon | 2 M NaOH | Cd
Studied cell characteristics: Single crystal photoelectrode, Pt counter electrode, cell conﬁguration is similar to Fig. 4, Eredox of Se2− /Se2 2− = −800 mV vs SCE, Eredox of Cd/Cd(OH)2 = −1050 mV vs SCE.
Illumination 100 mW cm−2 Xe lamp, conversion efﬁciency = 4%, FF = 0.53, photopotential = −500 mV, and shortcircuit discharge current of the storage cell in the dark using Pt electrodes =
14.6 mA cm−2 .
Cell: 18 [55]. The next two cells use a polycrystalline n-CdSe photoanode. The following cell exhibited steady currenttime and voltage-time curves during the photoelectrochemical charging and dark discharging. The ﬂat discharge curve prevailed until the capacity of the sulﬁde electrolyte is exhausted. The storage reaction and cell conﬁguration are
−

Dark

−−→
−−
Cd + S + 2OH ←−− Cd(OH)2 + S2−
Light

(31) n-CdSe | 0.1 M in NaOH, Na2 S, 1 M in S,
Na2 Se, Se | Naﬁon | 2 M NaOH | Cd
Studied cell characteristics: Polycrystalline photoelectrode, Pt counter electrodes, cell conﬁguration is similar to
Fig. 4. Eredox of Sx 2 /S2− = −700 mV vs
SCE, Eredox of Cd(OH)2 /Cd = −1050 mV vs SCE, illumination 100 mV cm−2 Xe light, conversion efﬁciency = 4%, FF =
0.45, photovoltage = −400 mV during

discharge through Pt and Cd electrodes with a 100-ohm load, and a current of
8.3 mA cm−2 ﬂowed at cell voltage close to
175 mV.
Cell: 19 [59]. In this study, the possibility of using organic semiconductors to drive storage processes is demonstrated.
The process is in principle similar to a concentration cell. During photocharging, Prussian Blue (PB, Fe4 [FeII (CN)6 ]3 ) is reduced at the photocathode and PB is oxidized at the anode. In the dark, the redox process involving PB is reversed producing an electron ﬂow. Process ability, stability, and lack of photocorrosion make these low band gap organic materials very attractive for photoelectrochemical applications.
However, they are defect-based systems, and the very low conversion efﬁciencies and self-discharge appear to outweigh these beneﬁts. The storage reaction and cell conﬁguration are bilayer electrode: FeII [FeII (CN)6 ]3 4− + 4h+
4
Light

−−→ FeIII [FeII (CN)6 ]3
−−
←−− 4

(32)

Dark

counter electrode: FeIII [FeII (CN)6 ]3 + 4e−
4
Light

−−→ FeII [FeII (CN)6 ]3 4−
−−
←−− 4

(33)

Dark

ITO | P3MT | PB | 0.2 M KCl, 0.1 M
HCl | PB | ITO
Studied cell characteristics: Illumination
500 W Xenon lamp, the ITO/P3MT electrode has open-circuit voltage = 0.44 V, short-circuit photocurrent 0.09 µA cm−2 , and charge efﬁciency of the storage cell =
40%.
Cell: 20 [60]. Metal ions introduced into a solid β -alumina lattice behave like ions in solution. This study illustrates a compact solid-state storage cell that can be charged using solar energy. During charging, Fe

4.2 Photoelectrochemical Solar Energy Storage Cells

and Ti change their oxidation state and the charge balance is maintained by the migration of Na+ ions from one phase to the other. In the actual cell design, an n-type semiconductor is connected to the alumina phase containing Ti and p -material is connected to the phase containing Fe.
Limitations are the comparatively slow diffusion of ions in the solid electrolyte and resistance to ionic movement at various phase boundaries, and lower the energy output during discharge. In this device, back wall illumination demands the use of very thin semiconductor layers to minimize absorption losses and has the general form n-semiconductor | Na2 O.11(AlFeO3 ) |
Na2 O.11(Al2 O3 ) | Na2 O.11(AlTiO3 ) | p -semiconductor
Cell: 21 [47]. In this detailed study, selection of a Naﬁon-315 membrane was done on the basis of (1) stability in high alkaline sulﬁde solutions, (2) low IR drop, and (3) low permeability to sulﬁde. Maintaining an area ratio of 1 : 8 between photo and storage electrodes has minimized polarization at the storage electrode. The storage system was driven by three semiconductor PEC devices connected in series. Charging was done up to 90% of the capacity followed by complete discharge. Overall observed charge efﬁciency was 83%. Although the system was not fully optimized with respect to photoelectrode, electrolyte, and storage, voltage efﬁciency of 75% was obtained during discharge. Discharge curves were ﬂat until the stored active material was fully consumed. The storage reaction and cell conﬁguration are
Light

−−→
−−
S2− + Zn(OH)4 2− ←−− S0 + Zn + 4OH−
Dark

(34)

n − CdSe | 1 M in NaOH, Na2 S,
S Naﬁon-315 | 0.1 M ZnO, 1 M NaOH | C

Studied cell characteristics: Polycrystalline photoelectrode, Ni counter electrode, basic cell conﬁguration is based on Fig. 4, Eredox of Sx 2 /S2− = 0.500 V vs
SHE, Eredox of Zn/Zn(OH)4 2− = −1.25 V vs SHE, artiﬁcial illumination, conversion efﬁciency = 3%, photovoltage = −0.50 V, during discharge through 75-ohm load between C and Ni discharge current =
10 mA, and voltage = 0.6 V.
Cell: 22 [61]. This cell takes advantage of photocorrosion to drive a storage cell.
Under illumination, n-CdSe is decomposed and p -CdSe is electroplated, and the reverse occurs during cell discharge.
However, photoactivity depends on an optimized semiconductor surface, and in an environment where the surface is changed constantly, the surface optimization is lost.
This and the poor kinetics of the p -type photoreduction result in a continual deterioration of the photoactivity and cause low photoefﬁciency and low-discharge power density. The storage reaction and cell conﬁguration are
Light

−−→
−−
CdSe + 2h+ ←−− Se0 + Cd2+ (35)
Dark

and the other electrode in photoelectroplated by Cd
Light

−−→
−−
CdTe + 2e− ←−− Cd0 + Te2− (36)
Dark

n-CdSe | 0.1 M CdSO4 | p -CdTe
Cell: 23 [46]. This is a detailed study of a thin ﬁlm cell with moderately high outdoor solar efﬁciency, high storage efﬁciency, and an output that is highly invariant despite changing illumination. This study provides extensive details of the choice

341

342

4 Solar Energy Conversion without Dye Sensitization

of photoelectrode, membrane, and electrochemistry of the tin–tin sulﬁde redox storage. Cd(Se,Te) electrodes, compared to CdSe, improve the band gap match and increase solar-conversion efﬁciency. Two photoelectrodes in series were used to provide a voltage match to the storage redox couple in a cell of the form of Fig. 12.
The conversion and storage reactions and cell conﬁguration are presented as
Light

−−→
−−
SnS + 2e ←−− S2− + Sn Storage (37)
Dark

Light

−−
S2− −−→ S + 2e Photo electrode (38)
←−−
Dark

n-Cd(Se,Te) | 2 M in NaOH, Na2 S, S |
Redcad Membrane | 2 M in NaOH, Na2 S |
SnS | Sn
Studied cell characteristics: Bipolar series polycrystalline photoelectrode, CoS counter electrodes, cell conﬁguration is as shown in Fig. 4 without the need of switches E or F . Eredox of S/S2− =
−0.48 V vs NHE, Eredox of SnS/Sn,S2− =
−0.94 V vs NHE. Illumination sunlight,
500 mWhr cm−2 per day, conversion efﬁciency 6–7%, photovoltage = −600 mV, and storage efﬁciency >90%. After two weeks of continuous operation the overall solar to electrical efﬁciency (including conversion and storage losses) is 2–7%.
Cell: 24 [4]. The earlier cell is improved by a series of solution phase optimizations
(cesium electrolyte with low hydroxide and optimized polysulﬁde), to provide a higher photopotential and improved stability and also the use of a single crystal, rather than thin ﬁlm, Cd(Se,Te) to also improve photopotential and cell efﬁciency, as described earlier in Fig. 12.
Because of the higher photopotential, only a single photoelectrode is required to match the storage potential and high overall efﬁciencies are observed. The cell

has the design as shown in beginning of the chapter (Fig. 2) and uses conversion and storage reactions described in the earlier cell and a conﬁguration n-Cd(Se,Te) | 0.8 M Cs2 S, 1 M Cs2 S4 |
Redcad Membrane | 1.8 M Cs2 S | SnS | Sn
Studied cell characteristics: The PEC had a power conversion efﬁciency of 12.7% under 96.5 mW cm−2 insolation and voltage at maximum power point was of
−1.1 V vs SHE, sufﬁcient to drive the
SnS/Sn storage system. Under direct illumination, the 0.08 cm2 single crystal photoelectrode generated more than 1.5 mA through the 3 cm3 SnS electrode driving
SnS reduction while supporting 0.33 mA through a 1500 load simultaneously at a photogenerated 0.495 V. In the dark spontaneous oxidation drive, the load with storage efﬁciency over 95%. The total conversion efﬁciency, including conversion and storage losses, was 11.8%.
PECS Cells Incorporating
Intercalation
In photointercalation, illumination drives insertion storage into layer type compounds [62]. The photointercalation process can be characterized as
4.2.6.3

TX2 + e− (h) + p+ (h) + Msol + − −
−→
TMIN X2 + p+

(39)

where TX2 is generally a nonintercalated transition metal dichalcogenide. For this process to occur without the assistance of an external power source, a counter electrode is driven at an electrode potential negative to that of the layer type intercalating electrode. The process is generally restricted to p -type materials. The development of this concept has been slow because of dearth of materials that are stable semiconductors and at the same time behave as

4.2 Photoelectrochemical Solar Energy Storage“hbox Cells
Bipolar Cd Se Tel [polysulfide]lCoS PEC with in-situ tini[sulfide]ll[polysulfide]lCos storage hν hν
Cd Se Te photo
Cd Se Te photo
CoS counter
A2
CoS counter
A
Membrane
B
Load

Tin storage

Fig. 12 A bipolar thin ﬁlm photoelectrochemical solar cell with in situ storage.
Compartments A and A2 contain alkali polysulﬁde solution and compartment B contains alkali sulﬁde solution.

intercalating compounds that are able to exchange guest ions and molecules with an electrolyte in a reversible manner, and yet that is not disruptive to photon absorption.
Cell: 25 [63]. In this cell, Eredox of copper thiophosphate is variable depending on the degree of intercalation. A limitation of this system is poor-discharge kinetics and low-energy density of the discharge. The cell conﬁguration is given by
Cu3 PS4 | 0.02 M CuCl | CH3 CN | Cu2 S
Studied cell characteristics: Eredox of
Cu+ /Cu0 = −0.344 vs NHE, illumination
117 mW cm−2 Xe lamp, photopotential = l 00 mV, charging current < 50 µA cm−2 , and discharge current < 10 µA cm−2 .
Cell: 26 [64]. This cell illustrates another all solid state design for a thin storage cell. p -Cux S changes its electrode potential with changes in its composition. During charging, Cu is oxidized at n-CdS surface while it is reduced at the Cu electrode.
Between the two electrodes Cu+ ion transport process takes place in the solid state electrolyte. The cell conﬁguration is given by
Cu | n-CdS | p -Cu2 S | RuCl4 I5 Cl3.5 | Cu
Cell: 27 [64]. As with the earlier cell, this ﬁnal cell requires a very thin design

because to reach the junction, light has to travel several layers. The cell functions in the same manner as the earlier cell, and the conﬁguration is given by
Conductive Glass | Cu | Cu+ Conducting solid electrolyte | p -Cu2 Te | n-CdTe | Mo
4.2.7

Summary

Conversion and storage of solar energy is of growing importance as fossil fuel energy sources are depleted and stricter environmental legislation is implemented.
Although society’s electrical needs are largely continuous, clouds and darkness dictate that photovoltaic solar cells have an intermittent output. Photoelectrochemical systems have the potential to not only convert but also store incident solar energy. Design component and system considerations and a number of photoelectrochemical solar cells with storage have been reviewed in this chapter.
Acknowledgment

S. Licht is grateful to Dharmasena Peramunage and for support by the BMBF
Israel–German Cooperation.

343

344

4 Solar Energy Conversion without Dye Sensitization

References
1. S. Licht, O. Khaselev, T. Soga et al., Electrochem. Solid State Lett. 1998, 1, 20.
2. (a) S. Licht, O. Khaselev, T. Soga et al., J.
Phys. Chem. 102, 2536; (b) ibid, 2546.
3. B. O’Regan, M. Gr¨tzel, Nature 1991, 353, a 737.
4. S. Licht, G. Hodes, R. Tenne et al., Nature
1987, 326, 863.
5. G. De Maria, L. D’Alesso, E. Coffari et al.,
C. A. Tiberio, Solar Energy 1985, 35, 409.
6. M. Gr¨tzel, Ber. Bunsenges. Phys. Chem., a 1980, 84, 981.
7. K. Kalyanasundaram, Coord. Chem. Rev.
1982, 46, 159.
8. A. Harriman, Photochemistry 1986–1987, 19,
509.
9. K. Kutal, Chem. Educ. 1984, 60, 882.
10. L. J. Heidt, A. F. McMillan, Science 1953,
117, 712.
11. R. Memming, Top. Curr. Chem. 1988, 143,
79.
12. A. Fujishima, K. Honda, Nature 1972, 238,
37.
13. J. M. Bolts, A. B. Bocarsly, M. C. Palazzotto et al., J. Am. Chem. Soc. 1979, 101, 1378.
14. H. P. Maruska, K. A. Ghosh, Solar Energy
1978, 20, 443.
15. G. Hodes, L. Thompson, J. Dubow et al., J.
Am. Chem. Soc. 1983, 105, 324.
16. R. C. Kaintala, B. Zelenay, J. OM. Bockris, J.
Electrochem. Soc. 1986, 133, 248.
17. R. C. Kaintala, J. OM. Bockris, Int. J. Hydrogen Energy 1988, 13, 3712.
18. O. Khaselev, J. Turner, Science 1998, 280,
4212.
19. S. Licht, B. Wang, T. Soga et al., H. Tributsch, J. Phys. Chem. B. 2000, 104, 8920.
19. S. Licht, B. Wang, T. Soga et al., H. Tributsch, J. Phys. Chem. B. 2000, 104, 8920.
20. J. Manassen, G. Hodes, D. Cahen, CHEMTECH 1981, 112.
21. A. J. Bard, M. S. Wrighton, J. Electrochem.
Soc. 1977, 124, 1706.
22. A. Heller, G. P. Schawartz, R. G. Vadimisky et al., J. Electrochem. Soc. 1978, 25, 1156.
23. G. Hodes, J. Manassen, D. Cahen, Nature
1976, 261, 403.
24. K. W. Frese, Appl. Phys. Lett. 1982, 40, 2712.
25. F.-R. F. Fan, B. Wheeler, A. J. Bard et al., J.
Electrochem. Soc. 1981, 128, 2042.

26. B. Miller, S. Licht, M. E. Orazem et al., Photoelectrochemical Systems, Crit. Rev. Surf.
Chem. 1994, 3, 29.
27. D. Cahen, G. Hodes, J. Manassen et al., ACS.
Symp. Ser. No. 1980, 146, 369.
28. G. Hodes, in Energy Resources Through Photochemistry and Catalysis, (Ed.: M. Gratzel),
Academic Press, New York, 1983, p. 242.
29. A. Heller, K. C. Chang, B. Miller, J. Electrochem. Soc. 1977, 124, 697.
30. R. Tenne, G. Hodes, Appl. Phys. Lett. 1980,
37, 428.
31. M. Tomkiewicz, I. Ling, W. S. Parsons, J.
Electrochem. Soc. 1982, 129, 2016.
32. J. Reichman, M. A. Russak, J. Appl. Phys.
1982, 52, 708.
33. B. A. Parkinson, A. Heller, B. Miller, J. Electrochem. Soc. 1979, 126, 954.
34. G. Hodes, J. Manassen, D. Cahen, J. Electrochem. Soc. 1980, 127, 544.
35. A. J. McEroy, M. Etman, R. Memming, J.
Electroanal. Chem. 1985, 190, 2212.
36. J. Reichman, M. A. Russak, J. Electrochem.
Soc. 1984, 131, 796.
37. D. Cohen, Y. W. Chen, Appl. Phys. Lett. 1984,
45, 746.
38. S. Licht, N. Myung, R. Tenne et al., J. Electrochem. Soc. 1995, 142, 840; S. Licht,
N. Myung, ibid, L129; ibid, 8412.
39. S. Licht et al., J. Electrochem. Soc. 1985, 132,
1076; ibid, 1986, 133, 52; ibid, 272; ibid, 277; ibid, 269.
40. S. Licht, Solar Energy Mat. Solar Cells 1995,
38, 3012.
41. S. Licht, Nature 1987, 330, 148.
42. S. Licht, D. Peramunage, Solar Energy 1994,
52, 197.
43. S. Licht, D. Peramunage, Nature 1990, 345,
330.
44. B. Keita, L. Nadjao, J. Electroanal. Chem.
1984, 163, 171.
45. S. Licht, F. Forouzan, J. Electrochem. Soc.
1995, 142, 1539; ibid, 1546.
46. S. Licht, J. Manassen, J. Electrochem. Soc.
1987, 134, 1064.
47. P. Bratin, M. Tomkiewicz, J. Electrochem. Soc.
1982, 129, 2649.
48. C. H. Henry, J. Appl. Phys. 1980, 51, 4494.
49. S. Licht, B. Wang, T. Soga et al., Appl. Phys.
Lett. 1999, 74, 4055.
50. B. Wang, S. Licht, T. Soga et al., Solar Energy
Mat. Solar Cells. 2000, 64, 311.
51. J. White, F-R. Fan, A. J. Bard, J. Electrochem.
Soc. 1985, 132, 544.

4.2 Photoelectrochemical Solar Energy Storage Cells
52. G. W. Murphy, Solar Energy 1978, 21, 403.
53. M. Sharon, S. Kumar, N. P. Sathe et al., Solar
Cells 1984, 12, 353.
54. M. Sharon, A. Singha, Int. J. Hydrogen Energy
1982, 7, 557.
55. P. G. Ang, P. A. F. Sammells, Faraday Discussions of the Chemical Society, General Discussions 1980, No. 70, Photoelectrochemistry, St. Catherine’s College, Oxford,
Sept 8–10.
56. F. -R. F. Fan, H. S. White, B. L. Wheeler et al., J. Am. Chem. Soc. 1980, 102, 5142.
57. H. Hada, K. Takaoka, M. Saikawa, Y. Yonezawa, Bull. Chem. Soc. Jpn. 1981, 54, 1640.

58. Y. Yonezawa, M. Okai, M. Ishino et al., Bull.
Chem. Soc. Jpn 1983, 56, 2873.
59. M. Kaneko, K. Takagashi, E. Tsuchida, J.
Electroanal. Chem. 1987, 227, 2512.
60. A. Sammells, A. Ang, Patent 1980, US
4,235,9512.
61. H. J. Gerritsen, W. Ruppel, P. Wurfel, J. Electrochem. Soc. 1984, 131, 2037.
62. H. Tributsh, Appl. Phys. 1986, 23, 61.
63. G. Betz, S. Fiechter, H. Tributsch, J. Appl.
Phys. 1987, 62(11), 4597.
64. T. Tonomura, K. Teratoshi, 1986, JP
Patent 62,249,366; ibid., JP Patent 62,249,
3612.

345

346

4 Solar Energy Conversion without Dye Sensitization

4.3

Solar Photoelectrochemical Generation of
Hydrogen Fuel
Maheshwar Sharon
Indian Institute of Technology, Bombay, India
Stuart Licht
Israel Institute of Technology, Technion, Israel
4.3.1

Introduction

Photoelectrolysis is a vast ﬁeld and it is difﬁcult to cover all its aspects in this chapter.
Therefore, this chapter conﬁnes its discussions to photoelectrolysis of water to obtain hydrogen gas. Hydrogen (H2 ) is an important renewable source of energy, as water is the main precursor for hydrogen.
Water (H2 O), on electrolysis produces hydrogen and oxygen (O2 ) in the ratio of
2 : 1. Moreover, when hydrogen is utilized as a fuel and oxidized to release its heat chemically (burnt) or electrochemically (in a hydrogen or air fuel cell), it produces water by consuming similar ratio of hydrogen and oxygen. Unlike other sources of energy such as coal, gas, and oil, hydrogen is a very clean source of energy. Hence, the hydrogen ⇔ water cycle seems to provide the most appropriate path toward a renewable source of energy.
4.3.2

Theoretical Consideration for Water
Electrolysis

Solar energy–driven water splitting combines several attractive features for energy utilization. Both the energy source (sun) and the reactive media (H2 O) are readily available and are renewable, and the resultant fuel (generated hydrogen) and the emission with fuel consumption (H2 O) are

each environmentally benign. Insolation
(solar radiation) on semiconductors can generate signiﬁcant electrical, electrochemical, or chemical energy [1–3]. Efﬁcient solar-driven water splitting requires a critical balance of the energetics of the solar conversion and the solution phase redox processes. The UV and visible energy-rich portion of the solar spectrum is transmitted through H2 O (Fig. 1 [4] and
Fig. 2 [5]). Therefore, sensitization, such as via semiconductors, is required to derive the water-splitting process. In a solar photoelectrolysis system, the redox-active interfaces can be in indirect or direct contact with the photosensitizer and can comprise either an ohmic or a Schottky junction.
Independent of this interface composition, the various parameters in models predicting solar water-splitting conversion efﬁciency may be combined into general parameters: (1) related to losses in optical energy conversion, ηphoto or (2) related to losses in redox conversion of H2 and O2 , ηelectrolysis . A combination of these parameters yields an overall solar to electrolysis efﬁciency (excluding storage and utilization losses) as ηphotoelectrolysis = ηphoto × ηelectrolysis (1)
Early photoelectrolytic attempts used solution redox species such as Ce3+/4+ and displayed poor quantum yield [6]. Further studies utilized semiconductors, such as
TiO2 (band gap, Eg = 3.0 eV) [7] or SrTiO3
(Eg = 3.2 eV) [8]. The wide Eg excludes longer wavelength insolation, leading to poor efﬁciency. H2 and O2 evolutions have been enhanced using large surface area or catalyst addition, but energy conversion efﬁciency remains low [9]. Early photoelectrolysis systems also combined p -type and n-type photoelectrodes [10, 11]. A two or more band gap conﬁguration can provide efﬁcient matching of the solar spectra.

4.3 Solar Photoelectrochemical Generation of Hydrogen Fuel
Fig. 1 Absorption spectrum of water (after K. W. Atanabe and
M. Zelikoff, J. Opt. Soc. Am.
1953, 43, 753).

ε cm−1 Atm−1

A
500

250

0
150
nm

100

Solar spectral irradiance
(w m−2 nm−1 ) at air-mass ratio
(a/m) of 0, 1, and 2 (after
Nikola Getoff, Int. J. Hydrogen
Energy 1990, 15(6), 407).
Fig. 2

a/m = 0

2.0
W m−2 nm−1

200

a/m = 1 a/m = 2

1.6
1.2
0.8
0.4
0
500

1000

1500

2000

nm

p -InP and n-GaAs were demonstrated at
8.2% efﬁciency to generate H2 and O2 [12].
A GaInP2 /GaAs photoelectrolysis cell was demonstrated at 12.4% efﬁciency [13] and a AlGaAs/Si photoelectrolysis cell was demonstrated at 18.3% efﬁciency [14].
A limited fraction of incident solar photons has sufﬁcient energy greater than
Eg to initiate charge excitation within a semiconductor. A semiconductor conﬁguration can drive water electrolysis under the conditions in which the generated photovoltage, Vphoto , is greater than the electrolysis potential, VH2 O . A complex energetic challenge exists to choose a semiconductor system with band gap at a combined maximum power point voltage tuned to VH2 O , in a system also providing high ηphoto (high ﬁll factor, photopotential, and photocurrent). Incident light of sufﬁciently energetic photons are absorbed; less energetic photons are transmitted.
Absorbed photons, hν > EGw , can stimulate e− / h+ pair excitation. A pn-type

or a Schottky junction electrochemical or a solid-state junction can inhibit charge recombination, driving charge at photopotential, Vphoto . In these junctions, Vphoto is constrained by the saturation current, jo , through to yield a maximum power photopotential, Vpmax , less than 60% of the band gap in efﬁcient devices [15].
Vphoto (jphoto ) = Vw + Vs = nkT e

ln 1 +

jphoto jo−w × 1+

jphoto jo−s (2)

Vpmax < 0.6(EGw + Egs )

(3)

To initiate electrolysis Vphoto must be greater than VH2 O . The thermodynamic o potential, EH2 O , for the water-splitting reaction is given by
−→
H2 O − − H2 + 1 O2 ;
2
o o o
EH2 O = EO2 − EH2 ;
◦

o
EH2 O (25 C) = 1.229 V

(4)

347

348

4 Solar Energy Conversion without Dye Sensitization

The subsequent challenge is to optimize sustained water electrolysis, without considerable additional energy losses. Effective water electrolysis must occur at a potential, VH2 O , near the photocell point of maximum power. VH2 O is greater than o EH2 O (= − GH2 O /nF ) as a result of overpotential losses, ξ , in driving an electrolysis current density, j , through both the O2 and the H2 electrodes:
VH2 O (j ) = EO2 (j ) − EH2 (j ) o o
= [EO2 + ξO2 (j )] − [EH2 + ξ H2 (j )] (5)

Planar platinum and Ptblack are examples of effective H2 electrocatalysts.
Minimization of ξO2 is a greater challenge. In the absence of competing redox couples, the faradic efﬁciency of H2 and
O2 evolution approaches 100%, and the ηelectrolysis is determined by the current limited VH2 O (j ): ηelectrolysis =
◦

ηelectrolysis (25 C) =

o
EH2 O

VH2 O (j )
1.229 V
VH2 O (j )

;
(6)

Thermodynamic heat considerations can be applied to photoconversion. Some amount of heat, Q, is always be lost due to charge carrier relaxation and vibrations.
This heat difference involves an entropy turnover Q/T = S . Photoenergy conversion efﬁciencies can be optimized by minimizing Hph . For band gap excitation,
Hph corresponds to Eg . If the absorbed photon energy is much higher than Eg , then Hph is correspondingly higher. If
Eg is much larger than the required electrolysis energy, GH2 O , then the predicted efﬁciency is low. The efﬁciency is higher if the photovoltaic system is matched to the chemical one. Consequently, there is

a thermodynamic photoefﬁciency describing the generation of electrochemical free energy for electrolysis:
GH2 O
Hph

(7)

In reality, how small can Hph be made, as compared with GH2 O ?
For photosynthesis, it has been estimated that of the 1.8-eV excitation energy of chlorophyll, at least 0.6–0.8 eV are lost before the energy can be stored in stable chemical products. Photosynthesis, however, requires many subsequent electron transfer steps, all of which contribute to efﬁciency losses. The molecules involved are also complicated and quite unstable compounds maintained through self-organization. Photoelectrolysis systems with many fewer components, and with fewer degrees of freedom in terms of chemical reactivity, can better approach the ideal energy conversion efﬁciency.
Contrarily, when fuel is consumed to reduce power (resulting in a fuel cell with efﬁciency G/ H ) the optimal efﬁciency for electrolysis must include entropy considerations, and this strongly depends on the temperature enthalpy change
( HH2 O (T )): ηelectrolysis−opt =

HH2 O
GH2 O

(8)

Practical electrolysis has to include losses so that the efﬁciency becomes ηelectrolysis−opt =

HH2 O
GH2 O + losses

(9)

A thermoneutral potential, Etn (Etn =
HH2 O /zF ), is deﬁned in which no heat turnover is observed and Etn = 1.48 V for water electrolysis [16]. If, because of effective catalysis, the total electrolysis cell voltage is close to 1.48 V, then ηelectrolysis ≈ 1.

4.3 Solar Photoelectrochemical Generation of Hydrogen Fuel

If we return to Eq. (1) and substitute
Eqs. (7) and (9), we obtain ηphotoelectrolysis = ηphoto × ηelectrolysis
=

GH2 O
Hph

1 + losses
GH2 O

(10)

Two conclusions arise from Eq. (10).
First, it is seen that efﬁciency can be maximized if electrochemical ‘‘losses’’ can be made small as compared with the Gibbs free energy change for water electrolysis,
GH2 O . Second, it is seen that HH2 O , the enthalpy of water electrolysis (with little temperature dependence), is involved in determining the overall efﬁciency as well as Hph , the enthalpy of photogenerated charge carriers. If these two can be properly matched, a maximum overall efﬁciency may be accomplished.
4.3.3

Photoelectrochemical Cell for
Photoelectrolysis

Conversion of solar energy into electrical energy can be achieved by using a

photoelectrochemical (PEC) solar cell conﬁguration similar to that described in the earlier chapter. In brief, a PEC cell consists of two electrodes separated by a suitable redox electrolyte. Both electrodes could either be n-type and p -type semiconductors or one electrode could be a semiconductor (either n-type or p type) and the other electrode could be a noncorrosive metal. If both electrodes are made of semiconductor, then the anodic electrode should be made from n-type semiconductor and the cathodic electrode should be made from p -type semiconductor. This last conﬁguration adds the additional requirement of photocurrent matching through the two photoelectrodes. Regarding n-type semiconductor
(Fig. 3), during its illumination photogenerated carriers are generated, which assist in the oxidation of electrolyte at the interface of the semiconductor and electrolyte.
A reverse phenomenon takes place if a p -type semiconductor is used, that is, photogenerated electrons are generated at the semiconductor–electrolyte interface to
RL

A
V

0.3 cm

A schematic representation of a photoelectrochemical cell. A, B, and C are counter electrode, electrolyte, and semiconductor electrode, respectively.
RL is used to vary the potential (V ) to measure photocurrent (A) ﬂowing across two electrodes.

Fig. 3

C

A

hν
B

349

350

4 Solar Energy Conversion without Dye Sensitization

initiate reduction of the electrolyte. Thus, in a PEC cell (Fig. 3), if water is used instead of a redox electrolyte, oxidation of water can take place at the illuminated n-type semiconductor (C) and reduction can take place at counter electrode (A).
However, for photodecomposition of water rather than a single redox couple, two different redox couples are used at the two immersed electrodes, with a provision to collect hydrogen and oxygen separately.

apply an external bias (Eb ) to facilitate the photoelectrolysis of water. The energy level of semiconductor and other photoelectrochemical reactions are shown in Fig. 4(b).
The semiconductor anode was illuminated with UV radiation. On illumination of
TiO2 electrode, the photogenerated holes oxidize water to produce oxygen and the photogenerated electrons are transferred to counter electrode to perform the reduction of water at the platinum electrode.

4.3.4

4.3.4.1

Energetics of Photodecomposition of Water
To accomplish photodecomposition of water, so that the PEC cell supplies the entire potential, energetic of the cell should meet the requirements as shown in Fig. 5. It is assumed that Fermi energy of n-type semiconductor (or its ﬂat band potential) is equivalent to water reduction potential. Alternatively, instead of using metal as a counter electrode, one could use a p -type semiconductor (Fig. 6). Because

Photoelectrolysis of Water

Fujishima and Honda [7, 17] were the ﬁrst to show the possibility of decomposing water through a PEC cell (Fig. 4a). They used a photoelectrochemical cell with an anode made up of a semiconductor electrode of n-TiO2 connected to a platinum black counter electrode through an external load.
Because photopotential developed by the
TiO2 electrode was insufﬁcient, they had to
E-bias

E-bias e− O2

e−

H2

e−

H2 e− EC

hν
Pt
Na2SO4

NaOH
Frit

n -TiO2
Anode
(a)

e−

Cathode

EF
EV

EG
P+

H

e−

EF
Eb

H2O e− OH + H+

O2
Electrolyte
n -TiO2
Helmholtz
Metal layer contact
(b)

Pt

(a) Fujishima-Honda cell with n-TiO2 photoanode and Pt-cathode. (b) Schematic energy level diagram of the cell. EV -valence band, EC -conduction band, EF -Fermi level,
EG -energy gap (for n-TiO2 ; EG = 3.0 eV), Eb -bias voltage, p+ -hole (after Nikola Getoff,
Int. J. Hydrogen Energy 1990, 15(6), 407).

Fig. 4

4.3 Solar Photoelectrochemical Generation of Hydrogen Fuel
−
−

Conduct band
−
−

EF,sc

nEF

∆V

hν

2+

+

+

n- semiconductor

2− −

−

Fermi-level

Stored energy H2O

pEF

Valence band

2H+

H2

1 O + 2H+
22

Electrolyte

Metal

Energy scheme of a cell with one n-type semiconductor electrode for photoelectrolysis of water. V is stored energy for electrolysis. p EF is Fermi level of photogenerated holes known as quasi-Fermi level, n EF is Fermi level of electron.
(after Heinz Gerischer, Pure Appl. Chem. 1980, 52, 2649).

Fig. 5

Load h+ 2H2O + 2e−

e

2OH− + H2

e−

EC

hν e EC
EF

EF

∼1.7 V
+

hν

Eν

+

2h + 2OH−

Eν

H2O + 1/2 O2

Energy scheme of a cell utilizing n-type semiconductor and p-type semiconductor. The difference between the energy levels of valence band edge of n-type and conduction band edge of p-type should be approximately equal to potential needed to electrolyze water (VH2 O ).

Fig. 6

the objective is to electrolyze water with a freely available energy, light source for the illumination has to be solar energy. Considering factors responsible to produce photocurrent or photopotential generated by semiconductors of various band gaps utilizing solar energy of various wavelengths and intensities, one can

calculate theoretical achievable efﬁciency for various band gaps. This relationship gives a parabolic curve with a maximum theoretical efﬁciency of about 30% for a band gap of 1.4 eV (Fig. 7). But, a semiconductor with a band gap of 1.4 eV would give a maximum photopotential of
∼700 mV (assuming its ﬂat band potential
=

351

4 Solar Energy Conversion without Dye Sensitization
35
Black-body limit (AMO)
30

25
Efficiency
[%]

352

AM1.5
Si

20

Cu2S

GaAs

a-Si : H a-Si : H : F

AMO

15
T = 300 K
CdS

Ge
10

5
0.5

1.0

1.5
Semiconductor band gap
[eV]

2.0

2.5

Theoretically calculated conversion efﬁciency of solar cell materials versus band gap for single junction cells (after Adolf Goetzberger, Christopher Hebling, Sol. Energy
Mater. Sol. cells 2000, 62, 1).

Fig. 7

matches with the reduction potential of water), which is not sufﬁcient to electrolyze water. On the contrary, for obtaining the necessary electrolysis photopotential using a single semiconductor, a semiconductor of much larger band gap is needed (Fig. 5).
With such material, biasing is needed, as is case with n-TiO2 (Fig. 4). Moreover, as a result of high band gap, n-TiO2 absorbs only about 2% of the solar spectra (Fig. 2).
Therefore, a semiconductor of large band gap would correspondingly give very low photocurrent (hence, lowphotoconversion efﬁciency), making the photoelectrolysis processes uneconomical.
Multiple-type PEC Cells
The calculations from the earlier section suggest that there is a need to introduce some improvements in the PEC cell to attain a sufﬁcient solar energy conversion
4.3.4.2

with photopotential larger than VH2 O .
Sharon and Rao [19] developed a photoelectrochemical cell with a semiconductor separating two types of electrolytes: one electrolyte forming an ohmic contact and the other forming a Schottky-type contact.
It is postulated that like the formation of a Schottky-type barrier, if the magnitude of the Fermi level of the semiconductor and the redox electrolyte is same, then the contact between them should be ohmic.
This provided an opportunity to visualize a
PEC cell in which the semiconductor acts like a separator between two types of redox electrolytes. The electrolyte is selected such that the front side of the semiconductor (i.e. the side to be illuminated) forms a Schottky junction. The backside of the semiconductor is kept in contact with another redox electrolyte, which gives an ohmic contact. The front side of the

4.3 Solar Photoelectrochemical Generation of Hydrogen Fuel

electrons migrate in opposite directions to arrive at the metal electrode (lead). Because this metal gives an ohmic contact, photogenerated electrons are easily transferred to the redox electrolyte to perform reduction of the electrolyte. Thus, while the front side (i.e. the illuminated side) performs oxidation of the redox electrolyte, the backside performs reduction of the electrolyte
(and also acts like a counter electrode).
When more than one such electrodes are used (Fig. 8a), a vectorial migration of holes takes place from left to right side while vectorial migration of electrons takes place from right to left side. These carriers
(electrons and holes) are transferred to the load by the two outer electrodes, that is,

semiconductor acts as a normal PEC cell and its backside acts as a counter electrode.
This type of cell has been classiﬁed as a
Sharon-Schottky-type cell [20]. To simplify this type of cell, this concept was extended to make a multiple-type PEC cell. It is assumed that a metal forms an ohmic contact with its own oxide. In this cell, three electrodes made of n-type semiconductor (n-Pb3 O4 ) deposited over a metal
(lead) are arranged in an array (Fig. 8a).
The front side of each electrode contains n-Pb3 O4 , which is dipped in a suitable redox electrolyte to form a Schottky-type barrier. On illumination of n-Pb3 O4 , electrode photogenerated holes perform oxidation of redox electrolyte. Photogenerated
Load

hν

Semiconductor electrode hν

Counter electrode hν

e

A

h+

B

e

e

e− + OX−
Red
carbon electrode
Z
h+

h+

h+

PEC cells

Red

e + OX−
Pb

e
Pb3O4

h+ − Red

Pb3O4

OX−

+

X
Lead

h+

Y

Schematic representation of multiple
PEC cells connected by the help of redox electrolyte. Inset X shows the energy level of semiconductor Pb3 O4 electrode deposited over metal lead. Photoelectrochemical reactions occurring at two ends of the semiconductor is shown. Inset Y shows the ohmic contact

Fig. 8

between the metal and the semiconductor and the direction of ﬂow of photogenerated electron and hole. Electron is extracted to the load via metal A and the hole is extracted via inert metal
B. Inset Z describes reduction process occurring at inert carbon electrode-B.

353

354

4 Solar Energy Conversion without Dye Sensitization
Fig. 9 A photograph of a multiple PEC cells (7 in number) connected in series by the help of redox electrolyte as described in Fig. 7. The output of this cell can be given to a normal electrolytic cell for electrolysis of water (after
M. Sharon et al., Electrochim. Acta 1991,
36(7), 1107–1126).

by the backside of the last left side electrode (A) and metallic electrode on the extreme right side of the cell (B). The total photopotential of such system is equal to the photopotential of one PEC cell times the number of semiconductor electrodes used [20]. This arrangement gives an opportunity to get the required electrolysis photopotential, even with the semiconductor of band gap ≈1.4 eV. In addition, the requirement for matching the conduction band edge (or ﬂat band potential) with the electrode potential for hydrogen evolution and the valence band edge with electrode potential of oxygen evolution (Figs. 5 and 6) becomes redundant. Photocurrent, however, would depend on the area of the individual semiconductor exposed to solar

n -SC
1/4 O2

1/2 H2O
(1)
Salt bridge

(2)

Bipolar Cell
The bipolar cell of Sharon-Schottky cell has been further developed. Bard and coworkers [21, 22] developed a bipolar semiconductor photoelectrode array (Fig. 10) and studied its application to light-driven water splitting and electrical power generation.
They used ﬁve n-TiO2 electrodes in series.
In the subsequent developments [23], they devised a similar bipolar cell (Fig. 11) with
CoS/CdSe. In this cell, also a salt bridge is used to complete the electrical circuit.
CoS is used to make ohmic contact with
4.3.4.3

Bipolar electrode O

R

radiation (assuming each semiconductor to be of the same area). Employing this cell as a source of electrical power, it also can be used for electrolyzing water (Fig. 9).

O

O

1/2 H2
H+

R

R
(3)

(4)
DC

DC(H2)

Salt bridge

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

Fig. 10

4.3 Solar Photoelectrochemical Generation of Hydrogen Fuel
Visible light flux
Hydrogen
collector

Oxygen collector CdSe
1
A
Pt

2
BC

3
DE

Pyrex tube
F

4

5
G

H

6

I

J
Pt

CoS

E

−

CdSe/Electrolyte interface hν
E ° 2− s2 s /2
+

CB
*
n EF

CdSe Pt
*
pEF

+

−

EF

VB

E°
H2 / H2O
Pt / Electrolyte interface Direction of electron flow

(a) Schematic representation of water photoelectrolysis cell. A, J: Pt; B, D, F, H: CoS; C, E, G, I: CdSe.
Solutions 1,6: KOH (1M); 2–5 Na2 S (1M), S (1M), KOH (1M).
For H2 and O2 generation solutions 1 and 6 are connected with KOH bridge. (b) Expansion shows energetics of bipolar panel (after A. J. Bard et al., J. Phys. Chem. 1987, 91, 6).
Fig. 11

n-CdSe as well as with the electrolyte. Four
CoS/CdSe electrodes are used. Polysulﬁde is used as an electrolyte in these PEC cells.
In this arrangement, the PEC cell is formed among (B, C), (D, E), (F, G), and (H, I).
Sides C, E, G, and I form Schottky-type barriers while sides B, D, F, and H act like a counter electrode. The ﬁrst and the last electrodes are made of platinum. The energy level diagram to show the ﬂow of electron is shown in Fig. 11.
Several oxides (e.g. TiO2 [7], SrTiO3 [8], n-SiC [24], p -GaP [25], Fe2 O3 [26], etc.) and chalcogenides such as CdSe [23] have been used as anodes for water photodecomposition. n-TiO2 , even today seems to be one of the most important materials, because it has been possible to extend its spectral response into visible portion of the solar spectrum through sensitization with organic dyes. A separate chapter is devoted

for sensitization of semiconductor electrode. Hence, this will not be discussed here any further.
4.3.5

Recent Developments

Principal solar water-splitting models had predicted similar dual band gap photoelectrolysis efﬁciencies of only 16% and
10–18% [3, 28], respectively, whereas recently dual band gap systems were calculated to be capable of attaining over
30% solar photoelectrolysis conversion efﬁciency [14]. The physics of the earlier models were superb, but their analysis was inﬂuenced by dated technology, and underestimated the experimental ηphoto attained by contemporary devices or underestimated the high experimental values of ηelectrolysis which can be

355

356

4 Solar Energy Conversion without Dye Sensitization

attained. For example, Ref. 3 estimates low values of ηphoto (less than 20% conversion) because of the assumed cumulative relative secondary losses that include 10% reﬂection loss, 10% quantum-yield loss, and 20% absorption loss. Experimentally, a cell containing illuminated AlGaAs/Si
RuO2 /Ptblack was demonstrated to evolve
H2 and O2 at record solar–driven water electrolysis efﬁciency. Under illumination, bipolar conﬁgured Al0.15 Ga0.85 As (Eg =
1.6 eV) and Si (Eg = 1.1 eV) semiconductors generate open circuit and maximum power photopotentials of 1.30 and 1.57 V, respectively, well suited to the water electrolysis thermodynamic potential:
H2 O − − H2 + 1 O2 ;
−→
2 o EH2 O = EO2 − EH2 ;
◦

o
EH2 O (25 C) = 1.229 V

(11)

o
The EH2 O /photopotential-matched semiconductors are combined with effective water electrolysis O2 or H2 electrocatalysts, RuO2 , or Ptblack . The resultant solar photoelectrolysis cell drives sustained water-splitting at 18.3% conversion efﬁciencies [14]. These recent developments in hydrogen generation at high solar energy conversion efﬁciency are detailed in this volume in the chapter titled ‘‘Optimizing Photoelectrochemical Solar Energy
Conversion: Multiple Band Gap and Solution Phase Phenomena.’’

4.3.6

Conclusion

In this section, efforts are made to discuss the thermodynamics of photoelectrolysis of water using a PEC cell. To facilitate the generation of required potential for the photoelectrolysis of water, discussions are made on some modiﬁed PEC cells popularly known as bipolar cell. Potential

developed in bipolar cells, amounts to connecting large number of PEC cells in series. The advantage of bipolar cells is that it can operate with low band gap semiconductor, yet provide desired photopotential and high solar to electrical efﬁciency. Although photopotential of such device is equal to the photopotential of one-photoelectrode times the number of PEC cells connected in series, photocurrent of system depends on the intensity of solar radiation falling on an individual photoelectrode. But it is important to realize that none of semiconductors so far developed can be used for prolonged photoelectrolysis of water, because of their inherent instability towards photocorrosion. Success of making bipolar cells as a commercial viable system entirely depends on the development of photoelectrochemically stable low band gap semiconductor.
Acknowledgment

S. Licht is grateful to Helmut Tributsch for his review and suggestions on theoretical sections of this chapter and thankful for the support by the BMBF Israel-German
Cooperation and the Berman-Shein Solar
Fund. Maheshwar Sharon is thankful to his students and in special to G. Ranga
Rao who contributed toward development and modiﬁcation of PEC cell.
References
1. M. A. Green, K. Emery, K. Bucher et al.,
Propgr. Photovolt. 1999, 11, 31.
2. S. Licht, B. Wang, T. Soga et al., Appl. Phys.
Lett. 1999, 74, 4055.
3. J. R. Bolton, S. J. Strickler, J. S. Connolly,
Nature 1985, 316, 495.
4. K. W. Atanabe, M. Zelikoff, J. Opt. Soc. Am.
1953, 43, 753.
5. Nikola Getoff, Int. J. Hydrogen Energy 1990,
15(6), 407.

4.3 Solar Photoelectrochemical Generation of Hydrogen Fuel
6. L. J. Heidt, A. F. McMillan, Science 1953,
117, 75.
7. A. Fujishima, K. Honda, Nature 1972, 238,
37.
8. J. M. Bolts, M. S. Wrighton, J. Phys. Chem.
1976, 80, 2641.
9. R. Memming, Top. Curr. Chem. 1988, 143,
79.
10. A. Nozik, Appl. Phys. Lett. 1976, 29, 150.
11. J. White, F. -R. Fan, A. J. Bard, J. Electrochem.
Soc. 1985, 132, 544.
¨
12. R. C. Kaintala, J. O. M. Bockris, J. Int. Hydrogen Energy 1988, 13, 375.
13. O. Khaselev, K. Turner, Science 1998, 280,
425.
14. S. Licht, B. Wang, S. Mukerji et al., J. Phys.
Chem. B 2000, 104, 8920.
15. R. Memming in Photochemical Conversion and Storage of Solar Energy, (Eds.: E. Pelizzetti,
M. Schiavello), Kluwer Academic Publishers,
Netherlands, 1991, pp. 193–212.
16. F. Gutman, O. J. Murphy in Modern Aspects of Electrochemistry (Eds.: White, Bockris,
Conway) 1983, P5.

17. A. Fujishima, K. Honda, Bull. Chem. Soc.
Jpn. 1971, 44, 1148.
18. Heinz Gerischer, Pure Appl. Chem. 1980, 52,
2649.
19. Maheshwar Sharon, G. Ranga Rao, Indian J.
Chem. 1986, 25A, 170–172.
20. M. Sharon et al., Electrochim. Acta 1991,
36(7), 1107–1126.
21. A. J. Bard et al., J. Phys. Chem. 1986, 90,
4606.
22. A. J. Bard et al., J. Electrochem. Soc. 1988, 135,
567.
23. Tooru Inoue, Toshihiro Yamase, Chem. Soc.
Japan, Chem. Lett. 1985, 869.
24. H. Honeyama, H. Sakamoto, H. Tamura,
Electrochim. Acta 1979, 277, 637.
25. Lynn C. Schumacher, Suzanne MamicheAfara, Michael F. Weber et al., J. Electrochem.
Soc. 1985, 132(12), 2945.
26. M. F. Weber, M. J. Digman, Int. J. Hydrogen
Energy 1986, 11, 225.
27. Ibid, J. Electrochem. Soc. 1984, 131, 1258.
28. Adolf Goetzberger, Christopher Hebling,
Sol. Energy Mater. Sol. cells 2000, 62, 1.

357

358

4 Solar Energy Conversion without Dye Sensitization

4.4

Optimizing Photoelectrochemical Solar
Energy Conversion: Multiple Bandgap and
Solution Phase Phenomena
Stuart Licht
Technion – Israel Institute of Technology,
Haifa, Israel
4.4.1

Introduction

This chapter focuses on two concerted efforts to achieve stable, high solar energy conversion efﬁciency using a variety of photoelectrochemical systems. The ﬁrst effort explores the use of multiple band gap semiconductor systems. Limiting constraints of multiple band gap photoelectrochemical energy conversion as well as practical conﬁgurations for efﬁcient solar to electrical energy conversion have been probed [1–6]. Such systems are capable of better matching and utilization of incident solar radiation (insolation). Efﬁcient solar cells, solar storage cells, and solar hydrogen generation systems are discussed and demonstrated. The principles of photoelectrochemical phenomena systems can differ substantially from conventional solid-state physics. Photoelectrochemical systems can be characterized not only by semiconductor but also often by electrolytic limitations. The second half of the chapter focuses on the substantial improvements to photoelectrochemical energy conversion attained by understanding and optimizing of such solution phase phenomena [7, 8].
4.4.2

Multiple Band Gap Photoelectrochemistry
Theory of Multiple Band Gap Solar
Cell Conﬁgurations
Radiation incident on semiconductors can drive electrochemical oxidation or
4.4.2.1

reduction and generate chemical, electrical, or electrochemical energy. Energetic constraints imposed by single band gap semiconductors had limited values of photoelectrochemical solar to electrical energy conversion efﬁciency to date to 12–16% [9,
10]. Multiple band gap devices can provide efﬁcient matching of the solar spectra [11–15]. A two or more band gap conﬁguration will lead, per unit surface area, to more efﬁcient solar energy conversion, and in the solid-state, multiple band gap solar cells have achieved more than 30% conversion efﬁciency of solar energy [14].
The fundamental beneﬁts of multiple band gap photoelectrochemistry have been recognized [16]. In this section an overview of the energetics of distinct multiple band gap photoelectrochemical solar cell (MPEC) conﬁgurations is introduced. The MPEC conﬁgurations can lead to higher conversion efﬁciency than previously observed for single band gap solar cells.
A limited fraction of incident solar photons have sufﬁcient (greater than band gap) energy to initiate charge excitation within a semiconductor. Because of the low fraction of short wavelength solar light, wide band gap solar cells generate a high photovoltage but have low photocurrent. Smaller band gap cells can use a larger fraction of the incident photons but generate lower photovoltage. As shown in Fig. 1, schematic, multiple band gap devices can overcome these limitations. In stacked multijunction systems, the topmost cell absorbs
(and converts) energetic photons but is transparent to lower energy photons. Subsequent layer(s) absorb the lower energy photons. Conversion efﬁciencies can be enhanced, and calculations predict that a
1.64-eV and 0.96-eV two-band gap system has an ideal efﬁciency of 38% and 50% for 1 and 1000 suns concentration, respectively.

4.4 Optimizing Photoelectrochemical Solar Energy Conversion:
In stacked multijunction solar cells, the top cell converts higher energy photons and transmits the remainder onto layers, each of smaller band gap cell than the layer above, for more effective utilization of the solar spectrum. Fig. 1

The ideal efﬁciency increases to a limit of
72% for a 36-band gap solar cell [11].
Present solid-state (photovoltaic) multijunction photovoltaics exist in one of two modes, either splitting (in which the solar spectrum is optically separated before incidence on the cells) or stacked devices. The latter mode has been developed as either monolithic (integrated) or mechanically (discrete cells layered with transparent adhesive) stacked cells. Most monolithic stacked multijunction photovoltaics under development use III–V semiconductors in at least one of the layers and take advantage of the variations in band gap and close lattice match achieved with other related III–V alloys. For example, GaInP has been used as a wide gap top cell, or GaSb and
GaInAsP as small gap lower cells, for
III–V top-layer multijunction cells; other cells use silicon or CIS cells as a lower layer [12–15].
The photopotential of a single pn junction is:
Vpn =

nkT e ln

1 + jph,pn jo,pn (1)

where Vpn is constrained by the photocurrent density through the junction, jph,pn . jo,pn is the saturation current (of a reversely polarized diode) and is described by the Shockley equation [17].
The photopotential of a single liquid
Schottky junction is given by:
VSch =

nkT e ln

1 + jph,Sch jo,Sch (2)

Eg1 >

Eg2 > Eg3

Cell 1 (Eg1)

Cell 2 (Eg2)

Cell 3 (Eg3)

Depending on the relative rates of charge transfer, jo,Sch may be constrained by either solid-state or electrochemical limitations, and is respectively termed the saturation current or the equilibrium exchange current [17]. A single representation of either a pn or Schottky junction is schematized in the upper center of
Fig. 2, by a junction generating a voltage V .
A variety of distinct MPEC conﬁgurations are possible, each with advantages and disadvantages [1]. The simplest MPEC conﬁgurations contain two adjacent band gaps. Adjacent band gap layers can be aligned in the cell in either a bipolar or a less conventional inverted manner. In either the bipolar or inverted cell conﬁguration, the PEC solid–electrode interface can consist of either an ohmic or a Schottky interface. The ohmic interface can consist of either direct (semiconductor–electrolyte)

359

Storage
BGIO
PEC
Regenerative
BGDO
PEC
Storage
BGDO
PEC

Bipolar gap direct (semiconductorsolution) ohmic contact
(BGDO PEC)

∆V

Regenerative
BGS
PEC
Storage
BGS
PEC

Bipolar gap
Schottky contact photoelectrochemical solar cell
(BGS PEC)

EFermi

Schottky sc Electrolyte
Eredox

Ohmic junction Regenerative
IGIO
PEC

Storage
IGIO
PEC

Inverted gap indirect (semiconductormetal-solution) ohmic contact
(IGIO PEC)

pn junction ∆V

Ohmic
Electrolyte
Eredox

Ohmic contacts Semiconductor(sc) /electroyte junctions

Regenerative
IGDO
PEC

∆V small

Regenerative
IGS
PEC

Storage
IGS
PEC

Inverted gap
Schottky contact photoelectrochemical solar cell
(IGS PEC)

Inverted pnnp 2 band gap structure :
V = ∆V wide, ∆V small

∆V wide

Storage
IGDO
PEC

Inverted gap direct (semiconductorsolution) ohmic contact
(IGDO PEC)

Inverted gap ohmic contact photoelectrochemical solar cell
(PEC)

Inverted gap solar cell

Common node

Fig. 2 Relation of the twelve representative MPEC conﬁgurations comprising regenerative ohmic cells: (1) bipolar gap direct ohmic regenerative,
(2) bipolar gap indirect ohmic regenerative, (3) inverted gap direct ohmic regenerative, (4) inverted gap indirect ohmic regenerative; storage ohmic cells:
(5) bipolar gap direct ohmic storage, (6) bipolar gap indirect ohmic storage, (7) inverted gap direct ohmic storage, (8) inverted gap indirect ohmic storage; regenerative Schottky cells: (9) bipolar gap Schottky regenerative, (10) inverted gap Schottky regenerative; and storage Schottky cells:
(11) bipolar gap Schottky storage, and the (12) inverted gap Schottky storage conﬁguration.

Regenerative
BGIO
PEC

Bipolar gap indirect (semiconductormetal-solution) ohmic contact
(BGIO PEC)

Bipolar gap ohmic contact photoelectrochemical solar cell
(PEC)

Bipolar pnpn 2 band gap structure:
V = ∆V wide + ∆V small

∆V wide

∆V small

Bipolar gap solar cell

Configurations of Multiple Band Gap Photoelectrochemical Solar Cells
Single band gap p-Schottky or pn junction
∆V
representation

360

4 Solar Energy Conversion without Dye Sensitization

4.4 Optimizing Photoelectrochemical Solar Energy Conversion:

or indirect (semiconductor–metal and/or electrocatalyst–electrolyte) interfaces. Finally, each of these modes can be conﬁgured as either regenerative or storage solar cells. As shown in Fig. 2, these modes provide twelve alternate conﬁgurations available for MPECs.
Compared with the inverted case, the bipolar arrangement provides a conceptually simpler PEC but generates a large open-circuit photopotential, Voc . As shown for a regenerative MPEC in either scheme on the left side of Fig. 2, the generated bipolar photovoltage, Vphoto , is the sum of the potentials of the individual band gap layers minus cathodic and anodic polarization losses in driving a regenerative redox couple: Vphoto = Vw + Vs − (ηcathode + ηanode )
(3)
The energy diagram of a bipolar band gap photocathodic electrochemical
Schottky conﬁguration is presented in
Fig. 3(a). The scheme comprises a twophoton/one-electron photoelectrochemical process (2hν → e− ), which may be generalized for an n band gap conﬁguration, to an n photon process (nhν → e− ).
Light shown incident from the left side of the conﬁguration ﬁrst enters the wide band gap layer(s) in which more energetic photons are absorbed; less energetic photons are transmitted through this upper layer and are absorbed by the small band gap layer. The resultant combined potential of the photodriven charge sustains reduction at the photocathode interface, and drives extractable work through the external load,
Rload . The wide ‘‘w’’ and small ‘‘s’’ band gap layers are denoted with respective valence and conduction bands, EV and EC , and band gap:
EGw = ECw − EVw ;

EGs = ECs − EVs
(4)

Wide band gap layer charge separation occurs across a pn junction space charge ﬁeld gradient, while charge separation in the small band gap is maintained with a ﬁeld formed by the Schottky semiconductor–electrolyte interface. In the bipolar
Schottky MPEC conﬁguration, generated charge ﬂows through all layers of the cell, providing the additional constraint: jph,pn = jph,Sch

(5)

In an alternate bipolar regenerative conﬁguration, the bipolar (or multiple) band gap conﬁgurations may contain consecutive space charge ﬁeld gradients generated only via solid-state phenomena. This is presented in Fig. 3(b) for the case of two consecutive bipolar pn junctions. The lowest semiconductor layer (the small band gap n-type layer in the ﬁgure) may remain in direct contact with the electrolyte, but the contact is ohmic and is not the source of the small band gap space charge ﬁeld.
A similar indirect ohmic contact bipolar regenerative conﬁguration may also be derived from this ﬁgure. In this case, the lowest semiconductor layer is restricted to electronic and not ionic contact with the electrolyte through use of an intermediate
(bridging) ohmic electrocatalytic surface layer. This can facilitate charge transfer to the solution phase redox couple, and prevent any chemical attack of the semiconductor. In the bipolar cases (including
Schottky, direct or indirect ohmic conﬁgurations), the photopower generated by a bipolar regenerative MPEC is given by the product [17]:
Pbipolar regenerative = jph [Vw + Vs
− (ηcathode + ηanode )]

(6)

Adjacent band gap layers in a multiple band gap conﬁguration can also be aligned

361

4 Solar Energy Conversion without Dye Sensitization
V = Vw + Vs − η cathode + η anode e− ECs
EFermi (ns)

e−

e−

EGs hν ECw
EFermi (nw = ps)
Vw
h ν EGw

EVs

e−
A → A+

A+ →

A ηcathode Vs

ηanode

Eredox

h+
EGw > hν > EGs⇒

Electrocatalyst anode

R load

Semiconductor

EFermi (pw)
+
h ν > EGw ⇒ h EVw

p n Wide gap

(a)

p-Schott. small gap

Ohmic junction Electrolyte

V = Vw + Vs − η cathode + η anode e− e− E
Cs
e−

EFermi(ns)
Vs

Ecw

hν

EFermi (nw=ps)
EGw > h ν > EGs ⇒ h+

Vw hν (b)

EVs

ηcathode

Eredox

e−
A → A+ ηanode EGw
Ohmic
junction

EFermi (pw) h ν > EGw ⇒ h+

EGs

A+ → A

Electrocatalyst anode

R load
Semiconductor or electrocatalyst cathode

362

Electrolyte

EVw

p n Wide gap

Ohmic junction p n Small gap

Fig. 3 Energy diagrams for multiple band gap photoelectrochemistry. Elements of bipolar or inverted band gap, Schottky, ohmic, regenerative, and storage conﬁgurations are illustrated.
(a): Bipolar band gap Schottky regenerative MPEC. (b): Bipolar band gap (direct or indirect) ohmic regenerative MPEC.(c): Inverted band gap indirect ohmic regenerative MPEC.(d): Inverted gap (direct or indirect) ohmic storage MPEC.

4.4 Optimizing Photoelectrochemical Solar Energy Conversion:

e−

EFermi (nw=ns)
Vw

E
Vs hν Gs

h ν EGw

EVs EFermi (ps)

EGw > h ν >EGs ⇒ h+

EFermi (pw)
+
h ν > EGw ⇒ h

ηas ηc

Eredox
A → A+

ηc ηaw

Eredox
A+ ← A

Electrocatalyst anode

e−

Electrocatalyst cathode

e−

ECw

Rload-s
Vs− η c+ ηas

Electrocatalyst anode

e−

ECs

Semiconductor/metal

Vw − ηc+ ηaw
Rload-w

A+ → A
Electrolyte

Ohmic junction EVw

Ohmic p n junction Small gap

n
Wide gap

p
(c)

Vw

Vs

hνE

hν > EGw ⇒ h+ EVw p n
(d)

Fig. 3

EVs EFermi (ps)

Gw

EFermi (pw)

hν EGs

EGw > h ν > EGs ⇒ h+

Ohmic junction n

B →B ηcB EredoxB ηaA EredoxA
A→

A+

Electrolyte

Ew = Vw
+

C →C ηcC EredoxC
D → D+ ηaD EredoxD

Semiconductor or electrocatalyst anode

EFermi (nw=ns)

+

e−

Electrocatalyst cathode

ECw

Es = V s

ECs

e−

Semiconductor or electrocatalyst anode

e−

Electrocatalyst cathode

e−

Electrolyte

p

Wide gap Ohmic Small gap junction (Continued)

in a less conventional inverted manner. As shown in Fig. 3(c), a dual gap inverted cell generates two smaller photopotentials, which can be separately applied to the same PEC, minimizing electrolysis losses. This ﬁgure presents an inverted pn/np (wide band gap or small band gap)

regenerative ohmic conﬁguration, compared to the pn/pn in Fig. 3(b). As seen for the inverted pn/np case, a quasi fermi level is shared by the conduction bands of the wide band gap n-type and the small band gap n-type layers. Light incident from the ﬁgure left side generates charge through

363

364

4 Solar Energy Conversion without Dye Sensitization

a common intermediate node, comprising a two photon/two electron photoelectrochemical process (2hν → 2e− ):
2h+ e− + hν(hν > EGw ) + hν(hν > EGs )
− − 2h+ + 2ew −∗ + 2es −∗
−→

(7)

Photoinduced holes, generated in accord with Eq. (7), drive solution phase oxidation at separate electrocatalytic anodes. A challenge to bipolar MPEC use is that all generated charge must ﬂow through subsequent cell layers, as seen in the diagrams in the left side of Fig. 3. This imposes a current matching constraint on each of the individual junctions of any monolithic bipolar device. This constraint is removed with inverted band gaps, in which photocurrent is independent for each band gap, ﬂowing through the common node illustrated in the ﬁgure. Unlike the bipolar cases, for the inverted regenerative
(Schottky, direct or indirect ohmic) conﬁgurations, the photopower generated is a separate sum of the different layers, where for the two band gap case [17]:
Pinverted regenerative = jph−w Vw + jph−s
× Vs − (ηcathode + ηanode )

(8)

Electrochemical energy storage conﬁgurations provide an energy reservoir that may compensate for the intermittent nature of terrestrial insolation. When different anodic and cathodic redox processes are driven, electrochemical energy storage can be accomplished. Storage can consist of both in situ or photoelectrolysis cells. Until recently, photoelectrolysis cells, in which the electrolyte solvent is oxidized and/or reduced to provide a chemical fuel, had been inefﬁcient [6]. In situ secondary redox couples, added to the electrolyte, have been simpler to optimize, and efﬁcient single band gap in situ PEC storage cells have been demonstrated [18].

Secondary redox storage must be accomplished at sufﬁciently low potentials to prevent losses due to simultaneous
(undesired) solvent electrolysis. However, bipolar band gap photoelectrochemistry imposes large photopotentials. These are avoided through the inverted band gap conﬁguration, as exempliﬁed in Fig. 3(d), in which the photopotentials generated in the respective small and wide band gap portions of the tandem cell, Vw and Vs drive two separate electrochemical storage processes: D + C+ − − D + + C ;
−→
EC+ /C − ED+ /D < Vw and
−→
A + B+ − − A+ + B;
EB+ /B − EA+ /A < Vs

(9)

Bipolar Band Gap PECs
A bipolar gap direct ohmic photoelectrochemical system comprises either a bipolar band gap pnpn/electrolyte ohmic photoelectrochemical cell, with reduction occurring at the photoelectrode–electrolyte interface and regenerative oxidation occurring at the electrolyte–counter electrode (anode) interface or alternately: a npnp/electrolyte bipolar band gap with oxidation occurring at the semiconductor–electrolyte interface and regenerative reduction occurring at the electrolyte–counter electrode interface. In the bipolar gap direct ohmic photoelectrochemical system, direct refers to the direct contact between semiconductor and solution, and ohmic indicates this interface is an ohmic rather than a Schottky junction.
This facilitates study of several characteristics of bipolar multiple band gap systems, without the added complication of simultaneous parameterization of a direct
Schottky barrier at the electrolyte interface.
4.4.2.2

4.4 Optimizing Photoelectrochemical Solar Energy Conversion:

The examples presented here utilize combine multijunction solid-state layers consisting of a bipolar AlGaAs (EGw =
1.6 eV) wide band gap, overlaid on a
Si (EGs = 1.0 eV) small band gap, and used in an electrolytic cell [2]. Light absorption by the electrolyte can interfere with the cell and should be avoided.
Figure 4 overlays the optical characteristics of the solid and solution phase of the AlGaAs/Si solid-state and V3+/2+ electrolyte optimized components within a bipolar gap photoelectrochemical cell. Solution transmission is measured through a pathlength that is typical (1 mm) of many experimental front wall photoelectrochemical cells. As is evident, light-transmission

interference will occur for the top AlGaAs layers, and bottom Si layers through this or substantially shorter electrolyte pathlengths. The solid-state component includes a graded band emitter, varying from
Al(0.3 – 0.15) Ga(0.7 – 0.85) As with overlayers of p + -Alx Gax −1 As on n-Alx Gax −1 As. The growth sequence and graded band emitter layer improve collection efﬁciency [2]. The
Si bottom cell consists of a p + -Si, n-Si, and n+ -Si multijunction. The band edges observed in the ﬁgure at approximately
800 nm and 1100 nm are consistent with the respective AlGaAs and Si band gaps.
For efﬁcient electron/hole pair charge generation, incident photons need to be localized within the multiple band gap
0

0.8

20

Q(AlGaAs layers)

Q(Si layers)

0.6

40

0.4

60

0.2

0.0
400

T(0.35 M V 2 + /3 +)
1 mm pathlength

500

600

Transmittance
[%]

Quantum efficiency

1.0

80

700

800

900

1000

1100

100
1200

Wavelength
[nm]
Overlay of the optical characteristics of the solid and solution phase of the AlGaAs/Si solid-state and V3+/2+ electrolyte constituents within a bipolar gap photoelectrochemical cell.
Transmission of the V3+/2+ electrolyte is

Fig. 4

measured through a pathlength of 1 mm. As described in the text, the Si bottom cell consists of a p+ -Si, n-Si and n+ -Si multijunction. The
Al(0.3 – 0.15) Ga(0.7 – 0.85) As top cell utilizes a graded band emitter.

365

4 Solar Energy Conversion without Dye Sensitization

GaAs layer. Internally, a bridging GaAs buffer layer provides an ohmic contact between the wide band gap AlGaAs junctions and the lower Si layers. An intermediate contact layer indicated as ‘‘Au’’ is used only for probing separated characteristics of the wide and small band gap junctions, and is not utilized in the complete cell. Photo generated charge at the indicated silicon electrolyte interface induces solution phase vanadium reduction, and a carbon counter electrode provides an effective (low polarization) electrocatalytic surface for the reverse process in a regenerative cell, in accord with:

semiconductor small and wide band gap regions, rather than lost through competitive electrolyte light absorption. As seen in Fig. 4, the vanadium electrolyte can signiﬁcantly block light, over a wide range of visible and near infrared wavelengths, from entering the wide and small band gap layers of the multiple band gap photoelectrochemical cell. This deleterious effect is prevented by use of the back wall multiple band gap photoelectrochemical cell presented in Fig. 5. Light does not pass through the solution. As shown, illumination enters directly through antireﬂection ﬁlms of 50 nm ZnS situated on 70 nm
MgF2 . An evaporated Au-Zn/Au grid provides electrical contact to the wide gap
AlGaAs layers through a bridging p + =

V3+ (+hν) − − V2+ + h+ ;
−→
V2+ − − V3+ + e−
−→

V 2+ → V 3+ + e−
Carbon

350 µm
800 nm
(8 × 1015 cm−3)

(4 × 1019 cm−3)

n - Si

E°,V 2+/3+
Electrolyte
− 0.3 V vs H2

+

n - Si

10 nm
1.0 µm
(1 × 1019 cm−3)

n - GaAs
GaAs (Buffer layer)

p+ - S i

1.7 nm

20 nm

(1 × 1018 cm−3)

n+Al(0.15)Ga(0.85)As

300 nm

1.0 µm
(2 × 1017 cm−3)

n - Al(0.15)Ga(0.85)As

50 nm

p+ - Al0.8Ga0.2As
(1 × 1018 cm−3) p+ - Al(0.3 − 0.15)Ga(0.7− 0.85)As (1 × 1018 cm−3)

V 3+ + e− → V 2+

Illumination ⇒

p+ - GaAs

Au-Zn/Au

Au

e−

AR coating [ZnS(50 nm)/MgF2(70 nm)]

366

Schematic description of the components in the bipolar gap direct ohmic AlGaAs/Si-V3+/2+ PECs.

Fig. 5

(10)

4.4 Optimizing Photoelectrochemical Solar Energy Conversion:

Heller, Miller and coworkers have previously shown that the p -Si surface is capable of sustaining minority carrier injection into the solution and stabilizes reduction of the V3+/2+ redox couple [19]. However, at n-Si this does not appear to be the case for reduction of the V3+/2+ redox couple by majority carrier injection into the solution.
Thermodynamically silicon can be spontaneously oxidized in aqueous solution. This process may occur despite the photoinduced generation of reductive charge at the semiconductor–electrolyte interface.
This is observed as the time-dependent decrease of photocurrent under constant illumination, during photoinduced reduction of V2+ (by majority carrier injection at n-Si) indicated in the inset Fig. 6. This is consistent with the onset of surface passivation forming a layer that is passive to charge transfer and diminishes the photocurrent. The addition of a low

concentration of HF to the electrolyte can remove this passivating layer and permit sustained photocurrent. This is illustrated in the inset of Fig. 6, in which the addition of 0.02 and 0.2 M HF improves the photocurrent stability of this bipolar direct ohmic MPEC. Also as shown, no further improvement in the photocurrent stability is observed when the HF concentration is further increased from 0.2 to 0.5 M HF. As presented in the main portion of Fig. 6, the
0.2 M HF modiﬁed electrolyte stabilizes the photocurrent at the n-Si interface for the measured period of several hours [2].
As will be subsequently shown, photocurrent stability is further improved through use of a electrode catalyst bridging the silicon–electrolyte interface.
As can be seen in Eq. (6), maximization of the photopower necessitates minimization of the anodic and cathodic polarization losses, ηanode and

12

Bipolar gap ohmic photoelectrochemical cell
AlGaAs/Si - V2+/3+ Electrolyte solar cell
Illuminated area : 0.22 cm2
Illumination : simulated AM 1.5
Cathode : 0.2 cm2 n+ - Si 14
Anode : 0.2 cm2C

Electrolyte :
0.35 M V(II)+V(III), 4 M HCl, 0.2 M HF

12

10

Photocurrent
−2
[mA cm ]

Photocurrent density
[mA cm−2]

14

8

8

No HF
0.02 MHF
0.2 MHF
0.5 MHF

10

Photocurrent stability, n-Si photoanode in direct contact with solution
0

0

1

10

2

20

30
Time
[min]

3

40

4

50

60

5

Time
[h]
Photocurrent stability in several V3+/2+ aqueous electrolytes of the bipolar band gap direct ohmic AlGaAs/Si-V3+/2+ photoelectrochemical cell (measured indoors using a tungsten halogen lamp to simulate outdoor AM 1.5 insolation).

Fig. 6

367

4 Solar Energy Conversion without Dye Sensitization

ηcathode , during charge transfer through the photoelectrode and counter electrode interfaces. In the current domain investigated, polarization losses are highly linear for both anodic and cathodic processes at
2.5 to 3.5 mV cm2 mA−1 , and can create small but signiﬁcant losses on the order of
10–100 millivolts in the MPEC.
Figure 7 presents the outdoor characteristics of the bipolar direct ohmic
AlGaAs/Si/V2+/3+ photoelectrochemical cell under solar illumination. The system comprises the individual components illustrated in Fig. 5 and uses a HF containing aqueous vanadium electrolyte to improve photocurrent stability (0.35 M
V(II)+V(III), 4 M HCl, 0.2 M HF). The photoelectrochemical characteristics of the cell were determined under 75mW cm−2 insolation. As shown under

illumination, the AlGaAs/Si/V2+/3+ electrolyte photoelectrochemical solar cell exhibits an open-circuit potential, Voc =
1.4V, a short-circuit photocurrent, Jsc =
12.7 mA cm−2 , a ﬁll factor, FF = 0.81, determined from the fraction of the maximum power, Pmax , compared to the product of the open-circuit potential and short-circuit current.
The multiple band gap solar to electrical conversion efﬁciency of 19.2% compares favorably to the maximum 15 to 16% solar to electrical energy conversion efﬁciency previously reported for single band gap
PECs [9, 10]. Small photoelectrochemical efﬁciency losses can be attributed to polarization losses accumulating at the solution interfaces. Under illumination, a photocurrent density of 13 mA cm−2 seen in Fig. 7 is consistent with polarization

15

15
Bipolar gap direct ohmic photoelectrochemistry
AlGaAs/Si - V2+/3+ electroyte solar cell

FF = 0.81
12

9

Current
Power

3

Indirect ohmic AlGaAs/Si V2+/3+ cell
Insolation: 80 mW cm−2
Illuminated area :
0.22 cm2
Anode + Cathode :
0.2 cm2 C
FF = 0.79
Solar conversion efficiency : 19.2%
Electrolyte : Aq.0.35 M V(II) + V(III), 4 M HCl
0

0
0.0

0.2

0.4

Photovoltage
[V]

0.0

0.6

0.8

1.0

1.2

1.2

Photovoltage
[V]
Fig. 7 Measured outdoor photocurrent/voltage characteristics of the bipolar gap direct ohmic AlGaAs/Si-V3+/2+ photoelectrochemical solar cell.
Inset: Measured outdoor photocurrent/voltage characteristics of the bipolar gap indirect ohmic AlGaAs/Si-V3+/2+ photoelectrochemical solar cell.

Photopowder density

15

15

6

Photopower density
[mW cm−2]

9

Electrolyte: Aq.0.35 M V(II)+V(III), 4 M HCl, 0.2M HF
Electrolyte cathode: 0.2 cm2 n+ - Si
Electrolyte anode: 0.2 cm2 C
Insolation: 75 mWcm−2
Illuminated area: 0.22 cm2
Solar electrical conversion
Efficiency: 19.1%
Photocurrent density

12

Photocurrent density
[mA cm−2]

368

6

3

0

0

1.4

4.4 Optimizing Photoelectrochemical Solar Energy Conversion:

losses of approximately 0.04 V, accumulating both at the anode and the cathode.
At the observed maximum power point photovoltage, in excess of 1.2 V in Fig. 7, this combined 0.08 V loss is consistent with an effective loss in efﬁciency of 0.3 to
0.4% compared to an analogous solid-state photovoltaic cell conﬁgurations.
A common disadvantage of photoelectrochemical systems is photoinduced corrosion of the semiconductor that originates at the semiconductor–solution interface.
The corroded surface inhibits charge transfer that diminishes photocurrent. A stable solid–solution interface, which both facilitates charge transfer and impedes semiconductor photocorrosion, is provided by an electrocatalyst placed between the semiconductor and the electrolyte. The multiple band gap photoelectrochemical cell can utilize this electrocatalyst interface, as well as a bipolar series arrangement of wide and small band gap semiconductors to enhance energy conversion. The photoelectrochemical characterization presented in the inset of Fig. 7 summarizes a modiﬁed
GaAs/Si-V3+/2+ MPEC. In this indirect ohmic photoelectrochemistry, electrolyte induced photocorrosion of the silicon is entirely inhibited by utilization of an electrocatalyst (a second carbon electrode) bridging charge transfer between the semiconductor and the electrolyte.
Bipolar gap indirect ohmic photoelectrochemistry comprises either: a bipolar band gap pnpn/electrolyte ohmic photoelectrochemical cell, with reduction occurring at the semiconductor–electrocatalyst–electrolyte interface and regenerative oxidation occurring at the electrolyte–counter electrode (anode) interface or alternately: an npnp/electrolyte cell, with oxidation occurring at the semiconductor–electrocatalyst–electrolyte interface and regenerative reduction occurring at

the electrolyte–counter electrode interface. In these systems, indirect refers to the catalyst interface that bridges the semiconductor and solution, and ohmic indicates this interface is an ohmic rather than a Schottky junction. The photocathodic bipolar direct ohmic MPEC photoelectrochemistry comprises a twophoton/one-electron photoelectrochemical process (2hν → e− ) injecting charge into an electrocatalyst and then into solution. Energy conversion can occur either through Fig. 8’s photocathodic driven redox process: hν pn(wide gap)|pn (small gap)| electrocatalyst|redox couple| electrocatalyst anode
(11)
or a photoanodic driven redox couple: hν np(wide gap)|np (small gap)| electrocatalyst|redox couple| electrocatalyst cathode
(12)
Two highly stable aqueous phase redox couples, iodide and polysulﬁde, are used in these studies to further this bipolar indirect multiple band gap photoelectrochemistry. Platinum and cobalt sulﬁde, respectively, provide an effective (low overpotential) electrocatalyst for a wide range of iodide and polysulﬁde electrolytes [2].
These electrolytes have also been utilized in bipolar band gap solar cells. Table 1 summarizes the results of a variety of bipolar two band gap solar cells [2, 3].
In this table, comparison of the solidstate and direct photoelectrochemical cells shows that energy conversion efﬁciency of the photoelectrochemical cell approach that of the solid-state device. In the bipolar direct MPEC cell, the majority of the photopower loss in these photoelectrochemical cells was attributed to polarization

369

4 Solar Energy Conversion without Dye Sensitization

20

10

12

9

6

Photocurrent density

8

Current
Power

5

0

3.2

FF = 0.74
4
Bottom cell characteristics Photovoltage

0

0.2

0

0.0

0.0

0.0

3

1.6

Photopower density
[mW cm−2]

15

Inverted gap direct ohmic photoelectrochemistry
GaAs/Si-lodide electrolyte solar cell
Electrolyte: Aqueous 10.4 M Hl, 0.01M I2, 0.8 M HF
Electrolyte electrodes:Top cell : 0.2 cm2 Pt
FF = 0.73
Bottom cell anode: 0.2 cm2 p+- Si
Bottom cell cathode: 0.2 cm2 Pt
Top cell
Insolation: 80 mW cm−2 characteristics Illuminated area: 0.22 cm2
Solar/Electrical conversion
Efficiency: 19.0%
Photopower density

Photocurrent density
[mA cm−2]

370

0.6

0.4

0.6

0.8

Photovoltage
[V]
Measured outdoor characteristics of the inverted band gap direct ohmic GaAs/Si/I3 − ,
3/2I− /Pt MPEC. The top cell consists of the
GaAs/Pti/I3 − , 3/2I− /Pt portion of the cell. The lower cell consists of the Si/I3 − , 3/2I− /Pt
Fig. 8

portion of the cell. Main ﬁgure: top layer cell photocurrent/voltage characteristics. Inset: bottom layer cell photocurrent/voltage characteristics. Tab. 1 Comparison of bipolar solid-state or bipolar regenerative ohmic PECs under solar illumination Cell conﬁguration Bipolar solid state
AlGaAs/Si
Bipolar direct V2+/3+
AlGaAs/Si/V2+ /V3+ /C
Bipolar indirect V2+/3+
AlGaAs/Si/C/V2+ /V3+ /C
Bipolar indirect sulﬁde
AlGaAs/Si/CoS/S2 2− /S4 2− /CoS
Bipolar indirect iodide
AlGaAs/Si/Pt/I3 − /I− /Pt

Voc
[V]

Isc
Pmax
[mA cm−2 ] [mW cm−2 ]

FF

Insolation
Conversion
[mW cm−2 ] [Efﬁciency %]

1.409

15.6

17.7

0.81

90.2

19.6

1.393

12.7

14.4

0.81

75.0

19.1

1.383

14.0

15.3

0.79

80.1

19.2

1.456

15.8

18.5

0.80

94.1

19.7

1.409

15.9

18.1

0.81

94.1

19.2

Note: Cells utilize multijunction wide band gap AlGaAs layers over smaller band gap Si layers and one of the indicated pairs of electrolyte/electrocatalyst electrodes indicated as Direct V2+/3+ : 0.35 M
V(II)+V(III), 4 M HCl, 0.2 M HF at carbon; Indirect V2+/3+ : 0.35 M V(II)+V(III), 4 M HCl at carbon;
Indirect Sulﬁde 1 M K2 S2 , 1 M KOH at CoS; Indirect Iodide: 10.4 M HI, 0.01, 4 M I2 at Pt.
PEC, photoelectrochemical solar cells.

4.4 Optimizing Photoelectrochemical Solar Energy Conversion:

losses arising at the solid–solution interfaces. In certain indirect MPEC cases, as seen in Table 1, the photoelectrochemical energy conversion efﬁciency is comparable to the solid-state cell process. In this case, combined electrolytic polarization losses are less than or equal to the potential drop from resistance losses in the back contact of the solid-state device.
Inverted Band Gap PECs
Inverted photoelectrochemistry permits efﬁcient energy, and maintained smaller photopotentials than in comparable bipolar MPECs. To probe inverted semiconductor direct ohmic regenerative electrochemistry, an electrolyte such as iodide was driven by a GaAs (EGw = 1.4 eV) wide band gap, and an Si (EGs = 1.0 eV) small band gap were utilized [3]. In the inverted conﬁguration, bottom and top cells are utilized simultaneously. Simultaneous to the photopower generated by the bottom portion (Si) of the MPEC, the top (GaAs) portion also generates photopower. A platinum counter anode, provides an effective
(stable, low polarization) electrocatalytic surface for the reverse process in both the top (wide band gap) and bottom (small band gap) driven regenerative cells. At the n-Si interface, photocurrent stability can be improved with the addition of HF to an acidic iodide electrolyte, and the electrolyte used was 10.4 M HI, 0.01 M I2 ,
0.8 M HF. Previous studies have shown that other electrolytes and surface modiﬁcations will also enhance photocurrent stability through the (single band gap) silicon–electrolyte interface [19–21]. The inset of Fig. 8 presents the inverted direct ohmic outdoor characteristics of the
(Si/I3 − /3/2I− ) bottom cell portion, and the main portion of the ﬁgure presents the characteristics of the GaAs driven
4.4.2.3

I3 − /3/2I− portion of the cell under solar illumination. Photoelectrochemical characteristics were determined under 80 mW cm−2 insolation and the generated photocurrent is highly stable. As shown under illumination, the Si bottom cell exhibits an opencircuit potential, Voc = 0.51V, a shortcircuit photocurrent, Jsc = 7.4 mA cm−2 a ﬁll factor, FF = 0.73, and a maximum power, Pmax = 2.8 mW cm−2 . Simultaneous to the photopower generated by the bottom portion of the MPEC, the top
(GaAs) portion also generates photopower.
The main portion of Fig. 8 presents the outdoor characteristics of the top cell
(GaAs/Pt/I3 − , 3/2I− /Pt) portion during solar illumination of the complete inverted direct ohmic GaAs/Si/I3 − , 3/2I− /Pt
MPEC. In this portion of the cell, GaAs is isolated from the iodide electrolyte via a Pt electrocatalytic anode. The Pt electrode is stable in this electrolyte, and hence the top cell portion of the photocurrent appears to be fully stable. Under the same 80 mW cm−2 insolation, photopower is generated simultaneous to the photopower presented in Fig. 8.
As shown under illumination, the GaAs cell exhibits an open-circuit potential,
Voc = 0.81 V, a short-circuit photocurrent,
Jsc = 21.0 mA cm−2 , a ﬁll factor, and a maximum power, Pmax = 12.4 mW cm−2 at FF = 0.74. The total power generated is the sum of the simultaneous extractable power generated by each component.
Table 2 compares inverted solid-state photovoltaics with direct and indirect regenerative ohmic PECs, each containing a GaAs wide band gap and an inverted aligned Si small band gap, and a variety of redox couples. As in the case of the bipolar systems summarized in Table 1, the solidstate device exhibits a marginally enhanced energy conversion efﬁciency compared to

371

372

4 Solar Energy Conversion without Dye Sensitization
Tab. 2 Comparison of inverted solid-state or inverted regenerative ohmic PECs under solar illumination Cell conﬁguration Inverted solid-state bottom: Si top: GaAs
GaAs/Si
Inverted direct iodide
Bottom: Si/I3 − /I− /Pt
Top: GaAs/Pt/I3 − /I− /Pt
GaAs/Si/I3 − /I− /Pt
Inverted indirect iodide
Bottom: Si/Pt/I3 − /I− /Pt
Top: GaAs/Pt/I3 − /I− /Pt
GaAs/Si/Pt/I3 − /I− /Pt
Inverted indirect sulﬁde
Bottom: Si/CoS/S2 2− /S4 2− /CoS
Top: GaAs/CoS/S2 2− /S4 2− /CoS
GaAs/Si/CoS/S2 2− /S4 2− /CoS
Inverted indirect V2+/3+
Bottom: Si/C/V2+ /V3+ /C
Top: GaAs/C/V2+ /V3+ /C
GaAs/Si/C/V2+ /V3+ /C

Voc
Isc
Pmax
FF Insolation Conversion
[V] [mA cm−2 ] [mW cm−2 ]
[mW cm−2 ] [Efﬁciency %]

0.552
0.855

0.510
0.813

0.541
0.866

0.556
0.855

0.523
0.823

22.8
8.7

7.4
21.0

9.2
24.5

8.9
24.0

7.7
20.8

3.6
14.2
17.7

0.74
0.74

2.8
12.4
15.2

0.74
0.73

3.7
14.9
18.5

0.74
0.70

3.8
14.9
18.7

0.76
0.73

3.0
12.3
15.3

0.73
0.72

90.2

19.6

80.1

19.0

94.1

19.7

94.1

19.8

80.1

19.2

Note: Cells utilize multijunction wide band gap GaAs layers over smaller band gap Si layers and one of the indicated pairs of electrolyte or electrocatalyst electrodes, indicated as Direct Iodide: 10.4 M
HI, 0.01, 4 M I2 , 0.8 M HF at Pt; Indirect Iodide: 10.4 M HI, 0.01, 4 M I2 at Pt; Indirect Sulﬁde 1 M
K2 S2 , 1 M KOH at CoS; Indirect V2+/3+ : 0.35 M V(II)+V(III), 4 M HCl at carbon.

the analogous direct regenerative MPEC.
This difference may be attributed to diminished charge transfer through the semiconductor–electrolyte interface. However these differences are small and are only slightly larger than those caused by the uncertainty of ±1 mW cm−2 in the measured insolation level. Each of the sulﬁde or iodide indirect regenerative MPECs probed in this study exhibit conversion efﬁciency comparable to, or better than, the solid-state device in Table 2. This indicates that electrolytic polarization losses in these cells are comparable to resistance losses of the semiconductor back contact in the solid-state MPEC. The marginally lower conversion efﬁciency of

the V2+/3+ indirect regenerative MPEC in this table is consistent with the higher polarization losses of 3 mV cm2 mA−1 that we have measured for V2+/3+ at carbon, compared to those of less than
2 mV cm2 mA−1 previously measured at
Pt and CoS, respectively, for the iodide and sulﬁde MPECs [2, 3].
Bipolar Band Gap Solar Storage
Cells
One limitation to solar cells is that, while most of our electrical needs are continuous, clouds and darkness dictate that solar energy is intermittent in nature.
PECs can generate not only electrical but also electrochemical energy, and provide
4.4.2.4

4.4 Optimizing Photoelectrochemical Solar Energy Conversion:

a single device to solve the problem of the intermittent nature of solar energy.
In 1987 we presented a highly efﬁcient single band gap photoelectrochemical cell combining in situ electrochemical storage and solar conversion capabilities, providing continuous output insensitive to daily variations in illumination. The 1987 conﬁguration was demonstrated with a n-Cd(Se,Te)/polysulﬁde electrolyte conversion half cell and a Sn/SnS storage system, resulting in a single cell operating continuously at an overall efﬁciency of 11% [18].
Under illumination, photocurrent droves an external load. Simultaneously, a portion of the photocurrent was used in the direct electrochemical reduction to a metal (Sn) in the device storage half cell. In darkness or below a certain level of light, the storage compartment delivers power by metal oxidation. Recently, high solar conversion and storage efﬁciencies have been attained with a system that combines efﬁcient multiple band gap semiconductors, with a simultaneous high-capacity electrochemical storage [4, 5]. The energy diagram for one of several multiple band gap cells is presented in Fig. 3(d), and as described in
Fig. 2, several other conﬁgurations are also feasible. In the ﬁgure, storage occurs at a potential of Eredox = EA+/A − EB/B+ . On illumination, two photons generate each electron, a fraction of which drives a load, while the remainder (1/x e− ) charges the storage redox couple. Without light, the potential falls below Eredox and the storage couple spontaneously discharges. This dark discharge is directed through the load, rather than through the multijunction semiconductor’s high dark resistance.
In Fig. 9 is presented an operational form of the solar conversion and storage cell described by the Fig. 2 energy diagram.
The single cell contains both multiple

band gap and electrochemical storage, which unlike conventional photovoltaics, provides a nearly constant energetic output in illuminated or dark conditions. The cell combines bipolar AlGaAs (Eg = 1.6 eV) and Si (Eg = 1.0 eV) and AB5 metal hydride/NiOOH storage. The NiOOH/MH metal hydride storage process is near ideal for the AlGaAs/Si because of the excellent match of the storage and photocharging potentials. The electrochemical storage processes utilizes MH oxidation and nickel oxyhydroxide reduction [22]:
−→
MH + OH− − − M + H2 O + e− ;
EM/MH = −0.8 V versus SHE

(13)

−→
NiOOH + H2 O + e− − −
Ni(OH)2 + OH− ;
ENiOOH/Ni(OH)2 = 0.4 V vs SHE (14)
As shown in Fig. 10, the cell generates a light variation insensitive potential of 1.2–1.3 V at total (including storage losses) solar/electrical energy conversion efﬁciency of 18%. Over an eight-month period of daily cycles, under constant 12-hour
(AM0) illumination, the long-term indoor cycling cell generated a nearly constant photocurrent density of 21.2 (constant to within one percent or ±0.2 mA cm−2 ), and a photopower that varied by ±3% [5]. The cell is a single physical/chemical device that generates load current without any external switching.
Bipolar Band Gap Solar Hydrolysis
(hydrogen generation) Cells
Solar energy–driven water splitting combines several attractive features for energy utilization. Both the energy source (sun) and the reactive media (water) are readily available and are renewable, and the resultant fuel (generated hydrogen) and the emission with fuel consumption (water)
4.4.2.5

373

18
−3
(1 × 10 cm ) 50 nm

−3

are each environmentally benign. The UV and visible energy rich portion of the solar spectrum is transmitted through
H2 O. Therefore sensitization, such as
Metal hydride anode

−3

(4 × 10 cm ) 800 nm
Au-Sb/Au

19

(8 × 1015 cm−3) 350 µm

n -Si n+-Si (1 × 1019 cm−3) 1.0 µm

10 nm

20 nm

(1 × 10 cm ) 1.7 nm

18

17
−3
(2 × 10 cm ) 1.0 µm

p+-Si

GaAs(Buffer layer)

n -GaAS

n -Al(0.15)Ga(0.85)As

+

n -Al(0.15)Ga(0.85)As

p+-Al(0.3 – 0.15)Ga(0.7– 0.85)As (1 × 1018 cm−3) 300 nm

p+-Al0.8Ga0.2As

p+-GaAs

−

18
−3
(1 × 10 cm ) 50 nm

18

−3

10 nm

20 nm

(1 × 10 cm−3 ) 1.7 nm

17

(2 × 10 cm ) 1.0 µm

(1 × 1019 cm−3 ) 1.0 µm

Separator with KOH electrolyte

MHx−1 + H2O + e− → MHx + OH−

Metal hydride anode

(4 × 1019 cm−3 ) 800 nm
Au-Sb/Au

n-Si n+-Si (8 × 1015 cm−3 ) 350 µm

p+-Si

GaAs(Buffer layer)

n-GaAS

n Al(0.15)Ga(0.85)As

+

n-Al(0.15)Ga(0.85)As

p+-Al(0.3 – 0.15)Ga(0.7–0.85)As (1 × 1018 cm−3 ) 300 nm

p+-Al0.8Ga0.2As

Nickel cathode

Photo current Storage current Load current ←1 − 1/xe−

Nickel cathode

−

Separator with KOH electrolyte
NiOOH + H2O + e → Ni(OH)2 + OH

+

Au-Zn/Au
AR coating [ZnS(50 nm)/MgF 2(70 nm)] p -GaAs

−

Reverse resistance prevents significant back current
−

Au-Zn/Au

⇓

e−→

MHx + OH− → MHx−1 + H2O + e−

Fig. 9
AR coating [ZnS(50 nm)/MgF2(70 nm)]

Illumination

2hν

Rload

Ni(OH)2 + OH → NiOOH + H2O + e

In dark

374

4 Solar Energy Conversion without Dye Sensitization
←1/xe−

Day-time energy conversion of the
Storage/bipolar band gap solar cell.

−

e− →

←e

Rload

Night-time energy conversion of the
Storage/bipolar band gap solar cell.

The bipolar AlGaAs/Si/MH/NiOOH MBPEC solar cell.

via semiconductors, is required to drive the water-splitting process. In a solar photoelectrolysis system, the redox active interfaces can be in indirect or direct

4.4 Optimizing Photoelectrochemical Solar Energy Conversion:
28

Photocurrent density
[mA cm−2]

20

IV curve at full charge
Solar to electrical conversion efficiency: 19.6%
Illumination: Simulated AMO (135.5 mW cm2), Illuminated area 0.22 cm2

16

1.0

Generated cell potential is insensitive to variation in illumination

0.5

Illumination

12

I
[mA]

0

Dark

Illumination

12
IDischarge

IDark

-2

4

Dark

IDischarge

2

8

8

IDark

4

ICharge
0

6

12

18

24
30
Time
[hours]

0
0.2

20
16

IPhoto

4

0.0

24

1.5

Load potential
[V]

24

Bipolar gap in situ storage solar cell
AlGaAs/Si-MH/Ni cell

0.4

0.6

Photopower density
[mA cm−2]

28

0.8

1.0

36

42

48

0
1.2

1.4

Photovoltage
[V]
Fig. 10 Photoconversion current and power characteristics of the AlGaAs/Si/MH/NiOOH solar cell under fully charged AM0 conditions.
Inset: Two days conversion and storage characteristics of the AlGaAs/Si/MH/NiOOH

solar cell under simulated AM0 insolation. In the ﬁrst 12 hours of each of the two 24-hour cycles, the AM0 illumination is varied from 0 to AM0 back to 0, using a graded diffuse ﬁlter; the second 12 hours are without illumination.

contact with the photosensitizer, and comprise either an ohmic or Schottky junction.
Independent of this interface composition, the various parameters in models predicting solar water–splitting conversion efﬁciency, may be combined into two general parameters: (1) related to losses in optical energy conversion, ηphoto , or (2) related to losses in redox conversion of water to H2 and O2 , ηelectrolysis . Combined, these yield an overall solar electrolysis efﬁciency (excluding storage and utilization losses) as:

The subsequent challenge is to optimize sustained water electrolysis, without considerable additional energy losses. Effective water electrolysis must occur at a potential, VH2 O , near the photocell point of maximum power. VH2 O is greater than o EH2 O (=− GH2 O /nF) as a result of overpotential losses, ζ , in driving an electrolysis current density, j , through both the O2 and the H2 electrodes:

ηphotoelectrolysis = ηphoto × ηelectrolysis
(15)
The thermodynamic potential, E o H 2O , for the water splitting reaction is given by:
−→
H2 O − − H2 + 1/2O2 ;
◦

o o EH2 O = EO2 − EH2 ;
◦

o
EH2 O (25 C) = 1.229 V

(16)

VH2 O (j ) = EO2 (j ) − EH2 (j ) o o
= EO2 + ζO2 (j ) − (EH2 + ζH2 (j ))
(17)
A cell containing illuminated AlGaAs/Si
RuO2 /Ptblack is demonstrated to evolve H2 and O2 at record solar-driven water electrolysis efﬁciency [6]. Under illumination, bipolar conﬁgured Al0.15 Ga0.85 As (Eg =
1.6 eV) and Si (Eg = 1.1 eV) semiconductors generate open-circuit and maximum

375

4 Solar Energy Conversion without Dye Sensitization

cell are available [6]. Under illumination, this bipolar cell generates an open-circuit potential of 1.57 V, which is considerably larger the thermodynamic potential for the water-splitting reaction. Also included are the photopotential dependence of solar to electrical conversion efﬁciency. A portion of the high ηphoto domain lies at o a potential above EH2 O , and in principle is sufﬁciently energetic to drive efﬁcient photoelectrolysis. Planar platinum and Ptblack are effective
H2 electrocatalysts. At low current densities (0.5 mA cm−2 ), the observed Pt overpotential is low (ζH2 = −17 mV), and even smaller (ζH2 = −13 mV) with the Ptblack electrocatalyst, and smaller yet (ζH2 =

power photopotentials of 1.30 and 1.57 V, well suited to the water electrolysis thero modynamic potential of EH2 O (25 ◦ C) = o 1.229 V. The EH2 O /photopotential-matched semiconductors are combined with effective water electrolysis O2 or H2 electrocatalysts, RuO2 or Ptblack . The resultant solar photoelectrolysis cell drives sustained water splitting at 18.3% conversion efﬁciencies. Alternate dual band gap systems are calculated to be capable of attaining more than 30% solar photoelectrolysis conversion efﬁciency.
Figure 11 presents the measured current/voltage characteristics for the AlGaAs/Si photocell at AM0 illumination.
Further details and characterization of the

Jphoto

24

Area: 0.22 sq.cm
AM0 (135 mW/sq.cm) illumination
Open-circuit potential: 1.57 V
Short-circuit current: 5.2 mA
Fill factor: 0.772

16

Jphoto

E oH2O,the minimum water electrolysis potential AlGaAs/Si cell characteristics

20

[mA sq.cm]

376

ηphoto

12

8
Photoelectrolysis conversion efficiency upper limit
= η photo × E ° 2O/Vphoto
H

4

0

0

200

400

600

800

1000

1200

1400

1600

Vphoto
[millivolt]
AlGaAs/Si I–V characteristics at AM0 light intensity (135 mW cm−2 ). ηphoto = 100% × (Jphoto ∗ Vphoto )/Pillumination . Using the indicated, measured ηphoto and Vphoto , the upper limits of photoelectrolysis efﬁciency are calculated as ηphoto × 1.229 V/Vphoto .
Fig. 11

4.4 Optimizing Photoelectrochemical Solar Energy Conversion:

−7 mV) with a low level of convection
(stirring) to improve mass transport and prevent observed gas buildup on the electrode surface. Minimization of ζO2 is a greater challenge. In this case Pt is a poor electrocatalyst. ζO2 may be decreased
∼500 mV, utilizing an RuO2 electrode.
The kinetics of this effective oxygen electrocatalyst [23] have been attributed to catalysis by the intervening RuO4 /RuO2 redox couple [24, 25]. In the absence of competing redox couples, the faradaic efﬁciency of H2 and O2 evolution approaches
100%, and ηelectrolysis is determined by the current limited VH2 O (j ): o ηelectrolysis = EH2 O /VH2 O (j );
◦

ηelectrolysis (25 C) = 1.229V/VH2 O (j )
(18)
The Fig. 12 inset contains the Eq. (18)–determined ηelectrolysis . The limiting maximum ηphotoelectrolysis can be readily determined from the solar to electrical conversion efﬁciency. Expanding Eq. (15) with Eq. (18), ηphotoelectrolysis is diminished from ηphoto by the potential of the stored energy compared to the potential at which the water electrolysis occurred: ηphotoelectrolysis (T)
=

ηphoto (T) × EH2 O (T)
EO2 (T) − EH2 (T)

;

ηphotoelectrolysis (25 ◦ C) = 1.229V × ηphoto
VH2 O

(19) includes these limiting ηphotoelectrolysis values because the AlGaAs/Si system is determined from measured values of ηphoto at various values of
Vphoto . AlGaAs/Si solar photoelectrolysis at conversion efﬁciencies exceeding 20%
(Fig. 11) are in principle possible at potentials approaching the EH2 O limit. The main
Figure 12

portion of Fig. 12 presents the water electrolysis potential, VH2 O , determined under stirred or quiescent conditions, and measured in 1 M HClO4 using equal areas of the optimized RuO2 and Ptblack electrodes.
These measurements of the electrolysis current, as a function of potential, enable us to predict the AlGaAs/Si photocurrent at which photoelectrolysis will occur.
Our photoelectrolysis cell consists of illuminated AlGaAs/Si RuO2 /Ptblack electrolysis. With these active electrocatalysts, high values of ηelectrolysis are insured by using large surface areas of the electrolysis electrodes, compared to the illuminated area. This is accomplished without increasing the illuminated electrode area, as schematically represented in the lower portion of Fig. 13, by utilizing a large vertical depth of electrolysis electrodes compared to the cross section of illumination. Speciﬁcally, the 10 cm2 Ptblack and RuO2 electrodes utilized are large compared to the 0.22 cm2 illuminated area. In the Fig. 13(a) is the H2 and O2 electrolysis current generated by this cell as a function of time. The average photocurrent of 4.42 mA (generating a current density of 20.1 mA cm−2 at the illuminated electrode area, and 0.44 mA cm−2 at the electrolysis electrodes) corresponds to a photopotential of 1.36 V in Fig. 11 comparable to the equivalent electrolysis potential at 0.5 mA cm−2 in Fig. 12. The overall efﬁciency is determined by the 1.229 V energy stored as H2 and O2 and the incident photopower (135 mW cm−2 ): ηphotoelectrolysis = 100% × 20.1 mA cm−2
× 1.229 V/135 mW cm−2 = 18.3% (20)
Higher Solar Production Rates of
Hydrogen Fuel are Attainable
In Eq. (15), the signiﬁcance of the electrolysis compared to photo components
4.4.2.6

377

4 Solar Energy Conversion without Dye Sensitization

100

20

90 ηelectrolysis [%]

80
16
70
60
12

50
1200

1600
2000
Velectrolysis
[mv]

2400
8

Jelectrolysis
[mA sq.cm]

378

4
Electrolysis in 1 M HClO4
Anode: Pt black on Pt mesh
Cathode: 2.3 µm RuO2
1200

1300

Stirred
Quiescent
1400

1500

0
1600

Velectrolysis
[millivolt]
Measured variation of the VH2 O with current density in 1 M HClO4 , using equal area 2.3 µm RuO2 and Ptblack . Inset: Calculated ηphoto , in a 100% faradaic-efﬁcient process, as a function of VH2 O .

Fig. 12

of the conversion efﬁciency is evident on analysis of a report of a 12.4% photoelectrolysis cell. The cell contained a Pt coated GaInP2 hydrogen electrode in bipolar contact with a GaAs junction driving an oxygen electrode [26]. At their speciﬁed
11 sun–illuminated electrolysis, current density of 120 mA cm−2 , ζH2 will exceed
300 mV at Pt. Oxygen losses will be even larger, and the cell will have a total overpotential loss >700 mV. Two such cells will o have losses >EH2 O ; unnecessary losses if the cell was illuminated at 1 sun or employed larger surface area or more effective electrocatalysts. Under these conditions, two GaInP2 /GaAs cells placed in series should drive three water electrolysis cells

in series, effectively increasing the relative photoelectrolysis efﬁciency by 50%.
Previous principal solar water splitting models predict similar dual band gap photoelectrolysis efﬁciencies of only 16%, and
10–18%, respectively [27, 28]. Each are lower than our observed water splitting efﬁciency discussed below. The physics of these models were superb, but their analysis was inﬂuenced by dated technology and underestimated the experimental ηphoto attained by contemporary devices or underestimated the high experimental values of ηelectrolysis , which can be attained. For example, Boltonand coworkers, estimates low values of ηphoto (less than 20% conversion) due to assumed cumulative relative

4.4 Optimizing Photoelectrochemical Solar Energy Conversion:

Ptblack

3

2

2

4

6

8

Ru02

1M HCIO4

Ptblack

1

0

⇓

AR coating [ZnS(50 nm)/MgF2(70 nm)] section expanded P+-AI0.8Ga0.2As
50 nm
P+-AI(0.3−0.15)Ga(0.7−0.85)As 300 nm n -AI(0.15)Ga(0.85)As
1.0 µm n+-AI(0.15)Ga(0.85)As 1.7 nm n -GaAs
20 nm
GaAs (Buffer Layer)
10 nm
+
p -Si
1.0 µm n -Si
350 µm n+-Si 800 nm
Au − Sb/Au

RuO2

hν

hν

Solar water-splitting system
Illumination
Cross

4

P+-GaAs Au-Zn/Au

Photoelectrolysis current
[mA]

5

AIGaAs/Si RuO2/Ptblack Photoelectrolysis
Illumination power: 135 mW cm−2
Illuminated & electrolysis area: 0.22 cm2 & 10 cm2

10

12

Time
[hours]
Schematic representation (inset) and measured characteristics of the illuminated
AlGaAs/Si RuO2 /Ptblack photoelectrolysis cell. Further details of the layered AlGaAs/Si structure are given in Ref. 6.

Fig. 13

secondary losses that include: 10% reﬂection loss, 10% quantum-yield loss,
20% absorption loss. As summarized in
Table 3 demonstrated ηphoto are substantially higher than 20%. Many ηphoto that have been carefully reviewed or monitored are in excess of 30% [29].
Each of the cells in the Table 3 exhibits an open-circuit photopotential signiﬁcantly greater than the minimum potential necessary to split water. The majority of these cells can generate a photopotential in excess of 2 volts. An unnecessary limit of one multiple band gap photoexcitation per electrolysis would under-utilize Vphoto , diminishing ηelectrolysis . For example, a
GaInP/GaAs cell has a maximum power photopotential of 2.0 to 2.1 V and an open-circuit potential of 2.3 V [12]. Two

such cells in series will efﬁciently drive three 1.3 to 1.4 V water electrolysis cells in series, and as discussed here at a water electrolysis at 1.36 V (Fig. 12 inset) yielded electrolysis efﬁciencies of more than 90%. It is reasonable that with larger surface area, or more effective electrocatalysis, these efﬁciencies will approach 95%.
Using this range of ηelectrolysis = 90–95%: ηphotoelectrolysis (predicted maximum)
= ηphoto × 90–95%
(21)
Table 3 includes predicted maximum ηphotoelectrolysis using observed ηphoto of various dual band gap sensitizers. It is seen that solar water splitting efﬁciencies may be viable at up to double the amount of that previously predicted. Efﬁcient, three or more multiple band gap photoelectrolysis

379

380

4 Solar Energy Conversion without Dye Sensitization
Tab. 3

Predicted and measured photoelectrolysis efﬁciencies. Calculated ηphotoelectrolysis are from

Eq. (21)
Photovoltaic

GaInP/GaAs
GaInP/GaAs
GaAs/Si
GaAs/GaSb
InP/GaInAs
GaAs/GaInAsP
AlGaAs/Si
GaInP2 /GaAs
InP/GaAs

Light
Level
1
180
350
100
50
40
1
11
1

sun sun sun sun sun sun sun sun sun

ηphoto
Measured
30.3% (Ref.
30.2% (Ref.
29.6% (Ref.
32.6% (Ref.
31.8% (Ref.
30.2% (Ref.
21.2% (Ref.

30)
30)
30)
30)
30)
30)
15)

will be expected to be capable of attaining even higher efﬁciencies.
4.4.3

ηphotoelectrolysis
Predicted Maximum
27–29% (Ref.
27–29% (Ref.
27–28% (Ref.
29–31% (Ref.
29–30% (Ref.
27–29% (Ref.
19–20% (Ref.

6)
6)
6)
6)
6)
6)
6)

Measured

18.3% (Ref. 6)
12.4% (Ref. 27)
8.2% (Ref. 30)

phenomena. This systematically probes the approach in studies on polysulﬁde, ferrocyanide, polyselenide, and polyiodide
PECs.

Solution Phase Phenomena n-Cd Chalcogenide/Aqueous
Polysulﬁde Photoelectrochemistry
In 1976, the ﬁrst regenerative PECs with substantial and sustained solar to electrical conversion efﬁciency were demonstrated.
These PECs are based on n-type cadmium chalcogenide (S, Se or Te) electrodes immersed in aqueous polychalcogenide electrolytes. The cells were introduced by Hodes, Cahen, and Manassen [31],
Wrighton and coworkers [32], and Heller and Miller [33] and were capable of converting up to 7% of insolation to electrical energy. Most investigations of these systems focused on solid-state and interfacial aspects of these PECs and photodriven oxidation of polysulﬁde at the photoelectrode was represented:
4.4.3.2

Solution Phase Chemistry
Optimization
In PECs, the manifold conditions in which solution phase chemistry constrains the kinetics and thermodynamics of photoelectrochemical charge transfer have been probed. Modiﬁcation of the solution phase chemistry, can substantially impact on photoelectrochemical properties by (1) enhancing facile charge transfer, (2) suppressing competing reactions and suppressing both (3) electrode and
(4) electrolyte decomposition products, as well as (5) substantially affecting the opencircuit photovoltage. These limitations have been studied in a variety of redox couples and electrolytes. A chemical mechanism was presented [7, 8] for photoelectrochemical solar to electrical energy conversion which emphasized understanding of the distribution of species in photoelectrochemical electrolytes. The approach systematically modiﬁes observed photoelectrochemical kinetic and thermodynamic
4.4.3.1

−

n-cd(Se,Te)

−−
S− + 2hn −− − −→ S + 2e−

(22)

However as represented in Fig. 14(c), the number and type of species in polysulﬁde photoelectrolytes is considerably more complex than represented by Eq. (22). The

4.4 Optimizing Photoelectrochemical Solar Energy Conversion:
Photoelectrochemical solar cell solution phase studies

hν

Photoelectrode

In solution
H2S
HS−
S=
S2=
S2−
S3=
S4=
S5=
OH−
H+
Counter electrode

(a) Aqueous polysulfide PEC

Load

Load

Photoelectrode

hν

Photoelectrode

In solution
Fe(CN)64−
HxFe(CN)6(x − 4)
KyFe(CN)6(y − 4)
HxKyFe(CN)6(y + x − 4)
HxFe(CN)6(x − 3)
KyFe(CN)6(y − 3)
HxKyFe(CN)6(y + x − 3)
CN−
OH−
H+
Counter electrode

(b) Aqueous ferrocyanide PEC

Load

hν

In solution
H2Se
HSe−
Se=
Se2=
Se3=
Se4=
OH−
H+
Counter electrode

(c) Aqueous polyselenide PEC

Chemical species coexisting in polysulﬁde (a), ferrocyanide (b) and polyselenide (c) aqueous electrolytes used in regenerative PECs.
Fig. 14

addition to water of a simple soluble sulﬁde salt and sulfur to water gives rise to a wide distribution of species in solution. The equilibria constraining this distribution has been investigated by Licht and coworkers [34–39], Gigenbach [40], and Teder [41].
It has been shown that modiﬁcation of the distribution of species in aqueous polysulﬁde solution can be correlated to variations in transparency, conductivity, activity, and cadmium chalcogenide photoelectrochemistry. Speciﬁcally, the separate effects of hydroxide and pH modiﬁcation [42], sulfur [43] and sulﬁde [44], and cation [45, 46] were investigated in terms of speciation and photoelectrochemical phenomena. Previously cadmium chalcogenide polysulﬁde PECs had generally employed electrolytes composed 1 molar in sodium sulﬁde, sulfur and sodium hydroxide and resulted in solar conversion efﬁciencies of approximately 7% [31–33].
However, added hydroxide is to be minimized in these cells, cesium is the

preferred cation, and a sulfur to sulﬁde ratio of approximately 1.5 to 1 resulted in a near doubling of the conversion efﬁciency [47]. Figure 15 presents n-Cd(Se,Te) photocurrents measured at various applied potentials in a traditional and modiﬁed electrolyte. As seen in the ﬁgure, the cumulative effect of polysulﬁde electrolyte modiﬁcations on photoelectrochemical solar to electrical energy conversion by nCd(Se,Te)/aqueous polysulﬁde PECs can be considerable.
A greater percentage of photogenerated holes utilized in constructive oxidation of polysulﬁde results in enhanced photocurrents. This has the additional beneﬁt of fewer oxidizing holes available for attack on the semiconductor (photocorrosion).
As seen in Fig. 16, this results in enhanced photocurrent stability of the PEC.
This study showed that with solution optimization, not only the photocurrent but also the polysulﬁde electrolyte exhibits enhanced lifetime, both approaching oneyear operation outdoors [48].

381

4 Solar Energy Conversion without Dye Sensitization n-CdSe0.65 Te0.35 in aqueous polysulfide

1.6
7.7% efficiency
0.8

Photocurrent
[mA]

12.7% efficiency

Fig. 15 Potentiostatic photocurrent voltage characteristics for an illuminated n-CdSe0.65 Te0.35 single crystal immersed in either of two types of aqueous polysulﬁde electrolyte: Top curve in 1.8 M Cs2 S and 3 M sulfur:; bottom curve is in 1 M NaOH, 1 M
Na2 S, 1 M sulfur. The photocurrent voltage curves were obtained outdoors, and solar to electrical conversion efﬁciencies are indicated.

~AM1 illumination
−800

−600

−400

−200

Voltage
[mV]
0.5

0.4

Photovoltage over load
[Volts]

382

0.3

0.2

0.1

Jan

July

Jan

July

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

4.4 Optimizing Photoelectrochemical Solar Energy Conversion:

n-Cd Chalcogenide/Aqueous
Ferrocyanide Photoelectrochemistry
Reichman and Russak [49] and Freeze [50] had reported high (12 to 14%) conversion efﬁciencies for n-CdSe/ Fe(CN)6 3−/4− solar cells in electrolytes containing a 1 : 1 ratio of aqueous Fe(CN)6 4− to Fe(CN)6 3− salts in a highly alkaline environment.
However, Rubin and coworkers [51] had shown that these n-CdSe/Fe(CN)6 3−/4− solar cells can have a limited stability and photocurrents decrease within the order of hours. Their attempts to enhance the photoconversion stability of these systems have included cationic surface modiﬁcation of the n-CdSe photoelectrode and isolation of the wavelength dependence of the surface instability [51, 52]. In these studies, photo-oxidative processes in nCdSe/Fe(CN)6 3−/4− PECs were typically represented: 4.4.3.3

n-CdSe

Fe(CN)6 4− + hn − − → Fe(CN)6 3− + e−
−−
(23)

Photocurrent
[mA cm−2]

10

The role of the various redox active constituents in constraining n-CdSe/
Fe(CN)6 3−/4− photo-oxidative energy conversion had not been previously investigated, and ferricyanide and ferrocyanide salts exhibit a complex aqueous equilibria. However, in a manner analogous to the polysulﬁde electrolytes, and as illustrated in Fig. 14(b), these salts exhibit a complex aqueous equilibria, resulting in a variety of Fe(II) and Fe(III) species. As with the polysulﬁde electrolytes, we have probed the equilibria constraining the distribution of species in these solutions. This speciation can be controlled and substantially effects n-CdSe photo-oxidative energy conversion. The photoelectrochemical effect and solution stability effects of pH, cation modiﬁcation and ratio of ferrocyanide to ferricyanide, have been studied by Licht and Peramunage [9, 53–55], and also the systematic variation of the primary photo-oxidized species. As seen in Fig. 17, substitution of a single ligand in the

Photoelectrochemistry of n -CdSe in
[Fe(CN)5 -L]3−/2− aq
L

a
5

a) −CN−
b) −NH3

b

c) −NO2−

c

d) −Fe(CN)53−

d e 0
−0.8

−0.4

e) −NO+
0.0

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

383

4 Solar Energy Conversion without Dye Sensitization

hexacyanoferrate species has a substantial effect on photovoltage and photocurrent,
Licht and Peramunage [56].
As seen in Fig. 18, addition of potassium cyanide to an alkaline ferrocyanide electrolyte substantially enhances both photocurrent stability (ﬁgure inset) and photovoltage (ﬁgure outset), Licht and Peramunage [9]. It was proposed by Bocarlsy and coworkers [57] that in addition to
Fe(CN)6 4− , CN− was also photooxidized
(to CNO− ) by the photoelectrode. Our subsequent investigations indicated that although cyanide may be chemically oxidized to cyanate, no cyanide was photoelectrochemically oxidized by the semiconductor
Licht and Peramunage [53–56].

20

n-GaAs/Aqueous Polyselenide
Photoelectrochemistry
n-GaAs/aqueous polyselenide PECs have shown stable efﬁcient solar to electrical conversion. Photodriven oxidation of polyselenide can be written as
4.4.3.4

−

n-GaAs

−−
Se− + 2hn − − → Se + 2e−
−

n-GaAs

Iphoto (normalized)

θ

d b 0

0

1

3
Time
[days]
No Illumination

0
−1.2

−1.0

−0.8

−0.6

−0.4

(25)

Parkinson, Heller, and Miller [58] and
Lewis and coworkers [59] have shown that metal ion (Ru3+ , Os3+ ) treatment of the n-GaAs surface leads to high solar to electrical conversion efﬁciencies in these cells

1

5

−

2Se− + 2hn − − → Se2 − + 2e−
−−

40 mW cm−2 illuminated single crystal n-CdSe in
0.5 mK4Fe(CN)6, 25 mm K3Fe(CN)6, 0.5 m KOH without (dashed) or with (solid line) 0.1 m KCN

10

(24)

or:

15

Current Density
[mA cm−2]

384

−0.2

0.0

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

4.4 Optimizing Photoelectrochemical Solar Energy Conversion:
−

−

−

−
−

−
−

−
−

Se3 − + Se− ←−→ 2Se2 −

(with up to 15% conversion efﬁciencies reported). Lewis and coworkers [60] have further investigated 0.1 M to 1.0 M concentration K2 Se electrolytes. The ﬁrst and second acid dissociation constant of hydrogen selenide of pK1 = 3.9 and pK2 =
13.0 have been well characterized by several investigators including Myers and coworkers [61]. However, as represented in the right hand portion of Fig. 14, the fundamental equilibria constraining the distribution of polyselenide species was not characterized. Measurements of these equilibria are a ﬁrst step in determining the regenerative reactive species in polyselenide medium and in probing how they effect photoelectrochemical energy conversion. The equilibria constraining the formation of diselenide, triselenide, and tetraselenide may be written:

2Se4 + Se ←−→ 3Se3

(26)
(27)

with equilibrium constants for the equilibrium given by K23 and K34 , respectively.
These equilibria were probed by study of rest potential variation with solution composition and by isolation of the near UV absorption peaks of the various polyselenide species, and yield equilibrium constants of pK23 = −0.65 and pK34 = −4.2. When combined with K1 and K2 these provide a description of polyselenide speciation [62,
63].
Using the understanding of polyselenide speciation in solution, the rationale electrolyte modiﬁcation of aqueous polyselenide photoelectrochemical solar cell gas been investigated. n-GaAs photocurrent, photovoltage, and photopower are affected by the distribution of hydroselenide,

19

J, Photocurrent density
[mA cm−2]

J, Stable

18

J, Fluctuating
(Corrosion)
J, Diminishing

17

16

15

0

1

2
KOH
[M]

3

Short-circuit photocurrent density, Jsc , for a 75 mW cm−2 tungsten-halogen illuminated single crystal n-GaAs immersed 1 M K2 Se,
0.01 M Se and in varying KOH concentration.
Fig. 19

4

385

4 Solar Energy Conversion without Dye Sensitization

selenide, and polyselenide in solution.
Polyselenide electrolytes containing 1 to
2 m hydroxide and 1 m selenide (as K2 Se) enhance n-GaAs photocurrent density stability (Reported as ‘‘J’’ in Fig. 19), whereas photocurrent is unstable at either 0.5 m
KOH or 3.0 m KOH. The presence of polyselenide species in low concentrations (as
0.01 m dissolved selenium) are important for desorption of selenium from the nGaAs. Furthermore, electrolytes enriched in dissolved selenium, (as up to 0.2 m dissolved selenium) enhance n-GaAs photovoltage. As seen in Fig. 20, the photopower (the product of the photocurrent and the applied potential) as well as the photopower improves with an increase in selenium. However, these electrolytes diminish electrolyte transmittance necessitating use of an alternate back wall

cell conﬁguration discussed in Licht and
Forouzan [62, 63].
4.4.3.5

Aqueous Polyiodide Photoelectrochemistry

The aqueous polyiodide (I− /I3 − ) is the only redox couple shown to be compatible with efﬁcient oxidative photoelectrochemistry at n-type transition metal dichalcogenide including WSe2 , MoS2 ,
WS2 , and MoSe2 . Following the pioneering work of Tributsch and coworkers [64], several groups including Parkinson, Heller, and Miller [65] and Tenne and Wold [66] have reported n-WSe2 or n-MoSe2 /aqueous polyiodide solar to electrical energy conversion of more than 10% and/or photocurrent stability in excess of 105 coulombs cm−2 . The photodriven oxidation of iodine at tungsten diselenide

3.0

2.5

0.01 M Se
0.02 M Se
0.20 M Se

2.0

Power
[mW cm−2]

386

1.5

1.0

0.5

0.0

0.0

0.2

0.4

0.6

Photovoltage
[volts]
Photopower variation of a 92 mW cm−2 tungsten-halogen illuminated backwall n-GaAS/polyselenide PEC as a function of increasing selenium concentration.

Fig. 20

4.4 Optimizing Photoelectrochemical Solar Energy Conversion:

To a lesser extent the solution phase species IO3 − , HIO, H2 O I+ , I5 − , and
I6 2− can also occur. Figure 21 presents the relative variation of polyiodide speciation in solutions containing 0.01 molal iodine and 1 to 12 molal iodide and which may be compatible with n-WSe2 regenerative photoelectrochemistry. As emphasized in the ﬁgure inset, in these electrolytes the concentration of added iodide dominates iodine in these solutions. The distribution of species in these solutions is calculated in accordance with Eqs. (30) and (31). As can be seen in the ﬁgure inset, little (less than
0.3%) of the added iodine exists as solution

has been described in one of two manners: n-WSe2 ◦

−−
I− + hn − − → I + e−

(28)

or: n-WSe2 3I− + 2hn − − → I3 − + 2e−
−−

(29)

Signiﬁcant species in aqueous polyiodide solution can include: I− , I3 − , I4 2− , I− ,
OH− , and H+ . Constrained by pH and the equilibria [67, 68]:
I− + I2 ←−→ I3 −
I3 − + I− ←−→ I4 2−

K3 = 723

(30)

K4 = 0.20 (31)

100%
Aqueous polyiodide speciation
0.01 M iodine in the indicated concentration of sodium iodide
80%

Total iodine = 0.01 m = [I2] + [I3− ] + [I 4= ]

Percent of total

Total iodide = [Nal] = [I −] + [I3−] + 2[I4=]
60%
[I 4=] / Total iodine

[I3−] / Total iodine
40%
0.3%
% of total

20%

0.0%

0%

[I2] / Total iodine

0

2

Total iodine / Total iodide

0

4

4

NaI
[m]
6

8

12

8

10

12

NaI
[molal]
−

The fractions of I3 − and I4 − (compared to total iodine) in aqueous polyiodide solutions as a function of added iodide concentration. Inset: The fractions of solution phase iodine (compared to total iodine) and total iodine (compared to total iodide).

Fig. 21

387

388

4 Solar Energy Conversion without Dye Sensitization

phase iodine, and the maximum [I2 ] is
1 × 10−4 m. As seen in the ﬁgure outset, the bulk (greater than 99.7%) of the added iodine resides as the polyiodide species
I3 − and I4 2− . In iodide concentrations less than 5 m, I3 − is the predominant polyiodide, whereas at higher iodide concentrations, I4 2− predominates. Longer chain species, I5 − and I6 2− , if they exist, have upper limit concentrations of 1 × 10−6 m and 1 × 10−8 m, respectively. Hence, in these aqueous polyiodide electrolytes, I2 ,
I5 − and I6 2− are not at substantial concentrations in these photoelectrolytes, and a primary oxidizable ion is I− and a primary reducible ion is either I3 − or I4 2− .
Choice of cation in aqueous polyiodide solution can effect n-tungsten dichalcogenide photoelectrochemistry. The alkali cations do not substantially interact with solution phase iodide. However, other metal cations including Ag+ , or Zn2+ or
Cd2+ , will create a series of complexes with dissolved iodide including: AgI2 − ,
AgI3 2− , ZnI3 − , ZnI4 2− , CdI4 2 and so on. These complexes can shift rest potentials or either enhance or diminish charge transfer from the semiconductor surface. Because of the limitations on silver iodide solubility (Ksp (AgI) = 10−16 ) high (molal) level concentrations of silver will dissolve only in solutions that facilitate complex formation as in concentrated
NaI or KI solutions. In particular, we have shown that the presence of high concentrations of dissolved silver is advantageous to n-tungsten diselenide photoelectrochemistry. As seen in Fig. 22, silver dissolved as
AgNO3 enhances the voltage at maximum point. The extent of the improvement varies with the initial condition of the individual n-WSe2 crystal, and general improvements are a 15 to 45 mV increase in the voltage of maximum power, Vmax , and a 5 to 20% relative increase in power.

These improvements are sustained in the silver-bearing electrolytes. On returning the electrode to the polyiodide electrolyte without silver, the PEC gradually (on the order of hours) returns to the silver free
PEC behavior. This improvement appears to provide a mechanism for long-lasting suppression of n-WSe2 exposed edges and recombination sites as further discussed by Licht and Myung [67, 68].
4.4.4

Concluding Remarks

In the 1970s through 1990s health concerns (air pollution) and political concerns
(localized shortages) were cited in the need for development of alternative energy sources. Today, in addition to these concerns, growing awareness of carbon dioxide emissions as a green house gas, as well as the economic realization that oil and coal are better allocated as raw materials for pharmaceuticals and plastics, than for electrical energy generation, provide impetus to the technological development of renewable energy sources.
Solar energy remains the principal symbol of a clean, abundant renewable energy source. Society’s energy needs are continuous. Unlike purely solid-state, photovoltaic processes, the advantage of combined solar/electrochemical processes is that energy is not only converted but also may be stored for future use when sunlight is not so frequently available.
Limiting constraints of multiple band gap photoelectrochemical energy conversion, as well as practical conﬁgurations for efﬁcient solar to electrical energy conversion have been probed [1–6]. Such systems are capable of better matching and utilizing of incident solar radiation (insolation).
Efﬁcient solar cells, solar storage cells,

4.4 Optimizing Photoelectrochemical Solar Energy Conversion:

Illuminated n-Wse2 in 6 m iodide, 0.01 m iodine

1.0

Relative power

0.8

0.6

0.4

0.2

0.0 in Nal

Without silver

With 0.7 m silver nitrate
−0.6

in Kl

Without silver

With 1.0 m silver nitrate

−0.4

−0.2

0.0 −0.6

−0.4

−0.2

0.0

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

and solar hydrogen generation systems are discussed and demonstrated.
Along with conventional parameters constraining photovoltaic devices, modiﬁcation of the electrolyte (solution phase) chemistry is a key to understanding the mechanism of energy conversion and device characteristics of PECs. The fundamental importance of modiﬁcation of the electrolyte in terms of the distribution of species in photoelectrolytes and the pragmatic importance in terms of enhanced solar to electrical conversion efﬁciency is reiterated by studies

in polysulﬁde, ferrocyanide, polyselenide, and polyiodide electrolyte regenerative
PECs.
The advantage of multiple band gap semiconductor processes is in the ability to improve the energy match to the solar spectra, and hence improved utilization/conversion of the incident solar radiation (insolation). Limiting constraints of multiple band gap semiconductor/electrolyte energy conversion, as well as practical conﬁgurations for solar to electrical energy conversion have been probed. Such combined systems are capable of highly

389

390

4 Solar Energy Conversion without Dye Sensitization

efﬁcient utilization of solar energy. Efﬁcient solar cells, solar storage cells, and solar hydrogen generation systems are discussed and demonstrated. Alternate processes discussed include both an inverted (1 photon per e− ) or bipolar (n ≥ 2 photons per e− ) arrangement of successive band gaps, which can contain either a
Schottky or an ohmic photoelectrochemical solution interface, and can drive either regenerative single, or multiple different, redox reactions. The latter case permits solar energy storage or solar water splitting evolving H2 , and provides an energy reservoir which may compensate for the intermittent nature of solar energy. Generated
H2 is attractive as a clean, renewable fuel.
Experimental conﬁgurations examined include GaAs/Si or AlGaAs/Si photodriven redox couples, each with a solar to electrical conversion efﬁciency of 19 to
20%. A related solar cell is conﬁgured with electrochemical storage, and which provides a nearly constant energetic output in illuminated or dark conditions. Similarly, multiple band gap semiconductors can also be utilized to generate hydrogen fuel by solar-driven water splitting.
A cell containing illuminated AlGaAs/Si
RuO2 /Ptblack evolves H2 and O2 at record
(18.3%) efﬁciency. Contemporary models underestimate the attainable efﬁciency of solar energy conversion to water splitting, and future multiple band gap systems are calculated as capable of attaining more than 30% solar efﬁciency.
Acknowledgment

S. Licht is grateful for support by the BMBF
Israel-German Cooperation.
References
1. S. Licht, O. Khaselev, T. Soga et al., Electrochem. Solid State Lett. 1998, 1, 20.

2. S. Licht, O. Khaselev, P. A. Ramakrishnan et al., J. Phys. Chem. B 1998, 102, 2536.
3. S. Licht, O. Khaselev, P. A. Ramakrishnan et al., J. Phys. Chem. B 1998, 102, 2546.
4. S. Licht, B. Wang, T. Soga et al., Appl. Phys.
Lett. 1999, 74, 4055.
5. B. Wang, S. Licht, T. Soga et al., Sol. Energy
Mater. Sol. Cells 2000, 64, 311.
6. S. Licht, H. Tributsch, B. Wang et al., J.
Phys. Chem. B 2000, 104, 8920.
7. S. Licht, Sol. Energy Mater. Sol. Cells 1995,
38, 305.
8. S. Licht, Nature 1987, 330, 148.
9. S. Licht, D. Peramunage, Nature 1990, 345,
330.
10. B. Miller, S. Licht, M. E. Orazem et al., Photoelectrochemical systems, Crit. Rev. Surface
Chem. 1994, 3, 29.
11. C. H. Henry, J. Appl. Phys. 1980, 51, 4494.
12. D. J. Friedman, S. R. Kurtz, K. Bertness et al., Prog. Photovolt. 1995, 3, 47.
13. J. P. Benner, J. M. Olson, T. J. Coutts, in
Advances in Solar Energy, Chap. 4 (Eds.:
K. W. Boer), American Solar Energy Society, Inc., 1992, 125, Vol. 7.
14. M. A. Green, K. Emery, K. Bucher et al.,
Progr. Photovolt. 1996, 4, 321.
15. T. Soga, T. Kato, M. Yang et al., J. Appl.
Phys. 1995, 78, 4196.
16. J. White, F-R. Fan, A. J. Bard, J. Electrochem.
Soc. 1985, 132, 544.
17. R. Memming, in Photochemical Conversion and Storage of Solar Energy, (Eds.:
E. Pelizzetti, M. Schiavello), Kluwer Acad.
Pub., Netherlands, 1991, pp. 193.
18. S. Licht, G. Hodes, R. Tenne et al., Nature
1987, 330, 148.
19. A. Heller, H. J. Lewerenz, B. Miller, J. Am.
Chem. Soc. 1981, 103, 200.
20. C. L´ vy Cl´ ment, A. Lagoubi, M. Neumanne e Spallart et al., J. Electrochem. Soc. 1991, 138,
L69.
21. H. Kobayashi, T. Kubota, N. Toshikawa et al., J. Electroanal. Chem. 1995, 398, 165.
22. S. R. Ovshinsky, M. A. Fetcenko, J. Ross,
Science 1993, 260, 176.
23. G. Lodi, E. Sivieri, A. Debattisti et al., J.
Appl. Electrochem. 1978, 8, 135.
24. N. Alonso-Vante, H. Colell, U. Stimming et al., J. Phys. Chem. 1993, 97, 7381.
25. P. Salvador, N. Alonso-Vante, H. Tributsch,
J. Electrochem. Soc. 1998, 145, 216.
26. O. Khaselev, J. Turner, Science 1998, 280,
425.

4.4 Optimizing Photoelectrochemical Solar Energy Conversion:
27. J. R. Bolton, S. J. Strickler, J. S. Connolly,
Nature 1985, 316, 495.
28. M. F. Weber, M. J. Digman, Int. J. Hydrogen
Energy 1986, 11, 225.
28a. M. F. Weber, J. Electrochem. Soc. 1984, 131,
1258.
29. M. A. Green, K. Emery, K. Bucher et al.,
Prog. Photovolt. 1999, 11, 31.
30. R. C. Kainthla, B. Zelenay, JO’M. Bockris, J.
Electrochem. Soc 1986, 133, 248.
31. G. Hodes, J. Manassen, D. Cahen, Nature
1976, 261, 403.
32. A. B. Ellis, S. W. Kaiser, M. S. Wrighton, J.
Am. Chem. Soc. 1976, 98, 1635.
33. B. Miller, A. Heller, Nature 1976, 262, 680.
34. S. Licht, G. Hodes, J. Manassen, Inorg.
Chem. 1986, 25, 2486.
35. S. Licht, J. Manassen, J. Electrochem. Soc.
1987, 134, 918.
36. S. Licht, J. Electrochem. Soc. 1988, 135, 2971.
37. F. Forouzan, K. Longo, D. Peramunage et al., Anal. Chem. 1990, 62, 1356.
38. K. Longo, F. Farouzan, D. Peramunage et al., J. Electroanal. Chem. Interfacial Electrochem. 1991, 318, 111.
39. D. Peramunage, F. Forouzan, S. Licht, Anal.
Chem. 1994, 66, 378.
40. W. Giggenbach, Inorg. Chem. 1971, 10,
1333.
41. A. Teder, Sr. Papperstodn 1969, 72, 245.
42. S. Licht, Joost Manassen, J. Electrochem.
Soc. 1985, 132, 1077.
43. S. Licht, G. Hodes, Joost Manassen, J.
Electrochem. Soc. 1986, 133, 272.
44. S. Licht, J. Manassen, J. Electrochem. Soc.
1986, 133, 277.
45. S. Licht, R. Tenne, H. Flaisher et al., J.
Electrochem. Soc. 1986, 133, 52.
46. S. Licht, R. Tenne, H. Flaisher et al., J.
Electrochem. Soc. 1984, 131, 5920.
47. S. Licht, R. Tenne, G. Dagan et al., Appl.
Phys. Lett. 1985, 46, 608.

48. S. Licht, J. Phys. Chem. 1986, 90, 1096.
49. J. Reichman, M. Russak, J. Electrochem. Soc.
1984, 131, 796.
50. K. W. Freeze Jr., Appl. Phys. Lett. 1982, 40,
275.
51. H. D. Rubin, B. D. Humphrey, A. B. Bocarsley, Nature 1984, 308, 339.
52. H. D. Rubin, D. J. Arent, A. B. Bocarsley, J. Electrochem. Soc. 1985, 132,
523.
53. S. Licht, D. Peramunage, Nature 1991, 354,
440.
54. S. Licht, D. Peramunage, J. Electrochem. Soc.
1992, 139, L23.
55. S. Licht, D. Peramunage, J. Electrochem. Soc.
1992, 139, 1791.
56. S. Licht, D. Peramunage, Sol. Energy 1994,
52, 197.
57. G. Seshardi, J. K. M. Chun, A. B. Bocarsly,
Nature 1991, 352, 508.
58. B. A. Parkinson, A. Heller, B. Miller, J. Electrochem. Soc. 1979, 126, 954.
59. B. J. Tufts, I. L. Abrahams, P. G. Santengelo et al., Nature 1987, 326, 861.
60. L. G. Casasgrand, B. J. Tufts, N. S. Lewis, J.
Phys. Chem. 1991, 95, 1373.
61. D. E. Levy, R. J. Myers, J. Phys. Chem. 1990,
94, 7842.
62. S. Licht, F. Forouzan, J. Electrochem. Soc.
1995, 142, 1539.
63. S. Licht, F. Forouzan, J. Electrochem. Soc.
1995, 142, 1546.
64. W. Kautek, H. Gerischer, H. Tributsch, J.
Electrochem. Soc. 1981, 128, 2471.
65. B. A. Parkinson, A. Heller, B. Miller, Appl.
Phys. Lett. 1978, 33, 521.
66. R. Tenne, A. Wold, Appl. Phys. Lett. 1985,
47, 707.
67. S. Licht, N. Myung, J. Electrochem. Soc.
1985, 142, 845.
68. S. Licht, N. Myung, J. Electrochem. Soc.
1995, 142, L129.

391

397

5.1

5.1.2

Dye-Sensitized Regenerative Solar Cells

Device Concepts

..
Augustin J. McEvoy and Michael Gratzel
Ecole Polytechnique F´d´rale de Lausanne, ee Lausanne, Switzerland
5.1.1

Overview

At present, dye-sensitized photosystems provide the only technically and economically credible alternative to solid-state photovoltaic devices. The concept reconciles the electrochemical stability of wide band gap photoelectrodes with the efﬁciency of photovoltaic devices having a more extended absorption spectrum in the visible range. Optical absorption and charge separation take place on distinct sites within these photovoltaic cells.
Hence, oppositely charged species are restricted to separate phases, so conventional recombination losses are inhibited. In consequence, device photoconversion efﬁciency is better maintained at low light levels than with conventional semiconductor solid-state junction devices. Various conﬁgurations of the dye-sensitized concept are under investigation, including electrochemical and heterojunction variants. It should be remembered that the ﬁrst observation of the photovoltaic effect by A. -E. Becquerel [1] in 1839 was at a solid–liquid interface, with a semiconductor photoelectrode contacting an electrolyte, and was technically a photoelectrochemical cell. Nevertheless, the modern photovoltaic industry is established exclusively on solid-state devices, with semiconductors of appropriate band gap as optical absorbers and junctions to solids with different conduction mechanisms as the sites of charge separation and photovoltage generation. In commerce are found homojunction devices, in which the semiconductors are chemically identical, differing only in impurity content and therefore in free carrier polarity; heterojunctions, in which the semiconductors of opposite carrier polarity are chemically different, for example, cadmium sulﬁde or copper indium-gallium selenide contacts; and ﬁnally Schottky junctions in which the contacting phase is a metal. However, all these are solid-state devices in which conduction throughout is by an electronic mechanism. None is a photoelectrochemical device that has an electrolyte as the contacting phase or that uses an ionic conduction mechanism. Awareness of

398

5 Dye-Sensitized Photoelectrochemistry

photoelectrochemistry in the photovoltaic world therefore remains marginal, although this awareness is rapidly growing. Intensive research over the past two decades has led to an inescapable conclusion that those semiconductors, whose band gaps are sufﬁciently wide to give a long-term stability under illumination, in contact with an electrolyte and a technically useful photovoltage are insensitive to visible light, requiring higher-energy ultraviolet photons to excite charge carrier pairs. A narrower band gap, compatible with the photoconversion of visible light, is indicative of weaker chemical bonding of the semiconductor and hence liability to photocorrosion and incompatibility with a stable extended lifetime as an energyconversion device. The resolution of this dilemma lies in the separation of light absorption and charge-separation functions by sub–band gap sensitization of the semiconductor with an electroactive dye.

The realization in practice of this option is simply presented. Analogous to the band gap of the semiconductor is the
HOMO-LUMO (highest occupied molecular orbital ⇒ valence band/lowest unoccupied molecular orbital ⇒ conduction band) gap of a molecule. When the latter lies above the conduction band edge of a semiconductor substrate, an electron entering the LUMO on photoexcitation of the dye by a photon within its absorption spectrum energy range can be injected into the semiconductor. The positively charged oxidized dye molecule can then be neutralized and its ground state restored by reaction and charge exchange with the electrolyte, which in turn receives an electron from a metallized counterelectrode.
With a closed external circuit and under illumination, the device then constitutes a photovoltaic energy-conversion system, which is regenerative and stable, and is functionally comparable to its solid-state

Conducting glass TiO2
Dye Electrolyte
E /V vs NHE
Injection
S*
−0.5

Maximum voltage 0

hν
Red

0.5

ox
Mediator

Interception
1.0

Cathode

Diffusion

S°/S+

Schematic of the structure and function of the dye-sensitized electrochemical photovoltaic cell. The nanoporous sensitized semiconductor photoanode receives electrons from the photoexcited dye, which is thereby oxidized and which in turn oxidizes the mediator, a redox species dissolved in the electrolyte. The mediator is regenerated by reduction at the metallic cathode, while the electrons circulate in the external circuit.

Fig. 1

5.1 Dye-Sensitized Regenerative Solar Cells

junction analog. A typical wide band gap semiconductor for photoelectrochemistry with stability against photocorrosion is titanium dioxide. Titanium dioxide has a band gap of 3.1 eV and is therefore insensitive to visible light. The selected dye must have a HOMO-LUMO gap corresponding to a photon energy in the visible or infrared parts of the spectrum, with an optimum near the solar spectra maximum of 1.4 eV. This mechanism is illustrated schematically in Fig. 1. However, before presenting the relevant characteristics of a semiconductor electrode that is apt for chemisorption of a dye and therefore photovoltaic application, a background of the history and development of the sensitization of semiconductors is necessary.
5.1.3

Photography and Photoelectrochemistry

Although artisanal ‘‘virtual reality’’ by painting is as old as human culture, synthetic ‘‘virtual reality’’ by technical reproduction of images is relatively recent, with the ﬁrst photography by Daguerre in 1837 and the introduction of the silver halide process by Fox-Talbot in 1839.
Thereafter, the art advanced so rapidly that the nineteenth century events – the wars, the explorations, the early factory age and the introduction of steam power – are vividly documented by photographs. Photography remained an art for a century, becoming a science only with the theoretical analysis of the process by Gurney and Mott [2] in 1938. Nevertheless, an empirical skill in the formulation of photographic emulsions had evolved in the interval, even in the absence of a fundamental understanding. Early emulsions, referred to as ‘‘orthochromic,’’ had a distinct insensitivity to midspectral and red light, so that their gray scale reproduction

was not realistic of the visual impression of the same scene. So many early photographs therefore give an impression of an unnatural ‘‘cleanliness’’ of the scene imaged that considerable efforts were expended to ﬁnd emulsions capable of more realistic response. The problem is now recognized to be due to the semiconductor nature of the halide grains responsible for optical absorption in photography, which are activated essentially by the photoelectric effect. They have band gaps typically in the range 2.7 to 3.2 eV, and the photoresponse is effectively negligible for wavelengths longer than 460 nm. It was however noted that the use of gelatine as a support medium for the alkali halide grains in the photographic emulsion could signiﬁcantly modify the ﬁlm spectral response. Only in the last century was it ﬁnally demonstrated that an organo-sulfur compound present in calf-skin gelatine was responsible for this extension of sensitivity [3]. The mechanism now recognized is the induction of a nanostructure of silver sulphide on the halide surface of each grain, giving the ﬁrst application of the sensitization of a semiconductor heterojunction! Even more signiﬁcant in the history of sensitization is the work of Vogel, Professor of photochemistry, spectroscopy and photography at the then K¨ nigliche Technische o Hochschule of Berlin, who in the 1870s investigated the effect of dyes on emulsions and found that the halide spectral response extended through the visible and even into the infrared region [4]. Not only did this facilitate his spectroscopic research and contribute to astronomy but also made possible the commercial ‘‘panchromatic’’ broad-spectrum black-and-white ﬁlm with realistic gray scale reproduction, and later with spectrally selective dyes – the modern era of color photography.

399

400

5 Dye-Sensitized Photoelectrochemistry

Meanwhile, the photoelectric effect was a focus of scientiﬁc interest, with a history synchronized with that of photography. The concept of dye enhancement of the photoeffect was carried over from photography to the sensitization of an electrode already in 1887 by Moser [5] (Fig. 2) using the dye erythrosin on silver halide electrodes and conﬁrmed by Rigollot in 1893 [6]. In the retrospectively quaint report, written substantially before the Einstein theory of the photoelectric effect, Moser records his observations on dye-induced enhancement as an increased photopotential (V ) rather than the more fundamental current (A), despite the title – ‘‘Strengthened
Photoelectric Current Through Optical
Sensitization.’’

This parallel evolution of sensitization in photography and in photoelectrochemistry still seems to surprise each generation of photochemists. That the same dyes were effective for both processes was recognized among others by Namba and Hishiki [7] at the 1964 International Conference on Photosensitization in Solids, which can now be seen as a seminal event in the history of dyes in photochemistry. It was also recognized then that the dye should be adsorbed on the semiconductor surface as a closely packed monolayer for maximum sensitization effectiveness [8]. The conference also provided the venue for the identiﬁcation of the sensitization mechanism through charge transfer from a dye excited state; until then, the matter was in dispute, with an energy-coupling process being

A copy of the ﬁrst paper on sensitization published by
James Moser in 1887.

Fig. 2

5.1 Dye-Sensitized Regenerative Solar Cells
The dilemma of 1968 – energy resonance or electron injection?

Fig. 3

proposed, as illustrated in Fig. 3 taken from the conference proceedings [9]. The understanding became deﬁnitive with the work of Gerischer and Tributsch on sensitization of zinc oxide [10]. However, the fact remained that sensitization efﬁciency remained low, as research systematically investigated dyes in solution or those adsorbed on single-crystal substrates. In retrospect, it appears that there was an excessive inﬂuence of solid-state engineering practice in that the disordered surfaces were expected to promote surface recombination and therefore loss of excited charge carriers, whereas single-crystal surfaces, being better deﬁned physically and chemically, make the effects of sensitization more evident. The results of this presumption were slow to emerge.
5.1.4

Sensitization of Powders and Rough
Surfaces

With the ongoing interest in sensitization during the 1970s, and the idea that photocatalysts could possibly carry out electrolytic reactions such as the evolution of hydrogen from water, a phase of experimental development evolved with the sensitization of semiconductor powders [11].
With the ecological awareness of time, photochemical production particularly of energy-related chemicals was seen as a form of ‘‘artiﬁcial photosynthesis.’’ The prototype energy-converting dye provided by nature is indeed chlorophyll, a molecule consisting of a central magnesium atom surrounded by a nitrogen-containing porphyrin ring. Variations are due to minor

0ev

Vacuum Level

I

I e Conduction band

hν
Br −
Br+

D
Dye

hν
Dye
Valence band
Energy
transfer

BrAg

Electron transfer modiﬁcations of certain side groups.
Chlorophyll is in turn similar in structure to hemoglobin, the oxygen-carrying iron-based pigment found in blood. Given that the development of the dye-sensitized cell arose out of an interest in artiﬁcial photosynthesis, the adoption of porphyrin-like organometallic dyes as sensitizers was logical. However, although nature conﬁnes itself to magnesium and iron for its principal pigments, the synthetic chemist can access the whole range of metallic elements. The use of ruthenium pyridyl complexes now has more than twenty years of development history. Clark and Sutin had already used a trisbipyridyl ruthenium complex in 1977 to sensitize titanium dioxide to sub–band gap illumination [12], but in solution only. Charge transfer could only occur after diffusion of the ion to the semiconductor, so the efﬁciency of the sensitization was very low. By 1980, the idea of chemisorption of the dye through an acid carboxylate group bonding to the metal oxide surface had also been established [11] so that the sensitizer was immobilized and it formed the required monomolecular ﬁlm on the semiconductor substrate which facilitated charge transfer by electron injection. The carboxylated trisbipyridyl dye

401

402

5 Dye-Sensitized Photoelectrochemistry

(‘‘RuL3 ’’) therefore became the prototype sensitizer in photochemistry. The subsequent development of sensitizers through to the panchromatic ‘‘black dye’’ that constitutes the current state of the art is dealt with in detail in a separate chapter.
As a substrate for the chemisorption of sensitizing dyes, titania, TiO2 , particularly in the anatase form, gradually became dominant [13]. It has many advantages for sensitized photochemistry and photoelectrochemistry, being a low-cost, widely available, nontoxic and biocompatible material, and as such is even used in health care products and domestic applications such as paint pigmentation.
The objective had also evolved to concentrate on photovoltaic devices rather than on photosynthesis. Progress thereafter was incremental until the announcement in
1991 [14] of the sensitized electrochemical solar cell, with the remarkable conversion efﬁciency at that time of 7.1% under solar illumination, a synergy of structure, substrate roughness, dye photochemistry, counterelectrode kinetics and electrolyte redox chemistry. That evolution has continued progressively since then, with certiﬁed efﬁciency now more than 10% [15].
There has been a continuous development of the electrolyte solvent, aqueous systems being eliminated in favor of organics and molten salts. However, ﬁnding substitutes for the original iodine-iodide redox system has proved to be difﬁcult although recent work in our laboratory shows that alternative redox couples are ﬁnally at hand.
5.1.5

Substrate Development and Fabrication

If molecular design and engineering has underpinned the evolution of efﬁcient, stable sensitizer dyes, the materials science

of nanoporous ceramic ﬁlms is the basis for the understanding of semiconductors.
The nanoporous structure permits the speciﬁc surface concentration of the sensitizing dye to be sufﬁciently high for the total absorption of the incident light that is necessary for efﬁcient solar energy conversion, since the area of the monomolecular distribution of adsorbate is 2 to 3 orders of magnitude higher than the geometric area of the substrate. Moreover, in the electrochemical system, this high degree of roughness does not promote charge carrier loss by recombination, since the electron and the positive charge ﬁnd themselves within picoseconds on opposite sides of the liquid–solid interface. In addition, the photoexcited carrier in the n-type TiO2 is an electron, a majority carrier, and therefore not subject to depletion by carriers of the opposite polarity. Solid-state junction cells, on the other hand, are minority carrier devices. The carrier loss mechanisms are comparatively slow [16], and although conventionally referred to as recombination by analogy with the solid-state process, the loss of a photoexcited electron from the semiconductor should rather be regarded as a recapture of the electron by an oxidized dye species, or a redox capture, when the electron reacts directly with the iodine in the electrolyte. Either can occur on a millisecond timescale, but much slower than electron injection or transport to the rear contact layer.
The original substrate structure used for our early photosensitization experiments was a fractal derived by hydrolysis of an organo-titanium compound but it has since been replaced with a nanostructure deposited from colloidal suspension.
This evidently provides a much more reproducible and controlled porous high surface area nanotexture. Further, since

5.1 Dye-Sensitized Regenerative Solar Cells

it is compatible with screen-printing technology, it anticipates future production requirements. While commercially available titania powders, produced by a pyrolysis route from a chloride precursor, have been successfully employed, the present optimized material is the result of a procedure described by Brooks and coworkers [17]. A speciﬁc advantage of the hydrothermal technique is the ease of control of the particle size and hence of the nanostructure and porosity of the resultant semiconductor substrate. The relevant preparation ﬂow diagram is given and the product is illustrated by the accompanying micrograph
(Fig. 4; Table 1). Figure 5 presents data on the control of substrate porosity by the powder-preparation parameters.
5.1.6

Dye Sensitization in Heterojunctions

Because the sensitizing dye itself does not provide a conducting functionality but is distributed at an interface in the form of immobilized molecular species, it is evident that for charge transfer, each molecule must be in intimate contact with both conducting phases. It is clear that this applies to the porous wide band gap semiconductor substrate into which the photoexcited chemisorbed molecules inject electrons. It is also evident that in the photoelectrochemical form of the sensitized cell, the liquid electrolyte penetrates into the porosity, thereby permitting the intimate contact with the charged dye molecule that is necessary for charge neutralization after

SEM image of the surface of a mesoporous ﬁlm prepared from the hydrothermal TiO2 colloid.

Fig. 4

the electron loss by exchange with the redox system in solution. In the analogous nonliquid electrolyte case, it is not immediately evident that an interpenetrating network of two conducting solids can be so easily established that an immobilized molecule at their interface can exchange charge carriers with both. However, initial results [18,19] are promising. In both cases, the charge transport materials are deposited by spin coating from the liquid phase in order to achieve the necessary intimate contact. In the latter case, a mixture of polymers was used, which phaseseparates spontaneously on removal of a solvent, whereas the Lausanne laboratory introduces a solution of the conducting compound into a previously sensitized nanostructure. The charge transfer material currently used is a spirobiﬂuorene, proprietary to Hoechst [20] as shown in
Fig. 6. As a matter of technical precision, if these materials function in the cell as a hole conductor, the device is a sensitized nanostructured heterojunction in which the charge transport is entirely of electronic nature (Fig. 7). However, if the molecules accept positive charge to become cations, for which there is initial spectroscopic evidence [18], the charge transfer mechanism within this organic phase can be considered a redox equilibration, and the device

100 nm

403

5 Dye-Sensitized Photoelectrochemistry
Tab. 1

Flow diagram for the preparation of TiO2 colloids and mesoporous ﬁlms [17]
Precursor preparation

Modify Ti-isopropoxide with acetic acid

Hydrolysis

Rapidly add precursor to water

Peptization

Acidify with HNO3, reflux
Autoclave: 12 hours at 230 °C

Hydrothermal growth
Concentrate colloid

Rotovap: 45 °C, 30 mbar

Solvent exchange,
Ethyl cellulose addition

Flocculation, centrifuging

3-roll mill, 15 minutes

Homogenize paste
Screen-print films

Anneal: 450 °C, 20 minutes

Dry and anneal films

9000
190 °C
210 °C
230 °C
250 °C
270 °C

8000

Differential volume
[(cm3/g)/nm] [10−5]

404

7000
6000
5000
4000
3000
2000
1000
0

0

10

20

30

40

50

60

70

80

Pore diameter
[nm]
Control of semiconductor substrate porosity by temperature of hydrothermal processing.

Fig. 5

5.1 Dye-Sensitized Regenerative Solar Cells
OCH3

N
OCH3

OCH3

OCH3
N
N

OCH3

OCH3

N
OCH3
OCH3

Fig. 6 Structure of spirobiﬂuorene conducting material. The bisﬂuorene structures are perpendicular, conjoined through a carbon site common to both.

is photoelectrochemical. This distinction may have practical implications.
5.1.7

Commercial Prospects

The status of the dye-sensitized device as the only veriﬁed alternative to solid-state junction devices has already been discussed. It must be recognized that the solid-state devices, particularly the silicon p -n junction cells beneﬁt from forty years of industrial and development experience, the technology transfer from the silicon-based electronics industry, and even the widespread availability of highquality silicon at low cost resulting from the expansion of that industry. The procedures for high-yield fabrication of silicon devices, both crystalline and amorphous, are well understood, with costing well established on the basis of decades of solid industrial experience. For dye-sensitized

cells, in contrast, fabrication procedures require development and costing based on estimates of the requirements of chemical processes rather than the silicon metallurgy with elevated temperatures and vacuum technology required for conventional cells. Equally, it is well known that the substitution of an established technology by an upcoming alternative requires that the new concept has deﬁnite advantages and no clear disadvantages. It is therefore noted with some satisfaction that several companies in Europe, Japan, and Australia have taken up the challenge and are currently engaged under license

cb

Inj.(D + /D*)

hν
Reg.

Charge transport by hopping in the dye-sensitized solid-state photovoltaic cell.

Hopping

(D + /D)

Fig. 7

TiO2

Dye

Au

405

406

5 Dye-Sensitized Photoelectrochemistry

in the venture of industrialization and commercialization of dye-sensitized photovoltaic cells.
5.1.8

Conclusion

The development of reproducible and stable photovoltaic devices adapted for manufacturing processes has proceeded in evolutionary steps, with each component optimized and veriﬁed for compatibility with system requirements. The dye-sensitized nanocrystalline solar cell, either the electrochemical device or the closely related sensitized heterojunction [18], provides a credible alternative to the conventional semiconductor junction solid-state cell. Time and the market will tell if it can compete successfully.
Acknowledgments

The team at EPFL greatly appreciates the vote of conﬁdence represented by the licenses taken up by the industry for this type of solar cell. In the present work, we do acknowledge the initiative and innovative spirit of our coworkers, past and present, in the areas of dye synthesis, catalysis and electrochemistry, and semiconductor thin ﬁlm fabrication.
References
1. A. -E. Becquerel, C. R. Acad. Sci. 1839, 9,
561.

2. R. W. Gurney, N. F. Mott, Proc. Roy. Soc. A
1938, 164, 151.
3. T. H. James, in The Theory of the Photographic
Process, 4th. Ed., MacMillan, New York,
1977.
4. W. West, Proc. Vogel Centenary Symp., Photogr. Sci. Eng. 1974, 18, 35.
5. J. Moser, Monatsh. Chem. 1887, 8, 373.
6. H. Rigollot, C. R. Acad. Sci. 1893, 116,
873.
7. S. Namba, Y. Hishiki, J. Phys. Chem. 1965,
69, 774.
8. R. C. Nelson, J. Phys. Chem. 1965, 69,
714.
9. J. Bourdon, J. Phys. Chem. 1965, 69, 705.
10. H. Gerischer, H. Tributsch, Ber. Bunsen-Ges.
Phys. Chem. 1968, 72, 437.
11. M. P. Dare-Edwards, J. B. Goodenough, A.
Hamnet et al., Faraday Discuss. Chem. Soc.
1980, 70, 285.
12. W. D. K. Clark, N. Sutin, J. Am. Chem. Soc.
1977, 99, 4676.
13. J. DeSilvestro, M. Gr¨tzel, L. Kavan et al., a J. Am. Chem. Soc. 1985, 107, 2988.
14. B. O’Regan, M. Gr¨tzel, Nature 1991, 335, a 737.
15. M. K. Nazeeruddin, P. Pechy, Th. Renouard et al., J. Am. Chem Soc. 2001, 123, 1613.
16. S. A. Hague, Y. Tachibana, R. Willis, J. E.
Moser, M. Gr¨tzel, D. R. Klug, J. R. Durrant, a J. Phys. Chem. B 2000, 104, 538.
17. K. G. Brooks, S. D. Burnside, V. Shklover et al., Ceramic Transactions 109: Processing and Characterisation of Electrochemical Materials and Devices, (Eds.: P. N. Kumta), American Ceramic Society, 2000, p. 115.
18. U. Bach, D. Lupo, P. Comte et al., Nature
1998, 395, 583.
19. J. J. M. Halls, C. A. Walsh, N. C. Greenham et al., Nature 1995, 376, 498.
20. D. Lupo, J. Salbeck, Intern. patent PCT/
EP96/03944.

5.2 Dyes for Semiconductor Sensitization

5.2

Dyes for Semiconductor Sensitization
..
Md. Khaja Nazeeruddin and Michael Gratzel
Swiss Federal Institute of Technology, Lausanne, Switzerland
5.2.1

Introduction
General Background
The extensive use of conventional energy supplies is causing many serious problems to the global environment. Moreover, sooner or later the fossil fuels are going to be exhausted and if nuclear power is not to be depended on solely, alternative methods for harnessing solar power, which is clean, nonhazardous, enormous, and ever lasting have to be developed.
There are numerous ways to convert the solar radiation directly into electrical power or chemical fuel [1]. However, the capital cost of such devices is not attractive for large-scale applications.
Dye-sensitized solar cell technology is an interesting, inexpensive, and promising alternative to those of traditional solid-state photovoltaics. The process of extending the sensitivity of transparent materials such as titanium dioxide (TiO2 ) to the visible spectra is called spectral sensitization. The topic of sensitization is very old [2–5] and without going into the history of the sensitization, the progress made during last 10 years on dye-sensitized solar cells is reviewed. The intention here is to restrict the discussion exclusively to inorganic dyes (transition metal complexes) based largely on our own work.
The photoelectrode of a dye-sensitized solar cell consists of 8–20 µm ﬁlm of mesoporous nanocrystalline TiO2 particles containing a monolayer of anchored
5.2.1.1

dye molecules, which absorb light, freeing electrons to carry electric current.
The low production costs, stability, and the efﬁciency of these cells has generated renewed interest in many groups around the world [6–15]. The added advantages of these cells are availability and nontoxicity of the main component that is TiO2 , which is even used in paints, cosmetics, and health care products like hip joints and other orthopedic implants. Efﬁcient sensitization of large band gap semiconductors to the visible and the near infrared (IR) solar spectrum can be achieved by engineering dyes at a molecular level, which is the topic of this chapter.
A dye-derivatized mesoporous titanium ﬁlm is one of the key components in dyesensitized solar cells. The electrochemical and photophysical properties of the ground and the excited state of the dye play an important role in the charge-transfer (CT) dynamics at the semiconductor interface
(Chapter 5.3 in Volume 6).
The photophysical and photochemical properties of group VIII metal complexes using polypyridyl ligands have been thoroughly investigated during the last three decades [16–20]. The main thrust behind these studies is to understand the energy and electron-transfer processes in the excited state and to apply this knowledge to potential practical applications such as dye-sensitized solar cells and light-driven information processing [21].
Particularly, ruthenium (Ru) (II) complexes have been used extensively as
CT sensitizers on nanocrystalline TiO2 ﬁlms [22]. The choice of ruthenium metal is of special interest for a number of reasons: (1) because of its octahedral geometry, one can introduce speciﬁc ligands in a controlled manner; (2) the photophysical, photochemical, and the electrochemical

407

408

5 Dye-Sensitized Photoelectrochemistry

properties of these complexes can be tuned in a predictable way; (3) it possesses stable and accessible oxidation states from I to IV; (4) it forms very inert bonds with imine nitrogen centers [23, 24].
Iron (Fe) that is in the ﬁrst row transition metal of the periodic table, and the same group as ruthenium, is an inexpensive and abundant metal.

However, the photophysical and the electrochemical properties of its coordination complexes are difﬁcult to tune in an expected fashion [25]. The other notable disadvantage of this metal is weaker ligand ﬁeld splitting compared to ruthenium and osmium (Os). On the other hand, osmium being in the third row transition metal of the periodic table has a

Load e− e−
S+/S∗

e−

e−

−1.0

cb e− e−

e−

(e−)

−0.6

e−

−0.2

∆v

0
(R /
TiO2

e−

R−)

0.2

Electrolyte
0.6

S+/S
1.0
TCO

TiO2

Counter electrode

V vs SCE

COOH

HOOC

N
N

=

N

Anchored dye
HOOC

NCS
Ru
NCS
N

HOOC

Operating principles and energy level diagram of dye-sensitized solar cell. S/S+ = Sensitizer in the ground and oxidized state;
S∗ /S+ = Sensitizer in the excited state R/R− = Redox mediator.
Fig. 1

5.2 Dyes for Semiconductor Sensitization

stronger ligand ﬁeld splitting compared to ruthenium. Moreover, the spin-orbit coupling in osmium complexes leads to enhanced spectral response in the red region [26]. Nevertheless, the low abundance of this metal restricts its use for large-scale applications. Operating Principles of the
Dye-Sensitized Solar Cell
The details of the operating principles of the dye-sensitized solar cell are given in
Fig. 1. The photo-excitation of the metalto-ligand charge transfer (MLCT) of the adsorbed dye leads to injection of electrons into the conduction band of the oxide. The original state of the dye is subsequently restored by electron donation from an electrolyte, containing the redox system (iodide/triiodide). The injected electron ﬂows through the semiconductor network to arrive at the back contact and then through the external load to the counter electrode. At the counter electrode, reduction of triiodide in turn regenerates iodide, which completes the circuit.

conduction band and the redox couple because of differences in recombination rates [28].
IPCE = photocurrent density
−2
(mA cm ) photon ﬂux wavelength (nm) ×
−2
(W m )
(1)

(1.25 × 103 ) ×

5.2.1.2

Incident Photon to Current
Efﬁciency and Open-Circuit Photovoltage
The incident monochromatic photon-tocurrent conversion efﬁciency (IPCE), deﬁned as the number of electrons generated by light in the external circuit divided by the number of incident photons as a function of excitation wavelength is expressed in the Eq. (1) [27]. The open-circuit photovoltage is determined by the energy difference between the fermi level of the solid under illumination and the
Nernst potential of the redox couple in the electrolyte (Fig. 1). However, the experimentally observed open-circuit potential using various transition metal complexes is smaller than the difference between the
5.2.1.3

5.2.2

Molecular Sensitizers
Formation of Complexes
One can view the formation of a complex between metal ion and ligands as an electrostatic attraction between positively charged metal and negatively charged ions or the negative ends of the dipoles of neutral ligands. A metal ion or atom can act as a discrete center about which a set of ligands is arranged in a deﬁnite way. The general concept of a ligand (unidentate) is donating of an electron pair to the central metal atom. The number of ligands per metal center is generally either four or six, with others being rare. The approach of the negative, or the neutral, ligands toward a charged metal repels electrons residing in d-orbitals and raises their energies both with respect to those residing on the ligands and the metal.
In an octahedral ligand ﬁeld, the ﬁve degenerate d-orbitals split into degenerate t2g (dxy , dxz , dyz ), and eg (dx2 −y2 , dz2 ) sets of orbitals. The splitting value increases by about 40% as one moves from the ﬁrst row (3d) to the second (4d) and third (5d) row transition metal ions [29]. A schematic representation for splitting of d-orbital and the position of ligand orbitals in an octahedral complex containing three bidentate 2,2 -bipyridyl ligands is shown in Fig. 2.
5.2.2.1

409

5 Dye-Sensitized Photoelectrochemistry

Energy

410

eg

∆Oh

d

Metal orbitals

π*

π*

π*

∆Oh

∆Oh

t2g

Fe2+

Ru2+

Fe2+ ,

Ru2+

Os2+

Os2+

A d-orbital splitting pattern for and in an octahedral ﬁeld. The relative eg levels with respect to t2g and ligand π ∗ -orbitals are shown arbitrarily. Fig. 2

Photophysical Properties
The photophysics and photochemistry of polypyridyl complexes of ruthenium can be understood with the aid of the energy diagram shown in Fig. 3. In these complexes there are three possible excited states: (1) metal centered (MC), which are due to promotion of an electron from t2g to eg orbitals; (2) ligand centered (LC) that are π – π ∗ transitions; and
(3) CT excited states, which are either
MLCT or ligand-to-metal (LMCT). An electronic transition from metal t2g orbitals to empty ligand orbitals without spinchange allowed, is called a singlet-singlet optical transition. The allowed transitions are identiﬁed by large extinction coefﬁcients. The transitions with spin change are termed singlet-triplet optical transitions, which are forbidden and are usually associated with a small extinction coefﬁcient. However, the excited singlet state may also undergo a spin ﬂip, resulting in an excited triplet state. This process
5.2.2.2

is called intersystem crossing (ISC). The possible deexcitation processes are radiative and nonradiative. The radiative decay of a singlet and triplet excited states are termed ﬂuorescence and phosphorescence, respectively.
The intense colors in 2,2 bipyridyl complexes of iron, ruthenium, and osmium are due to excitation of an electron from metal t2g orbitals to the empty π ∗ –orbitals of the conjugated 2,2 bipyridyl. The photoexcitation of this MLCT excited state can lead to emission. However, not all complexes are luminescent because of the different competing deactivation pathways. This aspect is beyond the scope of this chapter; the interested reader can refer to a number of publications on this subject [16–20]. The other potential deactivation pathways for the excited dye are donation of an electron (called oxidative quenching, Eq. 2) or the capture of an electron (reductive quenching, Eq. 3) or transfer of its energy to other molecules or

5.2 Dyes for Semiconductor Sensitization
MC
eg

LC

p

1

MLCT

π*
ISC

s

3MLCT

d

Em t2g LMCT

σ
Metal orbitals

Molecular orbitals

Ligand orbitals

Schematic presentation of a molecular orbital diagram for an octahedral d6 metal complex involving 2,2 -bipyridyl type ligands in which various possible transitions are indicated.

Fig. 3

process.

solvent (Eq. 4).
D∗ + Q − − D+ + Q−
−→
D∗ + Q − − D− + Q+
−→

(4)

Ground and Excited State Redox
Potentials
The ground state oxidation and reduction potentials of a complex can be obtained by cyclic voltammetric studies. An approximate value of the excited state redox potential can be extracted from the potentials of the ground state couples and the zero-zero excitation energy (E0 – 0 ) according to Eqs. (5) and (6). The zero-zero energy can be obtained from 77 K emission spectrum of the sensitizer [30]. The excited state redox potential of a sensitizer plays an important role in the sensitization
5.2.2.3

(5)

E(S+ /S∗ ) = E(S+ /S) − E0 – 0

(3)

−→
D∗ + Q − − D + Q∗

E(S∗ /S− ) = E(S/S− ) + E0 – 0

(2)

(6)

Requirements of the Sensitizers
The ideal sensitizer for a single junction photovoltaic cell should absorb all the solar photons below a wavelength of about
920 nm [31]. In addition, it should fulﬁll several demanding requirements: (1) It should possess suitable excited state redox properties; (2) It must ﬁrmly be grafted to the semiconductor oxide surface and inject electrons into the conduction band with a quantum yield of unity; (3) Its ground state redox potential should be sufﬁciently high that it can be regenerated rapidly via electron donation from the electrolyte or a hole conductor, (4) It should have a long-term photo and thermal stability to
5.2.2.4

411

5 Dye-Sensitized Photoelectrochemistry

sustain at least 108 redox turnovers under illumination corresponding to about 20 years of exposure to natural sunlight at Air
Mass (AM) 1.5.
On the basis of extensive screening of hundreds of ruthenium complexes, it was discovered that the sensitizer excited state oxidation potential should be at least – 0.9 V versus Saturated calomel reference electrode (SCE), to inject electrons efﬁciently onto TiO2 conduction band. The ground state oxidation potential should be about 0.5 V versus SCE, to be regenerated rapidly via electron donation from the electrolyte (iodide/triiodide redox system) or a hole conductor. A signiﬁcant decrease in electron injection efﬁciencies will occur if the excited and ground state redox potentials are lower than these values [32].

of ruthenium polypyridyl complexes can be tuned in two ways. First, by introducing a ligand with a low-lying π ∗ -molecular orbital and second, by the destabilization of the metal t2g orbitals with a strong donor ligand. The limiting conditions for an efﬁcient sensitizer makes these two extreme situations incompatible. Because, in the former case, the excited state is not sufﬁciently energetic to enable electron transfer into the conduction band of the TiO2 semiconductor, and in the latter case, the easily oxidizable complex cannot be reduced back by a suitable electron relay like I− .
To illustrate the tuning aspects of the MLCT transitions in ruthenium polypyridyl complexes, the well known
[RuL3 ] (L = 4,4 dicarboxylic acid-2,2 -bipyridine) type of complex can be considered. This complex shows strong visible band at 466 nm, because of CT transition from metal t2g highest occupied molecular orbitals (HOMO) to π ∗ -lowest unoccupied molecular orbitals (LUMO) of the ligand (Fig. 3). The Ru(II)/(III) oxidation potential is at 1.3 V, and the ligand based reduction potential is at −1.5 V

Absorption Spectral Properties of
Metal Complexes

5.2.2.5

5.2.2.5.1 Tuning of MLCT Transitions To absorb the light below 920 nm, the sensitizer’s redox level needs to be tuned. The
MLCT transitions and the redox properties

π *t

Ligand π*orbitals un

ing

∆E

Energy

412

∆E

∆E

ing

t 2g

tun

Metal t2g orbitals
[Ru(dcbpy)3]

[Ru(dcbiq)3]

[Ru(dcbpy)2(phpy)]

Tuning of HOMO (t2g ) and LUMO (π ∗ ) orbital energy in various ruthenium polypyridyl complexes.

Fig. 4

5.2 Dyes for Semiconductor Sensitization

versus SCE [33]. From spectro chemical and electrochemical studies of polypyridyl complexes of ruthenium, it has been concluded that the oxidation and reduction potentials are the best indicators of the energy levels of the HOMO and LUMO [34].
The energy between the metal t2g orbitals and π ∗ -orbitals can be reduced either by raising the energy of the t2g or by decreasing the energy of the π ∗ -orbitals with donor acceptor ligands, respectively
(Fig. 4). In the following sections, ways to tune HOMO and LUMO energy levels by introducing various ligands will be discussed. 5.2.2.5.2 Spectral Tuning in ‘‘Push-Pull’’
Type of Complexes The lowest energy
MLCT transition of ruthenium polypyridyl complexes of the type [RuL3 ], (L =
4,4 dicarboxylic acid-2,2 -bipyridine), can be lowered so that it absorbs more in the red region of the visible spectrum by substituting one of the bidentate ligand by a donor bidentate ligand (pushing ligand).
The remaining two electron-withdrawing bidentate ligands act as pull ligands. A list of complexes is included in Table 1, where the absorption maxima of the complexes

are tuned from 466 nm to 560 nm by introducing different donor ligands. The oxidation potential in these complexes were tuned from 1.3 V to 0.3 V versus SCE, going from [RuL3 ] to [RuL2 L ]
L = phenylpyridine). Thus, the energy of the HOMO is varied to an extent of 1 V in this series of complexes [35]. Meyer and coworkers have used the same concept to synthesize black absorbers by judiciously selecting push and pull ligands [36].
It is interesting to note the magnitude of the spectral shift for the lowest energy
CT transition is ≈0.5 eV, and the shift in the oxidation potential ≈1 eV. This clearly shows that not all the HOMO tuning is translated into the spectral shift of the complex. The apparent 0.5 eV, difference is involved in raising the energy of π ∗ orbitals of the pulling ligands caused by the pushing ligands [35].
5.2.2.5.3 Spectral Tuning in Complexes
Containing Hybrid Donor Ligands The absorption and luminescence spectral properties of complexes of the type [Ru(dmbip)
(dcbpy)(X)] and [Ru(dmbip)(dcbiq)(X)], where dmbip = 2,6-bis(1-methylbenzimidazol-2 -yl)pyridine, dcbpy = 4,4 -(COOH)2 -

Absorption, emission and redox properties of ruthenium complexes containing
‘‘push-pull’’ ligands

Tab. 1

Complex
[Ru(dcbpy)3 ]4−
[Ru(dcbpy)2 (dmg)]2−
[Ru(dcbpy)2 (Me2 bpy)]2−
[Ru(dcbpy)2 (2COOpy)]3−
[Ru(dcbpy)2 (bpe)]2−
[Ru(dcbpy)2 (dea-bpy)]2−
[Ru(dcbpy)2 (dh-phen)]2−
[Ru(dcbpy)2 (NCS)2 ]4−
[Ru(dcbpy)2 (phpy)2 ]3−

Abs. max.
(nm)

Emission max.
(nm)

V vs SCE

Reference

466
470
474
490
500
506
516
518
560

640
665
655
712
716
732
760
750
865

1.3
1.2
1.2
0.80
0.78
0.60
–
0.60
0.3

33
35
35
35
35
35
35
22
35

413

414

5 Dye-Sensitized Photoelectrochemistry
Absorption, emission, and redox properties of ruthenium complexes containing ‘‘hybrid donor’’ ligands

Tab. 2

Complex

[Ru(dmbip)(dcbpyH)(Cl)]
[Ru(dmbip)(dcbpy)(H2 O)]
[Ru(dmbip)(dcbpy)(OH)]
[Ru(dmbip)(dcbpy)(NCS)]
[Ru(dmbip)(dcbpy)(CN)]
[Ru(dmbip)(dcbpy)(N(CN)2 )]
[Ru(dmbip)(dcbpy)(Cl3 -pcyd)]
[Ru(dmbip)(dcbpy)(NCN)]
[Ru(dmbip)(5,5dcbpy)(NCN)]
[Ru(dmbip)(dcbiqH2 )(NCS)]
[Ru(dmbip)(dcbiqH2 )(Cl)] Cl

Abs. max.
(nm)
516
508
528
502
486
492
502
528
542
580
608

2,2 -bipyridine, dcbiq = 4,4 -(COOH)2 -2,2 biquinoline and X = Cl− , NCS− , CN− ,
N(CN)2 − , NCN− , and H2 O are summarized in Table 2. The 2,6-bis(1-methylbenzimidazol-2 -yl)pyridine ligand is a hybrid ligand that contains donor groups
(benzimidazol-2 -yl) and an acceptor group
(pyridine). The complexes containing these ligands show intense ultraviolet
(UV) bands at 362 and 346 due to intraligand π – π ∗ transition of 2,6-bis
(1-methylbenzimidazol-2 -yl)pyridine. The bands at 308 and 244 nm are because of the π – π ∗ transition of dcbpy ligand. On acidiﬁcation, the 308 nm band shifts to
314 nm as a result of the protonation of the carboxylate groups, whereas the 362 and 342 nm bands remain unchanged in acidic and basic pH.
The MLCT bands of these complexes are broad and red shifted by approximately
140 nm, compared to [Ru(dcbpy)3 ]. The lowest-energy MLCT transitions within this series were shifted from 486 to
608 nm, and the HOMO level varied over an extent of 0.45 V versus SCE. The energy of the MLCT transition in these complexes decreases with the decrease in

Emission max.
(nm)
–
780
800
770
750
760
–
–
–
900
>950

V vs SCE

Reference

0.67
0.71
–
0.80
0.92
0.60
0.74
–
–
1.15
0.83

32
32
32
32
32
38
38
39
38
32
32

the π -acceptor strength of the ancillary ligand, CN− , NCS− , H2 O, NCN− Cl− .
The red shift of the absorption maxima in complexes containing dcbiq compared to dcbpy acceptor ligand is because of the low π ∗ -orbitals of biquinoline.
The resonance Raman spectra of these complexes for excitation at 568 nm show bands predominantly associated with the dcbpy and dcbiq ligands indicating that the lowest excited state is a metal to dcbpy or dcbiq ligand CT state [32, 37, 38].
5.2.2.5.4 Inﬂuence of Nonchromophoric
Ligands on MLCT Transitions This section compares spectral response of the three complexes of the type [Ru(P-terpy)(Me2 bpy)(NCS)] 1, (P-terpy = 4-phosphonato-2,
2 : 6 ,2 -terpyridine, Me2 bpy] = 4,4 -di[Ru(H2 dcbpy)2 methyl-2,2 -bipyridine),
(NCS)2 ] 2, (H2 dcbpy = 4,4 -dicarboxy-2,
2 -bipyridine) and (TBA)2 [RuL(NCS)3 ] 3,
(L = 4,4 ,4 -tricarboxy-2,2 : 6 ,2 -terpyridine) where the number of nonchromophoric ligands such as NCS− are varied from one to three. The absorption and luminescence spectral properties of complex 1 – 3 are summarized in Table 3.

5.2 Dyes for Semiconductor Sensitization
Tab. 3

Absorption, emission, and redox properties of ruthenium complexes containing NCS

ligands
Complex

[Ru(pterpy)(Me2 bpy)(NCS)]
[Ru(tcterpy)(Me2 bpy)(NCS)]
[Ru(H2 dcbpy)2 (NCS)2 ]
[Ru(dcbpy)(Me2 bpy)2 (NCS)2 ]
[Ru(terpy)(NCS)3 ]
[Ru(pterpy)(NCS)3 ]
[Ru(tcterpy)(NCS)3 ]

Abs. max.
(nm)

Emission max.
(nm)

V vs SCE

Reference

500
510
535
527
580
580
620

780
800
830
760
820
820
950

0.86
0.94
0.85
0.78
0.54
0.52
0.60

39
39
22
39
40
43
40

The absorption maxima are listed for the intense lowest-energy MLCT bands in ethanol, although the spectra possess additional absorption features at lower and higher energies.
The interesting feature of these complexes is a broad MLCT absorption bands in the visible region. The most intense
MLCT transition maxima of complexes 1,
2, and 3 are at 500, 535, and 620 nm, respectively. The 120 nm red shift of 3 compared to 1 is the result of an increase in the energy of the metal t2g orbitals caused by introducing nonchromophoric ligands and a slight decrease in the LUMO of the 4,4 ,4 -tricarboxy2,2 : 6 ,2 -terpyridine. The electrochemical data are consistent with the above assignments [22, 39, 40].
5.2.2.5.5 Inﬂuence of the Position of Carboxyl Groups on MLCT Transitions Ruthenium polypyridyl compounds, of the type
[RuL2 (X)2 ], (where L = 4,4 -dicarboxylic acid-2,2 -bipyridine, 5,5 -dicarboxylic acid2,2 -bipyridine, and 6,6 -dicarboxylic acid2,2 -bipyridine; X = Cl− , CN− and NCS− ) have been reported [41–43]. The lowest
MLCT absorption maxima of [RuL2 (NCS)2 ] complex (L = 2,2 -bipyridine) is seen at
510 nm in ethanol. By substituting two

carboxylic acid groups at the 4,4 -position of 2,2-bipyridine ligand the MLCT maxima red shifted to 535 nm. However, substitution at the 5,5 -position of 2,2 -bipyridine shifted the maxima further into the red,
580 nm. On the contrary, the MLCT maxima blue shifted (500 nm) by substituting at the 6,6 positions of 2,2 -bipyridine.
The enhanced red response of complexes containing the 5,5 -dicarboxylic acid-2,2 -bipyridine is because of a decrease in the energy of the π ∗ -orbitals, which makes them attractive as sensitizers for nanocrystalline TiO2 ﬁlms. Bignozzi and coworkers found that the IPCE of complexes having the 5,5 -dicarboxylic acid2,2 -bipyridine ligands were lower than the analogous complexes that contain 4,4 dicarboxylic acid-2,2 -bipyridine [43]. They rationalized the low efﬁciency of the sensitizers containing 5,5 -dicarboxylic acid2,2 -bipyridine ligands in terms of low excited state redox potentials.
5.2.2.5.6 MLCT Transitions in Geometric
Isomers Isomerization is another exciting approach toward tuning the spectral properties of metal complexes (Fig. 5)
[44–46]. The electronic properties of cisand trans-isomers of [RuL2 (X)2 ], (L = 4,4 dicarboxylic acid-2,2 -bipyridine; X = Cl− ,

415

5 Dye-Sensitized Photoelectrochemistry
0.8

COOH
COOH

HOOC
HOOC

N

0.6

Absorbance

416

N

N

NCS

N
Ru

SCN
NCS

Ru
N

N

HOOC

N
NCS
N

0.4
COOH

HOOC
HOOC
0.2

0.0
300

400

500

600

700

800

Energy
[nm]
UV-visible absorption spectra of cis-, transbis(2,2 -bipyridine-4,4 -dicarboxylic acid) ruthenium (II) (NCS)2 complexes.

Fig. 5

Electronic Spectral Data of the cis- and trans-bis(4,4 -dicarboxylic acid-2,2 -bipyridine)Ruthenium (X2 ) Complexes (X = Cl− , H2 O, NCS− ) in DMF

Tab. 4

Complex

cis-[Ru(H2 dcbpy)2 (Cl)2 ] cis-[Ru(H2 dcbpy)2 (H2 O)2 ]a cis-[Ru(H2 dcbpy)2 (NCS)2 ] trans-[Ru(H2 dcbpy)2 (Cl)2 ] trans-[Ru(H2 dcbpy)2 (H2 O)2 ] trans-[Ru(H2 dcbpy)2 (NCS)2 ] cis-[Ru(bpy)2 (H2 O)2 ] trans-[Ru(bpy)2 (H2 O)2 ]

Abs. max.
(nm)

Emission max.
(nm)

Reference

590
500
535
690
530
581
480
495

847
750
830
>900b
800b
870
660
700

47
44
47
47
47
47
45
46

DMF, dimethylformamide. a in water at pH 4.5. b Weak and broad emission maximum.

H2 O and NCS− ) are collected in Table 4.
The UV-visible absorption spectrum of the trans-dichloro complex in DMF solution shows at least three MLCT absorption bands in the visible region at 690, 592,

and 440 nm. On the other hand, the cis-dichloro complex in DMF solution shows only two distinct broadbands in the visible region at 590 and 434 nm, which was assigned to a MLCT transitions.

5.2 Dyes for Semiconductor Sensitization

The lowest energy MLCT band in the trans-dichloro complex is signiﬁcantly red shifted compared to the cis-dichloro complex. This red shift is due to stabilization of the LUMO of the dcbpy ligand in the trans-species relative to the cis-species [47].
The red shift (108 nm) of the lowest energy MLCT absorption in the spectrum of the trans-dichloro complex compared to the spectrum of the trans-dithiocyanato is due to the strong σ donor property of the
Cl− compared to the NCS− ligand. The chloride ligands cause destabilization of the metal t2g orbitals, and raising them in energy relatively closer to the ligand π ∗ -orbitals result in a lower energy MLCT transition. According to Lever’s electrochemical parameterization, the chloride ligand has a parameter value of −0.24 and the thiocyanate ligand has a value of −0.06.
The electrochemical and the absorption data are in good agreement with reported parameterization scale for ligands [48].
5.2.2.5.7 Spectral Tuning in Heteroleptic
Sensitizers
The synthesis and characterization of heteroleptic complexes of ruthenium
[Ru(L1 )(L2 )(L3 )] starting from a polymeric complex Ru(CO)2 Cl2 have been reported [49–51]. The synthetic methodology of such heteroleptic complexes involves several steps, resulting in very low yields. Recently, a novel synthetic route has been developed to introduce different ligands on a ruthenium (II) precursor, namely, the RuCl2 (DMSO)4 complex, which is a versatile and well-characterized starting material [52]. A range of mixed ligand sensitizers of the type [Ru(L1 )(L2 )
(L3 )] have been prepared and their spectroscopic, luminescent, and redox properties are reported [53, 54]. The absorption spectral data of these complexes show a red shift with increasing donor strength of

the L3 ligands. In these complexes, L3 being the dithiocarbamate ligand, shifts the
Ru(II) t2g level upward without signiﬁcantly affecting the π ∗ -level of the dcbpy ligand thereby increasing the MLCT absorption in the red portion of the visible region. 5.2.3

Hydrophobic Sensitizers

The other important aspect in dyesensitized solar cells is water-induced desorption of the sensitizer from the surface. Extensive efforts have been made in our laboratory to overcome this problem by introducing hydrophobic properties in the ligand. The heteroleptic complexes containing hydrophobic ligands of the type [Ru(dcbpy)(mhdbpy)(NCS)2 ] 1,
[Ru(dcbpy)(dtdbpy)(NCS)2 ] 2 [Ru(dcbpy)
(mddbpy)(NCS)2 ] 3 (dcbpy = 4,4 -dicarboxy-2,2 -bipyridine, mhdbpy = 4-methyl4 -hexadecyl-2,2 -bipyridine and dtdbpy =
4,4 -ditridecyl-2,2 -bipyridine, mddbpy =
4-methyl-4 -didodecyl-2,2 bipyridine) have been synthesized (Fig. 6). The photocurrent action spectra of these complexes show broad features covering a large part of visible spectrum and displays a maxima around 550 nm, where the incident monochromatic IPCE exceeds 80%.
The performance of these hydrophobic complexes as CT photosensitizers in nanocrystalline TiO2 -based solar cell shows excellent stability toward waterinduced desorption [55].
The rate of electron transport in dyesensitized solar cell is a major element of the overall efﬁciency of the cells. The electrons injected into the conduction band from optically excited dye, can traverse the
TiO2 network and can be collected at the transparent conducting glass or can react

417

418

5 Dye-Sensitized Photoelectrochemistry

Complex 1

Complex 2

Complex 3

Structure of complexes [Ru(dcbpy)(dtdbpy)(NCS)2 ] 1,
[Ru(dcbpy)(mhdbpy)(NCS)2 ] 2, and [Ru(dcbpy)(mddbpy)(NCS)2 ] 3.

Fig. 6

with a redox mediator. The reaction of injected electrons into the conduction band with a redox mediator gives undesirable dark currents, signiﬁcantly reducing the charge-collection efﬁciency, and thereby decreasing the total efﬁciency of the cell.
This problem is drastically suppressed by using hydrophobic sensitizers of the type shown in Fig. 6. The most likely explanation for the dark-current suppression is that the long chains of the sensitizer interacts laterally to form an aliphatic network, thereby preventing triiodide from reaching the TiO2 surface.
5.2.4

current conversion yields remaining under 1%, which is insufﬁcient for solar cell applications [57–60]. One of the reasons for such low efﬁciencies is aggregation of the dye on the TiO2 surface. This association often leads to undesirable photophysical properties such as self-quenching and excited state annihilation. However, the advantage of this class of complexes is the near IR response which is very strong, having extinction coefﬁcients of close to 50 000 M−1 cm−1 at 650 nm compared to the polypyridyl complexes, which have small extinction coefﬁcients at this wavelength. Near IR Sensitizers
Ruthenium Phthalocyanines
The bis(3,4-dicarboxypyridine)Ru(II)1,4,8,
11,15,18,22,25-octamethyl-phthalocyanine
(JM3306) shows a visible absorption band at 650 nm (ε 49 000 M−1 cm−1 ) and a phosphorescence band located at 895 nm.
The triplet state lifetime is 474 ns under anaerobic conditions. The emission is entirely quenched when JM3306 is adsorbed onto a nanocrystalline TiO2 ﬁlm.
5.2.4.2

Phthalocyanines
Phthalocyanines possess intense absorption bands in the near IR region and are known for their excellent stability rendering them attractive for photovoltaic applications [56]. They have been repeatedly tested in the past as sensitizers of wide band gap oxide semiconductors and obtained poor incident photon to electric
5.2.4.1

5.2 Dyes for Semiconductor Sensitization

The very efﬁcient quenching of the emission of JM3306 was found to be due to the electron injection from the excited singlet/triplet state of the phthalocyanine into the conduction band of the TiO2 [61]. The photocurrent action spectrum is shown in
Fig. 7, where the incident IPCE is plotted as a function of wavelength. The feature is extending well into the near IR region displaying a maximum at around 660 nm where the IPCE exceeds 60%. These are by far the highest conversion efﬁciencies obtained with the phthalocyanine type sensitizers. It is fascinating to note that this class of dyes inject efﬁciently into the conduction band of TiO2 , despite the fact that the pyridyl orbitals do not participate in

the π – π ∗ -excitation which is responsible for the 650 nm absorption band. This phenomenon shows that the electronic coupling of the excited state of the dye to the Ti (3d) conduction band manifold is strong enough through this mode of attachment to render charge injection very efﬁcient. These results establish a new pathway for grafting dyes to oxide surfaces through axially attached pyridine ligands.
Phthalocyanines Containing 3d
Metals
Several zinc (II) and aluminum (III) phthalocyanines substituted by carboxylic acid and sulfonic acid groups were anchored to nanocrystalline TiO2 ﬁlms and tested for their photovoltaic
5.2.4.3

100

80

IPCE
[%]

60

40

20

0
400

500

600

700

800

Wavelength
[nm]
Photocurrent action spectra of nanocrystalline TiO2 ﬁlms sensitized by bis(3,4-dicarboxypyridine)Ru(II)
1,4,8,11,15,18,22,25-octamethyl-phthalocyanin. The incident
IPCE is plotted as a function of wavelength.
Fig. 7

419

420

5 Dye-Sensitized Photoelectrochemistry

behavior using LiI/LiI3 /propylene carbonate electrolyte [62]. Interestingly, zinc (II) 2,9,16,23-tetracarboxyphthalocyanine exhibited 45% monochromatic current conversion efﬁciency at 700 nm. It is shown that electron injection to TiO2 occurs from the excited singlet state of the phthalocyanine derivatives. The inherent problem of phthalocyanine aggregation is reduced considerably in 3d metal phthalocyanines by introducing 4tert-butylpyridine and 3α ,7α -dihydroxy-5β cholic acid (cheno) into the dye solution.
The added 4-tert-butylpyridine in all likelihood coordinates to the metal in the axial position and thereby prevents aggregation of the phthalocyanines.
This type of sensitizers opens up new avenues for improving the near IR response of nanocrystalline injection solar cell. In addition, important applications can be foreseen for the development of photovoltaic windows transmitting part of the visible light. Such devices would remain transparent to the eye, while absorbing enough solar energy photons in the near
IR to render efﬁciencies acceptable for practical applications.
5.2.5

Mononuclear and Polynuclear Metal
Complexes of Group VIII
Iron Complexes
The photoconversion efﬁciency for iron complexes [FeL3 ] and [FeL2 (X)2 ], (L =
4,4 -dicarboxylic acid-2,2 -bipyridine and
5,5 -dicarboxylic acid-2,2 -bipyridine) has been reported [63]. The [FeL2 (X)2 ] type complexes show two broad MLCT bands in the visible region. In both cases, the photoconversion efﬁciency of the high-energy band was reported in the range of 40%. Nonetheless, there is virtually no photocurrent contribution
5.2.5.1

from the lower energy MLCT band of the complexes containing 5,5 -dicarboxylic acid-2,2 -bipyridine ligand because of lack of driving force for electron injection.
Osmium Complexes
Although there are a large number of osmium polypyridyl complexes, very few have been used as sensitizers in dyesensitized solar cells. The osmium complexes have several advantages compared to their ruthenium analogues: (1) osmium has a stronger ligand ﬁeld splitting compared to ruthenium, (2) the spin-orbit coupling leads to excellent response in the red region because of enhanced singlettriplet CT transitions [26]. Heimer and coworkers have synthesized [OsL2 (CN)2 ] complex (L = 4, 4 -dicarboxylic acid-2,2 bipyridine) and found that the complex is extremely stable under irradiation in a homogeneous aqueous solution compared to the analogous ruthenium complex [64].
The greater photo-stability for osmium is consistent with a stronger crystal ﬁeld splitting of metal d-orbital that inhibits efﬁcient population of d-d states. However, a factor of 10 lower injection efﬁciency for the osmium complex when anchored on
TiO2 electrode was reported compared to the ruthenium analog and rationalized in terms of slow regeneration rates for the oxidized dye [65].
Lewis and coworkers have synthesized and characterized two osmium complexes [OsL2 (CN)2 ] and [OsL3 ] (L = 4,4 dicarboxylic acid-2,2 -bipyridine). In their studies they found nearly 80% incident monochromatic IPCE [66].
5.2.5.2

Polynuclear Complexes
Polynuclear systems in which more than one complex is connected by a bridging ligand have been studied in several laboratories [67–72]. Here studies have been
5.2.5.3

5.2 Dyes for Semiconductor Sensitization

restricted exclusively to the cyano-bridged
(CN-bridged) trinuclear ruthenium complexes of the type shown in Fig. 8. Scandola and coworkers have synthesized and characterized several of these complexes with respect to their photophysical and electrochemical properties [73]. In such complexes, one can invoke possible antenna effects in the light harvesting process by introducing various fragments in which the photophysical properties are tuned to have directionality in the excited state. The presence of numerous metal centers should allow for different tailored fragments containing various ligands [72, 74].
Potential application of these complexes as photosensitizers for light energy harvesting relies heavily on demanding synthetic schemes. The CN-bridged trinuclear complex 1 (Fig. 8) has been found to be an excellent photosensitizer for spectral sensitization of the wide band gap semiconductor TiO2 [75]. Among the possible MLCT excited states associated

with the three metal centers, the one with lowest energy is based on the central unit carrying the ligand 4,4 -dicarboxy2,2 -bipyridine (dcbpy). When coated on a TiO2 -electrode, the complex anchors through the carboxyl unit of this central unit. Photons absorbed by the higher energy peripheral units efﬁciently transfer excitation energy to this central unit, accounting for its excellent performance.
To understand the principles of operation of polynuclear complexes and possibly improve their performance, the ‘‘tunability’’ of photophysical and redox properties has been examined in a number of such complexes. These complexes are composed of polypyridyl units of Ru and
Os as chromophores and cyanide as the bridging ligand [76]. By appropriate synthetic procedures, it is possible to control the mode of coordination of the bridging cyanide to a given chromophoric unit as a nitrile (C-bonded) or isonitrile (Nbonded) [74–77]. By varying the nature of
R
COOH
R

R

N
HOOC

N

N
N
N

HOOC

Ru
C

C

N
Ru
N

N

N
C

N
N

N
R

C

R

N
Ru

HOOC
R

N

N
N
R
R

1 : R = H ; 2 : R = COOH ; 3 : R = CH3 ; 4 : R = C6H5
5 = [(CN)(bby)2 Os-CN-Ru(dcbpy)2-NC-Os(bpy)2(CN)]

Structure of CN-bridge trinuclear ruthenium complex .

Fig. 8

6 = [(H2O)(bby)2 Ru-CN-Ru(dcbpy)2-NC-Ru(bpy)2(H 2O)]

421

422

5 Dye-Sensitized Photoelectrochemistry

the metal center, the mode of coordination of the bridging ligand, and the spectator ligands, it is possible to vary the location and the direction of excited state electron transfer process.
5.2.6

Surface Chelation of Polypyridyl Complexes onto the TiO2 Surface
Anchoring Groups
The functional groups serve as grafting agents for the oxide surface of the TiO2 ﬁlms. The grafting of polypyridyl complexes onto the oxide surface, which allows for electronic communication between
5.2.6.1

HOOC

the complex and the substrate, is an important target in dye-sensitized solar cells. Several ruthenium complexes containing carboxylic acid and phosphonated polypyridyl ligands are described [78–83].
The carboxylic and phosphonate functionality serves as an anchoring group to immobilize the complex on the nanocrystalline TiO2 ﬁlms. The immobilized sensitizer absorbs a photon to produce an excited state, which efﬁciently transfers its electron onto the TiO2 conduction band. To achieve high quantum yields of the excited state electron-transfer process the dye ideally needs to be in intimate contact with the semiconductor

COOH
HOOC
N

COOH

N

N

4,4′-dicarboxyl-2,2′-bipyridine
HOOC

N

5,5′-dicarboxyl-2,2 ′-bipyridine

COOH
COOH
HOOC

COOH
N
N

N

N

N

4,4′-(paradicarboxyl)diphenyl-2,2 ′-bipyridine

4,4′,4′′-tricarboxylic-2,2′;6,2′′-terpyridine

P(O)(OH)2
P(O)(OH)2

P(O)(OH)2
N
N

N

4,phosphonate-2,2′;6,2′′-terpyridine
HOOC

4,4′-diphosphonate-2,2′;bipyridine

COOH
HOOC
N

COOH

N
N

4,4′-dicarboxyl-2,2′-biquinoline
Fig. 9

N

4,7-dicarboxyl-1,10′-phenanthroline

Structure of the ligands containing anchoring groups.

5.2 Dyes for Semiconductor Sensitization

surface. The ruthenium complexes that have carboxylic acid and phosphonic acid groups show electron-transfer processes in the range of 80 to 90%. The near quantitative electron injection efﬁciency indicate a close overlap of the ligand π ∗ -orbitals and the titanium 3d orbitals [84].
Figure 9 shows the polypyridine ligands with different anchoring groups that have been employed for surface derivation of ruthenium complexes. Goodenough and coworkers have derivatized a mixed ligand complex [Ru(bpy)2 (dcbpy)] on TiO2 and SnO2 ﬁlms in the presence of dicyclohexylcarbodimide [84].
The interaction between the adsorbed sensitizer and the semiconductor surface has been addressed using resonance Raman and Fourier transform infrared (FT-IR) spectroscopy. The carboxylic acid functional group could adsorb on the surface in a unidentate, bidentate, or bridging fashion. Yanagida and coworkers concluded that the sensitizer cis-dithiocyanato bis(2,2 -bipyridine4,4 -dicarboxylate) ruthenium (II) complex
(N3) binds to the surface using ester-like and chelating linkages [85]. Woolfrey and coworkers have reported that the N3 complex anchor on the surface of TiO2 as a bidentate or bridging mode using two carboxylate groups per dye [86]. However,
Fillinger and Parkinson studied the adsorption behaviour of the N3 sensitizer and found that the initial binding involves one carboxylate, with subsequent binding of two or more carboxylate groups on the surface [87].
It is noteworthy that, Bignozzi and coworkers achieved sensitization of TiO2 with a cyanobridged dimer of Ru-Re polypyridyl complex where the coupling of the MLCT exited state to the conduction band manifold was through space and did not involve the anchoring group [88].

Acid-Base Equilibria of the
Anchoring Groups
The grafting properties of complexes containing the anchoring group onto the TiO2 surface and the factors responsible for differences in the attachment of various anchoring groups are essential for the design of novel sensitizers [89]. The difference in binding properties of the complexes containing carboxyl and phosphonate anchoring groups may stem from the differences in the pKa values of the ligands.
Knowledge of the pKa values of ionizable groups may aid in indicating the possible nature of the attachment. The acid base properties of transition metal complexes containing ligands with protonatable functional groups can provide important information about the nature of binding properties on the oxide surface.
5.2.6.2

Acid-Base Equilibria of the
4,4-Dicarboxy-2,2 -Bipyridine and its
Complexes
The ground state pKa values of
[Ru(dcbpy)2 (NCS)2 ] (N3) were determined from the relationship between the change in the optical density or the peak maximum with the pH for a given wavelength [90].
When acid is added to an alkaline solution of N3, changes in the electronic spectrum occur as shown in Fig. 10. The plot of λmax change versus pH for N3 shows the expected sigmoidal shape, with the pH at the inﬂection point giving two ground state pKa values at 3 and 1.5 ± 0.1.
In a related complex of the type
[Ru(bpy)2 (dcbpy)], two inﬂection points at pH 2.7 and 0
High OD, V = 0
Low OD, V = 0
Low OD, V > 0

0.2
0.1
0
0

2

4
Depth
[µm]

6

8

Illustration of electron density satisfying the general diffusion equation for different optical densities and applied biases. It is assumed that light is incident from the negative × direction. In the diffusion picture the electron density is always depleted near to the interface with the TCO at x = 0, providing a statistical driving force for electron collection.
Fig. 7

where nV (0) is the electron concentration in the dark when a bias V is applied dn across the terminals, and dx = 0 at the extreme semiconductor–electrolyte interface. Equation (29) is more complex than the case of electrons in the p -type region of a p–n junction solar cell (Eq. 22) because of the generally nonlinear behavior of the recombination and diffusion, and the failure of the Boltzmann approximation. The diffusion coefﬁcient is expressed as a function of n(x) to reﬂect this. Some approaches to this problem are reviewed later. Elucidation of the mechanisms of recombination and diffusion to provide appropriate functional forms for Dn and
R is a major experimental challenge, and is discussed in the following section.
5.3.5

the most relevant to photovoltaic device performance, but time and frequency resolved measurements, usually on model systems, are the most revealing about the nature of charge transport.
DC Measurements
In the steady state the left-hand side of Eq. (13) is equal to 0. Measurables are the steady state photocurrent, photovoltage, light and dark current voltage characteristics, and the quantum efﬁciency spectrum (or IPCE). If the physical origin of Jn is known, the dependence of such
DC measurements on variations in intensity, wavelength, and bias, deliver the parameters controlling Jn (for example, the values of the diffusion length and diffusion constant in the case of diffusion limited transport).
5.3.5.1

Experimental Techniques and Results

We divide these into three groups: DC, time domain, and frequency domain.
DC measurements under illumination are

5.3.5.1.1 Incident Photon to Current
Efﬁciency IPCE, also called external quantum efﬁciency, is the ratio of electrons collected at the terminals to incident

451

5 Dye-Sensitized Photoelectrochemistry

photons under monochromatic illumination at short circuit. A high IPCE is a prerequisite for a useful photovoltaic device. At short circuit, recombination can usually be neglected and the continuity equation becomes, in the one-dimensional geometry, 1 dJn (E, x)
+ g(E, x) = 0 q dx

(30)

for light of a given photon energy, E . IPCE is deﬁned as η(E) =

1 Jn (E, 0) q b0 (E)

(31)

where Jn (E, 0) is the current passing through the SnO2 interface at x = 0. Because the recombination rate, R , does not enter, this is a direct probe of the current mechanism. It is convenient to think of the IPCE as the product of the probabilities of (1) photon absorption, (2) electron injection, and (3) electron transport to the terminals. Because (1) is a known function

Incident photon to current efficiency

452

of device thickness, sensitizer, and substrate and (2) is near unit efﬁcient, (3) is the key quantity probed. We will refer to this probability as ηcoll , the quantum efﬁciency for charge collection. IPCE’s of around 80% at the absorption maximum of the dye have been reported for several Ru based sensitizer dyes on ∼10 µm thick nanocrystalline TiO2 ﬁlms [14, 96,
97]. These are illustrated in Fig. 8.
Such high IPCE’s correspond to almost
100% efﬁcient collection because around
20% of the light is lost by absorption in the
SnO2 layer, by reﬂection and scattering.
In terms of a simple diffusion model of electron transport, this indicates that the diffusion length of the electron is comparable to the optical depth of the cell at that wavelength. The very high
IPCE’s ﬁrst reported for dye-sensitized nanocrystalline ﬁlms were a surprising and promising feature.
More information about the diffusion mechanism can be extracted by studying the IPCE as a function of direction of

Rul′(NCS)3
60
Rul2(NCS)2
Rul2[Ru(bpy)2(CN)2]2
40

TiO2

20

0

L = 2,2′ − dcbpy
L′ = tc − terpy

RuL3

400

600

800

1000

Wavelength
[nm]
IPCE’s of some different sensitizer dyes, including the most widely used Ru(bpy)2 (NCS)2 , and the recently developed ‘‘black’’ dye,
..
developed by Prof. Gratzel’s group at EPFL [97].
Fig. 8

5.3 Charge Transport in Dye-sensitized Systems

illumination. In normal operation the device is illuminated from the glass substrate side, so that light is absorbed preferentially close to the SnO2 : TiO2 interface. If the diffusion length is comparable with the absorption depth and collection at the interface is easy, then high quantum efﬁciencies are expected for illumination from this direction. For illumination from the rear side (electrolyte side) of a ﬁlm which is optically thick, most electrons have to cross the ﬁlm to be collected, and study of the IPCE as a function of ﬁlm thickness provides information on the electron diffusion length. Sodergren and coworkers [21,
98] have applied the diffusion solutions to ﬁt the IPCE of a non-dye-sensitized device for illumination from electrolyte side and substrate side, and derive an electron diffusion length of 0.8 µm. Their results also conﬁrm, importantly, that electrons generated close to the SnO2 : TiO2 interface are collected most efﬁciently, indicating the presence of an electric ﬁeld driving collection at that interface. Measurements by Lindstrom and coworkers indicate that the electron diffusion length is longer in dye-sensitized devices [95], as would be expected from the absence of photogenerated holes. The IPCE at the optimum wavelength can be interpreted as an indicator of the relative efﬁciency of electron transport in variations of the dye-sensitized TiO2 device. For instance, the addition of electron acceptors such as O2 and I2 to the electrolyte lead to reduced IPCE at all wavelengths [95], which can be interpreted in terms of a shorter electron lifetime or shorter diffusion length.
Under forward bias (electrolyte at a more negative potential relative to the electrode potential) the electron density in the ﬁlm increases and recombination becomes signiﬁcant. At a bias that corresponds to

the operating bias of the solar cell, IPCE at peak wavelength falls to about 45% of its saturation value [93]. This indicates competition between recombination and transport, although it is not trivial to disentangle the contributions from Jn and R .
The IPCE for sensitized ZnO ﬁlms is smaller [64], although it is not clear whether this is because of poorer electron transport or poorer electron collection at the interface with substrate [67].
Dark-current Voltage In the dark, G = 0 and the steady state continuity
Eq. (13) becomes

5.3.5.1.2

1 d Jn
−R =0 q dx

(32)

where R depends (usually strongly) on the applied bias V . The dark current
Jdark (V ) is the current passing though x = 0 when a forward bias V is applied between the terminals. Note that Jdark acts in the opposite sense to the short circuit photocurrent. It is normally this process which limits the performance of a solar cell. Analysis of Jdark (V ) can in principle teach us something about R .
In a conventional solar cell, R results from the recombination of injected minority carriers. The dark current is rectifying and can be approximated by the form
Jdark = J0 (eqV /mkT − 1)

(33)

where J0 is a constant and the ideality factor m normally lies in the range 1 ≤ m ≤
2. m and J0 are related to the mechanism for charge transport and recombination.
The same parameters have been used to characterize dark currents of dyesensitized cells, although their meaning is not clear.

453

5 Dye-Sensitized Photoelectrochemistry

In a dye-sensitized cell in the dark, R results mainly from the scavenging of electrically injected electrons by oxidized species
(usually I3 − ions) in the hole conductor.
Dark currents typically reach current densities of several mA cm−2 at the operating bias of the solar cell, and exhibit rectifying behavior before being taken over by series resistance at current densities of ∼10 mA cm−2 (Fig. 9a). Unlike solidstate junctions, the dark current does not saturate at negative biases because of the nonnegligible probability of electron transfer to the ﬁlm at high reverse biases where the ﬁlm is electron poor. The high series resistances are due to the series resistance of the conducting glass of some 10–15 ohmsquare. Ideality factors of around 1 [99] and 2 [92] have been reported. In principle
107

Current
[mA cm 2]

106
105

22 °C
40 °C
55 °C
70 °C

this number is related to the kinetics of the reduction reaction, but because interpretation in terms of transport parameters depends on the type of junction, for instance, p–n, p–i–n, Schottky barrier, it is not useful for DSSCs. Interpretations of dark-current behavior in terms of various barrier [99] or kinetic [92] models have been made but are speculative.
The magnitude of the dark current can be increased by various treatments that are expected to enhance electron recombination. These include the addition of electron scavengers such as oxygen [100] to the electrolyte, and pretreatment to increase the density of defects (oxygen vacancies) in the ﬁlm [54]. The latter effect indicates that oxygen vacancies act as recombination centers. Any such

dark

104
103
102
101
−0.2 −0.1 0.0 0.1 0.2 0.3 0.4

Potential
[V]

(a)
0.75

0.50

Uphoto
[mV]

454

slope = 0.085 V

0.25

0.00
1010 1011 1012 1013 1014 1015 1016
(b)

Photon flux
[cm−2 s−1]

(a) Dark current voltage characteristics of dye-sensitized cells at different temperatures, from Ref. 99. At high forward bias the current deviates from the exponential form on account of series resistance. Notice the bias dependence of the current at reverse bias. (b) Light intensity dependence of the open circuit voltage. At high intensities Voc varies as expected from
Eq. (36) (from Ref. 16).

Fig. 9

5.3 Charge Transport in Dye-sensitized Systems

short circuit, and that recombination is negligible in this bias range.
At larger forward bias, the net photocurrent drops because of increased electron recombination. In a conventional solar cell, the process responsible for the recombination current is normally the same in light and dark, and it is common to make the approximation that J (V ) = Jsc − Jdark (V ).
Although this may be good for semiconductor p–n junction devices, it is not reliable for dye-sensitized photovoltaic devices. Ionization of the sensitizer opens up a new path for recombination in light, which is not active in dark (Fig. 11). There is some evidence [69] that the reverse current is larger under illumination than in the dark.

treatment that enhances dark current will reduce the efﬁciency of charge collection.
5.3.5.1.3 Light-current voltage Although the most relevant measurement for solar energy conversion, steady state J (V ) is not the most useful for analyzing charge transport. The J (V ) curve for a typical dye-sensitized cell in standard solar illumination is presented in Fig. 10. The short-circuit photocurrent, Jsc = J (0), corresponds to the IPCE integrated over the spectrum, Jsc =

(34)

η (E)b0 (E) dE

In a good device, J (V ) remains constant into small forward bias, indicating that the photocurrent has indeed saturated at

X-25 IV System
4.0
Sample PL0710/2
Oct 30, 1998 tape aperture, black dye

3.5
3.0

Area = 0.1863 cm2
Irradiance = 1000 Wm−2

Current
[mA]

2.5
2.0

VOC = 0.7210 V
ISC = 3.824 mA
Vmax = 0.5465 V
Imax = 3.552 mA
JSC = 20.53 mA cm−2
Pmax = 1.941 mW
Fill factor = 70.41%
Efficiency = 10.4%

1.5
1.0
0.5
0.0
−0.5
−0.2

0.0

0.2

0.4

0.6

0.8

Voltage
[V]
Fig. 10 Current voltage response of dye-sensitized cell measured at NREL [101]. This cell has a short-circuit current density of over 20 mA cm−2 and an open circuit voltage of over
0.7 V. The relatively low slope of I–V near Voc indicates a signiﬁcant series resistance, largely due to the sheet resistance of the conducting glass substrate.

455

456

5 Dye-Sensitized Photoelectrochemistry
Dark

Light

Schematic illustration of charge recombination pathways in a dye-sensitized device in dark and light conditions. In the dark, electrons in the nanocrystalline semiconductor recombine only with the hole conductor. Under illumination, some of the dye molecules are ionized and electrons may recombine either with the hole conductor or with the dye cation. The second recombination route is often neglected, but may be signiﬁcant at forward bias.

Fig. 11

Study of the dark current is thus useful for understanding the reaction between electrons and redox couple, but not for understanding the limits to photovoltage of the DSSC. Nevertheless, simple models of J (V ) behavior derived from solid-state device physics can been applied to DSSC’s to derive values of effective transport parameters. For example, Sodergren and coworkers [21] and Ferber and coworkers [102] have used a model, based on diffusion-controlled current, to ﬁt the J (V ) behavior of DSSCs.
The open circuit photovoltage, Voc , corresponds to the point, at which the recombination current and photocurrent are equal,
1 dJn
− Rn (Voc ) + G(x) = 0 q dx

(35)

There is no net current at Voc . If local currents are also small – for instance, if the generation rate is constant so that there are no internal diffusion currents – then the ﬁrst term in Eq. (35) can be neglected and the variation of Voc with light intensity tells us something about the form of

R , as does the dark J (V ), but without the problem of series resistance. For a dye-sensitized cell it is a more relevant comparison as it is probing recombination under illumination, and not in the dark.
Variation of Voc with light intensity has been reported by various authors including
Huang and coworkers [69] and Fisher and coworkers [16] for TiO2 and by de
Jongh and coworkers [67] for ZnO. For both materials the relationship tends to the form
Voc =

mkT ln X + Voc (1) q (36)

at high light intensities, where X is the light concentration factor, Voc (1) is the open-circuit voltage at unit concentration, and m the apparent ideality factor. Values of m are in the range 1.3 to 1.5, but, as commented earlier, these values are not meaningful without a model for the recombination current.
Voc is inﬂuenced by surface treatments and chemical environment. Huang and coworkers show that Voc decreases with increasing concentration of triiodide, as expected for a recombination current which

5.3 Charge Transport in Dye-sensitized Systems

is dominated by the reaction of electrons with I3 − . Several authors have reported the variation of Voc by surface treatments of the TiO2 . Treatment with pyridine derivatives [69] and tetrabutylammonium phosphate [2] is observed to increase Voc .
The effect is attributed to the masking of electrons at the TiO2 surface from electron scavengers in the electrolyte by inert adsorbed molecules. Treatment of TiO2 surface by benzoic acid derivatives is observed to alter the Voc of solid-state devices in a way that is consistent with modiﬁcation of the surface dipole ﬁeld by the adsorbed acids [103].
5.3.5.1.4 Photoconductivity and Photovoltaic Properties Relevant to the light response of dye-sensitized systems are the photoconductivity and photovoltaic properties of nanocrystalline ﬁlms. Earlier it was noted that the conductivity of nanocrystalline TiO2 is very small in the dark, around 10−10 –10−8 −1 cm−1 , rising by several orders of magnitude in UV light. Sensitized ﬁlms are photoconductive in visible light, even in the absence of a hole conductor [10]. The relationship between light intensity and photoconductivity is superlinear, indicating increased electron mobility through trap ﬁlling [104].
It can be shown by simulation that this behavior result is consistent with dispersive transport (discussed below) in the presence of traps [105].
Steady state surface photovoltage (SPV) spectroscopy is useful for determining the nature of the junction between the nanocrystalline ﬁlm and substrate. SPV studies of TiO2 ﬁlms on SnO2 conﬁrm that an electric ﬁeld exists at the interface, driving electrons into the substrate. This is consistent with the values of the work functions of the two materials in vacuum (4.85 eV for SnO2 : F, 5.15 eV for

TiO2 ) [12, 103] and with evidence for directed photocurrents in dry ﬁlms [94].
A comparison of the SPV spectra of an anatase single crystal and nanocrystalline ﬁlm indicate the presence of a tail of states in the band gap below the conduction band edge [58].
Time Domain
Traditionally, time resolved measurements are used to study the mechanisms of charge transport and recombination in semiconductor materials.
5.3.5.2

5.3.5.2.1 Photocurrent Transients Timeresolved photocurrent at zero or reverse bias is, like IPCE, a probe of the forces driving conduction rather than recombination. If recombination can be neglected, the transient electron current resulting from an instantaneous or steady state light stimulus G(x, t), is deﬁned by

1 ∂Jn
∂n
=
+ G(x, t)
∂t
q ∂x

(37)

in one dimension, where the form of J is not known.
Several conﬁgurations have been studied exhibiting some common features. Steady state and square-wave illumination of dye-sensitized electrochemical junctions produces a slow photocurrent, which rises to a saturation level [73, 106]. Photocurrent rise times are extremely long and intensity dependent (ms to s, depending on intensity). Both features are attributed to the trapping and release of photogenerated electrons. Schwarzburg and Willig [107] show with a simple rate equation treatment that the ﬁlling of a single trap level leads to intensity-dependent photocurrent kinetics, which are qualitatively similar to those observed. In their case of a single trap level
Eq. (17) can be resolved into the sum of

457

5 Dye-Sensitized Photoelectrochemistry

conduction (nc ) and trapped (nt ) electrons, and two continuity equations can be constructed for nc and nt with well deﬁned rate constants for trapping and detrapping. Cao and coworkers [106] present transients at different light intensities and solve Eq. (37) for a light step (G(x, t) = G0 for t > 0) and dn diffusive electron current Jn = qDn dx .
Although diffusion leads to a photocurrent which saturates at a level proportional to the light intensity, it predicts kinetics, which are intensity independent. This can be seen by normalizing n with respect to
G0 in Eq. (37): the solution for n(x, t)/G0 is independent of G0 . In the data, however, a clear intensity dependence is seen
(Fig. 12). The authors propose that the effective diffusion coefﬁcient is n-dependent and demonstrate that the numerical solution of the diffusion equation when
D(n) = D0 n, leads to a rise time which decreases with increasing light intensity.
This form of D(n) is somewhat arbitrary, but is qualitatively consistent with trap

ﬁlling: at higher injection levels, diffusion should be easier because fewer deep traps are available. Observations [73] of faster rise times under bias illumination and kinetics which depend on the direction of illumination (effectively, on the intensity reaching the SnO2 interface) are consistent with intensity-dependent diffusion.
Transient photocurrents in response to a laser pulse have been studied in electrolyte supported and dry ﬁlms [57, 60,
74, 108]. Such transient experiments are most informative when the optical depth of the ﬁlm is high, and the light pulse effectively creates a sheet of photogenerated carriers, which travels through the ﬁlm to the absorbing boundary. This is achieved by exciting in the UV where absorption of the TiO2 is high. Solbrand and coworkers [74, 108] present laser pulse induced transient photocurrents for a TiO2 electrochemical junction as a function of ﬁlm thickness, electrolyte composition and direction of illumination (Fig. 13). The

0.1
0.09
0.08

I (t) / G 0 / A W1

458

0.07
0.06
0.05
0.04

G 0 = 0.05 mW cm−2

0.03

G 0 = 4 mW cm−2

0.02
0.01
0
0

0.1

0.2

0.3

0.4

Time
[s]
Fig. 12 Transient photocurrent in response to step illumination of different intensities normalized with respect to the light intensity. At higher intensities, the photocurrent rise time is faster, indicating an electron density dependent diffusion coefﬁcient. Replotted from Ref. 106.

5.3 Charge Transport in Dye-sensitized Systems
In a transient photocurrent experiment, charges created near the
TiO2 –electrolyte interface travel through the ﬁlm to deliver a current at the collecting interface, from Ref. 74.
The thickness dependence of the transient currents is consistent with diffusion, but contains some features of
(a)
dispersive transport.

Fig. 13

n(x,t)

Back-contact

x

W

7.5 µm
11.3 µm
14.9 µm
17.4 µm

Current
[a.u.]

observed ﬁlm thickness dependence is consistent with a diffusion picture, where the carrier proﬁle generated near the semiconductor–electrolyte interface at x = W spreads out over time to deliver a timedependent ﬂux of carriers through the absorbing boundary at x = 0. It is evident that the transit times are long (tens of ms to s) and the current signal becomes more dispersive with ﬁlm thickness. They ﬁt the slow portion of their transients (an initial fast decay is attributed to capacitive effects) to the solution of the diffusion equation for a delta function stimulus, and derive an effective diffusion coefﬁcient, D , of the order of 10−6 cm2 s−1 . The facts that (1) the experimental tail in the photocurrent is more persistent than predicted by simulation and (2) the kinetic shape of the normalized current, J (0, t)/G0 , is intensity dependent, indicate that simple diffusion is not a complete description. These features are, however, consistent with a model of ‘‘anomalous diffusion’’, a random walk through energetically distributed trap states, as discussed by Nelson [109]. A second and important observation by

t2

t3

0

0
(b)

t1

20

40

60

80

100

Time
[ms]

Solbrand and coworkers [74] is that the apparent diffusion coefﬁcient is dependent on the conductivity and composition of the electrolyte. This is consistent with an activation energy for electron hopping which depends on the energy required to polarize its environment, a phenomenon which is familiar from conduction in doped disordered materials [110]. It is also a clear signal that the electrochemical interface is very important to transport, and consistent with the idea that electrons move along or close to the surface.
Laser pulse photocurrent transients in solid-state Schottky barrier junctions between dry TiO2 and a platinum contact have been studied by Konenkamp and coworkers [57]. They observe a power law dependence of the collected photocurrent on time,
(38)
J (0, t) ∝ t −β which is characteristic of ‘‘dispersive’’ transport. Such algebraic photocurrent transients are familiar from the physics of disordered semiconductors and have been shown to result from the anomalous

459

460

5 Dye-Sensitized Photoelectrochemistry

diffusion of carriers within a medium containing a wide energetic dispersion of trap states. A continuous time random walk model due to Scher and
Montroll [111] has been adapted by Nelson [109] for the case of nanocrystalline electrodes, and shown to be qualitatively consistent with observed photocurrent decays. There is some evidence in the work cited earlier [57] for a dependence of the photocurrent kinetics on chemical environment, suggesting that the different environments or surface treatments (air, vacuum, and pH of electrolyte) inﬂuence the distribution of states acting as electron traps. Related evidence is provided by
Hagfedlt and coworkers [112], who use time-resolved absorption at 900 nm to study the formation and disappearance of electrons (in absorbing Ti3+ states) following electron-hole pair generation by a UV laser pulse. They observe that in air, the signal initially decays rapidly
(tens of ns), whereas in ethanol, which is an efﬁcient hole scavenger, the signal persists for µs at least, implying that once holes have been removed the injected electron density decays very slowly. This suggests that it may be the presence of a hole-scavenging electrolyte, and not dye sensitization, which is the necessary condition for majority carrier injection in nanocrystalline ﬁlms.
5.3.5.2.2 Potential Step Transients This refers to the current transient in response to a potential step. In solid-state devices a technique known as junction recovery is applied to barrier junctions as a probe of the charge density in localized states, particularly in amorphous materials [104]. A large negative potential step is applied to a forward biased junction and

the recovered charge is measured as a function of forward current level. Konenkamp and coworkers [55, 60] have applied the technique to Schottky barrier nanocrystalline TiO2 junctions, and extract an electron mobility-lifetime product of approximately 10−10 cm2 V−1 , by ﬁtting their data with a simple model. They ﬁnd that the mobility-lifetime product is apparently independent of injection level, whereas the recombination time, which is in the ms– µs range, falls with increasing injection. This behavior is typical of disordered materials such as amorphous semiconductors, where charge transport is dominated by the ﬁlling of localized states. The observation is explained in terms of trap ﬁlling by arguing that the mobility µ varies like µ = µc

nc n (39)

where nc is the free (conduction band) electron density, n the total including trapped electrons, and µc represents the mobility of conduction band electrons in the bulk crystal. The ratio of nc to n is calculated using an exponential density of trap states, which appears to be consistent with observations of sub band gap photoconductivity.
The authors show that the junction recovery current transients are both intensity dependent and tend toward a power law form, as shown in Fig. 14. They argue that this behavior is consistent with the thermal stimulation of electrons out of an exponential distribution of trap states.
A variant of the potential step technique, chronoamperometry, has been applied to electrochemical junctions in the dark to estimate the charge stored in the ﬁlm.
The transient current resulting from an applied potential step is measured and integrated to yield a stored charge. Hoyer and Weller used this technique to estimate the charge stored in nanocrystalline ZnO

5.3 Charge Transport in Dye-sensitized Systems
−1

Log10 (Recovery current)
[A]

Fig. 14 Transient currents from a junction recovery experiment for a
TiO2 – Schottky barrier device [60]. The currents are dependent on injection level, and tend to a power law at long times, as expected for dispersive transport. 2 mA

α = 0.4

−2

0.4 mA

−3

0.1 mA

−4
−5
−6
−8

20 µA

−7

−6

−5

−4

−3

Log10 (Time)
[s]

ﬁlms [63] and others have used it to estimate charge stored in TiO2 [113]. Results indicate an accumulated electron density of several electrons per ∼15 nm diameter nanoparticle, and are consistent with – but do not prove – an exponential density of surface states. A variant of the junction recovery technique, where recovered current from an illuminated electrochemical junction at open circuit is measured as a function of light intensity, has been applied to nanocrystalline TiO2 by Duffy and coworkers [114]. These authors also infer an exponential density of states from their data and ﬁnd that the recovered charge varies algebraically with time, which they attribute to a reaction of electrons with I3 − , which is second order in n.
5.3.5.2.3 Transient Absorption Time-resolved absorption can be used to monitor the appearance or disappearance of the reduced or oxidized states of molecular species. Ultra fast time-resolved absorption was ﬁrst used by Tachibana and coworkers to measure the rate of injection from dye-excited state into semiconductor, probing the absorption of the

dye cation [90]. These and subsequent [82,
91] studies showed that injection occurs in sub picosecond time scales in a wide range of chemical environments, supporting the assertion made earlier that absorption is equivalent to photogeneration in a DSSC.
Although transient absorption is not a direct probe of transport, it can be used to probe the kinetics of recombination processes, which compete with transport.
Haque and coworkers [115] studied the recombination between electrons in the TiO2 electrode and photoionized dye molecules by monitoring the concentration of the dye cation following a laser pulse. By using a redox inactive environment, all other mechanisms to reduce the dye cation were suppressed. They found that this recombination reaction is slow (∼ms) at zero applied bias, but it is very strongly dependent on the bias applied to the junction, indicating that electrically injected electrons are capable of reducing the dye cation. Importantly, they ﬁnd that this recombination reaction becomes sufﬁciently fast (ns-µs time scale) to compete with the reduction of dye cation by iodide at voltages near open circuit, and so should

461

462

5 Dye-Sensitized Photoelectrochemistry

not be neglected in estimating the limits to performance of the DSSC. Simultaneous determination of the electron density in the ﬁlm using the known optical absorbance of electrons trapped at Ti3+ states showed that the half life for decay, t50% , varies like t50% ∝ n−1/α

(40)

where n is the total initial electron density including optically and electrically injected electrons, and α a fraction in the range from 0.25 to 0.5, which depends on chemical environment [11].
Neither the electron density dependence nor the shape (which is approximately stretched exponential) of the kinetics can be explained with second order reaction kinetics, where it is assumed that the reaction is controlled only by the concentrations of electrons and dye cations, nor are they consistent with simple electron transfer theory. An explanation was proposed by Nelson based on the continuous time random walk [109]. In the CTRW, electrons perform a random walk on a lattice, which contains trap sites distributed in energy, according to some distribution function, g(E). In contrast to normal diffusion, where the mean time taken for each step is a constant, in the
CTRW the time taken for each electron to move is determined by the time for thermal escape from the site currently occupied, τhop ∝ β e(EC −E)/kT

(41)

where EC is the conduction band edge energy, E is the energy of the trap site occupied, and β is the capture cross section deﬁned in Eq. (26) earlier.
The lattice is populated with a number of mobile electrons, n, and a number of static dye cations, [D+ ], and it is supposed that a dye cation is reduced whenever it meets an electron. If the trap

energies are distributed according to an exponential density of states function with characteristic temperature T0 g(E) =

1 −(EC −E)/kT0 e kT0

(42)

then, in the limit where n exceeds [D+ ], the cation population evolves with time like a stretched exponential [116]
[D+ ] ∝ e−n(t/τ )

α

where τ and α are constants and
T
α= .
T0

(43)

(44)

It can be shown from Eq. (43) [117] that in these conditions the half life for cation decay, t50% , varies with n like t50% ∝ n−1/α

(45)

as observed. Nelson and coworkers [11] have used Monte Carlo simulations of the
CTRW to model observed recombination kinetics in different chemical environments with different values of α , (Fig. 15).
If this is a valid description, it appears that recombination kinetics are strongly inﬂuenced by the presence and energetic distribution of trap states, and therefore may be controlled by controlling the ﬁlm treatment and chemical environment. In a separate study of the kinetics of the recombination reaction between electrons in TiO2 electrodes and rhenium sensitizer dyes, Hasselmann and Meyer [118] arrived at the conclusion that charge recombination is limited by the diffusional encounters of the electron with the dye cation, and not by the kinetics of the interfacial electron transfer step.
A less-used form of transient absorption is the monitoring of the electron density within the metal oxide electrode by measuring changes in the optical density in the red-IR, where Ti3+ states are

5.3 Charge Transport in Dye-sensitized Systems
1
0.9
0.8

∆OD(t) / OD0

0.7
−200 mV

0.6
0.5

−300 mV

0.4

−400 mV

0.3
−500 mV

0.2
0.1

0
10 000

100000

1000000

10000000

100000000

Time
[ns]

(a)

Electrons per cation

1000

n = At−0.46

100

10

1
10
(b)

1000

100000

10000000

t50%
[ns]

(a) Kinetic decay curves for the rereduction of the dye cation by electrons in the TiO2 ﬁlm. The strong bias dependence indicates that electrically injected electrons are capable of reducing the dye. The smooth lines are ﬁts using a continuous time random walk model of electron transport and assuming diffusion limited recombination. (b) Power law relationship between the electron density and the half life, t50% , for reduction of the dye cation. In the random walk model, the exponent α is related to the characteristic temperature of the density of trap states.
Fig. 15

known to absorb. This has been used by Hagfeldt and coworkers to observe electron recombination kinetics following injection by a UV laser pulse [112] and by

Olson and coworkers to observe the kinetics of electron accumulation following the application of a potential step in the dark [113]. Another novel probe of

463

464

5 Dye-Sensitized Photoelectrochemistry

transient electron density is the use of time-resolved microwave conductivity to sense changes in electron density following application of light or bias. Salafsky has applied this technique to deduce an electron transit time of 100s of ms to seconds for a dye-sensitized device at short circuit [119].
Frequency Domain
Here we refer to a class of techniques where the response of a system to small, periodic modulations of the driving force is studied as a function of the frequency of the disturbance. These give the same information as time-domain studies but without the need for large amplitude disturbances. This is particularly useful for dye-sensitized – and for a wide range of complex systems – because the optical and electronic response is nonlinear in n, and therefore diffusion coefﬁcient and recombination times are not constants.
The use of small modulations permit linearization of the transport problem, and yields effective diffusion and recombination parameters, which can be related to the injection level through the underlying steady state driving force.
We will distinguish various modes. The techniques most widely used in DSSCs are: intensity modulated photocurrent spectroscopy (IMPS) which is, like IPCE, concerned with electron transport under short circuit conditions; Intensity modulated photovoltage spectroscopy IMVS, which probes the competition between transport and recombination at open circuit; and electrical impedance spectroscopy (EIS), which probes the bias-dependent electrical response, analogous to dark-current transients. Other variants, such as frequencyresolved transmittance, the frequency domain analogs of transient absorption, have been developed [50]. The techniques and
5.3.5.3

their application to nanocrystalline systems are reviewed in Ref. 120.
In general, a driving force, which may be optical or electrical is applied and the phase and amplitude of the response, which may be a current, voltage, or optical density, is measured as a function of modulation frequency. For compactness, the results are often presented in the complex plane, as the locus of the complex response as frequency is varied, although phase and amplitude data plotted against frequency may be easier to interpret. For a simple system with a single time constant, τ , the response is a semicircle in the complex plane, and it is easy to show that the frequency at the point where the imaginary component of the response is a minimum, ωmin , is equal to 1/τ . Distorted semicircles imply multiphasic responses, which would be modeled with several time constants in the time domain.
5.3.5.3.1 Intensity Modulated Photocurrent Spectroscopy IMPS was used on unsensitized nanocrystalline TiO2 electrodes by de Jongh and Vanmaekelbergh [121].
In this conﬁguration the driving force is a modulated light source and the response is the modulated short-circuit photocurrent. These authors showed that ωmin is dependent on light intensity, X , varying approximately like ωmin ∝ X 0.5 (as shown in Fig. 16) and were able to attribute this to the trapping and detrapping of electrons in a distribution of trap states which is wide compared to kT . They deduced a density of trap states of around
1017 cm−3 and from this value concluded that the traps are located on the large
TiO2 –electrolyte interface and not at the
TiO2 –TCO contact [122]. The trapping hypothesis is supported by studies on porous
GaP electrodes, where ωmin is intensity independent and the trap density is known

5.3 Charge Transport in Dye-sensitized Systems

0.000
1 KHz
−0.005

100 Hz

0.1 Hz
1 Hz
8 Hz

−0.010

60
40
20
(c)

(c)
0.00

(a)

ω min/rad s−1

Im [∆i (ω)/e Φ(ω )]

80

0.01

0
0

0.02

Re [∆i (ω )/e Φ(ω)]

(b)

1

2

3

(Φ/1016 cm−2s−1)0.5

Fig. 16 (a) Complex plane plot of intensity-modulated photocurrent spectrum for a nanocrystalline TiO2 ﬁlm. A characteristic time is assigned from the frequency at which the imaginary part of the response function is greatest. (b) Intensity dependence of the

characteristic frequency, which varies approximately like the square root of the light intensity. In a trap-free system this frequency would be independent of light intensity (from
Ref. 121).

to be low; in this case multiple scattering is believed to dominate transport. The theory of trap limited IMPS response is developed in Ref. 123, where a distinction is made between multiple trapping and dispersive transport in the frequency domain, and in Ref. 124.
Dloczik and coworkers [92] later used
IMPS to obtain an estimate of the electron diffusion coefﬁcient in sensitized nanocrystalline TiO2 electrodes. This is done by transforming the time dependent diffusion equation (Eq. 22) into frequency space and ﬁtting the result for the photocurrent modulation to the measured values to extract values for an effective diffusion coefﬁcient, Dn and electron lifetime, τ . This analysis does not explicitly account for the trapping and detrapping of electrons. However, it is pointed out that for the case of a single trap level, the effective diffusion coefﬁcient is related to the true diffusion coefﬁcient, Dc , for conduction band electrons through nc Dn = Dc
(46)
n in analogy with Eq. (39). As injection level increases, the proportion of free electrons

increases and so does Dn . Later work by the same authors [50, 16] showed that the diffusion coefﬁcient derived this way varies with light intensity approximately like
Dn ∝ X 0.6

(47)

Fisher and coworkers combine this observation with the observed (exponential) dependence of photovoltage on light intensity to derive a relationship between Deff and electron Fermi level,
Dn ∝ e(1−β)EFn /kT

(48)

which is consistent with an exponential density of trap states of the form, g(E) ∝ eβE/kT

(49)

as Eq. (42), with β = α . They deduce a value of 0.57 for the exponent β from their data, which corresponds to a characteristic temperature T0 of 526 K [16].
IMVS In IMVS, the driving force is again a modulated light source and the measured response is the modulation in open circuit photovoltage. The technique was ﬁrst applied to dye-sensitized devices by Schlichthorl and coworkers [125].

5.3.5.3.2

465

466

5 Dye-Sensitized Photoelectrochemistry

The time constant, 1/ωmin , delivered by
IMVS, corresponds to an effective lifetime for recombination, τrec , whereas the time constant delivered by IMPS results from a combination of recombination and charge collection, τIMPS =

1 τrec +

1

−1

(50)

τcoll

although at short circuit the time constant for collection, τcoll , often dominates. It is clearly important to be able to relate these experimentally accessible time constants to the functionally relevant quantity, the quantum efﬁciency ηcoll for charge collection, but that is not trivial in a nonlinear system. Schlichthorl and coworkers [126] showed that although the linear approximation, ηcoll =

1 + τcoll τrec −1

(51)

is not adequate, a model based on electron diffusion with nonlinear relationships between electron density and recombination rate
(52)
R ∝ nm1 where m1 ≈ 2.2, and between total and free electron density ncb ∝ nm2

(53)

where m2 ≈ 2.7, is able to explain observed
IMVS and IMPS behavior. The two power law relationships are deduced from observations of τrec versus light intensity using IMVS, and Jsc versus electron density using IMPS. The strong power law dependence of ncb on n implied by the comparison is consistent with an exponential density of states, again of the form (40), but with α = 0.37 or
T0 = 810 K [127].

The origin of the power law dependence of recombination time on light intensity, consistent with Eq. (52) is not clear. Such behavior has been observed elsewhere but attributed alternatively to a reaction between electrons and triiodide, which is second order in n [16], and to charge trapping [11, 128].
5.3.5.3.3 Electrical Impedance Spectroscopy (EIS) EIS measures the current response to a modulation in applied bias, and can be used in the light or dark. Traditionally, EIS results are interpreted in terms of an equivalent circuit and information extracted about relevant capacitive, resistive, and faradaic processes. The technique has been applied to nanocrystalline TiO2 junctions in the dark by various groups [19, 22] and a ﬁlm capacitance extracted and interpreted using a simple equivalent circuit. Results conﬁrm that a potential drop does exist within the ﬁlm in the dark. In reverse bias, where the ﬁlm is depleted, the ﬁlm capacitance appears to be bias independent and dominated by the potential drop at the TiO2 –TCO interface. In forward bias the ﬁlm becomes populated and a high and a bias dependent capacitance develops at the large TiO2 –electrolyte interface, suggesting an increased charge storage capacity as surface or conduction band states become ﬁlled.
Van de Lagemaat and coworkers [93] apply EIS to illuminated nanocrystalline
TiO2 electrodes and again observe a capacitance which increases, approximately exponentially, with applied forward bias. They attribute this to the charging of surface states and show that the behavior is consistent with an exponential density of states. They further conclude that the potential drop across the TiO2 –TCO interface extends over a narrow region, and

5.3 Charge Transport in Dye-sensitized Systems

drops under illumination, which would be expected as the Fermi level of the TiO2 is raised. EIS under illumination provides information about the time constant for recombination, as does IMVS, but is not limited to open circuit. This means that with an appropriate model for the extraction of ηcoll from τrec and τcoll , EIS and
IMPS can provide an alternative measure of the bias dependent quantum efﬁciency of a dye-sensitized device. Those authors ﬁnd ηcoll = 0.45 at open circuit.
Frequency-resolved experiments in nanocrystalline electrodes often reveal a high degree of frequency dispersion in the underlying processes. In simple planar interface models this may be dealt with using a constant phase element (CPE) in the equivalent circuit. The CPE contains a degree of frequency dispersion but is hard to relate to physical processes.
Bisquert [129] has taken the analysis of electrical impedance spectra one stage further by invoking a transmission line model of the interface. The model allows for two resistive channels representing charge ﬂow in the two media, and distributed charge transfer at different points on the interface by means of a resistor-CPE element. They obtain excellent agreement with measured EIS in light and dark, and suggest that two characteristic relaxation frequencies can be resolved.
5.3.6

Summary

In summary, we can make the following observations about the nature of charge transport in dye-sensitized systems.
There is virtually no electric ﬁeld within the porous ﬁlm. There is an electric ﬁeld driving charge collection at the interface with TCO, and an electric ﬁeld at the
Helmholtz layer. Under illumination, a

photopotential is developed mainly across the semiconductor–electrolyte interface.
In electrolyte-supported systems, transport is rate limited by electron transport.
This has some characteristics of diffusion, but is not well described by a simple diffusion equation. Both the apparent diffusion coefﬁcient and the apparent recombination time are electron density dependent.
The n-dependence of D is consistent with the concept of increasing ease of transport as deep traps are ﬁlled.
There is widespread evidence for electron traps in the nanocrystalline ﬁlms, largely Ti3+ states because of oxygen vacancies and states because of adsorbed and intercalated species.
Nonlinear diffusion has been approached with a number of models:
(1) n-dependent diffusion coefﬁcient [106], which can be related to the fraction of free electrons in a medium containing traps, speciﬁcally, Dn ∝ nc /n [16, 55, 75,
127]; (2) rate-equation approach involving a distribution of trap states [123]; (3) Monte
Carlo simulations of anomalous diffusion in the presence of trap states [109]. Numerous authors have invoked an exponential distribution of trap states with a characteristic temperature of between 500 K and
810 K. There appears to be some consensus on this point.
Nonlinear recombination such as τrec ∝ n−β has been observed by several groups [11, 16, 60, 93, 115, 125]. This has been attributed to different origins: recombination reactions which are second order in n, and diffusion limited recombination, where the electron motion is limited by transport through traps.
The dispersion of time constants for conduction and recombination lead to dispersive features in both time-resolved and frequency-resolved measurements. These

467

468

5 Dye-Sensitized Photoelectrochemistry

are not easily addressable with equivalent circuit models for planar interfaces.
The fact that recombination is slow and transport relatively efﬁcient at short circuit, is qualitatively attributed to the effects of screening of electrons by cations in the electrolyte; and to the absence of mobile photogenerated holes. Recombination with the oxidized species in the hole conductor is normally assumed to be the main loss process for electrons, but recombination with dye cations cannot be ruled out.
Continuity equation approaches involving nonlinear forms for the electron diffusion and recombination terms are capable of reproducing some aspects of observed behavior [123, 126]. The challenge is to predictively relate the forms for recombination and diffusion to known material properties of the system.
Chemical environment strongly inﬂuences transport. Adsorption and intercalation inﬂuence the potential distribution at the interface, affecting the tendency for electrons to accumulate in the ﬁlm and the photovoltage. Composition of electrolyte inﬂuences rate of electron scavenging.
Film history inﬂuences the density of traps and recombination centers.
In nanoparticulate semiconductors other than TiO2 , quantum conﬁnement may become important.
In solid-state systems charge transport within the electrolyte is expected to be rate limiting. Even in liquid electrolyte systems there is some evidence for an inﬂuence of electrolyte on electron transport. This is expected from the screening and reorganization energy. In solid-state systems the absence of macroscopic ﬁelds cannot be assumed and drift may be expected to become signiﬁcant.
Most models applied to date use an effective medium in which electron transport

is considered only in one dimension.
Some consider transport within a single nanoparticle, or other highly simpliﬁed geometry. There is clearly a need to move on to models incorporating the interpenetrating features of the morphology.
Looking to the future, the following issues remain to be addressed quantitatively: the relationship between material properties and the apparent density of states; the effect of hole conductor composition and mobility on electron transport; Coulombic effects of nanoparticle charging and electron-hole interactions. An improved description of charge transport is needed to replace the simple diffusion model. Techniques developed to deal with disordered conductors in the past may prove useful here. Multiple-trapping models such as the continuous time random walk can explain certain dispersive features of current and recombination kinetics. Another tool that offers the potential to describe some features of the behavior of these systems is percolation theory. This has already been used to describe hole conduction in a polymeric hole conductor [130]; to describe charge hopping between quantum sized
ZnO nanoparticles [65]; and to describe hole migration between dyes [83, 84].
Acknowledgment

The author is grateful to James Durrant,
Sven Sodergren, Thierry Lutz, Carol Olson, Mili Biswas, and Ellen Moons for helpful discussions, and acknowledges the ﬁnancial support of the Engineering and
Physical Sciences Research Council and the Greenpeace Environmental Trust.
References
1. M. Gr¨tzel, Nanocrystalline electronic junca tions, Stud. Surf. Sci. Catal. 1997 103,
353–375.

5.3 Charge Transport in Dye-sensitized Systems
2. A. Hagfeldt, M. Gratzel, Light-induced redox reactions in nanocrystalline systems,
Chem. Rev. 1995, 95(1), 49–68.
3. B. O’Regan, M. Gratzel, A low-cost, highefﬁciency solar-cell based on dye-sensitized colloidal TiO2 ﬁlms, Nature 1991, 353(6346),
737–740.
4. N. Vlachopoulos, P. Liska, J. Augustynski et al., Very efﬁcient visible-light energy harvesting and conversion by spectral sensitization of high surface-area polycrystalline titanium-dioxide ﬁlms, J. Am. Chem. Soc.
1988, 110(4), 1216–1220.
5. C. Bechinger, S. Ferrer, A. Zaban et al.,
Photoelectrochromic windows and displays,
Nature 1996, 383(6601), 608–610.
6. J. Sotomayor, G. Will, D. Fitzmaurice, Photoelectrochromic heterosupramolecular assemblies, J. Mater. Chem. 2000, 10(3),
685–692.
7. E. Topoglidis, A. E. G. Cass, G. Gilardi et al., Protein adsorption on nanocrystalline
TiO2 ﬁlms: an immobilization strategy for bioanalytical devices, Anal. Chem. 1998,
70(23), 5111–5113.
8. S. Y. Huang, L. Kavan, I. Exnar et al., Rocking chair lithium battery based on nanocrystalline Tio2 (anatase), J. Electrochem. Soc.
1995, 142(9), L142–L144.
9. C. Bechinger, B. A. Gregg, Development of a new self-powered electrochromic device for light modulation without external power supply, Sol. Energy Mater. Sol. Cells 1998,
54(1–4), 405–410.
10. P. Hoyer, R. Konenkamp, Photoconduction in porous Tio2 sensitized by Pbs quantum dots, App. Phys. Lett. 1995, 66(3), 349–351.
11. J. Nelson, S. A. Haque, D. R. Klug et al.,
Observation of trap limited recombination in dye-sensitized nanocrystalline metal oxide electrodes, Phys. Rev. B 2001, 63,
20 5321.
12. D. Cahen, G. Hodes, M. Gratzel et al., Nature of photovoltaic action in dye-sensitized solar cells, J. Phys. Chem. B 2000, 104(9),
2053–2059.
13. C. A. Kelly, F. Farzad, D. W. Thompson et al., Cation-controlled interfacial charge injection in sensitized nanocrystalline TiO2 ,
Langmuir 1999, 15(20), 7047–7054.
14. M. K. Nazeeruddin, A. Kay, I. Rodicio et al.,
Conversion of light to electricity by cisX2bis(2,2 -Bipyridyl-4,4 -Dicarboxylate)
Ruthenium(Ii) charge-transfer sensitizers

15.

16.

17.
18.

19.

20.

21.

22.

23.

24.

25.

(X = Cl-, Br-, I-, Cn-, and Scn-) on nanocrystalline Tio2 electrodes, J. Am. Chem. Soc.
1993, 115(14), 6382–6390.
E. Topoglidis, T. Lutz, R. L. Willis et al., Protein adsorption on nanoporous TiO2 ﬁlms: a novel approach to studying photoinduced protein/electrode interactions, Faraday Discuss. 2000, 116, 35–46.
A. C. Fisher, L. M. Peter, E. A. Ponomarev et al., Intensity dependence of the back reaction and transport of electrons in dyesensitized nanocrystalline TiO2 solar cells,
J. Phys. Chem. B 2000, 104(5), 949–958.
N. Myung, S. Licht, J. Electrochem. Soc.
1995, 142, L129–L132.
D. Fitzmaurice, Using spectroscopy to probe the band energetics of transparent nanocrystalline semiconductor-ﬁlms,
Sol. Energy Mater. Sol. Cells 1994, 32(3),
289–305.
F. Cao, G. Oskam, P. C. Searson et al., Electrical and optical-properties of porous nanocrystalline Tio2 ﬁlms, J. Phys. Chem.
1995, 99(31), 11974–11980.
J. Bisquert, G. GarciaBelmonte, F. FabregatSantiago, Modelling the electric potential distribution in the dark in nanoporous semiconductor electrodes, J. Solid State Electrochem. 1999, 3(6), 337–347.
S. Sodergren, A. Hagfeldt, J. Olsson et al.,
Theoretical-models for the action spectrum and the current-voltage characteristics of microporous semiconductor-ﬁlms in photoelectrochemical cells, J. Phys. Chem. 1994,
98(21), 5552–5556.
A. Zaban, A. Meier, B. A. Gregg, Electric potential distribution and short-range screening in nanoporous TiO2 electrodes,
J. Phys. Chem. B 1997, 101(40), 7985–7990.
D. Matthews, P. Infelta, M. Gratzel, Calculation of the photocurrent-potential characteristic for regenerative, sensitized semiconductor electrodes, Sol. Energy Mater. Sol.
Cells 1996, 44(2), 119–155.
B. A. Gregg, A. Zaban, S. Ferrere, Dye sensitized solar cells: energetic considerations and applications, Zeitschrift Fur Physikalische Chemie-International, J. Res. Phys.
Chem. Chem. Phys. 1999, 212(Pt1), 11–22.
F. Pichot, B. A. Gregg, The photovoltagedetermining mechanism in dye-sensitized solar cells, J. Phys. Chem. B 2000, 104(1),
6–10.

469

470

5 Dye-Sensitized Photoelectrochemistry
26. H. Lindstrom, S. Sodergren, A. Solbrand et al., Li+ ion insertion in TiO2 (anatase) chronoamperometry on CVD ﬁlms and nanoporous ﬁlms, J. Phys. Chem. B 1997,
101(39), 7710–7716.
27. C. J. Barbe, F. Arendse, P. Comte et al.,
Nanocrystalline titanium oxide electrodes for photovoltaic applications, J. Am. Ceram.
Soc. 1997, 80(12), 3157–3171.
28. V. Shklover, M. K. Nazeeruddin, S. M.
Zakeeruddin et al., Structure of nanocrystalline TiO2 powders and precursor to their highly efﬁcient photosensitizer, Chem.
Mater. 1997, 9(2), 430–439.
29. A. Hagfeldt, B. Didriksson, T. Palmqvist et al., Veriﬁcation of high efﬁciencies for the Gratzel-cell – a 7-percent efﬁcient solarcell based on dye-sensitized colloidal TiO2 ﬁlms, Sol. Energy Mater. Sol. Cells 1994,
31(4), 481–488.
30. P. A. Cox, Transition Metal Oxides, in International Series of Monographs on Chemistry,
Oxford, Oxford University Press, 1992.
31. H. Tang, H. Berger, P. E. Schmid et al.,
Optical-properties of anatase (TiO2 ), Solid
State Commun. 1994, 92(3), 267–271.
32. H. Tang, K. Prasad, R. Sanjines et al., Electrical and optical-properties of TiO2 anatase thin-ﬁlms, J. Appl. Phys. 1994, 75(4),
2042–2047.
33. H. Tang, F. Levy, H. Berger et al., Urbach tail of anatase TiO2 , Phys. Rev. B: Conden.
Matter 1995, 52(11), 7771–7774.
34. R. Sanjines, H. Tang, H. Berger et al.,
Electronic-structure of anatase TiO2 oxide,
J. Appl. Phys. 1994, 75(6), 2945–2951.
35. H. Z. Zhang, J. F. Banﬁeld, Thermodynamic analysis of phase stability of nanocrystalline titania, J. Mater. Chem. 1998, 8(9),
2073–2076.
36. L. Kavan, M. Gr¨tzel, S. E. Gilbert et al., a Electrochemical and photoelectrochemical investigation of single-crystal anatase, J.
Amer. Chem. Soc. 1996, 118(28), 6716–6723.
37. W. C. Mackrodt, E. A. Simson, N. M. Harrison, An ab initio Hartree-Fock study of the electron-excess gap states in oxygendeﬁcient rutile TiO2 , Surf. Sci. 1997,
384(1–3), 192–200.
38. R. G. Egdell, S. Eriksen, W. R. Flavell, Oxygen deﬁcient SnO2 (110) and TiO2 (110) – a comparative-study by photoemission, Solid
State Commun. 1986, 60(10), 835–838.

39. U. Kirner, K. D. Schierbaum, W. Gopel et al., Low and high-temperature TiO2 oxygen sensors, Sens. Actuators B-Chem.
1990, 1(1–6), 103–107.
40. J. Muscat, N. M. Harrison, G. Thornton,
First-principles study of potassium adsorption on TiO2 surfaces, Phys. Rev. B: Conden.
Matter 1999, 59(23), 15457–15463.
41. S. Sodergren, H. Siegbahn, H. Rensmo et al., Lithium intercalation in nanoporous anatase TiO2 studied with XPS, J. Phys.
Chem. B 1997, 101(16), 3087–3090.
42. A. Vittadini, A. Selloni, F. P. Rotzinger et al.,
Structure and energetics of water adsorbed at TiO2 anatase (101) and (001) surfaces,
Phys. Rev. Lett. 1998, 81(14), 2954–2957.
43. P. J. D. Lindan, N. M. Harrison, M. J.
Gillan, Mixed dissociative and molecular adsorption of water on the rutile (110) surface,
Phys. Rev. Lett. 1998, 80(4), 762–765.
44. H. Haerudin, S. Bertel, R. Kramer, Surface stoichiometry of ‘‘titanium suboxide’’ – Part I–Volumetric and FTIR study,
J. Chem. Soc., Faraday Trans. 1998, 94(10),
1481–1487.
45. R. F. Howe, M. Gr¨tzel, Electron-paraa magnetic-res observation of trapped electrons in colloidal TiO2 , J. of Phys. Chem.
1985, 89(21), 4495–4499.
46. U. Kolle, J. Moser, M. Gr¨tzel, Dynama ics of interfacial charge-transfer reactions in semiconductor dispersions – reduction of cobaltoceniumdicarboxylate in colloidal
TiO2 , Inorg. Chem. 1985, 24(14), 2253–2258.
47. B. O’Regan, M. Gr¨tzel, D. Fitzmaurice, a Optical electrochemistry steady state spectroscopy of conduction-band electrons in a metal-oxide semiconductor electrode,
Chem. Phys. Lett. 1991, 183(1–2), 89–93.
48. G. Rothenberger, D. Fitzmaurice, M. Gratzel, Optical electrochemistry spectroscopy of conduction-band electrons in transparent metal-oxide semiconductor-ﬁlms – optical determination of the ﬂat-band potential of colloidal titanium-dioxide ﬁlms, J. Phys.
Chem. 1992, 96(14), 5983–5986.
49. G. Redmond, D. Fitzmaurice, Spectroscopic determination of ﬂat-band potentials for polycrystalline Tio2 electrodes in nonaqueous solvents, J. Phys. Chem. 1993, 97(7),
1426–1430.
50. G. Franco, J. Gehring, L. M. Peter et al.,
Frequency-resolved optical detection of photoinjected electrons in dye-sensitized

5.3 Charge Transport in Dye-sensitized Systems

51.

52.

53.

54.

55.

56.

57.

58.
59.

60.

61.

62.

nanocrystalline photovoltaic cells, J. Phys.
Chem. B 1999, 103(4), 692–698.
G. Boschloo, D. Fitzmaurice, Electron accumulation in nanostructured TiO2 (anatase) electrodes, J. Phys. Chem. B 1999, 103(37),
7860–7868.
L. Forro, O. Chauvet, D. Emin et al., Highmobility N -type charge-carriers in large single-crystals of anatase (TiO2 ), J. Appl.
Phys. 1994, 75(1), 633–635.
B. Poumellec, J. F. Marucco, F. Lagnel,
Electron-transport in TiO2 -X at intermediate temperatures 300-K less- than T lessthan 1500-K, Phys. Status Solidi A-Appl. Res.
1985, 89(1), 375–382.
J. Weidmann, T. Dittrich, E. Konstantinova et al., Inﬂuence of oxygen and water related surface defects on the dye sensitized TiO2 solar cell, Sol. Energy Mater. Sol. Cells 1999,
56(2), 153–165.
R. Konenkamp, Carrier transport in nanoporous TiO2 ﬁlms, Phys. Rev. B 2000, 61(16),
11057–11064.
T. Dittrich, J. Weidmann, V. Y. Timoshenko et al., Thermal activation of the electronic transport in porous titanium dioxides,
Mater. Sci. Eng., B-Solid State Materials
For Advanced Technology 2000, 69(SISI),
489–493.
R. Konenkamp, R. Henninger, P. Hoyer,
Photocarrier transport in colloidal TiO2 ﬁlms, J. Phys. Chem. 1993, 97(28),
7328–7330.
E. Moons, M. Gr¨tzel, Photovoltage in Nanoa crystalline TiO2 Films 1998, Unpublished.
T. Dittrich, E. A. Lebedev, J. Weidmann,
Electron drift mobility in porous TiO2
(anatase), Phys. Status Solidi A-Appl. Res.
1998, 165(2), R5–R6.
R. Konenkamp, R. Henninger, Recombination in nanophase TiO2 ﬁlms, Appl. Phys.
A-Materials Science & Processing 1994, 58(1),
87–90.
S. Burnside, J. E. Moser, K. Brooks et al.,
Nanocrystalline mesoporous strontium titanate as photoelectrode material for photosensitized solar devices: increasing photovoltage through ﬂatband potential engineering, J. Phys. Chem. B 1999, 103(43),
9328–9332.
J. M. Stipkala, F. N. Castellano, T. A.
Heimer et al., Light-induced charge separation at sensitized sol-gel processed

63.

64.

65.

66.

67.

68.

69.

70.

71.

72.

73.

semiconductors, Chem. Mater. 1997, 9(11),
2341–2353.
P. Hoyer, H. Weller, Potential-dependent electron injection in nanoporous colloidal
ZnO ﬁlms, J. Phys. Chem. 1995, 99(38),
14096–14100.
H. Rensmo, K. Keis, H. Lindstrom et al.,
High light-to-energy conversion efﬁciencies for solar cells based on nanostructured ZnO electrodes, J. Phys. Chem. B 1997, 101(14),
2598–2601.
E. A. Meulenkamp, Electron transport in nanoparticulate ZnO ﬁlms, J. Phys. Chem. B
1999, 103(37), 7831–7838.
A. vanDijken, E. A. Meulenkamp, D. Vanmaekelbergh et al., The kinetics of the radiative and nonradiative processes in nanocrystalline ZnO particles upon photoexcitation, J. Phys. Chem. B 2000, 104(8),
1715–1723.
P. E. de Jongh, E. A. Meulenkamp, D. Vanmaekelbergh et al., Charge carrier dynamics in illuminated particulate ZnO electrodes, J. Phys. Chem. B 2000, 104,
7686–7693.
P. E. de Jongh, D. Vanmaekelbergh, J. J.
Kelly, Photoelectrochemistry of electrodeposited Cu2 O, J. Electrochem. Soc. 2000,
147(2), 486–489.
S. Y. Huang, G. Schlichthorl, A. J. Nozik et al., Charge recombination in dye-sensitized nanocrystalline TiO2 solar cells, J. Phys.
Chem. B 1997, 101(14), 2576–2582.
A. Hauch, R. Kern, J. Ferber et al., Characterisation of the electrolyte-solid interfaces of dye-sensitized solar cells by means of impedance spectroscopy, 2nd World Conference on Photovoltaic Solar Energy Conversion,
Vienna, European Communities, 1998.
N. Papageorgiou, M. Gratzel, P. P. Infelta,
On the relevance of mass transport in thin layer nanocrystalline photoelectrochemical solar cells, Sol. Energy Mater. Sol. Cells 1996,
44(4), 405–438.
Z. Kebede, S. E. Lindquist, The obstructed diffusion of the I-3(−) ion in mesoscopic
TiO2 membranes, Sol. Energy Mater. Sol.
Cells 1998, 51(3–4), 291–303.
P. M. Sommeling, H. C. Rieffe, J. M. Kroon et al., Spectral response and response time of nanocrystalline dye-sensitized TiO2 solar cells, 14th European Photovoltaic Solar Energy Conference, Barcelona, H.S. Stephens and Associates, 1997.

471

472

5 Dye-Sensitized Photoelectrochemistry
74. A. Solbrand, H. Lindstrom, H. Rensmo et al., Electron transport in the nanostructured TiO2 -electrolyte system studied with time-resolved photocurrents, J. Phys.
Chem. B 1997, 101(14), 2514–2518.
75. N. Kopidakis, E. A. Schiff, N.- G. Park et al.,
Ambipolar diffusion of carriers in electrolyte ﬁlled, nanoporous TiO2 , J. Phys.
Chem. B 2000, 104(16), 3930–3936.
76. F. Cao, G. Oskam, P. C. Searson, A solidstate, dye-sensitized photoelectrochemical cell, J. Phys. Chem. 1995, 99(47),
17071–17073.
77. S. A. Haque, S. Sodergren, A. Holmes et al.,
Solid state dye-sensitized photovoltaic cells employing a polymer electrolyte, International conference on photochemical conversion and storage of solar energy, Colorado, 2000.
78. T. J. Savenije, J. M. Warman, A. Goossens,
Visible light sensitization of titanium dioxide using a phenylene vinylene polymer,
Chem. Phys. Lett. 1998, 287(1–2), 148–153.
79. U. Bach, D. Lupo, P. Comte et al., Solidstate dye-sensitized mesoporous TiO2 solar cells with high photon-to-electron conversion efﬁciencies, Nature 1998, 395(6702),
583–585.
80. B. O’Regan, D. T. Schwartz, Efﬁcient dyesensitized charge separation in a wide-bandgap p–n heterojunction, J. Appl. Phys. 1996,
80(8), 4749–4754.
81. K. Tennakone, G. Kumara, A. R. Kumarasinghe et al., A Dye-sensitized nano-porous solid-state photovoltaic cell, Semicond. Sci.
Technol. 1995, 10(12), 1689–1693.
82. Y. Tachibana, S. A. Haque, I. P. Mercer et al., Electron injection and recombination in dye sensitized nanocrystalline titanium dioxide ﬁlms: a comparison of ruthenium bipyridyl and porphyrin sensitizer dyes, J.
Phys. Chem. B 2000, 104(6), 1198–1205.
83. A. Fillinger, B. A. Parkinson, The adsorption behavior of a ruthenium-based sensitizing dye to nanocrystalline TiO2 – coverage effects on the external and internal sensitization quantum yields, J. Electrochem. Soc.
1999, 146(12), 4559–4564.
84. P. Bonhote, E. Gogniat, S. Tingry et al., Efﬁcient lateral electron transport inside a monolayer of aromatic amines anchored on nanocrystalline metal oxide ﬁlms, J. Phys.
Chem. B 1998, 102(9), 1498–1507.

85. R. Vogel, P. Hoyer, H. Weller, Quantumsized Pbs, Cds, Ag2s, Sb2s3, and Bi2s3 particles as sensitizers for various nanoporous wide-bandgap semiconductors, J. Phys.
Chem. 1994, 98(12), 3183–3188.
86. A. Zaban, O .I. Micic, B. A. Gregg et al.,
Photosensitization of nanoporous TiO2 electrodes with InP quantum dots, Langmuir 1998, 14(12), 3153–3156.
87. M. Shur, in Physics of Semiconductor Devices, Prentice-Hall, Englewood Cliffs, 1990, pp. 86–102.
88. M. Lundstrom, in Fundamentals of Carrier
Transport, Addison Wesley, Wokingham,
1990.
89. S. J. Fonash, in Solar Cell Device Physics,
Academic Press, New York, 1981.
90. Y. Tachibana, J. E. Moser, M. Gratzel et al.,
Subpicosecond interfacial charge separation in dye-sensitized nanocrystalline titanium dioxide ﬁlms, J. Phys. Chem. 1996,
100(51), 20056–20062.
91. T. Hannappel, B. Burfeindt, W. Storck et al.,
Measurement of ultrafast photoinduced electron transfer from chemically anchored
Ru-dye molecules into empty electronic states in a colloidal anatase TiO2 ﬁlm, J.
Phys. Chem. B 1997, 101(35), 6799–6802.
92. L. Dloczik, O. Ileperuma, I. Lauermann et al., Dynamic response of dye-sensitized nanocrystalline solar cells: characterization by intensity-modulated photocurrent spectroscopy, J. Phys. Chem. B 1997, 101(49),
10281–10289.
93. J. van de Lagemaat, N. G. Park, A. J. Frank,
Inﬂuence of electrical potential distribution, charge transport, and recombination on the photopotential and photocurrent conversion efﬁciency of dye-sensitized nanocrystalline TiO2 solar cells: a study by electrical impedance and optical modulation techniques, J. Phys. Chem. B 2000, 104(9),
2044–2052.
94. B. Levy, W. Liu, S. E. Gilbert, Directed photocurrents in nanostructured TiO2 /SnO2 heterojunction diodes, J. Phys. Chem. B
1997, 101(10), 1810–1816.
95. H. Lindstrom, H. Rensmo, S. Sodergren et al., Electron transport properties in dye-sensitized nanoporous-nanocrystalline
TiO2 ﬁlms, J. Phys. Chem. 1996, 100(8),
3084–3088.
96. M. K. Nazeeruddin, P. Pechy, M. Gr¨tzel, a Efﬁcient panchromatic sensitization of

5.3 Charge Transport in Dye-sensitized Systems

97.

98.

99.

100.

101.
102.

103.

104.

105.

106.

107.

108.

nanocrystalline TiO2 ﬁlms by a black dye based on a trithiocyanato-ruthenium complex, Chem. Commun. 1997, 18, 1705–1706.
M. Gr¨tzel, Photoelectrochemical solar ena ergy conversion by dye sensitization, Future
Generation Photovoltaic Technologies: First
NREL Conference, Denver: AIP, 1997.
A. Hagfeldt, U. Bjorksten, S. E. Lindquist,
Photoelectrochemical studies of colloidal
TiO2 -ﬁlms – the charge separation process studied by means of action spectra in the
UV region, Sol. Energy Mater. Sol. Cells 1992,
27(4), 293–304.
T. Dittrich, P. Beer, F. Koch et al., Barrier lowering in dye-sensitized porous-TiO2 solar cells, Appl. Phys. Lett. 1998, 73(13),
1901–1903.
A. Stanley, B. Verity, D. Matthews, Minimizing the dark current at the dyesensitized TiO2 electrode, Sol. Energy Mater.
Sol. Cells 1998, 52(1–2), 141–154.
A. Hagfedlt, M. Gr¨tzel, Acc. Chem. Res. a 2000, 33, 269–277.
J. Ferber, R. Stangl, J. Luther, An electrical model of the dye-sensitized solar cell, Sol.
Energy Mater. Sol. Cells 1998, 53(1–2),
29–54.
J. Kruger, U. Bach, M. Gr¨tzel, Modiﬁcaa tion of TiO2 heterojunctions with benzoic acid derivatives in hybrid molecular solidstate devices, Adv. Mater. 2000, 12(6), 447 (6 pages). R. A. Street, in Hydrogenated Amorphous
Silicon, Cambridge University Press, Cambridge, 1991.
J. A. Anta, J. Nelson, N. Quirke, Charge transport model for disordered materials: application to sensitized TiO2 , Phys. Rev. B
2000(Submitted).
F. Cao, G. Oskam, G. J. Meyer et al., Electron transport in porous nanocrystalline
TiO2 photoelectrochemical cells, J. Phys.
Chem. 1996, 100(42), 17021–17027.
K. Schwarzburg, F. Willig, Inﬂuence of trap ﬁlling on photocurrent transients in polycrystalline TiO2 , Appl. Phys. Lett. 1991,
58(22), 2520–2522.
A. Solbrand, A. Henningsson, S. Sodergren et al., Charge transport properties in dyesensitized nanostructured TiO2 thin ﬁlm electrodes studied by photoinduced current transients, J. Phys. Chem. B 1999, 103(7),
1078–1083.

109. J. Nelson, Continuous-time random-walk model of electron transport in nanocrystalline TiO2 electrodes, Phys. Rev. B: Condens. Matter 1999, 59(23), 15374–15380.
110. N. F. Mott, Conduction in glasses containing transition metal ions, J. Non-cryst. Solids
1968, 1, 1–17.
111. H. Scher, E. W. Montroll, Anomalous transit time dispersion in amorphous solids,
Phys. Rev. B 1975, 12, 2455.
112. A. Hagfeldt, S. E. Lindquist, M. Gratzel,
Charge-carrier separation and charge-transport in nanocrystalline junctions, Sol. Energy Mater. Sol. Cells 1994, 32(3), 245–257.
113. C. L. Olson, R. L. Willis, T. Lutz et al. Current Transients, Optical Absorbance and
Charge Recombination in Dye-sensitized
Nanocrystalline Titanium Dioxide Electrodes, 13th International conference on photochemical conversion and storage of solar energy (IPS-2000), Snowmass, Colorado,
2000, Unpublished.
114. N. W. Duffy, L. M. Peter, R. M. G. Rajapakse et al., Electrochemistry Communications 2, 2000, 658–662.
115. S. A. Haque, Y. Tachibana, D. R. Klug et al.,
Charge recombination kinetics in dyesensitized nanocrystalline titanium dioxide ﬁlms under externally applied bias, J. Phys.
Chem. B 1998, 102(10), 1745–1749.
116. A. Blumen, G. Zumofen, J. Klafter,
Target annihilation by random walkers,
Phys. Rev. B: Condens. Matter 1984, 30(9),
5379–5382.
117. A. Blumen, J. Klafter, G. Zumofen, Models for reaction dynamics in glasses, in Optical
Spectroscopy of glasses, (Ed.: I. Zschokke),
D. Reidel, Dordrecht, The Netherlands,
1986, pp. 199–266.
118. G. M. Hasselmann, G.J. Meyer, Diffusionlimited interfacial electron transfer with large apparent driving forces, J. Phys.
Chem. B 1999, 103(36), 7671–7675.
119. J. S. Salafsky, W. H. Lubberhuizen, E. vanFaassen et al., Charge dynamics following dye photoinjection into a TiO2 nanocrystalline network, J. Phys. Chem. B 1998,
102(5), 766–769.
120. L. Peter, D. Vanmaekelbergh, Time and frequency resolved studies of photoelectrochemical kinetics, in Advances in Electrochemical Science and Engineering, (Eds.:
R. Alkire, D. Kolb), Wiley WCH, Weinheim,
1999, pp. 77–163.

473

474

5 Dye-Sensitized Photoelectrochemistry
121. P. E. de Jongh, D. Vanmaekelbergh, Traplimited electronic transport in assemblies of nanometer-size TiO2 particles, Phys. Rev.
Lett. 1996, 77(16), 3427–3430.
122. P. E. de Jongh, D. Vanmaekelbergh, Investigation of the electronic transport properties of nanocrystalline particulate TiO2 electrodes by intensity-modulated photocurrent spectroscopy, J. Phys. Chem. B 1997,
101(14), 2716–2722.
123. D. Vanmaekelbergh, P. E. de Jongh, Electron transport in disordered semiconductors studied by a small harmonic modulation of the steady state, Phys. Rev. B 2000,
61(7), 4699–4704.
124. J. J. Kelly, D. Vanmaekelbergh, Charge carrier dynamics in nanoporous photoelectrodes, Electrochim. Acta 1998, 43(19–20),
2773–2780.
125. G. Schlichthorl, S. Y. Huang, J. Sprague et al., Band edge movement and recombination kinetics in dye-sensitized nanocrystalline TiO2 solar cells: a study by intensity modulated photovoltage spectroscopy, J.
Phys. Chem. B 1997, 101(41), 8141–8155.
126. G. Schlichthorl, N. G. Park, A. J. Frank,
Evaluation of the charge-collection efﬁ-

127.

128.

129.

130.

ciency of dye-sensitized nanocrystalline
TiO2 solar cells, J. Phys. Chem. B 1999,
103(5), 782–791.
J. van de Lagemaat, A. J. Frank, Effect of the surface-state distribution on electron transport in dye-sensitized TiO2 solar cells: Nonlinear electron-transport kinetics, J. Phys. Chem. B 2000, 104(18),
4292–4294.
N. W. Duffy, L. M. Peter, K. G. U. Wijayantha, Characterisation of electron transport and back reaction in dye-sensitized nanocrystalline solar cells by small amplitude laser pulse excitation, Electrochem.
Commun. 2000, 2(4), 262–266.
J. Bisquert, G. GarciaBelmonte, F. FabregatSantiago et al., Doubling exponent models for the analysis of porous ﬁlm electrodes by impedance. Relaxation of TiO2 nanoporous in aqueous solution, J. Phys. Chem. B 2000,
104(10), 2287–2298.
J. S. Salafsky, Exciton dissociation, charge transport, and recombination in ultrathin, conjugated polymer-TiO2 nanocrystal intermixed composites, Phys. Rev. B:
Condens. Matter 1999, 59(16), 10885–
10894.

5.4 Solid State Dye-sensitized Solar Cells – An Alternative Route Towards Low-Cost Photovoltaic Devices

5.4

Solid State Dye-sensitized Solar Cells – An
Alternative Route Towards Low-Cost
Photovoltaic Devices
Udo Bach
NTera Ltd., Dublin, Ireland
5.4.1

Introduction

To date, the ﬁeld of applied photovoltaics is dominated exclusively by conventional inorganic semiconductor technologies such as silicon or thin-ﬁlm solar cells. Since the invention of silicon solar cells in the
1940s, signiﬁcant improvements in device manufacture and the introduction of novel thin-ﬁlm techniques have led to a gradual increase in energy conversion efﬁciencies and made their production more energyand cost-effective. However, classical inorganic p-n junction high-efﬁciency solar cells most probably will always demand the use of extremely pure starting materials and somewhat sophisticated deposition procedures. A very promising alternative to classical, inorganic p-n junction solar cells is the concept of nanoporous, dye-sensitized photoelectrochemical solar cells, which was ﬁrst introduced by Gr¨tzel and coworka ers [1]. In these cells, dye molecules, adsorbed onto a nanostructured TiO2 electrode sensitize photoinduced charge separation across the interface. These cells exhibit external quantum efﬁciencies of up to 10% and can be made from less expensive starting materials and by simple printing techniques [2]. Therefore, they represent a promising alternative to classical inorganic p-n junction solar cells.
Both industry and academia have research efforts toward development and commercialization of dye-sensitized solar cells. However, despite the very promising

performance of these cells, the corrosive and volatile nature of the electrolyte represents an important drawback of this technology. Practical advantages might be gained by the replacement of the liquid electrolyte with a solid charge transport material (CTM). Organic materials and inorganic p -type semiconductors have been tested in this regard. The major focus of this chapter is to describe these concepts and to compare them to other recent ideas for combining new materials and concepts to solid-state photovoltaic junctions.
5.4.2

Solid-state Dye-sensitized Systems

When replacing the liquid electrolyte in dye-sensitized solar cells with a solid p type conductor, a range of requirements have to be fulﬁlled to preserve the high external quantum efﬁciencies common to the original photoelectrochemical devices.
First, the p -type material needs to have electronic levels into which holes can be injected from the oxidized or excited state of the dye. The redox levels of the dye and the p -type material therefore have to be adapted carefully. An intimate contact between the sensitized metal oxide and the p type material is vital to assure fast injection and regeneration processes (Fig. 1). This implies either the growth or deposition of one semiconductor inside a preformed, sensitized porous ﬁlm of its counterpart or the in situ formation of the sensitized composite. Direct formation of the sensitized junction would be appreciable; however, charge collection within the two independent semiconductor networks, in which at least one semiconductor is formed from nanometer-sized inorganic semiconductor particles, demands intimate contact between the particles. Reduced

475

476

5 Dye-Sensitized Photoelectrochemistry
Scheme for the electron transfer processes (injection, regeneration, and recapture) that occur in solid-state dye-sensitized heterojunctions.

Fig. 1

φ (V )

−1 cb 0

inj. (D+/D*)

rec.

hν reg. 1
(D+/D)
TiO2

Dye

HTM

Au

interparticle contact was speciﬁed as a reason for the poor photovoltaic performance of early sensitized TiO2 /polymer composites [3–5]. However, Greenham and coworkers realized a CdSe/MEH-PPV
(polymethoxyethyl hexyloxy-p -phenylenevinylene) solar cell made via simple spin coating techniques, with a maximum incident photon-to-electron conversion efﬁciency (IPCE) of 12%, suggesting efﬁcient electron percolation in the CdSe network [4]. Decapping of the nanoparticles before the formation of the composite was essential for the functioning of the device, a fact that has to be considered in future attempts to form composites with dye-derivatized nanometer-sized particles.
Once a positive charge is injected successfully into the p -type material, it has to be transported efﬁciently to the contacting electrode before interfacial recombination can occur. To allow for such efﬁcient charge collection, both the p - and n-type material need sufﬁcient conductivities so as not to impose high internal resistance and thereby limit charge collection.
Screening of charges by the displacement of ions plays an important role in the charge collection process in photoelectrochemical solar cells. Solid-state dye-sensitized heterojunctions also exhibit increased photovoltaic performance, when ionic species introduced in the form of

lithium salts were added to the phase of organic p -type conductors [6, 7]. It was postulated that the lithium salt might screen space charge layers, which are expected to build up in the illuminated heterojunction. The elimination of the space charge control of the photocurrent could thereby explain the observed increase in photovoltaic performance.
The contacting electrodes also have to be carefully chosen. The cathode material needs to form an ohmic contact to the p -type material so as to avoid injection barriers that would impede efﬁcient charge collection. The anode needs to form a blocking contact to the p -type material so as to avoid internal shortcircuits. Generally, transparent conducting metal oxide layers such as FTO (ﬂuorine doped tin(IV)oxide) or ITO (indium doped tin oxide) are used as the backcontact onto which the nanocrystalline ﬁlm is applied. Iodine is unique, because it shows a very high overpotential at these materials. However, most solid p -type materials will form ohmic contacts with these transparent metal oxides causing internal short circuits at the points of contact [8].
This applies both to organic and inorganic materials. Consequently, whenever transparent conducting metal oxide layers are used in solid heterojunctions, care has to be taken to avoid direct contact between

5.4 Solid State Dye-sensitized Solar Cells – An Alternative Route Towards Low-Cost Photovoltaic Devices

the p -type component of the sensitized heterojunction and the back contact of the photoanode. Organic CTMs Applied to
Nanoporous Dye-sensitized Metal Oxide
Electrodes
Organic electronic materials nowadays are successfully replacing inorganic semiconductor components in a growing number of electronic devices such as xerographic ink drums or light emitting diodes
(LEDs) [9, 10]. Further applications will soon follow including transistors [11], or entire integrated circuits fabricated on plastic [13]. Very recently, these materials also have been scrutinized with regard to their applicability in photovoltaic devices.
In the following section, the application of organic materials in conjunction with dye-sensitized mesoporous metal oxide electrodes is reviewed, whereas alternative conceptual approaches toward organic solar cells that do not directly involve dyesensitization are brieﬂy discussed in the ﬁnal section of this chapter.
Organic synthetic chemistry is a powerful tool used to develop materials with very speciﬁc electronic and optical properties. A large variety of electronically active materials with excellent, reproducible material properties became available over the last few years. These organic materials are still lagging their inorganic counterparts regarding their semiconducting properties such as conductivity and mobility.
Organic materials, however, offer an almost unlimited potential to vary their speciﬁc properties. Furthermore, most of them can be prepared easily in their amorphous state. This is important because crystallization would most probably impair the formation of a good contact between the mesoporous surface of the sensitized
5.4.2.1

metal oxide electrode and the CTM inside the sensitized heterojunction.
The application of organic CTMs in photovoltaic devices is a new research ﬁeld, which still is in its infancy. Figure 2 overviews the organic charge transfer materials that were scrutinized with regard to their application in solid-state dyesensitized solar cells. A much wider range of materials is mentioned in the patent literature [13–15].
The charge transfer events taking place in dye-sensitized heterojunctions are depicted in Fig. 1. Visible light is absorbed by the sensitizing dye. Electron injection from the excited state of the dye into the conduction band of TiO2 (1) is followed by subsequent hole transfer from the photooxidized dye to the organic CTM, regenerating the dye’s original ground state (2).
Dye∗ − − Dye+ + e− (cb TiO2 )
−→
+

Dye + CTM − − Dye + CTM
−→

+

(1)
(2)

The latter process competes with recapture of the injected electron by the oxidized dye (3):
Dye+ + e− (cb TiO2 ) − − Dye
−→

(3)

The double injection process described by Eqs. (1) and (2) generates mobile majority carriers in both the TiO2 and the hole conductor phases, which are subsequently collected at the contact electrodes. A second mechanism can be postulated that involves photoinduced hole transfer from the electronically excited dye to a neighboring charge transport molecule, followed by regeneration of the photoreduced state by electron injection into the TiO2 . Both mechanisms lead to the same initial charge separated state, which results in an electron in the conduction band of the TiO2 and a positive charge, localized on a hole conductor molecule.

477

478

5 Dye-Sensitized Photoelectrochemistry
OMe
H
N
N

N

n
Polypyrrol
MeO

p-OMeTPD
OCH3

N

OCH3
OCH3

OCH3
N
N
OCH3

OCH3
N

OCH3
OCH3
spiro-OMeTAD

Fig. 2

Different CTMs used in solid-state dye-sensitized solar

cells.

It is important to note, that hole conduction to the counter-electrode does not involve mass-transfer. When using lowmolecular-weight CTMs, the underlying mechanism for charge transport generally is assumed to be a hopping mechanism as described by B¨ssler and coworkers [16], a whereas charge transport in conjugated polymers is generally explained via a polaron mechanism [17].
Similar to photoelectrochemical solar cells, the concept of solid-state sensitized heterojunctions has the great advantage over conventional inorganic solar cells that only majority carriers are involved in the photoelectric conversion process. Photoelectric conversion in conventional inorganic p-n junction solar cells involves minority carriers whose lifetime is restricted

because of recombination. Because they are generated throughout the semiconductor and away from the junction, expensive high-purity materials are required to maintain the minority carrier diffusion length at a level at which current losses caused by recombination are avoided.
The ﬁrst to describe a solid-state device based on an organic CTM (p -OMeTPD,
Fig. 2) were Hagen and coworkers [18].
The organic hole conductor was applied to a Ru(dcbpy)2 (SCN)2 -sensitized nanocrystalline TiO2 electrode via thermal evaporation. A thin layer of gold, applied via vacuum deposition techniques, formed the back contact. IPCEs of these devices were still low (IPCE = 0.2%). Space charge effects were found to play a dominating role in this type of devices. The authors

5.4 Solid State Dye-sensitized Solar Cells – An Alternative Route Towards Low-Cost Photovoltaic Devices

attribute the low mobility of electrons in the ﬁnal device responsible for the poor performance. Incomplete pore ﬁlling and low conductivities might have been further reasons for the low efﬁciency, because thermal evaporation is not likely to transport organic material deep into the porous structure and because the charge carrier density inside the hole conductor matrix was low and merely inﬂuenced by accidental doping as a result of the presence of oxygen. The ﬁrst solid-state dye-sensitized heterojunction of TiO2 and a semiconducting polymer was reported by Murakoshi and coworkers [19, 7]. They formed a solid-state heterojunction by photoelectrochemical polymerization of pyrrole into a nanoporous TiO2 ﬁlm, sensitized with Ru(dcbpy)2 (SCN)2 . Direct contact of polypyrrole and the FTO back contact was suppressed by electrodeposition of a thin
TiO2 ﬁlm before the adsorption of the dye onto the nanocrystalline TiO2 electrode.
The doping density of the polymer was controlled electrochemically. Solar energy conversion efﬁciencies were somewhat

higher than previously reported, but still did not exceed 0.1%.
Signiﬁcant improvement in efﬁciency were achieved recently in a novel solar cell device applying the amorphous organic CTM 2,2 ,7,7 -tetrakis(N ,N -di-pmethoxyphenylamine)9,9 -spirobiﬂuorene
[7] (spiro-OMeTAD; Fig. 2). Photoinduced charge carrier generation was shown to be very efﬁcient. A solar cell based on spiro-OMeTAD converts photons to electric current with a strikingly high yield of 50% [20]. The new hole conductor contained a spiro-center, which was introduced to improve the glass-forming properties and prevent crystallization of the material. Its glass-transition temperature of Tg = 120 ◦ C, measured by differential scanning calorimetry, is much higher than that of comparable hole conductors. The methoxy groups were introduced to match the oxidation potential of the
CTM to that of the applied sensitizer
Ru(dcbpy)2 (SCN)2 .
Figure 3 shows a cross-section scheme of the sensitized TiO2 /spiro-OMeTAD heterojunction mentioned earlier. The e− −

+

h+ e− Light

Thin gold electrode
Dye-sensitized nanocrystalline heterojunction Dense TiO2 layer
Transparent, conducting layer
(F-doped SnO2)
Glass support
Organic hole conductor
Dye molecule
Semiconductor particle

Fig. 3 Cross-section scheme for a solid-state dye-sensitized solar cell as described in by Bach and coworkers [6].

479

480

5 Dye-Sensitized Photoelectrochemistry

working electrode consisted of conducting glass (F-doped SnO2 ) onto which a compact TiO2 layer was deposited by spray pyrolysis [21]. This avoids direct contact between the hole conductor and the SnO2 , which would short-circuit the cell. A 4.2µm thick mesoporous ﬁlm of TiO2 was deposited by screen printing onto the compact layer [22] and derivatized with
Ru(dcbpy)2 (SCN)2 . The hole conductor was introduced into the mesopores by spin coating a concentrated solution of
OMeTAD onto the TiO2 ﬁlm and subsequent evaporation of the solvent. A semitransparent gold back contact was evaporated on top of the hole conductor under vacuum.
The conductivity of the device was controlled by partial chemical oxidation of the hole conductor, before deposition onto the sensitized TiO2 substrate to increase the number of free charge carriers [23]. Addition of the lithium salt Li[(CF3 SO2 )2 N] to the spin coating solution of the hole conductor resulted in a strong performance increase in the ﬁnal device. The underlying mechanism remained unidentiﬁed although charge screening due to partial ionic mobility inside the hole conductor matrix and/or the effect of the present lithium ions on the ﬂat band potential of TiO2 [24] were postulated as possible mechanisms. Figure 4 shows the photocurrent action spectra of three typical cells under short circuit conditions [20]. Three different dyes were used to sensitize the heterojunction. In all three cases the action spectrum matches closely the absorptivity spectrum of the sensitizing dye. Solid-state heterojunctions sensitized with the merocyanine dyes (structure see Fig. 4) used in this study showed higher peak IPCE values when compared to Ru(II)L2 (SCN)2 sensitized junctions. Strongly improved

photovoltaic performances were reported recently for comparable systems [25].
Picosecond transient absorption laser spectroscopy was used to scrutinize the dynamics of the photoinduced charge separation process [26]. The dye regeneration process [2] could be time-resolved, and was shown to proceed primarily on the picosecond timescale with multiphasic kinetics.
Therefore it is at least one order of magnitude faster than the dye regeneration process involving iodine or iodide electrolytes [2, 27]. The high hole injection rate insures that the quantum efﬁciency for the overall charge separation process across the junction is close to unity. This may be a key factor behind the high external quantum efﬁciency for charge carrier collection obtained for spiro-OMeTAD compared to other solid-state hole conducting materials studied to date. Transient absorption spectroscopy also gave indications that in the speciﬁc junction, electron injection into
TiO2 proceeds before hole injection into spiro-OMeTAD. Analysis of the front and backside illumination photoresponse spectra proved to be a powerful tool to probe interfacial charge recombination during charge collection [20]. Figure 5 shows a typical example of a set of photoresponse spectra obtained on front and backside illumination, compared to the absorptivity of the sensitized TiO2 layer. At the absorption maximum of the dye, the IPCEfront (illumination through the conducting glass) exhibits a maximum, whereas the IPCEback spectrum (illumination through the semitransparent gold layer) exhibits a minimum. Even when corrected for the transmission losses of the conducting glass and gold electrode, respectively, the IPCE value for backside illumination mode is one order of magnitude smaller than the corresponding value for IPCEfront . This

5.4 Solid State Dye-sensitized Solar Cells – An Alternative Route Towards Low-Cost Photovoltaic Devices

50

IPCE
[%]

40

30

20

10

0
400

Mc2
Mc3
Ru(II)L

2(SCN) 2

500

600

700

800

Wavelength
[nm]
OH

O
O
N
S
S

S

N

O

OH

O
N
S
S

S

N
Mc3

S
O

N
OH
O

Photocurrent action spectra (IPCE) obtained for solid
TiO2 /spiro-OMeTAD heterojunctions, sensitized with three different sensitizers. (For device structure, see Fig. 3.) The dyes were absorbed from above solution for 48 hours at room temperature. All devices are based on
4-µm thick nanocrystalline TiO2 electrodes and spiro-OMeTAD as hole conducting layer (0.070 mol% spiro-OMeTAD++ (PF6 ) in respect to spiro-OMeTAD; the spin coating solution contained 15 mM
Li[(CF3 SO2 )2 N]). The counter electrode was a 10-nm thick gold layer.
Fig. 4

481

5 Dye-Sensitized Photoelectrochemistry
100
80

IPCE
[%]

482

60
40
20
0
400

500

600

700

Wavelength
[nm]
front-IPCE front-IPCEcorr back-IPCE

Fig. 5 Set of photoresponse spectra obtained under front and backside illumination conditions. The dotted curves show the IPCE spectra as measured, whereas the full curves corresponds to the corrected IPCE spectra, which take adsorption and reﬂection losses through the corresponding window materials (FTO glass or semitransparent gold layer, respectively). The top curve represents the absorptivity of the device, which corresponds to a calculated IPCE value assuming no reﬂection or recombination losses.

back-IPCEcorr
(100-T) [%]

ﬁnding strongly suggests that recombination of photogenerated electrons while migrating toward the collecting electrode is a major loss mechanism. Applying a simple recombination model, a free electron diffusion length of about 1 µm could be determined for the observed junction.
Charge recombination occurring during charge collection therefore could be identiﬁed as major loss mechanism of the cell.
A strong increase in energy conversion efﬁciency could be observed when tertiarybutylpyridine was introduced into the matrix of the organic hole conductor [20].
Although the typical open circuit voltage (Voc ) in absence of tertiarybutylpyridine reaches approximately 350 mV, it can easily exceed 700 mV in the presence of tertiarybutylpyridine. This leads to an increase in white light conversion efﬁciency from 0.74% to 1.75% (AM1.5,
10 mW cm−2 ). Very similar effects were reported for classical sensitized I − /I3 − electrolyte/TiO2 junctions [2]. The authors rationalized the effect in terms of 4-tertbutylpyridine adsorption at the TiO2 surface blocking surface states that are active as intermediates in the interfacial charge recombination. They presumed that these

sites are represented by Ti (IV) ions on the surface, which, because of their Lewis acidity, are prone to interact with the pyridine derivative. Similar results were obtained by
Huang and coworkers [28], who attributed the observed increase in Voc to either a charging of surface states or a shift of the conduction band edge.
Different techniques have been proposed to ﬁll sensitized nanoporous structures with organic material. Evaporation [19], electrodeposition [7] and spin coating [6] techniques were described in this respect. Little is known regarding intimacy of the contact between the sensitized metal oxide surface and the organic CTM when simple spincoating procedures are used. Therefore complete pore ﬁlling still remains a critical problem. Although spin coating or dip coating techniques are very simple and inexpensive, electrodeposition might still be a more controlled technique.
In situ polymerization of monomeric precursors appears to be the most reliable way to ﬁll nanoporous electrodes with large-molecular-weight polymers. Very recently it was shown that low-molecularweight organic semiconductors also can be electrodeposited into nanoporous TiO2

5.4 Solid State Dye-sensitized Solar Cells – An Alternative Route Towards Low-Cost Photovoltaic Devices

electrodes [29], which thereby represents a very powerful technique to deposit poorly soluble materials sush as perylenes or phthalocyanines into porous structures.
Many questions regarding the function of these sensitized inorganic-organic hybrid materials still remain unsolved. Most importantly, the factors controlling the electronic properties of the junction have to be identiﬁed and understood. First steps toward a better control of the interface have been made. The dipole ﬁeld, present at the interface because of the self-assembled monolayer of organic molecules proved to strongly affect the junction’s electronic properties [30]. Further systematic studies will hopefully help to widen the understanding of such devices, leading the way to increased device performances.
Inorganic Nanoporous
Dye-sensitized Heterojunctions
In dye-sensitized inorganic heterojunction solar cells, a monolayer of dye is sandwiched between two wide band gap semiconductors; one of them exhibits a p type and the other an n-type conduction mechanism. Inorganic p -type semiconductors were successfully applied in an attempt to replace the liquid electrolyte in photoelectrochemical solar cells, and in fact the ﬁrst solid-state dye-sensitized photovoltaic device described in the literature was based on a wide band gap p -type semiconductor material [31].
Such junctions mainly are of inorganic character, however, the dye present at the junction’s interface still plays the central role in the initial formation of the charge separated state. The sensitizer takes on the role of both the light absorber and acts as the link between the two inorganic materials. The choice of inorganic wide band gap semiconductors applicable in such junctions is small, and only few
5.4.2.2

combinations of wide band gap p -type and n-type semiconductors are energetically favorable. Organic and organo-metallic dyes easily can be adjusted with regard to their redox potentials and absorption characteristics to match the band positions of these materials. In principle, the dye also might be chosen to function as a ﬂexible buffer between the two nanocrystalline phases, reducing strain effects, which arise as a result of the lattice mismatch of the two semiconductor materials. A potential advantage of inorganic semiconductors over their novel organic counterparts is their generally high charge mobility.
However, the very limited choice of potentially interesting materials is a clear drawback, compared to the wide choice of organic CTMs.
Two mechanisms can contribute to the formation of the initial charge separated state. The electronic excitation of the dye either could be followed by electron injection into the n-type semiconductor or by hole injection into the p -type semiconductor. Both mechanisms have been shown to occur and both are likely to contribute to the charge separation mechanism [32].
However, no studies were undertaken to differentiate which of the two possible mechanisms mainly is responsible for the initial formation of the charge separated state. Although dye sensitization of wide band gap n-type semiconductors has been comprehensively studied, little is known regarding the sensitization of p -type materials. Sensitized electron injection into an ntype semiconductor and hole injection into a p -type material are analogous processes.
The misbalance might be explained simply in terms of the wider availability of stable wide band gap n-type materials, which allow for the preparation of nanoporous ﬁlms with excellent electronic properties.

483

484

5 Dye-Sensitized Photoelectrochemistry

Only a few materials were studied concerning their applicability to dye-sensitized hole injection processes. Among those are different copper(I) compounds (e.g.
Cu(I)SCN, Cu(I)I, Cu2 (I)O [33–35]) and nickel(II) oxide [36]. Photovoltaic performances of such devices are orders of magnitudes poorer than those of classical dye-sensitized photoelectrochemical solar cells based on n-type materials. Substantial advantages could arise if an efﬁcient photo–hole injection process would be available. The formation of solid-state tandem solar cells would become feasible, and a quantum step in device efﬁciency of dye-sensitized solar cells could be at reach.
However, because of the poor performance of all known photocathodes, a combination of available photoanodes and photocathodes to a tandem device always results in a device that is photovoltaically less efﬁcient than the photoanode on its own. The concept for electrolyte-based tandem cells exists. However, it contains strong potential to improve the photovoltaic performance in both electrolytic and in solid-state, dyesensitized solar cells.
Analogous to their photoelectrochemical counterparts, efﬁcient dye-sensitized solid-state solar cells require a nanoporous junction structure. The ﬁrst solar cell of this type was reported in 1995 [33] in form of a TiO2 /Cu(I)I heterojunction, sensitized with a cyanidine dye. When forming an inorganic interpenetrating network on the nanometer scale, the crystal lattice mismatch of the two inorganic semiconductors is likely to impair the formation of a good contact between the two nanocrystalline structures. Recently,
O’Regan and coworkers reported on a sensitized nanocomposite of ZnO and CuSCN forming a bicontinuous network heterojunction [37]. Nanoporous ZnO ﬁlms with

a columnar morphology, oriented perpendicular to the substrate surface could be realized, using an electrodeposition technique. These highly anisotropic dyesensitized ﬁlms were shown to form a complete inner surface electrical contact when p -type CuSCN was electrodeposited into the pores (Fig. 6). These junctions exhibited external quantum efﬁciencies of up to 50%, and incident energy conversion efﬁciencies of 1.5%. Dye-sensitized heterojunctions made from colloidal ZnO particles on the other hand, showed maximum external quantum efﬁciencies of only
20%. It could thereby be shown that lattice mismatch is not an intrinsic problem of dye-sensitized inorganic heterojunctions but can be overcome by choosing appropriate device assembly techniques and materials. The sensitizing dye might also exhibit a certain buffer function between the two crystal structures, especially if the dye is substituted with two sets of attachment groups, each of them strongly interacting with one of the semiconductor components. Solid-state dye-sensitized TiO2 /CuI heterojunction were recently realized by Tennakone and coworkers [38]. The authors reported on surprisingly high solar energy conversion efﬁciencies, showing that in the future it might be feasible to form solidstate dye-sensitized solar cells with similar efﬁciencies as their photoelectrochemical counterparts. 5.4.3

Alternative Organic Solid-state Approaches
Organic Solar Cells
Thin ﬁlms of organic pigments are known to exhibit semiconducting properties from their use in xerography; Morel, Ghosh,
Feng and coworkers [39, 40] reported on a solar cell in 1978 based on the Schottky
5.4.3.1

5.4 Solid State Dye-sensitized Solar Cells – An Alternative Route Towards Low-Cost Photovoltaic Devices
Scheme for a sensitized nanocomposite of ZnO and CuSCN, as described by O’Regan and coworkers [37]. (a) Layered structure of the dye-sensitized heterojunction.
(b) Fractured edge of a complete DSH, showing the SnO2 layer, the porous
ZnO/dye/CuSCN layer, and some of the overlying CuSCN. (c) Fractured edge of a porous ZnO layer, without CuSCN, showing the columnar structure perpendicular to the substrate.
(d) Schematic of the ZnO/dye/CuSCN interface with dye molecules shown to scale, but at low packing density.
Fig. 6

SnO2

−
Dense ZnO

Glass support

CuSCN
ZnO + dye +
CuSCN

h+

+

e−
Aluminium

Graphite

Light

(a)
1 µm

junction of aluminum and a merocyanine dye, which showed an overall sunlight conversion efﬁciency of 0.7% – about 35 times higher than any efﬁciency that had been reported on organic ﬁlms before. Following that, interest in organic solar cells strongly increased, particularly driven by the research effort undertaken by companies such as Exxon, Xerox,
Kodak, and Shell in the 1980s. However, white light conversion efﬁciencies never exceeded 1%, so that the general interest in this technology decreased after a number of years. Several comprehensive reviews describing organic solar cells in great indepth were published since [41–44]. A short summary follows, to allow for the comparison with alternative technologies.
In thin-ﬁlm organic solar cells, light is absorbed by a dye layer, typically of a thickness of several tens of nanometers. Light absorption leads to the formation of an exciton (an electron/hole pair, which is strongly bound). These excitons can give rise to free charge carriers by dissociation under the inﬂuence of a strong electric ﬁeld. In general, such ﬁelds develop close to the interface of two materials exhibiting different work functions because of

(b)
100 nm

(c)
20 nm e− ZnO

h+

CuSCN
(d)

• = Dye

the formation of a charge carrier depletion layer. According to the origin of this builtin ﬁeld, one can distinguish three different types of organic solar cells. Most organic photodiodes are based on the rectifying junction of an organic semiconductor and

485

486

5 Dye-Sensitized Photoelectrochemistry

a metal, forming a Schottky type cell such as the cell reported by Morel and coworkers. A rectifying contact can also be formed by the formation of a p-n junction between the light-absorbing, semiconducting dye and a second semiconductor of opposite conduction type. This second semiconductor either can be an inorganic wide band gap semiconductor or a second lightabsorbing, semiconducting dye, giving rise to an organic/inorganic or organic/organic p-n heterojunction.
The main limitation of all three device types is the limited diffusion length of excitons in organic materials, which usually does not exceed several tens of nanometers. Therefore the photovoltaically active region in a thin-ﬁlm solar cell is limited to a very thin layer close to the junction. This usually is too thin to assure good light harvesting over a wide spectral range. Very similar to conventional inorganic p-n junction solar cells, the exciton lifetime strongly decreases in the presence of bulk impurities, demanding highly puriﬁed dyes. Heterojunction formation by two different organic dyes can improve the light harvesting, especially if their absorption spectra are complementary to each other [45, 46]. However, layer thicknesses have to be optimized to prevent optical ﬁlter effects caused by photovoltaically inactive areas away from the junction. Attempts were made to widen the thickness of the photovoltaically active area by incorporation of a coevaporated layer of the two organic, semiconducting device components, resulting in a signiﬁcant increase in device efﬁciency [47].
Another limitation of organic thin-ﬁlm solar cells is the generally poor conductivity of the organic layer. Organic devices commonly show strong sublinearity of the photocurrent at illumination intensities close to AM1.5 (100 mW cm−2 ). However,

control of the charge carrier density via chemical doping with electron acceptors such as iodine for p -type materials [48] or with donors such as H2 for n-type materials [49] can improve drastically the conductivity and device performance.
An interesting approach partially overcoming the problem of low conductivity and limited exciton diffusion widths is to combine two organic p-n junctions in series, leading to a tandem cell with a doubled photovoltage, compared to the simple p-n structure device [50].
Interpenetrating Networks
As described in the previous section, the recombination of an exciton before it reaches the built-in ﬁeld region is the main loss mechanism in organic thin-ﬁlm solar cells.
This problem can be overcome by the formation of an interpenetrating p-n heterojunction in which the average width of the p -type and n-type domains are smaller than the exciton diffusion length, thereby guaranteeing that every exciton can reach the interface before it decays.
However, the role of the electric ﬁeld in these devices is questionable because the theoretical depletion layer width generally is much wider than the dimension of phase segregation. It might be more exact to assume the totality of the interpenetrating network to be depleted with exciton splitting occurring via an interfacial electron transfer mechanism. Similar to thin-ﬁlm solar cells, the rectifying junction can be formed by two organic semiconductors, or by a blend of an organic and an inorganic semiconductor. Semiconducting polymers are superior to their low-molecular-weight analogs, with regard to their applicability in interpenetrating junctions. This is due to the fact that charge transport in semiconducting polymers mainly proceeds along the
5.4.3.2

5.4 Solid State Dye-sensitized Solar Cells – An Alternative Route Towards Low-Cost Photovoltaic Devices

polymer chain, so that percolation starts already at low concentration thresholds (for a review on polymer solar cells see Ref. 51).
Poly(p -phenylenevinylene) derivatives
(PPVs) are the most prominent examples of polymers used in organic solar cells.
Depending on their substituents they either can exhibit p -type or n-type behavior.
The ﬁrst junction of this type was reported in 1995, using a phase-segregated thin ﬁlm of MEH-PPV as p -type material and CN-PPV (cyano-p -phenylenevinylene) as n-type material [52]. The phase segregation was on the scale of 10–100 nm and
IPCE values as high as 6% were recorded.
Recently, signiﬁcantly improved device performances were reported based on a polythiophene/PPV-derivative heterojunction, assembled via a lamination technique with IPCE values reaching 29% [53]. The cross-section scheme and the molecular structures can be seen in Fig. 7.
A different approach to form organic interpenetrating heterojunctions is to dope strongly a p -type conducting polymer with a low-molecular-weight acceptor until charge percolation occurs inside the acceptor phase. The blends attracting most attention are those of
PPV- or thiophene derivatives doped with fullerenes as acceptor, reaching 23% IPCE for polythiophene/C60 junctions [54] and
29% for PPV/C60 junctions [55]. Perylene nanocrystals were also used as doping acceptor [56]. Different theoretical models were developed to describe the functioning principals of these devices [57, 58].
Nonsensitized Inorganic/Organic
Junctions
Recently, increased interest was raised in the photovoltaic properties of composites of semiconducting polymers and nanosized semiconductor particles, in which either the polymer or the semiconductor
5.4.3.3

and the polymer take over the function of the light absorber. Among those, nanocrystalline heterojunctions of TiO2 either with
PPV [59] or polythiophene [60] derivatives have received particular interest. Several different techniques were applied to form the heterojunctions. Commonly the ﬁrst step was the formation of a nanoporous thin ﬁlm of TiO2 via spray pyrolysis or via deposition of a colloidal paste followed by a sintering step while the polymer was applied via a spincoating technique or (in the case of polythiophene) via electrodeposition. IPCE values reported on these systems are still very low (below 3%).
However, just recently, highly efﬁcient
TiO2 /polythiophene solar cells exhibiting peak IPCE values of up to 10% were accomplished [61]. Photovoltaic cells of comparable efﬁciency were obtained when polycrystalline ﬁlms of low band gap semiconductors such as CdSe, CdS, or CuInSe2 are contacted with a thin ﬁlm of polythiophene via electrodeposition [62]. Overall white light conversion efﬁciencies of 1.3% at 56 mW cm−2 were reported for a nanostructured CdSe/poly(3-methylethiophene) junction [63].
One particularly interesting aspect of nanoparticle/polymer solar cells is that charge ‘‘hopping’’ from one semiconductor particle to another takes place even in blends prepared from spin coating solutions containing both the polymer and the colloidal nanoparticles. Usually, those devices are not exposed to successive heat treatments. This is surprising because nanocrystalline semiconductor ﬁlms used in photoelectrochemical solar cells generally are believed to need such a heat treatment to induce interparticle contact and ‘‘necking.’’ Less expensive, ﬂexible dye-sensitized solar cells based on plastic substrates could be realized if this heat treatment could be made redundant.

487

488

5 Dye-Sensitized Photoelectrochemistry

O

O

CN n NC
S

n

O

O

MEH-CN-PPV

POPT
(a)
AI or Ca

Glass substrate

Acceptor layer
Donor layer
ITO or
PEDOT on Au
Glass substrate
(b)

Chemical structures and device structure for an organic
..
polymer photodiode as described by Granstrom and coworkers [53]. (a) Chemical structures of the polymers used in the devices. POPT is a regioregular phenyl-octyl substituted polythiophene. MEH/CN-PPV is a cyano-derivative of poly(p-phenylene vinylene), with a large electron afﬁnity as a result of the electron-withdrawing cyano groups. (b) Device structure.
Both polymers (see (a) for chemical structures) were dissolved in chloroform or toluene (5 mg ml−1 ). For the top half of the device, aluminum or calcium contacts were thermally evaporated on glass substrates and the acceptor material, MEH/CN-PPV (and a small amount of POPT, usually 5%), was spin-coated on top of the metal electrode. The other half of the device comprises the donor material, POPT (and a small amount of MEH/CN-PPV, usually
5%), either spin-coated on ITO substrates or glass-coated with poly(ethylene dioxythiophene) doped with polystyrene sulfonic acid. To ensure a low contact resistance, a thin layer of gold
(∼10 nm) was thermally evaporated on the glass slide before spin-coating the PEDOT ﬁlm from a water solution. The thickness of the PEDOT layer was 100 nm. To get the desired structure of the
POPT, this half of the device was heated to 200 ◦ C under vacuum before the device was laminated together by applying a light pressure while one half was still at elevated temperature. The total thickness of the organic semiconductive layer was 70–80 nm, and the active area 2.5 mm2 .
Fig. 7

5.4 Solid State Dye-sensitized Solar Cells – An Alternative Route Towards Low-Cost Photovoltaic Devices

Various examples of composite nanoparticle or polymer ﬁlms made via spin coating can be found in the literature [3,
4, 64]. Most examples actually show that charge percolation inside the nanoparticle network is reduced strongly, compared to nanocrystalline, sintered ﬁlms.
Arango and coworkers [3] reported a decrease in photoconductivity of 2 orders of magnitude when comparing a sintered polymer-coated TiO2 ﬁlm, to a composite made via spincoating of a
TiO2 -colloid/polymer solution. The reduced interparticle contact also leads to low photovoltaic performances. An exception to these observations is the CdSe/PPV composite solar cell reported by Greenham and coworkers [38]. Peak IPCE values of 12% were obtained from spin-coated composite thin ﬁlms with a weight ratio of CdSe (φ 5 nm): MEH-PPV = 9 : 1
(ﬁlm thickness = 300–500 nm). Removal of surface-adsorbed species from the colloid before the cell assembly was shown to be crucial for interparticle electron transfer to occur.
5.4.4

Conclusion and Outlook

A large number of different approaches toward solid-state dye-sensitized or organic solar cells were recently described in the literature, and some of these are brieﬂy described in this chapter. The efﬁciencies of these novel prototype devices are constantly rising, some of them showing energy conversion efﬁciencies of up to 2%.
Such concepts not only are of scientiﬁc but also of potentially economic interest, indicating that the realization of less expensive, efﬁcient plastic solar cells might soon be at hand. These recent developments strongly beneﬁt from the continued success in the synthesis of novel organic electronic

components. Those organic materials have found their way into ﬁelds, which until recently were purely dominated by inorganic device physics. Organic LEDs, for example, are already outperforming their inorganic counterparts in certain types of applications. Very often the pioneering work that was done to realize those novel applications ﬁnally gave access to materials that also showed very interesting photovoltaic properties. Examples are
PPV derivatives developed for their use in organic LEDs [51] and pentacene single crystals studied by Lucent Technologies for use in organic transistors [11]. It was only at a later stage that interesting photovoltaic properties of such materials were scrutinized [52–55, 65]. The organic compounds used in these early studies were poorly optimized with regard to their photovoltaic properties; nonetheless, those initial prototypes showed impressive performances.
Triggered by these initial promising results, new efforts currently are under way to develop tailor-made organic materials with improved optical and electronic properties for their use in solar cells.
These second-generation organic photovoltaic materials almost necessarily will lead to signiﬁcant improvements in device performances. Along with the development of novel materials, there is the need to gain a better understanding of the different concepts described in this chapter.
Most of these are only poorly understood, when compared to well-established systems such as photoelectrochemical solar cells. Nevertheless, very exciting advancements were realized over the last years and the ﬁeld of organic photovoltaics is steadily progressing. Organic solar cells have the potential for very low production costs at reasonable high efﬁciencies.
These properties make them interesting for their application in mobile electronic

489

490

5 Dye-Sensitized Photoelectrochemistry

consumables, in which high efﬁciencies and long-term stability in extreme environments are not an ultimate requisite. Therefore, the predescribed concepts emerge as a very interesting and viable option for future low-cost plastic solar cells.
Acknowledgment

Greg Smestad is gratefully acknowledged for proofreading this chapter.
References
1. B. O’Regan, M. Gr¨tzel, Nature 1991, 353, a 737–739.
2. M. K. Nazeeruddin et al., J. Am. Chem. Soc.
1993, 115, 6382–6390.
3. A. C. Arango, S. A. Carter, P. Brook, J. Appl.
Phys. Chem. 1999, 74, 1698–1700.
4. N. C. Greenham, X. Peng, A. P. Alivisatos,
Phys. Rev. B 1996, 54, 17628.
5. M. Kocher, T. K. D¨ubler, E. Harth et al., a Appl. Phys. Lett. 1998, 72, 650–652.
6. U. Bach, D. Lupo, P. Compte et al., Nature
1998, 395, 583–585.
7. K. Murakoshi, R. Kogure, Y. Wada et al., Sol.
Energy Mater. Sol. cells 1998, 55, 113–125.
8. U. Bach, Solid-state dye-sensitized mesoporous TiO2 solar cells, 2000, Ph.D. thesis
N◦ 2187, EPFL Lausanne, Switzerland.
9. G. Hadziioannou, P. F. van Hutten in Semiconducting polymers, Wiley-VCH, Weinheim,
2000.
10. K. M¨ llen, G. Wegner in Electronic Materials: u The Oligomer Approach, Wiley-VCH, Weinheim, 1998.
11. C. D. Dimitrakopoulos, S. Purushothaman,
J. Kymissis et al., Science 1999, 283, 822–824.
12. B. Crone, A. Dodabalapur, Y.-Y. Lin et al.,
Nature 2000, 403, 521–523.
13. D. Lupo, J. Salbeck 1999, US5885368.
14. L. Hausslinger et al., 1997, US5683833.
15. K. Shiratsuchi, H. Takizawa 2000, US
6084176.
16. P. M. Borsenberger, L. Pautmeier, H.
B¨ssler, J. Chem. Phys. 1991, 94, 5447–5454. a 17. M. N. Bussac, L. Zuppiroli, Phys. Rev. 1996,
B54, 4674.
18. J. Hagen, W. Schaffrath, P. Otschik et al.,
Synth. Met. 1997, 89, 215–220.

19. K. Murakoshi, R. Kogure, S. Yanagida, Chem.
Lett. 1997, 5, 471–472.
20. U. Bach, J. Kr¨ ger, M. Gr¨tzel, SPIE proceedu a ings San Diego, San Diego, Calif., 2000, in print. 21. L. Kavan, M. Gr¨tzel, Electrochim. Acta 1995, a 40, 643–652.
22. C. J. Barb´ et al., J. Am. Ceram. Soc. 1997, 80, e 3157–3171.
23. M. Abkowitz, D. M. Pai, Philos Mag. B 1986,
53, 193–216.
24. B. Enright, G. Redmond, D. Fitzmaurice,
J. Phys. Chem. 1994, 97, 1426–1430.
25. J. Kr¨ ger, R. Plass, L. Cavey et al., Appl. Phys. u Lett. 2001, 79, 2085–2087.
26. U. Bach, Y. Tachibana, J.-E. Moser et al.,
J. Am. Chem. Soc. 1998, 121, 7445–7446.
27. S. A. Haque, Y. Tachibana, D. R. Klug et al.,
J. Phys. Chem. B 1998, 102, 1745–1749.
28. S. Y. Huang, G. Schlichth¨ rl, A. J. Notzik o et al., J. Phys. Chem. B 1997, 101, 2576–2582.
29. A. Zaban, Y. Diamant, J. Phys. Chem. B 2000,
104, 10043–10046.
30. J. Kr¨ ger, U. Bach, M. Gr¨tzel, Adv. Mater. u a
2000, 12, 447–451.
31. K. Tennakone, K. P. Hewaparakkrama,
M. Dewasurendra et al., Semicond. Sci.
Technol. 1988, 3, 382–387.
32. B. O’Regan, D. T. Schwartz, Semicond. Chem.
Mater. 1998, 10, 1501–1509.
33. K. Tennakone, G. R. R. A. Kumara, A. R.
Kumarasinghe et al., Semicond. Sci. Technol.
1995, 10, 1689–1693.
34. C. A. N. Fernando, Sol. Energy. Mater. 1993,
30, 211–220.
35. C. R. N. Fernando, Sol. Energy. Mater. 1992,
28, 255–271.
36. J. He, H. Lindstr¨ m, A. Hagfeldt et al., o J. Phys. Chem. B 1999, 103, 8940–8943.
37. B. O’Regan, D. T. Schwartz, M. S. Zakeerudin et al., Adv. Mater. 2000, 12,
1263–67.
38. K. Tennakone, G. R. A. A. Kumara, I. R. M.
Kottegoda et al., J. Phys. D: Appl. Phys. 1998,
31, 1492–1496.
39. A. K. Ghosh, T. Feng, J. Appl. Phys. 1978, 49,
5982–5989.
40. D. L. Morel, A. K. Ghosh, T. Feng et al., Appl.
Phys. Lett. 1978, 32, 495–497.
41. G. A. Chamberlain, Sol. Cells 1983, 8, 47.
42. G. Horowitz, Adv. Mater. 1990, 2, 287.
43. D. W¨ hrle, D. Meissner, Adv. Mater. 1991, 3, o 129.

5.4 Solid State Dye-sensitized Solar Cells – An Alternative Route Towards Low-Cost Photovoltaic Devices
44. J. Simon, J. J. Andr´ , in Molecular Semicone ductors, Springer-Verlag, Berlin, 1985.
45. C. W. Tang, Appl. Phys Lett. 1986, 48, 183.
46. K. Kudo, T. Moriizumi, Jpn. J. Appl. Phys.
1981, 20, L553–L556.
47. M. Hiramoto, H. Fujiwara, M. Yokoyama,
J. Appl. Phys. 1992, 72, 3781–3787.
48. K. Yamashita, Y. Harima, T. Matsubayashi,
J. Phys. Chem. 1989, 93, 5311–5315.
49. M. Hiramoto, Y. Kishigami, M. Yokoyama,
Chem. Lett. 1990, 119–122.
50. M. Hiramoto, M. Suezaki, M. Yokoyama,
Chem. Lett. 1990, 327–330.
51. G. Hadziioannou, P. F. van Hutten, in Semiconducting Polymers, Wiley-VCH, Weinheim,
2000.
52. J. J. M. Halls, C. A. Walsh, N. C. Greenham et al., Nature 1995, 376, 498–500.
53. M. Granstr¨ m, K. Petrisch, A. C. Arias et al., o Nature 1998, 395, 257–260.
54. L. S. Roman, W. Mammo, L. A. A. Pettersson et al., Adv. Mater. 1998, 10, 774–777.
55. G. Yu, J. Gao, J. C. Hummelen et al., Science
1995, 270, 1789.

56. J. J. Dittmer et al., Sol. Energy Mater. Sol. Cells
2000, 61, 53–61.
57. A. K¨ hler, D. A. D. Santos, D. Beljonne et al., o Nature 1998, 392, 903–906.
58. L. A. A. Pettersson, L. S. Roman, O. Ingan¨s, a J. Appl. Phys. 1999, 86, 487–496.
59. T. J. Savenije, J. M. Warman, A. Goossens,
Chem. Phys. Lett. 1998, 287, 148–153.
60. K. Kajihara, K. Tanaka, K. Hirao et al.,
Jpn. J. Appl. Phys. 1997, 36, 5537–5542.
61. S. Spiekermann, S. Mayer, W. Blau et al., unpublished results.
62. S. A. Gamboa, H. Nguyen-Cong, P. Chartier et al., Sol. Energy Mater. Sol. Cells 1998, 55,
95–104.
63. P. Chartier, H. N. Cong, C. Sene, Sol. Energy
Mater. Sol. Cells 1998, 52, 413–421.
64. J. S. Salafsky, W. H. Lubbrthuizen, R. E. I.
Schropp, Chem. Phys. Lett. 1998, 290,
297–303.
65. J. H. Schoen, C. Kloc, E. Bucher et al., Nature
2000, 403, 408–410.

491

497

6.1

Fundamentals of Photocatalysis
Akira Fujishima and Donald A. Tryk
The University of Tokyo, Tokyo, Japan
6.1.1

Introduction

Remarkable progress has been made in the area of photocatalysis in almost 25 years since the earliest work appeared [1–3]. It was recognized that TiO2 is a highly suitable material for photocatalysis because of its complete resistance to photocorrosion. This virtue of TiO2 had been recognized only a few years earlier in the work on photoelectrochemical (PEC) water splitting [4–7]. Other virtues include the low cost and nontoxicity of TiO2 . These have allowed TiO2 to be used in an increasing number of applications involving both indoor and outdoor air puriﬁcation, water puriﬁcation, self-cleaning surfaces, and antifogging surfaces. This introductory chapter focuses on the basic aspects of photocatalysis.
6.1.2

Background

As outlined by Heller [8], the extensive work that was carried out in the ﬁeld of photoelectrochemistry during the 1960s,

1970s, and 1980s provided an excellent basis for work on photocatalysis. Each of the basic principles related to photoelectrochemistry is also involved in photocatalysis. To a ﬁrst approximation, the framework of photocatalysis diverges from other
PEC effects in two aspects: (1) scale, that is, micrometer or nanometer instead of millimeter; and (2) in most cases, lack of an external electrical circuit. The latter situation is analogous to that of electrochemical corrosion, in which anodic and cathodic electrochemical reactions short-circuit locally on a conductive metal surface. In the case of photocatalysis, a PEC reaction, usually an oxidation reaction, is short-circuited with a purely electrochemical reduction reaction, for example, that of molecular oxygen, as discussed later in detail in this chapter. In a solar PEC cell, photogenerated electron-hole pairs are driven efﬁciently in opposite directions by an electric ﬁeld existing at the interface between a semiconductor and an electrolyte solution.
The same processes can occur on a single semiconductor particle, although the electric ﬁeld may be vanishingly small [9, 10]. Thus, the carriers must be separated by means of diffusion, with nanoscale sources and sinks. In a system where semiconductor particles are suspended in a liquid solution, excitation

498

6 Nonsolar Energy Applications

of the semiconductor can lead to redox processes in the interfacial region around each particle. These types of systems have drawn the attention of a large number of investigators over the past twenty years.
For example, in the area of photocatalytic water splitting, there has been a very active effort, as reviewed recently [11, 12].
In the early work on PEC solar cells for water electrolysis, one of the ﬁrst types of electrode materials examined was
TiO2 , partly because it has a sufﬁciently positive valence band edge to oxidize water to oxygen [4–7]. It is also an extremely stable material in the presence of aqueous electrolyte solutions. The possibility of solar photoelectrolysis was demonstrated for the ﬁrst time with a system in which an immersed n-type TiO2 semiconductor electrode, connected through an electrical load to an immersed platinum black counter electrode, was exposed to nearUV (ultraviolet) light. When the surface of the TiO2 electrode was irradiated with light consisting of wavelengths shorter than ∼415 nm, photocurrent ﬂowed from the platinum counter electrode to the
TiO2 electrode through the external circuit.
The direction of the current revealed that the oxidation reaction (oxygen evolution) occurs at the TiO2 electrode and the reduction reaction (hydrogen evolution) at the Pt electrode.
When the conduction band energy, ECB , is higher (i.e. closer to the vacuum level or more negative on the electrochemical scale) than the hydrogen evolution potential, photogenerated electrons can ﬂow to the counter electrode and drive the reduction of protons, resulting in hydrogen gas evolution without an applied potential. Hence, for this to occur, ECB should be at least as negative as −0.4 V
[standard hydrogen electrode (SHE)] in

acid solution or −1.2 V (SHE) in alkaline solution. Among the oxide semiconductors, TiO2 , SrTiO3 , CaTiO3 , KTaO3 ,
Ta2 O5 , and ZrO2 satisfy this requirement, although in most cases the band gaps are relatively large (>3.0 eV), and thus the efﬁciency of solar energy conversion is very low.
Following early TiO2 -induced water splitting studies, it was discovered that
TiO2 particularly excels in photocatalytically breaking down organic compounds and other pollutants to innocuous products, neutralizing foul odors, and killing bacteria. Frank and Bard in 1977 were among the ﬁrst to examine the possibilities of using TiO2 to decompose pollutants; in their case, cyanide, in water [2, 3]. Since then, there has been an increasing interest in environmental applications. These authors quite correctly pointed out the implications of their result in the ﬁeld of environmental puriﬁcation. Their prediction has indeed been borne out, as evidenced by the extensive international efforts in this area [8, 12–23].
6.1.3

TiO2 Characteristics
General
Titanium dioxide occurs in nature in three crystalline forms, rutile, anatase, and brookite. The crystal structures are shown in Fig. 1 [24]. Some of the physical properties are given in Table 1. Much of the fundamental work on TiO2 has been carried out with rutile single crystals because these are relatively straightforward to produce. Anatase single crystals are more difﬁcult to produce, and some work has been carried out with natural crystals but it is possible to grow such crystals with the use of the vapor-phase transport method [25]. Photocatalytic reactions
6.1.3.1

6.1 Fundamentals of Photocatalysis c TiO2 crystal structures:
(a) rutile; (b) anatase;
(c) brookite. Structure data were taken from Wyckoff [24].

Fig. 1

b

a
(a)
c

a

b

(b) b (c) c

have been shown to exhibit larger rates on anatase compared to rutile, and it will be very useful to carry out further comparative studies with single crystals.
Band Edge Energies
The energies of the valence and conduction bands of TiO2 are determined by the chemical termination of the material, together with the interactions of the
6.1.3.2

a

surface with gas or liquid environment.
The electrochemical ﬂat band potentials,
Efb , can be used to estimate the band edge energies for titanium dioxide. Efb values for rutile single crystals have been reported by several authors [26]. In acid solution
(pH 2.1), a value of −0.05 V versus SHE has been proposed by Morrison, based on the results of several workers [27]. In alkaline solution, values of −0.6 [27] and

499

500

6 Nonsolar Energy Applications
Tab. 1

Selected properties of TiO2

Property
Crystal structure
Space group
Density, g cm−3
Band gap, eV
Type

Rutile

Anatase

Tetragonal
P42/mnm
(136)
4.250
3.02
Indirect

Tetragonal
I41/amd
(141)
3.894
3.2
Indirect

Note: Data taken from Finklea [24] and
Wyckoff [26]

−0.76 V [28] have been given. Recently,
Kavan and coworkers have measured Efb values for both rutile and synthetically grown anatase single crystals in 1 M
H2 SO4 and obtained values of +0.04 and
−0.16 V against SHE, respectively [25].
The authors point out that the latter value is sufﬁciently negative so that no external bias needs to be applied to photoelectrochemically evolve hydrogen.
On the basis of the ﬂat band potential, the actual location of the conduction band edge energy, ECB , then depends on the doping level, according to the following equation: n
ECB − EF = ξ = −kT ln
(1)
NC where EF is the Fermi level energy, n is the occupation of conduction band states with electrons, and NC is the density of conduction band states at the band edge [29]. For normal semiconductor doping levels, ξ could be in the range of 0.1 to 0.2 V [27]. In subsequent discussions we will refer to the Efb values, which should then be considered to be lower (energy) limits for the ECB values.
As already stated, the band positions depend on the environment external to the solid surface, for example, the hydrogen ion concentration or pH. This is a situation common to all oxide semiconductors and

arises from the fact that the surface oxygens of the oxide interact strongly with protons and hydroxide ions [27]. To understand the processes underlying photocatalysis on TiO2 , it is useful to take the pH dependencies of the band edge energies into perspective. The relationships of these energies to those corresponding to various redox couples, for example, those for various forms of reduced oxygen (see next section) or for various organic oxidations, then become apparent (Fig. 2).
Here we have adapted the commonly used
Pourbaix representation in a modiﬁed form (i.e. with negative potentials uppermost), to be consistent with the prevailing practice in energy band diagrams [30].
Therefore, the ubiquitous lines for the
H+ /H2 and O2 /H2 O redox couples are inverted and slope upward with increasing pH (more negative electrochemical potentials). Likewise, the valence band (VB) and conduction band (CB) energies for TiO2 slope upward (higher electron energies) with increasing pH.
The VB energies lie approximately
1.8 eV below those necessary to evolve molecular oxygen from water over the entire pH range. Electrons in the CB for rutile are very close in potential to that necessary to evolve molecular hydrogen from water but not at appreciable rates. In early TiO2 -based PEC solar cells, this problem was circumvented through the use of a two-compartment cell. The TiO2 electrode was immersed in a more alkaline half-cell and a Pt counter electrode was immersed in a more acid electrolyte in the second half-cell. The electrons generated at the photoelectrode diffuse at the back of the electrode and then toward the counter electrode. In this case, electrons photogenerated at the alkaline TiO2 electrode have a relatively high electrochemical energy, which is sufﬁcient to evolve hydrogen in

6.1 Fundamentals of Photocatalysis
−1

TiO2 band edge energies. Data were taken from
Finklea [26] and from the data given in Table 2.

Fig. 2

TiO2 CB

Potential
[V vs. SHE]

0

H+/H2
O2/O2−
O2/H2O2

O2/HO2−

O2/H2O

+1

O−/OH−
+2
OH/H2O
TiO2 VB

+3

0

2

4

6

8

10

12

14

pH

the more acidic solution. In the case of photocatalysis, the same type of effect may also be operative: speciﬁcally, if the local pH is high, for example, at a reducing site at a particular location on a TiO2 particle, additional reducing power is available for the reduction of molecular oxygen to superoxide. In contrast, the oxidizing power of photogenerated holes decreases with increasing pH but the standard potential for the one-electron oxidation of any given organic compound RH to its radical cation
•
RH+ is not pH-dependent. For example, the standard potential for this reaction for n-octane in acetonitrile is +2.71 V against
Ag/Ag+ [34], which can be converted to
+2.84 V against SHE. This value may not be sufﬁciently positive because solvent reactions make elucidation of the reversible potential problematic. At present,

the topic of the actual energetics of various oxidation reactions pertinent to photocatalysis is beginning to be explored in detail [35]. For example, because of the presence of water, the actual reaction could be considered to be an electron transfer followed by an extremely fast chemical reaction with water.
R − − • R+ + e−
−→
•

R+ + H2 O − − • RH + OH−
−→

(2)
(3)

The effective redox potential of this reaction would be shifted signiﬁcantly in the negative direction because of consumption of the oxidized product. Thus, it is quite possible that trapped holes with redox potentials of +1.5 V [36] to +1.75 [37], have sufﬁcient oxidizing power to induce the production of the neutral radical, which would involve Eqs. (2) and (3).

501

502

6 Nonsolar Energy Applications
Standard redox potentials for various electrochemical reactions involving molecular oxygen and its reduced forms

Tab. 2

Reaction

Acid dissociations
HO•
2
H2 O 2
•
OH
Redox reactions
O2 + e−
O2 + H+ + e−
O2 + H+ + e−
O2 + 2H+ + 2e−
O2 + 2H+ + 2e−
O2 + H+ + 2e−
O2 + H+ + 2e−
O2 + 4H+ + 4e−
O2 + 4H+ + 4e−
O2 + 4H+ + 4e−
•
O2 − + 2H+ + e−
HO• + e−
2
HO• + H+ + e−
2
H2 O2 + H+ + e−
H2 O2 + 2H+ + 2e−
•
OH + H+ + e−
•
OH + H+ + e−
•
OH + e−
•−
O + 2H+ + e−
•−
O + 2H+ + e−

pH

E0
[V]

G
[eV]

•

←−→
←−→
←−→
←−→
←−→
←−→
←−→
←−→
←−→
←−→
←−→
←−→
←−→
←−→
←−→
←−→
←−→
←−→
←−→
←−→

•

O2 −
HO•
2
HO•
2
H2 O 2
H2 O 2
HO2 −
HO2 −
2H2 O
2H2 O
2H2 O
H2 O 2
HO2 −
H2 O 2
•
OH + H2 O
2H2 O
H2 O
H2 O
OH−
H2 O
H2 O

all
0
7
0
7
14
7
0
7
14
7 all 7
7
7
0
7
14
7
14

−0.284
+0.005
−0.409
+0.693
+0.279
−0.065
+0.142
+1.229
+0.815
+0.401
+0.842
+0.694
+0.967
+0.433
+1.351
+2.683
+2.269
+1.857
+2.559
+1.733

References

4.88
11.63
11.9

O 2 − + H+
HO2 − + H+
•−
O + H+

←−→
←−→
←−→

pKa

31
31
32

+0.284
−0.005
+0.409
−1.386a
−0.558a
+0.130a
−0.284a
−4.916a
−3.260a
−1.603a
−0.558a
−0.558a
−0.991a
−3.260a
−2.233b
−0.991b
+0.252b
−0.701b
+0.128b

33 c c c c

33 c 33 c c c c c c c 32
32
32 c c

a The

free energy is for the product. free energy is for • OH or • O− . c Calculated.
Note: Data were taken from Hoare [31], Buxton and coworkers [32], and Tarasevich and coworkers [33]. b The

Solely considering Eq. (2), it would be expected that the reaction of a hole in the TiO2 valence band with the organic compound would become less favorable at high pH. However, considering Eq. (3), a normal pH dependence, that is, about
60 mV/decade at 25 ◦ C should exist. Overall, it might be expected that there should not be signiﬁcant pH effects in terms of changes in excess oxidizing power of valence band holes or trapped holes. In fact, the rates of photocatalytic oxidation

reactions are often higher at higher pH [18]. For organic oxidation reactions in aqueous solution, not involving radical species, standard redox potentials are known (Table 3) and many exhibit pH dependencies of about 60 mV pH−1 , although the behavior can be somewhat more complex, with at least one change in slope [38] because of involvement of acid-base equilibria.
The majority of the VB and CB values
(the latter being approximated by Efb as

that the actual G0 values are more negative than those given. from values versus ferrocene/ferrocinium [135]. c converted from values versus Ag/Ag+ (0.1 M) [135].

b converted

a Indicates

0
0
0
0
0
0
0
0
NA
NA
NA
NA
NA
NA
NA

CO(g) + H2 O
HCOOH(aq)
HCOH(aq) + H2 O
CH3 OH(aq) + H2 O
CH4 (g) + 2H2 O
CH3 OH(aq) + H2 O
CH3 OH(aq)
HCHO(aq) + H2 O
CH3 OH(CH3 CN)
CH3 CH2 OH(CH3 CN)
2-propanol(CH3 CN) n-pentane(CH3 CN) n-hexane(CH3 CN) n-heptane(CH3 CN) n-octane(CH3 CN)

CO2 (g) + 2H+ + 2e−
CO2 (g) + 2H+ + 2e−
CO2 (g) + 4H+ + 4e−
CO2 (g) + 6H+ + 6e−
CO2 (g) + 8H+ + 8e−
HCOOH(aq) + 4H+ + 4e−
HCHO(aq) + 2H+ + 2e−
HCOOH(aq) + 2H+ + 2e−
•
CH3 OH + (CH3 CN) + e−
•
CH3 CH2 OH + (CH3 CN) + e−
•+
2-propanol (CH3 CN) + e−
•
n-pentane + (CH3 CN) + e−
•+
n-hexane (CH3 CN) + e−
•
n-heptane + (CH3 CN) + e−
•+
n-octane (CH3 CN) + e−
←−→
←−→
←−→
←−→
←−→
←−→
←−→
←−→
←−→
←−→
←−→
←−→
←−→
←−→
←−→

pH

Selected free energies and standard potentials for CO2 reduction reactions

Reaction

Tab. 3

+20.06
+38.32
+27.49
−17.24
−130.62
−55.96
−44.77
−10.81
> −268.23a
> −256.65a
> −246.04a
> −281.74a
> −277.88a
> −274.02a
> −274.02a

G0
[kJ mol−1 ]
−0.104
−0.199
−0.071
+0.030
+0.169
+0.145
+0.232
+0.056
> +2.78b
> +2.66b
> +2.55b
> +2.92c
> +2.88c
> +2.84c
> +2.84c

E0 , V vs. SHE

39
38, 40
38, 40
38, 40
39
40
40
40
41
41
41
34
34
34
34

References

6.1 Fundamentals of Photocatalysis
503

504

6 Nonsolar Energy Applications

already discussed) shown in Fig. 2 are for rutile single crystals. It has already been pointed out that the crystalline form more commonly used in photocatalysis, anatase, has a higher CB energy (more negative electrochemical potential). A point is included for single crystal anatase, approximately 200 mV above the line drawn through the points for rutile, and a point is also shown for a nanocrystalline anatase thin ﬁlm: its CB energy is elevated approximately 100 mV with respect to that for rutile. Lying approximately 0.6 to 0.8 eV below the CB energy, there are surface states on TiO2 ; three experimental points are included in Fig. 2 at speciﬁc pH values but additional experimental evidence exists for such surface states under conditions with no pH control, for example, in the gas phase, and therefore such data cannot be included in the ﬁgure. A reasonable characterization for this type of surface state is that it involves a trapped electron with an accompanying proton, viz.:
H
O
Ti (IV)

O
Ti (IV)

Ti (IV)

Ti (III)

(4)
This type of species is important for photocatalysis because it provides a surface sink for photogenerated electrons. Additionally, this type of species is particularly important with respect to the wetting properties of TiO2 , which is of importance to self-cleaning surfaces.
6.1.4

Photocatalytic Reactions
Oxygen Electrochemistry
The active chemical species that are involved in photocatalytic reactions are
6.1.4.1

generally forms of reduced oxygen, including superoxide • O2 − , the hydroperoxyl radical • O2 H, hydrogen peroxide H2 O2 , the peroxide anion HO2 − , the hydroxyl radical • OH, and the oxyl radical anion
•−
O . The relevant reactions involving these species are shown in Table 2. The pH dependences of these reactions are shown in the same ﬁgure with the TiO2 band edge energies (Fig. 2). It is interesting that the reducing power of the conduction band electrons is barely sufﬁcient to reduce O2 to • O2 − at pH values below the pKa for • O2 H (4.88). With increasing pH, however, there is an increasing excess of reducing power for this reaction.
The pH dependence of the oxidizing power of the • OH radical parallels that for valence band holes, except that there is a discontinuity at pH 11.9, which is the pK for the equilibrium
•

OH ⇐ ⇒ H+ + O−

(5)

At pH 0, the potential for the oneelectron reduction of • OH is +2.683 V [32]
(Table 2). As already mentioned, the redox potential for trapped holes has been estimated to be signiﬁcantly less positive,
+1.50 to +1.75 V against SHE.
The standard potentials for the various redox processes, involving molecular oxygen, can be conveniently visualized in a Frost diagram, given here for pH 7
(Fig. 3) [33, 43, 44]. In this type of diagram, the relative state of reduction of the oxygen is plotted along the horizontal axis, with O2 itself being taken as zero and H2 O being taken as four. The vertical axis is the free energy of formation in units of electron volts. The slopes of the lines between species of differing oxidation states are the negative of the potential for the corresponding redox process. The free energy change for the process is the slope multiplied by the number of electrons involved

6.1 Fundamentals of Photocatalysis
Fig. 3 Frost diagram for O2 reduction at pH 7. Data were taken from Table 2.

HO2

O2−

Free energy
[eV]

0.0
−1.0

HO2−
H2O2

O2

O−
OH

−2.0
H2O

−3.0
0

1

2

3

4

State of reduction electrons in the oxidation state change. At any particular oxidation state, there can be more than one state of protonation depending on the difference of pH 7 and the pKa value for a particular equilibrium. The free energy change is obtained from the relationship

follows: hν − − h+ + e−
−→
−→ h+ + RH − − RH+
+

−→
RH − − R + H
•

(7)
(8)

+

(9)

−→ h+ + H2 O − − • OH + H+ (10)
2.303RT
G
=
(pH − pKa )
F
F

(6)

where G/F gives the free energy change directly in eV. This diagram shows that each of the species with oxidation states
−1, −2 and −3 are unstable with respect to disproportionation. It shows why the hydroxyl radical is such a powerful oxidant: it is simply because it gains so much stability by being reduced to water. It also shows that superoxide is not a very strong one-electron oxidant and that peroxide is a moderately strong two-electron oxidant.
The redox potentials are given in Table 2.
6.1.4.2

Basic Reactions, Kinetics, and Active
Species

The basic photocatalytic reactions have been reviewed by several authors. A generic set of reaction steps for a given organic compound RH can be written as

−→
RH + • OH − − R• + H2 O

(11)

As shown here, the initial attack on the organic compound can be carried out either by the hole (h+ ) or the hydroxyl radical (• OH). The question of which species is more important has been a subject of much debate. One reason for the uncertainty is that it is often difﬁcult to distinguish between products of reactions with holes and products of reactions with hydroxyl radicals. Similar to other aspects of photocatalysis, the answer must be given in very speciﬁc terms. One must specify the precise conditions under which a particular result is obtained. For example, one must consider the surface coverage of the organic compound when discussing the active species. If the coverage is high, direct attack by holes appears to be possible. For relatively low coverages, it appears that • OH is the active species.
Lawless and coworkers argued that an adsorbed • OH radical is identical

505

506

6 Nonsolar Energy Applications

to a trapped hole [36]. By reacting colloidal TiO2 (13 nm) in aqueous suspension with radiolytically generated • OH, they observed a broad absorption band
(ca. 350 nm). The intensity of this signal exhibited ﬁrst-order kinetics, with a rate that was proportional to the • OH concentration as a result of the self-reaction of
•
OH to produce hydrogen peroxide:
•

OH + • OH − − H2 O2
−→

(12)

with a time constant of about 50 ms.
In homogeneous solution, this reaction is second-order (Table 4) but becomes ﬁrst-order on the two-dimensional surface, as discussed later. The trapped hole can be represented as TiIV − O− . On the basis of the reaction with SCN− , the E 0 for this species was estimated to be about 1.5 V (pH 3), which is approximately 1.3 eV above the valence band.
These workers conclude that the existence of adsorbed • OH (surface-trapped holes) can explain most of the observed results, including the production of H2 O2
Tab. 4

and hydroxylated products of organic compounds. The desorption of free • OH is considered to be unlikely because of the extremely large rate constant for the reverse reaction. A later study by Goldstein and coworkers involved phenol as the reactant [45].
The photocatalytic reaction products were compared with those of radiolytically gen• erated • OH on one hand and SO4 − on the other. The latter is a strong one-electron oxidant and would mimic the effect of a hole (see section on mechanisms). At low phenol concentrations, the oxidation products of the photocatalytic reaction are similar to those for • OH, whereas at high phenol concentration, they are similar to
•
those for reaction with SO4 − , indicating involvement of holes. This study shows that the trapped hole acts as a • OH transfer agent. At high phenol concentrations, phenol adsorbs on the TiO2 surface and competes for mobile holes, decreasing the rates of hole localization and electron-hole recombination. Homogeneous reaction rate constants for selected reactions involving hydroxyl

radicals
Reactants

HO• + • OH
2
•
O2 − + • OH
•
OH + • OH
CH3 CHO + • OH
CH3 COCH3 + • OH
CH3 CH2 CH3 + • OH
2-propanol + • OH
1,4-C6 H4 (CO2 H− ) + • OH
Coumarin + • OH
C6 H5 OH + • OH
DMPO + • OH

Products

−−
−→
−−
−→
−−
−→
−−
−→
−−
−→
−−
−→
−−
−→
−−
−→
−−
−→
−−
−→
−−
−→

H2 O + O2
OH− + O2
H2 O 2
H2 O + CH3 CO
Products
Products
Products
Products
Products
dihydroxycyclohexadienyl
DMPO-OH

Rate constant
[M−1 s−1 ]
1 × 1010
1 × 1010
4.2 × 109
3.6 × 109
1.3 × 108
2.3 × 109
1.6 × 109
3.3 × 109
2 × 109
6.6 × 109
4.3 × 109

Note: Values were taken from A. B. Ross, NDRL/NIST Solution Kinetics Database, v. 2.0,
Standard Reference Data, NIST, Gaithersburg, MD 20899, USA (http://allen.rad.nd.edu), also found in Ref. 32. DMPO stands for 5,5-dimethyl-1-pyrroline N-oxide.

6.1 Fundamentals of Photocatalysis

Bahnemann and coworkers have distinguished between shallow and deeply trapped holes [46]. They point out that the absorption features commonly used to indicate trapped electrons (650 nm, 1.7 eV) and holes (430 nm, 2.7 eV), could not be the actual ones involved with photocatalytic reactions because they would correspond to energy levels near the middle of the band gap, for example, 1.7 eV below the CB or 2.7 eV above the VB.
The latter is assigned to an excitation of a surface peroxide. The active oxidizing species is thought to be a shallow trapped hole, which can equilibrate with
VB holes and would have a similar oxidizing power. These authors showed that strongly adsorbing species, such as citrate and acetate, react with free holes, whereas more weakly adsorbed species, such as ethanol, could react with deeply trapped holes (i.e. surface-bound • OH).
They also showed that dichloroacetate and thiocyanate adsorb strongly and react with either freely or shallowly trapped holes but not deeply trapped holes.
Cermenati and coworkers used a very interesting approach to assess the relative importance of holes and hydroxyl radicals [47]. For a model reactant, they employed quinoline, which has two aromatic rings with differing electron density, the nitrogen-containing ring with less.
Because • OH is electrophilic, the attack should occur on the benzene ring. On the basis of a detailed analysis of products as a function of time, the major mechanistic pathways involving • OH were discounted. Instead, a mechanism involving oxidation with holes, together with superoxide, was proposed (see section on mechanisms). A third mechanism was proposed in which initial oxidation by either holes or • OH is followed by reaction

with neutral molecular oxygen O2 . The effect of raising the pH from 3 to 6 was consistent with the second pathway because both the • Q− and • O2 − would be stabilized at the higher pH.
A different approach was taken by
Kesselman and coworkers, who examined the oxidation of some organic anions
(chloroacetates) at heavily Nb-doped TiO2 electrodes. Because the selectivity for reaction of these species with • OH in homogeneous solution is known, as is the ordering of the electrochemical redox potentials, it is possible to distinguish between pathways. In addition, the extent to which the oxidation current depends on the reactant concentration is a measure of the involvement of the direct-hole pathway. If there is no dependence on concentration, this indicates that • OH is produced and then reacts with the organic species. These authors found a full range of behavior from 12% to 97% direct-hole involvement for formate and
4-chlorocatechol, respectively. The results were found to vary signiﬁcantly from one electrode to another (15% vs. 65% directhole involvement for acetate).
From the work of Nosaka and coworkers with electron paramagnetic resonance
(EPR) spectroscopy, the proportion of the direct hole and • OH-mediated pathways can be determined for various types of
TiO2 samples. These workers used the photocatalytic oxidation of acetate as a probe reaction and found mixtures of EPRactive products. They have proposed the following schemes:
CH3 COOH + H+ − −
−→
•

CH3 + CO2 + H+

(13)

CH3 COOH + OH − −
−→
•

•

CH3 + CO2 + H2 O

(14)

507

508

6 Nonsolar Energy Applications

H2 O + h+ − − • OH + H+
−→

•

Ti4+ O− Ti4+ OH− + H2 O − −
−→

(15)

Ti4+ O2 − Ti4+ OH2 + • OH

−→
CH3 COOH + • OH − −
•

CH2 COOH + H2 O

(16)

The methyl radical • CH3 can be produced by reaction with either h+ or
•
OH, whereas the carboxymethyl radical
•
CH2 COOH can be produced only by reaction with • OH. Thus the ratio of • CH3 to
•
CH2 COOH radical yields the proportion of the direct h+ path against the • OH path.
It was found that the • OH path was favored for TiO2 samples with higher coverages of
OH groups. On such surfaces, the type of active surface species has been proposed by Howe and Gr¨tzel [48], based on EPR a measurements, to be
•

Ti4+ O− Ti4+ OH−

(S1)

On TiO2 surfaces that contain fewer
OH groups, perhaps as a result of heat treatment, the active species may instead be [49, 50]:
•

Ti4+ O2 − Ti4+ O−

(S2)

Hirakawa and coworkers propose that the reaction
A

B

C

takes place more easily than the reaction
•

Ti4+ O2 − Ti4+ O− + H2 O − −
−→
Ti4+ O2 − Ti4+ OH− + • OH

D

E

•

OH

O

Ti

Ti
O

O

O•

Ti
O
Ti

+Ti

Fig. 4

O
O

Ti

O
Ti

O
−
•

O
O

Ti

Ti
O

O

O
Ti

O
Ti

O

O
O

Ti

O

O

Ti

Ti
O

Ti

O

O
O

Ti

O

O

O

Ti
O

Ti

O

O
O

Ti

O

O

Ti
O

Ti
O

Ti

O

O

O

Ti

O

O

•

+

O

Ti

O

O

O

Ti

O

Ti
Ti

−

Ti

O
O

Ti

O
O

Ti

•

Ti
O

O

O

O
O

O

Ti

O
O

−O

OH

O

(18)

The structural arrangements of these species are shown in Fig. 4.
From work in our laboratory, it was found that the yields of ﬂuorescent products from the reactions of • OH with coumarin and 1,4-phthalate could be measured as a function of time, with a linear dependence being found for ﬂuorescence intensity versus illumination time [51]. It was concluded that the • OH pathway predominates for the low reactant concentrations used (10−4 to 10−3 M) in aqueous solution. In later work, it was found that the quantum yield for the reaction of • OH with 1,4-phthalate was relatively small (7 ×
10−5 ) under the experimental conditions used (4 × 10−4 M, 2 × 10−3 M NaOH; UV intensity, 1 mW cm−2 ) [52]. In contrast,

O
OH

(17)

Ti

O
Ti

TiO2 surface species, from Nakaoka and coworkers [45].

6.1 Fundamentals of Photocatalysis

the quantum yield for the photocatalytic oxidation of iodide I− to iodine I2 was found to be relatively high (5.7 × 10−2 ) for the conditions used (10−2 M KI; UV intensity, 1 mW cm−2 ). The quantum yields usually observed for photocatalytic reactions at similar light intensities are more in line with the latter, and therefore it was concluded that the hole is the active species in most cases.
As already mentioned, the desorption of active oxygen species, particularly • OH, from the TiO2 surface, with subsequent reaction in homogeneous aqueous solution, has been downplayed as a signiﬁcant pathway for TiO2 photocatalysis, although not by direct experiments. Nevertheless, our group has found it possible to observe the oxidative degradation of organic ﬁlms, for example, methylene blue, that are separated from the illuminated TiO2 surface by air gaps of up to 125 µm [53]. In an effort to identify the active species, various types of organic reactant ﬁlms were examined [54].
Changes in mass were also monitored by use of the quartz-crystal-microbalance
(QCM) technique. In addition to methylene blue, a polyaniline derivative, poly(2methoxyaniline-5-sulfonic acid) (PMAS), polystyrene (PS), octadecyltriethoxysilane
(ODS)-coated glass, and polyvinyl alcohol were examined. With the ODS-coated glass separated from the illuminated TiO2 ﬁlm by 50 µm, oxygen-containing functional groups were produced, as observed with
X-ray photoelectron spectroscopy, creating a hydrophilic surface. For PMAS, PS, and
ODS, the QCM results showed decreases in frequency, that is, increases in mass due to oxygenation. The fact that aliphatic hydrocarbons are oxidized suggests that the active oxidizing agent may not be HO• ,
2
H2 O2 , or 1 O2 . Thus the most likely candidate is • OH. However, it may also be possible that HO• or H2 O2 could be the
2

diffusing species and could then be converted to • OH on reaching the organic ﬁlm, as in the following reactions:
2HO• − − H2 O2 + O2
2 −→ hν H2 O2 − − 2• OH
−→

(19)
(20)

In experiments with chemically generated • OH (Fenton reagent), similar effects were observed in water contact angle for
PS and ODS.
In earlier work, it was found that
Escherichia coli cells were killed in aqueous suspensions in contact with illuminated
TiO2 ﬁlms, even when separated by a 50µm-thick porous membrane [55]. On the basis of the inhibitory effect of catalase, it was concluded that the active species is hydrogen peroxide but there appeared to be a cooperative effect, possibly due to superoxide, which signiﬁcantly enhanced the lethal effect.
Now that many of the important photogenerated species have been identiﬁed, it is useful to consider the kinetics of formation and decay of some of these species.
The absorption of a photon and creation of an electron-hole pair in a TiO2 particle occurs in a time range that has, thus far, been too short to measure, that is, a few femtoseconds. These charge carriers are initially considered to exist in the conduction band and valence band, respectively but are then quickly trapped.
Skinner and coworkers have determined the trapping time for electrons to approximately 180 fs based on measurements on TiO2 sols in acetonitrile, which was used to avoid the absorbance of hydrated electrons [56]. The nanoparticle surface consists of many different types of local geometries and corresponding trap energies, leading to a broad range for the absorbance spectrum. 509

510

6 Nonsolar Energy Applications

Holes have been estimated to be trapped within 1–10 ps by Serpone and coworkers, and these workers ﬁnd that 90–100% of all electron-hole (e-h) pairs have undergone recombination −→ e− + h+ − − heat

(21)

within 10 ns for colloidal TiO2 particles from 2–27 nm in diameter [57]. However,
Colombo and coworkers estimate the halflife of an e-h pair to be approximately
30 ps [58]. This is just one example of the wide variety of kinetic results that have been reported (Table 5). Second-order kinetics are observed particularly at short times and high light intensities, that is, when there are multiple e-h pairs existing simultaneously in a single particle [57, 59].
Results for time-resolved microwave conductivity measurements, which reﬂect the number of charge carriers, indicate that the decay kinetics span many orders of magnitude in time, for example, from ns to s [60, 61, 63]. Striking differences were observed, comparing anatase and rutile, with the latter exhibiting much faster decay kinetics [60]. Major differences were observed among various types of colloidal TiO2 preparations and in comparison to Degussa P25 in terms of recombination kinetics [63]. However,
Colombo and coworkers, using diffuse reﬂectance ﬂash photolysis experiments, have found the behavior of TiO2 colloidal solutions versus powder form, rutile versus anatase, and nanoparticle versus P25, all to be similar [58, 65]. These workers point out that the diffuse reﬂectivity results are speciﬁc for trapped electrons, rather than for carriers in general and assign the long-lived species observed in this and other studies to such electrons [65]. They also show that the effect of a hole scavenger, such as thiocyanate, is to

greatly increase the lifetime of the trapped electrons [58].
It had already been observed in early work with ﬂash photolysis, albeit on much longer timescales, that recombination could be greatly inhibited in the presence of either hole scavengers or electron scavengers [67]. Lifetimes of seconds or even hours have been reported for trapped electrons in the presence of hole scavengers such as 2-propanol [68].
Lifetimes of milliseconds (in acid) or microseconds (in base) have been reported for trapped holes in the presence of an electron scavenger such as methylviologen [67].
The decay process is usually assumed to involve recombination of electrons and holes in various states of trapping or detrapping. Because at short time intervals and relatively high light intensities, there may be approximately equal numbers of each type of carrier in a single particle, the recombination process can be expected to be second-order. When there is on the order of only a single eh pair in a particle, the decay process is known to become ﬁrst-order. This has been explained in terms of stochastic processes [59, 69–71], that is, because of the discrete nature of the species and corresponding events, they must be treated statistically rather than continuously. The simple fact that reactions are taking place on a two-dimensional surface is also important [72]. However, it is well known in the ﬁeld of solid-state electronics that direct recombination in a doped semiconductor behaves as a ﬁrst-order process for low injection rates, simply because the magnitude of the squared term in the rate expression becomes small [73]. This is the analog of pseudoﬁrst-order kinetics. In the case of indirect recombination (i.e. mediated by a band gap

OH decay

Sol
Sol
Sol
Sol
Sol
Anatase
powder
Rutile
powder
P25
powder
Sol

P25
P25

S13
P25

Q-size

e, h-decay e− -decay

e− -decay e− -decay

e− -decay

e− -decay
P25
P25 e− -decay e, h-trapping P25

•

e− -decay

e− -decay

e− -trapping h+ -trapping e− -decay e− -decay e− -decay e− -decay

TiO2 type aqueous

dry

dry

aqueous aqueous aqueous aqueous aqueous dry Medium

2–4 nm

–
29 nm dry paste dry 29 nm dry 29 nm
2-propanol
29 nm aqueous (124 nm)
10 ns aqueous 29 nm paste 13 nm

29 nm

–
–
–
–
–

Particle size, nm

–

–

Inert
Inert
Inert
Inert
Inert
–

Gas

–

–
–

2.5
–

–
–
2.5

–

–
–

O2
–

–
–
O2

Acid Inert

–

–

2.7
2.7
2.7
2.7
2.7
–

pH

PR-TRMC
PR-TRMC
LFP

PR

TRMC

TRMC

LFP
LFP
LFP
LFP
LFP
TRMC

–

TRMC

TRMC
TRMC

–
–
–
–

LFP
TRMC

0.8
0.8
–

–

–

–

67
67
67
6
0.85
–

e-h/
Techniquea
particle

Kinetics of photogenerated charge carrier trapping and decay

Reaction

Tab. 5

OH selfreaction

Rothenberger
Rothenberger
Rothenberger
Rothenberger
Rothenberger
Schindler

Author

1991 Lawless

1990 Schindler

1990 Schindler

1985
1985
1985
1985
1985
1990

Year

64

63
64

62
63

61
61
62

36

60

60

59
59
59
59
59
60

References

(continued overleaf )

trans-decalin 1994 Martin
Various
1994 Martin adsorbates Various
1994 Martin adsorbates 1993 Lepore
1994 Martin

∼10 ns
100 ns to
10 ms
>60 ms
230 ns to
1.6 ms
1.6 µs to
11 ms

trans-decalin

1991 Warman
1991 Warman
1993 Lepore

•

Comments

100 ns
25 µs
0.25–1 ns

20 ns

1 µs–10 s

1–5 µs

Photoelectrochemistry

Similar Documents

Electronic Music

Popular Essays